Geophysics

Geophysics and
Geosequestration
Geosequestration involves the deep geological storage of carbon dioxide from
major industrial sources, providing a potential solution for reducing the rate
of increase of atmospheric concentrations of carbon dioxide and mitigating
climate change. This volume provides an overview of the major geophysical
techniques and analysis methods for monitoring the movement and
predictability of carbon dioxide plumes underground. Comprising chapters
from eminent researchers, the book is illustrated with practical examples and
case studies of active projects and government initiatives, and discusses their
successes and remaining challenges. A key case study from Norway
demonstrates how governments and other stakeholders could estimate
storage capacity and design storage projects that meet the requirements of
regulatory authorities. Presenting reasons for embracing geosequestration,
technical best practice for carbon management, and outlooks for the future,
this volume provides a key reference for academic researchers, industry
practitioners, and graduate students looking to gain insight into subsurface
carbon management.
Thomas L. Davis is Professor Emeritus at the Colorado School of Mines and

founder of the Reservoir Characterisation Project (RCP), which received the
Distinguished Achievement Award from the Society of Exploration
Geophysics (SEG) in 2014. He is an active member of the SEG as an organizer
for conferences, workshops, and education programs, and has served as
SEG’s Second Vice President, Technical Program Chairman, and
Distinguished Lecturer. He has received the C. J. Mackenzie Award from the
Engineering College of the University of Saskatchewan, the Milton B. Dobrin
Award from the University of Houston, and the Deans Excellence and Melvin
F. Coolbaugh Memorial Awards from the Colorado School of Mines.
Martin Landrø is Professor at the Norwegian University of Science and

Technology. Past positions include working as a research geophysicist and
manager at IKU Petroleum Research and as a specialist at Statoil’s research
centre in Trondheim. He has won numerous awards including the Norman
Falcon Award, Louis Cagniard Award, and Conrad Schlumberger Award
from the European Association of Geoscientists and Engineers; the
Norwegian Geophysical Award; Statoil’s Research Prize; the SINTEF Award
for outstanding pedagogical activity; the ENI Award (New Frontiers in
Hydrocarbons); and the IOR Award from the Norwegian Petroleum
Directorate. His research interests include seismic inversion, marine seismic
acquisition, and 4D and 4C seismic surveys.
Malcolm Wilson has worked within government, academia, and the

petroleum research industry, and currently serves on the advisory council for
the Reservoir Characterization Project (Colorado School of Mines) and on
the International Advisory Board for the New Energy Coalition in the
Netherlands. He was instrumental in the creation of the IEA Greenhouse Gas
Weyburn CO2 Monitoring and Storage Project – an extensive research
program for monitoring the storage of CO2 in an oilfield. He was also part of
the Intergovernmental Panel on Climate Change (IPCC) special report on
capture, transport, and storage of CO2, and of the IPCC team that received
the Nobel Peace Prize in 2007. Dr. Wilson is an adjunct professor at the
Faculty of Engineering and Applied Science at the University of Regina, and
has received the Alumni of Distinction Award from the University of
Saskatchewan.
Geophysics and
Geosequestration
Edited by
Thomas L. Davis
Colorado School of Mines
Martin Landrø
Norwegian University of Science and Technology, Trondheim
Malcolm Wilson
New World Orange BioFuels
University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
314–321, 3rd Floor, Plot 3, Splendor Forum, Jasola District Centre, New Delhi – 110025, India
79 Anson Road, #06–04/06, Singapore 079906
Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning, and research at the highest international levels of excellence.
www.cambridge.org
Information on this title: www.cambridge.org/9781107137493
DOI: 10.1017/9781316480724
© Cambridge University Press 2019
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2019
Printed in the United Kingdom by TJ International Ltd. Padstow Cornwall
A catalogue record for this publication is available from the British Library.
ISBN 978-1-107-13749-3 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party internet websites referred to in this publication
and does not guarantee that any content on such websites is, or will remain,
accurate or appropriate.
Contents
List of Contributors vii
Preface ix
Acknowledgments xi
Abbreviations xii
Part I Introduction 10 Microseismic Imaging of CO2 Injection 168

Shawn Maxwell
1 Climate Change and the Role of Carbon
Capture and Storage in Mitigation 1 11 Well Logging 181
John Gale and Malcolm Wilson Zaki Bassiouni
2 The Role of Geophysics in Carbon Capture and

Storage 12 Part III Case Studies
David Lumley
12 Offshore Storage of CO2 in Norway 195
3 Goals of CO2 Monitoring: Why and How to Eva K. Halland
Assess the Subsurface Changes Associated
with Carbon Capture and Storage 54 13 Twenty Years of Monitoring CO2 Injection at
Thomas M. Daley and William Harbert Sleipner 209
Ola Eiken
14 Case Studies of the Value of 4D,
Part II Geophysical Techniques Multicomponent Seismic Monitoring in CO2
4 Rock Physics of CO2 Storage Monitoring in Enhanced Oil Recovery and
Porous Media 71 Geosequestration 235
Thomas M. Daley Thomas L. Davis, Scott Wehner, and
Trevor Richards
5 Multicomponent Seismic Monitoring 83
Thomas L. Davis and Martin Landrø 15 Integrated Geophysical Characterization and
Monitoring at the Aquistore CO2 Storage
6 Monitoring the Deformation Associated with Site 257
the Geological Storage of CO2 93 Don White
Donald W. Vasco, Alessandro Ferretti,
Alessio Rucci, Sergey V. Samsonov, and 16 Development and Analysis of a Geostatic
Don White Model for Shallow CO2 Injection at the Field
Research Station, Southern Alberta,
7 Gravity: Surface and Borehole 115 Canada 280
Ola Eiken Donald C. Lawton, Jessica Dongas, Kirk
Osadetz, Amin Saeedfar, and Marie
8 Estimating Saturation and Density Changes
Macquet
Caused by CO2 Injection at Sleipner 134
Martin Landrø and Mark Zumberge 17 Seismic and Electrical Resistivity
Tomography 3D Monitoring at the Ketzin
9 Electrical and Electromagnetic
Pilot Storage Site in Germany 297
Methods 154
Christopher Juhlin, Stefan Lüth, Monika
Erika Gasperikova and Michael Commer v
Ivandic, and Peter Bergmann
Contents
18 Time-Lapse Seismic Analysis of the CO2 Part IV Summary

Injection into the Tubåen Formation at
Snøhvit 319 20 What Next? 371
Sissel Grude and Martin Landrø Thomas L. Davis, Martin Landrø, and
Malcolm Wilson
19 Illinois Basin–Decatur Project 339
Robert A. Bauer, Robert Will, Sallie E.
Greenberg, and Steven G. Whittaker
Index 375
vi
Contributors
Zaki Bassiouni, Louisiana State University Martin Landrø, Norwegian University of Science and
Technology
Robert A. Bauer, Illinois State Geological Survey
Donald C. Lawton, University of Calgary
Peter Bergmann, GFZ German Research Centre for
Geosciences David Lumley, University of Texas at Dallas
Michael Commer, Lawrence Berkeley National Stefan Lüth, GFZ German Research Centre for
Laboratory Geosciences
Thomas M. Daley, Lawrence Berkeley National Marie Macquet, University of Calgary
Laboratory
Shawn Maxwell, Itasca-Image Company
Thomas L. Davis, Colorado School of Mines
Kirk Osadetz, CMC Research Institutes Inc.
Jessica Dongas, University of Calgary
Trevor Richards, Denbury Resources
Ola Eiken, Quad Geophysical Company
Alessio Rucci, TRE ALTAMIRA
Alessandro Ferretti, TRE ALTAMIRA
Amin Saeedfar, CMC Research Institutes Inc.
John Gale, IEA Green House Gas R&D Programme
Sergey V. Samsonov, Canada Centre for Mapping and
Erika Gasperikova, Lawrence Berkeley National Earth Observation, Natural Resources Canada
Laboratory
Donald W. Vasco, Lawrence Berkeley National
Sallie E. Greenberg, Illinois State Geological Survey Laboratory
Sissel Grude, Norwegian University of Science and Scott Wehner, Denbury Resources
Technology
Steven G. Whittaker, Illinois State Geological Survey
Eva K. Halland, Norwegian Petroleum Directorate
Don White, Geological Survey of Canada
William Harbert, University of Pittsburgh
Robert Will, Schlumberger Carbon Services
Monika Ivandic, Uppsala University, Sweden
Malcolm Wilson, New World Orange BioFuels
Christopher Juhlin, Uppsala University, Sweden
Mark Zumberge, Scripps Institute of Oceanography
vii
Preface
Two famous scientists in the late nineteenth century The enthusiasm for sequestration in the subsur-
laid the groundwork for this book. Svante Arrhenius face is tempered by fear of the unknown. What
developed the concept of the greenhouse effect based happens if the CO2 leaks into my basement, under-
on the chemistry of the atmosphere and initiated our ground parking, or other such confined spaces? This
understanding of the results of changing atmospheric is the same question that plagued the nuclear indus-
chemistry. Emil Johann Weichert, the first geophysics try and the storage of spent nuclear fuel in man-
professor, was examining the use of “artificial earth- made caverns. As an example, Shell was unable to
quakes” to look into the subsurface. Over the last move ahead with a proposed storage project in
dozen decades, the science of geophysics has moved Barendrecht, Netherlands, some years ago. Fear of
forward substantially, primarily as a result of demand the CO2 (which is known to have asphyxiated people
from the resource industry, and the current state of in old coal mines, for example) and concerns about
the art is examined in this book. The issue of the housing prices prevented this venture. In areas
greenhouse effect, now termed global warming, has where the oil and gas industry is important and
taken a different path. Concern about global warming people understand enhanced oil recovery (EOR),
came to public attention with the publication in the issues are far less pronounced. Hence Shell was
1987 of Our Common Future by the Brundtland able to go ahead with the Quest project in Alberta
Commission and the Toronto Conference of 1988 and there were only insignificant problems with EOR
that tried to develop a pathway for emissions reduc- in Weyburn, Saskatchewan.
tions. In subsequent years, the importance of global The key to public acceptance is the ability to
warming and the need to reduce its impact on econo- demonstrate the safe sequestration of the CO2 in the
mies and ecosystems, including human ecosystems, subsurface. In addition, the regulator must be able to
was recognized by the award of part of the 2007 confirm the volumes of CO2 sequestered to be able to
Nobel Peace Prize to the scientists working on the calculate national emissions. Voluntary standards are
United Nations reports on climate change, the in the process of being developed in recognition of
Intergovernmental Panel on Climate Change [IPCC]. this need for confirmation. We can measure the CO2
Since the late 1980s much effort has gone into the being injected into the subsurface with a great deal of
development of ways to reduce the emissions of green- accuracy, but we need to confirm that the CO2 stays
house gases, particularly carbon dioxide (CO2), into where it is intended. This important fact underlies the
the atmosphere. One of the ways identified is to inject significance of this volume: using geophysics, particu-
CO2 either into the deep subsurface as a solvent for larly seismic surveying, as the only way to “watch” the
enhanced oil recovery (Sacroc Field in Texas has been CO2 in the subsurface and ensure its stable sequestra-
receiving CO2 for more than 40 years) or into deep tion in its intended injection location. From the per-
saline aquifers (the first of these projects is offshore spective of the resource sector, geophysics has proven
Norway, the Sleipner project, started in 1996). This irreplaceable as a technique for improving the effi-
book contains case studies of both types of geoseques- ciency with which CO2 is used to extract oil resources.
tration of CO2. Indeed, the International Energy Regardless of environmental issues, CO2 has a cost
Agency (IEA) estimates that by 2050, some 13% of and is, therefore, an expensive solvent to put in the
reductions of greenhouse gas emissions will need to ground. Understanding where the CO2 is going is
be realized through geosequestration or the cost of important to minimizing CO2 utilization and effec-
emissions reductions will increase significantly. tively recovering the oil. This is vitally important for
ix
Preface
sustaining high-quality jobs in what are often under- contributor, has a long history with pushing the limits
developed parts of the world. of geophysics in EOR projects). The authors of the
The focus of this volume is, therefore, to evaluate various chapters have all developed distinguished
the state of the art in seismic work to determine the reputations in the field of geophysics, and are worthy
location and movement of CO2 in the subsurface. This successors to the likes of Weichert, with a particular
is a truly exciting venture, not just from the perspec- focus on the use of CO2 in the fossil fuel industry or
tive that Weichert would have recognized – the use sequestration in the deep subsurface. Chapter 16 by
of seismic surveys to ensure effective and safe exploi- Lawton et al. also examines our ability to track small
tation of fossil fuel resources – but also to meet volumes of CO2 at shallow subsurface levels should
a growing and important need to ensure to the reg- the CO2 escape from its storage location (this is also
ulator and the public that this is being done safely. mentioned in Chapter 13 by Eiken on the Sleipner
In this latter regard, the technology is meeting the goal storage project), thereby demonstrating our ability to
of the Brundtland Commission to reduce emissions to take action should the remote probability of leakage
the atmosphere. Chapter 3 by Daley and Harbert actually occur. Chapter 19 by Bauer et al. shows the
examines in more detail why we need to monitor sequestration of CO2 under an area with significant
what is happening in the subsurface. surface activity, demonstrating not only an ability to
This volume discusses the technology as it exists survey under less than optimal conditions, but also
today (for example, the excellent review by Lumley in the acceptance by the public in a situation where care
Chapter 2) and then puts the basics of seismic and is taken to understand the processes deep beneath the
other geophysical techniques work into the real world public’s feet.
with a series of important case studies including This volume will be a major contribution to stu-
Sleipner and Snohvit, among others, as pure seques- dents and practitioners of geophysics and to those
tration projects (Landrø, one of the editors and con- concerned with the issue of global warming and tak-
tributors, has played a significant role in this work) ing concrete action to reduce emissions to the atmo-
and also a number of EOR projects where the CO2 sphere in a safe and sustainable manner. We thank
can be monitored in oil fields (Davis, as editor and you for reading the book.
x
Acknowledgments
Publishing a book gives one a new appreciation for and very much appreciated. There is little reward to
not just the science but also the human factor that offer other than “job well done.”
goes into its preparation. From the beginning, the To those who helped us review the manuscripts
people at Cambridge University Press have been and to provide timely reviews and encouragement
very supportive. We would like to acknowledge to our authors to make their contribution better,
Susan Francis and Zoe Pruce for working with us we say thanks. Specifically, we would like to recog-
and providing the encouragement and guidance to nize Malcolm Wilson, Sean O’Brien, and Jared
see the project through. Atkinson. Malcolm volunteered to do the final
To our authors from around the world we say editing before the production phase. Thanks,
thanks. Your willingness to present your work and Malcolm, for stepping forward and taking one for
then to follow through as you have is quite remarkable the team.
xi
Abbreviations
B magnetic field ERT electrical resistance tomography

CSEM controlled source electromagnetics H magnetic field (B ¼ μH)
CO2 carbon dioxide HED horizontal electric dipole
DC direct current TDS total dissolved solids
E electric field VMD vertical magnetic dipole
EM electromagnetic
xii
Part I Introduction
Chapter
Climate Change and the Role of Carbon
1 Capture and Storage in Mitigation

John Gale and Malcolm Wilson
1.1 Climate Change and the Need for ppm during the Pliocene Epoch, between 5 and 3
million years ago. In that period, global average
Adaptation and Mitigation temperatures have been estimated to be 3–4°C
The atmosphere of the Earth contains, among others, and as much as 10°C warmer at the poles than
gases that are referred to as the greenhouse gases current levels. Sea levels have been estimated to
(GHGs). These gases are so called because, in the have ranged between 5 and 40 m higher than today.
atmosphere, they both absorb and emit radiation. It While CO2 has the highest atmospheric concen-
is this process of absorption and emission of radiation tration of the GHGs and its contribution to total
in the atmosphere that is the fundamental cause of radiative forcing is the largest of all the gases, that of
what is termed the greenhouse effect (Houghton et al., the other GHGs cannot be ignored. The atmospheric
1990). The main GHGs in the Earth’s atmosphere are concentrations of other GHGs have also risen signifi-
carbon dioxide (CO2), water vapor, methane (CH4), cantly since the late 1980s. The atmospheric concen-
nitrous oxide (N2O), and ozone. Collectively the trations of CH4 and N2O in 2011 were 1803 ppb and
greenhouse gases significantly affect the Earth’s tem- 324 ppb, respectively, exceeding preindustrial levels
perature; scientists predict that without them, the by about 150% and 20% (IPCC, 2013).
temperature at the Earth’s surface would average The Intergovernmental Panel on Climate Change
about 33°C colder than the present average of 14°C (IPCC) was established by the United Nations
(Le Treut et al., 2007). Environment Programme (UNEP) and the World
Since the late eighteenth century, with the Meteorological Organization (WMO) in 1988 to pro-
beginning of the Industrial Revolution in 1750, vide the world with a clear scientific view on the
the atmospheric concentrations of these GHGs, current state of knowledge of climate change and its
CO2 in particular, have risen substantially (Le potential environmental and socioeconomic impacts
Treut et al., 2007). Since that time, the atmospheric (IPCC, 1988). The IPCC reviews and assesses the most
concentration of CO2 has increased from 280 to recent scientific, technical, and socioeconomic informa-
more than 400 parts per million (ppm). Since tion produced worldwide relevant to the understanding
1957, the atmospheric concentration of CO2 has of climate change. It does not conduct any research nor
been measured at the Mauna Loa Observatory in does it monitor climate-related data or parameters.
Hawaii and is presented in what is known as the The IPCC publishes its results in a series of
Keeling Curve. In May 2016, the National Assessment Reports, the latest of which is Fifth
Oceanographic and Atmospheric Administration Assessment Report (AR5), which was published in
of the USA reported the highest ever monthly early 2014 (IPCC, 2014).
level of CO2 in the air: 407.7 ppm. The increase The AR5 indicates that global climate change has
in the concentration of CO2 in the atmosphere is already had observable effects on the environment.
attributed to the increased use of fossil fuels, com- Points highlighted by the report include:
bined with the extensive deforestation observed
• Warming of the climate system is unequivocal.
since the Industrial Revolution. Atmospheric con-
centrations of CO2 have not been observed at such • Each of the past three decades has been
successively warmer than any preceding decade
levels for millennia. Scientific analyses suggest that
since 1850.
atmospheric CO2 levels reached as much as 415 1
• Over the last two decades, the Greenland and climate change far outweigh the costs of not acting.
Antarctic ice sheets have been losing mass, glaciers According to the Review, without action the overall
have continued to shrink almost worldwide, and costs of climate change will be equivalent to losing at
Arctic sea ice and Northern Hemisphere spring least 5% of global gross domestic product (GDP) each
snow cover have continued to decrease in extent. year, now and in all the years to come. Including a
• The rate of sea level rise since the mid-nineteenth wider range of risks and impacts could increase this to
century has been larger than the mean rate during 20% of GDP or more, also indefinitely. Stern believed
the preceding two millennia. at the time that 5–6°C of temperature increase was “a
Over the coming decades, it is expected that real possibility.”
We then have two options to follow: mitigation
• Global surface temperature changes for the end of and adaptation. Mitigation addresses the root causes
the twenty-first century will likely exceed 1.5°C by reducing greenhouse gas emissions while adapta-
relative to 1850 to 1900 and may even exceed 2°C. tion seeks to lower the risks posed by the conse-
• Warming over the twenty-first century will cause quences of climatic changes
nonuniform responses in the global water cycle, (IPCC, 2007b). In reality, we will follow a twin
increasing the contrast in precipitation between process because we expect that some 1 to 1.5°C of
wet and dry regions and between wet and dry warming is already “locked in” (World Bank, 2014).
seasons. Humans have been adapting to climatic changes
• Continued warming of the global ocean will affect throughout their evolution, but we probably now
ocean circulation as heat penetrates from the face a greater challenge to adapt than ever before.
surface to the deep ocean. This chapter, however, concentrates on the issue of
• The Arctic sea ice cover and the Northern mitigation.
Hemisphere spring snow cover extents, and the
global glacier mass will all decrease further as
global mean surface temperature rises.
• Increased ocean warming and loss of mass from
1.2 What Mitigation Options Are
glaciers and ice sheets will cause global mean sea Needed?
level to rise at a rate that will very likely exceed that The largest sources of global emissions are the power
observed from 1971 to 2010. and industry sectors, which represent 56% of global
• Climate change will affect carbon cycle processes GHG emissions, including fugitive emissions from
in a way that will exacerbate the increase of CO2 in fossil fuel mining, refining, and transportation. Of
the atmosphere. Further uptake of carbon by the the global greenhouse gases, CO2 accounts for 65%
oceans will increase ocean acidification. of these emissions, primarily from the use of fossil
A recent report coauthored by the International fuels. The discussion that follows therefore concerns
Geosphere–Biosphere Programme (IGBP), the the mitigation of emissions of CO2 from the use of
Intergovernmental Oceanographic Commission fossil fuel in the power and industrial sectors. It is
(IOC-UNESCO), and the Scientific Committee on important at this junction to stress that there are
Oceanic Research (SCOR) highlights the impacts of multiple mitigation courses of action in both of
increased ocean acidification, the economic impact of these sectors, and the aim of this discussion is not to
which could be substantial. select individual ones but to recognize that all the low-
Ocean acidification causes ecosystems and marine carbon technology options will be needed in combi-
biodiversity to change. It has the potential to affect nation to meet strict emission targets. It also follows
food security and limits the capacity of the ocean to that the portfolio and balance of low-carbon technol-
absorb CO2 from human emissions. ogy options will vary in different regions to account
The Stern Review (2006) stated that climate for national considerations. To be clear, there is no
change is the greatest and widest-ranging market fail- single low-carbon technology option that will reduce
ure ever seen, presenting a unique challenge for eco- GHGs sufficiently on its own and there is no “one size
nomics. The Review’s main conclusion was that the fits all” low-carbon technology portfolio option
2 costs and benefits of strong and early action on either.
Climate Change and the Role of CCS in Mitigation
Technologies
60 Renewables 30%
50
CCS 13%
40 Power generation efficiency and
GtCO2
fuel switching 1%
30
End-use fuel switching 10%
20
End-use fuel and electricity
10 efficiency 38%
0 Nuclear 8%
2012 2020 2030 2040 2050
Cumulative reductions by sector and technology

Renewables
Other...
Buildings CCS
Transport Fuel switching
Industry Energy efficiency
Power Nuclear
0 50 100 150 200 250 300 350
GtCO2
Figure 1.1 Key technology options to reduce global greenhouse gas emissions from the power sector for the IEA’s 2-degree scenario.
To assess the mitigation needs, the international series. ETP’s annual analyses aim to set out sustain-
community has set itself temperature targets to limit able energy transition pathways that incorporate
the impact of global warming. These targets are detailed and transparent quantitative analysis and
framed and agreed on through the United Nations thus provide targeted reading for experts in the energy
Framework Convention on Climate Change field, policy makers, and heads of governments.
(UNFCCC), whereby countries meet annually at the The 2DS lays out an energy system deployment
Conference of the Parties (COP) to agree on interna- pathway and an emissions trajectory that are consis-
tional targets for climate change mitigation. Prior to tent with an at least 50% chance of limiting the aver-
COP21 in Paris, France in November 2015, the inter- age global temperature increase to 2°C. The 2DS
nationally agreed on target for limiting the increase in limits the total remaining cumulative energy-related
global average surface temperature rise target was 2 CO2 emissions between 2015 and 2100 to 1000
degrees centigrade (2°C Scenario [2DS]). The 2°C goal GtCO2. The 2DS reduces CO2 emissions (including
is achieved by limiting the concentration of GHGs in emissions from fuel combustion and process and
the atmosphere to around 450 ppm of CO2. feedstock emissions in industry) by almost 60% as
Considerable work has been done by many orga- compared to 2013 levels by 2050, with carbon emis-
nizations that have modeled the energy sector and sions being projected to decline after 2050 until car-
looked at the technology options that could meet bon neutrality is reached.
such a temperature target. One such organization An example of the modeling outputs as provided
that has been very active is the International Energy in ETP 2016 is given in Figure 1.1. The figure shows
Agency (IEA). For example, 2DS is the main focus of the key technologies that can be used to mitigate GHG 3
IEA’s Energy Technology Perspectives (ETP) book emissions from the power sector and the relative
contribution between 2013 and 2050 to meet the 2DS fuel power plants, transporting it to a storage site by
compared to a business-as-usual scenario (6DS). As pipeline or ship, and injecting it where it will be
indicated earlier, to achieve the required GHG emis- prevented from entering the atmosphere, normally
sion savings a portfolio of technologies will be in an underground geological formation. In this way
deployed that include nuclear power, fuel switching release of large quantities of CO2 into the atmosphere
and energy efficiency, carbon capture and storage can be prevented (IPCC, 2015).
(CCS), and various renewable options. The analyses by IEA and GEA discussed earlier
For industry, there is a similar picture that CO2 have shown that CCS is a key low-carbon technology
emissions in the 2DS will be achieved through a com- and has a strong role to play as part of the global low-
bination of lower carbon fuels, energy efficiency carbon technology portfolio in reducing global GHG
improvements, and CCS. emissions.
The IEA’s analysis is not a unique one. Another The IPCC Fifth Assessment report made two
example is the recent Global Energy Assessment (GEA) strong points regarding the importance of CCS
2012, which also works from setting a “business–as- (IPCC, 2014). First, it showed that most of the global
usual” base case but uses a different assessment assessment models could not reach the 2°C tempera-
approach than the IEA. The GEA report shows that ture rise target without the inclusion of CCS in the
there are many combinations of energy resources, end- global technology portfolio. Second, the IPCC analy-
use, and supply technologies that can simultaneously sis indicated that the cost of meeting the 2°C tempera-
address the multiple sustainability challenges. One of ture target would be 138% higher without the
the report’s key findings is that energy systems can be incorporation of CCS.
transformed to support a sustainable future through With the discussion beginning to unfold regarding
(a) radical improvements in energy efficiency, espe- a 1.5°C target and the need for negative emissions,
cially in end use; and (b) greater shares of renewable both BioCCS and Direct Air Capture (DAC) become
energies and advanced energy systems with CCS for relevant to future demand for CCS. Recent work to
both fossil fuels and biomass (Johansson et al., 2012). assess the global potential for BioCCS has suggested
At a recent COP, COP21 in Paris, the delegates the technical potential is large and, if deployed, could
agreed to a new temperature rise target of less than 2° result in negative emissions up to 10 Gtonnes of CO2
C, with the intent of limiting global surface temperature equivalents annually. The key obstacle to the imple-
rise to 1.5°C. This sets an even stronger need to reduce mentation of the technology is the absence of a price
emissions, and the technologies discussed previously for stored biomass-based CO2. There is, therefore, a
will all be required. Discussions in the post-Paris envir- need for policy developments in this area to assist
onment indicate that negative emission technologies global adoption of the technology (IEAGHG, 2011a).
(NETs) will also be required. These are technologies The implementation of DAC will require the utiliza-
that remove more emissions from the atmosphere tion of geological storage space to store CO2 that has
than they put in. A variety of NET options are being been directly removed from the atmosphere.
considered that include afforestation, enhanced weath- Consequently, the development of the transportation
ering, ocean fertilization, BioChar, direct air capture, and storage components of CCS will aid in advancing
and BioCCS, among others. However, unlike those dis- this technology.
cussed under the 2DS model, most of the NET technol-
ogies have not been tested at any realistic scale and thus
require considerable development in the coming cen- 1.4 Geological Storage of CO2
turies if they are to be successful in significantly miti- Geological storage of CO2 is generally accomplished
gating GHG emissions. by injecting it in dense form into rock formations
below the Earth’s surface (it may well be possible to
inject the CO2 into shallower formations as a gas as
well; certainly publications such as Chapter 17 by
1.3 CCS as a Global Mitigation Option Juhlin et al. in this book demonstrate the safety of
before and after the Paris Agreement shallower traps). Porous rock formations that hold or,
4 CCS is the process of capturing produced carbon as in the case of depleted oil and gas reservoirs, have
dioxide CO2 from large point sources, such as fossil previously held fluids, such as natural gas, oil, or
Table 1.1 Storage capacity for several geological storage since the IPCC SRCCS, the fact that the largest CO2
options
storage potential globally lies in deep saline forma-
Reservoir type Lower estimate Upper estimate tions still remains a core conclusion to this day.
of storage of storage Since the IPCC SRCCS, our knowledge on how
capacity capacity these storage resources can be developed has
(GtCO2) (GtCO2) advanced. Gas fields hold a greater storage potential
Oil and gas fields 675a 900a than oil fields. Compared to deep saline formations,
Unminable coal 3–15 200
both gas and oil fields are much better explored and
seams have a background data set of both geological and
(enhanced coal- operational/production data. On this basis, they
bed methane) both should be more suitable for early application of
Deep saline 1000 Uncertain but CO2 storage than deep saline formations. Storage in
formations possibly 104 oil fields is typically carried out as part of enhanced oil
recovery operations (EORs). In such systems, the
From Johansson et al. (2012). injection of CO2 is used to maximize oil production,
a
These numbers would increase by 25% if “undiscovered” oil and not for storage of the injected CO2 (see, e.g., Chapter
gas fields were included in this assessment. Source: IPCC SRCCS
2005.
14 in this book by Davis et al.). However, incidental
storage does occur within the reservoir that can
amount to 90% of the CO2 injected (Whittaker and
Perkins, 2013).
brines, are potential candidates for CO2 storage. From the time of the release of the IPCC SRCCS
Suitable storage formations can occur in both onshore there has been little research in the potential for geo-
and offshore sedimentary basins (natural large-scale logical storage in coal seams. In the past few years,
depressions in the Earth’s crust that are filled with interest in using shale has also surfaced as a potential
sediments). Coal beds also may be used for storage of storage option. Again, it is too early to decide whether
CO2 where it is unlikely that the coal will later be this is a promising option for the future or not
mined and provided that permeability is sufficient (IEAGHG, 2013a).
(IPCC, 2005). While deep saline formations represent a tantaliz-
The IPCC Special Report on CO2 Capture and ing resource for global storage of CO2, they remain
Storage (IPCC SRCCS, 2005) undertook the first relatively unexplored in most regions of the world.
review of the global potential for CO2 storage in Deep saline formations require much more extensive
geological formations (IEAGHG, 2011a). The IPCC characterization because, in general, they are “virgin”
SRCCS considered in detail three types of geological formations not previously investigated. Because of
formations that had at that time received extensive this, they require much longer lead times, potentially
consideration for the geological storage of CO2. The up to 15 years of preexploration, to be considered as
three options were storage in oil and gas reservoirs, viable for geological storage of CO2 (IEAGHG,
deep saline formations, and unminable coal beds. 2011b). Chapter 12 by Halland in this book demon-
At the time of the IPCC SRCCS several other strates the importance of having a catalog of storage
possible geological formations or structures were con- opportunities in anticipation of the need for geose-
sidered, such as basalts, oil or gas shales, salt caverns, questration in the future.
and abandoned mines. However, it was believed at the Storage efficiency, for example, depends on the
time that these represented only niche opportunities characteristics of the storage aquifer and confining
or had not been sufficiently studied at that time to caprock, operational characteristics of CO2 storage,
assess their potential. This conclusion is still largely and regulatory constraints. Based on these com-
valid today, although interest in shale formations for bined factors, storage efficiency can vary widely,
CO2 storage is growing. with values ranging from less than 1% to greater
The estimates of the technical potential for differ- than 10%. This wide variation in storage efficiency
ent geological storage options from the IPCC SRCCS correspondingly impacts storage capacity. In the
are summarized in Table 1.1. While there have been IPCC SRCCS storage capacity estimates were 5
numerous studies on the individual storage options based on a conservative 1% estimate of pore
Table 1.2 IEA GHG CO2 monitoring selection tool
Dark pink = method suitable; pink = less suitable; white = not applicable.
From IEA GHG (2010).
6
volume utilization. Research related to pressure a high degree of accuracy, but quantification of the
buildup and brine displacement has demonstrated amount in place in the subsurface is difficult, if not
that injection-induced pressure changes can propa- impossible, to measure with any accuracy. These tech-
gate in the injection aquifer much farther than the niques do, however, allow for the identification of the
CO2 plume itself and can therefore limit storage CO2 plume and will provide a high degree of certainty
capacity. Geomechanical effects of pressure buildup on the movement of the plume and hence the predict-
can include microseismicity and ground deforma- ability of the plume migration in the subsurface for
tion, but pressure management strategies exist. The eventual handover to public authorities.
knowledge base on capillary, or residual, trapping The first requirement of any monitoring program is
of CO2 has increased substantially. Laboratory to create a baseline such as was undertaken at Sleipner
observations confirm that CO2 will occupy at least (Chadwick et al., 2009) and Weyburn (Wilson and
10%, and more typically 30%, of the pore volume. Monea, 2004) against which future surveys can be com-
Dissolution of CO2 at the interface between CO2 pared. This is quite consistent with standard oil field
and aquifer water can be significantly accelerated. practices using a variety of techniques to understand the
Taken together, these factors have the effect of processes operating in the subsurface and applying his-
reducing the amount of free-phase mobile CO2 tory matching to predict future production. Indeed,
and increasing storage security because dissolved with enhanced oil recovery, the use of time-lapse seis-
CO2 is no longer buoyant and prone to leakage. mic surveys has become relatively routine (Weyburn,
In summary, the reviews in the Special Issue Wilson and Monea, 2004) and this type of process has
updating the IPCC SRCSS indicate that CO2 storage been applied in storage areas such as Sleipner
is by and large a safe operation if storage sites are (Chadwick et al., 2009). This process is clear in the
properly selected, characterized, and managed. This chapters, particularly the case studies, in this book.
should go a long way to alleviating concerns by policy Recognizing the need to ensure storage integrity
makers, the public, and other stakeholders’ concerns requires that appropriate monitoring takes place and,
that geological storage is a safe and permanent option as noted, that all changes are referenced back to a
from the removal of CO2 from the atmosphere. baseline set of data. Table 1.2 identifies the monitor-
ing technologies that can be used. These technologies
1.5 Ensuring Storage Integrity and include surface techniques to measure possible CO2
leakage at the surface (ocean bottom or land surface),
the Need for Monitoring of Injected CO2 shallow CO2 presence (particularly in potable water
Storage integrity has been and will continue to be a zones), and remote sensing of the deeper subsurface
critical consideration for the storage of CO2 in the to identify leaks and leakage pathways early. These
subsurface. In part this is a public issue, as people pathways are identified in the IPCC SRCCS 2005
have concerns about the safety of storage, and in part report. While direct measurement techniques can be
this relates to the ability to demonstrate predictable applied at the surface and near-surface, this is not
conditions in the subsurface prior to any transfer of effective in the deeper subsurface. It is here in the
the project to the public sector in the future. To do this, subsurface that geophysical techniques, and particu-
good monitoring of the project is needed to determine larly seismic surveys, demonstrate their value to
the fate of the injected CO2 (IPCC SRCCS, 2005). ensure that we have storage integrity and that we
Table 1.2, taken from the IEA Greenhouse Gas can predict movement of the CO2 in the subsurface.
R&D Programme, is a general list of the techniques Seismic technology also allows for an evaluation of
available for monitoring the state of the CO2 plume potential leaks and leakage pathways, particularly
and, in particular, evaluating possible leakage from the understanding such features as sealed versus open
storage horizon. The table identifies a number of points faults, thinning of caprocks, and other potential nat-
salient to this book. The first is the usefulness of geo- ural hazards that may result in CO2 crossing forma-
physical techniques both onshore and offshore. The tional boundaries and migrating toward the surface,
second is the lack of applicability of all the techniques economic zones, or potable water zones. Active mon-
when it comes to quantifying the amount of CO2 that itoring is a way of understanding the flow path of the
exists in the subsurface. In other words, the amount CO2 and the potential for it reaching man-made or 7
of CO2 entering the subsurface can be measured with natural routes out of the storage formation.
1.6 Practical Experience in various geological formations within a wide range

of depositional environments
Monitoring CO2 Storage Formations 3. Development Phase (2008–2018+):
At the time of the IPCC SRCCS there were three opera- Implementation of large-scale field testing
tional commercial scale CO2 injection projects globally: involving at least 1 million metric tons of CO2 per
Sleipner in the North Sea, the Weyburn CO2 EOR project
Project (with an associated research project, the The RCSP program, as of April 2010, has six
IEAGHG Weyburn CO2 Capture and Storage Project) operational 1 Mtonne CO2 injection projects, with
in Canada, and the In Salah Project in Algeria. Two of two more in preparation. This is the largest geological
these, Sleipner and InSalah, were injecting CO2 into storage demonstration program in the world
deep saline formations, while Weyburn was a CO2 (IEAGHG, 2013b).
flood in a depleting oil field (IEAGHG, 2011a). Together the pilot projects around the world have
All three of these projects had substantive monitor- played a fundamental role in enhancing our substan-
ing projects running alongside their commercial tive knowledge of monitoring the storage integrity of
operations. the CO2 storage component of the CCS system. Once
Since that time the number of CO2 injection pro- again, the experience gained in monitoring over the
jects has grown considerably. A recent analysis under- last 10 or so years has been summarized in the Special
taken by the IEA Greenhouse Gas R&D Programme Issue of the International Journal of Greenhouse Gas
has shown that there were, as of mid-2012, 45 small- Control (IJGGC, 2015). The present book seeks to
scale injection projects and 43 large-scale projects enhance our understanding of the seismic aspects of
(IEAGHG, 2013b). Small-scale projects were consid- monitoring still further.
ered to be those injecting less than 100 000 tonnes, The Special Issue of IJGGC has shown that
though many projects inject considerably less (less
than 15 000 tonnes). Large-scale projects were inject- 1. Monitoring and verification have developed many
ing more than 100 000 tonnes/year. shallow monitoring methods in parallel with the
Figure 1.2 provides an overview of the global dis- assessment of environmental impacts, reflecting
tribution of the small or pilot CO2 injection projects. societal concerns about leakage to the near-
While there are several individual projects in surface.
Australia, China, Europe, and Japan, the majority of 2. Very significant progress has been made in the
these projects are in North America, principally in deep-focused monitoring techniques, particular
the USA. The reason for the large number of projects examples being marine seismic monitoring at
in the USA is that in 2003, the U.S. Department Sleipner and the combination of pressure and
of Energy (DOE) awarded cooperative agreements to seismic imaging at Snøhvit.
seven Regional Carbon Sequestration Partnerships 3. In the case of Sleipner, both conformance and
(RCSPs). The seven RCSPs were tasked to determine containment have been convincingly demonstrated
the best geological and terrestrial storage approaches by innovative analysis of an impressive data set.
and apply technologies to safely and permanently Snøhvit is a textbook case of pressure monitoring
store CO2 for their specific regions (Rodosta, 2016). detecting nonconformance and a mitigation
The RCSP Initiative has been implemented in strategy being successfully adopted.
three phases: 4. Another success for monitoring of reservoir-level
processes was at In Salah, where ground surface
1. Characterization Phase (2003–2005): Initial
displacements were detected by the new (in terms
characterization of their region’s potential to store
of application to CCS) method of interferometric
CO2 in different geological formations
synthetic-aperture radar (InSAR), the measurement
2. Validation Phase (2005–2011): Evaluation of
of ground surface displacement from satellite
promising CO2 storage opportunities through a
platforms. The interpretation of those results in
series of small-scale (less than 500 000 metric tons
terms of geomechanical processes was subsequently
CO2) field tests to develop understanding of
shown to be consistent with microseismic
injectivity, capacity, and storability of CO2 in the
observations and time-lapse seismic imaging,
8
Figure 1.2 Global distribution of pilot-scale CO2 injection projects as of December 2013.
all suggesting initiation and reactivation of regional level, which is important for the success of
fractures. projects and CCS globally.
The common theme to these examples, which has
now emerged from many projects at all scales, is the
ability of the available techniques for monitoring and
References
interpretation to test containment and conformance. Chadwick, R. A., Arts, R., Bentham, M., et al. (2009). Review
Shallow-focused monitoring methods have also been of monitoring issues and technologies associated with
exploited extensively and have played an important the long-term underground storage of carbon dioxide.
In Special Publications, Vol. 313: 257–275. London:
role in countering leakage allegations at Weyburn and Geological Society.
providing assurance that environmental impacts of
CO2CRC. (2011). A review of existing best practice manuals
hypothetical leakages are undetectable above natural
for carbon dioxide storage and regulation. A desktop
variability in key parameters. study prepared for the Global CCS Institute by
The pilot projects have led to the production of CO2CRC.
several best practice documents and guidelines, which
ENERGY.GOV. (2012). Regional partnership U.S.
vary in scope and technical detail. A number of non– Department of Energy Office of Fossil Energy.
site-specific best practice guides have also been pro-
European Council (2014). Conclusions on 2030 climate and
duced, such as the National Energy Technology energy policy framework. Copenhagen: European
Laboratory (NETL)’s risk assessment and site selection Environment Agency.
manuals and the World Resources Institute (WRI)’s
Houghton, J. T., Jenkins, G. J., and Ephraums, J. J. (Eds.).
CCS guidelines that outline the entire process. There (1990). Climate Change: The IPCC Scientific Assessment
are also best practice guidelines that consider learnings Report prepared for Intergovernmental Panel on Climate
taken from particular projects, such as the Saline Aquifer Change by Working Group I. Cambridge University
CO2 Storage Project (SACS) best practices for the sto- Press, Cambridge and New York.
rage of CO2 in saline aquifers, which uses learnings from IEAGHG. (2011a). Potential for biomass and carbon
the Sleipner storage site in the North Sea. Other exam- dioxide capture and storage. Report 2011/06.
ples of best practice guides are the QUALSTORE best Cheltenham: IEAGHG.
practice guide and the EU Guidance documents. There IEAGHG. (2011b). Global storage resource gap analysis
are several documents outlining issues regarding public for policy makers. Report 2011/10. Cheltenham:
communication including guidelines from NETL IEAGHG.
and WRI. The Global CCS Institute commissioned IEAGHG. (2012). Quantification techniques for CO2
CO2CRC to produce a summary of best practice guides, leakage. Report 2012/02. Cheltenham: IEAGHG.
including a summary of the varying areas of coverage IEAGHG. (2013 a). Potential implications on gas
and technical detail (CO2CRC, 2011). production from shales and coals for geological
Pilot projects have played a key role in helping storage of CO2. Report 2013/10. Cheltenham:
build public confidence in geological storage IEAGHG.
(Romanak et al., 2013). It is safety/integrity of storage IEAGHG. (2013b). The process of developing a CO2 test
sites that principally gets raised in public debates on injection: Experience and best practice. Report 2013/13.
CCS. These pilot projects have assisted through Cheltenham: IEAGHG.
• Establishing visitor centers at sites so the public IEAGHG. (2015). Carbon capture and storage cluster
can get first-hand experience of storage operations projects: Review and future opportunities. Report
2015/03. Cheltenham: IEAGHG.
• Enabling direct local dialogue with farmers and
other key stakeholders IGBP, IOC, SCOR (2013). Ocean acidification summary for
policymakers. In Third Symposium on the Ocean in a
• Enabling the public to meet the scientists involved High-CO2 World: International Geosphere-Biosphere
so people can learn and speak openly about their Programme. Stockholm: Ocean Acidification
concerns International Coordination Centre.
• Providing the opportunity to disseminate IJGGC (2015). Special Issue commemorating the 10th year
information at a local level anniversary of the publication of the Intergovernmental
Through these actions the pilot projects have Panel on Climate Change Special Report on CO2
10 helped build public confidence in CCS at a local/ Capture and Storage. Edited by J. Gale, J.C. Abanades,
S. Bachu, and C. Jenkins. International Journal of Change 2007: The Physical Science Basis. Contribution of
Greenhouse Gas Control, 40: 1–458. Working Group I to the Fourth Assessment Report of the
IPCC. (1988). Organization. www.ipcc.ch/organization/or Intergovernmental Panel on Climate Change [Solomon,
ganization.shtml#.UvyrqmJ_v_E S., D. Qin, M. Manning, Z. Chen, M. Marquis, K. B.
Averyt, M. Tignor, and H. L. Miller (eds.)]. Cambridge
IPCC. (2005). IPCC Special Report on Carbon Dioxide University Press, Cambridge, United Kingdom and
Capture and Storage. Prepared by Working Group III New York, NY, USA.
of the Intergovernmental Panel on Climate Change:
Cambridge University Press, Cambridge, United NOAA. (2016). Comment on recent record-breaking CO2
Kingdom and New York, NY, USA. concentrations. https://scripps.ucsd.edu/programs/kee
lingcurve/2016/04/20/comment-on-recent-record-brea
IPCC. (2007a). Summary for Policymakers. In Climate king-co2-concentrations/#more-1406
Change 2007: Impacts, Adaptation and Vulnerability.
Contribution of Working Group II to the Fourth OECD, IEA (2012). Energy technology perspectives 2012:
Assessment Report of the Intergovernmental Panel on Pathways to a clean energy system. Energy Technology
Climate Change [M. L. Parry, O. F. Canziani, J. P. Perspectives. Paris: OECD Publishing.
Palutikof, P. J. van der Linden, and C. E. Hanson OECD/IEA. (2016). Energy technology perspectives 2016:
(eds.)]. Cambridge University Press, Cambridge, Towards sustainable urban energy systems. Energy
United Kingdom and New York, NY, USA. Technology Perspectives. Paris: OECD Publishing.
IPCC. (2007b). Summary for Policymakers. In Climate Romanak, K., Sherk, G.W., Hovorka, S., and Yang, C.
Change 2007: Mitigation. Contribution of Working (2013). Assessment of alleged CO2 leakage at the Kerr
Group III to the Fourth Assessment Report of the Farm using a simple process-based soil gas technique:
Intergovernmental Panel on Climate Change [M. L. Implications for carbon capture, utilization, and
Parry, O. F. Canziani, J. P. Palutikof, P. J. van der storage (CCUS) monitoring. Energy Procedia, GHGT-
Linden, and C. E. Hanson (eds.)]. Cambridge 11 37: 4242–4248.
University Press, Cambridge, United Kingdom and Stern, N. (2006). Stern review on the economics of climate
New York, NY, USA. change (pre-publication edition). Executive Summary).
IPCC. (2013). Summary for Policymakers. In Climate HM Treasury, London. Archived from the original on
Change 2013: The Physical Science Basis. Contribution of January 31, 2010. http://mudancasclimaticas.cptec.inpe
Working Group I to the Fifth Assessment Report of the .br/~rmclima/pdfs/destaques/sternreview_report_com
Intergovernmental Panel on Climate Change [T. F. plete.pdf
Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, Whittaker, S., and Perkins, E. (2013). Technical aspects
J. Boschung, A. Nauels, Y. Xia, V. Bex, P. M. Midgley of CO2 enhanced oil recovery and associated
(eds.)]. Cambridge University Press, Cambridge, carbon storage. Docklands, VIC: Global CCS
United Kingdom and New York, NY, USA. Institute.
IPCC. (2014). Summary for Policymakers. Fifth Assessment Wilson, M., and Monea, M., eds. (2004). IEA GHG
Report (AR5). Cambridge University Press, Weyburn CO2 Monitoring & Storage project summary
Cambridge, United Kingdom and New York, NY, USA. report 2000–2004: 7th International Conference on
Johansson, T. B., Nakicenovic, N., Patwardhan, A., Gomez- Greenhouse Gas Control Technologies, Vancouver,
Echeverri, L., eds. (2012). Global Energy Assessment Canada. Vol. 3.
(GEA): Toward a Sustainable Future. Cambridge: World Bank Group. (2014). Turn down the heat:
Cambridge University Press. Confronting the new climate normal. Washington,
Le Treut, H., Somerville R., Cubasch U., et al. (2007). DC: World Bank. https://openknowledge.worldbank
Historical Overview of Climate Change. In Climate .org/handle/10986/20595
11
Chapter
The Role of Geophysics in Carbon Capture
2 and Storage
David Lumley
Introduction a potential future global market in CO2 sequestration

(Figure 2.1).
Geosequestration
Geosequestration is the capture, transport, injection,
and long-term storage of industrial carbon dioxide The Role of Geophysics
(CO2) in deep geological reservoirs, thus accelerating In addition to the initial storage site and reservoir
the Earth’s natural carbon cycle, while simultaneously characterization requirements, there will be a major
reducing manmade CO2 emissions to the atmosphere. demand for ongoing geophysical monitoring and ver-
Geosequestration represents one of the famous “cli- ification of CO2 injection during the life of CO2 sto-
mate stabilization wedges” proposed by Pacala and rage projects. As operators inject CO2 into the
Socolow (2004), and may be able to capture and subsurface, there will be a requirement to image and
store approximately 15% or more of the global man- monitor the evolving CO2 plume to ensure that it is
made CO2 emissions required to keep atmospheric being injected at the correct depths and locations, that
CO2 levels below 500 ppm. More details on the driv- it fills the storage reservoir efficiently as predicted,
ing forces and methods associated with geosequestra- that it does not interfere with other subsurface
tion in specific, and carbon capture and storage (CCS) resources (e.g., groundwater, hydrocarbons, geother-
in general, can be found in other chapters of this book. mal, etc.), that it does not flow toward high-risk areas
The basic approach for geosequestration is to cap- such as major faults with uncertain seismic and flow
ture CO2 at industrial point sources (e.g., coal-fired properties, and that it remains sealed in the storage
power plants, mineral and agricultural processing reservoir over time. Furthermore, in order for an
plants, cement plants, liquefied natural gas facilities, operator to effectively transfer any potential long-
etc.) and inject it into nearby deep geological forma- term liability associated with a CO2 sequestration
tions. Subsurface storage reservoirs include depleted project to a government agency, operators will likely
hydrocarbon reservoirs and saline aquifers. An inter- be required to develop a 3D subsurface flow model
national process is underway to map and rank subsur- that accurately predicts and matches the CO2 injec-
face storage potential for CO2 sequestration among tion and monitoring data for a significant period of
the world’s major sedimentary basins, and future time (e.g., 25–50 years), and thus “close the loop” in
work will require detailed geological and geophysical terms of reconciling the subsurface geology, flow
site characterization of these basins, reservoirs, and simulations, and geophysical images of the subsurface
saline aquifers in terms of defining storage capacity CO2 distribution. This will require advances in CO2
(volume, porosity, etc.), injectivity (permeability, sequestration geophysics, including core measure-
pressure regime, etc.), and sealing efficiency (struc- ments of the fluid-flow and geochemical effects of
tural and stratigraphic traps, impermeable geological CO2 on rock properties, computational simulations
layers and faults, pore-space CO2 trapping mechan- and imaging algorithms to predict and image the
isms and capillary pressures, geochemistry, etc.). For effects of CO2 injection in the subsurface, and inver-
example, Australia has recently become the first coun- sion of geophysical data to make accurate quantitative
try in the world to open up offshore exploration leases estimates of the amount of CO2 in the ground. Owing
12 specifically for the purpose of finding and testing to the complexity of CO2 geophysics, achieving these
subsurface CO2 storage reservoirs in preparation for goals will likely require combining the powerful
The Role of Geophysics in CCS
10ºS
Bonaparte
Browse
Carnarvon Canning
20ºS
Bowen
Eromanga
Surat
Perth
30ºS
Basin Ranking
Highly Suitable Gas Fields Otway Gippsland
Suitable Gas Pipelines Bass
0 1000 km
Possible Oil Fields
Field outlines and pipeline information
Unlikely Oil Pipelines are provided by GP Info, an Encom
40ºS Petroleum Information Pty Ltd product.
Unsuitable
110ºE 120ºE 130ºE 140ºE 150ºE 160ºE
Figure 2.1 Map of Australian sedimentary basins ranked for CO2 storage potential (Carbon Storage Taskforce, DRET, Australia, 2009).
http://hub.globalccsinstitute.com/publications/potential-impacts-groundwater-resources-co2-geological-storage/21-assessments-potential
subsurface monitoring technique of time-lapse seis- 2005; Johnson, 2013; other chapters in this book).
mology with complementary geophysical methods Time-lapse seismology involves repeating active-
such as passive seismology, electromagnetics (EM), source seismic surveys in time-lapse mode, or con-
gravity, and interferometric synthetic aperture radar tinuously recording passive seismic (micro) earth-
(InSAR) satellite imaging. quake and ambient noise data, in order to image and
Detailed CO2 injection case studies are presented estimate physical property changes in the subsurface
in other chapters of this book, which show specific over time. To first order, seismic body waves measure
examples of various topics presented in this overview the compressibility of the subsurface porous rock–
chapter on seismic monitoring. fluid system. Thus, in order to create a time-lapse
seismic signal, the fluid-saturated rock compressibil-
ity will likely change over time. In general, the more
Time-Lapse Seismology compressible the rock matrix is (e.g., unconsolidated
This chapter is intended to provide a comprehensive sand), and the larger the compressibility contrast
overview of the subsurface monitoring method between the fluids of interest (e.g., water vs. super-
known as time-lapse (or 4D) seismology; the reader critical CO2), or the larger the change in injection
is referred to the following publications for additional pore-pressure/stress, the larger the resulting time-
detailed review material (Lumley, 1995, 2001; Calvert, lapse seismic signal will be. Conversely, stiff rocks, 13
David Lumley
fluids with similar compressibility, and small injec- 3. Discuss how the foregoing can be used to perform
tion pressure/stress changes tend to produce weak or time-lapse simulation/modeling of seismic data
nondetectable time-lapse signals. The dependency of for a wide variety of CO2 injection scenarios.
the time-lapse seismic signal response to changes in 4. Review key aspects of time-lapse seismic
rock and fluid properties caused by CO2 injection will acquisition and imaging in terms of suppressing
be reviewed in this chapter, and additional details can sources of nonrepeatable noise, and enhancing the
be found in other chapters of this book. detection of time-lapse seismic signals generated
As with the measurement of any physical phe- by CO2 injection.
nomena, the ability to detect a time-lapse signal in 5. Review multiple strategies for time-lapse seismic
field data depends both on the magnitude of the signal inversion, and discuss their relative strengths and
and on the noise level in the data. The variable physi- weaknesses in terms of making accurate
cal properties of the injected CO2 with reservoir pres- quantitative estimates of physical properties
sure and temperature, and the variable properties of associated with CO2 injection (fluid saturation,
the reservoir rocks saturated with CO2–fluid mixtures pressure, etc.).
after injection, determine the strength of the time- 6. Discuss recent advances in time-lapse passive
lapse seismic signal. The quality of the time-lapse seismology that may allow us to monitor CO2
seismic data and images, especially the level of non- injection and storage reservoirs without the
repeatable noise and complexity of wavefields, deter- need for manmade active source efforts, by
mines whether the CO2 can be accurately detected, instead using seismic data from (semi)
imaged, and quantified. The largest source of noise in permanent sensor arrays that continuously
time-lapse seismic data tends to be nonrepeatable record passive (micro) earthquake energy and
noise, which results from the fact that we cannot (ocean) ambient noise.
perfectly repeat the seismic imaging experiment and
its environmental conditions from one survey to the
next. Issues related to nonrepeatability in time-lapse Theory
seismic data acquisition, image processing, and inver- It is helpful to review the general mathematical and
sion will be comprehensively reviewed in this chapter. physical theory that forms the basis of time-lapse
seismic monitoring. We begin with a simple equation
that describes how we can model synthetic seismic
Outline for This Chapter data:
This chapter is intended to provide a comprehensive
overview and discussion of time-lapse seismology as it d ¼ Ffmg ð2:1Þ
specifically relates to monitoring CO2 injection and
where d(x, t) are the seismic data traces for the source
storage, and summarizes my experiences gained from
and receiver positions located at x = (x, y, z), and
working on 200+ time-lapse seismic projects (of all
recording time t. The forward modeling operator F
types) during the past 25+ years. As much of this book
generates seismic data at the receivers given a seismic
focuses on CO2 injection projects in Europe and
source and an earth model m. F can be as simple as a
North America, I have included a significant amount
1D linear convolutional operator that convolves a
of Australian content in this chapter to add variety,
seismic wavelet with a reflection coefficient series,
having spent the previous decade as Professor and
or F can be a fully nonlinear 3D computational
Chair in Geophysics at the University of Western
operator that solves a complex wave equation includ-
Australia.
ing acoustic, elastic, viscoelastic, anisotropic, etc.
The outline for the remainder of the chapter pro-
seismic properties of the earth (e.g., Aki and
ceeds as follows:
Richards, 1980).
1. Review the fundamentals of time-lapse seismic The earth model m(x) contains the spatial distri-
theory that form the mathematical and physical bution of rock and fluid properties in the earth
basis for time-lapse seismology. required to generate seismic data. These rock and
2. Review the key time-lapse rock and fluid physics fluid properties typically include the elastic bulk
14 aspects of CO2 injection, and the relationships to modulus K (stiffness, incompressibility), elastic shear
key time-lapse seismic physical properties. modulus G (shear strength), and the density ρ. These
three fundamental seismic properties can be recom- numerical representation of the forward modeling
bined to calculate P- and S-wave propagation veloci- operator (e.g., acoustic versus elastic, convolution vs.
ties Vp and Vs, P- and S-wave impedances Ip and Is, finite difference, etc.), and also errors and approxi-
and other seismic physical properties (Mavko et al., mations (em) we make in representing and parame-
2009): terizing the earth model (e.g., choice of model mesh
qffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffi size, petro-physical relationships, etc.). This leads to
Vp ¼ Kþ4G=3 ρ ; Vs ¼ Gρ ; Ip ¼ ρVp ; Is ¼ ρVs : the important topic of uncertainty analysis, in which
errors and nonuniqueness in the modeling domain
For time-lapse seismic monitoring purposes, we con-
often make it challenging to compare synthetic data
sider that the earth model m is not static, but instead
to real data (signal + noise) in the data domain (e.g.,
may vary with calendar time τ (to distinguish from
Lumley, 2006). A full discussion of error analysis,
recording time t). We can thus write the time-varying
uncertainty analysis, and nonuniqueness is beyond
earth model as m(x, τ). There are many reasons that
the scope of this chapter. However, it is helpful to
the rock and fluid properties of the earth can vary over
keep in mind that these concepts arise in many
time, including but not limited to
important ways throughout time-lapse seismic data
• Changes in pore fluid type, saturations, or analysis, especially when trying to determine the
mixtures with fluid flow (e.g., CO2 injection) validity of a reservoir flow model by trying to
• Changes in pore fluid pressure, changes in tectonic match the observed time-lapse monitoring data at a
or confining stress specific CO2 injection and storage site.
• Changes in temperature due to advective heat/ Since the earth model is time-varying, we can
fluid flow, diffusion, geothermal activity therefore record different seismic data sets at different
• Geomechanical deformation, compaction, calendar times:
dilation, shearing, or fracturing d1 ¼ Ffmðx; τ 1 Þg
• Geochemical reactions due to changes in the rock– d2 ¼ Ffmðx; τ 2 Þg
fluid chemical equilibrium ð2:3Þ
…
• And other physical phenomena . . . dn ¼ Ffmðx; τ n Þg
The foregoing sources of time-varying rock and fluid and hence form a variety of time-lapse difference data
properties can be either natural, or manmade, and sets:
both may occur together. When we inject CO2 and
other fluids into deep geological rock formations, we Δd2;1 ¼ d2 d1
can change the rock and fluid properties in a time- … ð2:4Þ
varying manner, especially the pore fluid saturations, Δdk; j ¼ dk dj :
pore pressure, and fluid temperature. CO2 injection Note the time-lapse polarity convention used in Eq.
may also cause subsurface changes due to geochem- (2.4); I always recommend forming time-lapse data
ical and geomechanical effects. differences by subtracting an earlier survey from a
As an aside, it may be useful to mention that Eq. later survey . . . this convention proves to be very
(2.1) can be rewritten more generally as useful for subsequent time-lapse interpretations and
ðF þ eF Þfm þ em g ¼ d þ ed ; ð2:2Þ analysis. As Figure 2.2 shows, I also recommend that
positive differences be colored blue, and negative differ-
where e* are the respective errors in the earth model ences be colored red. Then, with these time-lapse
m, the forward modeling operator F, and the result- polarity and color conventions, a positive/blue differ-
ing data d. If we have the correct earth model, and ence typically implies that the reservoir rock has har-
generate synthetic seismic data to compare with real dened, for example, via increased water saturation
data, we know that the data sets may not match (hence blue), and a negative/red difference typically
because of noise present in the real data. However, implies that the reservoir rock has softened, for exam-
Eq. (2.2) also shows that even with noise-free data ple, via increased pore pressure, or increased gas/CO2
and the correct model, the synthetic and real data saturation (hence red).
sets may still not match because of errors and In addition to modeling data, we are also interested
approximations (eF) we make in the physics and in imaging and inversion of seismic data. Seismic
15
David Lumley
Baseline Monitor M-B
Red – Softening of Reservoir, caused by either:

- Gas coming out of solution
- Increase in pressure
Baseline Monitor M-B
Blue – Hardening of Reservoir, caused by either:

- Oil replaced by water or gas replaced by oil
- Decrease in pressure
Figure 2.2 Recommended time-lapse seismic polarity convention (Lumley, 1995, 1996). Graphic prepared in collaboration with Guy Duncan,
BHP Billiton.
imaging is the process of using seismic data to image

geological and fluid-flow features within the earth, for EðmÞ ¼ ½dobs dmod ðmÞ2 þ constraints ð2:6Þ
example the storage containment reservoir and the in which the data misfit error E(m) is iteratively
injected CO2 plume. Seismic inversion is the process minimized until a model m is found such that the
of using seismic data to make quantitative estimates of synthetic data matches the real data, within some
the physical properties of rocks and fluids within the error tolerance. Other optimization methods can be
earth, for example, the amount of residual CO2 used instead of least-squares; for example, nonlinear
trapped within the storage reservoir, and its spatial techniques based on Lp norms (e.g., L1), random
and temporal distribution. model realizations (e.g., Monte Carlo), statistical
We can mathematically invert Eq. (2.1) to estimate methods (e.g., Bayesian), genetic or evolutionary
an earth model from the seismic data as follows: algorithms, particle swarm optimization, etc. These
m ¼ F −1 fdg ð2:5Þ and related optimization approaches form the basis
–1
for many seismic inversion methods, ranging from
where F is known as the inverse operator. For simple 1D poststack acoustic impedance inversion to
simple forward modeling problems, for example complex 3D and 4D wave-equation full waveform
when modeling a set of data points yi using the inversion (FWI).
equation of a line yi = mxi + b with model parameters A classical solution to the least-squares problem
slope m and intercept b, we can easily find an inverse defined in Eq. (2.6) can be written as
operator F–1 that accurately estimates the model
parameters (m, b) from the data. Unfortunately, for m ¼ ðF FÞ−1 F d ¼ H −1 g ð2:7Þ
most realistic seismic data problems it is impossible where F* is known as the adjoint operator to the
to explicitly estimate F–1 because the seismic pro- forward modeling operator F. Any adjoint operator
blem is too large, ill-posed, underconstrained, non- F* is formally defined by the representation theorem
unique, and thus the inverse operator may not in fact < F*d, m > = < Fm, d >, where < a, b > is the
exist. For these reasons, we instead implicitly deter- mathematical inner (dot) product between two quan-
mine the inverse operator by finding a model m that tities a and b. For seismic wave propagation, the adjoint
generates synthetic seismic data dmod which matches, operator F* propagates seismic data in reverse time
in some sense, the observed real data dobs. A com- from the surface receivers back into the earth. In opti-
mon approach to solving this inverse problem uses mization theory, the term F*d is known as the gradient
one of many variations of the least-squares (L2 g of the misfit energy E(m), and the term F*F is known
16 norm) data fitting method: as the Hessian H (curvature) of E(m). If the adjoint
operator F* is a reasonable approximation in some the seismic data into signal and noise, d = s + n; if the
sense to the inverse operator F–1, then H–1 is approxi- noise sources n1 and n2 are not repeatable (equal),
mately unitary, and Eq. (2.7) reduces to then the errors they cause in the image/model results
m ≈ F d: ð2:8Þ will not cancel, and in fact generally become
statistically additive. This means that with the
The equation above shows that the adjoint F* is some Parallel method, the error in the difference
sort of seismic imaging operator which uses the data image/model Δm will be larger than the individual
d to form an estimated “image” of the earth model m. image/model errors caused by nonrepeatable seismic
It can be shown that F* makes an estimate of only the noise.
high-wavenumber components (reflectivity) of the For the Double Difference (DD) method, the time-
full earth model, and thus only an image of the reflec- lapse data sets are not imaged/inverted separately;
tive boundaries caused by geological layer boundaries instead, the difference of the data sets is imaged/
(e.g., a CO2 storage reservoir) and fluid contacts (e.g., inverted directly:
an injected CO2 plume) in the earth. The imaging
operator F* cannot directly recover the low- Δm2;1 ¼ F fΔd2;1 ; m0 g: ð2:10Þ
wavenumber components (smooth velocity) of the The DD method is more robust to nonrepeatable
full earth model, but the imaging process can be noise than the Parallel method, especially if the
repeated in an iterative manner until a smooth velo- noise is random and uncorrelated. In this case the
city model is found which generates a well-focused imaging/inversion operator is able to somewhat “see
seismic image. This iterative process forms the basis through” the nonrepeatable noise and extract
for modern integrated seismic image processing and the correct image/model difference by taking advan-
velocity analysis. tage of the constraint that most of the image/model
Given multiple time-lapse seismic data sets, remains unchanged between the time-lapse surveys.
it becomes evident that there are multiple strategies However, the DD method is highly sensitive
available for time-lapse imaging and inversion, to changes in time-lapse acquisition, especially non-
including the following methods (Lumley, 1995, repeatable source-receiver positioning and environ-
1996; Lumley et al., 2003a; Kamei and Lumley, mental conditions. The DD method assumes that the
2017): time-lapse data sets share identical geometries, and
“Parallel” imaging/inversion identical wave paths, in order to form accurate
•
difference data Δd2,1. In practice for many real-life
• “Double Difference” imaging/inversion
scenarios these assumptions are not valid; for
• “Bootstrap” or “Sequential” imaging/inversion
example, marine streamer positioning is notoriously
• “Simultaneous” imaging/inversion nonrepeatable, and land data surveys can be subject
For the Parallel method, the time-lapse data sets to changing near-surface conditions (dry, wet), and
are imaged/inverted independently in parallel using permit variations regarding where sources and recei-
an initial reference velocity model m0, after which the vers can be deployed from survey to survey.
desired image/model differences are created by sub- For the Bootstrap method, the time-lapse data sets
tracting the results: are imaged/inverted in a special sequential manner,
m1 ¼ F fd1 ; m0 g by estimating the incremental changes between each
m2 ¼ F fd2 ; m0 g ð2:9Þ seismic survey, while simultaneously constraining the
Δm2;1 ¼ m2 m1 overall image/model to be similar:
m1 ¼ F fd1 ; m0 g
where it is henceforth understood that F* is the
m2 ¼ F fd2 ; m1 g ð2:11Þ
imaging operator for seismic imaging, or the inverse
Δm2;1 ¼ m2 m1 :
operator F–1 for seismic inversion. The Parallel
method can accommodate nonrepeatable source The Bootstrap method has proven to be fairly robust
and receiver acquisition geometries for d1 and d2, to both nonrepeatable seismic data noise, and non-
but is highly sensitive to nonrepeatable seismic repeatable acquisition geometry conditions, as will be
noise conditions. We can see this by decomposing shown later.
17
David Lumley
Finally, the Simultaneous method solves for the • Rock geomechanical properties caused by CO2
time-lapse image/model estimates using the time- injection
lapse data sets in a simultaneous manner: • Rock and fluid geochemical reactions caused by
m1 ; m2 ¼ F fd1 ; d2 ; m0 g CO2 injection
ð2:12Þ To first order, seismic waves primarily measure the
Δm2;1 ¼ m2 m1 :
elastic stiffness (or compressibility) of the subsurface
The Simultaneous method is the most robust of all porous rock–fluid system. Therefore, in order to
the methods described earlier because it performs a create a time-lapse seismic signal, the fluid-saturated
highly constrained global imaging/inversion proce- rock compressibility is likely to change over time. In
dure using all of the time-lapse data sets together, general, the more compressible the rock matrix (e.g.,
rather than a localized step-by-step imaging/inver- unconsolidated sand), the larger the pore pressure
sion procedure one data set at a time. The change, and the larger the compressibility contrast
Simultaneous method is not widespread but has between the fluids of interest (e.g., water and injected
been used successfully in time-lapse image proces- CO2), the larger the resulting time-lapse seismic sig-
sing for at least a decade (Lumley et al., 2003a), and is nal. Conversely, stiff rocks, small pore pressure
an active area of time-lapse inversion research today changes, and/or fluids with similar compressibility
(e.g., Kamei and Lumley, 2017, among others). properties, produce a much weaker time-lapse sig-
nal. Injected CO2 tends to be much more compres-
Rock and Fluid Physics sible than formation water, but may be of a similar
The time-lapse seismic responses to changes in the compressibility to hydrocarbon gas (methane) and
earth caused by CO2 injection are controlled primarily high-GOR (live) oil. In situations where storage
by the physical properties of the dry and fluid-saturated reservoir rocks are sufficiently soft, CO2 injection
rock, and of the injected CO2 at reservoir pressure and into saline aquifers is likely to produce a strong
temperature conditions. The three primary elastic rock time-lapse signal, whereas CO2 injection
properties to be considered are the bulk modulus (stiff- into depleted hydrocarbon reservoirs is more com-
ness, incompressibility) K, the shear modulus (shear plex and may produce a much weaker time-lapse
strength) G, and the density ρ. These three elastic signal.
parameters result from an approximation that rocks The dry and saturated rock properties of a CO2
behave like a linear isotropic elastic material, which by storage reservoir can be determined by core and fluid
definition returns to its original state after weak defor- lab measurements, geophysical borehole logging mea-
mation by an applied force (e.g., a seismic wave). surements, and theoretical rock and fluid physics
Additional seismic rock and fluid physics properties relationships.
may also change over time, in both a linear elastic and
nonlinear nonelastic manner (e.g., anisotropy para- Seismic Properties of Dry Rocks
meters, the attenuation factor Q, etc.), but for CO2 The dry rock bulk and shear moduli (K, G) are a
injection these changes are often second order in com- strong function of the rock’s porosity φ and grain
parison to (K, G, ρ) and thus will not be discussed in cementation c, among other factors. For narrow por-
much detail here (but are of great interest in the time- osity ranges K and G are approximately linear with
lapse seismic research community). porosity ðK ∼ Kmax kφÞ, and can be accurately esti-
For CO2 injection projects, the most likely sources mated from core measurements and well logs, per
for changes in the primary rock and fluid physical rock facies type. As shown in Figure 2.3, for large
properties will be caused by changes in porosity ranges K and G are approximately inversely
• Rock pore fluid saturations and mixtures due to proportional to porosity (K ~ k/φ) and fall between
CO2 injection the Reuss lower bound [an iso-stress harmonic aver-
• Rock pore fluid pressure and rock stress age (Σ 1/K) of the constituent K values] and the upper
conditions due to CO2 injection Voigt bound [an iso-strain arithmetic average (Σ K) of
• Rock and fluid temperature due to (cold) CO2 the constituent K values]. For rocks with somewhat
injection into a (hot) reservoir spherical grain shapes, the Hashin–Shtrikman
18
6000 Figure 2.3 Plot of P-wave velocity versus Porosity, showing various
core sample measurements falling within the with Voigt and Reuss
bounds (Mavko et al., 2009).
5000
P-Velocity (m/s)
4000
3000 Voigt Avg.
2000
Reuss Avg.
(Wood’s Relation)
1000
0 20 40 60 80 100
suspensions Porosity (%)
sand-clay mixt.
sand
clay-free sandstone
clay-bearing sandstone
material physics bounds for K and G can provide a previous models such as Hertz–Mindlin and others.
more narrow and accurate set of bounds than the It is worth noting that injection of highly compressi-
Voigt–Reuss approach. The Hashin–Shtrikman ble supercritical CO2 is unlikely to cause increases in
method can also accommodate the inclusion of grain pore pressure much greater than 0–5 MPa, and thus
cement, as can the improved cementation model by the pressure effects will typically be much smaller than
Dvorkin and Nur (1996). Further details regarding water injection, which can easily create pore pressure
various rock physics K, G bounds can be found in increases of 5–35 MPa or more.
Mavko et al. (2009). The dry rock density ρ is approximately linear
The elastic dry rock bulk and shear moduli (K, G) with porosity, and a function of the volume-weighted
can also vary with effective (differential) stress σij, or arithmetic average of the individual density values of
more simply, effective pressure Peff ¼ ðPconf Ppore Þ, the mineral grain constituents. The dry rock density is
where Pconf is the confining (overburden) pressure, easily measured in a core lab, or via borehole density
and Ppore is the pore pressure of the fluid in pore and saturation logs, and can be accurately fit with a
space. A rock tends to weaken/soften with increasing linear trend versus porosity. To first order, the dry
pore pressure such that K (and G) follow nonlinear rock density does not vary much with pressure; how-
(approximately exponential) relationships of the form ever, there are often exceptions when the rock is
KðPeff Þ ∼ Kmin þ αKmax ½1 expðPeff =Pmax )], extremely unconsolidated and compressible (e.g.,
Meadows et al., 2005; Saul and Lumley, 2013).
where Kmin is the K value at the highest pore pressure
when the rock begins to hydraulically fracture (the
“frack point”), and Kmax is the K value at the lowest Seismic Properties of Pure and Impure CO2
pore pressure when all of the micro-cracks in the rock For most CO2 storage reservoirs, the pressure and
have been closed under confining pressure such that temperature regime will be, by design, such that the
the rock cannot become elastically stiffer. Saul and injected CO2 will exist in the supercritical state of its
Lumley (2013) give a good discussion of the pressure phase diagram, thus having the complex supercritical
sensitivity of reservoir rocks, both elastic and none- physical properties of both a gas and a fluid.
lastic, and provide a new formulation that fits a wide Supercritical CO2 is approx. 500 times denser than
19
range of core lab data much more accurately than gaseous CO2 and thus much more efficient for CO2
David Lumley
Table 2.1 PCC CO2 mixture composition power plant capture process. Note that the compres-
Fluids Mole fractions
sibility and density of pure CO2 can vary as much as
one order of magnitude across the storage reservoir
CO2 0.938 pressure–temperature range. CO2 is much more com-
SO2 0.005 pressible than water ðKwater ∼ 2 GPaÞ but can have a
Nitrogen 0.038 similar large range of variability as hydrocarbon oil
Argon 0.007 and gas, and the density of CO2 varies from gas-like to
Oxygen 0.01
oil or water-like conditions depending on the exact
pressure and temperature. Adding impurities such as
Methane 0.002
SOx, NOx, and methane to pure CO2 shifts and blurs
Composition of CO2 + impurities resulting from an end-member the critical point of the CO2 mixture properties in the
example of a low-efficiency Post-Combustion Capture (PCC) detailed phase diagram, from (Pc, Tc) = (7.39 MPa,
process. 31.10°C) for pure CO2 to (8.32 MPa, 28.39°C) for CO2
with the PCC impurities listed in Table 2.1, but the
overall range of compressibility and density variation
storage. The critical-point pressure and temperature remains similar.
of pure CO2 occurs at (Pc, Tc) = (7.39 MPa, 31.10°C),
and can vary somewhat for CO2 mixtures that contain Seismic Properties of Injected CO2 + Fluid
small amounts of impurities. For typical geothermal
gradients (approx. 25°C/km) and hydrostatic pressure Mixtures
gradients (approx. 10 MPa/km), CO2 storage reser- When CO2 is injected into a storage reservoir, it mixes
voirs must therefore be deeper than approx. 1 km with the preexisting pore fluids (brine, oil, gas, etc.)
depth in order to ensure that the injected CO2 The bulk modulus Kfluid of the resulting CO2 + fluid
remains in a dense supercritical fluid state. mixture can be calculated using Wood’s Law, which is
A liquid natural gas (LNG) plant, using a liquid a harmonic average of the saturation-weighted
P con-
amine capture process, results in a CO2 stream that is stituent fluid values ð1=Kfluid ¼ Si =Κ i Þ, and the
typically >95% pure CO2. The bulk modulus and density of the CO2 + fluid mixture ρfluid can be calcu-
density (and other physical properties) of pure CO2, P average of the constituent fluid
lated by the arithmetic
can be calculated using Equation of State (EoS) meth- densities ðρfluid ¼ Si ρi Þ.
ods with algorithms publicly available from National Typical maximum values of residual/trapped CO2
Institute of Standards and Technology (e.g., saturation are on the order SCO2 = 0.2–0.6.
SUPERTRAPP, Span-Wagner), GERG-2008, and the CO2 is likely to be injected and stored in saline
more recent EoS–CG method developed specifically aquifers containing brine (very salty water), or in
for CCS applications. A coal-fired power plant, using depleted hydrocarbon (HC) reservoirs containing a
a low efficiency post-combustion capture (PCC) pro- mixture of residual oil, water, and/or HC gas. In the
cess, may generate a CO2 stream that contains resi- case of saline aquifers, the injection of CO2 into brine
dual impurities such as SOx, NOx, nitrogen, argon, will make the resulting fluid mixture much more
and methane (e.g., Table 2.1). The physical properties compressible and somewhat less dense, depending
of CO2 mixtures that contain such impurities can be on the exact pressure and temperature. Because
calculated most accurately using the new EoS–CG time-lapse seismic is sensitive to the fluid compressi-
model (Gernert and Span, 2016), which is a recently bility contrast of the in situ fluid versus the injectant,
published improvement to the GERG-2008 EoS CO2 injection into saline aquifers is highly favorable
model. for seismic monitoring given sufficiently soft reservoir
Figure 2.4 shows plots of the CO2 bulk modulus rocks (e.g., Sleipner).
KCO2 and CO2 density ρCO2 versus fluid pressure and In depleted HC reservoirs, the situation is more
temperature calculated using the EoS–CG method, complex because oil has widely variable physical
for two realistic end-member CO2 capture cases: properties, and CO2 reacts with oil over short time
pure CO2 representing an ideal liquid-amine capture scales to change the physical properties of the residual
process for an LNG plant, and post-combustion cap- oil, which is why CO2 injection is a well-established
20
ture (PCC) representing a low-efficiency coal-fired method for enhanced oil recovery (EOR). Because
Bulk Modulus Difference [MPa]

Pure CO2 Bulk Modulus [MPa] PCC Mixture Bulk Modulus [MPa] (PCC mixture - Pure CO2)
40 40 700 40 0
700
35 35 35 –20
600 600
–40
Pressure [MPa]
Pressure [MPa]
Pressure [MPa]
30 30 30
500 500 –60
25 25 25
400 400 –80
20 20 20
300 300 –100
15 15 15
200 –120
10 200 10 10 –140
5 100 5 100 5 –160
20 40 60 80 100 20 40 60 80 100 20 40 60 80 100
Temperature [0C] Temperature [0C] Temperature [0C]
Bulk Modulus Difference [MPa]

Pure CO2 Bulk Modulus [MPa] PCC Mixture Bulk Modulus [MPa] (PCC mixture - Pure CO2)
0
14 700 14 700 14
–20
12 600 12 600 12 –40
Pressure [MPa]
Pressure [MPa]
Pressure [MPa]
10 500 10 500 10 –60
8 400 8 400 8 –80
300 300 –100
6 6 6
–120
200 200
4 4 4 –140
100 100 –160
2 2 2
10 20 30 40 50 10 20 30 40 50 10 20 30 40 50
0 0 0
Temperature [ C] Temperature [ C] Temperature [ C]
Density Difference [kg/m3]

Pure CO2 Density [kg/m3] PCC Mixture Density [kg/m3] (PCC mixture - Pure CO2)
40 1000 40 1000 40
35 900 35 900 35 –100
800
Pressure [MPa]
Pressure [MPa]
800
Pressure [MPa]
30 30 30 –200
700 700
25 25 600 25 –300
600
20 500 20 500 20 –400
15 400 15 400 15 –500
300 300
10 10 10 –600
200 200
5 100 5 100 5
–700
20 40 60 80 100 20 40 60 80 100 20 40 60 80 100

Density Difference [kg/m3]
Pure CO2 Density [kg/m3] Pure CO2 Density [kg/m3] (PCC mixture - Pure CO2)
14 1000 14 1000 14
900 900 –100
12 12 12
Pressure [MPa]
Pressure [MPa]
Pressure [MPa]
800 800 –200

10 700 10 700 10
600 600 –300
8 500 8 500 8 –400
6 400 6 400 6 –500
300 300
4 200 4 200 4 –600
2 100 2 100 2 –700
10 20 30 40 50 10 20 30 40 50 10 20 30 40 50
Figure 2.4 Bulk modulus (top rows) and density (bottom rows) of pure CO2 (left), and CO2 with impurities after post-combustion capture
(PCC) (center), and the difference (right), as a function of pressure and temperature. The second and fourth rows are zooms of the first and third
rows respectively. (Figures prepared in collaboration with Paul Connolly and Eric May, University of Western Australia.)
time-lapse seismic is sensitive to fluid compressibility when the residual oil has a low GOR value, for exam-
contrasts, the most favorable conditions for seismic ple, at the Weyburn project (GOR = solution gas/oil 21
monitoring of CO2 in depleted HC reservoirs are ratio; i.e., the amount of dissolved HC gas in the oil).
David Lumley
In contrast, conditions which are less favorable for alternately, saturated (versus dry) core samples can
seismic monitoring of CO2 in a depleted HC reservoir be measured directly in the lab while being core-
occur when the residual oil is high-GOR and thus may flooded with CO2 to simulate the CO2 injection pro-
be of a similar compressibility to the injected CO2, or cess, but this requires an advanced lab setup and
when there is highly compressible residual HC gas expertise that is commonly found only in research
saturation in pore space (e.g., the CO2CRC Otway environments. Figure 2.5 shows P- and S-wave velo-
Stage 1 project). city measurements versus effective pressure for an
Otway core sample under dry, brine-saturated, and
Seismic Properties of Rocks Saturated with CO2-saturated lab conditions; note that Vp decreases
Mixtures of CO2 and In Situ Fluids/Gas approx. 5% from 3550 m/s at the original brine-
saturated conditions, to 3390 m/s after core-flooding
Once we have calculated the bulk modulus Kfluid of the to a maximum CO2 saturation of SCO2 = 0.5 (Lebedev,
CO2 + fluid mixture, the saturated rock bulk modulus 2012).
Ksat can be calculated by combining Kfluid with the dry As we have discussed, time-lapse seismic is pri-
rock properties using the well-known Gassmann marily sensitive to the elastic compressibility of the
equation (e.g., Mavko et al., 2009). The saturated dry rock frame, and the compressibility contrast of the
density is simply ρsat ¼ ρdry þ φ ρfluid . Gassmann the- pore fluid properties. A strong time-lapse seismic
ory works well when the reservoir rock fabric and signal requires rocks with high values of porosity
fluid distributions are fairly homogeneous, for exam- and dry-frame compressibility (i.e., low Kdry). High
ple, a clean well-sorted sandstone with a uniform porosity increases rock compressibility and the
distribution of injected CO2. In addition, or volume of injected CO2 in the rock to better affect
Figure 2.5 Otway core lab measurement data showing P-wave velocity of dry rock versus effective pressure (top left), brine-saturated rock
versus effective pressure (top center; dashed line is Gassmann theory prediction from dry measurements), CO2-saturated rock versus effective
22 pressure (top right), and CO2-saturated rock versus CO2 saturation values (bottom); (Lebedev 2012).
Vp - CO2 Saturation (Critical saturation : 0.6) Figure 2.6 P-wave velocity versus CO2 saturation.
Time-lapse geophysical well log data (blue dots) mea-
2.8
uniform sured during the Nagaoka CO2 sequestration project
2.7 injection, exhibiting patchy saturation behavior (Lumley,
Patchy 2010; Caspari et al., 2011).
2.6
Voigt
2.5
Brie(e=3)
P-wave Velocity(km/s)
2.4
Brie(e=2)
2.3
log(18–26th)
2.2
2.1
2
1.9
1.8
1.7
1.6
1.5
0 0.1 0.2 0.3 0.4 0.5 0.6
CO2 saturation
the seismic response, and large dry-frame compressi- Patchy Saturation

bility allows the seismic waves to better sense (via fluid
A first major complication is the issue of patchy
compressibility contrast) the type of fluid in pore
saturation: the fact that the compressibility of a rock
space. Rocks that are favorable for time-lapse mon-
as a function of CO2 saturation depends on the spatial
itoring of CO2 injection include unconsolidated
size of the CO2 fluid “patches” in the rock pore space
sands, turbidites, and heavily fractured material, for
compared to the wavelength of the seismic measure-
example. At the Sleipner CO2 project, the injection of
ment. When the seismic wavelength is much larger
supercritical CO2 into brine-saturated unconsoli-
than the average CO2 patch size (i.e., low-frequency
dated sand creates a huge P-wave velocity (Vp)
limit), the saturated rock compressibility tends to
decrease of up to 60%. Rocks that are less favorable
follow the highly nonlinear Reuss lower bound.
for time-lapse monitoring of CO2 injection include
When the seismic wavelength is much smaller than
well-cemented sandstones, low-porosity tight sands,
the average CO2 patch size (i.e., high-frequency limit),
and hard carbonates, for example. At the Weyburn
the saturated rock compressibility follows a quasi-
CO2 project, the injection of CO2 into low-GOR,
linear upper bound; and when the average CO2
residual-oil-saturated carbonate rocks creates a small
patch size is similar to the seismic wavelength, the
Vp decrease of only a few percent, which is at or near
saturated rock compressibility plots along a “patchy
the time-lapse seismic noise level for detection. At the
saturation” curve in between the two bounds, as
CO2CRC Otway Stage 1 project, the depleted HC
Figure 2.6 shows for time-lapse log data measured
reservoir rocks were moderately favourable, but the
during the Nagaoka CO2 injection project (Lumley,
residual HC gas in pore space made it difficult to
2010; Caspari et al., 2011). The patchy-saturation
detect injected CO2 with time-lapse seismic due to
values for Ksat can be calculated using Hill’s equation
the weak fluid compressibility contrast between the
(Mavko et al., 2009), which tends to produce a fairly
injected CO2 and the residual HC methane gas.
linear relationship between the initial and final Ksat
values as a function of injected CO2 saturation. It is
Additional Complications . . . still an open research question as to whether patchy
It is worth mentioning a few important complications saturation effects measured in the lab at MHz fre-
to be aware of, with specific regard to time-lapse quencies, and in log data at 10–100 kHz, are a sig-
seismic monitoring of CO2 injection. nificant effect in surface seismic data at 10–100 Hz. 23
David Lumley
Figure 2.7 Geochemical fluid sample monitoring at 45 boreholes pre- (a, c, e) and post- (b, d, f) CO2 injection at the Weyburn project: (a, b)
Carbon-13 isotope ratio (‰); (c, d) dissolved calcium (mg/L); (e, f) alkalinity (mg/L). These plots show clear evidence that major geochemical
changes are occurring due to CO2 injection into the Weyburn carbonate reservoir, causing significant dissolution of calcium and possibly other
minerals (Emberley et al., 2004).
However, quantitative calibration of the velocity- porosity and weakening the cementation and/or dry
saturation relationship is very important for design- frame modulus. This is especially true for carbonate
ing high-quality monitoring experiments, interpret- reservoir rocks (e.g., Figure 2.7; Emberley et al., 2004),
ing field data results, and quantifying the amount of but may also be important for clastic reservoir rocks
CO2 present in the subsurface from time-lapse seismic that contain carbonate grains and/or cement (e.g.,
images and inversions. Figure 2.8; Vanorio, 2015). The injected CO2 can
react with calcium carbonate (CaCO3) or other
minerals that may be present in the reservoir rock to
Geochemical Effects produce soluble bicarbonate [Ca(HCO3)2] which is a
A second major complication arises because CO2 is much weaker material than the original hard carbo-
not an inert fluid. My own personal experiences mon- nate. In addition, the CO2 can react to produce car-
itoring CO2 injection for EOR and CO2 storage pro- bonic acid (H2CO3), which can further dissolve
jects over the years have shown that CO2 can react in minerals and grain contact cement to weaken the
the pore space to alter or dissolve the rock matrix or reservoir rock. Recent lab measurements on CO2-
24 pore-filling diagenetic material on very short time- flooded cores (e.g., Vialle and Vanorio, 2011;
scales (weeks or months), thus creating secondary Vanorio, 2015) seem to support our field observations
e)
Figure 2.8 Geochemical reaction effects in Tuscaloosa sandstone core samples from the Cranfield CO2 project: P- and S-wave velocity versus
effective pressure for (a, b) shaly sandstone and (c, d) cleaner sandstone; (e) SEM image of sandstone before (top) and after (bottom) CO2
injection (Vanorio, 2015). The open and closed symbols refer to the velocities of the dry frame measured, respectively, before and after the
injection of 37 PV of aqueous CO2. Note that the injection of CO2 causes dissolution and removal of iron-rich chlorite from the pore space, thus
weakening the sandstone rock matrix.
that CO2 can be geochemically reactive with the reser- diagenetic pore-filling material, thus weakening the
voir rock over seismic monitoring timescales, and rock matrix of Tuscaloosa sandstone cores associated
thus cause significant decreases in dry frame stiffness. with the Cranfield CO2 storage project. These com-
Figure 2.8 shows that CO2 injection can cause the bined geochemical effects can make it difficult to iso-
25
dissolution and removal of grain contact cement and late the effects of CO2 fluid saturation alone, which in
David Lumley
turn makes it difficult to accurately quantify the seismic difference data sets can achieve NRMS
amount of supercritical CO2 present in pore space values of <0.4. In this case, from my
using only seismic images. Many time-lapse seismic experience, reservoir P-impedance changes
data sets at CO2 injection projects seem to exhibit >5% are often clearly visible in the time-lapse
geochemical reaction effects in the reservoir, which seismic data sets.
may be why observed time-lapse CO2 anomalies are • Reservoir properties
sometimes quite different (typically larger) than pre-
dicted in advance using non-geochemically reactive • Let’s assume that the saline aquifer is at a depth
modeling methods. of approx. 2 km, with a normal hydrostatic
pore pressure of approx. 20 MPa and a
reservoir temperature of approx. 60°C.
Geomechanical Effects Figure 2.4 shows that the injected CO2 will
A third major complication can arise due to geome- clearly be in its supercritical phase.
chanical effects; reservoir rocks may undergo irrever- • Let’s further assume that the reservoir rocks
sible nonlinear nonelastic plastic, ductile, or brittle are clastic (e.g., sandstone) and sufficiently
mechanical deformation during CO2 injection due compressible (e.g., Kdry, Gdry <10 GPa) and
to changes in pore pressure, in situ stress conditions, porous (e.g., φ>0.2) such that supercritical
and/or geochemical reactive effects. Geomechanical CO2 saturations SCO2 >0.3 will create
effects include porosity loss or increase, grain rota- P-impedance changes >5%.
tion, displacement or cracking, cement bond damage,
compaction, dilation, stress arching, shearing, frac- Given the foregoing assumptions, we may expect the
turing, and other nonelastic phenomena. Perhaps following time-lapse seismic responses.
the best examples of geomechanical effects that have • Saturation effects
been observed in time-lapse seismic data are in the • Per Figure 2.4, the injected supercritical CO2
North Sea, where both geomechanical and geochem- will be much more compressible (approx.
ical effects of seawater injection into carbonate (espe- 600%) than the saline water, and somewhat
cially chalk) reservoirs can cause significant reservoir less dense (approx. 30%). In regions of
compaction, often accompanied by ductile stretching maximum CO2 saturation, especially near the
of the overburden rocks, and subsidence at the sea- injection well, the saturation effect will cause a
floor (e.g., Guilbot and Smith, 2002; Hatchell and significant “softening” (P-impedance and Vp
Bourne, 2005). decrease >5%) of the reservoir rock, a
moderate density decrease, and a small/
negligible shear strengthening (S-impedance
Nonlinearly Coupled and Combined Time- and Vs increase) due to the density effect.
Lapse Seismic Responses to CO2 Injection • If a significant volume of the injected CO2
Often the individual rock and fluid physics effects should flow upward past the top seal to a depth
discussed in the foregoing become combined in a shallower than 1 km, it would no longer
nonlinearly coupled manner to produce the resulting remain in its supercritical phase, and would
full time-lapse seismic response. To provide a more “flash” from a supercritical liquid to a highly
intuitive understanding for the potential complexities compressible low-density gas; this would cause
of the various time-lapse seismic responses, let us an extremely bright time-lapse seismic
consider the following hypothetical scenario of super- response in the overburden due to a major
critical CO2 injection into a saline aquifer storage decrease in Ip, Vp, and density (>10–20%).
reservoir.
• Pressure/stress effects
• Repeatability • CO2 injections are highly likely to be pressure-
• Let’s assume that the time-lapse seismic controlled so that important stress thresholds
surveys are sufficiently repeatable in terms of are not exceeded (e.g., at the Snøhvit and
acquisition geometry, noise and near-surface/ Aquistore projects, the injection rate is
26 environmental conditions such that the constrained such that the pore pressure does
not exceed 90% of the reservoir and cap rock • Since the saline aquifer at 2 km depth will be at a
fracture pressure values). In addition, because normal geothermal gradient temperature of
supercritical CO2 is highly compressible approx. 60°C (see earlier), the injected CO2 will
compared to water, only small pore pressure be approx. 25–30°C colder than the reservoir at
changes are anticipated due to CO2 injection. the injection point, and will progressively heat
However, unexpected pressure changes up as it flows away into the distant parts of the
sometimes can, and do, occur due to reservoir.
unanticipated geological complexity, • Injection of “cold” CO2 into a “hot” reservoir
decreased injectivity, reservoir may cause thermal fracturing of the rock near
compartmentalization, etc. the injection point, which may be visible as a
• In general, there will likely be a modest time-lapse “bulls-eye” anomaly centered on
(0–5 MPa) pressure increase near the injection the injection well, especially in S-wave or
well, which may moderately (<2%) or AVO/A data. This may explain the anomalous
negligibly soften the reservoir rock. If there are S-wave results in the time-lapse cross-well data
water production (withdrawal) wells to help acquired by Daley et al. (2008) for the Frio CO2
manage reservoir pressure, and/or to steer the injection project. Such time-lapse bulls-eyes
injected CO2 plume, then modest (0–5 MPa) are often seen in 4D seismic data at water
pressure decreases can be expected near the injection wells that inject very cold seawater
water production wells, but these will tend to into very hot hydrocarbon reservoirs.
have a negligible effect because rocks are • Injection of cold CO2 into hot saline aquifer
asymmetrically much more sensitive to water will advectively cool the brine, but this
pressure increases than pressure decreases. fluid temperature effect on its own is likely to
• If the reservoir rock was initially fractured, or if be weak or negligible (<1–2%).
the maximum horizontal confining stress is • Injection of cold CO2 into a hot reservoir will
much greater than the minimum horizontal diffusively cool the rock near the injection well
stress, then the reservoir may exhibit initial (beyond the thermal fracture radius), but
azimuthal anisotropy (HTI or orthorhombic). thermal diffusion is a very slow and gradual
The change in pore pressure due to CO2 process and so the time-lapse effect is also
injection can open or close fracture sets, and/or expected to be weak or negligible (<1–2%).
modify the effective confining stresses, which
• Geochemical effects
can thus cause the seismic anisotropy to change
over time. These time-lapse anisotropy effects • The injection of CO2 into a saline aquifer may
are most likely to be observed in S-wave data, cause geochemical reactions if the reservoir
especially shear-wave splitting, or in azimuthal sandstone contains carbonate minerals, and/or
amplitude versus offset (AVO)/A responses at carbonate (versus quartz) grain cement. In this
large offsets/angles which are sensitive to shear case the dry rock frame may weaken due to
impedance changes. This topic is an area of dissolution of grain contact cement, carbonate
active research, and few field data examples have mineral grains and/or carbonate pore-filling
been observed to date (e.g., Davis et al., 2003). material (e.g., Vanorio, 2015). These
geochemical reaction effects, if they occur,
• Temperature effects could easily cause a significant time-lapse
• Industrial CO2 is likely to be pressurized and decrease in the P-wave and S-wave velocities of
heated (if necessary) to become a supercritical 5–10% or more, which would be readily visible
fluid at its capture source point, and then in time-lapse seismic data.
transported in special pipelines to an injection
• Geomechanical effects
and storage site. Thus for energy efficiency
purposes, the temperature of the supercritical • The injection of CO2 into a saline aquifer is
CO2 at the injection point in the storage highly unlikely to cause significant
reservoir is likely to be slightly above the geomechanical stress effects since the injection
critical point, i.e., about 32–35°C. rate will likely be controlled to not increase the 27
David Lumley
Top Muderong
Lower Gearle Formation
Top Pyrenees (gas)

Depth (m)
Windalia Radiolarite
Gas gas-oil-contact
Oil
Muderong Formation
Pyrenees Sands oil-water-contact
West East
Figure 2.9 Idealized example of a seismic well tie. (Left) Marine seismic image of a hydrocarbon reservoir offshore Australia. (Right) Extracted
seismic trace overlain on background seismic image. (Figure courtesy of Guy Duncan and colleagues, BHP Billiton.)
original reservoir pressure by more than time-lapse seismic data at the Ekofisk Field in
0–5 MPa, and certainly to not exceed 90% of the Norwegian North Sea due to injection of
the rock fracture pressure value. seawater into a chalk reservoir and production
• However . . . sometimes unanticipated of hydrocarbons (Guilbot and Smith, 2002);
geological reservoir complexity and fault and measurable surface ground uplift of
compartmentalization can lead to strong approx. 10 mm/year or more near CO2
pressure transients, which may be detected in injection wells, and ground subsidence of
time-lapse seismic data as decreases in P- and approx. 5 mm/year near gas production wells,
S-wave velocity, and wide-aperture TL AVO/A has been observed in time-lapse InSAR
responses. Such unanticipated pressure satellite data for the CO2 injection project at In
transients may also induce seismicity (micro- Salah, Algeria (Mathieson et al., 2009).
earthquakes) on any faults within the • Combined effects
pressurized region of the injected plume, if the
faults are originally close to the “critical stress • Any of the foregoing individual effects can be
point,” i.e., the critical stress value at which a coupled and combined nonlinearly in any
particular fault is ready to slip and release number (few or all), and can vary in both space
seismic energy. and time, during the life of a CO2 injection and
• If injected CO2 geochemical effects have storage project. Hopefully this simple
significantly weakened the reservoir rock hypothetical saline aquifer example provides
matrix per the preceding discussion, the some intuition for the potential complexity
altered reservoir rock may be much more and nonuniqueness of the full time-lapse
susceptible to geomechanical stress effects seismic response to CO2 injection.
such as reservoir compaction and overburden
stretching in the area of the CO2 plume,
reservoir dilation near the CO2 injectors, and
Time-Lapse Seismic Modeling
perhaps even ground/seafloor subsidence or Given the preceding discussion regarding the physical
uplift. For example, extreme reservoir properties of the storage reservoir rocks, the in situ
compaction and seafloor subsidence (>10 m pore fluids, and the injected CO2, it is evident that
28 over 40+ years) has been observed in seismic modeling to predict the expected time-lapse
response can be performed for virtually any CO2 layer of CO2, and two layers of CO2, respectively. An
reservoir injection scenario. Time-lapse seismic mod- accurate time-lapse imaging algorithm would need
eling can be done at a single point in the reservoir the ability to correctly image these complex time-
(“zero-D” modeling), at a well location (1D model- lapse coda waveforms into a single (or double) layer
ing), along a reservoir cross section (2D modeling), of CO2 . . . quite a challenge!
and in a full 3D reservoir model coupled with the Figure 2.11 shows the result of applying a state-of-
results from a flow simulation (e.g., Lumley, 1996), the-art prestack depth migration (PSDM) image-
and optionally may include flow-coupled geochemical processing flow to the synthetic data for the single
and geomechanical simulations. Seismic modeling CO2 layer (blue arrow), including elastic wavefield
can range from simple 1D convolution of a seismic separation, multiple suppression, depth migration
source wavelet with a reflection coefficient series, to velocity analysis, etc. The synthetic seismic difference
full-blown Finite Difference (FD) or Finite Element image is compared to the real marine data (PSTM)
(FEM) modeling, and can represent the physics of seismic difference image at Sleipner, the latter of
first-order acoustic wave propagation (P-waves which is interpreted to contain many layers of CO2.
only), higher order elastic waves (P- and S-waves), The blue dot is the location of the Sleipner CO2
viscoelasticity (including Q attenuation), and various injection point. The imaging artifacts in the synthetic
forms of anisotropy (polar, azimuthal, orthorhombic, data, which look similar to the seismic differences
etc.). observed in the real data, are caused by strong multi-
A vital starting point to the time-lapse seismic ple wave-scattering and mode-conversion effects
modeling process is to ensure that a high-quality (time-lapse coda) that are not properly handled by
“seismic well tie” is achievable. A seismic well tie conventional wave-equation imaging operators, and
combines the geological overburden and reservoir thus can be easily misinterpreted as false layers of
model, the rock and fluid property relationships CO2, and therefore may contaminate the image of
from core and log data, and a seismic forward model- the true CO2 distribution (Lumley et al., 2008;
ing method, to generate synthetic seismic data that Lumley, 2010).
matches or “ties” with the actual seismic data at a Figure 2.12 shows an example of 3D FD modeling
given physical location (usually a borehole). If a using publicly available Sofi3D software to predict
good quality well tie is obtained, this builds confi- time-lapse synthetic seismic data with realistic addi-
dence that the earth model and physical property tive noise for Stage 2C of the CO2CRC Otway CO2
relationships are reasonably correct, and that the seis- injection project. The modeling process includes
mic data have been imaged correctly. If the synthetic building a detailed 3D geological model, and generat-
seismic does not match the real data, this highlights ing a flow simulation of the injected CO2 plume. The
that there are major discrepancies to be resolved in the 3D synthetic shot gathers replicate the identical
geological model, physical properties, core and log source and receiver geometries of the real seismic
data, or seismic image processing. Figure 2.9 shows surveys, and realistic noise is estimated from the real
an example of an extracted seismic volume trace data to be added to the synthetic data. The synthetic
representing an idealized seismic well tie from a reser- data sets are then passed through the same image
voir offshore Western Australia. processing flow as the real data in order to make
Figure 2.10 shows an example of 2D FD viscoelas- realistic predictions and comparisons regarding the
tic seismic wavefield data generated from an earth ability to detect, image and interpret the injected CO2
model based on Sleipner reservoir rock and CO2 plume at various volumes of injection, and under
properties, respectively containing no CO2, one layer various noise conditions (Glubokoskikh et al., 2016).
of CO2, and two layers of CO2, near the top of the Figure 2.13 shows an example of 3D FD anisotro-
Utsira unconsolidated sand reservoir. The panels pic viscoelastic modeling using custom developed
show the preinjection seismic shot gather, the differ- UWA software to generate full 9-component (9C)
ence between the preinjection and one-layer CO2 3D seismic elastic wavefield data sets. This example
gather, and the difference between the preinjection uses a full seismic source moment tensor, both for
and two-layer CO2 gather. Note the rich and complex manmade surface sources, and for natural or induced
wavefield scattering differences generated by just one (micro) earthquakes at depth, in a 3D geological 29
CDP
210 210 210
OFFSET
050 –1300 –550 200 950 1700 –1750 –1000 –250 500 1250 2000 –1450 –700 50 800 1550
200 T0 T1-T0 T2-T0 200
400 400
600 600
800 800
1000 1000
1200 1200
Time (ms)
Time (ms)
1400 1400
1600 1600
1800 1800
2000 2000
2200 2200
2400 2400
2600 2600
2800 2800
Figure 2.10 Reservoir model (top), and finite-difference elastic wavefield shot gathers (bottom) at same scale generated from the reservoir
model containing (left) no CO2, (center) one layer of CO2, and (right) two layers of CO2, near the top of an unconsolidated sand storage
reservoir. The left panel shows the preinjection seismic shot gather; the center panel shows the difference between the preinjection and one-
layer CO2 gather; and the right panel shows the difference between the preinjection and two-layer CO2 gather. Note the rich and complex
wavefield scattering differences generated by only one layer of CO2 (center) and two layers of CO2 (right), respectively. An accurate time-lapse
imaging algorithm would require the challenging ability to correctly image these complex time-lapse wave coda into a single (or double) layer
of CO2 (Lumley et al., 2008; Lumley, 2010).
receiver positioning differences between surveys; nat-

ural variations in the near-surface conditions on land
(wet/dry, water table, ground coupling, etc.); natural
variations within the water column at sea (tides, wave
heights, water temperature, salinity, etc.); and varia-
tions in source waveforms, receiver functions, and
equipment hardware specifications.
Many of the manmade nonrepeatability factors
can be reduced or eliminated by careful time-lapse
seismic survey planning and data preprocessing (e.g.,
source and receiver positions). Many of the natural
nonrepeatability factors cannot be similarly con-
trolled, but can be quantitatively measured and mon-
itored during each survey, and then used to
deterministically compensate or correct the data in
processing (e.g., tidal variations). However, it is
important to be aware that there will always be some
nonrepeatablity factors that we cannot easily control
or quantify (e.g., ambient noise, weather, overburden
Figure 2.11 (Left) Synthetic PSDM time-lapse seismic difference variations, etc.) and therefore there will always be
image from FD data using a reservoir model with only a single CO2 some level of nonrepeatable noise that we will have
layer (blue arrow). (Right) Marine data PSTM time-lapse seismic
difference image at Sleipner interpreted to contain multiple layers of to include and manage as part of our time-lapse seis-
CO2. The blue dot is the location of the true CO2 injection point. The mic imaging, inversion, and interpretation analysis.
imaging artifacts in the left panel are caused by strong multiple Nonrepeatable (NR) noise can be quantified in
wave-scattering and mode-conversion effects (time-lapse coda) that
are not properly handled by conventional wave-equation imaging many ways, but the most common measure is
operators, and thus may be misinterpreted as false layers of CO2, or known as “normalized RMS” or NRMS:
may contaminate the image of the real CO2 distribution (Lumley
et al., 2008; Lumley, 2010). 2 RMSðData2 – Data1Þ
NRMS ¼ RMSðData2Þ þ RMSðData1Þ
where RMS is the standard definition of “root-mean-

model consisting of more than 5 billion grid points, squares.” Kragh and Christie (2002) provide a very
for the South West Hub CO2 storage project, Western good discussion of NRMS with respect to measuring
Australia (Lumley et al., 2016). This type of realistic nonrepeatable noise in time-lapse seismic surveys.
complex 3D modeling requires a massive computa- Nonrepeatability can be measured at any stage of
tional effort, and thus the resources of a Top 100 time-lapse seismic acquisition and image processing,
world-wide high-performance computation (HPC) starting from raw shot gather traces, to final processed
facility, like the Pawsey Centre in Perth, Australia, image volumes. From my monitoring experience, a
which was used to create the Otway Stage 2C and reasonable rule of thumb is that the final processed
South West Hub project seismic simulations. image volume data can be expected to have an NRMS
value that is approximately half of the initial shot
gather NRMS values (with very basic filtering
Time-Lapse Seismic Acquisition applied).
The ability to image injected CO2 with time-lapse The normalization inherent in the NRMS calcula-
seismic data depends on the magnitude of the seismic tion is intended to provide an unbiased estimate
response to CO2 (i.e., the time-lapse signal), in com- between data sets, but in fact the NRMS values can
parison to the time-lapse seismic noise level. The be significantly biased depending on the specific para-
largest sources of time-lapse seismic noise are typi- meters chosen by the user: time window, location,
cally caused by nonrepeatability in the acquisition of length, width, and data contents (reflection signal
the time-lapse seismic data surveys. Seismic acquisi- and noise strengths). From my experience, I recom-
tion nonrepeatability can be caused by source– mend choosing a time window approx. 250–500 ms 31
David Lumley
5734000 658000 SSTVD
CRC-2
0 0
–400
–400
VP, m/s
–800
–800
4000.00
–1200
–1200
3250.00
–1600
–1600
2500.00
5735000
657000
g
5734000
in
1750.00 658000
UT
rth
M
no
ea
st 5733000
M
in 659000
UT
g
ILINE_NO
120 120 120
XLINE_NO
1 41 81 121 161 201 241 11 51 91 131 171 211 251 21 61 101 141 181 221 261
300 300
400 400
500 500
600 600
Time (ms)
Time (ms)
700 700
800 800
900 900
1000 1000
1100 1100
1200 1200
1300 1300
Figure 2.12 3D modeling of time-lapse synthetic seismic data with realistic noise for the CO2CRC Otway Stage 2C project. (Top) 3D
geological model and flow simulation of injected CO2 plume. (Bottom) Synthetic seismic images from 3D earth model. (Left) Preinjection
seismic image. (Center) Postinjection seismic image. (Right) Difference seismic image showing injected CO2 plume and background
nonrepeatable noise level (Glubokoskikh et al., 2016).
long for typical 10–100 Hz seismic data, in a wind- compaction/dilation effects in the overburden), and
owed zone just above the reservoir that is unlikely to that contains similar reflection event characteristics
have changed between surveys (beware of unexpected and strengths as in the reservoir zone. In this case,
32
real fluid/pressure leakage effects or geomechanical NRMS usually provides a quantitative value that is a
Figure 2.13 Finite difference viscoelastic modeling of the 3D seismic wavefield generated by a seismic source in the 3D earth model, South
West Hub CO2 storage project, Western Australia (Lumley et al., 2016). (Figure prepared in collaboration with Taka Miyoshi.)
Table 2.2 NRMS values versus 4D data repeatability source–receiver distance. For this reason, modern
time-lapse marine streamer surveys often employ over-
NRMS Comment sampling methods (more cables, wider arrays) for each
<0.1 Outstanding repeatability (permanent successive survey, and take advantage of steerable
arrays) streamers to help reduce variable cable feather. Two-
<0.2 Excellent repeatability (very low NR boat surveys which separate the source boat indepen-
noise) dently from the streamer array boat, also make it easier
0.2–0.4 Very good repeatability (common for to repeat the surveys in the presence of variable and
marine surveys) often conflicting current, tides, and wave directions
0.4–0.6 Reasonable repeatability (common for (Ridsdill-Smith et al., 2008). Ocean bottom cables
land surveys) (OBC), nodes (OBN), and broadband seismographs
0.6–0.8 Poor repeatability (but may be ok for (OBS), deployed by subsea remotely operated vehicles
qualitative work) (ROVs), can greatly increase the repeatability on the
0.8–1.2 Highly nonrepeatable (time-lapse data receiver side of the equation to within a few meters.
may not be useable) (Semi) permanent arrays, which can be deployed long
term by burying cables/sensors in the seafloor mud,
1.40 Equivalent to two random data sets!
provide the maximum receiver repeatability available
2.00 Data sets are identical, but polarity
today. At the Valhall project in the Norwegian North
reversed
Sea, BP and partners used a permanently entrenched
OBC seafloor array to achieve repeatability levels of
0.05–0.10 for raw (unprocessed) seismic repeat shot
gather data (van Gestel et al., 2008).
very good indication of the (non)repeatability of the For land time-lapse seismic surveys, near-surface
time-lapse seismic data sets and their amenability for variations are a dominant source of nonrepeatable
reservoir analysis. Table 2.2 summarizes my rules of noise. As demonstrated in Figure 2.14 by Pevzner et al.
thumb for time-lapse data quality versus NRMS value (2011), variations between surveys in the near-surface
(Lumley, 1996). moisture conditions (wet, dry, frozen, mud . . .) can
For marine time-lapse seismic surveys, source– greatly affect the coupling of the sources and receivers
receiver positioning differences are a dominant source to the ground, and therefore the source strength and
of nonrepeatable noise. With differential GPS waveform, the frequency content of the data, and the
(DGPS), we can often locate and repeat seismic source strength of Rayleigh surface wave “groundroll” noise,
locations to within a few meters; however, the among other factors. At the land-based Weyburn and
nonrepeatability of receiver (hydrophone group) Aquistore CO2 projects in Canada, White et al. (2015,
locations can often be on the order of 10s to 100s and this book) were able to reduce the raw shot gather
data nonrepeatability from average NRMS values of
33
of meters, generally increasing proportional to
David Lumley
b) WD 2008 c) WD 2007
CDP 0 500 1000 0 500 1000 Offset, m
0
dry wet A B
200
150
TWTT, ms
100
1000
50 mb2008/wd2007
mb2008/wd2008
wd2008/wd2007
0
0 20 40 60 80 100
Frequency, Hz
2000
Figure 2.14 Acquisition nonrepeatability example: repeated weight-drop source (WD) shot gathers at the Otway CO2 injection test site. (Left)
Repeated shot gather during dry surface conditions in 2008. (Center) Original shot gather during wet surface conditions in 2007. (Right)
Nonrepeatability NRMS (%) plotted versus seismic data frequency (green curve corresponds to the two shot gathers displayed here). (Extracted
from Pevzner et al., 2011.)
1.20 to 0.50, by switching from surface-based geophone The development of highly repeatable seismic
arrays to geophones buried at 20 m depth, in order to acquisition systems for both land and marine envir-
suppress the variable near-surface layer effects. At the onments (sources, receivers, permanent arrays, high-
land-based CO2CRC Otway project in Australia, precision navigation, quantitative measurements of
Pevzner et al. were able to reduce the prestack shot environmental meta-data, etc.), is providing a con-
gather nonrepeatability NRMS values within the main tinuously improving ability to detect and image very
20- to 80-Hz seismic passband from approx. 0.75 in weak time-lapse signals generated in the earth. In the
Stage 1 (Figure 2.14), to 0.20 in Stage 2C, by burying a near future, we may not only be able to image and
900+ geophone array (Figure 2.15) at just 4 m below the monitor the injected CO2 plume as it evolves over
surface within the permanent groundwater-saturated time, but we may also be able to image more subtle
zone (Shulakova et al., 2015; Pevzner et al., 2017). phenomena such as the weak injection pressure/stress
For borehole seismic surveys, the major sources front as it propagates away from the CO2 injection
of nonrepeatability are source and receiver position- well and ultimately defines the regulatory Region of
ing, source and receiver instrumentation, and the Influence (ROI) for the overall CO2 storage and con-
coupling between the sources and/or receivers with tainment project. The increasing use of (semi)perma-
the borehole fluid, casing, or formation. In many nent arrays is also allowing us to develop and use
instances, the best results for time-lapse vertical innovative passive seismic and ambient noise techni-
seismic profiling (VSP) and cross-well surveys have ques to image the storage reservoir, as will be shown
been achieved by cementing in the receiver arrays, later in this chapter, thus potentially minimizing the
although interestingly some observations have been need for active manmade source efforts, and thereby
reported of significant changes over time due to reducing the “environmental footprint” of the seismic
suspected long-term curing effects of cement in the surveying operations that are required to monitor a
34 borehole. CO2 injection project over its long life.
Figure 2.15 Buried geophone array at the CO2CRC Otway Stage 2C project site (coordinates courtesy of R. Pevzner and CO2CRC). Prestack
shot gather repeatability was improved from approx. NRMS = 0.75 to NRMS = 0.2 within the main 20- to 80-Hz seismic passband by burying
the 900+ geophone array at just 4 m below the surface within the permanent groundwater-saturated zone (Shulakova et al., 2015; Pevzner
et al., 2017).
Time-Lapse Seismic Imaging

Time-lapse image noise, errors, and artifacts can be depth, and whether the time-to-depth conversion
created during the process of seismic imaging due to process is accurate. For these reasons, I recommend
(sometimes unavoidable) variations in the image pro- that time-lapse data sets be imaged directly in depth
cessing algorithms and parameters, the approximate with amplitude-preserved algorithms.
physics of certain imaging operators that do not Figure 2.16 shows a comparison of the resulting
account for full wave propagation effects, and the time-lapse difference images for various image-
sensitivity of certain nonlinear statistical and data- processing and data combinations. The use of an
dependent signal processing and imaging operators explicit time-lapse image processing flow (e.g., the
to the presence of nonrepeatable seismic noise. Parallel method), applied to two time-lapse data sets,
The theory for time-lapse seismic imaging is dis- produces a good quality difference image of the real
cussed in the Theory section of this chapter. In gen- fluid anomaly, with modest levels of nonrepeatable
eral, I recommend that time-lapse seismic data sets be noise in the background. Using a single 3D seismic
imaged using prestack depth migration (PSDM) that data set (i.e., no time-lapse data!), with two similar,
is amplitude-preserved such that the image gathers but nonidentical, 3D image processing flows (e.g.,
have the correct angle-dependent reflectivity (AVO/ somewhat different algorithm and parameter choices
A) in depth. Often time-lapse seismic data sets are by different data processors), produces a nonzero
imaged in vertical travel time (PSTM), as these algo- difference image with artifacts that could be falsely
rithms are faster and cheaper; however, when we get interpreted as real fluid anomalies. Using two time-
to the detailed analysis of fluid flow features within lapse seismic data sets, but with a nonidentical image
the reservoir, we inevitably find ourselves asking processing flow and parameters, produces a poor-
whether the images we are interpreting have the cor- quality difference image in which background noise
rect structure, closure, stratigraphy, flowpaths, etc. in and image artifacts obscure the true fluid anomaly. 35
David Lumley
4D processing flow, Similar 3D processing flows, Similar 3D processing flows,

4D datasets (Mon-Base) same 3D dataset (no 4D!) 4D datasets (Mon-Base)
060 065 000 095 910 925 940 955 860 865 880 895 910 925 940 955 860 865 880 895 910 925 940 955
750 750 750 750 750 750

800 800 800 800 800 800
850 850 850 850 850 850
900 900 900 900 900 900
950 950 950 950 950 950
1000 1000 1000 1000 1000 1000
1050 1050 1050 1050 1050 1050
1100 1100 1100 1100 1100 1100
1150 1150 1150 1150 1150 1150
1200 1200 1200 1200 1200 1200
1250 1250 1250 1250 1250 1250
1300 1300 1300 1300 1300 1300
1350 1350 1350 1350 1350 1350
1400 1400 1400 1400 1400 1400
1450 1450 1450 1450 1450 1450
1500 1500 1500 1500 1500 1500
1550 1550 1550 1550 1550 1550
1600 1600 1600 1600 1600 1600
1650 1650 1650 1650 1650 1650
1700 1700 1700 1700 1700 1700
1750 1750 1750 1750 1750 1750
1800 1800 1800 1800 1800 1800
1850 1850 1850 1850 1850 1850
1900 1900 1900 1900 1900 1900
1950 1950 1950 1950 1950 1950
2000 2000 2000 2000 2000 2000
2050 2050 2050 2050 2050 2050
2100 2100 2100 2100 2100 2100
2150 2150 2150 2150 2150 2150
Real 4D fluid flow anomaly
Figure 2.16 Comparison of time-lapse (4D) difference images for various time-lapse image-processing flows. (Left) Using a time-lapse
consistent image processing flow with two time-lapse data sets produces a good quality difference image of the real fluid anomaly, with
modest nonrepeatable noise in the background. (Center) Using a single 3D seismic data set (i.e., no time-lapse data!), with two similar, but
not identical, 3D image processing flows (e.g., different parameter choices by different data processors), produces a nonzero difference image
with artifacts that could be falsely interpreted as real fluid anomalies. (Right) Using the two time-lapse seismic data sets in the left panel, but
the nonidentical image processing flow parameters of the center panel, produces a poor-quality difference image in which background noise
and artifacts obscure the true reservoir fluid anomaly. (Figure adapted from an image prepared by Jim Keggin of BP, while on assignment at
Statoil.)
This set of results shows that, if time-lapse image 3. Lastly, calculate and apply statistical data-
processing is not done specifically and carefully to dependent operators only if absolutely necessary,
optimize difference images, real time-lapse anomalies and only for residual “clean-up” (e.g., FXY or
can be obscured, and false anomalies can be created, prediction error filter [PEF] noise suppression).
by the image artifacts created by nonrepeatable image These data-dependent operators are highly risky
processing methods. for time-lapse imaging, especially “Match filters,”
In general, a time-lapse specific image processing as they can remove real anomalies, and also create
flow should avoid statistical data-dependent opera- false anomalies (Rickett and Lumley, 2001). In my
tors, and focus on deterministic and physical model- experience, a high-quality time-lapse image
based operators, in this general order of priority: processing flow will never (rarely) need to rely on
image processing operators that fall within this
1. First, apply all imaging corrections using
category (#3).
deterministic methods and auxiliary (meta) data
measurements (e.g., navigation data, tidal The quality of the time-lapse processing flow and
corrections, etc.). results should be monitored step-by-step with various
2. Second, calculate and apply further imaging quality control measures (NRMS, 4D SNR, Coherency
corrections using physical model-based operators etc.) to ensure that (1) the individual seismic images
(e.g., surface-consistent source and receiver are well focused and (2) the time-lapse difference
corrections, migration velocity analysis, Radon images contain minimal NR noise (e.g., Lumley et
36 al., 2003a, among others). In the Acquisition section
multiple suppression, etc.).
RMS
0.4
0.3
Northing
CRC 2 CRC 2 CRC 2
0.2
0.1
0.0
Easting
0m 500 m
Figure 2.17 Time-lapse seismic images of injected CO2 at the CO2CRC Otway Stage 2C project. The top panels are 4D seismic map views at
reservoir level showing the injected CO2 plume growing over time, and being contained by a sealing fault. The bottom panels are 4D seismic
cross-section images for a vertical plane passing through the CRC-2 injection well, showing preinjection baseline data and three difference
volumes corresponding to 5000, 10 000, and 15 000 tons of injected CO2/CH4 (Pevzner et al., 2017a,b). These 4D seismic field data images can
be compared to the modeled data images in Figure 2.12 predicted as part of the original feasibility study.
of this chapter, I provide advice on how to calculate example of time-lapse seismic images for CO2 injec-
NRMS to obtain reasonable unbiased estimates of NR tion at the CO2CRC Otway Stage 2C project. The top
noise, and to be wary of real time-lapse changes above panels are 4D seismic map views at reservoir level
the reservoir due to fluid/pressure leakage or geome- showing the injected CO2 plume growing over time,
chanical compaction/dilation effects. and being contained by a sealing fault. The bottom
There are many excellent examples of time-lapse panels are 4D seismic cross-section images for a ver-
seismic images of CO2 injection from around the tical plane passing through the CRC-2 injection well,
world, especially in Europe and North America, pre- showing preinjection baseline data and three differ-
sented as case studies in other chapters of this book. In ence volumes corresponding to 5000, 10 000, and 15
keeping with the theme of highlighting Australian 000 tons of injected CO2/CH4 (Pevzner et al., 2017a,
content, Figure 2.17 shows a spectacular recent b). These 4D seismic field data images can be 37
David Lumley
Baseline Parallel Simultaneous

1999 2001 – 1999 Before 2001 – 1999 After
650 650
700 700
750 750
800 800
850 850
900 900
950 950
1000 1000
1050 1050
Time (ms)
Time (ms)
1100 1100
1150 1150
1200 1200
1250 1250
1300 1300
1350 1350
1400 1400
1450 1450
1500 1500
1550 1550
Figure 2.18 Cross-section view of time-lapse seismic difference images using the Parallel image-processing method (center) compared to
the Simultaneous image-processing method (right). Left panel is the 3D baseline (preinjection) seismic image. Note the improvement in the
imaging of the weak CO2 injection anomalies in the right panel compared to the more conventional Parallel image processing result in the
center panel (Lumley et al., 2003a; Lumley, 2010).
compared to the modeled data images in Figure 2.12 nonrepeatable noise level. However, the wavefields
predicted as part of the original feasibility study. generated may be complicated by strong internal scat-
As discussed previously, the strength of the time- tering and wave-mode conversions that are not prop-
lapse signal is related to the compressibility of the erly handled by current image-processing operators,
reservoir rocks, and the compressibility contrast and thus can lead to artifacts in the resulting time-
between pore fluid types. Weak time-lapse signals lapse seismic images (Figure 2.11).
resulting from CO2 injection can be due to hard Most time-lapse seismic imaging is performed
rocks (e.g., Weyburn) and/or weak fluid compressi- using only the P-wave components of the data.
bility contrast (e.g., Otway Stage 1), and in these cases However, most seismic sources generate both P- and
extra effort must be invested to optimize repeatability S-waves, and even pure acoustic or compressive
in order to suppress time-lapse noise and thus sources generate converted P–S waves along the pro-
enhance the detection of a weak CO2 signal. pagation path. Because S-waves are first-order sensi-
Figures 2.18 and 2.19 show examples for a weak tive to pressure/stress changes rather than fluid
time-lapse signal (Weyburn), and compare the saturation changes, using both P- and S-wave data
Parallel and Simultaneous time-lapse imaging meth- information can allow for more accurate determina-
ods. Note the significant improvement in the imaging tion of both fluid saturation and pressure/stress
of the CO2 injection anomalies using the change due to CO2 injection. Figure 2.20 shows a
Simultaneous method, compared to the Parallel multicomponent time-lapse seismic example, with
method result (Lumley et al., 2003a; Lumley, 2010). map-view time-lapse seismic images of the Weyburn
In the opposite case of very strong time-lapse CO2 project at reservoir depth. Horizontal CO2 injec-
38 responses to CO2 (e.g., Sleipner), the presence of tion wells are labelled as red forked lines, horizontal
injected CO2 is easily detected above the producing wells as white lines, and vertical wells as
Figure 2.19 Map view of time-lapse seismic

difference images (Figure 2.18) at reservoir
depth using the Parallel image-processing
method (left) compared to the Simultaneous
image-processing method (right). Note the clear
improvement in the imaging of the weak CO2
injection anomalies in the right panel compared
to the conventional image processing result in
the left panel (Lumley et al., 2003a).
amplitude difference maps are primarily sensitive to

pressure and stress changes only, caused by CO2
injection (Davis et al., 2003).
In this chapter we are primarily considering
surface-seismic surveying methods, for both marine
and land environments. However, as we have seen, it
can be very advantageous to place the seismic sources
and/or receivers in the subsurface closer to the reser-
voir for improved signal-noise conditions and resolu-
tion. The ultimate example of this is using borehole
seismic surveys, especially VSP and cross-well (well to
well) geometries. Figure 2.21 shows a crosswell tomo-
graphic image of the CO2 plume during injection at
the Frio CO2 project. The CO2 plume is imaged by the
time-lapse change in the P-wave velocity, which
matches well with the RST (saturation) logs displayed
for each borehole. The RST logs were repeated over
time during CO2 injection and the changes are shown
in yellow (Daley et al., 2008).
There is still an ongoing need to develop improved
seismic imaging and inversion algorithms specifically
for time-lapse objectives. A concept that has shown
very encouraging results is to treat the image processing
of a series of N time-lapse seismic data surveys as a global
optimization problem by addressing all N of the seismic
Figure 2.20 Multicomponent time-lapse seismic data example. data sets simultaneously – as with the Simultaneous
Map view time-lapse seismic images for the Weyburn CO2 project. approach – constrained by the fact that some areas of
Horizontal CO2 injection wells (red forked lines), horizontal produ- the subsurface are changing whereas most other areas
cing wells (white lines), and vertical wells (white circles). P-wave (top)
and S-wave (bottom) time-lapse seismic amplitude difference maps are not (e.g., Lumley et al., 2003a; Kamei and Lumley,
at reservoir depth (Davis et al., 2003). 2017). This contrasts with the current common
approach of processing N time-lapse data sets as a series
white circles. The P-wave (top) time-lapse amplitude of paired data sets, two-by-two, using the same or similar
difference image is sensitive to both pore pressure and processing flows and parameters for each (analogous to
39
CO2 saturation changes, whereas the S-wave (bottom) parallel processing by local optimization).
David Lumley
Injection
• The effects of saturation changes and injection
Distance (m) Monitor
Well 0 30 Well
pressure in the reservoir, and the possible presence
1505 of multiple phases of CO2, are combined in the
total seismic response and thus not easy to
separate uniquely.
• Pressure–saturation effects may be complicated by
CO2 miscibility and reactive effects on the rock
Depth (m)
matrix (e.g., dissolution of minerals and grain

1530
cement) or in pore fluids (e.g., miscible oil
fractionation).
• Rock and fluid physics analysis shows that the
seismic response is not very sensitive to CO2
saturation levels beyond SCO2 approx. 30% when
1550 CO 2 Plume
present as a supercritical “fluid,” and beyond SCO2
approx. 10% when present as a supercritical “gas.”
• The density effect of injected CO2 is difficult to
isolate and estimate from seismic data alone unless
Figure 2.21 Cross-well tomographic image of the injected CO2 there are very large amounts of CO2 present, and
plume at the Frio CO2 project. The CO2 plume is imaged by time- the seismic data is exceptionally clean (low NR
lapse change of the P-wave velocity, displayed along with the RST
(saturation) logs for each well. The RST logs had multiple runs and
noise environment) and contains ultrafar offsets
the changed zones are highlighted in yellow (Daley et al., 2008). and wide-aperture reflection angles.
• The time-lapse seismic response may be
significantly complicated by additional
Another major opportunity and challenge is to geochemical and geomechanical effects.
extract more of the information contained in the • The full time-lapse seismic response to CO2 can be
time-lapse wavefields, especially using both the P- highly nonlinear and nonunique, thus making it
and S-wave arrivals (e.g., Davis et al., 2003), as well difficult to extract specific information accurately
as the full complex scattered wavefield and the asso- from prestack seismic data or images, which may
ciated time-lapse “4D coda” (e.g., Lumley and be required as input for many CO2 inversion
Shragge, 2013; Shragge and Lumley, 2013; Kamei algorithms.
and Lumley, 2017). Despite these challenges, there are several different
approaches to performing time-lapse seismic inver-
Time-Lapse Seismic Inversion sion in order to make quantitative estimates of the
injected CO2 distribution.
Time-lapse Seismic Inversion to Quantify
CO2 Saturation AVO/AVA Inversion
Whereas seismic imaging is the process of using seis- A first broad category of time-lapse inversions is
mic data to image geological and fluid-flow features based on the time-lapse changes in elastic reflection
within the earth, for example the storage containment amplitude versus source-receiver offset or reflection
reservoir and the injected CO2 plume; seismic inver- angle (AVO/A). In general, the reflection coefficients
sion is the process of using seismic data to make for a seismic wave are nonlinear and depend on gra-
quantitative estimates of the physical properties of dients of the earth’s rock and fluid properties. In
rocks and fluids within the earth, for example the practice, a more simple linearized (Bortfeld) form of
amount of residual CO2 trapped within the storage the full nonlinear Zoeppritz equations is often used
reservoir, and its spatial and temporal distribution. (e.g., Aki and Richards, 1980):
The ability to quantify the amount of injected CO2
present in the reservoir using time-lapse seismic data Rðx; φÞ ≈ c1 ðφÞ m1 ðxÞ þ c2 ðφÞ m2 ðxÞ þ c3 ðφÞ m3 ðxÞ
40 would be extremely useful. However, this quantifica- where R(x, φ) is the angle-dependent P-wave reflec-
tion goal faces several challenges, including tivity at a given point x in the earth, (c1 c2 c2) are
Figure 2.22 Time-lapse seismic

inversion estimate of the 3D CO2
saturation distribution in the Utsira
sand storage reservoir for the Sleipner
project (4th Wave Imaging).
trigonometric functions that include the geometry of geometry, the foregoing equation can be rewritten as
the P-wave reflection angle φ at the reflection point x, a time-lapse change in reflectivity ΔR:
and m = (m1, m2, m3) is a set of three elastic properties
ΔRðx; φÞ ≈ c1 ðφÞΔm1 ðxÞþc2 ðφÞΔm2 ðxÞþc3 ðφÞΔm3 ðxÞ
at the reflection point, for example relative contrasts
in P- and S-wave impedances Ip, Is, and density ρ. For where the time-lapse changes in rock physics proper-
AVO/A methods, it is assumed that the R(x, φ) values ties Δm are explicit functions of the time-lapse
can be accurately estimated from angle-dependent changes in CO2 injection fluid-flow properties
amplitude-preserved seismic image gathers that are ΔSCO2, ΔPp, etc. Given the equations above, and
correctly imaged in depth. time-lapse seismic data measurements of ΔR, an
In the rock physics section of this chapter we inverse problem can be defined in several ways to
discussed how CO2 injection can cause time-lapse estimate the Δm values, and hence the time-lapse
changes in fluid saturation and pore pressure, ΔSCO2, ΔPp, etc. fluid-flow property changes, that fit
among other physical parameters, which can in turn the observed time-lapse seismic data responses.
create time-lapse changes in rock and fluid properties Time-lapse AVO/A inversion methods were initi-
such as bulk modulus K, shear modulus G, density ρ, ally developed in the mid–late 1990s by Lumley,
seismic velocities Vp and Vs, and seismic impedances Landrø, and others (e.g., Lumley, 1995b; Tura and
Ip and Is. If the seismic surveys are reasonably repea- Lumley, 1998, 1999; Landrø, 1999, 2001; Cole et al.,
2002; Lumley et al., 2003b,c). For a single seismic
41
table in terms of source, receiver and wavepath
David Lumley
survey, the AVO response has long been used to make from the seismic image data D(t), via a local 1D solu-
estimates of the coupled rock type and pore fluid tion to a convolutional model (not a wave equation),
content at a given reservoir reflection location (e.g., and is essentially a causal integration of the source-
Ostrander, 1984). For time-lapse seismic data sets, the deconvolved seismic reflection image data D̂ . A more
time-lapse (TL) changes in the AVO response (TL accurate form involves the logarithm of the impe-
AVO, or 4D AVO) can be used to estimate the time- dance derivative and an exponentiated integral.
lapse changes in fluid saturation and pore pressure, as These equations can be generalized to include angle-
summarized in Figure 2.2. For example, by injecting dependent reflection data D(t, φ), both in time or in
CO2 into a saline aquifer storage reservoir, the depth, in which case three independent seismic prop-
increased CO2 saturation with respect to the original erties can be estimated, for example (Ip, Is, ρ), which is
water will cause the rock to soften, and the small injec- known as “prestack” or “elastic” impedance inversion.
tion pore pressure increase will likely have a negligible Because the seismic source is band limited, and the
seismic effect; the combined TL AVO response would earth is naturally attenuative, the seismic data and
therefore be a negative amplitude change (P- hence the foregoing impedance estimates are missing
impedance decrease) at zero offset/angle that brightens both low- and high-frequency information. The
with offset/angle. Such TL AVO responses can thus be source deconvolution in the impedance inversion
used as a basis to develop inversion methods that equation attempts to recover some of the high-
separately estimate the changes in CO2 saturation and frequency impedance information, within seismic
pore pressure at every reflection point along, or within, noise limits. Several methods can be used to recover
the reservoir. Figure 2.22 shows an example of a time- the low-frequency background impedance trend, for
lapse seismic AVO/A inversion estimate of the 3D CO2 example, by using constraints from well-log informa-
saturation distribution in the Utsira sand storage reser- tion and seismic imaging velocities. For impedance
voir for the Sleipner project. inversion methods, it is implicitly assumed that the
reflectivity image data D can be accurately estimated
Impedance Inversion from angle-dependent amplitude-preserved seismic
imaging methods correctly in depth.
A second broad category of time-lapse inversions is
The foregoing impedance inversion method can be
based on the 1D convolutional P-wave reflection seis-
applied to time-lapse seismic data sets using any of the
mic model approximated as
approaches discussed in the Theory section (Parallel,
DðtÞ ≈ SðtÞ ¤ RðtÞ ≈ SðtÞ ¤ ∂t Ip ðtÞ Double Difference, Bootstrap, Simultaneous, etc.).
where D is a seismic reflection image data trace, S is Figure 2.23 shows an example of time-lapse impedance
the source wavelet, R is the reflectivity (reflection inversion to estimate the water saturation change (ΔSw)
coefficient series), Ip is the P-impedance log converted in a hydrocarbon reservoir that is being produced with
from depth to time, ∂t is the time derivative operator, horizontal wells. Each panel shows a depth slice map of
and “¤” is the convolutional operator. This equation the time-lapse change in P-impedance, with blue indi-
can be rewritten in the frequency (w) domain as cating increased water saturation, starting at the gas–
oil contact (GOC) at 1295.5 m depth, and stepping
DðwÞ ≈ SðwÞ RðwÞ ≈ ðiwÞ SðwÞ Ip ðwÞ downwards in approx. 3 m depth increments to the
And solved or “inverted” for Ip as original oil–water contact (OOWC) at 1307.7 m depth.
There has been significant vertical water movement of
Ip ðwÞ ≈ DðwÞ = ½ðiwÞ SðwÞ approx. 6 m from the OOWC depth of 1307.7 m up to
and then inverse Fourier-transformed back to the 1301.5 m depth, and additional water movement in
time domain: some areas of approx. 12 m up to the level of the
ðt GOC at 1295.5 m depth. Although this is a hydrocar-
Ip ðtÞ ≈ ĎðτÞ dτ: bon production example, it is not difficult to imagine
that similar maps could be made for a CO2 injection
0
project, to monitor supercritical CO2 that has buoy-
The approximate equations above form the basis antly migrated upwards from a deep injection point
for all “impedance inversion” methods, in which the through a saline aquifer or depleted hydrocarbon sto-
42 P-wave (or “acoustic”) impedance Ip(t) is estimated rage reservoir, toward the top seal.
At GOC1295.5m TVDSS (+/–1m) At 1–3m below GOC
VANCOGH-1 ST1 VANCOGH-1 ST1
Pimp_Diff_Percent Pimp_Diff_Percent
THEO-1 4.50
THEO-1 4.50
4.00 4.00
3.50 3.50
3.00 3.00
2.50 2.50
2.00 2.00
1.50 1.50
1.00 1.00
0.50 0.50
0.00 0.00
0 200 400 600 800 1000m 0 200 400 600 800 1000m
At 1301.5m TVDSS (+/–1m) At OOWC 1307.7m TVDSS (2–4m below)

(6m below GOC & 6m above OOWC)
VANCOGH-1 ST1 VANCOGH-1 ST1
Pimp_Diff_Percent Pimp_Diff_Percent
THEO-1 THEO-1 4.50
4.50
4.00 4.00
3.50 3.50
3.00 3.00
2.50 2.50
2.00 2.00
1.50 1.50
1.00 1.00
0.50 0.50
0.00 0.00
0 200 400 600 800 1000m 0 200 400 600 800 1000m
Figure 2.23 Time-lapse seismic impedance inversion example. Each panel shows a depth slice map of the time-lapse change in P-
impedance, with blue indicating increased water saturation, starting at the gas–oil contact (GOC) (top left) at 1295.5 m depth, and stepping
downwards in approx. 3 m depth increments to the original oil–water contact (OOWC) at 1307.7 m depth (lower right). There has been
significant vertical water movement of approx. 6 m from the OOWC depth of 1307.7 m up to 1301.5 m depth, including additional water
movement in some areas right up to the level of the GOC at 1295.5 m depth. (In collaboration with and courtesy of Paul Bouloudas and
colleagues, Quadrant Energy.)
Full Waveform Inversion FWI methods are very promising because they can
include far more realistic nonelastic wave propagation
A third broad category of time-lapse inversion is
and scattering effects, do not require a local 1D earth
based on solving a wave equation such that an esti-
assumption, and have the potential to achieve much
mated model of the time-varying earth properties
higher resolution than the AVO/A and impedance-
m(x, τ) creates synthetic seismic shot gathers that
inversion methods, all of which are extremely impor-
match the real data shot gathers to within some spe-
tant for detailed reservoir monitoring objectives.
cified error tolerance. The theory that forms the basis
Figure 2.24 shows an example of time-lapse full
for these time-lapse full waveform inversions (TL
waveform inversion (TL FWI) for a near-surface
FWI, or 4D FWI) is discussed in the Theory section
microbubble injection monitoring experiment
of this chapter.
(Kamei et al., 2017). The microbubble injection
TL FWI methods are currently the subject of much
method is a new technique being developed and tested
active research in my group and others; they are not
in Japan, in which water infused with tiny bubbles is
yet mature. FWI methods are also approx. 10–100
injected into shallow sediments to help “cushion”
times more computationally expensive than the
them in order to suppress earthquake surface-wave
AVO/A and impedance inversion methods (including
their prerequisite PSDM/PSTM costs). However, TL
damage to buildings and infrastructure, and help 43
David Lumley
15
20
Depth [m]
25
30
35
0 10 20 30 40 50 60
Distance [m]
15 15 15
20 26h 20 28h 20
42h
Depth [m]
Depth [m]
Depth [m]
25 25 25
30 30 30
35 35 35
0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60
15 15 15
20
48h 20 50h 20 66h
Depth [m]
Depth [m]
Depth [m]
25 25 25
30 30 30
35 35 35
0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60
15 15 15
20 71h 20 73h 20 74h
Depth [m]
Depth [m]
Depth [m]
25 25 25
30 30 30
35 35 35
0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60
Distance [m] Distance [m] Distance [m]
1.550 1.600 1.650 1.700 1.750 1.800

km/s
15 15 15
20
26h 20
28h 20
42h
Depth [m]
Depth [m]
Depth [m]
25 25 25
30 30 30
35 35 35
0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60
15 15 15
20 48h 20 50h 20
66h
Depth [m]
Depth [m]
Depth [m]
25 25 25
30 30 30
35 35 35
0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60
15 15 15
20 71h 20 73h 20 74h
Depth [m]
Depth [m]
Depth [m]
25 25 25
30 30 30
35 35 35
0 10 20 30 40 50 60 0 10 20 30 40 50 60 0 10 20 30 40 50 60
Distance [m] Distance [m] Distance [m]
–0.010 –0.008 –0.006 –0.004 –0.002 0.000 0.002 0.004 0.006 0.008 0.010 –0.010 –0.008 –0.006 –0.004 –0.002 0.000 0.002 0.004 0.006 0.008 0.010 –0.010 –0.008 –0.006 –0.004 –0.002 0.000 0.002 0.004 0.006 0.008 0.010
–0.010 –0.008 –0.006 –0.004 –0.002 0.000 0.002 0.004 0.006 0.008 0.010
km/s
Figure 2.24 Time-lapse full waveform inversion (TL FWI) example using repeated cross-well seismic surveys, showing <1% velocity changes
as injected plume evolves during 74 hours of microbubble injection (injector location marked in white). Top panels are individual FWI velocity
inversion results; bottom panels are time-lapse FWI velocity-change inversion results (Kamei et al., 2017, and unpublished work).
44
avoid loss of life. Under certain conditions, the injec- resolution quantitative estimates of the volume and
tion of compressible microbubbles into the original mass of CO2 present in the storage reservoir. These
formation water may be similar to injecting compres- are topics of much active research in the geophysical
sible CO2 into a saline aquifer storage reservoir. For community today, as partly discussed in other chap-
this research project, 15 time-lapse cross-well seismic ters of this book, and thus deserve a more detailed
surveys were acquired over a period of 74 hours discussion that is beyond the scope of an overview
(3 days) before, during, and after microbubble injec- chapter on seismic monitoring.
tion (the injection well is shown in white). Figure 2.24
shows the results of our TL FWI for the baseline Passive Seismic Monitoring
preinjection survey, and 9 of the 14 repeated time-
In this final section, I will present recent “hot off the
lapse surveys. The top panels are the individual FWI
press” research advances in time-lapse passive seis-
velocity estimates for each cross-well seismic survey,
mology that may allow us, in the future, to monitor
the bottom panels are the time-lapse velocity-change
CO2 injection and storage reservoirs without the need
estimates obtained using the TL FWI Bootstrap
for active manmade source efforts, by instead using
method. These spectacular results show that we are
seismic data from (semi)permanent sensor arrays that
clearly able to see the effects of the microbubble injec-
continuously record (micro) earthquake energy and
tion plume evolving over time to within less than 1%
(ocean) ambient noise.
P-wave velocity variation (Kamei et al., 2017, unpub-
Up until this point in the chapter, we have been
lished work).
considering time-lapse seismic surveying techniques
that use active manmade sources, for example dyna-
Integrated Inversion with Nonseismic Data mite shots or Vibroseis trucks on land, or compressed
To help overcome the limitations of using time-lapse airgun arrays at sea. For each time-lapse seismic sur-
seismic alone to quantify the CO2 saturation distribu- vey, thousands of source locations are typically
tion in the storage reservoir, it is likely that seismic required to form high-quality 2D and 3D images of
inversions would benefit from being combined with, the Earth. This manmade source effort requires a
and constrained by, complementary geophysical data major mobilization of crew personnel, equipment,
sets such as electromagnetic (EM), gravity, and InSAR and resources, and can have a significant financial
satellite radar. EM data are highly sensitive to CO2 cost and environmental impact on the survey area
saturation when the CO2 is electrically resistive com- (e.g., by cutting hundreds of source-line access swaths
pared to the brine formation water. Gravity data are through fields or farm land). For many CO2 storage
approximately linearly dependent on CO2 saturation projects, time-lapse seismic surveys will need to be
because the density of the injected CO2 is less than acquired every 1–10 years depending on the stage in
that of water. InSAR satellite data measures the the project’s life. In many cases, the stakeholders who
ground deformation at the Earth’s surface caused by control the surface access rights to the CO2 storage
geomechanical deformations within and above the site may not welcome this magnitude or frequency of
reservoir, where small surface ground displacements seismic survey impact, and the project operators may
on the order of a few millimeters or centimeters can be not be able to afford the significant financial costs of
related to the injection pressure and stress changes in repeated active-source seismic surveying.
the reservoir via coupled flow-simulation and geome- One possible strategy to minimize the environ-
chanical modeling. mental impact, and the expense, of repeating time-
One strategy forward for enhanced seismic CO2 lapse seismic surveys at a CO2 storage site, may be to
inversion would be to combine the higher resolution deploy a (semi)permanent receiver array, and also to
but more qualitative estimates of CO2 saturation from avoid the use of manmade sources. In this scenario, a
seismic, with the lower resolution but more quantita- (semi)permanent sensor array could be deployed at
tive CO2 saturation estimates from EM and gravity, the site as a one-time event, buried in the shallow
and the high-resolution ground surface deformation subsurface/seafloor zone (e.g., the top 1–10 m); this
data measured by InSAR data. These in turn can be not only keeps the array secure from interference
coupled with, and constrained by, fluid-flow and geo- with, and damage by, manmade and natural activities,
mechanical modeling, in order to obtain high- but also typically enhances the time-lapse seismic data 45
David Lumley
Otway Grouped Data: 2015–12–19 00:25:45 Event: 19–12–2015 00:25:45 : LSQ–TT [ms]
800
120 600
400 48 48
100
40
False Northing [m]

200 32
Group Number
24
80
0
60 –200 8
32 16
32
–400
40
–600
20
–800
0 –1000
12.5 13.0 13.5 14.0 14.5 Event: 19–12–2015 00:25:45 : LSQ–TT [ms]
Elevation [m]
Time [s] –1300

35
15
30
20
10
–1400
25
30
–1500 +
5
–1600 10
25
15
20
–1700
–800 –600 –400 –200 0 200 400 600 800 1000
False Easting [m]
Figure 2.25 Example of microseismic event detected with the passive seismic array during the 15 000-ton CO2 injection for the Otway
Stage 2C project. The top panel shows two detected microseismic events (A2 and A3) compared to the background noise level (A3 event
S/N>>10) during 1 week of recording (December 14–21, 2016). The lower left panel shows the high-quality seismic trace data for the A3
microseismic event. The lower right panel shows the A3 estimated microseismic source location, which is very close to both the CO2 injection
well located at the center of the array (star in Figure 2.26), and to the CO2 injection point at approx. 1500 m depth (Drew, Saygin, and
Lumley, unpublished work, 2017).
quality by 20 dB or more (van Gestel et al., 2008; Advantages of using a passive seismic array for
Pevzner et al., 2017). And, instead of imaging with monitoring of CO2 injection projects may include
seismic waves generated by active manmade sources,
• Low environmental impact on the survey area
seismic images of the subsurface can be made by using
• Low cost of repeating many time-lapse seismic
(micro) earthquake seismic energy, and/or interfero-
surveys
metric images created from (ocean) ambient noise,
continuously recorded in passive “listen” mode by • Ability to provide continuous, real-time
monitoring and imaging of the reservoir
the permanent sensor array (Claerbout, 1968;
Schuster, 2009; Wapenaar et al., 2010; Issa and • Ability to detect weak (unfelt) seismicity and
Lumley, 2015; Issa et al., 2017). adjust the CO2 injection program to avoid the
46
possibility of inducing felt earthquake events, and for a previous surface geophone array, to NRMS less
thus maintain public confidence and support than 0.20 within the main 20–80 Hz seismic passband
(social license) for the project by burying the 900+ geophone array at just 4 m below
• Ability to use microseismic events to monitor the the surface within the permanent groundwater-
shape, location, and progress of the injected CO2 saturated zone (Shulakova et al., 2015; Pevzner et al.,
plume and pressure front over time 2017). As part of the Otway Stage 2C research project
• Ability to use ambient seismic noise to image the funded by the CO2CRC, my research group has been
shape, location, and progress of the injected CO2 using passive seismic data recorded with this array for
plume and pressure front over time long continuous periods (weeks, months) to try and
Disadvantages of using a passive seismic array for monitor the Otway Stage 2C injection of 15 000 tons
monitoring of CO2 injection projects may include of CO2/CH4, using both micro-seismic energy, and
the following: ocean ambient noise energy.
Figures 2.25 and 2.26 show examples of our ongoing
• Initial cost and environmental impact of research results for microseismic events detected during
deploying the sensor array may be high. CO2 injection using the Otway Stage 2C passive seismic
• Ongoing power, access, maintenance, and repair array data (Drew, Saygin, and Lumley, unpublished
requirements for a (buried) permanent array may work, 2017). Figure 2.25 shows two detected microseis-
become increasingly challenging over long periods mic events (A2 and A3), which are stronger than the
of time. background noise levels (A3 event S/N ≫ 10), during a
• No ability to control the source frequency week of continuous recording (December 14–21, 2016).
spectrum of the passive seismic energy, and thus The lower left panel shows the high quality of the passive
the image resolution at reservoir depth, as it seismic data for the A3 microseismic event. The lower
depends on the microseismic events and ambient right panel shows the estimated microseismic source
noise naturally present in the area. location, which is very close to the CO2 injection well
• May require very long recording times (days, located at the center of the array (star in Figure 2.26),
weeks, months, or even years) to accumulate and is estimated to be at approx. 1500 m depth, which is
enough passive seismic body-wave energy to form the true CO2 injection depth. Figure 2.26 shows the
high-quality images at the reservoir depth. source mechanism estimated for the A3 microseismic
Harnessing passive seismic energy to monitor subsur- event. The estimated event magnitude is Mw = 0.0, and
face reservoirs is an active area of research in my the detection noise threshold of the array is estimated to
group, among others. Over the past five years, my be approx. Mw >–1.5. Given the low CO2 injection
group has been deploying various (semi) permanent pressures, this relatively large Mw = 0 value suggests
sensor arrays at basin scales (100+ km spacing), this is a fault reactivation event for a fault that intercepts
regional scales (10–100 km spacing), and reservoir the reservoir, and that is at or near critical stress. The
monitoring project scales (0.1–10 km spacing), using event waveform energy and polarity is plotted in an
high-sensitivity exploration geophones with an overlay on the passive seismic array; note the event
approximate 1–500 Hz flat spectral response, and also polarity reversal and nodal plane across the array. This
broadband seismographs with a 0.01–200 Hz flat spec- microseismic source mechanism corresponds to a SW–
tral response. With these arrays, we have been con- NE striking normal fault dip–slip motion, which is also
tinuously recording large distant and moderate consistent with weak fault reactivation induced by CO2
regional earthquakes of magnitude Mw >2, microseis- injection.
mic earthquakes of magnitude Mw >–2, and ambient Figures 2.27–2.29 show our ongoing research
seismic noise of all types, especially strong ocean ambi- results related to ambient seismic noise imaging and
ent noise with spectral peaks at about 0.05–0.2 Hz. time-lapse reservoir monitoring of CO2 injection
Figure 2.15 shows a buried geophone array at the using the Otway Stage 2C passive seismic array.
Otway Stage 2C project site, deployed by Roman Figure 2.27 shows an interferometric ambient seismic
Pevzner et al. at Curtin University and the CO2CRC. noise image created from 1 hour of continuous
Pevzner et al. were able to reduce the average prestack recording with an initial linear 2D test array, plotted
shot-gather NRMS values from approximately 0.75 against two vintages of conventional active-source 3D 47
David Lumley
Figure 2.26 Source mechanism estimated for the A3 microseismic event, detected with the passive seismic array during the Otway Stage 2C
CO2 injection. (Left) estimated magnitude of Mw = 0.0 for the A3 microseismic event, versus a detection noise threshold of Mw >–1.5 in the
main frequency band of 10–100 Hz. Given the small injection pressures, this relatively large Mw value suggests a fault reactivation event for a
fault at or near critical stress. (Right) Waveform energy and polarity of the microseismic event arrivals overlain on the passive seismic array (note
the polarity reversal and SW–NE nodal plane across the array); this source mechanism corresponds to a SW–NE normal fault dip-slip motion,
also consistent with fault reactivation (Drew, Saygin, and Lumley, unpublished work, 2017).
seismic data, at the Otway CO2 project site (Issa and 2017). The top panel shows that interferometric
Lumley, 2015; Issa et al., 2017). Note that the 2D Green’s function seismic traces are highly repeatable
ambient noise image has good signal to about 2– before CO2 injection, but start to change significantly
3 km depth (deeper than the CO2 injection reservoir just before (well pad activity) and during CO2 injection
at approx. 1500 m depth), but is limited to a max- (especially in the later parts of the waveform coda). The
imum frequency content of approx. 25 Hz due to the bottom panel shows that the Green’s function coher-
nature of the ocean storm-wave noise sources. ency (repeatability) is close to 100% before CO2 injec-
Figure 2.28 shows beam-steering velocity-azimuth tion, and then significantly decreases during the activity
“radar” plots of ambient seismic noise recorded with immediately prior to, and during, CO2 injection.
the full 3D passive seismic array (Issa and Lumley, These passive seismic results are “hot off the press”
unpublished work, 2016). The top panel shows 2-Hz at the time of writing this book chapter, and are the
surface waves arriving from the SW direction of the subject of our ongoing active research. The results are
Southern Ocean and propagating horizontally along certainly highly encouraging and interesting; we look
the surface at approx. 700 m/s. The bottom panel forward to reporting further passive seismic monitor-
shows 14-Hz body waves arriving from the same SW ing results in journal publications and future editions
ocean direction but scattering off of the nearby coast- of this book.
line, propagating at approx. 3000 m/s and approx. 30
degrees from vertical incidence, as determined from
the array apparent velocities, and local 1D sonic log Acknowledgments
velocities. The author of this Chapter (David Lumley) kindly
Figure 2.29 shows preliminary results of time- acknowledges the support of his current employer
lapse reservoir monitoring with ambient seismic as Professor at the University of Texas at Dallas
noise interferometry at the Otway Stage 2C CO2 injec- (UTD), and the generous permanent endowment
48
tion project (Saygin and Lumley, unpublished work, made by Cecil and Ida Green (founders of Texas
funding through its industry members and research

partners, the Australian Government under the CCS
Flagships Programme, the Victorian State
Government and the Global CCS Institute.
Portions of this research material were created
using the high-performance computing (HPC)
resources of the Pawsey Supercomputing Centre in
Perth, Australia.
Finally, the author wishes to kindly acknowledge
the many research collaborations and discussions
with colleagues, staff, and graduate students over the
years that have contributed to some of the material in
this chapter, including but not limited to (apologies in
advance to anyone I may have inadvertently forgotten
to mention): The University of Western Australia
(especially Paul Connolly, Julian Drew, Nader Issa,
UGeun Jang, Rie Kamei, Eric May, Taka Miyoshi,
Matt Saul, Erdinc Saygin, Jeff Shragge et al.), CSIRO
(especially Andrew King, Joel Sarout, Linda Stalker et
al.), Curtin University (especially Boris Gurevich,
Figure 2.27 Interferometric ambient seismic noise image (center) Maxim Lebedev, Roman Pevzner et al.), WAERA,
created from 1 hour of continuous recording, compared to con- WA-DMP (especially Dominique van Gent, Sandeep
ventional active-source 3D seismic data images, at the Otway CO2
project site. Note that the ambient noise image has good signal to Sharma et al.), ANLEC (especially Rick Causebrook,
about 2–3 km depth (deeper than both reservoirs), but is limited to a Kevin Dodds et al.), CO2CRC (especially Charles
maximum frequency content of approx. 25 Hz due to the nature of Jenkins, Matthias Raab, Max Watson et al.),
the ocean storm wave noise sources (Issa and Lumley, 2015; Issa
et al., 2017). Geoscience Australia (especially Clinton Foster et
al.), Woodside (especially Neil Kavanagh, André
Gerhardt, Andrew Lockwood, Tom Ridsdill-Smith et
Instruments), which supports the author’s current al.), Chevron (especially Ron Behrens, Steve Doherty,
position as the Green Chair in Geophysics at UTD. Ray Ergas, Gary Hampson, Fred Herkenhoff, Harry
The author also kindly acknowledges the generous Martin, Jason McKenna, Don Sherlock, Joe Stefani,
support of Woodside Energy, and Chevron, for partial Mark Trupp, Wendy Young, Zee Wang et al.), BHP
support of the author’s previous position as Professor Billiton (especially Guy Duncan, Robin Hill et al.),
and Chair in Geophysics at the University of Western Quadrant Energy (especially Paul Bouloudas, Rob
Australia (UWA), where a significant portion of the Kneale, Fred Wehr et al.), Shell (especially Rodney
research and teaching material discussed in this chap- Calvert, Peter Chia, Paul Hatchell, Jan Stammeijer et
ter was developed. al.), 4th Wave Imaging (especially Don Adams, Steve
The author wishes to acknowledge research finan- Cole, Dave Markus, Mark Meadows, Rich Wright et
cial assistance provided through Australian National al.), Geological Survey of Canada (especially Don
Low Emissions Coal Research and Development White et al.), Stanford University (especially Jon
(ANLEC R&D). ANLEC R&D is supported by Claerbout, Jack Dvorkin, Gary Mavko, Amos Nur,
Australian Coal Association Low Emissions Lynn Orr, Tiziana Vanorio et al.), Statoil (especially
Technology Limited and the Australian Government Ola Eiken, Lars Kristian Strønen, Per Digranes et al.),
through the Clean Energy Initiative. NTNU (especially Martin Landrø et al.), LBNL (espe-
The author wishes to acknowledge the CO2CRC cially Tom Daley et al.), JOGMEC (especially Masashi
for their research financial assistance, and leadership Nakatsukasa, Takuji Mouri, Ayato Kato, Mamoru
in the conceptualization, design, and field operations Takanashi, Yuki Nakamura et al.), and many other
associated with the Otway 2C Project (especially colleagues I have worked with on 200+ time-lapse 4D
Roman Pevzner et al.), which received CO2CRC seismic projects over the past 25+ years. 49
Figure 2.28 Beam-steering azimuth-velocity “radar” plots of ambient seismic noise arriving at the Otway CO2 project site. (Top) 2-Hz surface
waves arriving from the (SW) Southern Ocean at approx. 700 m/s. (Bottom) 14-Hz body waves arriving from the same SW direction at approx.
3000 m/s and approx. 30 degrees from vertical incidence, with complex azimuthal scattering off the irregular coastline (Issa and Lumley,
University of Western Australia).
Pre-injection CO2 injection
Pre-injection CO2 injection
Figure 2.29 Time-lapse reservoir monitoring with ambient seismic noise interferometry at the Otway Stage 2C CO2 injection project (45 days
of passive seismic data displayed; 8 days preinjection and 37 days during injection). (Top) Interferometric Green’s function seismic traces are
highly repeatable before CO2 injection, then start changing significantly (especially the later parts of the coda) just prior to and during CO2
injection starting at red dash line. (Bottom) Green’s function coherency (repeatability) is close to 100% before injection, and then significantly
changes during the well pad activity just prior to, and during the CO2 injection (Saygin and Lumley, unpublished work, 2017).
Resources, Energy and Tourism (DRET),

References Australia.
Aki, K., and Richards, P. G. (1980). Quantitative seismology: Caspari, E., Müller, T. M., and Gurevich, B. (2011). Time-
Theory and methods. San Francisco: W. H. Freeman. lapse sonic logsreveal patchy CO2 saturationin-situ.
DOI:10.1017/S0016756800034439 Geophysics Research Letters, 38: L13301. DOI:10.1029/
Calvert, R. (2005). Insights and methods for 4D reservoir 2011GL046959.
monitoring and characterization. Society of Exploration Claerbout, J. (1968). Synthesis of a layered medium from its
Geophysicists. www.amazon.com/Insights-Methods-4D acoustic transmission response. Geophysics, 33(2): 264–
-Reservoir-Charterization/dp/1560801360 269. https://doi.org/10.1190/1.1439927
Carbon Storage Taskforce. (2009). National carbon Cole, S., Lumley, D., Meadows, M., and Tura, A. (2002).
mapping and infrastructure plan. Department of 51
Pressure and saturation inversion of 4D seismic data by
David Lumley
rock physics forward modeling. Technical Program Kamei, R., and Lumley, D. (2017). Full waveform inversion
Expanded Abstracts. Society of Exploration of repeating seismic events to estimate time-lapse
Geophysicists, 2475–2478. DOI:10.1190/1.1817221. velocity changes. Geophysical Journal International, 209
Daley, T. M., Myer, L. R., Peterson, J. E., Majer, E. L., and (2): 1239–1264.
Hoversten, G. M. (2008). Time-lapse crosswell seismic Kamei, R., Jang, U., Lumley, D., et al. (2017). Time-lapse
and VSP monitoring of injected CO2 in a brine aquifer. full waveform inversion for monitoring near-
Environmental Geology, 54(8): 1657–1665. surface microbubble injection. Expanded Abstracts,
DOI:10.1007/s00254-007–0943-z. European Association of Engineering Geosciece
Davis, T., Terrell, M. J., Benson, R. D., Cardona, R., Kendall, R. (EAGE), Paris, France. DOI:10.3997/2214–
R., and Winarsky, R. (2003). Multicomponent seismic 4609.201700956.
characterization and monitoring of the CO2 flood at Kragh, E., and Christie, P. (2002). Seismic repeatability,
Weyburn Field, Saskatchewan. Leading Edge, 22(7): 696– normalized RMS, and predictability. Leading Edge, 21
697. DOI:10.1190/1.1599699. (7): 640–647. DOI:10.1190/1.1497316.
Dvorkin, J., and Nur, A. (1996). Elasticity of high-porosity Landrø, M. (1999). Discrimination between pressure and
sandstones: Theory for two North Sea data sets. fluid saturation changes from time-lapse seismic data. In
Geophysics, 61: 1363–1370. DOI:10.1190/1.1444059. Expanded Abstracts, 69th Annual International
Emberley, S., Hutcheon, I., Shevalier, M., Durocher, K., Meeting. Society of Exploration Geophysicists, 1651–1654.
Gunter, W. D., and Perkins, E. H. (2004). Geochemical Landrø, M. (2001). Discrimination between pressure and
monitoring of fluid-rock interaction and CO2 storage at fluid saturation changes from time-lapse seismic data.
the Weyburn CO2-injection enhanced oil recovery site, Geophysics, 66(3): 836–844. http://dx.doi.org/10.1190/
Saskatchewan, Canada. Energy. https://ideas.repec.org/ 1.1444973
a/eee/energy/v29y2004i9p1393-1401.html Lebedev, M. (2012). Geophysics laboratory: Otway rock
Gernert, J., and Span, R. (2016). EOS–CG: A Helmholtz physics tests. In B. Evans, R. Rezaee, V. Rasouli, et al.,
energy mixture model for humid gases and CCS Milestone Report for ANLEC Project #3–1110-0122.
mixtures. Journal of Chemical Thermodynamics, 93: http://anlecrd.com.au/projects/predicting-co-sub-2-sub
274–293. https://doi.org/10.1016/j.jct.2015.05.015 -injectivity-properties-for-application-at-ccs-sites/
Glubokovskikh, S., Pevzner, R., Dance, T., et al. (2016). Lumley, D. (1995a). Seismic time-lapse monitoring of
Seismic monitoring of CO2 geosequestration: CO2CRC subsurface fluid flow. PhD thesis, Stanford University.
Otway case study using full 4D FDTD approach. Lumley, D. E. (1995b). Seismic monitoring of hydrocarbon
International Journal of Greenhouse Gas Control, 49: fluid flow. Journal of Mathematical Imaging and Vision,
201–216. https://doi.org/10.1016/j.ijggc.2016.02.02 5(4): 287–296. DOI:10.1007/BF01250285.
Guilbot, J., and Smith, B. (2002). 4D constrained depth Lumley, D. E. (2001). Time-lapse seismic reservoir
conversion for reservoir compaction estimation: monitoring. Geophysics, 66: 50–53.
Application to Ekofisk Field. Leading Edge, 21(3):
302–308. http://dx.doi.org/10.1190/1.1463782 Lumley, D. E. (2006). Nonlinear uncertainty analysis in
reservoir seismic modeling and inverse problems:
Hatchell, P., and Bourne, S. (2005). Rocks under strain: Technical Program Expanded Abstracts, Society of
Strain-induced time-lapse time shifts are observed for Exploration Geophysicists, 2037–2041. DOI:10.1190/
depleting reservoirs. Leading Edge, 24(12): 1222–1225. 1.2369936.
http://library.seg.org/doi/abs/10.1190/1.2149624
Lumley, D. (1996–present). 4D seismic reservoir monitoring:
Issa, N., and Lumley, D. (2015). Passive seismic imaging at Course book.
depth using ambient noise fields recorded in a shallow
buried sensor array. Australian Society of Exploration Lumley, D. (2010). 4D seismic monitoring of CO2
Geophysicists. library.seg.org/doi/pdf/10.1071/ sequestration. Leading Edge, 29(2): 150–155.
ASEG2015ab135 DOI:10.1190/1.3304817.
Issa, N., Lumley, D., and Pevzner, R. (2017). Passive seismic Lumley, D., and Shragge, J. (2013). Advanced concepts in
imaging at reservoir depths using ambient seismic active and passive seismic monitoring using full
noise recorded at the Otway CO2 geological storage wavefield techniques. Extended Abstracts, Australian
research facility. Geophysical Journal International, 209 Society of Exploration Geophysicists, 2013: 1–4. https://
(3): 1622–1628. https://doi.org/10.1093/gji/ggx109 doi.org/10.1071/ASEG2013ab167
Johnston, D. H. (2013). Practical applications of time-lapse Lumley, D., Adams, D. C., Meadows, M., Cole, S., and
seismic data. Society of Exploration Geophysicists. http:// Wright, R. (2003a). 4D seismic data processing issues
52 dx.doi.org/10.1190/1.9781560803126 and examples. Technical Program Expanded Abstracts,
Society of Exploration Geophysicists, 1394–1397. into the subsurface: CO2CRC Otway Project.
DOI:10.1190/1.1817550. International Journal of Greenhouse Gas Control, 63:
Lumley, D., Adams, D., Meadows, M., Cole, S., and Ergas, R. 150–157. https://doi.org/10.1016/j.ijggc.2017.05.008
(2003b). 4D seismic pressure-saturation inversion at Rickett, J. E., and Lumley, D. E. (2001). Cross-equalization
Gullfaks Field, Norway. First Break, 21, September, data processing for time-lapse seismic reservoir
European Association of Geoscientists and Engineers monitoring: A case study from the Gulf of Mexico.
(EAGE). Geophysics, 66, Special Section, 1015–1025.
Lumley, D., Meadows, M., Cole, C., and Adams, D. (2003c). DOI:10.1190/1.1487049.
Estimation of reservoir pressure and saturations by Ridsdill-Smith, T., Flynn, D., and Darling, S. (2008).
crossplot inversion of 4D seismic attributes. Technical Benefits of two-boat 4D acquisition, an Australian case
Program Expanded Abstracts, Society of Exploration study. Leading Edge, 27(7): 940–944. DOI:10.1190/
Geophysicists, 1513–1516. DOI:10.1190/1.1817582. 1.2954036.
Lumley, D., Adams, D., Wright, R., Markus, D., and Cole, S. Saul, M., and Lumley, D. (2013). A new velocity-pressure-
(2008). Seismic monitoring of CO2 geo-sequestration: compaction model for uncemented sediments.
Realistic capabilities and limitations. Technical Geophysical Journal International, 193(2): 905–913.
Program Expanded Abstracts, Society of Exploration DOI:10.1093/gji/ggt005.
Geophysicists, 2841–2845. DOI:10.1190/1.3063935. Schuster, G. T. (2009). Seismic interferometry. Cambridge:
Lumley, D., King, A., Pevzner, R., et al. (2016). Feasibility Cambridge University Press.
and design for passive seismic monitoring at the SW Shragge, J., and Lumley, D. (2013). Time-lapse wave-equation
Hub CO2 Geosequestration Site. Australian National migration velocity analysis. Geophysics, 78(2): S69–79.
Low Emissions Council (ANLEC) R&D Project Number
7-0212-0203. http://anlecrd.com.au/reports_storage/ Shulakova, V., Pevzner, R., Dupuis, J. C., et al. (2015).
Burying receivers for improved time-lapse seismic
Mathieson, A., Wright, I., Roberts, D., and Ringrose, P. repeatability: CO2CRC Otway field experiment.
(2009). Satellite imaging to monitor CO2 movement at Geophysical Prospecting, 63: 55–69.
Krechba, Algeria. Energy Procedia, 1(1): 2201–2209.
DOI:10.1016/j.egypro.2009.01.286. Tura, A., and Lumley, D. E. (1998). Subsurface fluid-flow
properties from time-lapse elastic-wave reflection data. In
Mavko, G., Mukerji, T., and Dvorkin, J. (2009). The rock 43rd Annual Meeting, SPIE, Proceedings, 125–138.
physics handbook: Tools for seismic analysis of porous
media. Cambridge: Cambridge University Press. Tura, A., and Lumley, D. E. (1999). Estimating pressure and
saturation changes from time-lapse AVO data.
Meadows, M., Adams, D., Wright, R., Tura, A., Cole, S., and Technical Program Expanded Abstracts, Society of
Lumley, D. (2005). Rock physics analysis for time-lapse Exploration Geophysicists, 1655–1658.
seismic at Schiehallion Field, North Sea. Geophysical
Prospecting, 53: 205–213. van Gestel, J-P., Kommedal, J. H., Barkved, O. I., Mundal, I.,
Bakke, R., and Best, K. D. (2008). Continuous seismic
Ostrander, W. (1984). Plane-wave reflection coefficients for surveillance of Valhall Field. Leading Edge, 27(12):
gas sands at non-normal angles of incidence. Geophysics, 1616–1621. DOI:10.1190/1.3036964.
49(10): 1637–1648. http://dx.doi.org/10.1190/1.1441571
Vanorio, T. (2015). Recent advances in time-lapse,
Pacala, S., and Socolow, R. (2004). Stabilization wedges: laboratory rock physics for the characterization
Solving the climate problem for the next 50 years with and monitoring of fluid-rock interactions. Geophysics,
current technologies. Science, 305: 968–972. 80(2): WA49–WA59. DOI:10.1190/geo2014-0202.1.
DOI:10.1126/science.1100103.
Vialle, S., and Vanorio, T. (2011). Laboratory measurements
Pevzner, R., Shulakova, V., Kepic, A., and Urosevic, M. of elastic properties of carbonate rocks during injection
(2011). Repeatability analysis of land time-lapse seismic of reactive CO2-saturated water. Geophysics Research
data: CO2CRC Otway pilot project case study. Letters, 38: L01302. DOI:10.1029/2010GL045606.
Geophysical Prospecting, 59: 66–77. DOI:10.1111/
j.1365–2478.2010.00907. Wapenaar, K., Draganov, D., Snieder, R., Campman, X., and
Verdel, A. (2010). Tutorial on seismic interferometry:
Pevzner, R., Urosevic, M., Tertyshnikov, K., et al. (2017a). Part 1 – Basic principles and applications. Geophysics,
Stage 2C of the CO2CRC Otway Project: Seismic 75(5): 75A211–227. https://doi.org/10.1190/1.3457445
monitoring operations and preliminary results. Energy
Procedia. www.sciencedirect.com/science/article/pii/ White, D. J., Roach, L. A. N., and Roberts, B. (2015). Time-
S1876610217317344 lapse seismic performance of a sparse permanent array:
Experience from the Aquistore CO2 storage site.
Pevzner, R., Urosevic, M., Popik, D., et al. (2017b). 4D Geophysics, 80(2): WA35–WA48. http://library.seg.org/
surface seismic tracks small supercritical CO2 injection 53
doi/abs/10.1190/geo2014-0239.1
Chapter
Goals of CO2 Monitoring
3 Why and How to Assess the Subsurface Changes

Associated with Carbon Capture and Storage
Thomas M. Daley and William Harbert
Why Monitor? characterized, along with the technology used, are

selected based on the goals for characterizing and
Nearly all subsurface engineering activities begin with
monitoring; i.e., they are chosen after answering the
an effort to characterize subsurface properties impor-
question: why are we monitoring? Therefore, under-
tant to the activity and have a goal to assess (and
standing the goals of monitoring is essential. The site-
reduce) risk. Over the last century, many methods
specific goals for each project will vary, but will follow
needed to characterize the subsurface have been
from the main drivers for monitoring: performance
developed, largely in support of extraction of
assurance, regulatory requirements, public assurance,
resources from the Earth (e.g., minerals or hydrocar-
and an overarching desire for project risk reduction.
bons). As the extraction methods became more com-
As we describe each of these goal drivers, it is impor-
plex and efficient, the characterization efforts merged
tant to remember that minimizing monitoring costs
into monitoring efforts in which measurements were
will also be an overarching goal – though not always
repeated over time. For example, active seismic ima-
the most important. While this chapter discusses the
ging started in the 1930s as two-dimensional (2D)
goals and some of the techniques of monitoring, other
cross sections of the Earth, expanded to three-dimen-
chapters look at case studies of projects storing CO2 in
sional (3D) images in the 1980s and then became 4D
the subsurface and some issues encountered in mon-
in the 1990s when 3D surveys were repeated over time
itoring as well as successes.
(with time as the fourth dimension). Often these early
4D data sets had only two or three time samples (with
each full repeat of a 3D survey being a time sample). Performance Assurance
Early success in detecting spatially localized changes A goal for operators of a geological storage project will
in subsurface properties (e.g., mapping where oil was be to confirm that the injection and growth of a CO2
being depleted or left in place) led to routine monitor- plume is progressing as expected and predicted by
ing of the subsurface with 4D seismic. Thus, 4D seis- their modeling of flow and transport within the reser-
mic is now considered a key monitoring technology in voir – this is a key aspect of performance assurance.
oil and gas production (e.g., Johnston, 2013). An example of performance assurance is monitoring
The initial characterization of a new subsurface pro- injection pressures and volumes, which are typically
ject will impact later monitoring, especially monitor- monitored on the surface at hourly to weekly inter-
ing that needs a baseline. In fact, modern subsurface vals. The correct performance of surface equipment
activities should view site characterization as an itera- (e.g., pumps and compressors) can be directly tied to
tive process linked to and updated by repeated mon- these measurements, and subsurface changes near the
itoring, and this is true of carbon capture and storage injection well may be reflected quickly. Initial perfor-
(CCS) projects. mance monitoring at the start of injection will likely
Geological carbon storage (GCS or geosequestra- focus on surface measurements. The subsurface is of
tion) of carbon dioxide (CO2) will require subsurface course heterogeneous and, as the CO2 plume grows,
characterization and monitoring. It is important to changes in reservoir performance are possible and
understand the goals of monitoring to guide decision- likely. The operator will want to understand how the
54 making. The specific subsurface properties to be entire reservoir, including the seal, is performing over
time. For this purpose, the monitoring will be focused pressure front tracking. AOR and PISC are defined
on the predicted extent of the CO2 plume at each as follows:
monitoring time interval, as well as the distribution
• Area of review: The region surrounding the
of CO2 within the plume. For performance assurance,
geological carbon storage/sequestration project
the monitoring will be expected to provide quantitative
where USDWs may be endangered by CO2.
values of properties, such as CO2 saturation in a given
The AOR is likely delineated by computational
volume, which can be compared with modeled values.
modeling, based on geological and geophysical site
A component of performance assurance is early
characterization that accounts for injected and
and definitive leakage detection and attribution (i.e.,
displaced fluids.
is the detected CO2 attributable to the storage pro-
ject?). Monitoring technologies need to be optimal for • Postinjection site care: Appropriate monitoring
and other actions needed following cessation of
leakage detection and, if necessary, they will need to
CO2 injection to ensure that USDWs are not
guide mitigation. It is likely that a separate set of
endangered.
monitoring technologies will be deployed in the
event of leakage and mitigation efforts. For example, Further EPA regulation, building upon the UIC pro-
Lawton et al. (this volume) look at a specific pilot gram for geological sequestration wells, is subpart RR
project to examine the identification of leakage in associated with the Clean Air Act which requires
a shallow subsurface zone. a US EPA-approved MRV plan and reporting of the
amount of CO2 sequestered using a mass balance
Regulatory Compliance approach (EPA, 2015, 2016).
Within the European Commission framework,
Regulatory compliance is a first-order goal for mon-
Directive 2009/31/EC (EU, 2009), the postinjection
itoring. The importance of regulation as a driver of
or closure/decommissioning activities for large sto-
monitoring and verification (MV) was recognized by
rage sites include the development of a monitoring
the intergovernmental panel on climate change
plan with targets and methods, postclosure monitor-
(IPCC) in their 2005 summary of CCS (IPCC, 2005)
ing, updated site characterization and risk assessment,
by the statement that the MV cost estimate “covers
and inspections by the Competent Authority (CA)
pre-injection, injection and post-injection monitor-
following the site closure. Before transfer of the site
ing, and depends on the regulatory requirements”
to the CA, long-term containment of CO2 must be
(IPCC, 2005). The United States Environmental
verified, the site sealed, and monitoring and assess-
Protection Agency (US EPA) and other regulators
ment demonstrated for 20 years.
require that carbon capture and storage projects
Storage projects will need to identify the monitor-
implement monitoring, reporting, and verification
ing technologies to meet the goals of regulatory
(MRV) systems that detect leakage of CO2 from either
requirements. Examples include defining and moni-
the containment reservoir or away from the storage
toring the entire AOR (while possibly reducing the
site (EPA, 2016). Beginning in 2011, the US EPA
AOR based on characterization or monitoring
Underground Injection Control (UIC) program
results), and selecting long-term monitoring for
established the Class VI CO2 injection well category
PISC while potentially reducing the PISC time-line.
for the purposes of protecting underground sources of
drinking water (USDW). The US EPA requires Class
VI wells to include a monitoring and testing plan Public Assurance
during the operational phase of the project and to The need to assure the general public that a CCS
set minimum technical criteria for injection permits project is operating safely may require different mon-
that include geological site characterization, area of itoring strategies and activities than those used for
review (AOR), well mechanical integrity testing technical or regulatory requirements; thus public
(MIT), monitoring, postinjection site care (PISC), assurance can be a separate goal. Public outreach
and site closure. Both AOR and PISC are examples programs for GCS projects will interact with local
of regulatory needs for site characterization and mon- constituencies and may lead to monitoring that spe-
itoring, possibly including monitoring ground water cifically addresses local concerns. For example, storage
quality above the confining zone and plume and under public (or private) land with water wells may 55
necessitate geochemical well sampling to assure the in the area of study. This model should include struc-
specific water well owners, separate from the required ture, material properties, pore fluid characteristics,
regulatory monitoring requirements. As another exam- pressure and stress, fracture and fault characteristics,
ple, atmospheric sampling of CO2 for some locations and material anisotropy characteristics within the
may not be effective from a technical leakage-detection area of study. These and other properties are the
cost/benefit analysis but may provide public assurance initial site characterization, typically including data
value greater than the cost. from an initial characterization well. After the initial
characterization has been completed, baseline data
Risk Reduction can be acquired and the characterization refined.
In addition to determining the suitability of
A site-specific risk assessment is likely to guide mon-
a storage site, characterization is used in the design
itoring. Reduction of the overall project risk will be
of monitoring to constrain the expected response of
a goal for the monitoring program. A quantitative risk
target formations to CO2 injection activities.
assessment will also allow a cost/benefit or value of
Geophysical survey techniques, such as those listed
information analysis to be used for selecting monitor-
in Table 3.1, have parameters that can be tuned to
ing technologies. The benefit from a given technology
better image specific spatial and depth zones, or alter-
is the detection ability, both spatially and temporally,
natively, if not properly designed, can introduce noise
i.e., how large a plume of CO2 is detectable and how
into geophysical data sets through inappropriate
quickly you can detect it, while the cost of each mon-
acquisition geometries. These acquisition noise com-
itoring technology includes data acquisition (includ-
ponents can be difficult to remove, or misinterpreted
ing permitting) and data processing/analysis.
as subsurface behavior. Site characterization allows
An example is a site with known abandoned boreholes
acquisition noise to be minimized by careful survey
that could be leakage pathways, and thus a technology
design, yielding higher quality time-lapse results.
such as soil gas monitoring at or near those boreholes
Only then, and as a distinct step, should the acquisi-
could be high benefit and also help reduce the risk
tion of additional monitoring data occur. It is impor-
profile. This type of monitoring may also address
tant to view the characterization process as a closed
goals of performance assurance. Risk assessment for
loop, beginning and ending with the goal of an accu-
subsurface activities often has high uncertainty
rate and thorough knowledge of the reservoir perfor-
because of geological heterogeneity and uncertainty
mance behavior with respect to the project objectives.
in characterization. Harbert et al. (2016) discuss the
Initial characterization can serve as baseline monitor-
importance of monitoring strategies for risk reduc-
ing, but ideally baseline surveys would be designed
tion and outline an approach to integrate monitoring
after initial characterization was complete.
strategies with risk assessment and reduction. Even
For geophysical monitoring, characterization pro-
research can be driven by risk reduction, as one of
vides bounds to the behavior of key parameters and
their conclusions is that “the need for field experiment
the expected geophysical response of monitoring meth-
testing that is risk-driven” (p. 260). While monitoring
ods in the presence of realistic noise conditions. Once
for performance assurance is focused on demonstrat-
these characterizations have been completed they can be
ing a predicted outcome, monitoring for risk reduc-
incorporated into the design of baseline surveys and
tion will be more focused on the unexpected or
assessments to yield the maximum value of information.
unwanted outcomes.
Monitoring is quite different from either characteriza-
tion or baseline surveys. In monitoring, detecting subtle
Characterization versus Monitoring subsurface changes from the baseline case is the princi-
The goals of monitoring overlap with the goals of site pal focus of analysis and interpretation (Fgure 3.1 from
characterization. Subsurface characterization is an Calvert [2005]). In the best case, an observed change will
iterative process based on, and including, specific have a high value of information because it ensures that
monitoring and geoengineering objectives related to the objectives are being successfully obtained (perfor-
injection. This process begins with conceptual and mance assurance), or because in a situation with anom-
numerical models being refined to produce the most alous behavior the observation can inform rapid
56 accurate possible quantitative model of the subsurface operational changes as needed (risk reduction).
Figure 3.1 The ability of carefully designed monitoring and baseline geophysical surveys to generate a quality difference image of reservoir
formation changes using reflection seismic methods is shown in this figure taken from Calvert (2005). The Draugen Field is located
approximately 160 km offshore of Norway and was discovered in 1984. These data were interpreted to better understand water flooding of
a target reservoir related to enhanced oil recovery (EOR). The region of time lapse change (interpreted with respect to the water flood changes)
is shown in map view on the right. The spatial topology of injection related changes, formation compartmentalization, and intensity of change
were constrained by this approach.
Monitoring can be considered a continuous sub- measurement monitoring tools include subsurface
surface characterization during the CO2 injection determinations of pressure or temperature, measured
activity, as shown in Figure 3.2. Characterization and at a specific in situ location of the subsurface.
monitoring activities include initial baseline surveys, Downhole fluid analysis would be another direct mea-
thorough reservoir characterization, risk assessment, surement approach, in which formation fluids from
developing specific injection engineering protocols, a specific location and depth are sampled and analyzed.
managing the CO2 injection, accurately modeling the In contrast, indirect measurements typically observe
expected plume configuration, developing metrics for a proxy, such as seismic velocity or electrical resistivity,
monitoring evolving risk, monitoring the injection and which is physically related to the subsurface property of
planning for site closure, and decommissioning (EU, interest, but is not that property itself. These observa-
2009; Wright et al., 2009;Wright, 2011). tions are then analyzed and interpreted to estimate
a subsurface property of interest (such as pressure or
gas saturation). For example, electrical or seismic mon-
What Properties to Monitor? itoring are indirect monitoring methods; the actual geo-
To address each of the goal drivers discussed in the physical observations, such as elastic wave velocity or
foregoing, a different set of subsurface properties may electrical conductivity, are interpreted to derive the para-
be targeted. Deciding what properties to monitor meters of interest (see Chapter 2 by Lumley, this volume,
requires understanding how they are measured, and for a more detailed discussion). Direct monitoring is
then the process of choosing the properties and mon- limited to the specific subsurface locations available
itoring tools can be considered. The properties can be from boreholes, while indirect monitoring, typically
measured in different ways. In general, subsurface mea- using geophysical methods, can survey large volumes
surements can be considered as direct or indirect. Direct of the subsurface with varying spatial resolution. 57
Possible site Probable site Proven site

High
Baseline Monitoring Injection Monitoring PISC
Uncertainty
Data Injection &

Total cost & uncertainty
acquisition monitoring
Construction Model
Detailed Includes time update
for plant
characterization construction
Closure
Design
Living model Equalization
Model driven Monitoring
Prelim. plan Long term
Prediction Environmental
study monitoring
Total cost Validation
Low 1yr Time 2–3 yrs 5–7 yrs 30 yrs 50 yrs 100+
Figure 3.2 Role of monitoring in the lifecycle of a commercial CO2 storage site. Using relative cost and uncertainty scales, it is suggested that
uncertainty decreases with time (after Peters 2007). That is, within a framework of minimum monitoring requirements, measurements such as
geophysical monitoring are made to assure decreasing uncertainty in performance of the GCS site and geological carbon storage.
Geophysical Monitoring It is important to understand the physical para-

meters that affect each geophysical technology. For
Geophysics is a fundamental part of subsurface obser-
example, seismic methods are dependent on effective
vation, change detection, and reservoir analysis.
stress, effective pressure, elastic properties, and den-
Geophysical technologies can be used to observe and
sity of solid and fluid components of rock materials.
monitor various subsurface characteristics relevant to
The elastic moduli are ratios of applied stresses and the
CO2 operations. Two common geophysical technolo-
resulting strain in linear elastic materials and vary
gies are seismic (Lumley, 2001, 2010; Lumley et al.,
with rock framework composition, porosity topology,
2008; Lumley, Chapter 2, this volume) and electrical
pore-filling phase, and effective pressure conditions
(Bergmann et al., 2012; Gasperikova and Commer,
(Sayers, 2006, 2010). Because of their sensitivity to
Chapter 9, this volume) surveys. A number of chap-
these and other properties, seismic imaging methods
ters in this volume discuss the application of these
are considered important tools for subsurface charac-
techniques in field cases. Elastic wave (seismic) based
terization and monitoring.
geophysical methods are sensitive to subsurface elas-
As mentioned earlier, following the characteriza-
tic moduli, which can be used to determine porosity,
tion of the potential site and the careful design of
density, and fracture geometries, among other prop-
a seismic survey to accurately image the target forma-
erties. The determination of these reservoir character-
tions, seismic tools are used for improving and adding
istics is referred to as seismic lithology. Electrical or
detail to the structural and stratigraphic model and
electromagnetic geophysical methods are sensitive to
confirming the expected baseline seismic responses.
conductivity, which can be used to determine porosity
An initial 3D reflection seismic survey can also serve
and accurately estimate the geochemical composition
as a structural framework for the spatial location of
of formation fluids. Conductivity can also be
other monitoring data and as a baseline for determin-
impacted by reactive geochemistry, formation perme-
ing temporal (i.e., time-lapse, 4D) changes in seismic
ability, and/or seal integrity, thus allowing the system
properties (Johnson, 2013) and their temporal uncer-
to be better understood via monitoring. Table 3.1
tainty (Kragh and Christie, 2002; Cantillo, 2011).
summarizes geophysical approaches, the physical
A key application of 3D/4D reflection seismic is the
property measured, and the subsurface properties
58 they may constrain.
mapping of changes in reflector amplitude (δA),
Table 3.1 Summary of geophysical methods used for subsurface CO2 monitoring
Method Principle Physical property Interpreted property

measured
Electromagnetic
Airborne electromagnetic Detects frequency domain- Electrical conductance and Geological structurea
frequency domain (FDEM) based variation in inductance Groundwater geometryb
electromagnetic flow in
subsurface materials
Airborne electromagnetic Detects time domain-based Electrical conductance and Geological structure
time domain variation in electromagnetic inductance Groundwater geometry
flow in subsurface materials
Subsurface pressure
changes CO2 geometry
Ground-based electrical and Detects electromagnetic Electrical conductance and Geological structure
electromagnetic 2D and 3D flow in subsurface materials inductance Groundwater geometry
surveys (resistivity sounding, Electrical resistivity and CO2 geometry
spontaneous potential, capacitance (induced
induced polarization, polarization and
electromagnetic, very low magneto-telluric methods)
frequency, long offset time Potential differences
domain electromagnetics, (spontaneous potential
and magnetotelluric) method)
Crosswell electromagnetics Detects electromagnetic Electrical conductance and Geological structure
flow in subsurface materials inductance Groundwater geometry
CO2 characteristics and
geometry
Potential Fields
Surface microgravity, Detects variations in the Mass density Geological structure
downhole microgravity Earth’s gravitational field Changes in pore-filling fluid
phases
Seismic
Surface survey-based Detects reflected seismic Elastic moduli and density Geological structure
reflection seismic; 2D, 3D, P- or S-waves Pore fluid identification
and 4D, including Geological structure
3-component Pore-pressure estimation
Fracture geometry
Vertical seismic profiling Detects refracted seismic Elastic moduli and density Geological structure
(VSP) P- or S-waves Pore fluid identification
Pore-pressure estimation
Crosswell tomography Detects seismic P and S ray- Elastic moduli and density Geological structure
path changes Pore fluid identification
Pore-pressure estimation
59
Table 3.1 (cont.)
Method Principle Physical property Interpreted property

measured
Microseismic monitoring Detects small failure events Elastic moduli, density, Geological structure
related to injection, stress Pore fluid location
production, or Fracture geometry
hydrofracturing In situ stress, hydraulic
diffusivity, failure surface
orientation
Surface Imaging
InSAR Satellite-based system Ground displacement High-resolution topography
generates and detects Phase shift of reflected wave Variation in topography
changes in reflected radar related to subsidence or
frequency electromagnetic inflation
waves.
Tilt measurement Change in orientation of Ground displacement Variation in topography
ground (surface or related to subsidence or
borehole) inflation
a
Geological structure includes geologic strata geometry, heterogeneity, and fault geometry.
b
Groundwater geometry includes presence or absence of groundwater and variation in groundwater conductivity that can be related to
interaction with CO2 and changes in pH.
reflector arrival time (δt), poststack seismic attributes, properties (Goodway et al., 1997). In addition, prestack
pore fluid properties such as water saturation (ΔSw), amplitude versus offset (AVO) analysis (Castagna and
and variations in effective pressure. Seismic data are Backus, 1993; Castagna and Swan, 1997; Castagna et al.,
analyzed for these properties both before imaging 1998) uses relationships that contain both rock frame-
(prestack) and after imaging (poststack) (Figure 3.3). work and fluid terms, allowing these properties of the
Changes in effective pressure (ΔPeff), defined as the injection reservoir to be investigated in detail.
difference between confining and pore pressure, can In general, prestack tools can efficiently utilize proper-
influence poroelastic characteristics that are resolva- ties of the near and far offset reflection seismic
ble using seismic methods. Perturbations in pore data prior to the reflection stacking and imaging
pressure have a variety of impacts, such as opening operations.
compliant cracks or pores that can be observed using An example associated with CO2 injection at
seismic methods. Such pore-pressure perturbations a Permian basin site had prestack AVO analysis iden-
have been related to induced seismicity (e.g., Mavko tifying a useful proxy for subsurface CO2 (Purcell
et al., 2009; Shapiro, 2015). There is a large body of et al., 2009, 2010; Harbert et al., 2011). In this analysis
knowledge related to estimating formation pore fluids the prestack AVO attribute proxy was defined within
and pressure using calibrated seismic data (Landrø the injection reservoir and constrained at the top by
et al., 2001, 2003; Sayers, 2010). a caprock unit. When extracted as a geobody from the
A variety of advanced seismic techniques can be data set and interpreted, this AVO-anomaly feature
used to calculate seismic attributes from well log cali- was observed to have a flat bottomed AVO attribute
brated 3D reflection seismic volumes (Simm and Bacon, expression through the reservoir and was interpreted
2014). These techniques can estimate acoustic or elastic to represent a pore-phase variation, based on compar-
impedance and can determine material properties, such ison with a rock-calibrated petrophysical model,
as Lamè parameters or combinations of parameters microcomputer tomography (μCT) analysis, and
60 such as λρ/μρ that are proxies for other important well logs (Mur et al., 2011).
Pre-stack Pre-stack
Incident Angle (deg)
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 Incident Angle (deg)
0
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50
AVO Response changes
ΔPeff
20
20
40
ΔSw
40
Time (ms)
Time (ms)
60
60
80
80 Gradient
100 100
Intercept
120 120
Stk B Stk B
NIp NIp
Response changes
Time: δt
Amplitude: δA
Attribute changes
Geometry
0–30 0–30 Instantaneous
Post-stack Post-stack Spectral Decompostition
Wavelet
Baseline Injection Monitoring Injection induced

Survey Survey 1 changes
Figure 3.3 Time-lapse seismic monitoring of an injection. Both prestack and poststack reflection amplitude versus offset (AVO) data can be
analyzed to detect changes in the subsurface. Time-lapse data should be interpreted within a framework of petrophysical models, differential
stress variation, and production effects (Duffaut and Landrø, 2007).
Well Logging logging tools are localized in their sensitivity to the well
region, these estimates are commonly extrapolated to
Well-based logging tools (e.g., wire-line well logging)
populate formation models in correlation with seismic
provide highly detailed measurements directly within
lithology inversion. Boreholes, and their well log data,
the formations of interest and are critical for AVO and
can be horizontal in orientation, allowing lateral for-
seismic inversion analyses and to calibrate and better
mation variation and zoning to be estimated.
understand all other reservoir associated geophysical
Incorporating structural data from borehole log data
data, such as electrical, electromagnetic, crosswell,
has been shown to considerably improve geophysical
VSP, and reflection seismic (Chapter 11 by Bassiouni,
electrical inversion results (Doetsch et al., 2012, 2013).
this volume). Log-based measurements include con-
ductivity, pressure, temperature, acoustic velocity,
nuclear magnetic resonance (NMR), electrical and Geochemical Monitoring
electromagnetic responses, borehole images, and for- Geochemical monitoring provides insight into both
mation fluid composition (Kirksey, 2012). Processing reservoir containment and active geochemical pro-
and analysis of various acoustic, electrical, and nuclear cesses that could impact injection efficiency.
measurements lead to estimates of other rock and fluid Geochemical tracer techniques provide a direct
properties, such as porosity, permeability, salinity, and measure of subsurface connectivity between the
anisotropic geomechanical properties (Schlumberger, points of tracer injection and measurement in the
1989, 1991). Analysis of pulsed neutron logging allows subsurface. Geophysical monitoring techniques are
indirect estimation of CO2 saturation and is commonly more accurately understood when constrained by
used in CO2 storage and EOR projects, although results the geochemistry of the subsurface systems.
are highly dependent on quality of the cement bond Geochemical alteration of the rock matrix can cause
(Nelson and Guillot, 2006). Because the observations of 61
notable error in geophysical analysis that assumes
constant frame properties. Above-storage-zone fluid data requires joint or coupled inversions to estimate
sampling is a direct measurement for storage secur- the “best” fit to the data and must address spatial
ity, and knowledge of fluid composition is important sampling variations between data sets and models.
to calibrate and aid in the interpretation of electrical Traditional single well hydrological data provide
and electromagnetic geophysical monitoring. high-resolution measurements for properties imme-
As an example of field testing the geochemical and diately surrounding boreholes. In the interwell region,
electrical geophysical monitoring of a leakage sce- geophysical cross-borehole measurements are sensi-
nario, Dafflon et al. (2012) used laboratory and field tive to subsurface properties over larger regions but
experiments to explore the sensitivity of time-lapse can have lower spatial or temporal resolution and can
complex resistivity response, and associated geo- be difficult to interpret quantitatively.
chemical transformations, to a dissolved CO2 plume Doetsch et al. (2012, 2013) combined the advan-
at Plant Daniel, Mississippi. Results showed that elec- tages of hydrological and geophysical data in a fully
trical resistivity and phase responses correlated well coupled hydrogeophysical inversion. Time-lapse
with dissolved CO2 injection processes. Specifically, seismic and electrical resistivity tomography (ERT)
resistivity initially decreased in response to an data (Carrigan, et al., 2009) from the Cranfield site
increase of bicarbonate and dissolved species, after were inverted using cross-gradient joint inversion.
which the resistivity rebounded toward the initial This joint inversion approach ensures structural
conditions due to transition of bicarbonate into similarity while not assuming any specific petrophy-
non-dissociated carbonic acid. Laboratory studies sical relations between model parameters. Inversion
confirmed that field observations of electrical resistiv- of seismic and ERT data for simultaneous changes
ity, similar to crosswell electrical resistivity tomogra- in seismic velocity and resistivity from baseline
phy, could detect dissolved CO2. Both laboratory and surveys to nine months after the start of the CO2
field experiments demonstrate the potential of field- injection were successful in detecting changes in
based geophysical methods for remotely monitoring formation pore fluids. In this method, the indivi-
changes in subsurface groundwater quality (geochem- dual time-lapse data sets are initially inverted: seis-
istry) due to CO2 leakage. mic inversion for changes in seismic velocity and
An innovative technology, the U-tube fluid electrical inversion for changes in conductivity –
sampler (Freifield et al., 2005) is a tool for direct both related to CO2 gas saturation. The joint inver-
fluid sampling that can be deployed at multiple sion results show a more clearly focused CO2
depths as a permanent installation to allow plume: in particular, the scatter plots indicate
a multiyear sampling program. An additional a link between change in velocity and ERT-derived
innovative type of analysis, which can significantly gas saturation. A joint inversion can estimate the
aid geophysical monitoring, is downhole fluid ana- gas saturation that best matches both the seismic
lysis (DFA) (Mullins, 2008). Downhole fluid ana- data and the electrical data (Kowalsky et al., 2016).
lysis (or direct sampling via wireline) draws fluids The results of the ERT time-lapse inversion were
directly from the reservoir of interest for immedi- then combined with the hydrological data (e.g., gas
ate geochemical analysis and allows complexities sampling and tracer tests) to test different concep-
of reservoir fluid composition, reservoir compart- tual aquifer models. Among the outcomes was the
mentalization and delineation, and the potential ability to estimate width of fluid flow paths that
for reactive geochemical processes that could alter connect the injection and monitoring wells.
formation permeability, porosity, or seal integrity,
to be better understood. How to Choose Monitoring Tools and
Joint Inversion from Multiple Methods Parameters
When data sets from multiple technologies are col- Cost/Benefit Assessment of Tools
lected for monitoring, we are often faced with very
Once the goals of monitoring are chosen, monitoring
different measures of subsurface properties, requiring
tools can be assessed for their ability to aid in achiev-
balancing of uncertainty. For example, an integrated
62 ing those goals, their benefit, and for their site specific
analysis of hydrological and geophysical monitoring
costs. A large number of monitoring tools are
Figure 3.4 Initial review of monitoring technologies presented in a Boston Square for the Krechba Reservoir at In Salah, Algeria (modified
after Mathieson et al., 2010). The green box in the lower left were technologies identified in the initial planning as having high estimated
benefit and lower costs. On the Boston Square plot these positions were not static during the project and the final techniques used included
dynamic modeling, surface flux, geochemistry, cement CO2 work, well logging, water chemistry, geomechanics, aquifer studies, 4D seismic,
annulus sampling, geochemical tracers, wellhead CO2 monitoring, wellbore sampling and satellite (InSAR) imaging. Technologies tested
included microbiology, tilt meters, and microseismic. The final benefit/cost locations can be reviewed in Mathieson et al. (2010).
available, as shown in Table 3.1, and reducing cost will consideration. In this Boston Square, repeat 3D
always be a goal, so assessing cost/benefit is impor- reflection seismic (4D seismic) was located in the
tant. The initial selection of monitoring tools to con- high-cost/high-benefit quadrant but was included
sider for cost/benefit analysis can be daunting. in the monitoring. Because CO2 reservoirs can be
The International Energy Agency Greenhouse Gas complexly heterogeneous, a high-resolution
(IEAGHG) program has developed an on-line selec- approach that includes reservoir characterization
tion process that can be a first step (IEAGHG, 2017). and modeling is preferred (Mathieson et al.,
One method to compare a large number of 2010). In addition to understanding the monitor-
monitoring techniques, using their estimated ben- ing techniques that reside in the “just do it” quad-
efits and costs for a specific site, is plotting them rant for a site, two important additional concerns
on a Boston Square (or Boston Chart) (Hannas, are the time line for postinjection site care and
2013). determining the impact of volume and rate of CO2
Figure 3.4 shows the initial format of such an injection on the monitoring design.
analysis by Mathieson et al. (2010) where high- As reviewed earlier, monitoring technologies are
benefit, low-cost monitoring options are in divided naturally between direct and indirect meth-
a quadrant labeled “just do it.” These techniques ods. Geophysical methods in general are indirect
were reviewed and an updated approach shown in methods of monitoring. Direct measurement moni-
Mathieson et al. (2010) was then used in monitor- toring tools include subsurface determinations of
ing. In this analysis, high-cost/low-benefit methods pressure or temperature, measured at a specific
were avoided, low-cost/high-benefit methods rou- in situ location of the subsurface. Indirect measure-
tinely included, and methods in the high-cost/ ments, well constrained by physical principals and
high-benefit quadrant were given careful models, are analyzed and interpreted to estimate the 63
subsurface material property of interest. Electrical or that sensitivity decreases with depth and the solu-
seismic monitoring methods can be interpreted to tion is nonunique. Microgravimetry is most useful
derive subsurface fluid saturation (between brine when combined with other methods such as surface
and supercritical CO2) using laboratory calibrated deformation and seismic.
variations in electrical conductivity or bulk modulus Borehole methods require wellbore access for
along with petrophysical models. A value of informa- receiver placement and provide information at gen-
tion (VOI) study by Trainor-Guitton et al. (2013) erally higher spatial resolution in a more limited sub-
provides another example in which calibrated electri- surface region. When the goals of monitoring need
cal resistivity was analyzed to accurately estimate sub- a localized high-resolution method, perhaps a specific
surface water characteristics such as total dissolved depth zone for risk reduction, borehole techniques are
solids (TDS) and pH. warranted.
In addition to direct and indirect methods, mon-
itoring techniques can be characterized as being
active, methods that require a source, or passive, What Depths/Zones to Monitor?
methods that simply observe the system. One aspect of developing monitoring goals is
An example of a high-resolution active monitoring determining what area and depths to monitor.
approach is borehole seismic methods such as vertical For performance assessment the reservoir depth
seismic profiles (VSP), 3D-VSP (Cheng et al., 2010), zone is one target. The reservoir depths are
and continuous active-source seismic monitoring defined by the top of the permeable/porous storage
(CASSM) (Daley et al., 2008). These methods have zone – or alternately as the base of the imperme-
higher spatial resolution, better signal to noise char- able seal above the reservoir – and by the bottom
acteristics, easier direct calibration, and more rapid of the reservoir formation (or the base of expected
time-lapse monitoring near wellbores (Coueslan, CO2 plume). There are likely to be goals requiring
2013) compared with surface seismic surveys. deeper monitoring, below the reservoir, such as
Technology developments such as fiber optic sensing monitoring fluid pressure change in basement
for VSP (e.g., Daley et al., 2015) are improving the rocks which may cause induced seismicity on
cost/benefit ratio of active source methods. unknown faults. Also, there is likely to be above
An example of a passive monitoring approach zone monitoring intervals (AZMI, e.g., Hovorka,
would be the use of microgravity. Microgravimetry 2013), including multiple seals above the reservoir
is a cost-effective and relatively rapid means of defined as part of a storage zone. For performance
observing changes in density distribution in the assessment, the spatial area (the “footprint”) of
subsurface, particularly those changes caused by reservoir monitoring will be determined by reser-
the migration of fluids with differing densities. voir/injection modeling, with the expected plume
This is a passive monitoring technique because the dimensions being a likely minimum size of mon-
variation within the earth system itself creates a time itoring. The depths monitored for performance
lapse anomaly observable without active probing. assessment could also include AZMI. For regula-
Time-lapse gravity has been used since 1961 (Allis tory compliance, the area and depths will be deter-
and Hunt, 1986), but substantial improvements in mined by regulatory definitions such as the
gravimeter technology, and the advent of highly US EPA’s AOR, which is also initially tied to
precise Global Positioning Satellite (GPS) systems, reservoir modeling. The AOR is largely controlled
have led to rapid growth in microgravity applica- by the pressure footprint, as well as the CO2
tions. This technology has been successfully applied plume, with the pressure area being larger.
to carbon sequestration at Sleipner (Alnes et al., Groundwater depths, especially those depth zones
2008), aquifer recharge studies in Utah and else- with usable drinking water, are also likely to need
where (Chapman et al., 2008; Davis et al., 2008), monitoring to meet regulatory requirements. For
and to hydrocarbon EOR surveillance in Alaska public assurance, areas of surface development,
(Ferguson et al., 2007). This method gives direct housing, industry, etc., may extend the area or
estimates of densities and its implementation cost cause more detailed effort on some subset of the
64 is low among geophysical methods. Drawbacks are footprint. The risk reduction goals will have site
specific areas of focus such as existing wellbores or An example of using qualitative versus quantita-
fault locations. tive monitoring is leakage assessment. Definitions of
The size of the monitoring area may also change leakage will probably be project specific and have
with time. Initial characterization may be larger than regulatory definitions. Migration of CO2 outside the
the expected plume footprint to allow for unex- initial injection zone is not necessarily defined as
pected flow and to assess the area of the pressure leakage, especially if the storage zone includes multi-
plume (which will be larger than the CO2 plume). ple seals. Here leakage means migration of CO2
Repeat monitoring, such as surface seismic, may beyond predetermined project boundaries. Leakage
start with a smaller area and expand to larger assessment is a likely goal for risk reduction, perfor-
areas as the CO2 plume and pressure fronts grow. mance assessment, regulatory compliance, and public
The goals will determine whether CO2 presence or assurance. Initial monitoring for leakage assessment
pressure or some other factor defines the areal can be qualitative – is a leak detected – rather than
extent of monitoring. The choice of monitoring quantitatively monitoring leakage amounts. However,
tools used will consider the depth and areal extent if a leak is detected then the monitoring will need to be
of required monitoring as well as the target. For more quantitative in order to guide decisions.
example, groundwater monitoring is more likely to Mitigation of leakage (if required) will have designs
use geochemical sampling and electrical geophysics, dependent on measurement of leakage rates, mass,
as opposed to seismic or deformation monitoring volume, location, etc. Monitoring assessment of leak-
(of course site-specific conditions and cost/benefit age mass flow rates will have a value of information
are deciding factors between tools). that may be considered with the cost of monitoring
tools to decide on both monitoring deployment and
Quantitative versus Qualitative Monitoring intervention techniques to stop/reduce the leakage.
When addressing the goals of monitoring with speci-
fic monitoring tools, one important differentiator How Often to Monitor?
is quantitative versus qualitative monitoring. The temporal sampling of monitoring results is an
An example of the difference is simple detection of important consideration. Monitoring tools can be
CO2 at any concentration (qualitative) versus estimat- separated between those that are essentially continu-
ing CO2 mass in a given volume (quantitative). ous and those that have some discrete time period
In general, qualitative monitoring is likely to have between measurements. For those tools with discrete
lower costs. Qualitative monitoring is also likely to time intervals, the goals of monitoring will inform the
be indirect monitoring, while direct monitoring (e.g., decision of how often to monitor. Because essentially
geochemical sampling) is more likely to provide all modern data collection is digital with some discrete
quantitative information. For some goals, qualitative time sampling, continuous monitoring can be defined
monitoring will be sufficient and will have lower cost/ as data collected frequently enough to capture all
benefit ratio. The same data may be used for both changes related to processes of interest. For example,
qualitative and quantitative assessment with the dif- a downhole (or surface) pressure gauge is digitally
ference being analysis effort. For example, 4D surface sampled, typically at intervals of seconds to minutes,
seismic monitoring can provide qualitative detection with all changes in pressure expected to happen at
with less costly processing. Quantitative assessment of intervals at least an order of magnitude longer than
CO2 saturation per unit volume can also be obtained the sampling interval. With modern digitizing and
from 4D seismic but more intensive processing recording systems, there is essentially no cost savings
and analysis is required, often including other infor- for longer intervals of sampling so we expect pressure
mation such as core studies and calibrated rock phy- to be a continuous recording. Another example would
sics models of the reservoir rock. Ultimately, the most be tilt meters used to measure surface deformation.
certain quantitative assessment involves drilling wells We expect tilt meters to record data at intervals at
to have subsurface access for direct measurements least an order of magnitude smaller than any induced
such as geochemical sampling and pressure measure- surface deformation (which is typically hours to
ment. Indirect or remote monitoring will have larger months).
uncertainty. 65
Noncontinuous monitoring will require selection benefit of time intervals. In some cases, regulatory
of a time interval. Two examples are surface seismic requirements will define the initial monitoring
and geochemical fluid sampling. Both require active interval for various tools. For example, regulatory
use of instrumentation and both have some time compliance for postinjection site care can deter-
required for data processing and analysis. These mine time intervals that are required for the pro-
types of monitoring are likely to have variable periods ject and site. More frequent monitoring may be
between data collection. Initially some interval will be used because of public assurance, performance
selected, perhaps selected because of regulatory assessment, or risk reduction goals. For perfor-
requirement or public assurance goals. When moni- mance assessment, the monitoring interval will be
toring results conform to predictions of reservoir considered in an iterative loop with the reservoir
models (or other process models) the sampling inter- model – as monitoring results confirm or modify
val will be maintained or may be extended to reduce the model assumptions, the monitoring interval
cost. If some anomaly is detected, frequent repeated and model parameters will change. Thus, the mon-
monitoring may be used to confirm and understand itoring interval, like the monitoring method used,
the anomalous results. is subject to change over the life of a storage
In many areas of monitoring, the separation project.
between continuous and discrete measurements is
being reduced or eliminated in current research. References
For example, the use of borehole fluid sampling Allis, R. G., and Hunt, T. M. (1986). Analysis
with a U-tube device (Freifeld et al., 2005) has of exploitation-induced gravity changes at
reduced the time interval and incremental cost of Wairakei Geothermal Field. Geophysics, 51(8):
downhole samples by using a permanent or semi- 1647–1660.
permanent installation which allows sampling on Alnes, H., Eiken, O., and Stenvold, T. (2008). Monitoring
hourly intervals. Monitoring tools in the oil and gas production and CO2 injection at the Sleipner field
gas industry are being considered in the context of using time-lapse gravimetry. Geophysics, 73(6):
“permanent reservoir monitoring.” In this context, WA155–WA161.
permanent typically means the instrumentation is Bergmann, P., Schmidt-Hattenberger, C., Kiessling, D.,
permanent but the data collection interval may et al. (2012). Surface-downhole electrical resistivity
still be variable depending on goals and costs. tomography applied to monitoring of CO2
A permanent installation of instrumentation does storage at Ketzin, Germany. Geophysics, 77(6):
B253–B267.
facilitate more frequent monitoring and, often,
higher quality results (an example for surface seis- Calvert, R. (2005). Insights and methods for 4D reservoir
mic monitoring of injected CO2 is in Urosevic monitoring and characterization. Society of Exploration
Geophysicists Distinguished Instructor Series, No. 8.
et al. [2010] and Pevezner et al. [2011, 2017]).
Advances in technology, such as distributed acous- Cantillo, J. (2011). A quantitative discussion on time-lapse
repeatability and its metrics: Expanded Abstracts,
tic sensing with fiber optics, have the potential to
Society of Exploration Geophysicists. http://dx.doi.org/
greatly reduce the cost of repeat surveys via low- 10.1190/segam2012-0719.1
cost permanent sensing cables (Daley et al, 2015).
Carrigan, C. R., Ramirez, A. L., Newmark, R. L., Aines, R.,
Some monitoring may have its own protocol
and Friedmann, S. J. (2009). Application of ERT for
for time intervals that can depend on other mon- tracking CO2 plume growth and movement at the
itoring data. For example, fluid sampling in an SECARB Cranfield site. In 8th Annual Conference on
observation well may trigger seismic monitoring Carbon Capture & Sequestration, Pittsburgh, PA
if CO2 is detected when not expected. Because (Vol. 4, No. 7).
the cost of seismic monitoring is typically higher, Castagna, J. P., and Backus, M. M. ( 1993). Offset-dependent
this is also an example of cost-driven monitoring reflectivity – theory and practice of AVO analysis.
intervals. Society of Exploration Geophysicists Investigations in
Because the selection of a monitoring interval Geophysics No. 8.
is likely cost constrained, the goals of the project Castagna, J. P., and Swan, H. W. (1997). Principles of AVO
will be important factors in considering the cost/ crossplotting. Leading Edge, 16(4): 337–344.
66
Castagna, J. P., Swan, H. W., and Foster, D. J. (1998). Guidance, Office of Water (4606 M), EPA 816-
Framework for AVO gradient and intercept R-13–001. www.epa.gov/safewater
interpretation. Geophysics, 63(3): 948–956. EPA. (2015). Subpart RR: Geologic sequestration of carbon
Chapman, D.S., Sahm, E., and Gettings, P. (2008). dioxide, Greenhouse Gas Reporting Program
Monitoring aquifer recharge using repeated (GHGRP). http://www2.epa.gov/ghgreporting/sub
high-precision gravity measurements: A pilot study in part-rr-geologic-sequestration-carbon-dioxide
South Weber, Utah. Geophysics, 73(6): WA83–WA93. EPA. (2016). Greenhouse Gas Reporting Program
Cheng, A., Huang, L., and Rutledge, J. (2010). Time-lapse (GHGRP), 2016. www.epa.gov/ghgreporting
VSP data processing for monitoring CO2 injection. EU. (2009a). Implementation of the CCS Directive,
Leading Edge, 29: 196–199. Guidance Documents 1, 2, 3 and 4. http://ec.europa.eu/
Coueslan, M. (2013). Monitoring CO2 injection at the clima/policies/lowcarbon/ccs/implementation/docu
Illinois Basin – Decatur Project: Second monitor mentation_en.htm
survey. Presentation at MGSC Annual Meeting EU. (2009b). Directive 2009/31/EC of the European
September 2013. http://sequestration.org/resources/P Parliament and of the Council of 23 April 2009 on the
AGOct2013Presentations/09-Time-lapse_3DVSP-Mco geological storage of carbon dioxide and amending
ueslan_Sept2013.pdf Council Directive 85/337/EEC, Official Journal of the
Dafflon, B., Wu, Y., Hubbard, S. S., et al. (2012). Monitoring European Union, 5.6.2009. http://eur-lex.europa.eu/Le
CO2 intrusion and associated geochemical transforma- xUriServ/LexUriServ.do?uri=OJ:L:2009:140:0114:0135:
tions in a shallow groundwater system using complex EN:PDF
electrical methods. Environmental Science and Techno- Ferguson, J. F., Chen, T., Brady, J., Aiken, C. L. V., and
logy, 47(1): 314–21. http://dx.doi.org/10.1021/es301260e Seibert, J. (2007). The 4D microgravity method for
Daley, T. M., Myer, L. R., Peterson, J. E., Majer, E. L., and waterflood surveillance II: Gravity measurements for
Hoversten, G. M. (2008). Time-lapse crosswell seismic the Prudhoe Bay reservoir, Alaska. Geophysics, 72(2):
and VSP monitoring of injected CO2 in a brine aquifer. I33–I43.
Environmental Geology, 54: 1657–1665. DOI:10.1007/ Freifeld, B. M., Trautz, R. C., Kharaks, Y. K., et al. (2005).
s00254-007–0943-z. The U-tube: A novel system for acquiring borehole fluid
Daley, T. M., Miller, D. E., Dodds, K., Cook, P., and samples from a deep geologic CO2 sequestration
Freifeld, B. M. (2015). Field testing of modular borehole experiment. Journal of Geophysical Research, 110:
monitoring with simultaneous distributed acoustic B10203.
sensing and geophone vertical seismic profile at Goodway, B., Chen, T., and Downton, J. (1997).
Citronelle, Alabama. Geophysical Prospecting, 64(5): Improved AVO fluid detection and lithology
318–1334. DOI:10.1111/1365–2478.12324. discrimination using Lamé petrophysical
Davis, K., Li, Y., and Batzle, M. (2008). Time-lapse gravity parameters; λρ, μρ and λ/μ fluid stack”, from P and
monitoring: A systematic 4D approach with application S inversions. Expanded Abstracts, Society of
to aquifer storage and recovery. Geophysics, 73(6): Exploration Geophysicists, 183–186.
WA61–WA69. Hannas, S. D. (2013). Monitoring the geological storage of
Doetsch, J., Kowalsky, M. B., Doughty, C., et al. (2012). Fully CO2. Geological storage of carbon dioxide (CO2).
coupled hydrogeophysical inversion of CO2 migration Geoscience, technologies, environmental aspects and legal
data in a deep saline aquifer. In SEG-AGU frameworks, 54. Duxford, UK: Woodhead Publishing
Hydrogeophysics Workshop, July 8–11, 2012, Boise Series in Energy, 68–96.
State University, Boise, Idaho. Harbert, W., Purcell, C., and Mur, A. (2011). Seismic
Doetsch, J., Kowalsky, M. B., Doughty, C., et al. (2013). reflection data processing of 3D surveys over an EOR
Constraining CO2 simulations by coupled modeling CO2 injection. Energy Procedia, 4: 3684–3690.
and inversion of electrical resistance and gas Harbert, W., Daley, T. M., Bromhal, G. Sullivan, C., and
composition data. International Journal of Greenhouse Huang, L. (2016). Progress in monitoring strategies for
Gas Control, 18: 510–522. risk reduction in geologic CO2 storage. International
Duffaut, K., and Landrø, M. (2007). Vp/Vs ratio versus Journal of Greenhouse Gas Control, 51: 260–275.
differential stress and rock consolidation: DOI:10.1016/j.ijggc.2016.05.007.
A comparison between rock models and time-lapse Hovorka, S. D. (2013). Three-million-metric-ton-
AVO data. Geophysics, 72(5): C81–C94. monitored injection at the Secarb Cranfield Project:
EPA. (2013). Underground Injection Control (UIC) Project update. Energy Procedia, 37: 6412–6423. http://
Program Class VI Well Testing and Monitoring dx.doi.org/10.1016/j.egypro.2013.06.571 67
IEAGHG. (2017). Monitoring selection tool. http://ieaghg Mur, A., Purcell, C., Soong, Y., et al. (2011). Integration of
.org/ccs-resources/monitoring-selection-tool1 core sample velocity measurements into a 4D seismic
IPCC. (2005). IPCC Special Report on Carbon Dioxide survey and analysis of SEM and CT images to obtain
Capture and Storage. Prepared by Working Group III pore scale properties. Energy Procedia, 4: 3676–3683.
of the Intergovernmental Panel on Climate Change Nelson, E. B., and Guillot, D. (2006). Well cementing, 2nd
[Metz, B., O. Davidson, H. C. de Coninck, M. Loos, and edn. Houston: Schlumberger.
L. A. Meyer (eds.)]. Cambridge University Press, Peters, D. (2007). CO2 geological storage: Methodology and
Cambridge, United Kingdom and New York, NY, USA, risk management process. Presented at NHA Hydrogen
442 pp. Conference, March 20, 2007.
Johnson, D. H. (2013). Practical applications of time-lapse Pevzner, R., Shulakova, V., Kepic, A., and Urosevic, M.
seismic data. Society of Exploration Geophysics, (2011). Repeatability analysis of land time-lapse seismic
Distinguished Instructor Series, No. 16. data: CO2CRC Otway pilot project case study.
Kirksey, J. (2012). Deep well monitoring and verification at Geophysical Prospecting, 59: 66–77.
the Illinois Basin Decatur Project. Presentation at Pevzner, R., Urosevic, M., Popik, D., et al. (2017). 4D surface
MGSC Annual Meeting September 2012. http://seques seismic tracks small supercritical CO2 injection into the
tration.org/resources/PAGSept2012Presentations/06- subsurface: CO2CRC Otway Project. International
JimKirksey_PAG2012.pdf Journal of Greenhouse Gas Control, 63: 150–157. https://
Kowalsky, M. B., Doetsch, J., Commer, M., et al. (2016). doi.org/10.1016/j.ijggc.2017.05.008
Coupled inversion of hydrological and geophysical data Purcell, C., Harbert, W., Soong, Y., et al. (2009). Velocity
for improved prediction of subsurface CO2 migration. measurements in reservoir rock samples from the
Lawrence Berkeley National Laboratory Report, SACROC unit using various pore fluids and integration
NRAP-TRS-III-004–2016, Level III Technical Report into a seismic survey taken before and after a CO2
Series. sequestration flood. Energy Procedia, 1: 2323–2331.
Kragh, E., and Christie, P. (2002). Seismic repeatability, Purcell, C., Mur, A., Soong, Y., McLendon, T. R.,
normalized RMS, and predictability. Leading Edge, 21: Haljasmaa, I. V., and Harbert, W. (2010). Integrating
640–647. velocity measurements in a reservoir rock sample from
Landrø, M. (2001). Discrimination between pressure and the SACROC unit with an AVO proxy for subsurface
fluid saturation changes from time-lapse seismic data. supercritical CO2. Leading Edge, 29(2): 192–195.
Geophysics, 66(3): 836–844. Sayers, C. M. (2006). An introduction to velocity-based
Landrø, M., Hafslund Veire, H., Duffaut, K., and stress changes in sandstones. Leading Edge, 24(12):
Najjar, N. (2003). Discrimination between pressure 1262–1266.
and fluid saturation changes from marine Sayers, C. M. (2010). Geophysics under stress:
multicomponent time-lapse seismic data. Geophysics, Geomechanical applications of seismic and borehole
68(5): 1592–1599. acoustic waves. Society of Exploration Geophysicists
Lumley, D. (2001). Time-lapse seismic reservoir Distinguished Instructor Series, No. 13.
monitoring. Geophysics, 66: 50–53. Schlumberger. (1989). Cased hole log interpretation:
Lumley, D. (2010). 4D seismic monitoring of CO2 Principles/applications. www.slb.com/resources/publi
sequestration. Leading Edge, 29(2): 150–155. cations/books/ch_lipa.aspx
Lumley, D., Adams, D., Wright, R., Markus, D., and Cole, S. Schlumberger. (1991). Log interpretation: Principles/
(2008). Seismic monitoring of CO2 geo-sequestration: applications. www.slb.com/resources/publications/boo
Realistic capabilities and limitations. Expanded Abstracts, ks/lipa.aspx.
Society of Exploration Geophysicists, 2841–2845. Shapiro, S. (2015). Fluid induced seismicity. Cambridge:
Mathieson, A., Midgley, J., Dodds, K., Wright, I., Cambridge University Press.
Ringrose, P., and Saoul, N. (2010). CO2 sequestration Simm, R., and Bacon, M. (2014). Seismic amplitude:
monitoring and verification technologies applied at An interpreter’s handbook. Cambridge: Cambridge
Krechba, Algeria. Leading Edge, 29(2): 216–222. University Press.
Mavko, G., Tukerji, T., and Dvorkin, J. (2009). The rock Trainor-Guitton, W. J., Ramirez, A., Yang, X., Mansoor, K.,
physics handbook, 2nd edn. Cambridge: Cambridge Sun, Y., and Carroll, S. (2013). Value of information
University Press. methodology for assessing the ability of electrical
Mullins, O. G. (2008). The physics of reservoir fluids: resistivity to detect CO2/brine leakage into a shallow
68 Discovery through downhole fluid analysis. Houston: aquifer. International Journal of Greenhouse Gas
Schlumberger. Control, 18: 101–113. DOI:10.1016/j.ijggc.2013.06.018.
Urosevic, M., Pevzner, R., Kepic, A., and Wisman, P. Capture and Sequestration, Pittsburgh, PA, May 2–5,
(2010). Time-lapse seismic monitoring of 2011.
CO2 injection into a depleted gas Wright, I., Ringrose, P., Mathieson, A., and Eiken, O. (2009).
reservoir-Naylor Field, Australia. Leading Edge, An overview of active large-scale CO2 storage projects.
29(2): 164–169. In Society of Petroleum Engineers, SPE 127096, SPE
Wright, I. (2011). In Salah CO2 storage JIP lessons International Conference on CO2 Capture, Storage, and
learned. In 10th Annual Conference on Carbon Utilization, November 2–4, 2009, San Diego, CA.
69
Part II Geophysical Techniques
Chapter
Rock Physics of CO2 Storage Monitoring in
4 Porous Media
Thomas M. Daley
Introduction depths considered (and thus also pressures and tem-

Remote geophysical monitoring of carbon dioxide peratures) to those most likely for storage, taken here
(CO2) injection and storage typically starts with qua- as 0.8–4 km. The minimum depth of approximately
litative analysis, i.e., changes in monitoring data are 0.8 km is that which would allow supercritical phase
interpreted for the presence of CO2 in the subsurface – CO2 (a limit previously used by IPCC, 2005). Storage of
simply showing that we can “see” the CO2. However, supercritical phase CO2 is assumed because the
such detection is often insufficient for regulatory or increased fluid density allows more efficient storage.
scientific use. Once CO2 is detected the next step is The maximum depth of storage is a practical limit set
quantitative analysis – deriving estimates of volume by the cost of drilling and storage operations and is
and/or mass of CO2 in a given volume of the reservoir. stated here, somewhat arbitrarily, as 4 km. Notably, we
Quantitative analysis requires rock physics – theories also limit our discussion with the assumption that the
and models. The physics and chemistry that impact preexisting pore fluid is brine and thus consider only
monitoring the injection and long-term storage of brine–CO2 mixtures following injection. Brine aquifers
CO2 in rocks are a broad topic, with a full summary are the target because they provide the largest potential
deserving its own book length treatment. In this chap- storage capacity for geological sequestration of CO2
ter we limit the discussion to rock physics that (IPCC, 2005). We are avoiding the issues related to
impacts the geophysical monitoring of CO2 injection injection in reservoirs with hydrocarbons including pre-
and storage in porous media, with specific limiting existing natural gas (typically methane [CH4]). While
assumptions described in the text that follows. This multiple pore fluids add complexity, much work is
topic is a subset of the broad use of geophysics for available from the oil and gas industry in this area.
time-lapse monitoring of subsurface processes. To date, all large-scale storage projects (not enhanced
Johnston (2013) presents a good summary of seismic oil recovery [EOR] projects) have targeted CO2 injection
monitoring and defines seismic rock physics as “the in brine formations, such as the long running Sleipner
link between the static and dynamic properties of project (e.g., Arts et al., 2004; Chadwick et al., 2016).
a reservoir and the elastic properties of the reservoir The Otway project Stage 1, is one small (60 000-tonne)
rock” (p. 23). This link is what allows us to use remote pilot that did target a depleted methane reservoir for
monitoring data to detect and quantitatively estimate storage, and this work is well documented by Cook
the amount of stored CO2 in the subsurface, and holds (2014). To further constrain the scope of this chapter,
for whatever fluid state (liquid or gas) and mixture of we limit ourselves to the rock physics controlling remote
fluids is contained in the reservoir. However, with geophysical monitoring (e.g., reflection seismic ima-
increasing number of fluids and fluid states the rock ging) and do not consider issues specific to core mea-
physics relationships become more complex and surements, well logging, or direct CO2 detection
uncertainty increases. To understand rock physics (geochemical sampling). However, it is important to
relationships, as typically used in storage projects, note that many of the parameters needed for rock phy-
we first describe the limitations and assumptions sics calculations, and thus for interpretation of remote
used. monitoring, come from core and well logging studies.
In this chapter we restrict ourselves to focusing on Other chapters in this book will consider some aspects
realistic CO2 storage scenarios and limit the range of excluded here. 71
Thomas M. Daley
The most widely used remote geophysical monitor- where the bulk modulus Ksat is modified by the satur-
ing technique is active-source seismic (both surface and ating fluids in the rock pore space, including CO2.
borehole based measurements) with additional work in The shear-wave velocity is given by
electrical monitoring techniques (mostly electrical qffiffiffiffiffiffiffi

resistance tomography [ERT], again surface and bore- Vs ¼ μ ρ ð4:2Þ
hole; see Chapter 9 by Gasperikova and Commer, this
volume) and deformation (geomechanical) monitoring, and depends on the shear modulus of the rock, μ,
typically with surface-based measurements (e.g., inter- which is assumed to be unaffected by fluids, and the
ferometric synthetic aperture radar [InSAR], GPS, and rock density, ρ, which is affected by fluids in the rock.
tilt monitoring surveys; see Chapter 6 by Vasco et al., Thus, understanding the effect of CO2 on seismic
this volume). Rutqvist (2012) has reviewed geomecha- velocity requires understanding the effect of CO2 on
nical monitoring of CO2, which looks at deformation elastic moduli K and ρ.
due to pore pressure changes. Gasperikova and Common use of elastic moduli for monitoring
Hoversten (2008) consider both electrical and gravity fluids in rocks began with Fritz Gassmann’s pioneer-
monitoring. Gravity monitoring uses changes in the ing study of elasticity in porous media (Gassmann,
Earth’s gravity field due to CO2 displacing pore fluid 1951) showing the impact of fluid-filled pores on the
of different density (Chapter 7 by Eiken, this volume). bulk modulus and the resultant seismic velocity.
Here we will focus on the rock physics important to Analysis of pore fluid changes (the “fluid substitution
active-source seismic and electrical monitoring. problem”) such as CO2 displacing brine typically is
called “Gassmann substitution” and has been restated
Seismic Monitoring: Elastic Moduli and summarized often, including by Berryman
A fundamental measurement in active seismic mon- (1995), Mavko et al. (1998), Wang (2001), and Smith
itoring is the travel time of a seismic wave propagating et al. (2003). These authors take effort to remind read-
between a controlled source and a sensor (typically ers of underlying assumptions (such as homogeneous
one of an array of sensors). This travel time is con- isotropic rock with fully connected pore space and
trolled by the elastic wave velocity of the subsurface. zero viscosity fluid) as well as usage pitfalls; these
A second important measurement is the amplitude of should be heeded as they often impact quantitative
waves reflected from velocity interfaces. Reflection analysis of seismic data. Wang (2001) summarizes the
amplitudes are controlled by the impedance contrast, assumptions as follows:
where impedance is the product of formation velocity 1. The rock (both the matrix and the frame) is
and density. For wave propagation in a linear elastic macroscopically homogeneous.
medium, the wave velocity is controlled by the elastic 2. All the pores are interconnected or
moduli, originally introduced by Robert Hooke as communicating.
Hooke’s “Law of Nature” for deforming materials 3. The pores are filled with a frictionless fluid (liquid,
(Hooke, 1678), which linearly related stress and gas, or mixture).
strain. Two elastic moduli for isotropic linear elastic 4. The rock–fluid system under study is closed
rocks, λ and μ, were defined by G. G. Stokes (1845), (undrained).
and used along with density, ρ, to define the velocity of 5. The pore fluid does not interact with the solid in
P- and S-waves. Further developments in elasticity, a way that would soften or harden the frame.
and the various elastic moduli, are summarized in
These assumptions are known to be problematic in
many books (e.g., Mavko et al., 1998; Aki and
experimental work (field and laboratory scale).
Richards, 2002). For seismic monitoringpffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi of CO2, the However, in general, Gassmann’s assumptions are
bulk modulus, K, where K ¼ ðλ þ 2μÞ, is perhaps
assumed to hold for unconsolidated clastic rocks at
the most important elastic moduli because of its sen-
low seismic frequencies (<100 Hz). This is the case for
sitivity to pore fluid. The P-wave seismic velocity (Vp)
typical surface seismic monitoring of saline storage of
of an isotropic rock volume is given by:
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi CO2, and thus we follow the Gassmann equation
ffi approach to accurately determine the relationship
4μ
Vp ¼ Ksat þ =ρ ð4:1Þ between the material properties of CO2 (such as com-
72 3
pressibility and density) and the corresponding
Rock Physics of CO2 Storage Monitoring
seismic wave velocity of rocks containing CO2. ρfl ¼ Sw ρw þ Sg ρg ð4:5Þ

The simplifying assumptions used allow estimation of
partial CO2 saturation with only seismic P-wave velo- Equations (4.1)–(4.5) show that changes in P-wave
city changes. If more information is available, then velocity from CO2 displacing brine in the pore space
more complete (and complex) rock physics theories are determined by the bulk modulus and density of
can be used. For example, the effects of fluid pressure brine and CO2 and their partial saturations (other
can be important for geophysical monitoring. Recently rock properties such as K* and μ are determined
Pride et al. (2016) developed “explicit analytical models from core or well logs and typically assumed to be
for how the seismic velocities, porosity, electrical con- static). While supercritical CO2 has a range of com-
ductivity and fluid-flow permeability vary with effec- pressibility (1/Kg) and density, for the typical pres-
tive stress,” where effective stress is the difference sures and temperatures of storage depths the
between confining stress and pore fluid pressure. compressibility is most often the dominant factor
The notation of elastic moduli varies somewhat controlling P-wave velocity change from brine displa-
between authors and problems addressed. We will cement (e.g., as described in Figure 4.1). It is funda-
start with notation following Smith et al. (2003). mental to seismic monitoring of CO2 storage that
The fundamental relationship in Gassmann fluid sub- changes in CO2 saturation due to injection can be
stitution expresses the bulk modulus of a saturated determined from changes in seismic (P-wave) velocity
rock, Ksat, in terms of the “dry frame” bulk modulus, using these equations, within the given assumptions.
K*, the bulk modulus of the minerals making the rock To apply these relationships for a specific site, knowl-
frame, K0, the bulk modulus of the pore fluid, Kfl, and edge of the brine and CO2 properties is needed.
the rock porosity, ϕ:

Brine and CO2 Properties
ð1 KK0 Þ2 For any given storage project, with site-specific sub-
Ksat ¼ K þ ϕ
ð4:3Þ
þ ð1ϕÞ

K0 K 2 0
K surface conditions, the properties of brine and CO2
Kfl
in the reservoir need to be understood for rock
The bulk rock density, ρB , is found from the frame physics analysis. For brine, the most commonly
or grain density, ρg , and the density of the saturating used estimates come from work published by
fluids, ρfl , as: ρB ¼ ρg ð1 ϕÞ þ ρfl ϕ. The shear mod- Batzle and Wang (1992), who developed relation-
ulus, μ, is generally assumed to be unaffected by fluid ships for the density and sonic velocity of brines as
properties, i.e., an assumption of no geochemical a function of pressure, temperature, and salinity (for
alteration of cement or grains by the fluids – an sodium chloride [NaCl]). These relationships are
assumption questioned for some CO2 injections and Eqs. 4.27b and 4.29 in Batzle and Wang (1992)
discussed later. and are also summarized in Mavko et al. (1998).
While Eq. (4.3) shows the effect of a single pore The properties of CO2 are most commonly taken
fluid on the bulk modulus Ksat and on Vp via Eq. (4.1), from the database of the National Institute of
we need an equivalent Kfl and ρfl for two fluids filling Standards and Technology (NIST). The NIST results
the pore space, such as for CO2 and brine. The fluids for CO2 density and sonic velocity are available in
will have variable partial saturations Sw (brine) and Sg tabular form from the website: http://webbook.nist
(CO2), where Sw + Sg = 1. Assuming the pore space has .gov/chemistry/fluid/, which is periodically updated
two homogeneous, uniformly distributed fluids (brine with recent improvements in data and equations of
and CO2), with individual bulk moduli Kw (brine) and state for CO2 and other fluids. An example of calcu-
Kg (CO2), the Reuss (isostress) average as described by lations of brine and CO2 properties for a storage site
Wood’s equation (Mavko et al., 1998) gives the com- is shown in Figure 4.1, which has isothermal density
bined-fluid bulk modulus as: and bulk modulus for a range of pressures corre-
sponding to the Cranfield, Mississippi, DAS project
Sw Sg 1 site (Hovorka et al., 2013). In this example we see
Kfl ¼ þ ð4:4Þ
Kw Kg that density contrast is large and variable over
a range of pore pressures, but the bulk modulus
and similarly, the fluid density is a saturation
contrast is an order of magnitude larger, which is 73
weighted average:
typical for supercritical CO2 at depth.
Thomas M. Daley
Isothermal : T = 120 C Figure 4.1 Isothermal (T = 120°C) density

1300 3500 (left) and bulk modulus (right) for brine (200
CO2 ppt NaCl) and CO2. The temperature, brine
1200 Brine (146000 ppm) properties, and pressure range are based on
3000 a CO2 storage site near Cranfield, Mississippi.
(Courtesy Jonathan Ajo-Franklin.)
1100
2500
1000
Bulk Modulus (MPa)

Density (kg/m )
3
900 2000
800 1500
700
1000
600
500
500
400 0
3000 4000 5000 6000 3000 4000 5000 6000
Pressure (psi) Pressure (psi)
In general, the acoustic properties of brine–CO2 impact when brine and CO2 are not homogeneously
solutions for conditions of geological storage have distributed in all pore space. This assumption is impli-
limited experimental data. Also, while the presence cit in Wang’s (2001) assumption 3 listed earlier. Rock
of pure pore filling phases of brine and CO2 are physics theories that consider inhomogeneous distri-
common assumptions for Gassmann equation based butions are typically termed “patchy saturation” mod-
calculations, it should be noted that CO2 will dissolve els because the individual fluids have isolated
into brines at variable rates, creating a mixed phase. “patches.” As we know that an assumption of homo-
This dissolution is typically considered as part of geneous fluid mixing is not valid for the naturally
reservoir modeling of injection and flow, and, impor- heterogeneous subsurface, using patchy saturation
tantly, leads to a slightly acidic solution including models is advisable when possible (for a case study,
carbonic acid (H2CO3). This can lead to geochemical see Chapter 18 by Grude and Landrø, this volume).
alteration of the reservoir rock frame (matrix), with These models lead to frequency dependence in both
implications discussed in the text that follows. seismic velocity and attenuation. Gassmann’s
Furthermore, some trace amounts of hydrocarbons assumption is valid only for very long wavelengths
(gas or liquid) can be found in reservoirs considered (low frequencies) as compared to the scale of fluid
“saline,” and this can impact the rock physics, and be patches. Because the patch size is unknown at field
a “pitfall” in quantitative interpretation. While we are scale, more information is needed to constrain the
not considering these complications here, they should velocity–saturation relationship. Measurement of
be assessed for each project site. attenuation along with velocity (and their frequency
dependence) is thus a way to obtain more accurate
estimates of partial CO2 saturation from seismic mon-
Patchy Saturation itoring data. The frequency dependence becomes
As stated earlier, Gassmann fluid substitution has important when using seismic data with higher fre-
many simplifying assumptions which can be modified quencies such as crosswell seismic measurements
when more information is available. One of the most (typically 500–2000 Hz) or sonic logging data (typi-
common extensions of the first-order Gassmann ana- cally 10 000 Hz or higher), rather than surface seismic
74
lysis described in the foregoing is to consider the (typically below 150 Hz).
0 0.18
WDO, 15 cm patch WDO, 15 cm patch
−50 WDO, 2.5 cm patch WDO, 2.5 cm patch
0.16
Gassmann + Reuss
Gassmann + Voigt
−100
BGH 0.14
−150
P−wave Attenuation
0.12
−200
0.1
ΔVp
−250
[A] 0.08
[B]
−300
0.06
−350
0.04
−400
−450 0.02
−500 0
0 0.5 1 0 0.5 1
CO2 Saturation (fraction) CO2 Saturation (fraction)
Figure 4.2 (A) Change in P-wave velocity as a function of CO2 saturation for various rock physics models and (B) P-wave attenuation as
a function of CO2 saturation for two patch sizes in the WDO model. (From Daley et al., 2011.)
An early patchy saturation model was proposed by attenuation on wavelength (and thus frequency).
White (1975), with corrections made by Dutta and A further extension allowing for variable (random)
Seriff (1979). White’s model assumes a homogeneous length scales of patch size has been developed
rock frame with concentric spherical fluid patches, by Müller and Gurevich (2004) and Toms et al.
a central sphere of one fluid and an outer spherical (2007).
shell saturated with a second fluid phase, in our case For patchy saturation models the change in Vp due
brine and supercritical CO2. to CO2 displacing brine will be less than the
White’s model also assumes that the seismic wave- Gassmann model with Reuss fluid mixing (as
length is larger than the characteristic patch (sphere) described earlier), so Gassmann with Reuss serves as
dimension and the results are dependent on seismic a lower bound on ΔVp and there is an upper bound
wavelength. defined by “Voigt” fluid mixing (Mavko et al., 1998),
Following the early results of White, the patchy where the bulk modulus is given by
saturation model has been generalized to more realis- Kfl ¼ Sw Kw þ Sg Kg ð4:6Þ
tic fluid distributions. Recent summaries of patchy
saturation models include Toms et al. (2007) which can be compared to Eq. (4.4). A typical com-
and Müller et al. (2010). Pride (2005) and Pride et al. parison of the difference between rock physics models
(2004) use a characteristic length scale, L, for patches for a CO2 /brine mix is shown in Figure 4.2.
of uniformly spaced but arbitrarily shaped fluid dis- Figure 4.2(A) compares White’s model (WDO),
tribution. This model allows for the pressure redis- with two example patch sizes, to three other poroelas-
tribution between fluids, over a length scale, during tic models used to predict the properties of rocks
compression and dilation of a passing P-wave, with- partially saturated with CO2 including Gassmann
out specifying the geometry of the distribution within fluid substitution with either Reuss or Voigt effective
L. The Pride model also allows for and requires esti- fluid assumptions (Mavko et al., 1998) and the
mates of fluid properties such as viscosity. This length so-called Biot–Gassmann–Hill model (BGH) (Hill,
1963), which is a quasi-static prediction for
75
L is then related to the dependence of velocity and
Thomas M. Daley
0.18
S = 15% 2.5 cm
7.5 cm
0.16 S = 10% 10 cm
15 cm
0.14
0.12 Increasing
Saturation
0.1
Δ 1/Qp
0.08
0.06
0.04
0.02
0
−500 −450 −400 −350 −300 −250 −200 −150 −100 −50 0
Δ Vp (m/s)
Figure 4.3 Change in P-wave attenuation (1/Qp) as a function of change in P-wave velocity for four patch sizes using White’s model. CO2
saturation has characteristic values as shown by the two dashed lines for saturations of 10% and 15%. Rock properties taken from Frio
formation (Daley et al., 2011). (Figure courtesy of Jonathan Ajo-Franklin.)
a partially saturated medium with macroscopic however, these scales are limited in their spatial
patches. This example comes from monitoring a range. Another option is to constrain the patch size
CO2 injection test in the Frio formation in southeast by field scale measurement of other complementary
Texas (Daley et al., 2007). The elastic properties of the geophysical properties such as P-wave attenuation,
rock frame were selected from log and core informa- typically referred to as Q (quality factor; see Mavko
tion and are stated in Daley et al. (2011), with a Vp = et al., 1998). The use of P-wave attenuation is attrac-
2700 m/s and porosity = 25%. CO2 and brine proper- tive since it has complementary information as shown
ties were calculated for in situ reservoir pressures and in Figure 4.2B. Combining the measurement of
temperatures (P = 15 MPa, T = 55°C). The BGH change in velocity with change in attenuation can
model is the quasi-static prediction for a partially lead to a unique value of saturation and patch size
saturated medium with macroscopic patches. All cal- from seismic data, as shown in Figure 4.3.
culations were made for a seismic frequency of Assuming that the impact of patchy saturation is
1100 Hz. This frequency is typical for crosswell seis- accounted for, or may not be needed due to use of
mic data, which is higher frequency than surface seis- low-frequency data, there is another assumption from
mic and lower frequency than sonic well logging. typical fluid-substitution models that can have first-
As seismic monitoring uses higher frequency to order impacts on seismic monitoring data: rock frame
obtain better spatial resolution, the need for patchy changes.
saturation models is increased. All patchy saturation
models invoke some patch size property, such as Rock Frame Changes: Geochemical
White’s radius or Pride’s characteristic length, which
then becomes another unknown in the rock physics Alteration
analysis. In practice, we typically have no prior knowl- In addition to complications and uncertainty intro-
edge of patch size, which leads to considerable uncer- duced by either the assumption of homogeneous fluid
tainty when attempting to estimate CO2 saturation mixing or invoking models for the effects of patchy
from changes in seismic velocity using patchy satura- saturation, the underlying assumption of constant
tion models. A patch size value could come from frame properties (Wang’s assumption 5 in the fore-
76
calibration experiments on the core or log scale; going) has been shown to be problematic for CO2
Compare CO2 with Porosity Increase Compare CO2 with Porosity Increase
3600 A 2000
Initial Velocity CO2 S
aturatio B CO2 Saturation
Pre-Injection n, Con
Error in Interpreted CO2 Saturation 1950 stant F Initial Velocity
rame
3500 C Pre-Injection
B C A
1900 e
patchy CO2 saturation ng
20% ha
ec
Constant Saturation, Frame change

3400 1850 m
CO2 S + Fra
aturatio on
n, Con ati
Vp (m/s)
Vs (m/s)
B’ 40% C’ 1800 stant F tur
rame a
2s
3300 B’’ all water, porposity +3% CO
D 1750
C’’ 70% D 100% water
Porosity change
3200 1700 +3%
uniform CO2 saturation
1650
3100
100% water
all water, porposity +5% 1600 D’
D’ Porosity change
+5%
3000 1550
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Sw Sw
Figure 4.4 A numerical model example of seismic velocity changes from patchy and uniform fluid distribution along with rock frame change
for CO2 displacing water in pore space. (A, left) P-wave velocity (Vp) versus water saturation (Sw). (B, right) S-wave velocity (Vs) versus water
saturation (Sw). (From Vanorio et al., 2010.)
injection/storage applications. Recent work has of rock frame change on P-wave and S-wave velocities
shown that constant frame assumption can lead to as a function of water saturation (Sw) for a water/CO2
significant errors in calculation of elastic moduli and mixture. Figure 4.4 is from a numerical model of
thus in the inverse calculation of saturation from a carbonate reservoir at a depth of 2,580 m. This
measured moduli (e.g., Vialle and Vanorio, 2011; type of calculation is often used to interpret observed
Lebedev et al., 2014). A summary of recent advances changes in seismic velocity (from time-lapse monitor-
in understanding of CO2 induced geochemical frame ing data) for CO2 saturation in each volume of the
changes is given in Vanorio (2015). reservoir, but rarely considers frame alteration.
When considering alteration of rock frames for Starting at point A in Figure 4.4(A) (the initial
CO2 storage, we can consider the common rock “preinjection” P-wave velocity) a change of 100 m/s
types of siliciclastic and carbonate. Carbonates are in velocity corresponds to a 20% change in CO2
found to be especially reactive to CO2 mixing with saturation for a patchy distribution (A to B, in blue)
brine as pore filling fluid. Vanorio et al. (2011) con- or <1% CO2 saturation change for a homogeneous
sider four geochemical “scenarios” that can alter the fluid distribution (A to C, in red). Comparing A to B0
carbonate rock frame: versus A to C0 , a larger decrease in observed velocity
(250 m/s) leads to interpreted CO2 saturation chan-
1. Mineral dissolution leading to porosity increase
ging either 40% or 1% for patchy or homogeneous
2. Mineral dissolution leading to matrix cementation
mixing, respectively. Now consider a geochemical
decrease
rock frame change, an increase in porosity of 3%.
3. Mineral dissolution leading to both porosity
The initial, preinjection velocity drops by nearly
increase and cementation decrease
300 m/s (from point A to D). This example shows
4. Salt precipitation leading to porosity decrease that a 3% change in porosity can lead to a 70% error in
Each of these has differing, but nonnegligible, estimated saturation, depending on the “patchyness”
impacts on elastic moduli and seismic velocity via of the CO2 /water distribution. A 5% change in por-
the rock frame properties. It is notable that rock osity (A to D0 ) causes a larger velocity change (550 m/
frame properties often are derived from preinjection s) than any constant-frame fluid substitution model
well logs and core samples and therefore do not allow could explain. As noted by Vanorio et al. (2010),
for any of these changes. Laboratory-scale core mea- a similar change could also be caused by reduction
surements and/or post-injection well logging are in frame moduli due to change in grain contact stiff-
needed to understand geochemical alterations. ness (e.g., mineral/cement dissolution) without
Figure 4.4 shows the impact of patchy versus invoking porosity change. Assessing and correcting 77
homogeneous fluid saturation along with the impact for changes in frame properties is clearly important
Thomas M. Daley
for storage monitoring and will require more than just developed an interpretation of rock frame change
P-wave seismic data. utilizing measured Vs changes (and using supporting
geological, well log, and geochemical data) and using
a grain cementation model (Dvorkin and Nur, 1996;
Use of Multiphysics Monitoring Avseth et al., 2000). They propose a variable reduction
The size of potential errors in interpreted CO2 satura- in contact cement (from initial 0.1% to as low as
tion indicated in Figure 4.4 demonstrates the impor- 0.01%) that leads to a reduction in frame moduli
tance of correctly accounting for all the components (bulk and shear) of 30–40%. From the modified
of rock physics impacting seismic monitoring. frame moduli, an improved estimate of CO2 satura-
Understanding time-varying frame changes requires tion is made using patchy saturation models for
models of reactive flow and transport to be used with a 0.2-m patch size and the 300-Hz seismic data.
geochemical models and laboratory data to interpret As stated by Al Hosni et al., “having P- and S-wave
time-lapse geophysical monitoring data. For example, velocity time-lapse data is key to improved saturation
recent work in geochemical modeling of CO2 miner- estimates.”
alization has utilized advanced imaging techniques to
understand the reactive surface area of CO2 reservoir
rocks (e.g., Landrot et al., 2012). Electrical Conductivity
Using multiphysics monitoring data can aid in the Electrical conductivity is an important rock prop-
estimation of frame changes induced by CO2 storage, erty for remote monitoring of CO2 storage, and it
as well as provide independent estimates of saturation, can be measured from a variety of electromagnetic
and thus improve saturation estimates from monitor- (EM) methods, as summarized by Nabighian (1991)
ing data. Examples of multiphysics monitoring include and Gasperikova and Commer (Chapter 9, this
acquiring seismic P-wave and shear-wave velocities, or volume). EM geophysical methods cover a wide
electrical resistivity and seismic P-wave velocity. range of frequencies from DC (0 Hz) to gigahertz.
Figure 4.4B shows the impact of having S-wave velocity Fundamental EM properties of interest include
data. Starting from a preinjection S-velocity at point A, conductivity, σ, and dielectric permittivity, ε, both
with no frame change, VS would increase with CO2 of which are sensitive to the electrical properties of
saturation, e.g., approx. 50 m/s from point A to point pore fluids. For earth materials, high-frequency
B/C (the same increase for patchy or homogeneous electromagnetic waves propagate following a classic
fluid distribution). This velocity increase is due to the wave equation largely controlled by permittivity,
decrease in bulk density due to lower density CO2 while at low frequencies propagation becomes
displacing water. However, when we include frame a diffusion process largely controlled by conductiv-
change, for example, a –3% porosity change (point ity. The transition frequency for typical crustal
D), the shear velocity would now decrease by about rocks is at about 100 kHz (Mavko et al., 1998).
150 m/s. This drop in Vs would be a very different CO2 storage is at depths typically accessible to
monitoring observation than a small Vs increase, and EM methods below this frequency. For example,
thus provides complementary information to Vp by a common high-frequency EM monitoring tool is
indicating changing frame properties. ground penetrating radar (GPR), typically at
An example of field data utilizing Vs has been 0.1–2 GHz, which is sensitive to permittivity.
described by Daley et al. (2008) and Al Hosni et al. However, the typical depth of penetration of GPR
(2016a) using time-lapse crosswell seismic velocity is centimeters to meters and thus has little applica-
tomography at the Frio Brine Pilot Project (Hovorka tion to CO2 storage monitoring. Lesmes and
et al., 2006). Analysis of both crosswell and vertical Friedman (2005) review theoretical and empirical
seismic profile (VSP) P-wave data indicated Vp models for electrical methods, with high-frequency
decreases following CO2 injection that were larger permittivity measurements having a focus on
than could be explained by fluid substitution for the hydrogeophysics and application to shallow soils.
siliciclastic Frio formation (Daley et al., 2008; Al For the depths of interest to CO2 storage (below
Hosni et al., 2016b). The Vs change measured by 800 m), lower frequency methods focus on measuring
crosswell showed a decrease in velocity, not the slight the change in conductivity due to electrically resistive
78
increase predicted by fluid substitution. Al Hosni et al. CO2 displacing electrically conductive brine in the
reservoir pore space. Lower frequency electrical mea- saturation SCO2 = 1 – Sw, and porosity and brine
surements, such as electrical resistivity [ER], and its conductivity are taken from separate measurement,
use for electrical resistance tomography [ERT], along thus
with electromagnetic induction methods, can have
R ¼ ϕm Rw ð1 Sco2 Þn ð4:9Þ
a depth of penetration of 102 to >103 m and have
been used for remote monitoring of injected CO2. and
For CO2 storage applications, EM methods use rock n1
ϕm Rw
physics developed to assess and monitor reservoir Sco2 ¼ 1 ð4:10Þ
fluids, typically assumed to be initially brine, and R
then mixed brine and CO2. For time-lapse monitoring using electrical con-
The monitoring of reservoir fluids with electrical ductivity measurements, the calculation is typically
methods began with well logging (c. 1927). simplified further (e.g., Nakatsuka et al., 2010) by
Widespread quantitative application followed the pub- considering the ratio of fully brine saturated resistiv-
lication by G. E. Archie of “Archie’s Law” (Archie, ity and partially saturated resistivity, termed the resis-
1942), which is not a law, but an empirical fit of core tivity index (Guéguen and Palciauskas, 1994),
and well log data for saturated porous rocks. Archie R
defines a “formation resistivity factor,” F, and relates RI ¼ ¼ Sw n ð4:11Þ
R0
F to the bulk resistivity of a brine saturated rock, R, the
resistivity of the brine, Rw, and the rock’s porosity, ϕ, as Thus
n1
R 1
F¼ ¼ ϕm ð4:7Þ; Sco2 ¼1 ð4:12Þ
Rw RI
where m is an empirical exponent fitted at m = 1.3 for Often n is simply assumed to be 2 based on general
unconsolidated laboratory sand packs and at m = use, e.g., Carrigan et al. (2013), who cite Nakatsuka
1.9–2.0 for consolidated sandstone cores. Archie’s et al. (2010). Bergmann et al. (2012) also follow
work has been given more theoretical underpinning, Nakatsuka et al. in using a ratio of time-lapse resistiv-
for example, by Sen et al. (1981) and then Berryman ity measurement to obtain CO2 saturation with the
(1995), who gives bounds on F and shows, using saturation index n obtained from laboratory experi-
a differential effective media (DEM) approach, that ments at in situ conditions (n = 1.62).
for nonconducting spherical beads in a conducting The use of Archie’s equations becomes proble-
fluid m = 1.5. Berryman also shows model dependence matic as the clay content increases. For example,
on porosity with the spherical DEM model is best at Nakatsuka et al. (2010: 209) state that “Resistivity
porosities of 25–30%, while a self-consistent effective values from the induction log increased during the
medium theory with needle-shaped pores is best for CO2 injection period but, when calculated by Archie’s
porosities of approx. 15%. equation it shows CO2 saturation of about 10% which
Importantly, Archie and others following him is lower than that estimated from neutron logging
assume that formation brine has such high conduc- data. This low saturation is assumed as the effect of
tivity that other formation fluids and gases can be clay inclusion in the reservoir.”
assumed to have no contribution to the conductivity. For reservoirs with measurable clay or shale
Archie’s data and analysis show that for water satura- content, surface conduction mechanisms on the
tion, Sw, and a fully saturated rock resistivity, R0, the rock frame become important. Various modifica-
in situ measured resistivity of a partially saturated tions to Archie’s equations have been proposed,
rock, R, is given by although there is not yet a fundamental under-
standing of the mechanism of surface conductance
R ¼ R0 Sw n ¼ ϕm Rw Sw n ð4:8Þ;
for arbitrary complex frame minerals (S. Pride,
where n is termed the saturation exponent and n = personal communication, 2017). To allow for clay/
approx. 2.0 for typical sandstone reservoir rocks. shale content, Archie’s Eq. (4.7) is often written as
For monitoring CO2 in brine-saturated forma- F ¼ aϕm , where a is near 1 for clean sands but is
tions, Archie’s Eqs. (4.7) and (4.8) form the basis of found empirically for more clay-rich sands (Mavko 79
quantitative interpretation of resistivity, where CO2 et al., 1998). Waxman and Smits (1968) proposed
Thomas M. Daley
a model using the cation-exchange capacity and Transactions of the American Institute of Mining,
charge per unit pore volume, Qv, and an ion mobility Metallurgical, and Petroleum Engineers, 146: 54–62.
term, B, such that BQ F ¼ σ s ; where σ s is the surface
v
Arts, R., Eiken, O., Chadwick, A., Zweigel, P., van der
conductivity. Sen et al. (1988), among others, have Meer, L., and Zinszner, B. (2004). Monitoring of CO2
developed further refinements to B and the injected at Sleipner using time-lapse seismic data.
Waxman–Smits model, as summarized by Lesmes Energy, 29: 1383–1392.
and Friedman (2005), Worthington (1985) and Avseth, P., Dvorkin, J., Mavko, G., and Rykkje, J. (2000).
Mavko et al. (1998). Rock physics diagnostic of North Sea sands: Link
between microstructure and seismic properties.
Further extensions include the recent work of
Geophysical Research Letters, 27: 2761–2764.
Pride et al. (2016), who consider the stress and fluid DOI:10.1029/ 1999GL008468.
pressure impacts on conductivity in the context of an
Batzle, M., and Wang, Z. (1992). Seismic properties of pore
integrated flow–mechanics–electrical theory.
fluids. Geophysics, 57: 1396–1408.
The stress-dependent Pride model is important for
crystalline rocks (where fractures have strong sensi- Benson, S. (2008). Multi-phase flow of CO2 and brine in
saline aquifers. Expanded Abstracts, Society of
tivity to effective stress) but is less likely to be impor- Exploration Geophysicists, 27: 2839. DOI:10.1190/
tant for CO2 storage applications in porous media. 1.3063934.
However, geochemical changes in rock frame proper-
Benson, S., Tomutsa, L., Silin, D., Kneafsey, T., and
ties due to CO2 displacing brine could have an impact Miljkovic, L. (2005). Core scale and pore scale studies of
on electrical properties, especially surface conduc- carbon dioxide migration in saline formations.
tance. Therefore, the use of electrical methods for Lawrence Berkeley National Laboratory Report, LBNL-
quantitative estimation of CO2 saturation should be 59082. http://repositories.cdlib.org/lbnl/LBNL-59082
considered as having uncertainty that is difficult to Bergmann, P., Schmidt-Hattenberger, C., Kiessling, D.,
characterize with existing rock physics models. et al. (2012). Surface-downhole electrical resistivity
In practice, electrical resistivity monitoring, espe- tomography applied to monitoring of CO2 storage at
cially crosswell ERT, has proven very useful in monitor- Ketzin, Germany. Geophysics, 77(6): B253–B267.
ing changes in CO2 saturation within a reservoir Berryman, J. G. (1995). Mixture theories for rock properties.
(Bergmann et al., 2012; Carrigan et al., 2013). Also, In T. J. Ahrens (ed.), Rock physics & phase relations:
Dafflon et al. (2012) demonstrated the use of electrical A handbook of physical constants. Washington, DC:
methods, including ERT, to monitor the migration of American Geophysical Union. DOI:10.1029/RF003,
205–228.
groundwater with dissolved CO2. As the electrical
response of rocks is largely independent of the mechan- Biot, M. A. (1956). Theory of propagation of elastic waves in
a fluid-saturated porous solid. 1. Low-frequency range.
ical response, seismic and electrical measurements pro-
Journal of the Acoustical Society of America, 28(2):
vide complementary estimates of CO2 saturation and 168–178.
should be able to reduce uncertainty in a joint inversion
Carrigan, C. R., Yang, X., LaBrecque, D. J., et al. (2013).
methodology.
Electrical resistance tomographic monitoring of CO2
movement in deep geologic reservoirs. International
Journal of Greenhouse Gas Control, 18: 401–408. http://
References dx.doi.org/10.1016/j.ijggc.2013.04.016.
Aki, K. and Richards, P. G. (2002). Quantitative seismology. Chadwick, R. A., Williams, G. A., and White, J. C. (2016).
Sausalito, CA: University Science Books. High-resolution imaging and characterization of a CO2
Al Hosni, M., Caspari, E., Pevzner, R., Daley, T. M., and layer at the Sleipner CO2 storage operation, North Sea
Gurevich, B. (2016a). Case history: Using time-lapse using time-lapse seismics. First Break, 34(2): 77–85.
vertical seismic profiling data to constrain velocity– Cook, P. J., ed. (2014). Geologically storing carbon: Learning
saturation. Geophysical Prospecting, 64(4): 987–1000. from the Otway Project experience. Melbourne: CSIRO
Al Hosni, M., Vialle, S., Gurevich, B., and Daley, T. M. (2016b). Publishing.
Estimation of rock frame weakening using time-lapse Dafflon, B., Wu, Y., Hubbard, S. S., et al. (2012). Monitoring
crosswell: The Frio Brine Pilot Project. Geophysics, 81: CO2 intrusion and associated geochemical
B235–B245. DOI:10.1190/GEO2015-0684.1. transformations in a shallow groundwater system using
Archie, G. E. (1942). The electrical resistivity log as an aid in complex electrical method. Environmental Science and
80 determining some reservoir characteristics. Technology 47(1): 314–321.
Daley, T. M., Solbau, R. D., Ajo-Franklin, J. B., and [Metz, B., O. Davidson, H. C. de Coninck, M. Loos, and
Benson, S. M. (2007). Continuous active-source L. A. Meyer (eds.)]. Cambridge University Press,
monitoring of CO2 injection in a brine aquifer. Cambridge, United Kingdom and New York, NY, USA,
Geophysics, 72(5): A57–A61. DOI:10.1190/1.2754716. 442 pp.
Daley, T. M., Myer, L. R., Peterson, J. E., Majer, E. L., and Johnston, D. H. (2013). Practical applications of time-lapse
Hoversten, G. M. (2008). Time-lapse crosswell seismic seismic data. Society of Exploration Geophysicists
and VSP monitoring of injected CO2 in a brine aquifer. Distinguished Instructor Series No. 16. http://dx.doi.org/
Environmental Geology, 54: 1657–1665. DOI:10.1007/ 10.1190/1.9781560803126
s00254-007–0943-z. Landrot, G., Ajo-Franklin, J., Yang, L., Cabrini, S., and
Daley, Thomas M., Ajo-Franklin, J. B., and Doughty, C. Steefel, C. I. (2012). Measurement of accessible reactive
(2011). Constraining the reservoir model of an injected surface area in a sandstone, with application to CO2
CO2 plume with crosswell CASSM at the Frio-II brine mineralization. Chemical Geology, 318–319: 113–125.
pilot. International Journal of Greenhouse Gas Control, Lebedev, M., Toms-Stewart, J., Clennell, B., et al. (2009).
5: 1022–1030. DOI:10.1016/j.ijggc.2011.03.002. Direct laboratory observation of patchy saturation and
Dutta, N. C., and Seriff, A. J. (1979). On White’s model of its effects on ultrasonic velocities. Leading Edge, 28:
attenuation in rocks with partial gas saturation. 24–27.
Geophysics, 44: 1806–1812. Lebedev, M., Wilson, M. E. J., and Mikhaltsevitch, V. (2014).
Dvorkin, J., and Nur, A. (1996). Elasticity of high-porosity An experimental study of solid matrix weakening in
sandstones: Theory for two North Sea data sets. water-saturated Savonnieres limestone. Geophysical
Geophysics, 61: 1363–1370. DOI:10.1190/1 .1444059. Prospecting, 62: 1253–1265. DOI:10.1111/1365-
Gasperikova, E., and Hoversten, G. M. (2008). Gravity 2478.12168.
monitoring of CO2 movement during sequestration: Lemmon, E. W., McLinden, M. O., and Friend, D. G. (2005).
Model studies. Geophysics, 73(6): WA105–WA112. Thermophysical properties of fluid systems.
DOI:10.1190/1.2985823. In P. J. Linstrom and W. G. Mallard (eds.), Chemistry
Gassmann, F. (1951). On elasticity of porous media. web book. NIST Standard Reference Database Number
Reprinted in M. A. Pelissier, H. Hoeber, N. van de 69. National Institute of Standards and Technology.
Coevering, and I. F. Jones (eds.), Classics of elastic wave Lesmes, D. P., and Friedman, S. P. (2005). Relationships
theory. Geophysics Reprint Series No. 24, Society of between the electrical and hydrological properties of
Exploration Geophysicists, 2007. rocks and soils. In Y. Rubin and S. Hubbard (eds.),
Guéguen, Y., and Palciauskas, G. (1994). Introduction to the Hydrogeophysics. Dordrecht, The Netherlands:
physics of rocks. Princeton, NJ: Princeton University Springer.
Press. Mavko, G., Mukerji, T., and Dvorkin, J. (1998).
Hill, R. (1963). Elastic properties of reinforced solids: Some The rock physics handbook: Tools for seismic
theoretical principles. Journal of the Mechanics and analysis in porous media. Cambridge: Cambridge
Physics of Solids, 11: 357–372. University Press.
Hooke, R. (1678). Potentia Restitutiva, or Spring. Reprinted Müller, T. M., and Gurevich, B. (2004). One-dimensional
in M. A. Pelissier, H. Hoeber, N. van de Coevering, and random patchy saturation model for velocity and
I. F. Jones (eds.), Classics of elastic wave theory. attenuation in porous rocks. Geophysics, 69(5):
Geophysics Reprint Series No. 24, Society of Exploration 1166–1172. https://doi.org/10.1190/1.1801934
Geophysicists, 2007, 55–67. Müller, T. M., Gurevich, B., and Lebedev, M. (2010). Seismic
Hovorka, S. D., Doughty, C., Benson, S. M., et al. (2006). wave attenuation and dispersion resulting from
Measuring permanence of CO2 storage in saline wave-induced flow in porous rocks:
formations: The Frio experiment. Environmental A review. Geophysics, 75(5): 75A147–75A164.
Geoscience, 13(2): 105–121. DOI:10.1190/1.3463417.
Hovorka, S. D., Meckel, T., and Treviño, R. H. (2013). Nabighian, M., ed. (1991). Electromagnetic methods in
Monitoring a large-volume injection at Cranfield, applied geophysics, Vol. 2: Applications. In Society of
Mississippi–Project design and recommendations. Exploration Geophysicists Investigations in Geophysics.
International Journal of Greenhouse Gas Control, 18: DOI:10.1190/1.9781560802686.
345–360. Nakatsuka, Y., Xue, Z., Garcia, H., and Matsuoka, T. (2010).
IPCC. (2005). IPCC Special Report on Carbon Dioxide Experimental study on monitoring and quantification
Capture and Storage. Prepared by Working Group III of stored CO2 in saline formation using resistivity
of the Intergovernmental Panel on Climate Change measurements. International Journal of Greenhouse 81
Thomas M. Daley
Gas Control, 4: 209–216. http://dx.doi.org/10.1016/j of elastic solids. Reprinted in M. A. Pelissier, H. Hoeber,

.ijggc.2010.01.001 N. van de Coevering, and I. F. Jones (eds.), Classics of
Pride S. R. (2005). Relationships between seismic and elastic wave theory. Geophysics Reprint Series No. 24.
hydrological properties. In Y. Rubin and S. S. Hubbard Society of Exploration Geophysicists, 2007. 125–161.
(eds.), Hydrogeophysics. Water Science and Technology Toms, J., Müller, T. M., and Gurevich, B. (2007). Seismic
Library, Vol. 50. Dordrecht: Springer, 253–290. attenuation in porous rocks with random patchy
DOI:10.1007/1-4020-3102-5_9. saturation. Geophysical Prospecting, 55(5): 671–678.
Pride, S. R., Berryman, J. G., and Harris, J. M. (2004). DOI: 10.1111/j.1365-2478.2007.00644.x.
Seismic attenuation due to wave-induced flow. Journal Vanorio, T. (2015). Recent advances in time-lapse,
of Geophysical Research, 109: B01201. DOI : 10.1029/ laboratory rock physics for the characterization and
2003JB002639. monitoring of fluid-rock interactions. Geophysics, 80
Pride, S. R., Berryman, J. G., Commer, M., Nakagawa, S., (2): WA49–WA59. DOI:10.1190/geo2014-0202.1.
Newman, G. A., and Vasco, D. W. (2016). Changes in Vanorio, T., Mavko, G., Vialle, S., and Spratt, K. (2010).
geophysical properties caused by fluid injection into The rock physics basis for 4D seismic monitoring of
porous rocks: Analytical models. Geophysical CO2 fate: Are we there yet? Leading Edge, 29: 156–162.
Prospecting, 65(3). DOI:10.1111/1365–2478.12435. Vanorio, T., Nur, A., and Ebert, Y. (2011). Rock physics
Rubin, Y., and Hubbard, S., eds. (2005). Hydrogeophysics, analysis and time-lapse rock imaging of geochemical
Water Science and Technology Library, Vol. 50. effects due to the injection of CO2 into reservoir rocks.
Dordrecht, The Netherlands: Springer. Geophysics, 76(5): 23–33. DOI:10.1190/ geo2010-
Rutqvist, J. (2012). The geomechanics of CO2 storage in 0390.1.
deep sedimentary formations. Geotechnical and Vialle, S., and Vanorio, T. (2011). Laboratory measurements
Geological Engineering, 30(3): 525–551. DOI:10.1007/ of elastic properties of carbonate rocks during injection
s10706-011–9491-0. of reactive CO2-saturated water. Geophysical Research
Saito, H., Nobuoka, D., Azuma, H., Xue, Z., and Tanase, D. Letters, 38: L01302. DOI:10.1029/2010GL045606.
(2006). Time-lapse crosswell seismic tomography for Wang, Z. (2001). Fundamentals of seismic rock physics.
monitoring injected CO2 in an onshore aquifer, Geophysics, 66: 398–412.
Nagaoka, Japan. Exploration Geophysics, 37: 30–36. Wang, Z., Cates, M. E., and Langan, R. T. (1998). Seismic
Sen, P. N., Scala, C., and Cohen, M. H. (1981). A self-similar monitoring of a CO2 flood in carbonate reservoir:
model for sedimentary rocks with application to the A rock physics study. Geophysics, 63: 1604–1617.
dielectric constant of fused glass beads. Geophysics, 46: Waxman, M. H., and Smits, L. J. M. (1968). Electrical
781–795. conductivities in oil-bearing shaly sands. Society of
Sen, P. N., Goode, P. A., and Sibbit, A. (1988). Electrical Petroleum Engineers Journal, 8: 107–122.
conduction in clay bearing sandstones at low and high White, J. E. (1975). Computed seismic speeds and
salinities. Journal of Applied Physics, 63: 4832–4840. attenuation in rocks with partial gas saturation.
Sethian, J. A., and Popovici, A. M. (1999). 3-D traveltime Geophysics, 40: 224–232.
computation using the fast marching method. Worthington, P. F. (1985). The evolution of shaly-sand
Geophysics, 64(2): 516–523. concepts in reservoir evaluation. Log Analyst, 26: 23–
Smith, T. M., Sondergeld, C. H., and Rai, C. S. (2003). 40, SPWLA-1985-vXXVIn1a2.
Gassmann fluid substitutions: A tutorial. Geophysics, Xue, Z., Tanase, D., and Watanabe, J. (2006). Estimation of
68: 430–440. CO2 saturation from time-lapse CO2 well logging in an
Stokes, G. G. (1845). On the theories of the internal friction onshore aquifer, Nagaoka, Japan. Exploration
of fluids in motion, and of the equilibrium and motion Geophysics, 37: 19–29.
82
Chapter
Multicomponent Seismic Monitoring
5 Thomas L. Davis and Martin Landrø
Introduction and Background The velocity difference between the two waves is
a measure of fracture density in a fractured medium.
The large-scale deployment of carbon capture and
A majority of our reservoirs are fractured and frac-
sequestration will benefit greatly from new methods
tures cause the reservoir to be inherently anisotropic.
of seismic monitoring based on multicomponent seis-
Rock fabric within a reservoir can also cause
mic data and rock physics modeling. Simply put, our
a reservoir to be anisotropic. Stresses in the subsurface
reservoirs and carbon dioxide (CO2) storage sites are
can also induce anisotropy. Thus anisotropy of
often too complex to monitor without multicompo-
a reservoir is linked to the rock frame and stresses in
nent seismic data. The rock frame can and does change
the subsurface. Both can change over time due to
in many of our reservoirs with injection and produc-
injection, depletion, or stimulation of a reservoir.
tion. The interaction between fluid and the rock frame
Monitoring changes in S-wave azimuthal anisotropy
is critical to understanding dynamic changes in the
is a very robust means of reservoir monitoring espe-
reservoir. As a result, new approaches based on robust
cially in fractured reservoirs.
measures of velocity and attenuation anisotropy are
Shear waves come from shear wave sources, such
critical to include in dynamic reservoir characteriza-
as horizontal vibrators. They can also be generated by
tion. These approaches require a solid foundation
mode conversion from compressional waves imping-
based on rock physics and geomechanical modeling.
ing on reflection boundaries at nonnormal incidence.
Multicomponent seismic data involves 3-compo-
This is the case for marine multicomponent acquisi-
nent geophones that enable the recording of shear or
tion, as it is considered impractical and costly to place
S-waves as well as compressional or P-waves.
the source at the seabed. Both pure shear and con-
In marine multicomponent seismic acquisition we
verted shear waves exhibit shear wave splitting in
measure the omnidirectional pressure component in
anisotropic media. Time and amplitude measures of
addition to the three geophone components. This is
S-wave splitting can be used as a measure of S-wave
referred to as 4-component seismic or simply 4C.
azimuthal anisotropy. Time measurements are used if
Recording both waves enables us to quantify elastic
the medium is above seismic resolution (thickness
properties more fully in the subsurface. These elastic
greater than 1/2 wavelength) and amplitude measure-
properties are anisotropic in nature by virtue of the
ments are used for thin beds (below 1/2 wavelength).
media through which seismic waves propagate. Shear
Warping algorithms are very useful for detecting and
waves are especially sensitive to azimuthally anisotro-
visualizing S-wave splitting in thin beds.
pic media as they polarize or undergo birefringence
What can be done with multicomponent seismic
which causes shear wave splitting. This phenomenon
data? One can best link this question back to princi-
is especially useful for seismic monitoring involving
ples of rock physics. Elastic moduli such as Poisson’s
CO2 enhanced oil recovery (EOR) and geosequestra-
ratio can be measured (Figure 5.2). By propagating P-
tion (Figure 5.1).
and S-waves through a medium we can measure Vp
and Vs and determine Poisson’s ratio. The equation
Properties and Characteristics of Shear written in Figure 5.2 is for an isotropic medium. For
Waves an anisotropic medium the symmetry of the medium
When a shear wave enters an anisotropic medium it must be defined and then the shear wave splitting
splits into two waves, fast (S1) and slow (S2). phenomenon can be used to define the symmetry 83
Thomas L. Davis and Martin Landrø
Multicomponent seismic
• PP-Compressional waves and SS shear waves
• PS (PS1, PS2)-Mode-converted waves
⏐γ(1) – γ(2)⏐ VS1 – VS2 S

γ(S) = ≈ Anisotropy
1 + 2γ(2) VS2
3C receiver Seismically
anisotropic
P-wave rock
source
S2
S Slow
P S1
Fast
PP PS
reflection conversion Davis and Martin,1987
point point
Figure 5.1 Seismic recording with 3-component (3C) receivers or 4C in the case of adding a hydrophone allow for the recording of
multicomponent seismic data including P and S body waves.
stress change in the subsurface. The dynamic changes

are important to monitor especially in the case of CO2
Vp/Vs What Does it Mean?
flood monitoring.
Vp2 = Compressibility + 4/3 Rigidity/Density

Multicomponent Seismic for
Vs2 = Rigidity/Density Characterization of Fractured
Vp/Vs = Compressibility/Rigidity
Reservoirs
Employing multicomponent seismology improves
Vp2 – 2Vs2 reservoir characterization because the technology
v= provides the only volume-based method to be able
2(Vp2 – Vs2)
to directly observe the fractures and their changes
over time. Multicomponent seismology enables the
recording of shear waves that are necessary for the
Figure 5.2 Vp/Vs links to Poisson’s ratio, which is a fundamental characterization of naturally fractured reservoirs.
geomechanical property. Greater emphasis needs to be placed on
determining anisotropic elastic parameters, for example, Poisson’s Shear waves are more sensitive than P-waves to
ratio in the subsurface. detecting velocity and attenuation changes associated
with fractured reservoirs. Multicomponent seismol-
ogy is suited to the monitoring of fractured reservoirs
planes and obtain Vp and Vs in those symmetry owing to the influence of pressure changes that cause
planes. Thus two Poisson’s ratios are determined changes in velocity and attenuation. Both compressi-
and these can be useful for characterizing the med- bility and rigidity influence the velocity of compres-
ium, a fractured reservoir, for example. They are very sional waves, whereas shear wave velocity is
useful for linkage to geomechanical properties of the influenced only by rigidity. Natural fractures affect
84 medium. Time-lapse changes in the medium relate to the rigidity of the rock causing velocity anisotropy
owing to the oriented nature of fracture sets. waves are stress sensitive, we can observe pressure
The advantage of shear waves over P-waves is the compartments.
ability to detect azimuthal anisotropy along
a singular raypath. With P-waves multiple ray paths Developments within Acquisition and
are necessary to detect velocity anisotropy.
Heterogeneous, anisotropic media above the target Processing of Multicomponent Data
level can interfere with P-wave detection and charac- New acquisition and processing technologies are
terization of anisotropy. evolving. The increase in channel count and dis-
Multicomponent seismic data provide for better tributed recording systems allows for increased
quantitative interpretation of the subsurface. fold and redundancy necessary to carry out suc-
The principal use of these data is in fracture detec- cessful multicomponent seismic acquisition. New
tion and better imaging and characterization of the vibrator technology allows for greater bandwidth
subsurface. Multicomponent seismology provides recording especially on the low end of the fre-
for better porosity determination in carbonates, quency spectrum. Increased bandwidth on the low
better imaging especially under gas clouds, end is especially important for shear wave record-
improved fault imaging in wrench fault terrains, ing and vector fidelity. New interpretive processing
and finally improved imaging of low-impedance involving better velocity modeling, resolution
contrast reservoirs. It requires little or no incre- enhancement through full waveform inversion
mental cost to record 3C versus 1C seismic data. (FWI), elastic inversion, and better imaging
The cost of recording 9C data (3-component through reverse time migration (RTM) is especially
sources and 3-component receivers) on land is encouraging in creating value through multicom-
almost always twice the cost of a 1C survey due ponent seismology. The detection of geopressure
to the two orthogonal source polarization directions and geomechanical changes in the subsurface will
that one needs to impart at each source point. be especially important in the future. Quantitative
Shear waves are generated as horizontally polar- interpretation of rock and fluid properties is neces-
ized waves and because they travel in near normal sary and multicomponent seismology can reduce
incidence with horizontal polarizations they are risk and uncertainty.
especially sensitive to open natural fractures. Completion and production processes induce
The advantage of recording shear waves in three dynamic changes in reservoir properties including
dimensions (3D) and four dimensions (4D) is to pressure, saturation, and permeability. The most
see fractures and the connectivity of the fracture accurate geophysical tool for monitoring these
networks. Displaying connectivity is important for changes is time-lapse multicomponent seismology,
well location in reservoir development. One wants specifically in delineating the spatial distribution of
to minimize the number of wells necessary to the changes. Studies by the Reservoir
develop the resource and not drill into compart- Characterization Project (RCP) document the power
ments that are already connected. Highly deviated of multicomponent seismology to observe and quan-
wells enable us to drill across the main fracture tify these dynamic changes. RCP’s studies illustrate
trends thereby landing wells with the potential for that more accurate quantitative measurements of
greater recovery. In some instances, however, frac- reservoir properties can occur through the time-
tures can also be conduits for water delivery to lapse multicomponent seismic monitoring and that
these wells so every reservoir is different and one these measurements when introduced into dynamic
needs to monitor reservoir behavior over time. reservoir characterization provide for more accurate
In wrench fault terrains it is common for sealing prediction of reservoir performance.
faults to occur as they can set up reservoir com- Both pressure effects and fluid composition
partmentalization. An advantage of shear waves is changes associated with a CO2 injection program
their use in detecting pressure cells associated with can be detected using multicomponent seismic data.
these compartments. Pressure cells are zones of This is a key result with major implications for using
weakness or low rigidity in the subsurface. Fluid dynamic reservoir characterization to manage and
pressures hold fractures apart and because shear improve the economics of an EOR project.
85
reservoir was affected by larger quantities of CO2

injected in multiple wells. The primary measurement
tool used was measurements of S-wave anisotropy.
S-wave anisotropy is important. Quantifying
S-wave anisotropy relates to the relative amount of
open fractures, microfractures and low aspect ratio
pore structure within the rock volume being interro-
gated by the S-waves. S-wave anisotropy can be
related to permeability, both magnitude and preferred
direction. Providing a quantitative measure of perme-
ability is critical for reservoir characterization.
Typically, permeability is not a quantity provided to
the reservoir engineering community by geophysicists
working with seismic data. S-wave anisotropy mea-
surements within the reservoir can be related to per-
meability. Coupled with a porosity estimate, we may
Figure 5.3 Map of Central Vacuum Unit including area of be able to determine the reservoir storage volume and
RCP study.
its preferred permeability pathways for optimum
recovery of hydrocarbons. A static estimate is deter-
mined from a single multicomponent 3D survey inte-
Figure 5.3 shows a map of the Central Vacuum grated with geological and petroleum engineering
Unit Field in Lea County, New Mexico. The study area data. With repeated surveys, a dynamic picture of
is an approximate circle with a radius of 3300 ft. the reservoir is possible (dynamic reservoir character-
(approx. 1000 m). Four multicomponent 3D surveys ization). Figure 5.4 shows on the left panel the S-wave
were acquired. Both reservoir pressure and fluid com- anisotropy difference between the first and second
position were altered between the first and second surveys, before and after the injection of 50 million
surveys over a small reservoir volume near a single standard cubic feet in a single well, and on the right
CO2 injection well. The third and fourth surveys were panel S-wave anisotropy difference between the third
conducted while reservoir pressure was held approxi- and fourth surveys before and after a billion standard
mately constant while a greater volume of the cubic feet of injection into six wells.
Figure 5.4 Anisotropy difference

associated with CO2 flood at Vacuum
Field, New Mexico. Phase I is a single-
Anisotropy Difference (Pre-CO2) - (Post-CO2) well pilot. Phase II is a six-well pilot.
CVU93 CVU194 CVU94
CVU97
CVU-97
CVU200
CVU99 CVU100
CVU-200
Fig.1a: Phase-I Fig.1b: Phase-II

86 –12 Percent (%) +12
2.25 Conventional 3D data S N Figure 5.5 Comparison of a PP-section

(P-wave down and P-wave up) and a
2.50 PS-section (P-wave down and converted to
S-wave at each interface; or the
X-component section). Notice the improved
2.75
Time (sec.)
data quality on the X-component data

between the two vertical solid lines (the area
3.00 affected by presence of gas in the
overburden). (From Granli et al., 1999.)
3.25
3.50
3.75 1/9–1 1/9–3R 3km
X-component
4.50
5.00
Time (sec.)
5.50
6.00
6.50
7.00 1/9–1 1/9–3R
Marine Multicomponent Seismic Data significantly higher than conventional marine strea-
mer acquisitions. As marine 4C acquisition means
Inspired by results obtained on land shear-wave data
that there is a regular receiver grid at the seabed and
acquisition, Statoil embarked on developing 4-com-
a regular shooting grid at the water surface, it is
ponent marine seismic data acquisition at the begin-
obvious that the number of measurements will
ning of the 1990s. After some field tests offshore
increase significantly compared to towed streamer
Norway, the first successful example was imaging
acquisition. Furthermore, it means that a continuous
below the gas chimney at the Tommeliten Field
azimuthal coverage is achieved, which is important
(Granli et al., 1999). Figure 5.5 shows that the
for anisotropy analysis. Rognø et al. (1999) demon-
X-component data images the area that is distorted
strated that this increase in azimuthal coverage is
by presence of gas better. If we assume that the con-
crucial for imaging of complex reservoirs, as shown
verted S-waves are close to vertical when they hit the
in Figure 5.6.
seabed receiver, they will dominate the response at the
X-component. This is why we for simplicity assume
that the X-component (or radial component) is prac- Multicomponent Marine PRM Systems
tically equal to PS-converted data. Presently, there are several permanent marine multi-
In the early days of 4C seabed acquisition nodes component systems installed. Offshore Norway, PRM
were planted into the seabed using remote operating systems have been installed at the Valhall, Ekofisk,
vehicles (ROVs). The node method has been further Snorre, and Grane Fields. Typical 4D sampling time
developed, and today several node-based systems are interval for these fields is six months. Offshore Brazil,
offered. Nodes are especially attractive for deep water the Jubarte Field has been instrumented by a PRM
acquisition and for fields with a lot of other seabed system operating at a water depth of 800 m.
installation. Later, four-component receiver cables It has been decided that the Johan Sverdrup Field
that were either placed directly at the seabed or (estimated reserves of 3 billion barrels of oil) also will
trenched into the seabed (for permanent monitoring be instrumented with a seabed multicomponent recei-
[PRM] systems were developed). For most marine ver array. There are multiple examples of how these
87
4C acquisitions the number of seismic traces is arrays are used in the daily reservoir management of
Figure 5.6 (Left) P/Z seismic data

(exploiting both the pressure component
and the vertical geophone component and
the multiazimuthal coverage). (Right)
Conventional 3D streamer data using only
the pressure component. Notice the
improved definition of the fault blocks at
the basal Cretaceous unconformity level
(like a staircase effect) on the P/Z data
compared to the conventional data to the
right. (From Rognø et al., 1999.)
these fields (Gestel et al., 2008; Bertrand et al., 2014; Z-component geophone data was located at approx.
Thompson et al., 2016; Elde et al., 2016). 2 km depth in a well, while an air gun source was used
Landrø et al. (2017) suggested that such large perma- to cover a nearly circular shooting pattern at the surface,
nent arrays can be exploited to detect leakage of gas or as shown in Figure 5.7. Ten thousand shots were fired,
CO2 from the subsurface into the water layer. and normalized root mean square (NRMS) errors were
Conventional air gun arrays emit significant amounts computed for 70,000 shot pairs. All shot pairs were
of high-frequency signals (Landrø et al., 2011, 2016) aligned prior to differencing and NRMS computation.
and these signals may be used to image relatively thin Figure 5.8 shows the average NRMS error as a function of
chimneys of gas leakage at the seabed. This possibility is source separation distance for Z- and X-component VSP
also of interest for monitoring CO2-storage sites, in order data. We clearly observe that the NRMS error increase is
to detect unwanted leakage of CO2 into the water layer as somewhat larger for the X-component compared to the
early as possible. As the public acceptance for offshore Z-component. Misaghi et al. (2007) showed that the
CO2-storage is higher than for land-based storage sys- nonrepeatability is caused by lenses and inhomogeneous
tems, it is crucial to develop cost-effective and reliable overburden geology. If we assume that these shallow geo-
monitoring systems for early leakage detection, both bodies have larger contrast in S-wave velocity than the
close to the storage reservoir and within the overburden, corresponding P-wave velocity, it is reasonable to expect
including the water layer. less repeatable data, especially for larger source separa-
tion distances, like we observe in Figure 5.8. If this
hypothesis is correct, we must expect that the difference
Repeatability Issues in repeatability between S-wave and P-wave data will
For all 4D seismic studies, it is crucial to repeat the vary with geological complexity and lithology contrasts
experiments performed at two different calendar times in the overburden. Large S-wave velocity contrasts in the
as accurately as possible. This is challenging and often overburden reduce the repeatability of data.
involves several steps during acquisition and processing Figure 5.9 from Oseberg Field, offshore Norway,
to achieve a high degree of repeatability. In a 4C seismic shows a comparison between standard 3D streamer
survey, the X-component is crucial because shear waves data and ocean bottom cable data (4C). We notice that
are more apparent on this component. It is therefore of the low frequency content is much higher on the OBC
interest to study the repeatability of both the Z- and the data. We interpret this to be mainly attributed to the fact
X-component of seismic data. Landrø (1999) used a 3D that when the receivers are lowered to the seabed, the
VSP dataset acquired over the Oseberg Field for this influence of ocean noise related to waves and swell noise
88 purpose. A five-level VSP-tool recording X-, Y-, and is significantly less. This is also one of the motivations
Figure 5.7 Shot positions of the 10,000 shots acquired in the 3D VSP experiment. The hole in the middle is the well position, and the
noncircular hole to the northwest of the well is a platform. Typical distance between two adjacent shots is 25 m (from Landrø, Geophysics,
1999). The size of the square is 4 by 4 km. Notice that several of the shot positions are very close, enabling repeatability quantification also for
source separation distances much less than 25 m.
Figure 5.8 Average normalized root mean square (NRMS) error as a function of source positioning errors for Z-component (left) and
X-component VSP-data acquired at 2 km depth at Oseberg (North Sea). (Figure from Landrø [1999], Geophysics, 64: 1673–1679.) Notice that the
repeatability is very similar for source separation distances less than 10 m, and thereafter the repeatability is worse for the X-component data. 89
Streamer Ocean Bottom Cable
Statoil
30/9–6
1
30/9–6
Figure 5.9 Comparison between conventional streamer data and 4-component ocean bottom cable data from the Oseberg Field, North Sea.
Notice that the Oligocene sands (1) is better imaged on the 4C seismic data, as well as the strong chalk layer (2) situated above the reservoir (3).
Furthermore, the horst-structure (4) is clearer, as well as some potential deeper sand layers (5). Both data sets are P-wave only.
owc
Top Shetland Lower Lista shale
Figure 5.10 Geological cross section of the Grane reservoir (yellow) and shale layers above and below. This example was presented by
J. P. Fjellanger et al. at the Society of Exploration Geophysicists meeting in New Orleans in 2006, and is presented here with permission from the
authors.
90
Figure 5.11 PP (left) and PS (right) image gathers from Grane. Notice that the top and base Heimdal events are continuous for a wider range
of offsets on the PS data.
Figure 5.12 PP and PS stack images without interpretation overlaid (top) and the same including interpretation. Notice that the reservoir
body (Heimdal sand) is better defined on the PS image.
for towing conventional streamers deeper today com- a geological depth section, including the reservoir
pared to what was the standard 10–15 years ago. and the surrounding shale layers. Figure 5.11 shows
In Figure 5.9 we clearly see that this richness in lower seismic image gathers of the PP and PS data.
frequencies results in a far better image: The horst struc- The major cause for the improved quality of the
ture which is penetrated by the well is clearly imaged as PS-seismic data in this case is shown in Figure 5.11,
well as the Oligocene sands in the overburden. where we observe that the PP-reflectivity dims versus
The Grane Field is one of the Norwegian fields that offset for both top and base reservoir. For the PS-
are monitored by a PRM field. The reservoir sand gather we notice that this dimming is not observed,
(Heimdal) is complex and it is a challenge to map yielding a more stable and noise-free stacked signal.
the reservoir by conventional P-wave seismic. Figure 5.12 shows a comparison of the stacked PP
Fjellanger et al. (2006) presented an interesting case and PS data. The two sections at the top are displayed
study using converted wave data acquired over the without the interpretation overlaid, and the two sections
Grane Field, and we will present some of the key at the bottom are displayed with the interpretation 91
results from this study here. Figure 5.10 shows overlaid.
Figure 5.13 Demonstrating the improved definition of the reservoir by using PS seismic data. The solid blue line is the interpretation of top
reservoir based on PP data only, while the solid yellow line is the same horizon interpreted based on the PS data. On the lower figure, the
Gamma log reading is overlaid with the well path trajectory, clearly showing that the yellow line interpretation is in better accordance with the
Gamma log.
We clearly see that the PS section is clearer and gives Gestel, J-P, Kommedal, J. H., Barkved, O. I., Mundal, I.,
a better interpretation of the reservoir. This is also Bakke, R., and Best, K. D. (2008). Continuous seismic
confirmed when the Gamma log is added to the well surveillance of Valhall Field. Leading Edge, 27:
1616–1621.
penetrating the reservoir, as shown in Figure 5.13.
We notice that the interpretation of top reservoir Granli, J. R., Arntsen, B., Sollid, A., and Hilde, E. (1999).
based on the PP data predicts reservoir before the Imaging through gas-filled sediments using marine
shear-wave data. Geophysics, 64: 668–677.
wells actually enters the reservoir. We also notice that
the Gamma log has a strong response in the region Landrø, M. (1999). Repeatability issues of 3-D VSP data.
Geophysics, 64: 1673–1679.
between the blue and the yellow lines in the lower
section in Figure 5.13. We observe a similar effect closer Landrø, M., Amundsen L., and Barker, D. (2011). High-
to the toe of the horizontal section of the well (to the frequency signals from air-gun arrays. Geophysics, 76:
Q19–Q27.
right), where the blue line again is above the yellow line.
Landrø, M., Ni, Y., and Amundsen, L. (2016). Reducing
high-frequency ghost-cavitation from marine air-gun
arrays. Geophysics, 81: P47–P60.
References
Landrø, M., Hansteen F., and Amundsen, L. (2017).
Bertrand, A., Folstad, P. G., Lyngnes, B., et al. (2014). Detecting gas leakage using high-frequency signals
Ekofisk life-of-field seismic: Operations and 4D generated by air-gun arrays. Geophysics, 82: A7–A15.
processing. Leading Edge, 33: 142–148.
Misaghi, A., Landrø, M., and Petersen, S. (2007).
Davis, T. L., and Martin, M. A. (1987). Shear wave Overburden complexity and repeatability of seismic
birefringence: A new tool for evaluating fractured data: Impacts of positioning errors at the Oseberg Field,
reservoirs. Leading Edge, 6: 22–28. North Sea. Geophysical Prospecting, 55: 365–379.
Elde, R., Roy, S. S., Andersen, C. F., and Andersen, T. Rognø, H., Kristensen Å., and Amundsen, L. (1999).
(2016). Grane permanent reservoir monitoring – The Statfjord 3-D, 4_C OBC survey. Leading Edge, 18:
meeting expectations! In 78th EAGE Conference, 1301–1305.
Expanded Abstract, Tu LHR2 02.
Thompson, M., Andersen, M., Skogland, S. M., Courtial, C.,
Fjellanger, J. P., Bøen, F., and Rønning, K. J. (2006). and Biran, V. B. (2016). Time-lapse observations from
Successful use of converted wave data for interpretation PRM at Snorre. In 78th EAGE Conference, Expanded
and well optimization on Grane. Expanded Abstracts,
92 Society of Exploration Geophysicists, 1138–1148.
Abstract Tu LHR2 01.
Chapter
Monitoring the Deformation Associated
6 with the Geological Storage of CO2

Donald W. Vasco, Alessandro Ferretti, Alessio Rucci,
Sergey V. Samsonov, and Don White
Introduction oldest branches of geophysics. In this chapter we will

discuss geodetic observations that are useful for mon-
The geological storage of greenhouse gases from
itoring the deformation associated with the geological
a power plant source requires the injection of large
storage of CO2. After an overview of methods we will
quantities of supercritical carbon dioxide (CO2), typi-
quickly focus on inteferometric synthetic aperture
cally into a small number of wells. A key factor in the
radar (InSAR), as it is the most practical approach
Earth’s response to such an injection is how fast the
for long-term, frequent monitoring, with dense spa-
injected fluid can redistribute within the volume sur-
tial coverage and millimeter-level accuracy.
rounding the injection wells. For a medium with
a well-connected pore space the redistribution is con-
trolled by the hydraulic diffusivity. In particular, the Overview of Geodetic Techniques
velocity of the fluid v, under the influence of Geodetic techniques have a long history, with meth-
a pressure gradient rP, depends on the porosity and ods such as chaining, where an object of known length
permeability of the medium and the viscosity of the is used to estimate the distance between points, dating
fluid. The variation in fluid pressure with time will back to ancient times (Smith, 1997; Torge and Muller,
influence the effective stress buildup and conse- 2012). The classic techniques of triangulation and
quently the deformation around the injection well. trilateration were used to establish geodetic networks
Heterogeneity can have a major influence on the in the seventeenth century. In triangulation, two
fluid redistribution and associated pressure varia- reference points are used to establish a base distance
tions. For example, high-permeability pathways, and the angles of sight-lines to a third point are used
such as conductive faults, can concentrate pressure to estimate additional distances. In trilateration, dis-
changes into a confined region leading to large pres- tances between three points are used to locate bench-
sure changes extending a considerable distance from marks with respect to a base point. The distances are
the injection site. Barriers to flow can lead to the usually measured electronically, with an electronic/
compartmentalization of the fluid buildup and sharp optical distance meter (EDM). Estimates of deforma-
spatial variations in fluid pressure. Heterogeneities in tion follow from repeated surveys over a network of
flow properties are typically accompanied by varia- stations. A helpful overview of geodetic methods,
tions in mechanical properties, leading to coupled examined in relation to monitoring volcanic deforma-
deformation and flow. Such heterogeneities and tion, is provided in Dzurisin (2007).
fluid pressure variations can lead to issues associated Here we shall focus on approaches that are useful
with long-term storage, such as seal damage or poten- for monitoring the fate of CO2 stored within the
tial leakage. Earth. For example, to serve as a useful monitoring
Anomalous pressure buildup around an injection technology, a geodetic technique should cover
well can also result in deformation that is observable a sufficient area, the sampling time between observa-
in the overburden and even at the surface. Geodetic tions should be a few months at most, and the cost of
methods, techniques for accurately measuring dis- data from a single survey should be of the order of
tances and angles between objects on the Earth, are a few hundred to a few thousand dollars to allow for
well suited to detect such deformation. These meth- cost-effective, long-term monitoring. Field methods
in geodesy can be roughly categorized into individual
93
ods have a long history, and geodesy is one of the
D. W. Vasco, A. Ferretti, A. Rucci, S. V. Samsonov, and D. White
point measurements and scanning systems. Point domain, with wavelengths ranging from 3 to 24 cm
measurements are obtained from stand-alone instru- (for the sensors currently in orbit), 100,000 times
ments, such as tilt meters (Wright, 1998) and Global longer than those of the visible spectrum.
Navigation Satellite System (GNSS) receivers (Misra Electromagnetic waves in this frequency band are
and Enge, 2001, Petrovski and Tsujii, 2012). One capable of penetrating clouds, fog, and rain. Being
characteristic of these devices is fine temporal an active system, a radar sensor can operate day and
sampling, often producing observations every few night, throughout the year, independent of solar
minutes. Scanning systems actively reflect electro- illumination.
magnetic energy from deforming surfaces, differen- InSAR techniques for geodetic applications rely
cing the returning signal at various times in order on the comparison of the phase values of two or
to estimate the displacement of points on the more SAR images acquired at different times from
surface. Examples include InSAR (Ferretti, 2014) a similar acquisition geometry to detect and monitor
and light detection and ranging, known as LiDAR surface deformation. InSAR data can be gathered
(McManamon, 2015). We shall have much more to using both airborne and space-borne sensors. Here
say about the former technique in the section that we will focus on the latter case and will provide a short
follows. Scanning systems typically produce a higher introduction to this technology. For a more detailed
spatial density of sample points, while sacrificing treatment, the interested reader can refer to the tutor-
some degree of temporal resolution. Still, large-scale ial papers available in the literature (e.g., Bamler and
scanning systems can sample the deformation field Hartl, 1998; Massonnet and Feigl, 1998; Rosen et al.,
every few weeks or months, a sufficient rate for mon- 2000; Ferretti, 2014).
itoring the ground motion associated with the geolo- The basic idea behind satellite InSAR measure-
gical storage of CO2. In addition to techniques that are ments is rather simple. Satellite SAR sensors orbit the
suitable for dry land, there are methods such as repeat Earth along near polar paths, at an altitude ranging
bathymetry and sensitive pressure transducers that from 500 to 800 km. However, the combination of the
function in aquatic environments. Owing to the lim- rotation of the Earth and the motion of satellite makes
ited number of large-scale carbon storage sites, few it possible for SAR sensors to acquire data over the
geodetic monitoring techniques have been applied to same area, from the same acquisition geometry, but at
date. InSAR has been the most commonly used different times. By definition, radar measures the dis-
approach and has been applied at a handful of sites, tance between the antenna source and a target on the
such as the three described in the text that follows. ground. InSAR is just a way to monitor range varia-
A network of tilt meters was also tested at the In Salah tions (changes in distance) using multiple SAR images
storage site, but it proved too expensive and did not from an area gathered at different times. To detect
provide a significantly better monitoring capability very small surface displacements, of the order of
than InSAR (Mathieson et al., 2011). a centimeter, compared to the meter-scale resolution
of the radar image, we must obtain accurate estimates
of the phase shifts between two successive SAR acquisi-
Overview of InSAR tions from the same nominal orbit (Figure 6.1).
To gain some insight into the complex-valued
Introduction to InSAR nature of SAR images, we should think of a system
A synthetic aperture radar (SAR) is a coherent, active illuminating each resolvable cell in the area of interest
system capable of recording the electromagnetic with a short, monochromatic, sinusoidal, radar pulse,
energy backscattered to the radar antenna by an illu- with a frequency equal to the central frequency of the
minated scene and generating a two-dimensional SAR sensor. For the sake of illustration, we suppose
complex-valued image, containing both amplitude that there is just one dominant scatterer in each cell.
and phase information. While the amplitude data When the backscattered signal is sampled by the
depend on the amount of backscattered energy, the acquisition system, the radar return can be fully char-
phase values of a SAR image are related to the distance acterized by two variables: the amplitude and the
from the radar antenna to the objects on ground and phase of the recorded pulse. While the amplitude
94 they can be used as the basis for geodetic measure- will depend on the so-called radar cross section of
ments. SAR sensors operate in the microwave the target (Ferretti, 2014), its phase value is related
Monitoring Deformation from the Storage of CO2
time T0 time T0 + Δt INTERFEROMETRY

λ
Δr
R1
LO
S
(Lin
R1 R2 ≠ R1 R2
eo
fS
λ - wavelength:
igh
t)
Δr C-band= 5.66 [cm]

X-band= 3.10 [cm]
L-band= 24.00 [cm]
1st Acquisition
2nd Acquisition
Figure 6.1 A schematic showing the relationship between ground displacement and signal phase shift. The numeric value of the wavelength
λ depends on the radar sensor and ranges from 3 cm (X-band) to 24 cm (L-band) for the sensors currently in orbit.
to the sensor-to-target distance, as it is proportional The phase values of a single SAR image are of no
to the time delay between pulse transmission and practical use, as it is impossible to separate the differ-
reception. Of course, the complete process for SAR ent contributions in Eq. (6.1) without additional
imaging is much more complicated because it information. Moreover, owing to the sinusoidal nat-
involves concepts such as antenna synthesis, sensor ure of the representation, the phase values wrap
velocity, range compression, signal bandwidth, and around and are insensitive to factors of 2π.
sampling frequency (Ferretti, 2014), but our simplis- Therefore, the SAR phase matrix looks very much
tic model can be useful for non-radar specialists and like a set of random numbers ranging from –π to
will suffice for this discussion. +π. The basic idea of SAR interferometry is to mea-
A detailed analysis of the phase φ associated with sure the phase change, or interference between two
a pixel (picture element) P from a radar image, can radar images, generating an interferogram I:
help to identify when InSAR will be successful as well 4π
as the current limitations of this technology. For I ¼ ΔφðPÞ ¼ Δϑ þ Δr þ Δα þ Δn ð6:2Þ
λ
example, the phase φ can be modeled as a mixture of
for the time interval between the two satellite passes.
four distinct contributions (Rosen et al., 2000; Ferretti
In fact, if we consider an idealized scenario where
et al., 2007; Ferretti, 2014):
both target characteristics, such as the nature and
4π location of all elementary scatterers within each cell,
φðPÞ ¼ ϑ þ rþαþn ð6:1Þ
λ and atmospheric conditions are the same in the two
acquisitions, then – if noise can be considered negli-
where ϑ is a term related to the nature and the location
gible – Eq. (6.2) reduces to
of all elementary scatterers within the resolution cell
associated with pixel P. In our simplistic model we 4π
I ¼ ΔφðPÞ ¼ Δr ð6:3Þ
have considered a single, dominant scatterer per cell, λ
but typically there are several such scatterers. that is, interferometric phase values are proportional to
The term 4πλ r is an important factor in geodetic appli- range changes. In other words, if a point on ground
cations, because it is associated with the sensor-to- moves during the time interval between the two radar
target distance r, the range. The term α, referred to as images gathered with the same acquisition geometry,
the atmospheric phase screen (APS), is related to the the distance between the sensor and the target may
delay introduced by the Earth’s atmosphere, where change (unless the motion is orthogonal to the satel-
the propagation speed might change. Amplitude lite line of sight), creating a phase shift proportional to
data are nearly immune to atmospheric disturbances, the displacement. Rather than computing the differ-
but phase values can be strongly affected by the atmo- ence of the phase values, radar specialists would say
sphere. The last term, n, is a contribution related to that an interferogram is computed by multiplying
system noise, such as thermal vibration, quantization the complex values of the first SAR image, called the 95
errors, and so on. master image, by the complex conjugate of the second
Interferogram
1st acquisition (Master)
AM ZM ZS ∗
nd
2 acquisition (Slave)
Bn = 3 m
Bt = 70 d
AS
N
10 km
Comparison
time –π 0 +π
Phase variation (φ)
Event
λ/2 = 28 mm
–14 0 +14
Range variation [mm]
Figure 6.2 Example of SAR interferogram. The two SAR images were acquired by the Envisat ASAR sensor operating at C-band (λ/2 = 28 mm).
The area of interest in southeastern Iran, where on December 26, 2003 an Mw 6.6 earthquake devastated the town of Bam. The images were
acquired one before and one after the event. The computation of phase variations highlights the coseismic deformation. Interferometric
fringes resemble contour lines of the displacement field. (Adapted from Ferretti, 2014.)
acquisition, called the slave image (Figure 6.2), but exhibiting completely random sets of values, with the
this is terminology not necessary for the purpose of coherent fringe pattern of their interferogram.
our introduction. To retrieve the range change for each pixel of the
Equation (6.3) can explain the extremely high image, the phase values need to be unwrapped, start-
sensitivity of InSAR measurements: the unit of length ing from a reference pixel, assumed to be stationary.
is the centimeter-scale wavelength, rather than the In Figure 6.2, each fringe corresponds to 28 mm of
meter-scale range resolution of the radar image. range displacement; it is not difficult to estimate about
A displacement of the radar target along the satellite 30 cm of total displacement in the lower right part of
line of sight of λ/2 creates a phase shift of 2π radians the “butterfly” by simply counting the number of
due to the two-way travel path of the radar beam. cycles with respect to the reference point P0. Some
Thus, a range variation of just 1 mm creates an easily areas are affected by phase noise, but in general the
detectable phase shift of more than 20 degrees quality of the interferogram is rather high.
between two SAR images acquired at X-band It is important to point out that InSAR data are
(λ = 3.10 cm). Figure 6.2 is an example of a SAR differential measurements, as displacement values are
interferogram (C-band data from Envisat, estimated with respect to a reference point that is
λ = 5.66 cm). The color fringes created by the inter- relatively stable, or a benchmark whose displacement
ferometric processing are due to the coseismic displa- is known. This is the reason why uncertainty in the
cement field originating from an earthquake in satellite position can be mitigated: systematic errors
96 Iran (Funning et al., 2005). It is informative to com- affecting the whole area can be removed or strongly
pare the phase matrix of the master and slave images, reduced thanks to the Double-Difference approach,
T1 T2 ...Tn Figure 6.3 A schematic showing the basic

idea for measuring ground displacement
using successive radar signals and providing
a time series of displacement of ground
points identified within the illuminated scene.
NO DATA
The analysis of range variations is carried out
R1 R2 Rn on a subset of image pixels only slightly
affected by decorrelation phenomena.
ΔR 12
ΔR n
similar to the method used in differential GPS mea- Multi-interferogram Techniques

surements (Ferretti, 2014). Note that InSAR measures
As mentioned previously, conventional InSAR, invol-
only a single component of the three-dimensional dis-
ving the computation of phase variations on a pixel-
placement field: the component along the satellite line
by-pixel basis between two radar images, can be used
of sight. However, a combination of two or more
successfully only if the radar signature of the reflective
different orbital geometries, such as ascending and
target does not change significantly with time, the
descending paths, can be used to obtain two
signal-to-noise ratio (SNR) is high enough, and the
components of displacement, as discussed in a later
atmospheric phase components are negligible in
section.
the two SAR images. Multi-interferogram techniques
From Eq. (6.2) one can see that the limitations of
aim to overcome the limitations of conventional
InSAR analyses based on just two SAR images are due
InSAR analyses, based on just two SAR acquisitions,
to (apart from noise) atmospheric effects and possible
by estimating a time series of displacement data for
changes of the nature and the locations of elementary
those pixels characterized by high SNR values.
scatterers within each resolvable cell or pixel, creating
The PSInSAR technique (Ferretti et al., 2000, 2001),
what are known as “decorrelation” effects. Indeed,
developed in the late 1990s, was the one of the first
over water, snow, vegetation, and agricultural areas,
such InSAR algorithms. The basic idea is to compare
it is difficult, if not impossible, to retrieve good fringe
many SAR images and to focus the analysis on the
patterns because the radar signature of the target, the
high-quality radar targets (Figure 6.3), usually
first term in Eq. (6.1), changes with time and range
referred to as permanent scatterers (PSs). That is, we
variations cannot be estimated accurately. While
focus on those pixels that are only slightly affected by
phase decorrelation, due to changes of radar signa-
phase decorrelation, such as buildings and bare out-
ture, produces spatially uncorrelated phase changes,
crops. Such targets, identified by signal processing
atmospheric disturbances are uncorrelated in time
algorithms, exhibit very stable radar signatures and
but well correlated in space. This feature is exploited
a high SNR, and allow for the implementation of
by more sophisticated multi-interferogram techni-
sophisticated filtering procedures for estimating and
ques. In general, conventional InSAR results based
removing atmospheric disturbances. Thus, for the
on just two SAR images acquired at different times
selected set of scatterers, very accurate displacement
should be treated with care. Without prior informa-
data can be obtained. As in conventional InSAR,
tion, even considering areas not affected by possible
range variations are computed with respect to
phase unwrapping errors and phase decorrelation, the
a reference point that is either stable or is monitored
precision of range displacement data does not exceed
by other means, such as a GNSS receiver.
1–2 cm if the area of interest is larger than a few
More recent algorithms, such as the SqueeSAR™
square kilometers, as the impact of atmospheric dis-
approach of Ferretti et al. (2011), allow one to increase
turbances increases with the distance from the refer-
the spatial density of measurement points over the
ence point (Ferretti, 2014). Improved results can be
area of interest by identifying two distinct families of
obtained working on a long time series of SAR data,
points: permanent scatterers (PSs) and distributed 97
using a multi-interferogram approach.
MP1
[in]
6
4
2
0
–2
2014/01/01 2015/01/01 2016/01/01 [date]
MP1
MP2
[in]
2
MP2
1
0
–1
MP3 –2
≤10
–3
2014/01/01 2015/01/01 2016/01/01 [date]
MP3
[in]
0
0
–5
–10
≥–10 –15
[km]
[mm/yr] 0 0.5 1
2014/01/01 2015/01/01 2016/01/01 [date]
Figure 6.4 SqueeSAR™ analysis performed over an oil field. Red areas indicate measurement points (MP) affected by subsidence (negative
displacement rates), while blue areas are affected by uplift. Some examples of MP displacement time series are displayed on the right side of
the picture.
scatterers (DSs). PSs are radar targets characterized by necessary to create a reliable statistical analysis of the
high reflectivity values usually corresponding to radar returns, making it possible to identify pixels
buildings, metallic objects, pylons, antennae, out- exhibiting good SNR. The higher the number of
crops, and such. PSs generally produce very bright, images, the better the quality of the results.
isolated, pixels in a SAR scene. Conversely, distribu- We reiterate that, for displacement data associated
ted scatterers are radar targets corresponding to to both permanent and distributed scatterers, a key
a collection of pixels in the SAR image, all exhibiting factor is the distance from the reference point.
very similar radar reflectivity. Phase decorrelation, The relative accuracy can be better than a few milli-
though still affecting DSs, can be usually managed meters for a distance less than the average correlation
by means of spatial averaging procedures, allowing length of the atmospheric components (about 4 km at
for a significant improvement of the quality of the mid-latitudes). Average displacement rates can
interferometric fringes. These pixels usually corre- usually be estimated with a precision better than
spond to rocky areas, detritus, and areas of short 1 mm/year, depending on the number of data avail-
vegetation. able and the temporal span of the acquisitions
Provided that enough images are available, one (Ferretti, 2014). Figure 6.4 is an example of
can estimate a time series of range change with a SqueeSAR™ analysis performed over an area in the
a precision dependent on the number of available United States, using images acquired by the high-
data and the distance from the reference point, but resolution, X-band, COSMO-SkyMed SAR sensors
typically better than 3–4 mm. This may be done between 2013 and 2016. The measurement points
regardless of the type of measurement point identified are color-coded according to the average displace-
by the algorithm, that is, for both permanent and ment rate. Subsidence patterns (negative displace-
distributed scatterers. To perform a successful multi- ment rates are indicated by yellow to red) produced
98 interferogram analysis, a minimum number of satel- by oil production are clearly visible, as well as areas
lite images, approximately 10–15, are required. This is affected by uplift due to fluid injection (in blue). Over
Figure 6.5 Example of motion

decomposition combining ascending and
descending acquisition geometry.
nonvegetated areas, the spatial density of measure- because the satellite orbit is almost parallel to the
ment points identified by multi-interferogram techni- Earth’s meridians, the sensitivity to possible motion
ques can exceed 10,000 measurement points per in the north–south direction is negligible.
square kilometer. For each measurement point iden- An example of two-dimensional measurements is
tified, a displacement time series is available and it can presented in Figure 6.6. The area of interest is
be used successfully to run history matching algo- a stacked carbonate reservoir in the Middle East
rithms and, in general, for reservoir monitoring (Klemm et al., 2010; Tamburini et al., 2010). For this
purposes. case study, one ascending and one descending SAR
data set have been acquired over a time period of four
Two-Dimensional Displacement years. The ascending time series was acquired by the
Envisat satellite, while the descending data set was
Decomposition acquired by RADARSAT-1. Over the 120-km2 area,
Satellite SAR interferometry measures only the pro- some 280,000 and 360,000 InSAR measurement
jection of a three-dimensional displacement vector points were identified in the ascending and the des-
only along the satellite’s line-of-sight. However, it is cending data set respectively.
possible to combine data acquired from different
acquisition geometries to estimate two-dimensional Multidimensional Small Baseline Subset
displacement fields (Fujiwara et al., 2000; Wright
et al., 2004; Teatini et al., 2011; Rucci et al., 2013). Methodology
In fact, all SAR sensors follow near-polar orbits and The time series of one-dimensional line-of-sight
every point on the Earth can be imaged by two differ- deformation can be computed from a set of highly
ent acquisition geometries: one with the satellite fly- coherent interferograms acquired with small tem-
ing from north to south (descending mode), looking poral and spatial baselines using the small baseline
westward (for right-looking sensors) and the other subset (SBAS) technique (Berardino et al., 2002;
with the antenna moving from south to north Lanari et al., 2004; Samsonov et al., 2011).
(ascending mode), looking eastward. This is the rea- The temporal resolution of the estimated time series
son why, by combining InSAR results from both is defined by satellite revisit time and their spatial
acquisition modes, it is possible to estimate two com- resolution and coverage are similar to those of sin-
ponents of displacement (Figure 6.5), the quasi-east– gle-pair interferograms. The multidimensional small 99
west and the quasi-vertical components. In fact, baseline subset (MSBAS) technique computes two-
(a) Ascending Descending (b)

+50
mm/year
(c) Vertical Horizontal (E-W) (d)
–50
Horizontal Displacement Rate [mm/year]

–20 0 +20
Figure 6.6 Example of estimation of vertical and horizontal (east–west) displacement components from ascending and descending radar
data acquired by two SAR sensors over an oil field in the Middle East.
dimensional, horizontal east–west and vertical time from multiple SAR sensors. MSBAS methodology has
series of ground deformation from ascending and been used successfully for measuring ground deforma-
descending DInSAR (differential InSAR) data tion due to carbon sequestration (Samsonov et al.,
acquired by various SAR sensors with different acqui- 2015; Czarnogorska et al., 2016), mining (Samsonov
sition parameters, such as azimuth and incidence et al., 2013a,b, 2014a), urban development (Samsonov
angles, spatial resolution, and wavelength: et al., 2014b, 2016a), permafrost aggradation and pingo
growth (Samsonov et al., 2016b), and volcanic activity
^
A Ve ^
Φ
¼ ð6:4Þ (Samsonov et al., 2012, 2014c,d).
λL V u 0
where the matrix A ^ consists of time intervals between Interpretation Techniques

consecutive SAR acquisitions and the east–west and For oil and gas applications, InSAR analysis can
vertical components of a line-of-sight vector, Ve and provide data on (1) reservoir compaction/expan-
Vu represent unknown east–west and vertical veloci- sion and surface subsidence or uplift, (2) fault
ties that are to be determined, Φ ^ represents observed reactivation, and (3) areas of possible well failure.
DInSAR data, geocoded, and resampled to a common In some cases, surface deformation measurements
grid, λ is a regularization parameter and L is a zero-, can provide insight about where extracted fluids are
first-, or second-order difference operator. The drawn from and where injected fluids flow to, two
unknown parameters Ve and Vu for each pixel are of the most important questions in reservoir man-
solved by applying the singular value decomposition agement. To link surface deformation to reservoir
(SVD) of the coefficient matrix (Press et al., 2007) and parameters, we need to use geomechanics and geo-
the deformation time series are reconstructed from the physical inversion methods. InSAR data are becom-
computed deformation rates by numerical integration. ing more and more important in projects related to
Produced two-dimensional time series have combined enhanced oil recovery, underground gas storage
temporal resolution, and near daily temporal resolu- projects, and carbon capture and storage (CCS),
100
tion can be achieved by simultaneous processing data where it is becoming a standard monitoring tool.
Both forward and inverse modeling can aid in the rate of change of an observation with respect to
interpretation of geodetic observations related to a change in a model parameter is used to iteratively
the injection and storage of CO2. In forward mod- update a given model. We illustrate this approach in
eling one is given the properties of the medium, the the text that follows.
injection rates, and the properties of the fluids and To make things concrete, we shall consider
calculates the resulting fluid saturations, fluid pres- a specific inverse problem. First, note that the inverse
sures, reservoir volume changes, and deformation problem will be as complicated as the forward model
by solving the governing equations. Inverse model- on which it is based. In this example, based on an
ing, or solving the inverse problem, involves using application from the In Salah storage project, we shall
observed saturation, well pressure, and surface assume that the injected CO2 migrates through
deformation to estimate the properties within the a fracture zone intersecting the well (Vasco et al.,
reservoir that are driving the deformation. 2010, Rucci et al., 2013). The injected CO2 leads to
The first task is to develop a conceptual model of aperture changes (fracture opening) distributed over
the reservoir and of the processes controlling flow the fracture zone. The aperture changes will be
within the reservoir. Such a model is usually based on assumed to vary spatially within the fracture zone
well logs and core samples gathered during the drilling and also as a function of time. The fluid will be
of injection and monitoring wells. The conceptual assumed to stay within the fault/fracture zone for
model may be based on additional geophysical data, the duration of the injection. The unknown model
such as seismic reflection surveys defining reservoir parameters for this inverse problem are the aperture
boundaries and imaging visible faults. The conceptual changes within the fault/fracture zone during a given
model is typically large scale and incomplete, with time interval. Note that we are not modeling the fluid
important features, such as the permeability distribu- flow within the fault/fracture zone and will not be
tion within the reservoir, poorly constrained. relating aperture change to fluid pressure changes
Therefore, one must be prepared to modify the con- within the zone. That is another level of complexity
ceptual model in the face of new information such as that we shall not attempt to address, as we do not
flow data and/or geophysical monitoring data. For know the exact nature of the fracture zone because no
example, at the In Salah site, an anomalous double- core samples or log data were retrieved from it.
lobed pattern of InSAR range change suggested fault or Aperture changes in the fracture zone drive deforma-
fracture flow, rather than uniform and homogeneous tion within the surrounding medium, possibly leading
fluid migration through the porous sandstone of the to deformation at the Earth’s surface. We will further
reservoir (Vasco et al., 2010). assume that the material surrounding the fault/frac-
The conceptual model, along with reservoir prop- ture zone behaves elastically during the weekly to
erties from available well information, can be used in monthly time intervals between observations.
conjunction with a coupled hydromechanical reser- Therefore, the displacements are governed by the set
voir simulator to calculate the fluid distribution and of equations (Aki and Richards, 1980):
deformation due to the injection of CO2. This is the
∂ ∂ui ∂uj ∂uk ∂S
essence of the forward problem. The resulting predic- μ þ þλ δij ¼ ð6:5Þ
∂xj ∂xj ∂xi ∂xk ∂xj
tions can be compared with observed flow data and
surface deformation. The flow and mechanical prop-
where S is the source term. Using the representation
erties of the medium can be adjusted in order to
theorem (Aki and Richards, 1980), we can write the
produce better agreement between observed and cal-
solution to these equations in terms of an integral over
culated quantities. Such trial-and-error adjustments
the surface of the fault/fracture zone. A Green’s func-
are the simplest approach for solving the inverse pro-
tion is the solution of the governing equations corre-
blem, where observations are used to estimate the
sponding to a delta-function (impulsive) source.
properties of the medium. There are much more
The expression for the i-th component of displace-
sophisticated techniques for solving the inverse pro-
ment at point x for time t is given by the integral:
blem, some of which are based on stochastic methods ð
that rely on trial-and-error approaches. Other meth-
ui ðx; tÞ ¼ aðς; tÞgi ðx; ςÞdΣ ð6:6Þ
ods are based on sensitivity calculations in which the Σ 101
where aðς; tÞ is the aperture change at location ς at especially because the matrix M is a smoothing opera-
time t, Σ is the fault/fracture surface, and gi ðx; ςÞ is the tor that averages the aperture changes in order to
Green’s function. The integral expression can be writ- produce calculated range changes. Because of the
ten as a discrete sum if we represent the aperture large number of InSAR observations, the linear sys-
changes defined over the surface as piecewise constant tem of equations may be formally overdetermined,
functions defined over rectangular patches. Then the with more equations than unknown parameters.
integral may be written as the sum: However, owing to the averaging nature of the matrix
XN M many of the equations are not independent and
ui ðxj ; tÞ ¼ n¼1 i
Gn ðxj Þan ¼ Gi ðxj Þ aðtÞ ð6:7Þ the system of equations will be effectively underdeter-
mined, with too few independent equations to resolve
where an is the average aperture change of the n-th each model parameter. Furthermore, because of
fault/fracture patch and Gi ðxj Þ is a vector with com- errors in the observations and approximations in the
ponents that are the integral of the Green’s modeling, the equations may be inconsistent and may
function: not be directly solvable. To overcome this difficulty,
ð one may adopt a least squares approach, whereby one
Gni ðxj Þ ¼ gi ðxj ; ςÞdΣ ð6:8Þ finds the solution that minimizes the sum of the
Rn
squares of the residuals, rather than trying to solve
over the rectangular patch Rn . The expression for every equation exactly. Thus, we seek the aperture
ui ðxj ; tÞ is suitable for vector displacement data, such distribution g that minimizes:
as that obtained from the GNSS. However, a range
change, as derived from InSAR observations, is ρ2 ¼ ðr M aÞt ðr M aÞ ð6:13Þ
a projection of the displacement vector onto the look
vector pointing in the direction of the satellite, l: or some weighted version of the original equations.
To treat the nonuniqueness that is introduced by the
rðxj ; tÞ ¼ l u ¼ li ui ð6:9Þ
fact that the equations are effectively underdeter-
where we are invoking the convention of summation mined, one may introduce additional terms, referred
over repeated indices. Substituting the representation to as penalty terms biasing the solution in some fash-
of ui ðxj ; tÞ as a sum, or its vector matrix equivalent ion. This approach is known as regularization. For
Gi ðxj Þ aðtÞ, we can write the range change as example, it may be desirable to have a solution that
rðxj ; tÞ ¼ mj aðtÞ ð6:10Þ does not stray from a prior model of aperture changes
ap ; thus we would minimize:
where ρ2 ¼ ðr M aÞt ðr M aÞ þ wp ða ap Þt ða ap ÞI
mj ¼ li Gi ðxj Þ ð6:11Þ ð6:14Þ
and xj signifies an observation point, such as
a permanent scatterer. Given a large collection of where wp is a weighting coefficient that controls the
measurements we obtain a linear system of equations trade-off between fitting the data and remaining close
relating the aperture changes over the fracture zone to the prior model. The conditions for a minimum, or
during a particular time interval, a, to get a maximum for that matter, are given by the vanishing
of the gradient of the augmented functional with
r¼Ma ð6:12Þ respect to a:
where we have suppressed the explicit time ra ρ2 ¼ Mt r wp ap þ ½Mt M þ wp Ia ¼ 0 ð6:15Þ
dependence.
In the forward problem we use a model of aperture
changes to calculate the resulting range change and leading to a linear system of equations for the aperture
this follows from simple matrix multiplication in Eq. changes. The estimated aperture changes are given by
(6.12). In the inverse problem we have to solve the the formal solution of Eq. (6.16):
linear system of equations for the vector a on the
102 â ¼ ½Mt M þ wp I½Mt r þ wp ap ð6:16Þ
right-hand side. This is a much more difficult task,
Other forms of regularization may be introduced

in order to emphasize other aspects of the solution.
Gas producer wells
For example, one may minimize the roughness of the KB-13
KB-15
solution by introducing a penalty term that represents KB-12
KB-501
one of the spatial derivatives of the model (Menke, KB-11
KB-14
KB-502 Surface
1989). Alternatively, it might be desirable to produce KB-503
~460m elevation
a model in which the aperture changes are largest near Top of caprock
the injection well, where the fluid pressure changes ~850m deep
are the greatest. Thus, one can introduce a penalty

term for which aperture changes that are far from the
well are penalized. Such terms can be written as quad- Faulted Reservoir Layer
ratic forms in the model parameters, leading to ~1850m deep
a linear system of equations when differentiated.

A critical component of the inverse problem is an 1 km
E S
W
N
assessment in which model parameter uncertainties
and resolution are estimated (Menke, 1989). There is
a well-established formalism for model assessment Figure 6.7 General structural setting and well placement at the
In Salah gas storage site in Algeria.
when the inverse problem is linear. We will not
describe these techniques here. Rather, we refer the
reader to in-depth discussions of inverse problems fracture planes, in order to maximize the injectivity
(Menke, 1989). and the influence of the fractures.
The general structure of the reservoir is that of
Applications a gentle anticline with a few degrees of dip. The gas
field and producing wells are located on the crest of
There are a number of sites worldwide where significant
the anticline and the injection wells are on the flanks
volumes of CO2 are stored within the Earth, including
of the anticline (Figure 6.7). The reservoir is overlain
Sleipner (Chadwick et al., 2012), Weyburn (White et al.,
by more than 800 m of mudstone, forming an exten-
2011), and Cranfield (Hovorka et al., 2013). Here we
sive caprock seal above the injection zone. Topping
will discuss three additional sites where InSAR plays
this is approx. 900 m of interbedded sands and shales
a significant role in the monitoring program.
that contains the regional Pan Saharan potable aqui-
fer, an important water source for central and south-
In Salah, Algeria ern Algeria.
At the In Salah storage project excess CO2 from three In 2005 the Joint Industry Project (JIP),
natural gas fields is separated, processed, compressed, a collaboration between various research institutions
and reinjected into the flanks of one of the reservoirs. and the operating companies, was set up with the goal
A total of three horizontal wells injected more than of monitoring the fate of the injected CO2 using
3 million tons of CO2 between 2004 and 2008. This geophysical, geochemical, and reservoir production
made In Salah one of the largest storage projects in the data (Mathieson et al., 2011). As part of this effort,
world (Mathieson et al., 2010, 2011). The reservoir Lawrence Berkeley Laboratory contracted with TRE
chosen to contain the CO2 was not ideal for the of Milan, Italy to provide estimates of range change
storage of a large amount of CO2. While the sandstone from InSAR observations of the European Space
reservoir was laterally extensive, it was only 20 m thick Agency’s (ESA) Envisat satellite. Fortunately, these
on average, and its porosity varied between 10% and data were available before the start of injection in
18%. Image logs and mud loss data suggested that the 2004, allowing for the establishment of baseline obser-
reservoir contained two or three northwest-trending vations with which to measure range changes due to
fractures per meter, with apertures ranging between the injection of carbon dioxide (Figure 6.8). In Salah’s
0.1 and 1.0 mm, and fracture lengths varying between rough, boulder-strewn ground surface, with little
6 and 25 m (Iding and Ringrose, 2010). The three movable sand, proved favorable for InSAR monitor-
horizontal wells were oriented in the direction of the ing. The injected volume of carbon dioxide has pro- 103
least principle stress, perpendicular to the most likely duced more than 30 mm of observable range change,
1261 days Figure 6.8 Range change after 1226

29.240 days of injection.
29.202
Distance north (degrees)
KB-502
29.164 KB-503
29.126
KB-501
29.088
29.050
2.050 2.104 2.158 2.212 2.266 2.320
Distance east (degrees)
–30. 30.
Range decrease (mm)
measured from roughly 300,000 permanent scatterers zones were required at all three injectors in order to fit
(Vasco et al., 2010). Uplift at the Earth’s surface causes the observations (Rucci et al., 2013).
reflectors on the ground to move closer to the refer- The Envisat C-band satellite was decommis-
ence point in space, resulting in a decrease in range. sioned and in October 2010 was no longer in service
Significant uplift is observed over all three injector to monitor the range change over In Salah.
wells (KB-501, KB-502, and KB-503), with amplitudes Fortunately, two constellations of X-band satellites:
that are much larger than the estimated 5-mm error TerraSAR-X and Cosmo-SkyMed, were available and
associated with the measurements. A double-lobed the monitoring was transitioned to them in 2009 and
pattern, first noted by collaborators from Pinacle 2010. The X-band satellites had better temporal sam-
Technologies, is visible in the uplift measured over pling, with return times of 8 and 11 days, in compar-
injector KB-502. Such a pattern is suggestive of the ison to the best possible 35-day repeat time of
opening over a near vertical planar surface, such as Envisat satellite. This allowed for better estimates
a fault or fracture (Davis, 1983). Based on this obser- of corrections to the phase estimates. TRE merged
vation, Vasco et al. (2010) constructed a damage zone the three data sets – Envisat, TerraSAR-X, and
model with variable aperture change over a vertical Cosmo-SkyMed – into a common time series and
planar feature. They conducted an inversion of the provided it to Lawrence Berkeley National
range changes with the goal of estimating the aperture Laboratory for analysis.
change on the fracture zone and volume change The merged data set suggests a double lobed pat-
in the reservoir itself. Subsequent work with two– tern over well KB-501 for selected time increments
104 component displacement estimates (quasi-east–west (Figure 6.9). This supports the findings of Rucci et al.
and quasi-vertical) suggested that subvertical damage (2013) that only a model containing aperture and
107 days 212 days

29.150 29.150
29.134

29.134
29.118 29.118
29.102 29.102
29.086 29.086
29.070 29.070
2.240 2.256 2.272 2.288 2.304 2.320 2.240 2.256 2.272 2.288 2.304 2.320
Distance east (degrees) Distance east (degrees)
–5.0 Range change (mm) 10.0 –5.0 Range change (mm) 10.0
317 days 702 days

29.150 29.150
29.134
29.134
29.118 29.118
29.102 29.102
29.086 29.086
29.070 29.070
2.240 2.256 2.272 2.288 2.304 2.320 2.240 2.256 2.272 2.288 2.304 2.320
Distance east (degrees) Distance east (degrees)
–5.0 Range change (mm) 10.0 –15 Range change (mm) 20.
Figure 6.9 Range change over well KB-501 from the stiched Envisat, TERRASar-X, and Cosmo-SkyMet data set.
volume change on a subvertical damage zone could developed a conceptual model of flow and deforma-
satisfy both the quasi-east–west and quasi-vertical tion associated with the injection at KB-501. A major
data sets simultaneously (Rucci et al., 2013). For component of the deformation appears to be due to
time intervals of 702 days and greater, the second aperture change over a vertical fault/fracture, or
lobe disappears, perhaps because flow in the reservoir damage zone, intersecting the well. Note that this
begins to dominate the source of the deformation. conclusion is also supported by the analysis of Vasco
The generally higher quality corrections to the et al. (2008) and the linear high-permeability feature
merged data set and some spatial averaging helped suggested by a sequence of inversions for flow in the
to bring out the subtle pattern in the measurements. reservoir layer. The concentrated flow was correlated
Based on the suggestion of a double lobed pattern in with a hypothesized fault/fracture zone observed as
105
Figure 6.9 and the results of Rucci et al. (2013), we a break in a seismic horizon.
107 DAYS 212 DAYS

1.5 1.5
1.6 1.6
1.7 1.7
1.8 1.8
1.9 1.9
2.0 2.0
2.1 2.1
–5.0 –3.0 –1.0 1.0 3.0 5.0 –5.0 –3.0 –1.0 1.0 3.0 5.0
0.0 APERTURE CHANGE (CM) 0.25 0.0 APERTURE CHANGE (CM) 0.25
317 DAYS 877 DAYS

1.5 1.5
1.6 1.6
1.7 1.7
1.8 1.8
1.9 1.9
2.0 2.0
2.1 2.1
–5.0 –3.0 –1.0 1.0 3.0 5.0 –5.0 –3.0 –1.0 1.0 3.0 5.0
0.0 APERTURE CHANGE (CM) 0.25 0.0 APERTURE CHANGE (CM) 0.50
Figure 6.10 Distribution of aperture change over the damage zone for four time intervals, corresponding to the range change shown in
Figure 6.9.
To study the evolution of the aperture change the fracture opening. Specifically, we can compute
with time, due to the injection of CO2, we can con- the time at which the aperture begins to change
struct a model of the damage zone following the from its background value (Figure 6.11). From this
approach described in the section on interpretation figure one observes that the earliest aperture
techniques. Specifically, we postulated that aperture change is in a region surrounding the well, as
change generating the surface deformation is distrib- might be expected due to the pressure changes
uted over a vertical planar zone. As indicated in the associated with the injection of CO2. With time
previous section, we can subdivide such a zone into the aperture changes extend along reservoir and
a grid of voxels encompassing the damage zone and into a weak zone overlying the reservoir. For
set up a least squares problem for the aperture a brief period the aperture change appears to
change within each element. The grid geometry is have migrated upward before extending laterally
illustrated in Figure 6.10 and the solution to the least outward.
squares problem for the aperture change is also The results of the inversions suggest that the acti-
plotted in Figure 6.10 for four time intervals. That vated or reactivated fracture has remained within the
is, the solutions minimize the misfit to the range reservoir and the immediately overlying formations.
changes while simultaneously minimizing aperture However, owing to the nature of geodetic data, the
changes that are far from the injection well (Vasco solution to the inverse problem is nonunique and an
et al., 2010). infinite number of solutions are possible. Ramirez and
Because we have range change estimates every Foxall (2014) used a stochastic approach to try to
106 month or so, we can calculate the time history of characterize the range of possible solutions associated
the previous sections, with its combination of PSs and

1.5
DSs, is a flexible method for treating such a wide variety
1.6 of land surfaces. The distribution of the approx. 109,000
1.7 scatterers found in the region surrounding the injection
1.8 well (Falorni et al., 2014) is shown in Figure 6.12.
Another difficulty is introduced by the seasonal
1.9
and atmospheric conditions in this region, including
2.0 long periods of snow cover. Some PSs such as steep
2.1 roofs and tall towers may not accumulate much snow,
–5. –3. –1. 1. 3. 5. but large areas will be covered, reducing the coherence
and introducing significant variations in the charac-
0.0 ARRIVAL TIME (DAYS) 300. teristics of the scattering surface. Artificial reflectors,
manmade, stable targets designed to remain free of
Figure 6.11 Onset of aperture change within the damage zone. snow, are one remedy and are commonly used in
areas prone to snowfall. To provide for year-round
with the deformation observed over well KB-502. coverage, 21 artificial reflectors, spaced 75 m apart,
They found that the majority of solutions lie within were constructed and emplaced in an open area close
at or below the hot shale, within 200 m of the to the injection well-pad (Figure 6.12). All 21 reflec-
reservoir. tors were found to have strong and stable reflectivity,
both for ascending and descending satellite orbits.
The acquisition of COSMO-SkyMed X-band data
Decatur Project, Illinois Basin, USA began in July 2011, before the start of injection
The Illinois Basin Decatur project, started in November. Falorni et al. (2014) have reported on
in November 2011, is a multiyear program managed roughly two years of observations from July 2011
by Archer Daniels Midland, the U.S. Department of until June 2013. They note that there is little or no
Energy, the Illinois State Geological Survey, and ground motion associated with the start of injection
Schlumberger Carbon Services (Falorni et al., 2014). with the exception of two points. For example, a time
This is the only carbon capture and storage effort in series consisting of the average range change from all
the United States that is currently injecting large 21 reflectors is plotted in Figure 6.13. The two points
volumes of greenhouse gas emissions into a regional showing some movement correlating with the start of
deep saline formation (Finley et al., 2011, 2013). injection were in the vicinity of the injection well and
The injection, from an ethanol production facility, may reflect some movement related to the pressuriza-
started at a rate of 1000 metric tons/day and is tion of the well bore itself (Figure 6.14).
planned to increase to more than 2000 metric tons/ The lack of injection-related surface deformation
day. The CO2 is stored in the 550 m thick Mount is an expected result. Owing to the high permeability
Simon sandstone at a depth of 2.1 km. The overlying and thickness of the reservoir sandstone, the bottom
Eau Claire shale forms a 100–150 m thick seal. hole pressure has changed very little during the injec-
The Mount Simon formation was the site of tion. Barring any unforeseen geological features, such
a functioning natural gas storage facility and had as an unmapped fracture zone, the CO2 is expected to
performed well in that capacity. Several different geo- migrate outward into the reservoir without generating
chemical, geophysical, and remote sensing technolo- large pressure changes or associated deformation.
gies are employed at the site to monitor the evolution
of the injected volume of CO2. Preliminary results
from microseismic and InSAR monitoring were Aquistore, Williston Basin, Canada
reported by Kaven et al. (2014) and Falorni et al. The Aquistore storage site acts as a buffer for the
(2014), respectively. world’s first commercial post-combustion CO2 cap-
The injection site is situated in a mixture of indus- ture plant. The operation of the plant began
trial sites, farmland, forest, and residential areas and in October 2014 at the Boundary Dam coal-fired
InSAR monitoring can be challenging in such a diverse power station in Saskatchewan, Canada (Worth
environment. The SqueeSAR™ algorithm described in et al., 2014). The injection of CO2 into the subsurface
107
Figure 6.12 (Left) Distribution of scatterers in the region surrounding the injection well (magenta star), roughly 4000/km2. (Right) Close-up
view of the region surrounding the injection well-pad. The grid of artificial reflectors is visible to the northwest of the wellhead.
A0001 - Averaging - coherence: 0.9238 - velocity: 0.1477 - velocity standard deviation: 0.4044
40
30
20
10
[mm]
0
–10
–20
–30
–40
2011/08/01
Start of Injection
2011/12/01
2012/04/01
2012/08/01
2012/12/01
2013/04/01
[Date]
Figure 6.13 Average range change from the 21 artificial reflectors installed just to the northwest of the injection well site (Falorni et al., 2014).
The start of injection is indicated by the vertical blue line.
began in April 2015 with maximum anticipated injec- sedimentary units in the Williston Basin, and are
tion rates of up to 1500 tonnes/day to brine-filled located below any oil-producing and potash-bearing
clastic strata of the Deadwood and Winnipeg forma- formations (Norford et al., 1994). Similar deep saline
tions, in the depth range of 3150–3350 m (Norford formations are found elsewhere in western Canada
108 et al., 1994). These two formations are the deepest and throughout North America and the world.
A3LVP - coherence: 0.6 - velocity: 3.73 - velocity standard deviation: 0.69 A3LVR - coherence: 0.71- velocity: 3.5 - velocity standard deviation: 0.61
40
30
20
10
[mm]
0
–10
–20
–30
–40
2011/08/01
Start of Injection
2011/12/01
2012/04/01
2012/08/01
2012/12/01
2013/04/01
[Date]
Figure 6.14 Time series of two measurement points in the vicinity of the injection well. The vertical blue line indicates the start of injection.
The rocks in these formations are well suited for the is a two-dimensional, vertical and horizontal east–
storage of CO2 because they are porous, permeable, west deformation, time series referenced to a stable
and thick, leading to huge storage volumes, and are point (R in Figure 6.10). Although MSBAS is capable
overlain by substantial geological seals that will of measuring all three components of deformation,
impede the upward migration of the injected fluid. such processing requires data from other than the
Various in situ and remote sensing data are col- near-polar orbiting sensors. The availability of
lected at the Aquistore site for the determination of space-borne SAR data has been steadily improving,
the fluid distribution in the subsurface, pressure making this remote sensing technique a valuable,
changes, and ground deformation associated with cost-effective alternative to GNSS receivers and level-
the injection. For measuring surface deformation, ing. In an effort to improve the precision of the defor-
high-resolution ascending and descending SAR mation measurements, especially during winter when
RADARSAT-2 Spotlight and Wide Ultra-Fine SAR the ground is covered by snow, a network of corner
data were acquired starting in June 2012. reflectors was installed with reflectors at nine moni-
The Spotlight data consist of ascending Spotlight 18 toring sites marked as NE1, NE2, SE1, SE2, SE3, SITE,
(SLA18) and descending Spotlight 12 (SLA12) images, SW1, NW1, and NW2 (Figure 6.15). These corner
acquired with the incidence angles of 44 and 40 reflectors, designed at the Canada Centre for
degrees and the range-azimuth spatial resolution of Mapping and Earth Observation, Natural Resources
1.6 by 0.8 m. The Wide Ultra-Fine data consist of Canada, consist of two trihedral reflectors positioned
ascending and descending Wide 2 Ultra-Fine 7 for ascending and descending imaging geometry,
(U7W2) images, acquired with an incidence angle of mounted on a single monument. To more accurately
37 degrees and the range-azimuth spatial resolution of capture ground deformation from a deep source such
1.6 by 2.8 m. The overall temporal resolution is as the injected CO2, and to reduce signals from the
increased by a factor of 4, from 24 (i.e., standard near-surface processes caused by seasonal changes
RADARSAT-2 revisit cycle) to 6 days, by combining and post-mining soil settlement, the corner reflector
four independent data sets. A 1-m resolution LiDAR monuments were anchored at a depth of 20 m.
DEM is used to remove the topographic phase RADARSAT-2 SAR data have been collected
contribution. since June 2012, prior to the injection that started
Differential interferograms are computed from in April 2015. The nearly three years of preinjection
SAR data, geocoded, resampled to a common grid, DInSAR observation allowed for the mapping of 109
and processed with the MSBAS technique. The result ground deformation due to background processes.
Figure 6.15 Location of Aquistore CO2 storage site in Saskatchewan, Canada (top right). Background is Advanced Spaceborne Thermal
Emission and Reflection Radiometer (ASTER) 30 m resolution Digital Elevation Model (DEM). RADARSAT-2 frames are outlined in black. Region
of interest is outlined in brown. Extent in top right corner shows location of study region in North America. Photograph of corner reflector
designed at Canada Centre for Mapping and Earth Observation, Natural Resources Canada by K. Murnaghan for ascending and descending
imaging (top left). Observed vertical (bottom left) and east–west horizontal (bottom right) deformation rates computed by applying
multidimensional small baseline (MSBAS) technique to RADARSAT-2 data spanning 20120615–20140923. Monitoring sites and reference
region R are plotted in black. Points P1–P9 experiencing fast ground deformation are plotted in red. SAR intensity image is exposed in
background in regions of low coherence.
The vertical and east–west horizontal deformation where construction activities have affected the sur-
maps are shown in Figure 6.15. During the face. Uplift was observed at manmade sites but also
2012–2015 time interval, active ground deformation in other locations influenced by groundwater
was observed across the region. Vertical deformation recharge and clay swelling. Horizontal motion of up
110 exceeded 0.8 cm/year and subsidence was predomi- to 0.4 cm/year was observed mainly along the steep
nantly observed in areas of post-mining reclamation, slopes of the diversion canal. Analysis of the
preinjection deformation maps documented that Canada. The InSAR monitoring described in this
some monitoring sites are affected by near-surface chapter provided insight, and some unexpected
deformation processes, in particular sites located results, regarding the migration of injected CO2 at
south of the injection well (SW1, SE2, SE3), empha- the In Salah gas development site. For example, the
sizing the importance of having preinjection baseline magnitude of surface deformation was significantly
observations. The InSAR observations serve to verify larger than anticipated. In addition, the geodetic
performance of the sequestration effort and to detect monitoring suggested that fault/fracture zones con-
any anomalous behavior. trolled the flow around each of the injectors at
In Salah. Such fault/fracture zones were not observed
Conclusions in the preinjection seismic survey and not detected
during the drilling of the injection wells.
Geodetic methods, particularly interferometric syn-
The monitoring results from the Decatur and
thetic aperture radar (InSAR), provide a cost-effective
Aquistore sites appear to indicate much less pressure
approach for monitoring the geological storage of
buildup and associated deformation due to the
CO2. Satellite-based techniques are nonobtrusive
injected CO2. This may be due to the fact that the
and, in many cases, do not require the placement of
target formations contain a much greater storage
instrumentation on the ground. Measurements of
volume and the flow around the wells is much more
surface displacement are particularly well suited for
uniform, leading to smoothly varying pressure
detecting the upward migration of CO2 under pres-
changes and smaller pressure gradients.
sure, due to the sensitivity of ground motion to near-
surface stress changes. The technology for geodetic
monitoring is improving over time, with satellite revi- Acknowledgments
sit times decreasing and new processing algorithms Portions of this work were performed under the aus-
under development. Inversion and interpretation pices of the GEOSEQ project for the Assistant
approaches are also advancing, increasing in sophis- Secretary for Fossil Energy, Office of Coal and Power
tication and efficiency. Systems, through the National Energy Technology
There are a number of issues associated with the Laboratory, U.S. Department of Energy by Lawrence
use of geodetic data for monitoring subsurface pro- Berkeley National Laboratory under contract number
cesses. Some techniques, such as InSAR, are sensitive DE-AC02-05CH11231 and by Lawrence Livermore
to the nature of the ground surface and to the pre- National Laboratory under Contract DE-AC52-
sence or absence of vegetation. Geodetic data gathered 07NA27344. All authors would like to thank the
at the Earth’s surface generally provide poor depth members of the Joint Industry Project and the Joint
resolution of subsurface processes, and issues of Venture, in particular BP, Sonatrach, and Statoil for
uniqueness are often significant. The development of their cooperation and support at every stage of this
a conceptual model is not always straightforward and work. Special thanks go out to Iain Wright, Allen
often requires additional data, such as seismic obser- Mathieson, Sue Raikes, John Midgley, Catherine
vations. Combining various data sets appears to be Gibson-Poole, David Roberts, Stephen Cawley, Phil
a promising approach for reducing the nonunique- Ringrose, Lykke Gemmer, and the other members of
ness of the inverse problem. However, such large- the Joint Industry Project. We would like to thank the
scale inversions and imaging can be a computational Canadian Space Agency for providing RADARSAT-2
challenge, particularly when combining disparate data.
data types, such as flow observations and geodetic
measurements.
The geological storage of CO2 is not yet a common References
occurrence and has been implemented only at Aki, K., and Richards, P. G. (1980). Quantitative seismology.
a handful of sites. Therefore, our experience with San Francisco: Freeman and Sons.
monitoring techniques is rather limited. Still, InSAR Bamler, R., and Hartl, P. (1998). Synthetic aperture radar
monitoring has been documented in a few sites, interferometry. Inverse Problems, 14: R1–R54.
including the three described here – In Salah, Berardino, P., Fornaro, G., and Lanari, R. (2002). A new
Algeria, Decatur, Illinois – and the Aquistore site in algorithm for surface deformation monitoring based on 111
small baseline differential SAR interferograms. IEEE operational experience at a one-million tonne CCS
Transactions on Geoscience and Remote Sensing, 40: demonstration project, Decatur, Illinois. Energy
2375–2383. Procedia, 37: 6149–6155.
Chadwick, R. A., Williams, G. A., Williams, J. D. O., and Fujiwara, S., Nishimura, T., Murakami, M., Nakagawa, H.,
Noy, D. J. (2012). Measuring pressure performance of Tobita, M., and Rosen, P. A. (2000). 2.5-D surface
a large saline aquifer during industrial-scale CO2 deformation of M 6.1 earthquake near Mt. Iwate
injection: The Utsira Sand, Nowegian North Sea. detected by SAR interferometry. Geophysical Research
International Journal of Greenhouse Gas Control, 10: Letters, 27: 2049–2052.
374–388. Funning, G. J., Parsons, B., Wright, T. J., Jackson, J. A., and
Czarnogorska, M., Samsonov, S., and White, D. (2016). Fielding, E. J. (2005). Surface displacements and source
Airborne and spaceborne remote sensing parameters of the 2003 Bam (Iran) earthquake from
characterization for Aquistore carbon capture and Envisat advanced synthetic aperture radar imagery.
storage site. Canadian Journal of Remote Sensing, 42: Journal of Geophysical Research, 110: B09406.
274–291. DOI:10.1080/07038992.2016.1171131. DOI:10.1029/2004JB003338.
Davis, P. M. (1983). Surface deformation associated with Hovorka, S. D., Meckel, T. A., and Trevino, R. H. (2013).
a dipping hydrofracture. Journal of Geophysical Monitoring large-volume injection at Cranfield,
Research, 88: 5826–5834. Mississippi-Project design and recommendations.
Dzurisin, D. (2007). Volcano deformation: Geodetic International Journal of Greenhouse Gas Control, 18:
monitoring techniques. Chichester: Springer. 345–360. DOI:10.1016/j.ijggc.2013.03.021.
Falorni, G, Hsiao, V., Iannaconne, J., Morgan, J., and Iding, M., and Ringrose, P. (2010). Evaluating the impact of
Michaud, J.-S. (2014). InSAR monitoring of ground fractures on the performance of the In Salah CO2
deformation at the Illinois Basin Decatur Project. storage site. International Journal of Greenhouse Gas
In Carbon dioxide capture for storage in deep geological Control, 4: 242–248. DOI:10.1016/j.ijggc.2009.10.016.
formations, 4. Thatcham, Berks: CPL Press. Kaven, J. O., Hickman, S. H., McGarr, A. F., Walter, S., and
Ferretti, A. (2014). Satellite InSAR data: Reservoir Ellsworth, W. L. (2014). Seismic monitoring at the
monitoring from space. Houten, The Netherlands: Decatur, IL, CO2 sequestration demonstration site.
EAGE Publications. Energy Procedia, 63: 4264–4272.
Ferretti, A., Prati, C., and Rocca, F. (2000). Nonlinear Klemm, H., Quseimi, I., Novali, F., Ferretti, A., and
subsidence rate estimation using permanent scatterers Tamburini, A. (2010). Monitoring horizontal and
in differential SAR interferometry. IEEE Transactions vertical surface deformation over a hydrocarbon
on Geoscience and Remote Sensing, 38: 2202–2212. reservoir by PSInSAR. First Break, 28: 29–37.
Ferretti, A., Prati, C., and Rocca, F. (2001). Permanent Lanari, R., Mora, O., Manunta, M., Mallorqui, J. J.,
scatterers in SAR inferometry. IEEE Transactions on Berardino, P., and Sanosti, E. (2004). A small-baseline
Geoscience and Remote Sensing, 39: 8–20. approach for investigating deformation on
full-resolution differential SAR interferograms. IEEE
Ferretti, A., Monti-Guarnieri, A., Prati, C., Rocca, F., and Transactions on Geoscience and Remote Sensing, 42:
Massonnet, D. (2007). InSAR Principles: Guidelines for 1377–1386.
SAR Interferometry Processing and Interpretation.
Noordwijk, The Netherlands: ESA Publications, Massonnet, D., and Feigl, K. L. (1998). Radar interferometry
TM-19. and its application to changes in the Earth’s surface.
Reviews of Geophysics, 36: 441–500.
Ferretti, A., Fumagalli, A., Novali, F., Prati, C., Rocca, F.,
and Rucci, A. (2011). A new algorithm for processing Mathieson, A., Midgley, J., Dodds, K., Wright, I.,
interferometric data-stacks: SqueeSAR. IEEE Ringrose, P., and Saoul, N. (2010). CO2
Transactions on Geoscience and Remote Sensing, 49: sequestration monitoring and verification technologies
3460–3470. applied at Krechba, Algeria. Leading Edge, 29(2):
216–222.
Finley, R. J., Greenberg, S. D., Frailey, S. M., Krapac, I. G.,
Leetaru, H. E., and Marsteller, S. (2011). The path to Mathieson, A. Midgley, J., Wright, I., Saoula, N., and
a successful one-million tonne demonstration of Ringrose, P. (2011). In Salah CO2 storage JIP: CO2
geological sequestration: Characterization, sequestration monitoring and verification technologies
cooperation, and collaboration. Energy Procedia, 4: applied at Krechba, Algeria. Energy Procedia, 4:
4770–4776. 3596–3603.
Finley, R. J., Frailey, S. M., Leetaru, H. E., Senel, O., McManamon, P. (2015). Field fuide to Lidar. Bellingham,
112 Coueslan, M. L., and Marsteller, S. (2013). Early WA: SPIE Press.
Menke, W. (1989). Geophysical data analysis: Discrete Samsonov S., d’Oreye N., and Smets, B. (2013b). Ground
inverse theory. San Diego: Academic Press. deformation associated with post-mining activity at the
Misra, P., and Enge, P. (2001). Global positioning system: French-German border revealed by novel InSAR time
Signals, measurements, and performance. Lincoln, MA: series method. International Journal of Applied Earth
Ganga-Jamuna Press. Observation and Geoinformation, 23: 142–154.
Norford, B., Haidl, R., Bezys, F.M., Cecile, M., McCabe, H., Samsonov, S., Gonzalez, P., Tiampo, K., and d’Oreye, N.
and Paterson, D. (1994). Middle Ordovician to Lower (2014a). Modeling of fast ground subsidence observed
Devonian strata of the Western Canada Sedimentary in southern Saskatchewan (Canada) during 2008–2011.
Basin. In G. Mossop and I. Shetsen (comp. eds.), Natural Hazards and Earth System Sciences, 14:
Geological Atlas of the Western Canada Sedimentary 247–257. DOI:doi:10.5194/nhess-14–247-2014.
Basin. Edmonton, Alberta: Canadian Society of Samsonov, S., d’Oreye, N., Gonzalez, P., Tiampo, K.,
Petroleum Geologists, Calgary, Alberta and Alberta Ertolahti, L., and Clague, J. (2014b). Rapidly
Research Council, 109–127. accelerating subsidence in the Greater Vancouver
Petrovski, I. G., and Tsujii, T. (2012). Digital satellite region from two decades of ERS-ENVISAT-
navigation and geophysics. Cambridge: Cambridge RADARSAT-2 DInSAR measurements. Remote Sensing
University Press. of Environment, 143: 180–191. DOI:10.1016/j.
rse.2013.12.017.
Press, W. H., Teukolsky, S. A., Vetterling, W. T., and
Flannery, B. P. (2007). Numerical recipes. Cambridge: Samsonov, S. V., Tiampo, K. F., Camacho, A. G.,
Cambridge University Press. Fernandez, J., and Gonzalez, P. J. (2014c).
Spatiotemporal analysis and interpretation of
Ramirez, A., and Foxall, W. (2014). Stochastic inversion of 1993–2013 ground deformation at Campi Flegrei, Italy,
InSAR data to assess the probability of pressure observed by advanced DInSAR. Geophysical Research
penetration into the lower caprock at In Salah. Letters, 41: 6101–6108. DOI:10.1002/2014GL060595.
International Journal of Greenhouse Gas Control, 27,
42–58. Samsonov, S. V., Trishchenko, A. P., Tiampo, K. Tiampo,
Gonzalez, P. J., Zhang, Y., and Fernandez, J. (2014d).
Rosen, P. A., Hensley, S., Joughin, I. R., et al. (2000). Removal of systematic seasonal atmospheric signal
Synthetic aperture radar interferometry. Proceedings of from interferometric synthetic aperture radar ground
the IEEE, 88: 333–382. deformation time series. Geophysical Research Letters,
Rucci, A., Vasco, D. W., and Novali, F. (2013). Monitoring 41: 6123–6130. DOI:10.1002/2014GL061307.
the geologic storage of carbon dioxide using Samsonov, S., Czarnogorska, M., and White, D. (2015).
multicomponent SAR interferometry. Geophysical Satellite interferometry for high-precision detection of
Journal International, 193(1): 197–208. ground deformation at a carbon dioxide storage site.
Rutqvist, J. (2011). Status of TOUGH-FLAC simulator and International Journal of Greenhouse Gas Control, 42:
recent applications related to coupled fluid flow and 188–199. DOI:10.1016/j.ijggc.2015.07.034.
crustal deformations. Computational Geoscience, 37: Samsonov, S., Tiampo, K., and Feng, W. (2016a). Fast
739–750. subsidence in downtown of Seattle observed with
Samsonov, S., and d’Oreye, N. (2012). Multidimensional satellite radar. Remote Sensing Applications: Society and
time series analysis of ground deformation from Environment, 4: 179–187. DOI:10.1016/j.
multiple InSAR data sets applied to Virunga Volcanic rsase.2016.10.001.
Province: Geophysical Journal International, 191: Samsonov, S. V., Lantz, T. C., Kokelj, S. V., and Zhang, Y.
1095–1108. DOI:10.1111/j.1365-246X.2012.05669.x. (2016b). Growth of a young pingo in the Canadian
Samsonov, S., van der Koij, M., and Tiampo, K. (2011). Arctic observed by RADARSAT-2 interferometric
A simultaneous inversion for deformation rates and satellite radar. The Cryosphere, 10: 799–810.
topographic errors of DInSAR data utilizing linear least DOI:10.5194/tc-10–799-2016.
square inversion technique. Computers and Shi, J.-Q., Sinayuc, C., Durucan, S., and Korre, A. (2012).
Geosciences, 37: 1083–1091. Assessment of carbon dioxide plume behavior within
Samsonov, S., Gonzalez, P., Tiampo, K., and d’Oreye, N. the storage reservoir and the lower caprock around the
(2013a). Methodology for spatio-temporal analysis of KB-502 injection well at In Salah. International Journal
ground deformation occurring near Rice Lake of Greenhouse Gas Control, 7: 115–126.
(Saskatchewan) observed by RADARSAT-2 DInSAR Smith, J. R. (1997). Introduction to geodesy: The history and
during 2008–2011. Canadian Journal of Remote Sensing, concepts of modern geodesy. New York: John Wiley &
39: 27–33. Sons.
113
Tamburini, A., Bianchi, M., Chiara, G., and Novali, F. carbon dioxide. Geophysical Research Letters, 37:
(2010). Retrieving surface deformation by PSInSAR L03303, 1–5. DOI:10.1029/2009GL041544.
technology: A powerful tool in reservoir monitoring. White, D. J., Meadows, M., Cole, S., et al. (2011).
International Journal of Greenhouse Gas Control, 4: Geophysical monitoring of the Weyburn CO2 flood:
928–937. Results during 10 years of injection. Energy Procedia, 4:
Teatini, P., Castelletto, N., Ferronato, M., et al. (2011). 3628–3635.
Geomechanical response to seasonal gas storage in Worth, K., White, D., Chalaturnyk, R., et al. (2014).
depleted reservoirs: A case study in the Po River basin, Aquistore project measurement, monitoring and
Italy. Journal of Geophysical Research, 116: 1–21. verification: From concept to CO2 injection. Energy
Torge, W. and Muller, J. (2012). Geodesy. Berlin: Walter de Procedia, 63: 3202–3208. http://dx.doi.org/10.1016/
Gruyter. j.egypro.2014.11.345.
Vasco, D. W., Ferretti, A., and Novali, F. (2008). Estimating Wright, C. A. (1998). Tiltmeter fracture mapping: From the
permeability from quasi-static deformation: Temporal surface and now downhole. Petroleum Engineer
variations and arrival time inversion. Geophysics, 73: International, 71: 50–63.
O37–O52. DOI:10.1190/1.2978164. Wright, T. J., Parsons, B. E., and Lu, Z. (2004).
Vasco, D. W., Rucci, A., Ferretti, A., et al. (2010). Satellite- Toward mapping surface deformation in three
based measurements of surface deformation reveal dimensions using InSAR. Geophysical Research Letters,
fluid flow associated with the geological storage of 31: 1–5.
114
Chapter
Gravity
7 Surface and Borehole

Ola Eiken
Introduction have been correctly predicted. Jacob et al. (2010),

Wilson et al. (2012), and Christiansen et al. (2011)
The technology of gravity monitoring has improved
all reported successful hydrology monitoring.
over the decades, with an increasing number and
Measurements on hydrocarbon fields with moving
range of applications. Time-lapse gravity data can be
gas–fluid fronts started at the giant Groningen gas
more straightforward to interpret than most other
field in 1978 (Gelderen et al., 1999; Eiken et al.,
geophysical data, because the signal is proportional
2017). Seven gas fields in the North Sea have been
to mass changes. This is a fundamental property to
monitored successfully (Alnes et al., 2008; Eiken et al.,
understand in a dynamic underground reservoir.
2008; Vevatne et al., 2012; Van den Beukel et al.,
The method, which also is termed microgravity, is
2014). Successive microgravity surveys over the gas
particularly powerful when the geological framework
cap of the Prudhoe Bay oil field from 2003 to 2008
is well known and fluids with large density differences
provided insights into the distribution of water
flow. Gas–water and air–water are such favorable
injected into the gas cap (Brady et al., 2008;
fluid contrasts, but also injected and stored carbon
Ferguson et al., 2008; Hare et al., 2008; Yin et al.,
dioxide (CO2) can provide significant negative density
2016).
contrast when it pushes away water. Further, CO2
injection into confined reservoirs will cause pore
expansion, mass increase, and thus gravity increases Techniques of Gravity Surveying
around it. At the same time, pore and reservoir expan- A basic limitation is set by measurement accuracy, or
sion will lift the surface up, which will cause gravity noise level. While measurements with precision down
reductions at fixed observation platforms. to a few μGal (one μGal is 10–8 ms–2) have been
Gravity changes have been monitored for a variety obtained at laboratory conditions for decades (e.g.,
of manmade and natural subsurface mass changes, Torge, 1989), recent advances have made it practically
including hydrologic, geothermal, and hydrocarbon possible to carry out field surveys at such accuracy,
gas extraction processes. The depth of the mass and stationary measurement at sub-μGal precision.
changes range from sinkholes a few meters below This increases the potential applications of gravity
the surface (e.g., Debeglia and Dupoint, 2002) to gas monitoring to smaller and deeper reservoirs, with
reservoirs kilometers down (e.g., Alnes et al., 2010). lower density contrasts and at shorter time-lapse
Gravity monitoring of geothermal reservoirs has intervals. High-precision measurements of gravity
helped determine the natural and injected water may be done by (1) absolute free-fall gravimeters, (2)
inflow distribution and phase since the 1960s, with superconducting gravimeters, or (3) relative spring
pioneering work in New Zealand (Allis and Hunt, gravimeters. The various instruments all have their
1986; Hunt and Kissling, 1994), Japan (Sugihara and advantages and limitations. Some key accuracy prop-
Ishido, 2008), Philippines (Nordquist et al., 2004), erties are listed in Table 7.1.
and Indonesia (Sofyan, 2011). Development of Absolute free-fall gravimeters measure the drop
magma chambers in volcanoes has been monitored time of a mass in a vacuum chamber. These are
(e.g., Battaglia et al., 2008; Carbone et al., 2017). Both immune to sensor drift. Repeated measurements are
campaign and continuous gravity observations have thus not required during a survey, and reference sta-
made valuable contributions, and volcanic eruptions tions outside the area of interest are not needed to see 115
Ola Eiken
Table 7.1 Summary of the most common gravity measurement techniques, and typical accuracies
Type of gravimeter Accuracy (μGal) Drift (μGal/yr) Typical duration of a single Environment
measurement (hours)
Absolute free-fall 2–10 none 6–24 Land
Superconducting 0.01–1 5–50 >240 Land
Relative spring 1–5 >1000 0.05–0.5 Land/sea/borehole
Figure 7.1 Absolute gravity instruments: FG5X (left) and A-10 (right). (From Micro-g LaCoste, http://www.microglacoste.com/absolutemeters
.php)
time-lapse changes. Semiportable instruments, the sensitive to measuring tides and atmospheric influ-
FG5 series (Figure 7.1), require setup and demobiliza- ence. Drift for state-of-the art instrumentation is
tion times of a few hours at an indoor site, and may give given as less than 0.5 μGal/month, and scale factor
2 μGal accuracy and 1 μGal repeatability. The smaller will be stable within 10–2 for years (Calvo et al., 2014).
and more easily portable absolute A-10 instrument can Even with these low values, for monitoring over sev-
give 10 μGal accuracy and repeatability in a 10-minute eral years, drift uncertainty can build up to several
reading on a quiet site. It can be operated outdoors, at μGal, and calibration by, for example, an absolute
temperatures ranging from –18°C to +38°C. meter is required. Sugihara et al. (2013) discussed
Superconducting gravimeters measure a spherical the use of superconducting gravimeters in CO2 sto-
superconducting mass, levitated using a magnetic rage monitoring, and considered it a promising,
force that exactly balances the force of gravity. though yet unproved, technique. Kim et al. (2015)
The iGRAV (Figure 7.2) requires a setup time of found that a CO2 signal larger than about 0.5 μGal
about five hours and a significantly longer sensor can be detected with an iGrav’s continuous record-
stabilization time, which makes the instrument best ings, by superposing real time series from supercon-
suited for months to years long continuous recording ducting meters onto modeled CO2 signals.
on a site. Resolution is better than 0.05 μGal after one- A relative spring meter utilizes a spring with
116
minute averaging, and it is therefore particularly a proof mass at the end, which extends proportional
Gravity: Surface and Borehole
drift can be controlled by repeated measurements

during a survey. The scale factor needs to be calibrated
from time to time. Repeatability is given at a few μGal.
A seafloor version of the CG5 gravimeter has been
built for offshore monitoring (Zumberge et al., 2008;
Figure 7.3).
Relative gravimeters can be used in campaigns/sur-
veys, which is when one instrument is doing sequential
measurements at a grid of stations. Sensor drift is con-
trolled by making repeat measurements at the same
station. Using multiple measurements at a site also
reduces other types of noise, such as random back-
ground noise. The uncertainty in final station values
can be significantly lower than the reading accuracy of
each measurement, and repeatabilities of 1–2 μGal have
been reported from several recent surveys (Vevatne
et al., 2012; Eiken et al., 2017; Lien et al., 2017).
Gravity data need several corrections before sta-
tion values and time-lapse changes can be estimated
(Table 7.2). Land data need to be corrected for varia-
Figure 7.2 iGrav instrument. (From GWR instruments, http://www
.gwrinstruments.com/pdf/principles-of-operation.pdf) tions in groundwater and shallow aquifers (hydrol-
ogy). Seafloor data need to be corrected for the
varying gravity attraction from the ocean above
to the force of gravity. The position of the mass is (ocean tides and seawater density effects). All data
accurately read off. The supplier Scintrex uses a quartz need to be corrected for the varying gravity attraction
spring with a drift of some μGal/hour, and the com- from the atmosphere, earth tides, and ocean loading,
pany ZLS uses metal zero-length springs (Figure 7.3). either by a good model or by actual recordings by
Most of the drift is linear, and much of the residual a stationary reference gravimeter.
Figure 7.3 Spring-based instruments. (Left) Scintrex land meter (http://www.scintrexltd.com/dat/content/file/CG-6%20Brochure_R15(1)

.pdf). (Middle) ZLS land meter (http://zlscorp.com/?page_id=33). (Right) Quad Geometrics seafloor meters (http://quadgeo.com/services/ 117
gravity-surveys/equipment/).
Ola Eiken
Table 7.2 Some gravity effects to be corrected for in time-lapse of 2000–2500 kg/m3. Nind and MacQueen (2013)
gravimetry processing
stated that depth-induced errors in Δg are typically
Effect Typical size of Residual noise comparable to the sensitivity of the gravity measure-
signal (μGal) (μGal) ments themselves. Depth measurements may be
Sensor drift 10–1000/day 1–10
obtained using high-resolution casing collar locator
logs.
Earth tides 200 <1
Borehole data are obviously limited by where the
Ocean loading 20 <1 boreholes are, their geometry, and whether they can
Atmospheric <10 <1 be accessed. CO2 injection operations will aim at
variations minimizing the number of injection wells, which
Microseism 50–5000 in the <1 may be relatively sparser than at many oil and gas
band >0.1 Hz fields.
Groundwater <50 1–10 Observation points may be above the reservoir,
and time-lapse changes will be in the same directions
as surface gravity but larger, due to closer proximity to
Stability of measurement platforms over years is the reservoir.
required for high-quality time-lapse data. This can be Borehole gravity can also be used as a density log-
achieved by installing benchmarks, made of, e.g., mas- ging tool with deep penetration into the formation.
sive concrete or metal. In the North Sea, conical con- When measurements are made at different depths, the
crete benchmarks have been residing on the seafloor gravity and height difference will tell the average for-
for this purpose for nearly 20 years (Stenvold et al., mation density. A Shuttle Sonde (trademark of Microg
2006). When vertical movements of the measurement LaCoste) can reduce the depth increment error to ±
platforms are measured independently, e.g., by optical 1 mm over a range of 2.5 m or less. When done
land leveling, inteferometric synthetic aperture radar repeatedly, the density changes can be used to infer
(InSAR), or GPS satellite measurements on land or by saturation changes in the depth interval. This has been
water pressure at the seafloor, the gravity data can be done successfully for producing gas reservoirs, with an
corrected for height changes, by using the appropriate estimate of residual gas saturation (e.g., Adams, 1991).
vertical gravity gradient. Such use of borehole gravity data “sees” much farther
A surface gravity monitoring program may be into the formation than most other logging tools, and is
designed as a survey of a grid with relative gravimeter- thus less vulnerable to hole breakouts, casing effects, etc.
(s), or with portable absolute meters, or a sparser grid Gravity gradiometry measures the spatial gravity
of stations with permanent superconducting gravi- gradients (in three spatial directions). If derived from
meters. Surveys using a combination of these types gravity measurements, the gradients offer no addi-
of instrumentation may be termed hybrid gravimetry. tional information or reduced noise level in the inter-
By measuring gravity in boreholes, the sensors get pretation stage. Some gradiometers can measure the
closer to the storage reservoir and thus stronger time- gravity gradients directly. So far, no surveying cap-
lapse signals can be observed. Recent advances have abilities have been reported that are comparable to or
made the instruments to work in smaller, more better than gravity sensors.
deviated boreholes (Ander and Chapin, 1997;
Chapin and Ander, 2000; Seigel et al., 2007; Nind CO2 Densities in the Subsurface
and MacQueen, 2013). Quoted measurement accu-
The phase and density of injected CO2 in the storage
racy range from 1 to 5 μGal. Borehole gravity data
reservoir will depend highly on temperature and pres-
will have a challenge with depth repeatability.
sure, as shown in the phase diagram (Figure 7.4).
The gravity gradient in a borehole is given by:
The critical point at 31.1°C and 7.39 MPa makes
Δg a supercritical state likely for depths exceeding 1 km.
¼ 3:086 0:838 ρ;
Δz CO2 density versus depth for three geothermal gradi-
in μGal=cm; when ρ is in kg=m3 ents are shown in Figure 7.5. The “cold case” gives
Seigel et al. (2007) estimated ± 5 cm depth uncer- sharp transition at about 500 m depth, while the war-
118
tainty, which transform to 5–7 μGal for rock densities mer situations have more gradual density transitions.
Figure 7.4 Phase diagram for CO2 (http://en.wikpedia.org/

wiki/Supercritical_fluid).
10,000
solid
1,000 supercritical
fluid
pressure
liquid
P (bar)
100
critical point
10
triple point gas
1
200 250 300 350 400
temperature
T (K)
CO2 will typically have high pressure at the

entrance of an injection well (wellhead).
The temperature may be the ambient surface (air or
water) temperature or higher, depending on the dis-
tance from the compression facility. As the CO2 gets
further compressed down the well by the weight of
CO2 above, the temperature will change (generally
increase) due to compression and heat transfer with
the surrounding rock (e.g., Singhe et al., 2013). For
instance, at Sleipner, the temperature has been esti-
mated to increase from 25°C at the injection point to
48°C at the perforation (Alnes et al., 2011). When
flowing into the reservoir, the CO2 will generally
take the formation temperature quickly, due to the
much greater heat capacity of the rock and formation
water.
If CO2 is pushing away and replacing formation
water in the reservoir, this storage mechanism may
involve only minor pressure changes. Neglecting pres-
sure changes, the density contrast can be described as
Δρ ¼ θ ðSw;i Sw;r Þ ðρw ρCO2 Þ ð7:1Þ
where
Δρ is rock density change
θ is porosity
Figure 7.5 Density of CO2 vs. depth shown as solid lines for three
different geothermal gradients, shown as dashed lines. The gradients
Sw;i is initial water saturation
are (1) 0°C + 25°C/km, (2) 15°C + 27.5°C/km, (3) 30°C + 30°C/km, and Sw;r is residual water saturation
shown as dashed lines. The effects of 10% methane impurities are ρw is water density
shown as fine dashed lines for the warmest and coldest cases. 119
ρCO2 is CO2 density
Ola Eiken
Figure 7.6 (Left) Point mass response to 1 Mtonne at various depths. (Right) Scaled to the same peak level.
In the case of CO2 pushing away water, both the reservoir. An example for a 10% mole fraction of
density and saturation changes may be significant, and methane is shown in Figure 7.5. This can reduce the
vary significantly from case to case. Saturation changes density to less than half in the most extreme cases.
will depend on the process of fluid replacement, and is
likely to be higher for high porosity and permeability. Overburden/Burial
Pressure increase during injection will cause the
The depth of the storage reservoir will not only influ-
storage formation to expand and create extra space.
ence pressure and temperature conditions and thus the
The pore volume increase may occur following linear
in situ density of CO2, but also the signal decay caused
elasticity:
by the distance between the reservoir and the observa-
V ¼ V0 ð1 þ cp ΔpÞ ð7:2Þ tion points on the surface. The scalar gravity change
from a mass change Δm is given by Newton’s law:
where
Δm
V is rock volume Δg ¼ G
d2
V0 is initial rock volume where
cp is pore compressibility
Δg is gravity change
Δp is pressure change
G is the gravitational constant; measured as 6.67408
Such expansion could occur as expansion of all × 10−11 m3 kg−1s−2
pores, or more heterogeneously, e.g., by a much larger d is the distance between the mass change and the
opening of a few fractures. In either case, the extra observation point
space will most often be filled with CO2, causing mass For a reservoir body of some extension, the surface
increase and gravity changes, as well as deformation gravity change can be found by volume integration of
of the reservoir and the surrounding rocks. each mass change in Newton’s formula. A gravimeter
The density of the formation water will generally be measures the scalar (vertical) gravity attraction. For
higher for higher salinity and pressure, and lower for small changes, therefore, only the vertical component
higher temperature. Borehole data on these will usually of change can be inferred. The measured gravity
provide sufficient information for the modeling. change, Δgz, then becomes
Most often the injected CO2 will have impurities,
Δm z
depending on the processes that have treated the CO2 Δgz ¼ G
prior to entering the wellhead. If the CO2 has been d3
separated from flue gas in a combustion plant or where z is depth.
separated from an initial mix with hydrocarbon An example of the point mass response of
gases, the impurities may be dominated by nitrogen 1 Mtonne CO2 at various depths is shown in
or methane. The volumetric fractions of such impu- Figure 7.6. The response will both have lower ampli-
120 rities can dramatically increase from the surface to the tude and longer wavelengths for deeper sources.
Gravity gradiometers represent another measure-

ment principle. Gravity gradients decrease faster than
gravity itself with increasing distance, by d13 , and there-
fore the strength of the change will decay more rapidly
with increasing reservoir depth. However, the shorter
wavelengths are relatively better preserved in a grid of
surface observations, and can thus provide higher
lateral resolution in data before inversion. After inver-
sion, the signal-to-noise ratio will determine resolu-
tion, and current gradiometers are mostly inferior to
gravimeters, except for very shallow reservoirs.
Surface Uplift/Height Changes

The reservoir pressure buildup associated with CO2
injection will cause mechanical deformations, usually
pore expansion. This will again cause deformations in Figure 7.7 Correlation between depth changes and gravity
the surrounding rocks, and likely uplift of the surface, changes for the outermost benchmarks of Sleipner 2005–2002.
The slope of the best fitting line is a gravity gradient of 180 μGal/m.
which may be measurable or too small to detect. Some (From Alnes et al., 2008.)
horizontal strain will also occur. Pore expansion will
create additional storage space, which in some cases
may be the primary storage mechanism, e.g., in tight be less significant offshore, as the seafloor topography
reservoirs (Kabiradeh and Sideris, 2016). In most situa- generally is much flatter than terrestrial landforms.
tions, except for very large and permeable formations Settlement of the benchmarks causes local height
with low injection rates, the pressure increase will be an changes, which may have another vertical gradient,
important parameter for managing the injection. dependent on the mechanisms of deformation. As an
Surface height changes may on land be monitored example, scouring around and beneath the benchmarks
by optical leveling, InSAR, or GPS. This is discussed in at Sleipner caused them to sink, with a gradient of 160
Chapter 15 by Don White in this volume. At sea, mGal/m estimated from the data (Figure 7.7).
water pressure surveys of stable seafloor benchmarks A significant uncertainty in the correction to the
can provide height change data with relative accuracy gravity data may arise from the subsidence measure-
of 2–4 mm (Stenvold et al., 2006), at water depths ments. Subsidence correction will often be
ranging from 80 to 1200 m and survey areas up to a significant part of the total gravity error budget.
1000 km2. For example, an uncertainty of 5 mm in height trans-
Height change (uplift or subsidence) of the sur- forms to 1.3–1.5 μGal in gravity. Some uncertainty
face, and thereby the gravity stations themselves, will also come from the vertical gradient, either the
cause a change in the observed gravity. The free-air land topography or the seawater density.
gravity gradient, and uplift or subsidence gradient, Surface elevation change is not only needed for
is on average about 308 μGal/m, but may vary gravity monitoring, but can, in itself, be an important
several tens of μGal/m across a survey area if it is monitoring tool. The size of the uplift will provide
influenced by the local topography (e.g., Hunt information about the size of the reservoir expansion
et al., 2002). When the seafloor is lifted up, there and thus the additional pore space available for sto-
will be less water above the observation point. rage. The aerial distribution of uplift can be used to
The influence on gravity will be a combination of infer the lateral distribution of pressure increase
the free-air gradient and the (opposite sign) and how it decays away from the injection well,
gravity attraction of a Bouguer plate of water, dependent on the reservoir geometry, barriers, etc.
Δg
z ¼ 2 π ρwater G, which on average reduces the
The In Salah injection project (Figure 7.8) is a good
vertical gradient to about 267 μGal/cm. Local varia- example of such an effect, and has been analyzed
tions in the gradient across a survey area are likely to (Ringrose et al., 2013). Another example is the
121
Ola Eiken
appear to have worked well. More advanced models

Map of surface uplift
may reveal further detail.
May 20mm
2009 uplift
Modeling Gravity Response for Various
Storage Mechanisms
+ KB-5
Time-lapse gravity changes are straightforward to cal-
+ KB-503 culate from flow models, based on the fluid saturations
+ KB-502
and pressures at each time step, as outputted from
reservoir simulators such as Eclipse (www.software.slb
.com/products/eclipse), TOUGH2 (Preuss, 1998; http://
esd1.lbl.gov/research/projects/tough/) or STAR
(Pritchett, 1995; www.cmgl.ca/stars). The temperature
Modelled rock strain (section)
may be included as a model parameter, as CO2 density
alters significantly with temperature. This would differ
from the common isothermal practice for modeling of
production from or injection into oil and gas reservoirs.
If flow models are not available, or not appropriate, less
Injection Unit
complex models, such as geometrical reservoir models
with sketched fluid front and pressure development,
may be used.
Figure 7.8 (Above) Surface uplift at the In Salah CO2 injection sites Forward modeling of scenarios spanning the
as inferred from InSAR data. (Below) Modeled rock strain. (From uncertainties will be the core of gravity feasibility
Ringrose et al., 2013.)
studies, which also should include sensor and survey
setup, expected accuracy, and interpretive aspects.
summer–winter cycling of a gas storage reservoir Different CO2 storage mechanisms will cause gravity
(Figure 7.9). anomalies with different characteristics. Mechanisms
Such inferences from surface data to reservoir may be divided into (1) aquifers, (2) depleted gas
require a working model for the overburden response fields, (3) overpressure space, and (4) enhanced oil
to deformations. Geomechanical modeling is recovery (EOR) projects.
a comprehensive topic of its own. Simple models as When injecting CO2 into large, usually saline,
the Geertsma (1973) or van Opstal (1974) assuming aquifers, pressure increases are likely to be modest.
homogeneous elastic properties are widely used and The main gravity effect will be replacement of water
Injected / extracted volume Surface displacement

15
10
DISPLACEMENT [mm]
–5
–10
–15
Summer Winter –20
–25
2004/01/01 2005/01/01 2006/01/01 2007/01/01 2008/01/01 2009/01/01
ACQUISITION [one acquisition each 30 days]
Figure 7.9 Summer injection and winter extraction of gas produce seasonal surface displacement. In the right graph, the displacement time
series obtained with SqueeSARTM shows a strong correlation with the injected/extracted gas volume. (From www.tre-altamira.com/oil-and-
122 gas/)
Figure 7.10 (Left) Map view of model of 1-km-side CO2 fronts. (Middle) Surface gravity response from the model. (Right) Density change by
inversion of the gravity data in the middle. (From Gasperikova and Hoversten, 2008.)
with CO2. The largest density changes will be at the causing a 4 μGal gravity change. To discriminate
CO2 front, and the dynamics of the front will be the between models, they concluded that a precision bet-
dominating effect to model and capture. Thermal ter than 0.1 μGal/year would be needed. Sugihara et al.
effects – heating or cooling of the injected CO2 when (2013) further discussed the feasibility of such detec-
filtering through the formation – may have tion using superconducting gravimeters (iGrav)
a secondary effect on the CO2 density. Dissolution of together with free-fall absolute calibrations, and con-
CO2 in the formation water may be a slow process, but cluded it would be challenging with currently avail-
ongoing for a long time after site closure. It will able instrumentation.
increase the density of the water and decrease the Jacob et al. (2016) modeled the injection of 18
volume of the CO2 plume, and thus cause a slight Mtonnes at 1800 m depth, developing a plume of
increase in gravity with time. Such information may 2600 m radius (Figure 7.11) and causing an up to 8
be valuable in a postinjection, site closure, phase. μGal gravity reduction. They inverted the plume out-
Three model studies of aquifer storage have line after 5 μGal random noise was added. They found
been published. Gasperikova and Hoversten (2008) generally that the shape could not be reliably mapped,
modeled the time-lapse gravity response for an but the total mass estimate was more reliable
advancing 1 km wide CO2 front in a 20 m thick (Figure 7.12).
brine formation with 20% porosity at a depth of For CO2 injection into depleted gas fields with
1900 m (Figure 7.10). With a maximum of 30% pressure much below initial or hydrostatic pressure,
CO2 saturation in the generated plume, the max- the CO2 is likely to mix with the remaining gas.
imum surface change was 10 μGal. The modeled If there is insignificant water inflow from aquifers,
data were inverted after 2.5 μGal random noise was the mass change will be equal to the injected CO2,
added. Inversion with the density variations con- and consequently the gravity change will be positive
strained to occur between the top and bottom (increase) and of a size roughly proportional to the
reservoir surfaces recovered the density changes injected mass. The point mass response shown in
within 30% of the true value. Gasperikova and Figure 7.6 is relevant for such situations. However,
Hoversten (2008) concluded that it is possible to the response must be scaled for the injected mass and
recover the general position of density changes adjusted for the reservoir geometry having a finite
caused by advancement of the CO2 front, but not extension. There might be less uncertainty and inter-
the absolute value of the change. They further est in the detailed spatial distribution of CO2 in
advocated gravity monitoring as a low-cost supple- a depleted gas field, as long as it is contained within
ment to or replacement of seismic monitoring. the depleted reservoir, and time-lapse gravity may
Sugihara et al. (2013) modeled the injection of 10 have more limited value than for some of the other
123
Mtonnes of CO2 into an aquifer at a depth of 2050 m, storage mechanisms.
Ola Eiken
Figure 7.11 (a) Model rock density changes after injection of 18 Mtonnes of CO2 at 1800 m. The black line represents the saturation front,
color scale is clipped. (b) Gravity attraction at the surface from the model densities. (From Jacob et al., 2016.)
a forward attraction, 5 μgal noise c Inverted densities (kg.m–3)

4000 4000
2000 2000
Northings (m)
Northings (m)
0 0
–2000
–2000
–4000
–4000
–4000 –2000 0 2000 4000 –4000 –2000 0 2000 4000
Eastings (m) Eastings (m)
–15 –10 –5 0 5 10
–20 –15 –10 –5 0 5
Figure 7.12 (Left) Gravity attraction at the surface with 5 μGal Gaussian noise added. (Right) Inverted densities for the L-curve method. Thick
black line: position of the plume in the forward model; white line: inverted plume position. (From Jacob et al., 2016.)
Sherlock et al. (2006) modeled the gravity change over 18 months, and pressure was assumed to be
above a depleted gas reservoir that was 29 m thick and hydrostatic throughout the injection period.
124 at a depth of about 2 km, with an active aquifer The injected CO2 partly mixed with the remaining
beneath. At the site, 100 000 tonnes were injected methane cap, partly created a plume with CO2
saturation >50% within the initial gas reservoir, and a maximum response of 4 μGal – on the edge of
pushed the gas–water contact downwards. The first detection. Thicker or shallower coal seams would
two effects increase reservoir density, while the latter clearly increase the signal.
reduces the density. The modeled surface gravity
changes summed up to a change of about 0.1 μGal, History Matching
much less than any realistic detection levels. A straightforward analysis of gravity monitor data is
Gravimeters in boreholes closely above the reservoir done by comparing with the modeled (predicted)
could detect signals of several μGals, but this would time-lapse changes. “History matching” is often the
likely be a cost-prohibitive survey (Sherlock et al., 2006). term used when flow models are adjusted to get
Gravity monitoring in the late gas extraction phase a better match to measured data. The traditional
and after gas depletion (postproduction) could give data used in the petroleum industry are well pressures
unique and valuable information about the natural and production rates/volumes. History matching may
water inflow into the reservoir when in a low-pressure also use gravity data, as has been demonstrated by
state. This can provide information about regional Glegola et al. (2012). They found that pressure data
aquifer strength and connectivity, which can be combined with gravity gave history matched models
important for basin-wide CO2 storage capacity. Such with less uncertainty in the initial gas mass and pro-
understanding would otherwise require costly duction prognosis than models that had been
acquired well pressure data, and be rare. matched to pressure only.
Petroleum reservoirs subject to CO2 injection for
EOR have potentially more complex flow patterns
than brine formations because they typically have Mass Change Estimates
reduced vertical extent and multiple in situ fluids The time-lapse version of Gauss’s theorem says that
(oil, hydrocarbon gas, brine, and CO2). Gasperikova the gravity change can be area-integrated to get the
and Hoversten (2006, 2008) modeled the gravity total mass change. This is irrespective of shape or
change of a 25–30 m thick reservoir at 1150–1350 m depth of the volume undergoing change (e.g., Goetz,
depth after 20 years when injecting in 22 wells. 1958; Campos-Enriquez et al., 1998), and the mass
Maximum gravity change, at the surface was on the change is
ð
order of 3 μGal, not much above current state-of-the-
ΔM ¼ 4πG1
Δgz dA
art noise levels, they concluded. A
In confined storage reservoirs, pore pressure will
increase during injection, and expansion of the for- where
mation will create extra storage space. The extra ΔM is the anomalous mass
injected CO2 will cause a positive gravity change at G is the gravitational constant
the surface, while the surface uplift will cause Δgz is the change in attraction of gravity
a negative change at the observation platforms. A is a surface that covers the surface gravity anomaly
Kabirzadeh et al. (2017) calculated the uplift effect to
Signal-to-noise ratios and spatial sampling will
easily dominate gravity for typical reservoir para-
limit the accuracy. For estimating mass changes
meters. When uplift is measured with sufficient pre-
from increasingly deeper reservoirs, an increasing
cision (as discussed earlier) this part of the gravity
survey area will be required.
signal can be corrected, and the reservoir changes can
A mass change estimate can be a valuable con-
be extracted. Real cases may be in between the end
straint in situations of CO2 injection and storage. This
cases of confined and unconfined reservoirs, and thus
may be of interest, for instance, when the injected
have less uplift and less gravity change. In such situa-
mass is known, and the mass flowing out of the reser-
tions, it will be important to monitor both gravity and
voir (area) can be calculated.
height changes.
Coal formations subject to methane production
and CO2 storage tend to be thin and shallow. Leakage Detection
Gasperikova and Hoversten (2008) modeled a 3 m An important purpose of gravity monitoring can be
thick zone at depths between 400 m and 700 m, with the detection of any leakage of CO2 upward from the 125
water in the fractures replaced by CO2, causing storage reservoir, toward the surface. The density of
Ola Eiken
PEAK GRAVITY CHANGE [MICROGAL] Figure 7.13 Peak gravity change for
a spherical CO2 plume (bubble) replacing
0.01 0.10 1.00 10.00 100.00 1000.00 saline formation water at various depths.
0
Assumptions are hydrostatic pressure,
200 geothermal gradient of 15°C + 27.5°C/km,
400 water salinity 3.5%. The relative size of the
600 CO2 bubble is shown for each depth.
800
1000
1200
1400
1600
DEPTH [M]
1800
2000
2200
2400
2600
2800
3000
3200
3400
3600
3800
4000
CO2 will decrease as pressure is lowered on the way Shallower sources of mass changes will require
upwards (Figure 7.5). This will be a particularly strong denser spatial sampling than the reservoir, in order
effect for depths shallower than 500–1000 m, dependent to keep an aerial coverage without holes. If predefined
on the temperature distribution. The density contrast areas with a higher risk of leakage, such as a mapped
and mass change when CO2 expels water will increase zone of fractures, can be identified, permanent gravi-
correspondingly. Greater proximity to the observation meters could be placed above for continuous moni-
points on the surface or seafloor will further increase the toring. For instance, superconducting gravimeters
sensitivity of gravity to small amounts of leaking CO2. could detect smaller and more abrupt changes than
The peak gravity response of a bubble (plume) of survey campaigns can.
100 ktonnes of CO2 that has replaced formation water
is shown in Figure 7.13. The size of each bubble is
illustrated, with a radius of 33 m at 4000 m depths, Sleipner
expanding to >100 m at 100 m depth. With a detection At the Sleipner CO2 storage site, seafloor micrograv-
level of 1–2 μGal, a 100 ktonne concentrated leakage ity surveys have so far been carried out in 2002,
could be detected at about 800 m depth, while 2005, 2009, and 2013, using the ROVDOG techni-
a superconducting meter with sensitivity 50 nGal que with relative spring meters (Sasagawa et al.,
could detect a sudden leak at 2.5 km depth. 2003; Zumberge et al., 2008). As described in
A leakage could occur along vertical cracks or Chapter 13 by Eiken (this volume), nearly 1
fissures with high permeability. Flow rates could be Mtonne of CO2 has been injected yearly since
high while only small amounts of CO2 are resident in 1996. The Utsira Fm. storage reservoir is situated
the overburden at a given time. This could be challen- 700–1000 m below the seafloor, at water depths of
ging to detect. On the other hand, some of the leaking 82–83 m. The initial purpose of the gravity moni-
CO2 may accumulate in pockets or strings of porous toring was to test the feasibility of observing a time-
rock, e.g., sand or silt, as is commonly found with lapse signal from the evolving CO2 plume. In 2005,
shallow gas pockets. Such shallow CO2 accumulations when the first gravity repeat measurements were
126
may have easily detectable time-lapse gravity signals. made, five time-lapse seismic surveys (Chadwick
vertical distribution of low- and high-saturation

zones of CO2 at scales below the seismic
wavelengths.
Benchmarks were deployed on the seafloor prior
to the first gravity survey, and serve as permanent
measurement platforms for measurements
(Figure 7.14). Stations were distributed in a main
profile across the CO2 plume, and with additional
stations giving an aerial cover of the plume
(Figure 7.15). In each survey, which lasted about one
week, three separate station visits were made with
three gravity sensors recording in parallel. After data
processing, estimated station time-lapse accuracies of
about 3 μGal were estimated (Alnes et al., 2011).
The seafloor benchmarks show relative height
Figure 7.14 Photograph of one of the seafloor benchmarks at changes exceeding 20 cm between surveys
Sleipner, in 2013, 11 years after the deployment. (Figure 7.16). The movements vary greatly between
neighboring stations, and can hardly be explained by
uplift caused by expansion of the storage reservoir or
et al., 2004, 2005) had provided rich information on subsidence caused by compaction of the gas reservoirs
the outline and internal stratification of the CO2 beneath. More likely, the benchmarks have been sub-
plume. However, the seismic data are limited at ject to scouring. This is supported by changes in
quantifying CO2 saturations and resolving the measured tilt of the top of the platform between
Figure 7.15 The layout of gravity stations, with

the rim of the CO2 plume as of 2008 in red, and
the rim of the Sleipner Øst Ty gas-condensate
reservoir in green. (From Alnes et al., 2011.)
127
Ola Eiken
a gravity reduction could be identified above the cen-

tral part of the CO2 plume (Nooner et al., 2007; Alnes
et al., 2008, 2011, 2015; Arts et al., 2008).
The estimated maximum gravity reduction increased
from about 10 μGal over 2002–2005 to more than 20
μGal over the time interval 2002–2013. The gravity
data have in various studies been used to invert for the
average CO2 density, dissolution rate of CO2 in the
brine and lately also the thickness of CO2. For such
inversions, the geometry of the CO2 plume as imaged
in 4D seismic data, and the injected amount of CO2,
are further constraining the models.
Figure 7.16 Depth changes 2002–2005. (From Alnes et al., 2008.) Both Nooner et al. (2007) and Alnes et al. (2008)
estimated the average CO2 density, in a joint inversion
with (1) the gravity height gradient and (2) a scale
factor for the Ty Fm gas reservoir flow model (to
surveys, by water-filled gaps between the bottom of adjust the water influx), as free parameters. Nooner
the benchmarks and the seafloor (Figure 7.14), and by et al. (2007) estimated an average CO2 density of 530 ±
shell fragments, possibly exposed after local erosion of 65 kg/m3 (95% confidence interval). Based on this,
the uppermost silt layer around the benchmarks. they argued that a relatively high geothermal gradient
After correcting the gravity for height changes, was required in the area to have most of the injected
using a vertical gravity gradient derived from the CO2 at such low densities. Alnes et al. (2008) esti-
data as shown in Figure 7.17, a time-lapse gravity mated a significantly higher average CO2 density: 760
increase of nearly 40 μGal was calculated at the wes- kg/m3, corresponding to a temperature of 29°C at
ternmost station, gradually decreasing to no change 870 m depth for pure CO2. This change in the estimate
over 2–3 km eastward. Alnes et al. (2008) explained based on the same data was explained by a series of
this strong increase at the end of the profile, well aside small changes: (1) reprocessing of the raw gravity
from the CO2 plume, as caused by water inflow to the data, (2) updated reservoir model, (3) new 4D seismic
Ty Fm. gas reservoir. The required water volumes data of the CO2 plume, and (4) a different treatment of
agreed with later acquired saturation logs in (unknown) dissolved CO2. Alnes et al. (2011) used the
a nearby well and with 4D seismic data. longer time span of 2002–2009 and reprocessed all
After corrections were made for both the time- gravity and depth data, and arrived at 720 ± 80 kg/m3
lapse height changes and water flow into Ty Fm, for average CO2 density. They used an additional
Figure 7.17 Estimated gravity changes caused

by CO2 injection, from 2002 to 2013 (from Alnes,
2015). Inverted CO2 thickness is shown as blue
colors.
128
similar conclusions as Alnes et al. (2011), as shown in

Figure 7.18. After the 2013 data were acquired, Alnes
(2015) updated the dissolution estimate to <2.7%
per year, based on the longer 2002–2013 time span.
Landrø and Zumberge (2017) combined time-
lapse seismic AVO data with gravity data to estimate
saturation changes also for values exceeding 0.3. They
claimed that a simple inversion procedure can be used.
Gravity Monitoring at Other Storage

Figure 7.18 Probability distribution of dissolution rate, based on
Sites
the 2002–2009 time span. (From Hauge and Kolbjørnsen, 2015.) The In Salah gas field started CO2 injection in 2004.
With depths of 1700 m below the surface, tempera-
tures of 80–100°C, and pressures between about
180 bar initially and up to 280 bar during injection
constraint on rock temperature from a well producing (Eiken et al., 2011), the CO2 is supercritical, with
water from the same formation about 10 km away, densities in the range of 500–700 kg/m3. However,
giving a geothermal gradient of porosities of 10–15% and reservoir thicknesses of
TðzÞ ¼ 31:7°C=km z þ 3:4°Cð 0:5°C=kmÞ only about 20 m cause the signal to hardly exceed
a few μGal. Elevation monitoring by InSAR has suc-
From this they estimated a dissolution rate of CO2 cessfully measured uplift to a few mm’s accuracy. No
in the formation water between zero and 1.8% per year. microgravity data have been acquired at the site.
Hauge and Kolbjørnsen (2015) performed a Bayesian Hydrocarbon gas has been injected in the Izaute
analysis of the in situ CO2 density and dissolution, gas storage reservoir during summer, and produced in
using the same 2002–2005–2009 data set, with the Ty the winter (Bate, 2005). Gravity measurements were
Fm and height change effects removed. They arrived at made over a period exceeding a year, with one
Figure 7.19 Diagram showing the

continuous time lapse stock pressure
time-lapse gravity (blue dots),
25 0 continuous gravity (blue trace), gas
stock levels (red trace), and reservoir
10 pressure (red dashed trace) at
15 a monitoring well over a 16-month
period (from Bate, 2005). The polarity
20
of the gravity data has been inverted
5 to facilitate easier comparison.
30
40
µgal –5 %
50
–15
60
–25 70
80
–35
90
–45 100
03 r0
3 03 03 t0
3 03 04 r0
4 04
Fe
b
Ap Ju
n
Au
g
O
c
De
c
Fe
b
Ap Ju
n 129
Ola Eiken
Table 7.3 Benefits and limitations of gravity monitor data

compared with other data References
Adams, S. J. (1991). Gas saturation monitoring in North
Benefit Limitation
Oman Reservoir using a borehole gravimeter. SPE
Direct and linear relation Detectability (need Journal, 21414: 669–678.
to mass change (better a minimum mass change, Allis, R. G., and Hunt, T. M. (1986). Analysis of
than the complicated rock which is depth exploitation-induced gravity changes at Wairakei
physics relations that dependent) Geothermal Field. Geophysics, 51: 1647–1660.
applies to 4D seismic data)
Alnes, H. (2015). Gravity surveys over time at Sleipner.
Areal coverage (better Limited lateral resolution Presentation at the 10th IEAGHG Monitoring Network
than wells) (depth and signal-to- Meeting, June 10–12, 2015. http://ieaghg.org/docs/
noise ratio dependent) General_Docs/8_Mon/6_Gravity_surveys_over_time_
at_SleipnerSEC.pdf
Alnes, H., Eiken, O., and Stenvold, T. (2008). Monitoring
permanent gravimeter installed near the expected gas production and CO2 injection at the Sleipner field
peak signal, as well as a grid that was surveyed about using time-lapse gravimetry. Geophysics, 73:
WA155–WA161.
once a month. The permanent station shows a clear
seasonal gravity signal that correlated well with gas Alnes, H., Stenvold, T., and Eiken, O. (2010). Experiences
injection/production and pressure fluctuations on seafloor gravimetrics and subsidence monitoring
above producing reservoirs. In Extended Abstract, 72nd
(Figure 7.19). As the reservoir is depleted, the gas/ EAGE Conference.
water contact rises, resulting in an increase in the
Alnes, H., Eiken, O., Nooner, S., Stenvold, T., and
time-varying gravity field.
Zumberge, M. A. (2011). Results from Sleipner gravity
monitoring: Updated density and temperature
Discussion and Summary distribution of the CO2 plume. Energy Procedia, 4:
Gravity’s unique strength compared to other moni- 5504–5511 (10th International Conference on
Greenhouse Gas Control Technologies).
toring techniques is the ability to measure mass
changes for a larger volume, directly with no trans- Ander, M. E., and Chapin, D. A. (1997). Borehole
formation except for Newton’s gravity constant. Such gravimetry: A review. In Extended Abstracts, 67th
Annual Society of Exploration Geophysicists Meeting,
data are thus straightforward to use in quantitative 531–534.
estimates and flow model matching. This could be of
Arts, R., Chadwick, A., Eiken, O., Thibeau, S., and
particular interest for CO2 storage, as the mass bal-
Nooner, S. (2008). Ten years’ experience of monitoring
ance will be a critical aspect of the injection and CO2 injection in the Utsira Sand at Sleipner, offshore
storage management. Gravity may be regarded as Norway. First Break, 26: 65–72.
a complementary rather than competing technique
Bate, D. (2005). 4D reservoir volumetrics: A case study over
to well time series (pressure and flow rates) and seis- the Izaute gas storage facility. First Break, 23: 69–71.
mic or electromagnetic techniques, giving different
Battaglia, M., Gottsmann, J., Carbone, D., and Fernández, J.
information. Some benefits and limitations, as com- (2008). 4D volcano gravimetry. Geophysics, 73(6):
pared with other reservoir monitoring data, are listed WA3–WA18.
in Table 7.3.
Brady, J. L., Hare, J. L., Ferguson, J. F., et al. (2008). Results
The rate of dissolution of CO2 into formation of the world’s first 4D microgravity surveillance of
water will be of high interest for decades- to centu- a Waterfloor – Prudhoe Bay, Alaska. SPE Journal,
ries-long storage performance, and can be difficult to 101762.
measure. Precise gravity changes offer unique data on Calvo, M., Hinderer, J., Rosat, S., et al. (2014). Time stability
this process. Microgravity data may also provide early of spring and superconducting gravimeters through the
warning for leakage to the surface, as the time-lapse analysis of very long gravity records. Journal of
gravity signal would quickly increase if CO2 should Geodynamics, 80: 20–33.
rise to shallower levels than the storage reservoir. Campos-Enriquez, J. O., Morales-Rodrigues, H. F.,
Mass changes will, regardless, be key to under- Domínguez-Mendez, F., and Birch, F. S. (1998). Gauss’s
standing the subsurface flow. theorem, mass deficiency at Chicxulub crater (Yucatan,
130
Mexico), and the extinction of the dinosaurs. Ferguson, J. F., Klopping, F. J., Chen, T., Seibert, J. E.,
Geophysics, 63(5): 1585–1594. Hare, J. L., and Brady, J. L. (2008). The 4D microgravity
Carbone, D., Poland, M. P., Diament, M., and Greco, F. method for waterflood surveillance: Part 3–4D absolute
(2017). The added value of time-variable microgravity surveys at Prudhoe Bay, Alaska.
microgravimetry to the understanding of how Geophysics, 73(6): WA163–WA171.
volcanoes work. Earth-Science Reviews, 169: 146–179. Furre, A.-K., Eiken, O., Alnes, H., Vevatne, J. N., and
Chadwick R. A., Arts, R., Eiken, O., Kirby, G. A., Kiær, A. F. (2017). 20 years of monitoring CO2
Lindeberg, E., and Zweigel, P. (2004). 4D seismic injection at Sleipner. Energy Procedia, 4: 5541–5548.
imaging of an injected CO2 plume at the Sleipner Field, Gasperikova, E., and Hoversten, G. M. (2006). A feasibility
central North Sea. In R. J. Cartwright, S. A. Stewart, study of nonseismic geophysical methods for
M. Lappin, and J. R. Underhill, (eds.), 3D seismic monitoring geologic CO2 sequestration. Leading Edge,
technology: Application to the exploration of sedimentary October: 1282–1288.
basins. Geological Society, London, Memoirs, 29: Gasperikova E., and Hoversten, G. M. (2008). Gravity
311–320. The Geological Society of London. monitoring of CO2 movement during sequestration:
Chadwick, R. A., Arts, R., and Eiken, O. (2005). 4D seismic Model studies. Geophysics, 73(6): WA105–WA112.
quantification of a growing CO2 plume at Sleipner, Geertsma, J. (1973). Land subsidence above compacting oil
North Sea. In A. G. Doré and B. A. Vining (eds.), and gas reservoirs. Journal of Petroleum Technology, 59
Petroleum geology: North-west Europe and global (6): 734–744.
perspectives: Proceedings of the 6th Petroleum Geology
Conference, 1385–1399. Gelderen, M. v, Haagmans, R., and Bilker, M. (1999).
Gravity changes and natural gas extraction in
Chadwick, R.A., Arts, R., Bentham, M., et al. (2009). Review Groningen. Geophysical Prospecting, 47: 979–993.
of monitoring issues and technologies associated with
the long-term underground storage of carbon dioxide. Glegola, M., Didmar, P., Hanea, R. G., et al. (2012). History
London: Geological Society, Special Publications, 313: matching time-lapse surface-gravity and well-pressure
257–275. data with ensemble smoother for estimating gas field
aquifer support: A 3D numerical study. SPE Journal,
Chapin, D. A., and Ander, M. E. (2000). Advances in deep- 161483.
penetration density logging. Society of Petroleum
Engineers Conference Papers, 59698. Goetz, J. F. (1958). A gravity investigation of a sulphide
deposit. Geophysics, 23(6): 606–623.
Christiansen, L., Lund, S., Andersen, O. B., Binning, P. J.,
Rosbjerg, D., and Bauer-Gottwein, P. (2011). Hare, J. L., Ferguson, J. F., and Brady, J. L. (2008). The 4D
Measuring gravity change caused by water storage microgravity method for waterflood surveillance: Part
variations: Performance assessment under controlled IV – Modeling and interpretation of early epoch 4D
conditions. Journal of Hydrology, 402: 60–70. gravity surveys at Prudhoe Bay, Alaska. Geophysics, 73
(6): WA173–WA180.
Debeglia, N., and Dupont F. (2002). Some critical factors for
engineering and environmental microgravity Hauge, V. L., and Kobjørnsen, O. (2015). Bayesian inversion
investigations. Journal of Applied Geophysics, 50: 435–454. of gravimetric data and assessment of CO2 dissolution
in the Utsira Formation. Interpretation, sp1–sp10.
Dodds, K., Krahenbuhl, R., Reitz, A., Li, Y., and Hovorka, S.
(2013). Evaluating time-lapse borehole gravity for CO2 Hunt, T. M., and Kissling, W. M. (1994). Determination of
plume detection at SECARB Cranfield. International reservoir properties at Wairakei Geothermal Field
Journal of Greenhouse Gas Control, 18: 421–429. using gravity change measurements. Journal of
Volcanology and Geothermal Research, 63: 129–143.
Eiken, O., Stenvold, T., Zumberge, M., Alnes, H., and
Sasagawa, G. (2008). Gravimetric monitoring of gas Hunt, T., Sugihara, M., Sato, T., and Takemura, T. (2002).
production from the Troll field. Geophysics, 73: Measurement and use of the vertical gravity gradient in
WA149–WA154. correcting repeat microgravity measurements for the
effects of ground subsidence in geothermal systems.
Eiken, O., Ringrose, P., Hermanrud, C., Nazarian, B., Geothermics, 31: 524–543.
Torp, T. A., and Høier, L. (2011). Lessons learned from
14 years of CCS Operations: Sleipner, In Salah and Jacob, T., Bayer, R., Chery, J., and Le Moigne, N. (2010).
Snøhvit. Energy Procedia, 4: 5541–5548. Time-lapse microgravity surveys reveal water storage
heterogeneity of a karst aquifer. Journal of Geophysical
Eiken, O., Glegola, M., Liu, S., and Zumberge, M. A. (2017). Research, 115: B06402.
Four decades of gravity monitoring of the Groningen
Gas Field. Extended Abstract, First EAGE Workshop Jacob, T., Rohmer, J., and Manceau, J.-C. (2016). Using
on Practical Reservoir Monitoring. surface and borehole time-lapse gravity to monitor CO2 131
Ola Eiken
in saline aquifers: A numerical feasibility study. Sasagawa, G., Crawford, W., Eiken, O., Nooner, S. L.,
Greenhouse Gas Science and Technology, 6: 34–54. Stenvold, T., and Zumberge, M. A. (2003). A new
Kabirzadeh, H., Kim, J. W., and Sideris, M. G. (2017). sea-floor gravimeter. Geophysics, 68(2): 544–553.
Micro-gravimetric monitoring of geological CO2 Seigel, H. O., Nind, C., Lachapelle, R., Choteau, M., and
reservoirs. International Journal of Greenhouse Gas Giroux, B. (2007). Development of a borehole gravity
Control, 56: 187–193. meter for mining applications. In B. Milkereit (ed.),
Kabirzadeh, H., Sideris, M. G., Shin, Y. J., and Kim, J. W. Proceedings of Exploration 07: Fifth Decennial
(2017). Gravimetric monitoring of confined and International Conference on Mineral Exploration,
unconfined geological CO2 reservoirs. Energy Procedia, 21143–21147.
114: 3961–3968. Sherlock, D., Toomey, A., Hoversten, M., Gasperikova, E.,
Kim, J. W., Neumeyer, J., Kao, R., and Kabirzadeh, H. and Dodds, K. (2006). Gravity monitoring of CO2
(2015). Mass balance monitoring of geological CO2 storage in a depleted gas field: A sensitivity study.
storage with a superconducting gravimeter: A case Exploration Geophysics, 37: 37–43.
study. Journal of Applied Geophysics, 114: 244–250. Singhe, A. T., Ursin, J. R., Pusch, G., and Ganzer, L. (2013).
Landrø, M., and Zumberge, M. (2017). Estimating Modeling of temperature effects in CO2 injection wells.
saturation and density changes caused by CO2 injection Energy Procedia, 37: 3927–3935.
at Sleipner: Using time-lapse seismic amplitude- Sofyan, Y., Kamah, Y., Fujimitsy, Y., Ehara, S., Fukuda, Y.,
variation-with-offset and time-lapse gravity. and Taniguchi, M. (2011). Mass variation in outcome to
Interpretation, T243–T257. high productionactivity in Kamojang Geothermal
Lien, M., Agersborg, R., Hille, L. T., Lindgård, J. E., Ruiz, H., Field, Indonesia: A reservoir monitoring with relative
and Vatshelle, M. (2017). How 4D gravity and and absolute gravimetry. Earth Planets and Space, 63:
subsidence monitoring provide improved decision 1157–1167.
making at a lower cost. Extended Abstract, First EAGE Stenvold, T., Eiken, O., Zumberge, M. A., Sasagawa, G. S.,
Workshop on Practical Reservoir Monitoring. and Nooner, S. L. (2006). High-precision relative depth
Nind, C. J. M., and MacQueen, J. D. (2013). The borehole and subsidence mapping from seafloor water-pressure
gravity meter: Development and Results: 10th Biennial measurements. SPE Journal, 11(3): 380–389.
International Conference & Exhibition. Sugihara, M., and Ishido, T. (2008). Geothermal reservoir
Nooner, S. L. (2005). Gravity changes associated with monitoring with a combination of absolute and relative
underground injection of CO2 at the Sleipner storage gravimetry. Geophysics, 73(6): WA37–WA47.
reservoir in the North Sea, and other marine geodetic Sugihara, M., Nawa, K., Nishi, Y., Ishido, T., and Soma, N.
studies. PhD thesis, University of California, San Diego. (2013). Continuous gravity monitoring for CO2
Nooner, S. L., Eiken, O., Hermanrud, C., Sasagawa, G. S., geo-sequestration. Energy Procedia, 37: 4302–4309.
Stenvold, T., and Zumberge, M. A. (2007). Constraints Torge, W. (1989). Gravimetry. Berlin: Walter de Gruyter.
on the in situ density of CO2 within the Utsira Van den Beukel, A. (2014). Integrated reservoir monitoring
formation from time-lapse seafloor gravity of the Ormen Lange field: Time lapse seismic, time
measurements. International Journal of Greenhouse Gas lapse gravity and seafloor deformation monitoring.
Control, 1: 198–214. The Biennial Geophysical Seminar, NPF, Kristiansand.
Nordquist, G., Protacio, J. A., and Acuna, J. A. (2004). Van Opstal, G. H. C. (1974). The effect of base-rock rigidity
Precision gravity monitoring of the Bulalo geothermal on subsidence due to reservoir compaction.
field, Philippines: Independent checks and constraints In Proceedings of the 3rd Congress of the International
on numerical simulation. Geothermics, 33: 37–56. Society for Rock Mechanics, Denver, II, Part B,
Preuss, K. ed. (1998). Proceedings of the TOUGH Workshop 1102–1111.
´98, Berkeley, California, May 4 –6, 1998. Lawrence Vevatne, J. N., Alnes, H., Eiken, O., Stenvold, T., and
Berkeley National Laboratory report LBNL-41995. Vassenden, F. (2012). Use of field-wide seafloor time-
Pritchett, J. W., and Garg, S. K. (1995). STAR: A geothermal lapse gravity in history matching the Mikkel gas
reservoir simulation system: Proceedings of the World condensate field. Extended Abstract, 74th EAGE
Geothermal Congress 1995, Florence, Italy, May 18–31, Conference.
International Geothermal Association, 2959–2963. Wilson, C. R., Scanlon, B., Sharp, J., Longuevergne, L., and
Ringrose, P. S., Mathieson, A. S., Wright, I. W., et al. (2013). Wu, H. (2012). Field test of the superconducting
The In Salah CO2 storage project: Lessons learned and gravimeter as a hydrologic sensor. Ground Water, 50
knowledge transfer. Energy Procedia, 37: 6226–6236. (3): 442–449.
132
Yin, Q., Krahenbuhl, R., Li, Y., Wagner, S., and Brady, J. Zumberge, M., Alnes, H., Eiken, O., Sasagawa, G.,
(2016). Time-lapse gravity data at Prudhoe Bay: New and Stenvold, T. (2008). Precision of seafloor
understanding through integration with reservoir gravity and pressure measurements for
simulation models. Expanded Abstract, Society of reservoir monitoring. Geophysics, 73(6):
Exploration Geophysicists Annual Meeting. WA133–WA141.
133
Chapter
Estimating Saturation and Density Changes
8 Caused by CO2 Injection at Sleipner

Martin Landrø and Mark Zumberge
Introduction are undertaken. For time-lapse seismic data, when oil

To estimate rock density directly from seismic data is or gas is replaced with water, for example, there is
both challenging and associated with high uncertain- a linear relationship between the saturation change
ties. Prestack inversion methods exploiting amplitude and the density change. Therefore, for 4D seismic data
versus offset (AVO) information have been tested to one can say that density estimation is to some extent
estimate P-wave velocity, S-wave velocity, and density equivalent to estimating saturation changes. There are
simultaneously by several researchers. Helgesen and several examples of saturation estimation from time-
Landrø (1993) formulated an AVO inversion scheme lapse seismic data (see, e.g., Landrø, 1999, 2001; Tura
where P- and S-wave velocities, densities, and layer and Lumley, 1999; Trani et al., 2011; Grude et al.,
thicknesses were inverted for. Buland et al. (1996) 2013; Bhakta and Landrø, 2014). If the saturation is
used a nonlinear AVO inversion in the tau-p domain not uniform (patchy), then the rock physics relations
to estimate these three parameters at the Troll Field, between saturation and seismic parameters should be
offshore Norway. Leiceaga et al. (2010) used multi- changed (see, e.g., Grude et al., 2013).
component data to reduce the uncertainty related to In the work presented here our purpose is to
inversion of density. Bai and Yingst (2014) tested full combine saturation estimates from time-lapse seismic
waveform inversion (FWI) on synthetic data and data with repeated gravity measurements in order to
found that it is challenging to avoid cross-talk reduce the large uncertainty related to time-lapse esti-
between velocity and density. Roy et al. (2006) mated density changes. For this purpose we will use
explored the possibility of using wide-angle data for data acquired at the Sleipner carbon dioxide (CO2)
density inversion and found that by incorporating storage site. Since Statoil launched this project in
incidence angles from 40 to 55 degrees, more stable 1996, more than 150 papers have been published on
and reliable density estimates can be achieved. various topics attached to it. Several seismic surveys
However, such data must be processed with care and and complementary data have been made available to
often involve anisotropic imaging and wavelet stretch researchers, and therefore it is expected that a variety
corrections. Despite the fact that there have been of analyses will continue to grow in the future. Arts
significant research efforts on how to estimate density et al. (2008) published an overview of the seismic
reliably from seismic data, it is still considered diffi- results achieved after 10 years of CO2 injection,
cult (see, e.g., Roy et al., 2008) and much more where they discuss combined use of time-lapse seis-
unstable compared to impedance inversion. mic and gravity. One major difference between our
For time-lapse seismic data we find more or less paper and Arts et al. (2008) is that we derive explicit
the same challenges; however, the possibility of expressions for saturation changes using time-lapse
extracting density information might be somewhat seismic AVO data as input. Furthermore, we con-
better in cases in which high-quality repeated surveys strain these estimates by using gravity measurements
in the inversion procedure. Queisser and Singh (2013)
used a 2D full waveform inversion scheme to invert
This chapter was published in Interpretation, 5(2) (May for P-wave velocity using prestack time-lapse seismic
2017), 1–15. DOI: 10.1190/INT-2016–0120.1. It is reprinted data from 1994 and 1999. Rabben and Ursin (2011)
here with the permission of The Society of Exploration used the 2001 seismic data and performed an ampli-
134 Geophysicists and American Association of Petroleum
Geologists.
tude variation with angle (AVA) inversion for the top
Utsira formation and estimated P- and S-wave impe- Table 8.1 An overview of available offset stacks
dances as well as density. They clearly show the poten- Data type 150–450 m 600–1050 m 1200–1650 m
tial for using amplitude information to distinguish offset stack offset stack offset stack
between acoustic impedance and density at Sleipner. NEAR FAR ULTRA FAR
Evensen and Landrø (2010) used a time-lapse tomo-
2001 3D Used Used Not used
graphic inversion method and seismic data sets from
seismic
1994 and 2001 to estimate P-wave velocity in a thin
CO2 layer. 2008 3D Used Used Not used
seismic
In 1996 Statoil and Scripps Institute of
Oceanography embarked on a project to develop Those that were actually used in this study are marked by green.
high-precision seafloor gravimeters focusing on accu- Note that what we refer to as far offset here (corresponding to an
rate measurements of density changes caused by average reflection angle of approx. 30 degrees) does not repre-
sent the offset stack including the largest offsets.
hydrocarbon production in a reservoir or storage of
CO2 in the subsurface (Sasagawa et al., 2003; Zumberge
et al., 2008). Here we combine the gravimetric results
simple ray tracing for the near offset stack (150–450 m)
published in Alnes et al. (2011) with seismic estimation
is approx. 11 degrees, and the corresponding angle for
of saturation changes between 2001 and 2008.
our far offset data (600–1050 m) is approx. 30 degrees.
Another way to constrain and help the interpreta-
A moderate Q-filter assuming a constant Q-value of
tion of 4D seismic data acquired above CO2 storage sites
300 was used in the processing of the seismic data.
is to use fluid flow simulation techniques. One of the
The remaining amplitude compensation was done con-
early published simulation results was published by
sistently for the 2001 and 2008 seismic data sets.
Lindeberg and Bergmo (2003), and a more recent exam-
We will therefore assume that apart from an additional
ple can be found in Cavanagh and Haszeldine (2014).
Q-compensation that we introduce, there should be
This chapter is organized as follows. First we pre-
only one global scalar necessary to convert the seismic
sent the input data, both time-lapse seismic and grav-
data into “true” reflectivity. As we use a root mean
ity data. An overview of the basic assumptions made
square (RMS) window technique to extract reflection
in the chapter is given in the next section. A simple
amplitudes, this scalar will vary with the length of the
rock physics model based on earlier work is presented
window we are using, and hence the global scalar will
as a basis for the analysis work. Then we derive
increase as the length of the RMS window increases.
a simple formula relating near and far offset time
The injection of CO2 at Sleipner started in 1999.
lapse changes directly to saturation changes, based
This means that the 2001 data set that we are using as
on the rock physics model. The calibration procedure
a baseline is not a true baseline survey because some
used to couple seismic amplitude changes directly to
changes had already occurred when the 2001 survey
saturation changes is described in a separate section.
was acquired. Hence, we have to estimate roughly the
A key section in the chapter is the formulation of
extent and saturation in 2001 based on the 2001 data
a simple inversion problem to combine measured
only, and this is done by assuming that the amplitude
time-lapse seismic data with the measured gravity
anomaly observed close to the injection well on the
anomalies. Before we conclude we discuss various
2001 seismic data is caused mainly by the CO2 injec-
limitations and precautions for the presented work.
tion. A better choice would of course be to use data
from 1994 (which unfortunately were not available to
The Input Data us during this project), and use the time-lapse seismic
The input seismic data used in this project are listed in difference between 2001 and 1994 to estimate the
Table 8.1. Note that what we refer to here as the far offset saturation distribution in 2001.
stack (450–1050 m) is not the stack containing the largest Time-lapse gravity data from 2002 and 2009 are
offsets. Offsets from 1200 to 1650 m were collected but also used, and here we have simply used the same data
are not used owing to significantly lower 4D repeatability as those presented by Alnes et al. (2011). The seismic
for this stack. In this project our aim is to study 4D data and the gravity data cover approximately the
changes in AVO, and hence we need to use data that same number of years (seven); however, the gravity
are highly repeatable. The average incidence angle using data have a delay of one year compared to the seismic 135
data. We will assume that this delay is zero, and that averaging. This is a rough approximation and we
the seismic and gravity time-lapse data were acquired regard it as the lowest order approximation to the
simultaneously. It is hard to assess or quantify the time-lapse seismic AVO tuning problem. This
error caused by this assumption. issue is discussed in more detail in Appendix B.
• We assume that the average reservoir thickness, or
Some Basic Assumptions the thickness of the Utsira sand layer, is approx.
As this is a combined methodology and case study, we 200 m. From the seismic data we estimate that the
have to make several assumptions. Some of these top reservoir varies by approx. 15 m, and that the
assumptions are based on scientific considerations base has somewhat larger variations over the area
and others are used in order to simplify the case where the CO2 plume occurs.
study. In the following we list such assumptions and • We assume that saturation effects dominate over
add some comments to each item: temperature and pore pressure on the time-lapse
seismic data. This is motivated mainly by the
• We assume that the empirical relations given by
dramatic decrease in P-wave velocity caused by
Span and Wagner for CO2 at various pressures and
a relatively small change in CO2 saturation, and
temperatures are valid. The initial reservoir
the fact that the pressure and temperature changes
temperature at the injection point is measured to
are moderate in the Sleipner CO2 project.
35.5°C (Alnes et al., 2011). The initial pore pressure
is hydrostatic, which means that it is 80 bar at
800 m depth, and hence it is reasonable to assume A Simple Rock Physics Model for CO2
that the injected CO2 is most likely supercritical, as Injection at Sleipner
the critical point is at 31°C and 74 bar. (For more We will assume that the CO2 that is injected into the
discussion on the temperature of injectant see Utsira sand layer at Sleipner is supercritical and does not
Chapter 13 by Eiken in this volume.) mix with the brine water that occupies the pore space
• We assume that when supercritical CO2 is injected prior to injection. In a P–T phase diagram, the typical
into the brine-filled sandstone rock, the two pore pressure in the Utsira formation is above 80 bar
liquids are immiscible, and that the CO2 pushes and the temperature is most likely above 30°C, which
the water away from the injection point. leads to a supercritical state of CO2. This is discussed by
• A calibration procedure is needed to convert seismic Alnes et al. (2011), who mentioned that the injected CO2
amplitudes into reflection coefficients that are used at Sleipner is close to the critical point. Furthermore, we
for the AVO inversion. Our choice is to use one assume that the porosity is constant and equal to 37%.
global scalar for this purpose. Ghaderi and Landrø Ghaderi and Landrø (2009) used the empirical relations
(2009) found a near offset reflection coefficient of – obtained by Span and Wagner (1996) to estimate the
0.06 for the top sand layer outside the plume. This bulk modulus and density of the injected CO2. It is
value was used to determine this global scalar (again important to notice that the density is strongly depen-
using seismic data outside the CO2 plume) to 0.02. dent on both temperature and pressure at the storage
• We use a simple constant Q-model to account for site. For example, a temperature increase from 30° to 40°
amplitude variations between near and far offset C reduces the CO2 density from 680 kg/m3 to 300 kg/m3,
stacks (we found that Q = 80 was a good choice). assuming that the pore pressure is 80 bar. Ghaderi and
This is explained in Appendix A. Landrø (2009) therefore present two curves (Figure 8.4
• Because we are using seismic data from 2001 and in their paper) representing the P-wave velocity change
2008, we need an estimate for the saturation change as a function of CO2 saturation. In this chapter we use
from 1996 (injection start) to 2001. We have a simple exponential decay curve to model how the
assumed that the average CO2 saturation was 0.1 in P-wave velocity (α) change with CO2 saturation (S)
2001. 3 Mtonnes of CO2 has been injected into the within the Utsira sand (using a porosity value of 37%):
reservoir in the period between 1996 and 2001.
α ¼ α1 þ α2 eκS ð8:1Þ
• We assume that the amplitude versus offset
response of many thin CO2 layers can be Here α1 ¼ 1437m=s, α2 ¼ 613m=s and κ ¼ 10 was
136 approximated by using one thick layer and Backus found by simple trial-and-error curve fitting
Figure 8.1 P-wave velocity versus CO2

saturation using Eq. (8.1).
Figure 8.2 Rock density versus CO2

saturation assuming that the density of CO2
is 675 kg/m3 (solid line) and 425 kg/m3
(dashed line).
compared to Figure 8.4 in Ghaderi and Landrø (2009). fluid pressure is then 105 bar). Alnes et al. (2011)
This empirical formula represents an average between estimated the well-bottom CO2 temperature to be
the two curves used in Ghaderi and Landrø (2009). 48°C and using a fluid pressure of 105 bar they
The important feature of Eq. (8.1) is that it captures found a density of 485 kg/m3. At top Utsira the corre-
the steep decrease in velocity when the CO2 saturation sponding density value is as low as 425 kg/m3.
increases from 0 to 0.1, as shown in Figure 8.1. However, they concluded that the average CO2 den-
Assuming nonmiscible mixture of CO2 and water in sity within the Utsira formation is 675 kg/m3. We will
the pore space the fluid density (ρF ) is a linear combi- therefore use this value in most of our calculations
nation of the two phases: and the 425-value as an extreme low-density value.
ρF ¼ SρCO2 þ ð1 SÞρW The density of the rock including the fluid in the
ð8:2Þ
pore space is given as
where W denotes water, which we assume has
ρ ¼ ϕρF þ ð1 ϕÞρS ð8:3Þ
a density equal to 1050 kg/m3. Alnes et al. (2011)
discussed the ranges for densities within the Utsira where ϕ ¼ 0:37 is porosity, and ρS ¼ 2650 kg=m3 is
formation. The initial reservoir temperature is 35.5°C the density of the rock matrix or in this case the
137
at the injection point at 1050 m (the corresponding density of the quartz minerals. Density versus CO2
Figure 8.3 Relative change (change

divided by original value) in P-wave velo-
city (solid line) and density assuming CO2
density of 675 kg/m3 (dashed line) and
density assuming CO2 density of 425 kg/
m3 (dotted line) versus CO2 saturation.
A relative velocity change of –0.2
corresponds to a saturation change of 0.11,
as shown by the dashed-dotted line.
The two double arrows indicate which
saturation bands we use for time-lapse
seismic and gravity data, respectively, in
this chapter.
saturation is shown in Figure 8.2 for the two end commonly accepted that it is robust and stable to
members of 675 and 435 kg/m3 of CO2 density. estimate seismic impedance. The same is the case for
The expected relative changes in P-wave velocity and time-lapse seismic data: There are few examples
density are summarized in Figure 8.3. We notice that where density changes have been estimated, and in
the relative velocity changes are large for saturation most cases this means estimating saturation changes
changes between 0 and 0.4. The relative density and then deriving density directly from the estimated
changes are small (less than 0.05) in most cases, and saturation changes is used. Landrø (2001) formulated
hence it is more appropriate to estimate velocity a direct inversion method using time-lapse AVO data
changes for the Sleipner CO2 case. We will therefore to estimate pressure and saturation changes in
estimate density changes caused by CO2 injection by a producing hydrocarbon reservoir, or alternatively
first estimating the saturation changes, and then an injection site for CO2. Using Eq. (8.7) in Landrø
estimate the density change by combining Eqs. (8.2) (2001) we find that the change in P-wave reflection
and (8.3): coefficient (DR) when the pore fluid saturation
changes, is given as
Dρ ¼ ϕDSðρCO2 ρW Þ ð8:4Þ
1 Dρ Dα Dα
Assuming that the density difference between CO2 DR ¼ þ þ tan 2 θ ð8:5Þ
and water is not varying spatially, we see that the 2 ρ α 2α
density change is directly proportional to the satura- where θ is the incidence angle, α the P-wave velocity, ρ
tion change. However, it is very likely (see discussion the density, and Δ represents time-lapse changes in
by Alnes et al., 2011) that the density of CO2 increases the parameters. We will use this to directly invert for
away from the injection point because the injected density changes. In the following sections we will
CO2 is gradually cooled by the surrounding rock as assume that the top reservoir has a clear amplitude
it propagates away from the injection point. For sim- increase when shale is overlaying CO2-filled sand-
plicity, we will assume that this effect is second order stone rock compared to when the same rock is water
and assume that observed density changes from time- filled. Next, we will assume that the time-lapse Δ in
lapse gravity can be directly compared to estimated Eq. (8.5) represents the difference between water-
saturation changes from time-lapse seismic data. filled and CO2-filled rock. This means that it is
straightforward to estimate changes in the near and
A Simple Method to Estimate Density far offset stacks (DN and DF) between the two regions.
We will further assume that these stacks have been
Changes calibrated to modeled reflection coefficients (as
It is inherently difficult to estimate density directly described in the section “Calibrating time lapse
from seismic data. In seismic inversion, it is AVO data”) so that they represent reflection
138
Figure 8.4 Modeling reflectivity

changes using Zoeppritz equation
(solid line), Eq. (8.5) (dotted line) and Eq.
(8.10) (dashed line). Input parameters
are listed in the first column in
Table 8.2.
coefficients. The time-lapse amplitude changes at near P-wave velocity effect is the dominant effect, an
and far offset are given (where θN and θF are the near acoustic approximation will be sufficient.
and far offset angles): Assuming that we can use an acoustic approxima-
tion for the Zoeppritz equation we get
1 Dρ Dα Dα
DN ¼ þ þ tan 2 θN ð8:6Þ
2 ρ α 2α 1 Dρ Dα Dα
DR ¼ þ þ sin 2 θ: ð8:10Þ
2 ρ α 2α
1 Dρ Dα Dα
and DF ¼ þ þ tan 2 θF : ð8:7Þ As Eqs. (8.10) and (8.5) differ only in the angle-
2 ρ α 2α
dependent term, we simply replace tangent by sine in
Subtracting Eqs. (8.7) and (8.6) we find an explicit Eqs. (8.8) and (8.9) to obtain the alternative and more
expression for the change in P-wave velocity: accurate equations for the Sleipner CO2 case.
Dα 2ðDF DNÞ A comparison between Eqs. (8.5) and (8.10) to the
¼ : ð8:8Þ exact Zoeppritz equation is shown in Figure 8.4, using
α tan 2 θF tan 2 θN
realistic values from the Sleipner CO2 case. We see
Substituting this back into Eq. (8.6) again we find that Eq. (8.10) is more accurate in predicting the AVO
an expression for the relative density change: behavior for the far offsets. Both approximations
show a minor discrepancy of approx. 6–7% at zero
Dρ ðDF DNÞ
¼ 2DN 2 ð1 þ tan 2 θN Þ: offset, which is caused by the simplification of repla-
ρ tan 2 θF tan 2 θN cing the zero-offset reflection coefficient by the first
ð8:9Þ term on the right-hand side in Eq. (8.10) (or 8.5).
The nice feature of this equation is that it is
simple and represents a direct seismic estimate of Estimating CO2 Saturation Changes
the density contrast between water-filled and CO2- If we assume that the CO2 saturation of the time for
filled reservoir rock. One obvious problem is, how- the first seismic survey (A) is SA and the correspond-
ever, that the assumption behind Eqs. (8.8) and ing saturation for the second survey (B) is SB, we find
(8.9) is that all relative contrasts should be much from Eq. (8.1) that the change in P-wave velocity is
less than 1. In many cases these assumptions hold
true, and in such cases these equations are valid. Dα ¼ α1 þ α2 eκSB α1 þ α2 eκSA ð8:11Þ
It is also possible to include higher order terms.
However, for our CO2-injection case at Sleipner, and hence the relative velocity change is given by
the above equations are not sufficiently accurate. Dα α2 ðeκSB eκSA Þ
The relative P-wave velocity change is expected to ¼ : ð8:12Þ
α α1 þ α2 eκSA
be larger than 20% for the Sleipner CO2 example.
Therefore, we will use a slightly different version Combining this with Eq. (8.8) we find that the CO2
saturation at time B is given as
139
of Eq. (8.5), where we assume that because the

1 κSA ðDF DNÞðα1 þ α2 eκSA Þ 0.1
SB ¼ ln e þ2 4.5
κ α2 ðsin 2 θF sin 2 θN Þ
ð8:13Þ 0.08
The difference between the near and far offset 4

seismic stacks is measured, and the angle span is 0.06
known, so the most critical issue is to determine the
Y (km)
CO2 saturation for the first survey. Estimating CO2 3.5
saturation from time-lapse seismic data becomes 0.04
unstable for saturation values above 0.3–0.4 because
there are practically no velocity changes when the
saturation is increased from 0.3 to 1 (Figure 8.3). 3 0.02
Therefore, this equation must be used with great
care, and it definitely must be limited to saturation 0
values below 0.3. Furthermore, we observe that 1 1.5 2
the saturation change is directly dependent on the X (km)
difference between the time-lapse far and near offset
differences. This means that it is critical to calibrate Figure 8.5 RMS amplitudes for near offset stack from 2001 using
a 26 ms time window. The injection point is shown as a white circle
the near and far offset differences prior to using and the dark red anomaly extending both north and south to the
Eq. (8.13). For example, if the argument to the loga- injection point is caused by the CO2 plume. North of approx. 3.5 km
rithmic function in Eq. (8.13) is negative, it means we assume that there is no CO2 in 2001, and we will therefore
assume that the reflectivity above this line should represent the
that the calibration of the near and far offset reflectiv- untouched reservoir. The color scale (0–0.1) represents absolute
ity changes is poor. In practice, to avoid problems values of the reflection coefficient for the top Utsira interface.
related to negative arguments for the logarithm in
Eq. (8.13), we simply set SB equal to 0.4 if this occurs.
The value of 0.4 is chosen based on trial and error and untouched part of the reservoir, as shown in
is also related to the point where the P-wave variation Figure 8.5.
with CO2 saturation flattens. Furthermore, we Near and far offset stacks of the data for two cross-
observe from Figure 8.4 that there is a minor discre- lines are shown in Figures 8.6 and 8.7, where the first
pancy between Zoeppritz modeling and Eq. (8.10) cross-line intersects the CO2 injection point.
(which is used to derive Eq. 8.13) for near offsets. We notice that the CO2 effect on the seismic data is
This discrepancy can be reduced by introducing significant, both on near and far offset data. This is
higher-order terms in Eq. (8.10) (which then will consistent with the rock physics model and corre-
lead to a corresponding modification of Eq. 8.13). sponds well with the modeled velocity and density
changes displayed in Figures 8.1–8.3. The top of the
upper sand layer has a depth variation of the order of
Calibrating Time-Lapse AVO Data 15 ms within the plume mapped in 2008 (Figure 8.8).
In Eq. (8.13) it is assumed that both near and far offset In our analysis this depth variation has been included
seismic data have been calibrated so that they both in a very simple way by adding a stepwise start time
represent true reflectivity changes. In our calibration for the 26-ms time window used from north to south,
procedure we will first scale both the near offset and so that the added time for the southern areas is
far offset seismic data by one constant scalar. From approximately 15 ms. In this way we ensure that the
rock physics modeling and well information, Ghaderi RMS signal captures the main seismic reflection asso-
and Landrø (2009) found a near offset reflection coef- ciated with the top sand layer. We also tried to use the
ficient of –0.06 in a water-filled part of the top sand minimum and maximum values within the time win-
layer (outside the CO2 plume). We found that by dow, and found similar spatial maps when using the
applying a global scalar of 0.02, the estimated RMS RMS window, and we therefore decided to use RMS as
amplitudes using a 26 ms time window gives a zero- a way to estimate reflection coefficients from the near
140 offset reflection strength of approx. 0.06 for the and far offset data.
0.8 10 0.8 10
5 5
Time (s)
Time (s)
0.85 0.85
0 0
0.9 –5 0.9 –5
–10 –10
0 1 2 3 0 1 2 3
X (km) X (km)
0.8 10 0.8 10
5 5
Time (s)
Time (s)
0.85 0.85
0 0
0.9 –5 0.9 –5
–10 –10
0 1 2 3 0 1 2 3
X (km) X (km)
Figure 8.6 Near offset stacks (left) from 2001 (top) and 2008 (bottom), and corresponding far offset stacks (right) for xline 1120 (see Figure
8.10). This cross-line intersects the injection point at approx. 1.5 km. The color bar represents the original seismic amplitudes after processing
into near and far offset stacks. The location of xline 1120 is shown in Figure 8.10, and the direction is from west to east.
The next step is to find the relative scaling offset data) occur approx. 1.4 km northeast of the
between the near and far offset stacks; we use injection point. At this point (or area) we observe an
a constant Q-model (see Appendix A) for the over- RMS difference change of approx. –0.14
burden to correct for the extra geometric damping of (Figure 8.10). This near offset value (shown by
the far offset data. Prior to this calibration step we a black star in Figure 8.9) corresponds very well
modeled four different scenarios based on the simple with the modeled black solid line in Figure 8.9.
rock physics model discussed earlier. The four sce- This means that the global scalar of 0.02 is consistent
narios are given in Table 8.2, which include CO2 with the observed RMS near offset differences, given
saturation changes from 0 to 0.2, 0 to 0.5, 0.2 to 0.5, that the saturation has changed from zero to 0.2 in this
and finally 0.5 to 1.0. The reflection coefficients ver- area. The next step is to calibrate the far offset dif-
sus angle for the four cases are shown in Figure 8.9. ference data. From the modeled curve we observe
We notice that for the two latter scenarios, both the that the far offset RMS amplitude difference should
overall differences and the AVO differences are be somewhat larger in magnitude and close to –0.16.
small, as expected. The seismic data are not very By using a constant Q-model and assuming
sensitive to saturation changes above 0.4. The two a Q-value of 80 we obtain an RMS difference value
first scenarios on the contrary show significant that is in perfect agreement with the modeled curve
reflectivity differences at zero offset as well as in Figure 8.9. From Figure 8.10 we see the effect of
a slight AVO increase with offset. In 2001, the extent applying this Q-compensation to the far offset data:
of the plume for the uppermost CO2 layer was lim- The Q-compensated far offset difference data (to the
ited to the area close to the injection point right) are somewhat stronger in amplitude compared
(Figure 8.11). In 2008 the extent of the seismic to the uncompensated data (in the middle).
anomaly caused by CO2 injection has increased sig- In summary, we have now used one global scalar
nificantly (Figure 8.10) and followed the structurally (equal to 0.02) to convert the data from seismic
higher areas toward the north. We find that the amplitude values into reflection coefficient values,
largest changes in reflectivity (both for near and far and another calibration step to calibrate the far offset 141
Figure 8.7 Near offset stacks (left) from 2001 (top) and 2008 (bottom), and corresponding far offset stacks (right) for xline 1258 (see Figure
8.10). This is approx. 1 km north of the injection point. The CO2 anomaly is observed for 2008 at x = 1.9 km. The color bar represents the original
seismic amplitudes after processed into near and far offset stacks. The location of xline 1258 is shown in Figure 8.10, and the direction is from
west to east.
data by using a constant Q-model. For the calibration (which is 0.02 for a 26 ms long window) should
area (1.4 km northeast of the injection well) the be changed accordingly.
observed near and far offset amplitude changes cor-
responds to a saturation change from 0 to 0.2.
It should be noted that if we assume that the
Estimating the Saturation Change
saturation change is 0.5 instead of 0.2, a similar from 1996 to 2001
correction (using Q = 80) would still give In Eq. (8.13) we need the initial saturation distribu-
a reasonable fit to this curve (shown by the red tion within the plume (SA). As we do not have the
solid line in Figure 8.9). This is because the two 1996 offset stacks, we used a simplified method to get
curves for the two first scenarios are close to par- an estimate of this, by using the near offset stack from
allel. The estimated near offset time lapse difference 2001, and simply assuming that the strong anomaly
after application of the 0.02 correction is shown in observed close to the injection point is caused by the
Figure 8.10 together with uncorrected and CO2 injection. This is a reasonable assumption, as we
Q-compensated far offset time-lapse difference. know that the pressure changes are close to zero. One
A time window of 26 ms has been used to estimate remaining cause is temperature changes, because we
RMS amplitudes prior to the calibration steps. are injecting relatively warm CO2 into a somewhat
After these two calibration steps we are ready to cooler reservoir (35.5°C). We will assume that the
use Eq. (8.13) to estimate saturation changes seismic impacts of such temperature changes are
between 2001 and 2008. It should be noted that if minor.
the analysis is extended by using longer time win- Figure 8.11 shows near offset seismic data from
142 dows for the RMS calculation, the scaling factor 2001, and we notice a clear anomaly close to the
Figure 8.8 Top of Utsira time surface. Deep purple

colors in the south correspond to 872 ms and red colors
in the north correspond to 857 ms. (Figure provided by
Statoil.) The size of the figure is approx. 1.8 km (horizontal)
by 5 km (vertical).
injection well. We interpret this as being caused by the amplitude RMS values, we obtain the RMS map
CO2 injection, and more precisely by the CO2 satura- shown to the right in Figure 8.11. We observe that
tion change. If we subtract the background average the average RMS level is close to 0.1 after subtracting 143
this background amplitude value. In our case the this chapter that the areal extent of the uppermost
zero-offset reflection coefficient is approximately layer is approx. 0.23 km2 in 2001. Assuming
equal to Dα2α , where α is the P-wave velocity. This a homogeneous distribution of CO2 from the top
means that a reflection coefficient of –0.1 corresponds layer to the injection point, that is over a depth
to Dα
α ¼ 0:2. From Figure 8.3 we observe that this range of 250 m, yields an available volume Va =
value corresponds to a CO2 saturation of approxi- 0.058 km3. We know that approximately 3 Mtonnes
mately 0.1 (which means that we use SA = 0.1). of CO2 has been injected at Sleipner between 1996 and
An alternative way to estimate the 2001 CO2 2001, corresponding to a compressed volume VC =
saturation is to use earlier publications to obtain 0.0044 km3 assuming that the CO2 density is 675 kg/
a rough estimate. Using Arts et al. (2008), we observe m3. Dividing these two volumes, yields a very rough
from their Figure 8.6 that the seismic response in estimate for the average CO2 saturation of 0.08, which
2001 is distributed over approximately the same area is not too far from our estimate of 0.1.
also in depth, and that the reflectivity strength is
fairly constant with depth, approximately down to Estimating the Saturation Change
the injection point (which is at 1050 m). From
Kiær et al. (2015) we observe from Figure 8.1 in from 2001 to 2008
Using the estimated saturation change in 2001 as
input to Eq. (8.13), and using the near and far offset
differences between 2001 and 2008 as input to
Table 8.2 Relative changes in P-wave velocity and density for Eq. (8.13), we obtain the saturation in 2008 (SB).
the four time-lapse scenarios If we subtract SA (equal to 0.1) from this, we get the
Change in Relative Relative Relative saturation change between 2001 and 2008 as shown
CO2 P-wave density density in Figure 8.12. Again, this estimate is meant to
saturation velocity change change represent the uppermost sand layer (often referred
change assuming assuming to as layer 9; see, e.g., Furre and Eiken, 2014).
rhoCO2 = rhoCO2 = A smoothed version of the saturation change is
675 kg/m3 425 kg/m3 also shown in this figure. We notice saturation
0 to 0.2 −0.24 −0.014 −0.023 changes to the north of the injection point, but
0 to 0.5 −0.28 −0.034 −0.056
also fairly close to the injection point at the east
and south side.
0.2 to 0.5 −0.06 −0.021 −0.034
In order to combine seismic and gravity data, we
0.5 to 1.0 −0.02 −0.035 −0.060 need to find a way to estimate saturation changes for
0 0.5–1.0
Figure 8.9 Zoeppritz AVO curves for
four CO2 saturation change scenarios:
0–0.2 (black line); 0–0.5 (red line);
–0.05 0.2–0.5 (blue line), and 0.5–1.0 (green
Reflectivity difference
0.2–0.5 line). Calibration of far offset stack

data was obtained by using the black
–0.1 curve: First the near and far offset data
0–0.2
were multiplied with 0.02 based on
–0.15 the near offset reflectivity outside the
plume. Notice that the near offset
difference data point shown by a star
–0.2 at 10 degrees is very close to the first
0–0.5
scenario. Then the far offset data
(shown by the second star at 30
–0.25 degrees) was scaled by using
0 5 10 15 20 25 30 35 40
a constant Q-value of 80, so that the
Incidence angle (degrees) far offset difference follow the
modeled trend predicted by the rock
144 physics curve.
Near 2001–2008 Far 2001–2008 Q=0 Far 2001–2008 Q=80

6 6 6 0.2
5 5 5
0.15
XL 1258
Y (km)
4 4 4
0.1
3 XL 1120 3 3
0.05
2 2 2
Injection point
0
0 1 2 3 0 1 2 3 0 1 2 3
X (km) X (km) X (km)
Figure 8.10 RMS amplitude (26 ms time window) for the 2001–2008 near offset stack difference (left), corresponding far offset amplitudes
(middle, no Q-compensation), and after application of a constant Q-model using Q = 80 (right). The white circle at approx. (1.5, 2.8) km is the
CO2 injection point. The location of the cross-lines shown in Figures 8.6 and 8.7 are shown as white horizontal lines on the left plot in this figure.
RMS 2001 RMS 2001-mean Figure 8.11 Near offset reflection

6 0.15 6 0.1 amplitude (RMS) from 2001 (left) and
same data after subtracting the average
value (right). The maximum value for this
difference is 0.1, which corresponds to
0.08
5 5 a relative velocity change of –0.2.
The corresponding saturation change
0.1 (see dashed line in Figure 8.3) is
0.06 approximately 0.1. This means that the
right figure can also serve as an initial
Y (km)
Y (km)
4 4
guess for the saturation changes
between 1996 and 2001, without further
0.04 scaling.
3 0.05 3
0.02
2 2
0 0
0 1 2 3 0 1 2 3
X (km) X (km)
the entire reservoir thickness. If we use a very long applied to the data, and the initial saturation
window instead of the short window used so far, the (Figure 8.13) was estimated by multiplying the near
algorithm should pick up some saturation changes offset amplitude change by 0.4. In this procedure we
also from deeper layers. We apply the same procedure assume that the time-lapse AVO behavior of a stack of
as described earlier for a time window ranging from thin layers follows the same AVO trend as one thick
850 ms to 1100 ms (which should include all layers layer represented by the RMS amplitude over the
above the injection point). A new calibration was entire layer. The accuracy of this assumption is dis-
done for this purpose and a single scalar of 0.08 was cussed in Appendix B. The estimated saturation 145
changes are displayed in Figure 8.13, showing both an

unsmoothed and a smoothed version. We observe Time-Lapse Gravity
that the algorithm suggests that there are practically Offshore time-lapse gravity measurements have
no saturation changes in an area close to the injection a relatively short history. This monitoring technique
point equal to approximately 1 square kilometer. This was possible by the development of high-precision
area corresponds more or less exactly to the estimated seafloor gravimeters (Sasagawa et al., 2003). By gra-
initial saturation change (left image in Figure 8.13). dual improvements of the seabed gravimeters the
It is very likely that the 4D seismic estimates are repeatability between two gravity surveys was pushed
inaccurate and underestimate the saturation changes from approx. 4 μgal to 2 μgal. For the Sleipner CO2
caused by both the shadow effect from the CO2 that project, we have gravity data from 2002 and 2009
was occupying this volume in 2001, and the fact that available. We will use these data to constrain our
saturation changes above 0.3 are hard to detect from saturation inversion, and especially put weight on
seismic measurements. One way to circumvent this the gravity data in areas where we assume that the
problem is to use the time-lapse gravity data as CO2 saturation is larger than 0.3. Our gravity signal is
a complementary source of information in this area. caused by the density difference between supercritical
smoothed Figure 8.12 Estimated saturation

6 0.3 6 0.3 changes (left) between 2001 and
2008 for the upper layer at Sleipner.
0.25 0.25 The white circle shows the injection
5 5 point. A smoothed version of the
same plot is shown to the right.
0.2 0.2
Y (km)
Y (km)
4 4
0.15 0.15
3 0.1 3 0.1
0.05 0.05
2 2
0 0
0 1 2 3 0 1 2 3
X (km) X (km)
ds from 1999–2001 ds from 2001–2008 smoothed Figure 8.13 (Left) Near off-
6 0.2 6 0.3 6 0.3 set RMS amplitude for a time
window from 850 to 1100 ms,
0.25 0.25 after subtracting the average
5 0.15 5 5 value, and scaled by 0.4. This
is used as an initial estimate of
0.2 0.2 the average CO2 saturation in
2001. (Middle) Estimated
Y (km)
4 4 4
0.1 0.15 0.15 saturation changes between
2001 and 2008 from the 4D
3 3 0.1 3 0.1 seismic data. (Right)
0.05 Smoothed version of the
estimated saturation
0.05 0.05 changes.
2 2 2
0 0 0
0 1 2 3 0 1 2 3 0 1 2 3
146 X (km) X (km) X (km)
CO2 and the brine water originally trapped in the pore

space within the Utsira sand layer.
Following for instance Keary et al. (2002; Eq. 8.9)
we can write down the expression for the gravity
anomaly caused by the CO2 injection. We find that
the gravity difference response modeled at the seabed
in (x, y) is simply estimated by computing the follow-
ing 2D integral:
ð
Dgmod ðx; yÞ ¼ G dx0 dy0 φΔρΔSðx0 ; y0 Þ
dz
32 : ð8:14Þ
ðx x Þ þ ðy y0 Þ2 þ z2
0 2
Here G is the gravity constant, φ the porosity,

Δρ the density difference between CO2 and water, Figure 8.14 Modeled gravity signal change based on the 4D
seismic estimate of saturation change (left). The color bar is in
z the depth to the reservoir, and d the thickness of μGal, and 10 μGal corresponds to a reduction of 10 μGal
the reservoir. ΔS is the CO2 saturation change between 2001 and 2008. The circles show the measured gravity
between 2002 and 2009. Figure 8.14 (left) shows the anomaly data. The circles are color-coded, so a huge color
difference between the circle and the surrounding modeled
estimated gravity signal based on the smoothed signal means discrepancy between modeling and observations.
saturation change estimate (Figure 8.13, left). As our The size of the circles is proportional to the gravity change
main concern here is the spatial distribution, we have signal. (Right) Corresponding modeled gravity change, after least
square error minimization. The injection point is shown as
simply scaled the final estimated gravity change to a white circle.
match the maximum measured gravity change.
We clearly observe from Figure 8.14 (left) that there
is a mismatch between the measured gravity data has to be determined on a trial-and-error basis.
(shown by circles) and the modeled response In this way we allow the estimated saturation
based on the estimated saturation changes from 4D changes based on time-lapse seismic to be scaled by
seismic. The white circle in Figure 8.14 is the injection k for instance in the shadow zone. This means that
point. we keep the spatial distribution of the saturation
Now we will use a simple inversion strategy: Let changes suggested by the time-lapse seismic inver-
the estimated saturation changes in the areas where sion, but let the measured gravity data determine
the estimated saturation changes are below a critical how much the estimated saturation change should
threshold (Sc) be scaled by a factor (k) in order to be up-scaled. In effect, we let the gravity data “talk” in
enhance the estimated saturation changes. This criti- the shadow zone, and simply assume that the satura-
cal threshold is estimated by trial and error and it is tion change suggested by the time lapse seismic
a practical number that identifies areas where the inversion should be strengthened by this scalar in
seismic shadow effect has been strong. This means order to compensate for the seismic shadow effect
that we change the modeled gravity difference given and the fact that there is low seismic sensitivity
in Eq. (8.14) to for saturation changes from 0.3 and above.
ð
To determine an optimal value for k we use the
Dgmod ðk; x; yÞ ¼ k G dx0 dy0 φΔρΔSðx0 ; y0 Þ following least squares norm:
dz N
X 2
32 ð8:15Þ LSðκÞ ¼ Δgmod ðκ; xi ; yi Þ Δgobs ðxi ; yi Þ
0 2 0 2
ðx x Þ þ ðy y Þ þ z2 i¼1
where k = 1 (meaning no change) if ΔS > SC . In the ð8:16Þ

present example we used Sc = 0.03. The risk of intro- where N is the number of gravity anomaly measure-
ducing noise in the final saturation image increases ments (denoted Dgobs ) made at locations (xi, yi).
as the critical threshold decreases, so this parameter By letting this factor ðκÞ vary between 1 and 7, we
147
1 Figure 8.15 Relative least squares error as

a function of the scaling factor used to enhance
0.95 the saturation changes in the “shadow zone.”
The optimal scaling factor is 2.4.
0.9
Relative LS error
0.85
0.8
0.75
0.7
0.65
1 2 3 4 5 6 7
Scaling factor
Before After The final estimated saturation changes are simply

6 6 0.3
given as DSfinal ¼ kΔS, where ΔSðx; yÞ represents the
0.25 estimated saturation changes based on seismic data
5 5 only. In this inversion we used only gravity measure-
0.2 ments that were larger than 4 μGal. The 4 μGal thresh-
Y (km)
4 4 old represents the accuracy of the time-lapse gravity

0.15
measurements (Alnes et al., 2011). A comparison
3 3 0.1 between the initial saturation change estimate
(between 2001 and 2008) and the inverted one is
0.05
2 2 shown in Figure 8.16. We clearly see the role of the
0 gravity data: to increase the saturation changes close
0 1 2 3 0 1 2 3
to the injection point.
X (km) X (km) On the lower right corners of the estimated satura-
tion changes shown in Figure 8.16, there is an anomaly
Figure 8.16 Initial saturation change estimate between 2001 and
2008 based on 4D seismic (left) and inverted saturation changes that we interpret as noise. If we study the RMS differ-
constrained by a least squares inversion using the gravity ences for near and far offset (Figure 8.17) we notice that
measurements. there is a weak difference signal in the lower right
corner on the far offset data. As we have not imple-
mented a cutoff related to the amplitude of the 4D
find a global least squares error minimum between the differences, and we are using the difference between
gravity data and the modeled gravity response from the near and far offset differences, this noise will show
the estimated saturation changes at approx. 2.4, as up as signal on the estimated saturation changes.
shown in Figure 8.15. The optimal modeled gravity However, the uncertainties and pitfalls associated
change is shown in Figure 8.14 (right), and we see that by using Eq. (8.13) for a long time window are numer-
the match between the measured gravity data (shown ous. First, there will be shadow effects and focusing/
as circles) and the modeled gravity data has improved. defocusing effects, causing the deeper reflections to
This is particularly true for the saddle area between be less accurate and actually misleading to be used by
the two main peaks in the modeled gravity anomaly. a simple AVO interface approach as proposed here.
For other measurements we can actually find Second, remaining multiples caused by the strong
areas where there is no improvement. However, the reflections associated with the top layer will cause
overall result is that the least squares error is reduced misleading results. Third, additional filling of a
148 by approximately 20%, as shown in Figure 8.15. lower CO2 layer will be observed only if the CO2
the final result. A downward migration of the CO2

6 6 0.2
that was in the plume in 2001 will also cause a positive
gravity change, but this is also very unlikely to hap-
5 5 0.15 pen. It is much more likely that injection of 5.88
Mtonnes of CO2 will cause decreased gravity, both
Y (km)
4 4 when measured right above the plume and when

0.1
measured at long distances away from the plume.
3 3 Therefore, we used a threshold of 4 μGal (correspond-
0.05
ing to the uncertainty of the time lapse gravity mea-
2
dNear
2
dFar surements) for the constrained inversion. Figure 8.18
0 shows a simple comparison between the measured
0 1 2 3 0 1 2 3
gravity data plotted against the radial distance from
X (km) X (km)
the injection point. In this figure we compare the
Figure 8.17 Near (left) and far (right) offset RMS-differences experimental data to simple gravity modeling assum-
(between 2001 and 2008) using a time window from 850 to 1100 ms. ing that the 5.88 Mtonnes of CO2 has been injected as
Notice the weak noise in the lower right corner on the far offset a point mass at the injection point (shown by solid
differences. These differences (although they are weak) create
a noisy saturation signal (see Figure 8.16). lines), and assuming that the same mass is distributed
as two point masses separated by 1.6 km (dashed
lines). Black curves represent a CO2 density of 675
saturation increases to a level that is less than 0.3. For kg/m3 while red curves correspond to a density of 720
instance, an increase in saturation from 0.3 to 0.6 will kg/m3. We clearly see that the simple monopole dis-
not give a large reflectivity change in the time-lapse tribution does not fit the data, and that the two-point
seismic data. distribution is closer to the measurements. It should
be noted that the dashed lines correspond to an in-line
Discussion two-point distribution. For other directions, the grav-
In the constrained inversion we used only gravity data ity signal will be weaker, and even closer to the mea-
that showed a decrease in gravity, and where the surements. Compared to the best estimate of CO2
absolute value of the gravity change is larger than 4 density changes obtained by the joint inversion
μGal. If we include all available measurements, some method (Figure 8.14), we observe that this estimate
of the weaker gravity signals and especially those is far from such simple distributions: It is more like an
measurements resulting in increased gravity change elongated structure with a weak concentration struc-
will give a very strong signal close to the injection ture in the north–south direction. This might indicate
point. In addition to this, gravity measurements that that the uppermost layers are being filled from several
are farther away from the Sleipner CO2 site and closer sources that are spread, and not only from the area
to the area of gas production will be more influenced close to the injection point.
by the gas production. The Sleipner gas field is situ- How sensitive is the time-lapse AVO method to
ated to the west of the Sleipner CO2 area. We think errors in the estimated calibration parameters?
this is misleading for the following reasons: Between A simple test where the global scaling parameter is
years 2002 and 2009, 5.88 Mtonnes of CO2 had been changed from 0.06 to 0.08 is shown in Figure 8.19.
injected, and therefore a decreased gravity should be The major effect is to increase the saturation by
expected, especially for those measurements that are a constant factor. Maybe the increase is slightly less
taken outside the plume (the farther away from the in the south compared to the north. A similar effect
plume the closer the gravity signal will be to a point (but opposite in sign) is observed if the Q-factor is
source). There are some scenarios that might create increased from 80 to 100: The estimated saturation
increased gravity, for instance, if the CO2 dissolves in changes have approximately the same spatial distribu-
water (after a time) and hence water goes back into the tion, but the overall saturation changes are less, as
plume again. In our opinion this is not very likely to shown in Figure 8.20. From these simple tests, we
happen at a large scale. However, in a detailed mass conclude that the calibration procedure does not
balance computation such an effect might influence influence the spatial distribution of the estimated 149
30 Figure 8.18 Change in observed gravity

3
anomaly signals (black dots) between 2002 and
720 kg/m 2009 plotted as function of the radial distance
25 Point source
675 kg/m3 from the injection point. Solid lines show point
source (assuming that 5.88 Mtonnes of CO2 has
Gravity signal (uGal)
20 Dipole source 1.6 km been injected) gravity signals assuming that

the density of CO2 is constant and equal to 675
15 kg/m3 (black line) and 720 kg/m3 (red line).
Density of water is assumed to be 1040 kg/m3.
10 Density of water: 1040 kg/m3 A depth of 900 m has been used for the point
Depth of source(s): 900 m source gravity modeling. Dashed lines show
corresponding (inline) result assuming two
5
point sources separated by 1.6 km.
0
–5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Radial distance (km)
6 6 6 6
Q = 80 Q = 100
0.25 0.25
5 5 5 5
0.2 0.2
Y (km)
Y (km)
4 4 4 4
0.15 0.15
3 3 0.1 3 3 0.1
0.05 0.05
2 2 2 2
0 0
0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3
X (km) X (km) X (km) X (km)
Figure 8.19 The effect of increasing the global scaling factor from Figure 8.20 The effect of increasing the Q-factor from 80 (left) to
0.06 (left) to 0.08 (right) on the estimated saturation changes based 100 (right) on the estimated saturation changes based on time-lapse
on time-lapse seismic data only. seismic data (2001 and 2008) only.
saturation changes significantly, but it changes the DC The total amount of CO2 injected between the
component or the absolute value of the estimated 2002 and 2009 seismic surveys is 5.88 Mtonnes.
changes. If we use the estimated saturation change values as
We have chosen to use Eq. (8.13) to estimate the shown in Figure 8.16b, we find that the total mass of
saturation changes and then use the direct relation- CO2 injected in this period is equal to 7.91 Mtonnes.
ship between saturation and density to get the density In this simple calculation we have assumed an average
changes. This choice is related to the fact that the reservoir thickness of 200 m, and that the density
sensitivity for density changes on the seismic data is difference between CO2 and water is 230 kg/m3.
significantly less than that for velocity (Figure 8.3). Furthermore, we have assumed that the average por-
However, if we use Eq. (8.9) to estimate density osity is 0.37. The injection point is not located at the
changes directly from the seismic data we get base of the Utsira sandstone unit, but approx. 50 m
a result (Figure 8.21) that is not too far from the above. Using 150 m for the reservoir thickness will
saturation estimate (Figure 8.16, left). It should be therefore reduce the total injected CO2 from 7.91 to
noted, however, that the estimated size of the density 5.93 Mtonnes, which is much closer to the actual
150 change – 30% – is not very realistic. value. However, the uncertainties coupled to these
6 6 Appendix A: Using a Simple Q-Model to

0.25
Calibrate Near versus Far Offset Stacks
5 5
From Eq. (8.13) it is evident that it is critical to
0.2 determine the ratio between the near and far offset
4 4
Y (km)
0.15
changes as precisely as possible. The time-lapse pro-
cessing of the data has been done as accurately as
3 3
0.1 possible, but the Q-compensation has been very mod-
erate, as a relatively high Q-value of 300 has been
2 2 0.05 applied to the data. Typical Q-values for overburden
0 sediments in the North Sea is more likely to be less
0 1 2 3 0 1 2 3 than 100 (Reid et al., 2001), and therefore we intro-
X (km) X (km) duce a simple Q-model correction term:
2πf ðcoszθ coszθ Þ=ðQVÞ
Figure 8.21 Estimated density changes using the time window AQ ¼ e F N ð8:A-1Þ
from 850–1100 ms (left) using seismic data from 2001 and 2008.
The smoothed version is shown to the right. where Q is the quality factor, z is the depth to the
Utsira formation, f is the average frequency, and V is
the average P-wave velocity for the overburden layers.
estimates are huge, and therefore this exercise should The far offset data should be divided by AQ to correct
not be taken too seriously. for this absorption effect. One way to implement Eq.
(8.A-1) is to simply test various Q-values and see how
Conclusion robust Eq. (8.13) is to such variations. For calibration
A calibrated time-lapse seismic method using near purposes we used a z-value of 800 m, an average
and far offset differences as input to estimate CO2 frequency of 50 Hz and an average P-wave velocity
saturation changes, has been tested on field data for the overburden (V) of 1800 m/s.
from the Sleipner CO2 injection site, offshore
Norway. We find that this method is of limited value Appendix B: Using Backus Averaging
if the CO2 is stored in several layers on top of each
other, as the time-lapse seismic data have less sensi-
to Estimate Time-Lapse Response of
tivity to detect saturation changes in such a multi- a Multilayered CO2 Storage Site
layered medium. By combining the seismic method Backus (1962) introduced a systematic way of averaging
with time-lapse gravity measurements we demon- finely layered media. Using this averaging technique
strate that a simple inversion procedure can be used Stovas et al. (2006) showed that it is possible to replace
to estimate saturation changes also in areas where a sequence of finely layered medium, with one thick
multiple CO2 layers are stacked on top of each other. layer being described by average parameters and the net
to gross (N/G) ratio, which is a number describing the
relative thickness of all CO2 layers within the Utsira
Acknowledgments sand divided by the total thickness. In Stovas et al.
We thank Statoil and the Sleipner license partners (2006) the N/G ratio is defined as the thickness of the
ExxonMobil and Total for permission to use the sand layer divided by the total thickness of the reservoir
data. Anne Kari Furre is acknowledged for assistance, layer. In this work we have adapted the N/G ratio for
discussions, and providing the seismic data to us. our purpose, and therefore we replace the sand thick-
Thanks also go to Alistair Harding for helping us to ness by thickness of CO2 layers. For the Sleipner case,
load and read the data. Matthew Dzieciuch is the N/G ratio is relatively small, as the thickness of each
acknowledged for assistance in writing Matlab layer is of the order of 5–15 m. To the lowest order, the
scripts. ML wants to thank Scripps for hosting him reflection coefficient versus offset can be written as
during his sabbatical stay, and the Norwegian
Research Council for financial support. RðθÞ ¼ R0 þ G sin 2 θ ð8:B-1Þ
151
where θ is the incidence angle, R0 is the zero angle temporal evolution of injected CO2 at the Sleipner
reflection coefficient, and G is the gradient. Field, North Sea. Journal of Geophysical Research, 117:
If we assume that the contrasts in all elastic constants B03309.
are small, Eq. (8.11) from Stovas et al. (2006) reads (note Buland, A., Landrø, M., Andersen M., and Dahl, T. (1996).
that in Stovas et al. [2006] Δρ and ΔVP are dimension- AVO inversion of Troll Field data. Geophysics, 61:
less entities in contrast to this chapter where Δρ=ρ is 1589–1602.
the corresponding dimensionless quantity): Cavanagh, A. J., and Haszeldine, R. S. (2014). The Sleipner
storage site: Capillary flow modeling of a layered CO2
N=G Δα Δρ plume requires fractured shale barriers within the
R0 ¼ þ
2 α ρ Utsira Formation. International Journal of Greenhouse
ð8:B-2Þ
N=G Δα β2 Δρ 2Δβ Gas Control, 21: 101–112.
G¼ 2 2 þ
2 α α ρ β Evensen, A. K., and Landrø, M. (2010). Time-lapse
tomographic inversion using a Gaussian
where α and β denote P- and S-wave velocities, respec- parameterization of the velocity changes. Geophysics,
tively. This means that to the lowest order, both R0 and 75: U29–U38.
G are directly proportional to the expressions for one Furre, A. K., and Eiken, O. (2014). Dual sensor streamer
thick layer (if we let N/G = 1 in Eq. 8.B-2 we are back to technology used in Sleipner CO2 injection monitoring.
the conventional AVO formula for PP reflections). For Geophysical Prospecting, 62: 1075–1088.
medium contrasts in all elastic parameters Stovas et al. Garcia Leiceaga, G., Silva, J., Artola, F., Marquez E., and
(2006) found expressions that include higher order Vanzeler, J. (2010). Enhanced density estimation from
terms in N/G ratios. These expressions will alter the prestack inversion of multicomponent seismic data.
analysis done in this chapter somewhat. However, if Leading Edge, 29: 1220–1226.
the N/G ratio is small, these terms will not change our Ghaderi, A., and Landrø, M. (2009). Estimation of thickness
estimates significantly. For N/G ratios close to 0.5 these and velocity changes of injected carbon dioxide layers
corrections will be more significant and that requires from prestack time-lapse seismic data. Geophysics, 74:
that we use Eq. 8.17 in Stovas et al. (2006) instead of Eq. O17–O28.
(8.B-2). Grude, S., Landrø M., and Osdal, B. (2013). Time-lapse
pressure saturation discrimination for CO2 storage at
the Snøhvit field. International Journal of Greenhouse
References Gas Control, 19: 369–378.
Alnes, H., Eiken, O., Nooner, S., Sasagawa, G., Stenvold, T., Helgesen, J., and Landrø, M. (1993). Estimation of elastic
and Zumberge, M. (2011). Results from Sleipner gravity parameters from AVO effects in the tau-p domain.
monitoring: Updated density and temperature Geophysical Prospecting, 41: 341–366.
distribution of the CO2 plume. Energy Procedia, 4: Keary, P., Brooks M., and Hill, I. (2002). An introduction to
5504–5551. geophysical exploration, 3rd edn. Oxford: Blackwell.
Arts, R., Chadwick, A., Eiken, O., Thibeau S., and Nooner, S. Kiær, A. F., Eiken O., and Landrø, M. (2015). Calendar time
(2008). Ten years’ experience of monitoring CO2 interpolation of amplitude maps from 4D seismic data.
injection in the Utsira sand at Sleipner, offshore Geophysical Prospecting, 64: 421–430.
Norway. First Break, 26: 65–72.
Landrø, M. (1999). Discrimination between pressure and
Backus, G. E. (1962). Long-wave elastic anisotropy fluid saturation changes from time lapse seismic data.
produced by horizontal layering. Journal of Geophysical Expanded Abstracts, Society of Exploration
Research, 67: 4427–4440. Geophysicists Technical Program, 1651–1654.
Bai, J., and Yingst, D. (2014). Simultaneous inversion of Landrø, M. (2001). Discrimination between pressure and
velocity and density in time-domain full waveform fluid saturation changes from time lapse seismic data.
inversion. Expanded Abstracts, Society of Exploration Geophysics, 66: 836–844.
Geophysicists Technical Program, 922–927.
Lindeberg, E., and Bergmo, P. (2003). The long-term fate of
Bhakta, T., and Landrø, M. (2014). Estimation of CO2 injected into an aquifer. In J. Gale and Y. Kaya
pressure-saturation changes for unconsolidated reservoir (eds.), Proceedings of the 6th International Conference
rocks with high Vp/Vs ratio. Geophysics, 79: M35–M54. on Greenhouse Gas Control Technologies (GHGT-6),
Boait, F. C., White, N. J., Bickle, M. J., Chadwick, R. A., Kyoto, Japan, October 1–4, 2002. Oxford: Pergamon,
152 Neufeld J. A., and Huppert, H. E. (2012). Spatial and 489–494.
Queißer, M., and Singh, S. C. (2013). Localizing CO2 at Span, R., and Wagner, W. (1996). A new equation of state
Sleipner: Seismic images versus P-wave velocities from for carbon dioxide covering the fluid region from the
waveform inversion. Geophysics, 78: B131–B146. triple-point temperature to 1100 K at pressures up to
Rabben, T. E., and Ursin, B. (2011). AVA inversion of the 800 MPa. Journal of Physical and Chemical Reference
top Utsira Sand reflection at the Sleipner field. Data, 25: 1509–1596.
Geophysics, 76: C53–C63. Stovas, A., Landrø M., and Avseth, P. (2006). AVO attribute
Reid, F. J. L., Nguyen, P. H., MacBeth, C., Clark R. A., and inversion for finely layered reservoirs. Geophysics, 71:
Magnus, I. (2001). Q estimates from North Sea VSPs. C25–C36.
Expanded Abstracts, Society of Exploration Trani, M., Arts, R., Leeuwenburgh O., and Brouwer J.
Geophysicists Technical Program, 440–443. (2011). Estimation of changes in saturation and
Roy, B., Anno, P., and Gurch, M. (2006). Wide-angle pressure from 4D seismic AVO and time-shift analysis.
inversion for density: Tests for heavy-oil reservoir Geophysics, 76: C1–C17.
characterization. Expanded Abstracts, Society of Tura, A., and Lumley, D. E. (1999). Estimating pressure and
Exploration Geophysicists Technical Program, saturation changes from time-lapse AVO data. In 69th
1660–1664. Annual International Meeting, The Society of
Roy, B., Anno, P., and Gurch, M. (2008). Imaging Exploration Geophysicists. Expanded Abstracts, Society
oil-sand reservoir heterogeneities using wide-angle of Exploration Geophysicists Technical Program,
prestack seismic inversion. Leading Edge, 27: 1655–1658.
1192–1201. Zumberge, M., Alnes, H., Eiken, O., Sasagawa G., and
Sasagawa, G., Crawford, W., Eiken, O., Nooner, S., Stenvold, T. (2008). Precision of seafloor gravity and
Stenvold T., and Zumberge, M. (2003). A new seafloor pressure measurements for reservoir monitoring.
gravimeter. Geophysics, 68: 544–553. Geophysics, 73: WA133–WA141.
153
Chapter
Electrical and Electromagnetic Methods
9 Erika Gasperikova and Michael Commer
Introduction Given a complete description of the physical sys-

tem, that is, with the subsurface structure in terms of
Monitoring approaches for geological carbon dioxide
its physical property distribution precisely known, the
(CO2) storage (GCS) or geosequestration depend on
field values could be predicted uniquely. This process
objectives, subsurface reservoir/site dimensions, and
is called the simulation or modeling problem. In geo-
the stage of a CO2 storage site. Different monitoring
physical surveying, the problem is the opposite: esti-
techniques would be selected for a site characterization
mate the properties of the subsurface structure on the
prior to CO2 injection, for monitoring while injecting
basis of actual field measurements. This procedure is
CO2, or for a postinjection stage. Electromagnetic
analogous to medical imaging. In medicine, radiolo-
(EM) and electrical techniques are used to map subsur-
gic or ultrasound methods are routinely employed to
face electrical resistivity. They are complementary to
make pictures of organs and structures inside the
seismic monitoring because they are sensitive to fluid
human body. Similarly, geophysical measurements
properties and hence able to detect elevated CO2
have the ultimate goal of producing structural pic-
saturations.
tures of the Earth’s subsurface. Both instances involve
Electrical and EM techniques measure electric (E)
the simulation of fields that propagate through the
and magnetic (B or H) fields caused by currents that
body/subsurface and field anomalies due to anoma-
are injected into the ground by contacting electrodes
lous material properties along the travel path.
or a time-varying magnetic field induced to flow into
Compared to medical imaging, the obvious difference
the ground by inductive sources. Electrical techni-
is the size of the investigated body. This has the major
ques, also called resistivity techniques, use only cur-
implication that geophysical imaging problems are
rent and voltage measurements at frequencies low
harder to solve, because the measurements at hand
enough at which EM induction effects are negligible.
are usually sparse in relation to the subsurface volume
EM techniques require frequency-dependent sources
of interest. Because of the sparsity issue, geophysical
to induce currents in the ground. Magnetic fields are
data analysis in the pioneering days was typically
produced from currents created from both types of
restricted to a trial-and-error repetition of the follow-
sources. The basic concept in both techniques is to
ing procedure:
measure these electric and magnetic fields and to infer
from these measurements the configuration and 1. Assume a certain subsurface structure by
amplitudes of the current in the subsurface and establishing and quantifying the relevant
hence the distribution of electrical resistivity. parameters that make up this structure.
Injected CO2 is expected to form lenticular lenses 2. Simulate the underlying physical processes that
or plumes of a finite size and change the subsurface are sensitive to these structure parameters.
resistivity resulting in a resistivity contrast with the 3. Predict the measurements that would be produced
enclosing formation. The goal of the survey is to by these simulations.
identify a local variation in resistivity relative to the 4. Compare these predictions to actual real-world
background geology. The changes in resistivity and measurements.
the associated perturbations in the measured electric
and magnetic fields are referred to as anomalies. The Depending on the number of structure parameters,
process of continuously measuring field variations this procedure may need to be repeated many times in
154 due to the anomalies is referred to as monitoring. the hope of closing in on a parameter set that leads to
a good match in step 4. The final step is to graphically

visualize the spatial distribution of the parameters
and obtain an interpretable subsurface image. The
procedure of carrying out steps 1–3 is also known
as forward modeling because a geophysical survey
scenario is modeled, usually by means of computer
simulations.
The aforementioned trial-and-error modeling
procedure of a complex subsurface would become
rather tedious because it would have to be repeated
many times. Therefore, a reversed approach, inverse
modeling or simply inversion, is more appropriate.
It takes given measurements as input with the goal
of predicting the “right” numbers for the structural
(modeling) parameters such that step 4 gives a match.
A graphically enhanced version of such an algorithm
would finally produce an easy-to-interpret layman’s
image of the “true” model parameters. However, the
forward modeling routine is an essential ingredient
of inversion codes. Internally, inversion algorithms Figure 9.1 Phase diagram of CO2.
carry out many forward modeling instances in order
to figure out the best search direction in the parameter
properties. Commonly used is Archie’s Law
space. An overarching optimization framework
(Archie, 1942), which describes the electrical resis-
controls this search by modifying model parameter
tivity ðρb Þ of sedimentary rocks as a function of
guesses in an iterative way until data predictions
water saturation (Sw), porosity ðϕÞ, and pore fluid
match observations.
resistivity ðρw Þ:
ρ
Rock Properties and Resistivity ρb ¼ a ϕm nw
Sw
The form and behavior of CO2 depend on tempera-
ture and pressure. CO2 behaves as a gas in air at where a is tortuosity, and m and n are constants with
standard pressure and temperature, as a solid (called 1.8 < m ≤ 2 and n ffi 2.
dry ice) when frozen, or as a supercritical fluid at During GCS, CO2 may be injected into
temperatures and pressures at or above the critical a formation originally filled with brine. Replacing
point (temperature of 31°C and pressure of 7.4 MPa) brine with CO2 results in a CO2 saturation SCO2 =
(Figure 9.1). Under supercritical conditions, which 1 – Sw. Figure 9.2 shows the rock bulk resistivity (ρb)
appear at depths greater than 800 m, CO2 takes as a function of CO2 saturation (SCO2) for the forma-
a much smaller volume than in the gas phase. tion with brine resistivity of 0.3 Ohm-m and 25%
The resistivity of CO2 is high, similar to gas or air, porosity. CO2, as well as all petroleum fluids (e.g.,
independent of its state. Brine-bearing formations oil, condensate, and hydrocarbon gas), is electrically
that are below and hydrologically separated from resistive; hence the relation shown in Figure 9.2 is
drinking water reservoirs have been widely recog- appropriate not only for CO2 but also for any combi-
nized as having high potential for GCS. The resistivity nation of oil, hydrocarbon gas, or condensate.
of brine depends on the amount of total dissolved The replacement of highly conductive saline fluids
solids (TDS), but in general is low. (TDS = 10 000 ppm and up; parts per million;
The electrical resistivity of the subsurface is 1 ppm = 1 mg of salt in 1 liter of water) with resistive
highly sensitive to changes in key formation prop- CO2 results in a resistivity increase in the storage
erties such as porosity, pore fluid resistivity, and reservoir. When CO2 is present at shallow depths,
fluid saturation. A wide range of empirical rela- dissolution of CO2 causes an increase in TDS and
tions exists for linking formation and electrical results in resistivity decrease. 155
Erika Gasperikova and Michael Commer
Figure 9.3 Schlumberger electrode array for a surface electrical

survey. C1 and C2 are current electrodes, P1 and P2 are voltage
electrodes, d1 is spacing between potential electrodes, d2 is spacing
between current and potential electrodes, L is total array length, I is
current, V is voltage.
Basic Principles of Electrical and EM

Techniques
Resistivity Techniques
Figure 9.2 Reservoir bulk resistivity (ρrock) as a function of gas Resistivity is a material property that describes the
saturation (Sg = 1 – Sw) for a reservoir with brine resistivity equivalent resistance to current flow when a voltage is applied
to sea water (ρbrine = 0.33 Ω-m) with 25% porosity. (Figure 1 from
Gasperikova and Hoversten, 2016.) across a sample of the material. The resistance R (in
Ohms) between the current (I) flowing through
Electrical resistivity can be used to determine CO2 a sample of length L and cross section A and the
saturation: applied voltage (V) across the sample is given by
aρ
SCO2 ¼ 1 – Sw ¼ 1 m w : R ¼ V=I
φ ρb
Complex mineral composition may affect bulk
rock parameters, and estimates of CO2 saturations
using Archie’s equation might not be accurate.
In such situations, another useful and simple relation-
ship between resistivity and brine saturation, the
resistivity index (Gueguen, 1994), can be used: When expressed in terms of the electric field E =
ρ V/L, in volts per meter (V/m), and current density J =
RI ¼ ¼ ðSw Þn ;
ρ0 I/A, in amperes per square meter (A/m2), this
becomes the resistivity ρ (Ohm-m):
where ρ is the resistivity of the rock partially saturated
with brine, ρ0 is the resistivity of fully saturated rock E A
ρ¼ ¼R
with brine, and n is the saturation exponent. Again, in J L
the case of CO2 injection, it is possible to estimate CO2 The relation J ¼ Eρ is Ohm’s law and it defines the
saturation from the initial resistivity of the fully satu- resistivity. It is often written as J ¼ σE, where σ ¼ 1ρ is
rated rock with brine and partial brine saturation the conductivity in Siemens per meter (S/m). For the
during CO2 injection using: sample cross sectional area of 1.0 m2 and the length
1=n 1.0 m ρ ¼ R.
1
SCO2 ¼ 1– The resistivity of the ground is measured by inject-
RI
ing electrical currents into the ground and measuring
When a formation contains a substantial amount of the resulting potential differences. The current source
clay, an additional parameter – the ratio of volume of is typically a wire and two current electrodes (C1, C2)
sand to volume of clay – is necessary (e.g., Nakatsuka connected to a battery or a generator, called an electric
156 et al., 2010). bipole (Figure 9.3). The voltage difference is measured
Figure 9.4 Schlumberger apparent resistivity

response for a layer 50 m thick and resistivity of
1000 Ohm-m in a 100 Ohm-m half-space as
a function of layer depth.
between two potential electrodes (P1, P2 in geometric effect, the observed voltage V for any
Figure 9.3). The electrodes can be on the surface as source–receiver array can be interpreted in terms of
well as in boreholes. The measured electric field, E, is the resistivity of a uniform half-space that would give
given by Ohm’s law. If the ground is of uniform rise to the measured value of V. This is called the
resistivity, and the injected current (I) and the spacing apparent resistivity ρa and it is the standard form of
of the electrodes (d1, d2) are known, then the mea- representing the results of resistivity surveys.
sured voltage (V) is linearly related to the ground Figure 9.4 shows the Schlumberger apparent resis-
resistivity through a simple formula. If the ground is tivity response for a layer 50 m thick and resistivity
not uniform, the currents will be distorted by the of 1000 Ohm-m in a 100 Ohm-m half-space as
inhomogeneities, and voltage measurements become a function of layer depth. The gray vertical line is
indirect measures of the resistivity distribution. a response of 100 Ohm-m half-space. This figure
The array shown in Figure 9.3 is called the shows that with increasing layer depth, the peak
Schlumberger array. As the array length L increases, response gets smaller and necessitates an increasing
the injected currents flow deeper into the earth, and current electrode spacing. The figure is illustrative of
the measured V becomes sensitive to deeper layers or a practical deployment problem with the technique;
features. In any resistivity survey, the value of V is the current electrode spacing has to be greater than
greatly influenced by the geometry of the source– 8 km (L/2 = 4 km) to define the response curve for the
receiver configuration as well as by the resistivity layer at 1.0 km depth. In addition to the logistical 157
distribution in the subsurface. To normalize the difficulty of deploying an 8 km long current-carrying
Figure 9.5 Dipole–dipole response

for deep conductive and resistive
bodies. (Figure 10 from Gasperikova
and Morrison, 2019.)
cable, it is very likely that on such a scale, the local potential electrodes and their ratios vary. The reader
shallow geological variations would produce resistiv- is referred to general geophysics textbooks (e.g.,
ity inhomogeneities that would distort the sounding Telford et al., 1990) for details about these arrays.
response. To achieve the best results in CO2 reservoir or
If the resistive layer at 1 km depth in Figure 9.4 is not plume monitoring, arrays are selected based on for-
continuous, but instead creates a thin slab of finite ward modeling and analysis of a model that captures
lateral extent (400 m), the response is significantly less the geological complexity at the monitoring site (e.g.,
(Figure 9.5). Figure 9.5 shows a model with resistive Zhou and Greenhalgh, 2000; Christensen et al., 2006;
(1000 Ohm-m) or conductive (1 Ohm-m) finite size Hagrey, 2012).
body at the depth of 1 km in 112 Ohm-m background.
The response in Figure 9.5 is for a traversing dipole–
dipole array with 200 m current (I) and potential (V) EM Techniques
dipoles and the separation of 2 km; the array moves in The current loops carrying an alternating current
200 m steps over the body. The dipole–dipole array or grounded current dipoles injecting alternating
combines sounding and lateral resolution, and has current into the ground have an associated chan-
a significant deployment advantage because the cable ging magnetic field that, through Faraday’s law,
needs to be laid only between relatively closely spaced induces currents to flow in the subsurface. A time-
electrodes. The depth of detection is controlled by the varying magnetic field passing through a circuit
dipole separation – the depth of the detection increases produces an electromotive force (emf), which is
with the dipole separation. The apparent resistivities are proportional to the time rate of change of the mag-
plotted at the midpoint of the array for the conductive netic flux threading the circuit. The total flux, Φ,
and resistive bodies. The conductive target causes 7% through the circuit is defined as the integral of the
apparent resistivity decrease, while the presence of the component of B normal to the surface contained by
resistive target causes only 1.6% apparent resistivity the circuit:
increase. Thus, the small resistive body (or CO2 ð
plume) would be very difficult to detect or to monitor. Φ ¼ B ds
Electrodes can be configured in many different
158 patterns where the distance between current and Faraday’s law then states that
ð (2 Ohm-m) at a depth of 1000 m in the resistive

d dΦ
emf ¼ B ds ¼ background (30 Ohm). The blue star indicates the
dt dt position of the source and inverted triangles indicate
In the circuit, the emf is the integral of the electric receiver positions. The curves in Figures 9.6b–i are
field around the circuit: plotted as a function of offset, which is the distance
I between the source and receiver. The response of the
dΦ
emf ¼ E dl ¼ resistor is shown in blue, while the response of the
dt
conductor is shown in red. The fields are plotted as
Applying Stokes’ theorem we get the differential form a ratio with and without the anomalous layer present.
of Faraday’s law: Two types of sources are considered: a horizontal
dB electric dipole (HED) (Figures 9.6d–i), and a vertical
rE¼ magnetic dipole (VMD) (Figures 9.6b, c). Depending
dt
on the orientation of the source, EM field ratios are
and Ampere’s law:
plotted for receiver components for which the ratios
r B ¼ μJ: are larger than 10%. For the conductor, all configura-
Ampere’s and Faraday’s laws, the constitutive relations produce a signal above this limit and therefore
tions, J ¼ σE, B ¼ μH, and the fact that r B ¼ 0 could be used for imaging this target. In contrast, only
are all the equations and relationships that are needed Figures 9.6d and e show a significant anomaly for the
to solve any problem in low-frequency electromag- resistive target. The poor detectability of resistors to
netic induction. most source–receiver geometries is related to currents
Both the current source (transmitter) and recei- preferably flowing within more conductive bodies.
vers are usually multiturn loops of wire. A field pro- The relatively good detectability of resistors to the
duced by a multiturn loop is proportional to its dipole HED–Ex and HED-Ez configurations is associated
moment, M, which is equal to the product of the with the guided-wave mode developing and propagat-
current, I, the area of the loop, A, and the number of ing within the resistive layer (Streich, 2016).
turns, N. The moment is a vector whose direction is Measured signal levels depend on the volume
normal to the plane of the loop (along the axis of the affected by CO2 or the target size. Responses calcu-
loop). The currents induced in the subsurface are lated using layered models give the upper limit of the
a function of the time rate of change of the primary response to be expected. However, in many practical
field at the inhomogeneity and of its size, shape, applications the storage reservoir or the target is of
resistivity ðρÞ, and magnetic permeability ðμÞ, and a finite size, which needs to be considered when
the resistivity and permeability of the surrounding designing a monitoring survey. Figure 9.7 illustrates
medium. The response of the ground or an anoma- a design of marine controlled source EM (CSEM)
lous structure (target) is defined as the measured field survey for a sensitivity analysis of a finite size CO2
for a given configuration of transmitter and receiver. plume with uniform or patchy saturations. The plume
The configuration of the sources and receivers, and dimensions are 2 km diameter and 40 m thickness,
the frequency of operation, need to be optimized for with its upper edge located 800 m below seafloor.
the target. CO2 is usually injected into relatively flat The plume has either uniform CO2 saturation or
lying formations and is expected to result in lenticular layered saturation (Figure 9.7b). The background
lenses or plumes of a finite size and of increased model is 1 Ohm-m. The resistivities of the CO2
resistivity, so this study will focus on the responses plume were estimated using the relationships of the
of resistive targets. section “Rock Properties and Resistivity” and
Figure 9.6 illustrates that a conductive (low resis- Figure 9.2, with 30% porosity, brine resistivity of 0.3
tivity) target produces a larger signal and therefore it Ohm-m, and 37% CO2 saturation for the uniform
is easier to detect than the resistive target. Figure 9.6a distribution, and 26%, 37%, and 85% CO2 saturations
shows two models: the resistor, which is a 100 m thick for the layered distribution. The sources are posi-
resistive layer (30 Ohm-m) at a depth of 1000 m in tioned on five north-heading tow-lines (Tx01–Tx05).
a conductive background (2 Ohm-m), and the con- The receivers (Rx001–Rx030) are 1 km apart and
ductor, which is a 100 m thick conductive layer cover the entire study area. Additional parameters in 159
(a) (b) (c) Resistor in conductive bg

Resistor in conductive bg
Resistor Conductor 2.5 Conductor in resistive bg
2.5 Conductor in resistive bg
0
|EXanom/Exbg|
|EXanom/Exbg|
2 2
1
30 Ωm 2 Ωm 1.5 1.5
1.1
1 1
2 Ωm 30 Ωm
Z (km)
0.5 0.5
0 5 10 15 0 5 10 15
Offset (km) Offset (km)
(d) (e) (f)
2.5 Conductor in resistive bg 2.5 2.5
|EXanom/Exbg|
|EXanom/Exbg|
|EXanom/Exbg|
2 2 2 Resistor in conductive bg
Conductor in resistive bg
1.5 1.5 1.5
1 1 Conductor in resistive bg 1
0.5 0.5 0.5

0 5 10 15 0 5 10 15 0 5 10 15
Offset (km) Offset (km) Offset (km)
(g) (h) (i)
Resistor in conductive bg Resistor in conductive bg
2.5 Conductor in resistive bg 2.5 Conductor in resistive bg
2.5 Conductor in resistive bg
|EXanom/Exbg|
|EXanom/Exbg|
|EXanom/Exbg|
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5

0 5 10 15 0 5 10 15 0 5 10 15
Offset (km) Offset (km) Offset (km)
Figure 9.6 (a) Canonical models of a resistor embedded in conductive background and a conductor embedded in resistive background. (b–i)
Ratios of fields for the models with the anomalous (anom) layers to the fields for the respective background (bg) models. (b) Vertical magnetic
dipole (VMD) source and field component Hx, (c) VMD and Hz, (d) x-directed horizontal electric dipole (HED) and inline Ex, (e) x-directed HED
and Ez, (f) x-directed HED and Hy, (g) y-directed HED and Ey, (h) y-directed HED and Hx, (i) y-directed HED and Hz. (Figure 1 from Streich, 2016.)
CSEM monitoring design are the frequency of the Electrical Resistance Tomography
operation and the configuration of the sources and
Electrical resistance tomography (ERT) is an indir-
receivers or the source–receiver offset, which depend
ect method for visualizing the movement of fluids
on the depth and size of the target. In this example, the
in porous media requiring the intermediate applica-
frequency is 1 Hz and results in Figure 9.8 are for the
tion of inversion algorithms that convert raw mea-
source–receiver offset of 2.5 km. The normalized elec-
surements of electrical resistance to a tomographic
tric field responses shown in Figure 9.8 are normal-
image (resistivity or concentration) of a fluid
ized to background responses at the upper left corner
plume. Figure 9.9 conceptualizes the monitoring of
(Rx005). The results show that the CO2 plume with
a CO2 reservoir using the ERT method. The place-
patchy and uniform saturations for the layered dis-
ment of the current (C1, C2) and potential (P1, P2)
tribution (Figures 9.8a, b) causes a 50% increase in the
electrode dipoles on either the surface or in bore-
electric field magnitude, while the plume with uni-
holes leads to a variety of possible ERT configura-
form CO2 saturation (Figures 9.8c, d) causes a 25%
tion types: surface-to-surface, surface-to-borehole,
160 increase of the electric field magnitude, and locates
borehole-to-surface, and borehole-to-borehole.
the plume with good lateral resolution.
usually favor the borehole-to-borehole method

Tx01 TX05 given the large application depths.
Within a monitoring context, we consider two
Rx005 RX030 types of electric field measurements. The first one
responds to the background geology, also called back-
ground field. It is recorded before the CO2 injection
starts in order to have a reference to any subsequent
reservoir state change. The second field type arises
when the CO2 enters the reservoir and changes the
resistivity of the background geology, thus called
the anomalous field. Given that CO2 behaves like an
electrical insulator, the reservoir’s bulk electrical
1 km resistivity will increase. The borehole-to-borehole
configuration in Figure 9.9 is designed to image the
CO2-saturated region between two wells. On CO2
Rx001 Tx03
Rx021 intrusion, anomalous voltage measurements will
a
translate to an increase of apparent resistivity, ρa .
Monitoring measurements, also called time-lapse
measurements, taken at regular intervals after injec-
Sea water 80 m 0.31 Ωm tion show how the evolution of the measured voltages
over time is linked to CO2 plume growth and struc-
tural change over time.
800 m The underlying physical processes that lead to the
spatial CO2 distribution over time are flow and trans-
port processes. They govern how CO2 accumulates
2000 m and travels in the subsurface and are primarily con-
trolled by two reservoir rock properties: porosity and
hydraulic conductivity. Basically, porosity governs
148/8.4/6 Ωm (P) 40 m
how much CO2 can be stored, whereas hydraulic
conductivity describes how easily the gas can travel
148/5.6/4.3 Ωm (H) through a rock’s pore space. While these parameters
are hard to measure right in the reservoir, electrical
Background 1 Ωm resistivity is an indirect indicator for their combined
effect, as it correlates with CO2 saturation changes.
Therefore, ERT time-lapse observations also provide
some capability of predicting preferential flow paths,
b
which may be helpful for risk assessment purposes.
Figure 9.7 (a) CSEM survey configuration. Receiver’s (Rx001–

Rx030) locations are shown by black circled cross markers. Gray lines Inverse Modeling
show five north-heading towlines (Tx01–Tx05). Footprint of the CO2
plume is shown by a large black circle. (b) CO2 plume model.
Resistivities of different gray shades of the CO2 plume are given on
Mathematical Principles of Inverse
the left, the background resistivity is 1 Ohm-m, and sea water
resistivity is 0.31 Ohm-m. (Figure 1 from Bhuyian et al., 2011.)
Modeling
Recall the following modeling steps:
1. Assume a certain subsurface structure by
The placement of the dipoles depends on survey establishing and quantifying the relevant
economics, logistics, and sensitivity to the target. parameters that make up this structure.
While forward modeling studies are typically con- 2. Simulate the underlying physical processes that
ducted to find the optimal layout, GCS scenarios are sensitive to these structure parameters. 161
Figure 9.8 Normalized

Normalized magnitude at 1.0 Hz Normalized magnitude at 1.0 Hz electric field responses for
Layered-patchy, offset = 2500 m Layered-homogeneous, offset = 2500 m (a) patchy and (b) uniform
saturations for layered
distribution, and (c) patchy
and (d) uniform saturations in
nonlayered distribution.
(Figure 3 from Bhuyian et al.,
2011.)
1.5 1.5
1.25 1.25
1.0 1.0
a b
0.75 0.75
Normalized magnitude at 1.0 Hz Normalized magnitude at 1.0 Hz

No layer-patchy, offset = 2200 m No layer-homogeneous, offset = 2500 m
1.5 1.5
1.25 1.25
1.0 1.0
0.75 c 0.75 d
Surface current electrodes Figure 9.9 Illustration of ERT

survey configurations within a
Surface
CO2 injection monitoring sce-
A potential
nario. The electrode pairs on the
electrodes V
left indicate current dipoles (C1,
C2), which can be placed either at
the surface or in a borehole.
C1 C2 CO2 P1 P2 The electrode pairs on the right
indicate potential electrodes (P1,
C1 P2), which can also be located
C2 either at the surface or in
a borehole. Electric fields are
A generated by the current dipole
P1 and are measured at the potential
CO2
P2 V dipole. Increasing CO2 saturations
Borehole
Reservoir in a reservoir cause anomalous
current electrodes
Borehole electric field measurements with
potential respect to the preinjection state.
162 electrodes
3. Predict the measurements that would be produced

1 λ 1
by these simulations. ϕðmÞ ¼ ϕd þ ϕm ðmÞ ¼ k Wd ðdobs F½mÞk2
4. Compare these predictions to actual real-world 2 2 2
λ
measurements. þ k Wm ðm mref Þk2 : ð9:2Þ
2
We will use them to outline the mathematical basics of
geophysical inverse modeling problems. The model- The second measure ϕm penalizes deviations from
ing step (1) involves the definition of a model para- a given reference model mref, thus encouraging model
meter vector m, to be fed into a modeling operator solutions that stay close to mref. Further, the matrix
F in step (2), which in practice is a simulation algo- Wm is constructed in a way that penalizes model
rithm in form of a computer code. The simulation roughness, that is, the product Wm becomes large in
step (3) delivers the final modeling product F(m), the the presence of large property contrasts between spa-
data predictions at a number of predefined receiver tially adjacent model parameters. The variable λ is
locations. To assess the quality of fit to an actually called the regularization parameter, providing
measured data vector dobs of N observations in step a means of finding a balance between the influence
(4), the misfit measure ϕd is commonly used: of ϕd and ϕm on the inverse solution.
Given a sufficiently simplified model
N obs 2
1X di Fi ½m 1 parameterization m that still honors the expected
ϕd ¼ ¼ k Wd ðdobs F½mÞk2 spatial property variation in the subsurface to
2 i¼1 εi 2
a sufficient degree, we can now solve the inverse
ð9:1Þ problem in a stable manner by minimizing ϕ subject
where the diagonal matrix Wd contains each datum’s to a predefined target misfit. In practice, this can be
standard deviation, εi. accomplished by setting the derivative of Eq. (9.2) to
Two major obstacles may hinder the process of zero. The resulting expression is then solved for the
finding an adequate solution m that minimizes the solution vector m. This process varies depending on
misfit ϕd. The first is related to a model’s level of detail. the nature of the forward modeling operator F, speci-
To honor true nature, that is, to completely model fically whether F is a linear or nonlinear function.
spatial property variations on even the smallest scale, Limiting this section to only an introduction, we
the size of m would be infinite. In contrast, geophysical highly recommend the tutorial by Oldenburg and Li
surveys can only provide observations that have a finite (2005) for both a more in-depth study and for a review
number of data points. This dilemma is referred to as of further readings on the broad subject of geophysi-
the nonuniqueness problem, because one may find cal inverse modeling.
many different vectors m with high granularity that
provide a fit to a data set dobs with insufficient spatial
coverage. The second obstacle is related to solution Field Data Time-Lapse Inversion Example
stability. A solution may not depend continuously on The inversion outcome of time-lapse borehole-to-
the data. In other words, m can exhibit large property borehole electrical field data from a pilot CO2 seques-
contrasts caused by only small variations in the data. tration experiment is summarized in Figure 9.10.
Other data deficiencies given by excessive noise or poor The experiment took place at the Cranfield site near
sensitivity to the physical property of interest aggravate Natchez, Mississippi, and involved the injection of
these difficulties. more than 1 million metric tons of CO2 into a deep
These two issues, the nonuniqueness and solution subsurface reservoir (approx. 3000 m). The recovered
instability, render geophysical inverse problems as change in resistivity is plotted as a function of distance
generally ill-posed. A common remedy is the incor- from the injector (x-axis) and depth (y-axis) for eight
poration of additional model knowledge into the solu- different days over a period of 103 days. Two mon-
tion process by means of a regularizing function, ϕm. itoring boreholes, marked as F2 and F3, were sepa-
Prior model information can be cast into mathema- rated by 30 m, with their ERT electrode locations
tical constraints in various ways. Here, we augment ϕ indicated by black dots. A string of 14 electrodes
by a so-called penalty term, arriving at the Tikhonov with 4.6 m vertical spacing and 61 m total length
functional (Tikhonov and Arsenin, 1977), was deployed in F2. Instrument failure led to only 163
BERT EMGeo BERT EMGeo

a Day 13 b Day 13 i Day 53 j Day 53
3160 3160 3160 3160
3180 3180 3180 3180

Depth [m]
Depth [m]
3200 3200 3200 3200
3220 3220 3220 3220
F2 F3 F2 F3 F2 F3 F2 F3
60 80 100 120 60 80 100 120 60 80 100 120 60 80 100 120
c Day 21 d Day 21 k Day 63 l Day 63

3160 3160 3160 3160
3180 3180 3180 3180

Depth [m]
Depth [m]
3200 3200 3200 3200
3220 3220 3220 3220
F2 F3 F2 F3 F2 F3 F2 F3
60 80 100 120 60 80 100 120 60 80 100 120 60 80 100 120
e Day 33 f Day 33 m Day 83 n Day 83

3160 3160 3160 3160
3180 3180 3180 3180

Depth [m]
Depth [m]
3200 3200 3200 3200
3220 3220 3220 3220
F2 F3 F2 F3 F2 F3 F2 F3
60 80 100 120 60 80 100 120 60 80 100 120 60 80 100 120
g Day 43 h Day 43 o Day 103 p Day 103

3160 3160 3160 3160
3180 3180 3180 3180

Depth [m]
Depth [m]
3200 3200 3200 3200
3220 3220 3220 3220
F2 F3 F2 F3 F2 F3 F2 F3
60 80 100 120 60 80 100 120 60 80 100 120 60 80 100 120

distance from injector [m] distance from injector [m] distance from injector [m] distance from injector [m]
-10 0 +10 +20 +30 +40 +50

Change in resistivity [%]
Figure 9.10 Comparative 3D ERT inversion results of a deep CO2 injection. The time-lapse ERT measurements span a period of over
100 days and involve a borehole-to-borehole configuration. The two different inversion algorithms BERT (images in columns 1 and 3) and
EMGeo (images in columns 2 and 4) indicate a similar plume evolution over time, ultimately reducing ambiguity in assessing preferential
flow paths.
164
seven deployable electrodes, hence the doubled verti- finding. This experiment demonstrated that even in
cal spacing in F3. The CO2 injector (F1) was located in the situation when the reservoir is relatively deep
another well 70 m to the left of F2 (not shown). The and geophysical measurements become logistically
dipole–dipole switching was scheduled in a way that demanding, this monitoring approach is feasible and
current (transmitting) and potential (receiving) elec- provides useful information about CO2 movement.
trode pairs moved along both boreholes, leading to
data with both pairs in one of either well, and data Advantages and Limitations of
with the electrode pairs across the two wells.
The field measurements were inverted using two Electrical and EM Techniques
different algorithms. Each inversion algorithm features Electrical and EM techniques are complementary to
a different kind of forward modeling engine as well as seismic monitoring because resistivity has a different
a different optimization driver. In this particular case, dependence on CO2 saturation than seismic velocity.
the usage of different inversion methods was motivated There is the potential in CSEM for providing impor-
by the prospect of reduced ambiguity: a similar model tant complementary information on its spatial distri-
parameter estimation produced by two different algo- bution in the subsurface (e.g., Eiken et al., 2000;
rithms might indicate a higher degree of geological Hoversten et al., 2003; Gasperikova and Chen, 2009).
meaningfulness, or less ambiguity. The employed 3D images of subsurface resistivity can be produced
methods are the inversion code BERT (Boundless routinely thanks to simultaneous development of
Electrical Resistivity Tomography) (Ruecker et al., acquisition hardware, computers, processing, and
2006) and EMGeo (ElectroMagnetic Geological map- modeling and inversion algorithms (e.g., Grayver
ping software) (Commer et al., 2011, 2016). For each of et al., 2014 and references within).
the eight selected days, BERT results are shown on the CO2 and all petroleum fluids (e.g., oil, conden-
left and EMGeo results are shown on the right. sate, and hydrocarbon gas) are electrically resistive;
The principal difference in their forward modeling therefore, many approaches developed for oil or gas
methods is the volume discretization, i.e., the process reservoir monitoring are applicable to CO2 monitor-
of transferring a continuous model into discrete ele- ing. However, because the resistive targets produce
ments. BERT uses the finite-element method, here much smaller signals than the conductive targets and
characterized by the mesh elements with tetrahedral are more difficult to detect, the choice of the techni-
geometry, while EMGeo is a finite-difference method, que will depend on the CO2 plume size/volume and
where typical volume discretizations are given by reg- its depth. Gasperikova and Morrison (2019) present
ular meshes of cubic cells. Each finite-element or finite- detectability analyses using a thin resistive layer and
difference mesh cell is assigned a Δ-parameter, repre- sphere models for a range of techniques. Although
senting the change of electrical resistivity with respect unlikely, the increased reservoir pressure can lead to
to the preinjection state. Recall that increased CO2 leakage of reservoir brine (high-salinity water) or
saturation causes a resistivity increase that further CO2 into shallow aquifers (Birkholzer et al., 2009).
translates into positive Δ-values, shown as a percent High-salinity reservoir brine can therefore increase
change in Figure 9.10. salinity levels in groundwater, which would be char-
Regarding the main features of the Δ-parameter acterized by a strong decrease in electrical resistivity,
distributions, the image sequences of the two differ- caused by the dissolved salt. In such a scenario,
ence methods show a similar CO2 plume development electrical techniques could be used to locate these
over time. These inversion results further confirmed conductive anomalies. The electrical signature of
a curiosity that emerged in an earlier ERT data inter- dissolved CO2 in groundwater can be either resistive
pretation (Doetsch et al., 2013) and also showed in gas or conductive, which creates challenges for detecting
and tracer concentration measurements near the F3 dissolved CO2 in freshwater aquifers. Fluid salinity
well (Lu et al., 2012): CO2 transport between F1 and and the complex geochemical reactions of CO2 with
F3 occurred faster than between F1 and F2, which is these fluids would have to be considered in the
likely due to differential flow paths in this reservoir electrical monitoring design.
region. The early onset (starting at day 21) of elevated Crucial requirements for monitoring are sufficient
resistivities near F3 in Figure 9.10 also supports this accuracy, repeatability of the measurements, and 165
sufficient sensitivity to the subsurface changes. Data Bhuyian, A. H., Ghaderi, A., and Landro, M. (2011). CSEM
errors have to be significantly smaller than the EM sensitivity study of CO2 layers with uniform versus
field anomalies resulting from changes within the patchy saturation distributions. Expanded Abstracts,
Society of Exploration Geohysicists Technical Program,
target structure. Repeatability errors may accumulate
655–659.
through repositioning errors of the acquisition equip-
ment, hardware changes, or temperature effects influ- Birkholzer, J. T., Zhou, Q., and Tsang, C.-F. (2009). Large-
scale impact of CO2 storage in deep saline aquifers:
encing hardware performance over time. A sensitivity study on pressure response in stratified
Data acquisition and monitoring on land face systems. International Journal of Greenhouse Gas
different challenges compared to marine environ- Control, 3: 181–194.
ments. Increasing levels of cultural noise pose a grow- Christensen, N. B., Sherlock, D., and Dodds, K. (2006).
ing challenge to EM surveys. The future success of Monitoring CO2 injection with cross-hole electrical
EM applications may thus depend on improving the resistivity tomography. Exploration Geophysics, 37:
techniques for handling the various types of noise 44–49.
encountered. The influence of noise may be reduced Commer, M., Newman, G. A., Williams, K. H., and
by recording field components less affected by noise Hubbard, S. S. (2011). Three-dimensional induced
or by defining source geometries and signals such that polarization data inversion for complex resistivity.
they are optimally separable from the noise. Further Geophysics, 76: F157–F171.
noise reduction may be achieved by developing Commer, M., Doetsch, J., Dafflon, B., Wu, Y., Daley, T. M.,
advanced processing schemes that exploit some and Hubbard, S. S. (2016). Time-lapse 3-D electrical
a priori knowledge of the noise at hand. Certain resistance tomography inversion for crosswell
types of signal undesired at the outset, such as the monitoring of dissolved and supercritical CO2 flow at
two field sites: Escatawpa and Cranfield, Mississippi,
effects of well casings, may have to be explicitly USA. International Journal of Greenhouse Gas Control,
accounted for in data interpretation. 49: 297–311.
Doetsch, J., Kowalsky, M. B., Doughty, C., et al. (2013).
Acronyms and Abbreviations Constraining CO2 simulations by coupled modeling
B magnetic field and inversion of electrical resistance and gas
BERT Boundless Electrical Resistivity composition data. International Journal of Greenhouse
Gas Control, 18: 510–522.
Tomography
CSEM controlled source electromagnetics Eiken, O., Brevik, I., Arts, R., Lindeberg, E., and Fagervik, K.
(2000). Seismic monitoring of CO2 injected into
CO2 carbon dioxide
a marine aquifer. Expanded Abstracts, Society of
DC direct current Exploration Geophysicists Technical Program,
E electric field 1623–1626.
EM electromagnetic
Gasperikova, E., and Chen, J. (2009). A resolution study of
EMGeo ElectroMagnetic Geological mapping non-seismic geophysical monitoring tools for
software monitoring of CO2 injection into coal beds. In L. I. Eide
ERT electrical resistance tomography (ed.), Carbon dioxide capture for storage in deep geologic
H magnetic field (B ¼ μH) formations. Results from the CO2 Capture Project, Vol. 3:
HED horizontal electric dipole Advances in CO2 capture and storage technology.
TDS total dissolved solids Berkshire: CPL Press, 403–420.
VMD vertical magnetic dipole Gasperikova, E., and Hoversten, G. M. (2006). A feasibility
study of nonseismic geophysical methods for
monitoring geologic CO2 sequestration. Leading Edge,
References 25: 1282–1288.
Archie, G. E. (1942). The electrical resistivity log as an aid in Gasperikova, E., and Morrison, H. F. (2019). Electrical
determining some reservoir characteristics. and electromagnetic techniques for CO2 monitoring.
Transactions of the AIME, 146: 54–62. In L. Huang (ed.), Geophysical monitoring for
geologic carbon sequestration. Hoboken, NJ: AGU
Bergmann, P., Schmidt-Hattenberger, C., Kiessling, D., et al. Wiley.
(2012). Surface-downhole electrical
resistivity tomography applied to monitoring of CO2 Grayver, A. V., Streich, R., and Ritter, O. (2014). 3D
166 inversion and resolution analysis of land-based CSEM
storage at Ketzin, Germany. Geophysics, 77: B253–B267.
data from the Ketzin CO2 storage formation. Oldenburg, D. W., and Li, Y. (2005). Inversion for applied
Geophysics, 79: E101–E114. geophysics: A tutorial. In D. K. Butler (ed.),
Gueguen, Y. (1994). Introduction to the physics of rocks. Investigations in geophysics, No. 13: Near-surface
Princeton, NJ: Princeton University Press. geophysics. Society for Exploration in Geophysics,
89–150.
Hagrey, S. A. (2012). 2D optimized electrode arrays for
borehole resistivity tomography and CO2 sequestration Rucker, C., Gunther, T., and Spitzer, K. (2006). Three-
modelling. Pure and Applied Geophysics, 169: dimensional modelling and inversion of dc resistivity
1283–1292. data incorporating topography. I. Modelling.
Geophysical Journal International, 166: 495–505.
Hoversten, G. M., Gritto, R., Washbourne, J., and Daley, T.
(2003). Pressure and fluid saturation prediction in Schmidt-Hattenberger, C., Bergmann, P., Bosing, D., et al.
a multicomponent reservoir using combined seismic (2013). Electrical resistivity tomography (ERT) for
and electromagnetic imaging. Geophysics, 68: monitoring of CO2 migration: From tool development
1580–1591. to reservoir surveillance at the Ketzin pilot site. Energy
Procedia, 37: 4268–4275.
Lu, J., Kharaka, Y. K., Thordsen, J. J., et al. (2012).
CO2rock-brine interactions in Lower Tuscaloosa Streich, R. (2016). Controlled-source electromagnetic
Formation at Cranfield CO2 sequestration site, approaches for hydrocarbon exploration and
Mississippi, U.S.A. Chemical Geology, 291: 269–277. monitoring on land. Surveys in Geophysics, 37: 47–80.
Morrison, H. F., and Gasperikova, E. The Berkeley Telford, W. M., Geldart, L. P., and Sheriff, R. F. (1990).
course in applied geophysics (interactive textbook). Applied geophysics. Cambridge: Cambridge University
http://appliedgeophysics.berkeley.edu Press.
Nakatsuka, Y., Xue, Z., Garcia, H., and Matsuoka, T. (2010). Tikhonov, A. V., and Arsenin, V. Y. (1977). Solution of ill-
Experimental study on CO2 monitoring and posed problems. Hoboken, NJ: John Wiley & Sons.
quantification of stored CO2 in saline formations using Zhou, B., and Greenhalgh, S. A. (2000). Cross-hole
resistivity measurements. International Journal of resistivity tomography using different electrode
Greenhouse Gas Control, 4: 209–216. configurations. Geophysical Prospecting, 48: 887–912.
167
Chapter
Microseismic Imaging of CO2 Injection
10 Shawn Maxwell
Passive microseismic monitoring is commonly used fractures with length dimensions of the order of 10s of
to image the geomechanical impact from a variety of meters. Larger faults could potentially also be asso-
industrial activities, including fluid or gas injection ciated with slip which leads to potential questions
(e.g., Maxwell et al., 2010). Stress or pressure changes around induced seismicity and possible seismic
during injection can result in either activation of hazard (Zoback and Gorelick, 2012).
preexisting fractures or potentially creation of new There are three main motivations for microseis-
fractures, both of which can result in induced seismi- mic monitoring of gas injection and geosequestration:
city associated with inelastic deformation of existing
1. Imaging the injection by detecting pressure-
fractures or the rock itself. Passive seismic monitoring
induced fracture movements
can be used to detect the transient seismic waves of
2. Monitoring for seal integrity
these sources, which can then be processed for loca-
3. Monitoring for induced seismicity
tion and source intensity, or magnitude, along with
other characteristics. Induced seismicity follows the These three objectives represent similar geomechani-
same Gutenberg–Richter law as tectonic earthquakes, cal processes, but with different scales and risk impli-
a power-law relationship between number of events cations to the injection.
and magnitude such that there is approximately an In this chapter, microseismic details will be
order of magnitude increase in the number of sources described, including geomechanics and source theory,
for one magnitude unit decrease in magnitude. acquisition, processing, and interpretation. Several
Therefore, there will be a relatively large number of case studies will also be presented, highlighting field
small magnitude sources, which are typically referred example of microseismic monitoring of geosequestra-
to as “microseismicity” generally with negative mag- tion and other gas injections.
nitudes (below zero) on the seismic magnitude scale.
The spatial and temporal characteristics of microseis- Geomechanics and Microseismic
micity can then be used to image the geomechanical
deformation and hence the subsurface characteristics Sources
of an injection. The classic investigation of brittle fracture stability
Geomechanical processes associated with micro- relies on a Mohr circle depiction of the stress state
seismicity during an injection operation include both (Hubbert and Rubey, 1959). Mohr circles represent
pressure changes and poroelastic stress changes. Pore the stress state on arbitrarily oriented fractures,
pressure increases resulting from injection can release decomposed into shear and normal stress compo-
the clamping stress on fractures and induce shear slip nents (Figure 10.1). Principal stress magnitudes cor-
(Hubbert and Rubey, 1959). Increased pore pressure respond to where the circle crosses the normal stress
can also lead to volumetric expansion and resulting axis (horizontal) such that the center of the circle is
local stress changes that can also lead to fracture positioned around the average stress and the radius
movement (Segall and Lu, 2015). During injection, of the circle is proportional to the deviatoric stress
including compressible gas as in the case of carbon difference between maximum and minimum stress.
dioxide (CO2) sequestration, direct pressure induced Conditions for fracture stability resulting in brittle
fracture slip is typically considered the main mechan- failure, frictional sliding, or slip of preexisting frac-
168 ism. Microseismic events are normally associated with tures, can be established using a specific failure
τ τ Figure 10.1 Mohr circle (red)

Slope Slope representation of stress state, relative to a
Mohr–Coulomb failure criterion. Left side
μ μ
shows an initial scenario of a stable
condition where the Mohr circle is to the
right of the failure or slip conditions. Right
C C side shows a perturbation of increased
pressure (P) causing slip on fractures in a
preferred direction by reducing the
σn σn
T σ3 σ1 T σ3 σ1 effective normal stress.
criterion, commonly the Mohr–Coulomb criteria for fractures. At these elevated pressures where the
frictional stability using frictional and cohesive Mohr circle is shifted to the left into negative normal
strength. When the Mohr circle intersects the failure stress conditions for tensional fractures, a signifi-
criteria, the combination of normal clamping force cantly higher rate of microseismic events would be
and shear slipping force will be such that a fracture in expected, with a faster growing pressure plume and
the appropriate orientation will undergo frictional more fractures encountered in orientations suscepti-
sliding or slip. ble to slip. Indeed, microseismic imaging is a common
During injection, increased pressure will reduce technology to map hydraulic fracture networks for
the effective normal stress and cause the Mohr circle stimulation treatments of geothermal and oil and gas
to translate to the left toward the failure condition wells, where the abundance of small magnitude
(Figure 10.1). Unless in a tectonically active region, microseismicity can provide a detailed image of the
initial stresses should be stable with the Mohr circle to resulting fracture network (e.g., Maxwell et al., 2010).
the right of the stability condition. During injection, In contrast, where injection operations are for
elevated pressures could lead to induced slip on pre- enhanced oil recovery (EOR) to either sweep pore-
viously stable fractures. Alternatively, poroelastic space hydrocarbon through flow of injected fluids,
stress changes can change the Mohr circle radius injection pressure would be locally offset by pressure
and increase the slip potential. Slip conditions would reductions associated with oil production. Therefore,
depend on the initial stress state and in situ pressure a relatively low microseismicity rate would be
changes, and orientations of preexisting fractures and expected during EOR injections.
their strength characteristics. It is important to note The most fundamental microseismic source char-
that fractures in a certain orientation will tend to slip acteristic is origin time and hypocentral location, and
under shear, a preferred direction being at an angle of indeed are important to microseismic imaging. The
approx. 30° to the maximum principal stress direc- so-called Brune source model (Brune, 1970) is often
tion. Fracture slip can then be associated with the used to characterize source deformation attributes
occurrence of a microseismic source, although the and can be related to the physical slip geometry
slip can potentially be aseismic (without detectable (Figure 10.2). The Brune model primarily uses two
seismicity), particularly if the slip is relatively slow signal attributes to quantify the source deformation:
such as cases with a small clamping force where aseis- dominant frequency of ground velocity is inversely
mic, stable slip can occur instead of stick-slip behavior proportional to the source radius of the slip (r) and
associated with seismicity. amplitude is proportional to the source strength nor-
Geosequestration and other fluid disposal or mally measured as the seismic moment (M0 in units of
storage operations typically involve a relatively low Nm), defined as
injection pressure. In contrast, however, hydraulic M0 ¼ μAd
fracturing operations involve a high injection pres-
sure slightly higher than the minimum principal where μ is the shear modulus, A is area of slip, and d 169
stress and above the pressure to create new tensile is the slip displacement. The logarithm of seismic
Shawn Maxwell
Figure 10.2 Schematic representation of shear slip associated with a microseismic event shown as two blue disks, with a displacement
(d) and area (A). The displacement creates a particular radiation pattern depicted as the four pulses in each surrounding quadrant where the
P-wave first motion changes polarity based on whether the rock experiences a compression (“push”) or dilation (“pull”). The radiation pattern is
represented by a “beach-ball” diagram representing a stereonet plot where the P-wave first motion is appropriately shaded.
Figure 10.3 Plot of magnitude versus fault size, and

corresponding amounts of slip (blue line) for
different stress release estimates. Fault characteristics
of a typical range of microseismicity are shown in
green and orange. A magnitude 4 earthquake is
shown for reference. (Modified after Zoback and
Gorelick, 2012, Figure 10.2.)
moment is used to compute a “moment magnitude” increase in earthquake occurrences in the central
(Mw), which is calibrated to be similar to the well- United States attributed to salt water disposal.
known “Richter magnitude” scale: Investigating the relative number of microseismic
M ¼ 2=3logðM0 Þ 6 events at different magnitudes is a common analysis
technique (Aki, 1965). The frequency–magnitude
The relationship between small magnitude microseis- relationship typically follows a power law as described
micity and larger earthquakes from both induced through the Gutenberg–Richter law:
and tectonic origins is of particular interest and
part of a seismology topic of “source scaling” (e.g., logðNÞ ¼ a bM
Abercrombie, 1995). Small and larger magnitude where N is the number of events. The slope is referred
events generally follow similar scaling, such that the to as the “b-value” and is typically approximately 1 for
larger the magnitude, the larger the source radius tectonic earthquakes but may be closer to 2 for fluid
(Figure 10.3). Slip on a larger fault could then be related induced seismicity. Equally important, but less
expected to have the potential of a larger magnitude commonly discussed is the “a-value” proportional to
induced seismic event (e.g., Zoback and Gorelick, the intrinsic microseismic activity rate. The a-value or
170 2012). Induced seismicity is a growing topic of con- activity rate varies significantly from site to site and
cern associated with fluid injections, following an with injection type. From an imaging perspective, the
lower the a-value the greater the challenge of record- three orthogonal orientations to record the complete
ing sufficient microseismicity to image the injection. seismic wavefield. Geophones are a common and
However, from an induced seismicity perspective a inexpensive sensor that typically respond to seismic
lower a-value relates to lower seismic hazard. Higher waves in the frequency range of 10’s of Hz and greater
b-values will also tend to relate to lower seismic and therefore are a convenient transducer.
hazard where the total seismic deformation takes Geophones can be deployed in deep or shallow
place with an abundance of smaller magnitude micro- boreholes or on surface in 2D patches, lines, or a grid
seismicity and proportionally fewer large-magnitude of single sensors (see Maxwell, 2014, for specifics). In
events. some instances, geophones permanently deployed for
A seismic source associated with a shear slip will time-lapse seismic surveys are used for passive moni-
also result in a specific radiation pattern of seismic toring between active surveys (see Chapter 15 by White,
waves (Jost and Hermann, 1989). Two types of seismic this volume). The number of sensors deployed on sur-
body waves will result from a seismic source: a com- face is variable, but in some cases very large channel
pressional P-wave and a slower traveling shear s-wave. counts (1000+ sensors) are used to stack and detect
The P-wave first motion and relative amplitude of low-amplitude signals that would otherwise be buried
each wave radiated by the source in different direc- in background noise (Duncan and Eisner, 2010).
tions depend on the seismic wave propagation angle However, these stacked surface arrays can have false
relative to the direction of slip and can be used to triggers and misidentify possible microseismic sources.
construct a “beach-ball diagram” or fault-plane solu- Downhole monitoring is also common in observation
tion (Figure 10.2). Such a fault-plane analysis can be wells, either cemented in place or temporarily clamped
used to infer the orientation of a slip plane (also a to casing. Geophones can also be deployed behind
secondary auxiliary plane orthogonal to the primary casing but requires careful planning during drilling,
slip plane results in an identical radiation pattern, or in the annular space behind tubing. These config-
creating a nonunique identification of which plane urations allow deployment in an injection well without
actually slipped) associated with individual microseis- the need for dedicated observation well, but often result
mic sources. In some instances, multiple events are in elevated background noise and loss of sensitivity to
grouped together in order to construct a composite record small-amplitude events. Regardless of the speci-
fault plane solution. In addition to constraining the fics of the well deployment, a number of 3C geophones
fault plane, the classic beach-ball diagram is also use- will typically be used so that there is sufficient array
ful to define the stress regime as normal/extensional, aperture to accurately estimate microseismic source
strike-slip or thrust/compressional. The beach-ball locations. There is potentially benefit monitoring
construction fundamentally assumes a shear slip microseismic from fiber optic, but the technology is in
source, but can be generalized using a “moment its infancy.
tensor inversion” to include other potential source While selecting a sensor type and corresponding
types including fracture dilation. array geometry for a specific application, the first
question should be related to the ability to address
the motivation for the monitoring. For example, will
Microseismic Acquisition passive microseismic imaging be used to image the
Microseismic data acquisition typically relies on a injection and potential loss of containment, or is it
digitizer connected to some type of seismic transdu- for seismic hazard? Once defined, subsequent seismic
cer, most commonly a geophone, seismometer, or array design questions commonly involve
accelerometer. Seismometers are useful for surface
1. Should a surface or a borehole array be used?
networks for monitoring induced seismicity, having
2. How many sensors to use and at what spacing?
the sensitivity and frequency response to accurately
characterize larger seismic events. Accelerometers are 3. How closely to the injection point should borehole
useful to quantify the ground motion associated with sensors be deployed and over what distance can
induced seismicity for seismic hazard applications. microseismicity be detected?
However, for monitoring small-magnitude microseis- Answers to these specific questions can be gained
micity, arrays of three component (3C) geophones through an analysis of the array performance 171
are most commonly used consisting of geophones in (Maxwell et al., 2003a), including comparisons of
Shawn Maxwell
logistics and costs of each monitoring option along 1. Detection of a microseismic signal
with estimates of the technical monitoring capabil- 2. Estimation of the hypocentral location
ities. Given a specific background noise level and 3. Computation of source attributes including
the intrinsic seismic attenuation characteristics, the magnitude
array sensitivity or smallest detectable seismic signal There are two general workflows that can be used for
amplitude and associated moment or magnitude can either surface or downhole data (see Maxwell, 2014,
be computed. If these parameters are not available, a and references therein for additional details). The
sensitivity study of potential characteristics can be classic method was adopted from earthquake meth-
performed. Seismic amplitudes decrease with distance ods, involving detection of a coherent signal above
according to the seismic attenuation characteristics, background noise followed by estimating the arrival
such that the smallest detectable magnitude will times of the seismic wave at each sensor and perform-
increase with distance. The probable magnitude ing an arrival-time inversion. Arrival-time inversion
range of recorded microseismicity is another involves finding the location that minimizes the time
unknown factor that ultimately will control the num- difference between the observed arrival times and
ber of events of the sensitivity limits and equally how computed arrival times by ray tracing through a velo-
far away from the array events will be detected. city model. Associated location uncertainties can
Magnitudes from injection in analogue scenarios can be determined through the arrival-time inversion,
be used as an estimate, or a geomechanical assessment resulting in uncertainties in different directions
can be made to compute fracture slip conditions and often represented by an error ellipsoid. Arrival times
likely microseismic magnitudes. can be automatically determined, or manually picked
Similarly, microseismic location uncertainties through visual inspection. For single monitoring
associated with a particular array can be estimated, wells, signal polarization of the 3C sensors is also
which is useful in setting expectations for the robust- used to constrain a unique location. An alternative is
ness of eventual microseismic data sets and ensuring a source scanning or “migration” method, which gen-
that the microseismicity will be sufficiently accurate erally combines detection and location. A series of
to resolve the intended targets. For example, depth locations are assessed by stacking the recorded signals
accuracy will be needed to a sufficient level to identify with time delays equivalent to differences in the com-
if microseismicity is occurring out of the injection puted arrival times for each location. The signals are
target interval for seal integrity applications. Lateral also sequentially scanned in time, with the stacked
and depth accuracy requirements will be scaled by the amplitude reaching a peak at a microseismic source
expected size of the injected CO2 plume, where micro- location and onset time. Generally, these methods are
seismicity is being used to image the injection. fully automated, but can be refined using arrival time
Monitoring array specifics will therefore be project picking constraints ideally once the data are manually
dependent; however, some general rules of thumb can reviewed.
be considered. The question of surface or downhole Both workflows require a seismic velocity model
monitoring will ultimately dictate the sensitivity of the to compute theoretical arrival times. Most methods
recording, where typically only larger magnitude use both p- and s-wave arrival times and so a velocity
events are detectable from a surface compared to a model is required for each. As injection is typically in
proximal downhole monitoring array. Depth location sedimentary rocks, depth varying velocity models are
uncertainties are also typically higher from a surface, needed. Velocities in each layer can be either isotropic
although tend to increase with distance from a down- or, better, anisotropic to account for layering effects.
hole array and may eventually become larger than sur- Lateral velocity variations can also be included in
face values. Horizontal uncertainties are typically these models, although often 1D models are common-
smaller for surface arrays. Surface arrays tend to be place. Initial velocity models are often constructed
ideal for monitoring induced seismicity during injec- from dipole sonic logs, but are ideally “calibrated”
tion, where larger potentially felt events are of interest. using travel time measurements from one or more
checkshots. Checkshots are most effective if located
Microseismic Processing near the expected microseismic locations and help
172 Microseismic processing typically involves three steps: verify the sonic logs, quantify anisotropy, and
mitigate any lateral velocity variability. The velocity are realistic, can be used to quantify the location con-
model construction and calibration are important fidence. Often these uncertainties are associated with
steps in microseismic processing, as inaccuracies in random errors associated with arrival time uncertain-
the velocity model can result in mislocation of the ties input to the location algorithm. Systematic location
microseismicity. Sensitivity studies can be made with uncertainty associated with the velocity model is
different velocity models to quantify the potential equally important and can be quantified through
mislocation. a velocity model sensitivity study, but is often not
Once detected and located, other microseismic undertaken. It is critical to properly interpret the
source characteristics can be estimated. The most microseismic locations to quantify the entire location
important is seismic moment, which can be esti- uncertainty, particularly when interpreting the depth
mated from a displacement spectrum of the recorded of the microseismic sources in relation to stratigraphic
signal. Beach-balls or source mechanisms can also be layers. Other QC attributes such as SNR can also be
determined, although often single monitoring wells useful to potentially identify a high confidence subset
are used that are unable to uniquely constrain the of the data and mitigate possible false triggered or
source mechanism for a single event. Although use- misidentified signals. Recording biases where the mini-
ful information, source mechanisms are not routi- mum detected magnitude varies in space and time are
nely determined. important to quantify and if possible to normalize. For
After an initial processing, advanced techniques downhole monitoring, a plot of magnitude versus dis-
can also be used to improve attributes such as location. tance between the monitoring stations and microseis-
One of these methods is a relative location technique, micity is useful to define a distance sensitivity bias
where arrival time differences between events are used and how far away events can be detected. A minimum
to estimate a location difference between sets of events. magnitude can be established, for instance, so that only
These techniques tend to improve precision although events above that magnitude are considered. For events
not absolute location and sometimes allow underlying above the minimum detectable magnitude, an analysis
patterns in microseismic location to emerge. of the magnitude of completeness is important for
The final processing product is a “catalog” of accurate comparison of activity rates between and
microseismicity consisting of data, time, 3D location, within specific projects.
magnitude, additional source characteristics, and The starting point of the interpretation is to exam-
quality control attributes. Useful QC parameters ine the spatial and temporal locations for specific
include location uncertainty ellipsoids, which will patterns. Microseismicity often forms clusters of
vary between sources, and signal-to-noise ratio events in space and time. The temporal evolution
(SNR). SNR is useful to define the signal quality and should be compared against the injection schedule,
hence a proxy for the confidence in arrival time and for correlations with changes in injection rate or
associated location confidence. A comprehensive doc- pressures. Spatial clustering is also important, speci-
umentation of suggested deliverables has been defined fically in relation to the injection well and geological
for hydraulic fracturing applications (Maxwell and structure. The depth of the microseismicity in relation
Reynolds, 2012), the majority of which are also suita- to the injection depth in a stratigraphic profile is a
ble to CO2 sequestration applications. fundamental consideration and vital for characteriz-
ing geological containment of the CO2. Lateral clus-
Microseismic Interpretation tering of the microseismicity is also informative,
Microseismic catalogs consist of a series of events providing insights into preferred CO2 plume direc-
that require an important interpretation step to tions. Microseismicity often illuminates interactions
address the original monitoring objectives. Prior to with preexisting fractures and faults and so integra-
interpretation, it is important to assess the location tion with a geological model and 3D seismic volumes
uncertainty and potential recording biases associated can assist with understanding heterogeneities. Time-
with the sensitivity of the monitoring (see Maxwell, lapse interpretation of the microseismic sources is
2014, and references therein for additional details). also an important aspect and can be more completely
Location uncertainties should be a specific component interpreted when integrated with time-lapse seismic
of the reported data set and, provided these estimates such as 4D seismic. 173
Shawn Maxwell
Figure 10.4 Schematic depiction of pressure profiles and associated microseismicity (orange) at increasing injection time steps.
Microseismicity would be expected behind the point of furthest pressure increase that moves with time relative to the injection point as the
pressure extends. Remnant microseismicity can be expected after the pressure is first increased to the critical level. Green colored
microseismicity shows the possibility of slip on less favorably oriented fractures, associated with higher pressure and therefore would occur
closer to the injection point at later times. The edge of the injected fluid will likely be somewhere between the extreme pressure change and
the microseismicity.
It is important while interpreting microseismicity CO2 Hydraulic Fracturing

to avoid forming a direct association between the CO2
During a hydraulic fracture treatment of a gas well,
plume and any recorded activity. As discussed in the
one depth interval was treated by pumping super-
geomechanics section, microseismic sources occur
critical CO2 instead of the typical water-based gel
when localized subsurface pressures increase to a
fracturing fluid used at other depths (Verdon et al,
level sufficient to cause slip on preexisting fractures
2010). For these injections above fracturing pressure,
optimally oriented in relation to the stress field
similar microseismicity rates and magnitudes (below
(Figure 10.4). The spatial position of the CO2 plume
magnitude –3) were found for both fluid types
relative to the critical pressure increase to generate
(Maxwell et al., 2008). Injection volumes, rates, and
microseismicity depends on the initial geomechanical
viscosity of the injected fluids were similar.
conditions. The microseismicity will also be concen-
Compressible gases could be expected to generate
trated in areas where optimally oriented fractures
less pressure increase than an incompressible fluid,
exist, resulting in the possibility of spatially distinct
but in this example the CO2 would be expected to
microseismic clusters. A coupled simulation of flow
stay in a supercritical state during pumping at least
and geomechanical response to injection can be used
within the high-pressure hydraulic fracture system.
to interpret the microseismicity and predict the CO2
Nevertheless, the similarity of the microseismic
saturations to identify the plume extents. Such a simu-
responses is particularly relevant for qualifying
lation requires specific input parameters, particularly
microseismicity as an imaging tool for gas injection.
the permeability characteristics to model flow and
pressure and the stress state and fracture characteris-
tics to model strains and microseismic sources. These Talco Oilfield Natural Gas EOR Injection
parameters are often challenging to robustly charac- A microseismic monitoring project was performed
terize, but the microseismicity along with pressure while injecting natural gas for EOR in the Talco field
recording and other plume images can be used to in Texas. As existing array, composed of three shallow
validate the modeling and potentially calibrate uncer- (extending down approximately 425 m from surface)
tain input parameters. string of geophones originally intended for time-lapse
monitoring of the reservoir injection at a depth of
1200 m below surface, was utilized. Seventy-seven
Case Studies microseismic events were detected during 6 months
In this section, specific case studies are reviewed. In of monitoring (Maxwell et al., 2003a) with moment
addition to pure CO2 sequestration examples, a few magnitudes between –1.6 and 0. More microseismi-
relevant microseismic examples of monitoring gas city could have been recorded if a deeper monitoring
174 injections for other types of wellbore operations are array had been used to get better sensitivity closer to
also included. the microseismically active region. The microseismicity
was confined within the oil reservoir and generally to the injection was not associated with induced
proximal to a bounding fault. seismicity.
Aneth Oilfield CO2 EOR Injection Longyearbyen CO2 Lab

A vertical string of 3C geophones was deployed in a In preparation for CO2 injection, water injection tests
monitoring well to monitor microseismicity during have been monitored with limited microseismicity
EOR operations in the Aneth field in Utah (Soma and detected (Albaric et al., 2014). Characteristics of the
Rutledge, 2013). Microseismicity was detected at dis- recorded seismicity during the injection test were
tances between 1 and 5 km from the geophones, and used to evaluate the initial monitoring array, leading
locations were computed using arrivals from refracted to the addition of a downhole array closer to the
raypaths to supplement direct arrivals and enhance injection depth to improve sensitivity and coverage
the accuracy. Specific microseismic clusters were to detect any microseismicity associated with future
identified, one spatially correlating with the strati- CO2 injections.
graphic depth of the CO2 injection, although the
majority of the activity was found at a depth of a In Salah CO2 Project
simultaneous injection for salt water disposal.
CO2 injection at the In Salah project in Algeria was
monitored with various methods. Surface deforma-
Weyburn CO2/Water EOR Injection tion was detected from inteferometric synthetic aper-
An EOR injection in the mature Weyburn oilfield in ture radar (InSAR) (Ramirez and Foxall, 2014), a
Saskatchewan, Canada was monitored using a vertical remote sensing technique using repeated radar satel-
string of 3C geophones in an abandoned well (White, lite measurements to potentially detect millimeter
2009). Very few microseismic events have been scale surface movements. Surface deformation is a
recorded, with the exception of limited time periods useful compliment to microseismic monitoring, and
where injection switched between CO2 and water can also be used to invert for reservoir geomechanical
(Maxwell et al., 2004). Although limited in number, strains (Maxwell et al., 2008). A coupled simulation of
the microseismic cluster between an injection and flow and geomechanics was performed and indicated
production well where a CO2 production increase a probable penetration into the caprock above the
was observed following the start of injection. The injection interval.
lack of microseismicity at Weyburn can be attributed Microseismic monitoring was also performed by
to the depleted nature of the field, where production passively recording 6 of 48-3C geophones deployed
has occurred since 1955. in a downhole monitoring well (Oye et al., 2013).
Instrumentation problems resulted in reliable record-
Lacq-Rousse CO2 Capture and Storage Pilot ing of signals from only two sensors and limited
directional information. The impact of the limited
Several shallow monitoring arrays, a surface seism-
recording meant that reliable locations could not be
ometer, and downhole fiber optic sensors were used to
determined, but the data have been analyzed for com-
monitor CO2 injection into a depleted gas field in
mon signal characteristics to identify distinct spatial
France (Prinet et al., 2013). During the first two and
clusters with possible locations based on S- to P-wave
half years of injection, three microseismic events with
arrival times. Although specific locations are uncer-
magnitudes below 0 were recorded. The microseismic
tain, the analysis was able to define a proximal dis-
quiescence was attributed to the depleted nature of the
tance to the sensors and associate some of the
field.
microseismicity with the distance to the injection well.
Despite the quality of the microseismic data, the
CO2CRC Otway Project quantity of the events is significant. Approximately
CO2 injection at the Otway project in Australia was 1500 microseismic events were initially detected over
monitored using downhole sensors (Daley et al., a 22-month period, increasing to about 5000 events
2009). A small number of microseismic events were using a signal matching technique to detect weak sig-
detected and, although not sufficient to image the nals based on signal characteristics consistent with the 175
plume, did provide an indication that a fault proximal larger originally detected events (Oye et al., 2013). The
Shawn Maxwell
50
Wellhead pressure [bar]
No.of events per day
cummulative number of events x10
injection rate [mcf/day]
150 40
WHP [bar] | No. of events/day]
CO2 injection rate [mcf/day]

30
100
20
50
10
0 0
Apr 01, 10 May 01, 10 Jun 01, 10 Jul 01, 10 Aug 01, 10 Sep 01, 10 Oct 01, 10 Nov 01, 10
Figure 10.5 Comparison of injection rate and microseismicity rate for In Salah. Note the increased microseismicity rate during the initial
period with a higher injection rate prior to August 2010. Also note the two shaded time periods where the injection pressure is interpreted to
exceed the fracture pressure (pink horizontal line). (After Oye et al., 2013.)
largest moment magnitude was 1. Over and above a injection was monitored over a period between 2011
background microseismicity rate, greater activity was and 2014, during which time approximately a million
found during periods of increased injection rate and tonnes of CO2 was injected into the Mount Simon
associated pressure increases (Figure 10.5). The injec- Sandstone located above Pre-Cambrian crystalline
tion pressure was interpreted to be above the hydraulic basement. A surface seismometer array consisting of
fracturing pressure, suggesting that hydraulic fractur- 13 stations was later deployed during the last half of
ing was occurring during these periods of increased the injection period and able to record the larger
activity. Notwithstanding the processing issues related magnitude events (Kaven et al., 2015).
to acquisition problems, the In Salah project appears The preinjection monitoring resulted in identifi-
to be the first to demonstrate that an adequate amount cation of eight microseismic events with magnitudes
of microseismicity could be detected to image the between –2.16 and –1.52, suggesting little background
injection. activity near the injection point (Smith and Jacques,
2016). During the injection, a total of 4747 microseis-
mic events were detected by the downhole arrays with
Illinois Basin–Decatur Project moment magnitudes between –2.12 and 1.17 (Bauer
A comprehensive characterization and monitoring et al., 2016). The microseismic was found to be dis-
project was performed to investigate CO2 injection tributed in depth between the upper Pre-Cambrian
(see also Chapter 19 by Bauer et al., this volume). and lower Mount Simon. No significant activity was
Microseismicity was initially monitored using three found above a clay layer above the injection point,
geophones deployed in the injection well along with which is believed to act as a flow baffle and contain the
a string of the 3C geophones used for both active and injection. Laterally from the injection point, the
passive imaging (Will et al., 2014). Baseline monitor- microseismic formed several discrete clusters primar-
ing of the background microseismicity was performed ily to the north of the injection (Figure 10.6). Each
176 over a one–and-half-year period before injection cluster tended to start at a specific time with clusters
(Smith and Jacques, 2016). Three years of CO2 both at greater and lesser offsets beginning at different
times. Within each cluster, the number of microseis-

mic tended to increase to a peak and then decline.
The surface monitoring of the second half of the
injection detected 179 locatable events (Figure 10.7)
in the moment magnitude range between –1.13 and
1.26 (Kaven et al., 2015). The surface seismometer
network was not able to record relatively small mag-
nitude events, consistent with increased noise levels
and greater offset from the activity that is expected for
a surface compared to a downhole array. The depth
distribution and spatial clustering were consistent
with the results of the contemporaneous downhole
monitoring, although the hypocentral depths were
spread over a larger depth interval consistent with
larger depth uncertainty expected for the surface
array geometry. Focal mechanisms could be com-
puted for individual events and indicated an approx-
imate NE–SW strike-slip mechanism (Figure 10.8). A
similar mechanism was found with the downhole data
as a composite solution of events within specific
clusters.
Figure 10.6 Map view of downhole detected microseismicity
resulting from injection into well CCS1 at the Illinois Basin–Decatur
3D seismic volumes were analyzed for preexisting
Project showing clusters proximal to the injection. The timing of the fractures using “edge detection” methods, which indi-
onset of the microseismic clusters is labeled sequentially. (From Will cated lineation aligned with the microseismic source
et al., 2014.) (Courtesy Illinois Geological Survey.)
Figure 10.7 Map view of

1 km surface detected micro-
seismicity at the Illinois
Basin–Decatur Project
depicting additional clus-
ters detected more distal to
the injection. (From Kaven
39.90 N et al., 2015.)
Injection well
01/2015
Mw = 0.8
39.85 N 0.3
–0.4
MEQs
Surface station
Borehole station 177
07/2013
88.92 W 88.88 W
Shawn Maxwell
Figure 10.8 Fault-plane solution for

select events detected by the surface
array at the Illinois Basin–Decatur
39.90 N SHmax Project. (From Kaven et al., 2015.)
Injection well
Mw = 0.8
0.3
–0.4 1 km
88.92 W 88.88 W
mechanisms (Will et al., 2014). A flow simulation was and microseismic activity rates. Microseismic data
performed to estimate pore pressure changes result- quality can be partially addressed using reliable
ing in an approximately circular pressure increase recording equipment, and using a sensitive monitoring
centered on the injection well (Figure 10.9), which array to improve the sensitivity of the recording.
was then used in a geomechanical model to examine However, microseismic rates are clearly highly variable
fault slip conditions (Lee et al., 2014). Preexisting and depend on the injection rates and associated pres-
faults were assumed based on the microseismic cluster sures, but also the geomechanical reservoir conditions.
geometry, location, and orientation. Relatively low At sufficiently high injection pressures, CO2 can result
friction had to be assumed on these faults in order in detectable microseismicity, particularly if hydraulic
for them to slip, with frictional angles of 10°. Based on fractures are created. Even below the fracture pressure,
the modeled slip, equivalent seismic moments were microseismic can occur if in situ pressures are suffi-
estimated, but were approximately two orders of mag- cient to cause slip on preferentially oriented fractures.
nitude larger than the observed moments. Differences For some depleted reservoirs, the injection pressure
between modeled and observed seismic moments does not appear to be sufficient to generate significant
could be attributed to energy partitioning between levels of microseismicity, although this can be seen as
seismic and aseismic deformation, and dynamic favorable from an induced seismic hazard perspective.
aspects of the slip associated with individual sources. Discernible microseismicity have been reported during
The modeling also did not attempt to simulate seis- some gas EOR injections, although the reservoirs are
mically quiescent regions, which is an important also depleted and the injection is being balanced by
potential aspect to understand the spatial clustering. hydrocarbon production. Potential for microseismicity
Nevertheless, the Decatur project is the most thor- appears to be highest during injection into saline aqui-
ough to date from a microseismic perspective, with fers, although still dependent on the site specifics.
significant numbers of high-quality microseismic Although no felt seismicity has been attributed to
sources leading into an integrated interpretation. CO2 injection to date, injection induced seismicity con-
cerns are increasing with experiences in geothermal
and salt water disposal. As sequestration projects grow
Conclusion in scale with larger injection volumes, the seismic
In conclusion, passive microseismic monitoring is an hazard question will likely be under increased scrutiny.
important tool for characterizing hazard of induced With the challenges associated with sufficiently charac-
seismicity and potentially to image the injection. The terizing the injection target to confidently estimate
various published examples of microseismic monitor- seismic potential, the motivation for passive monitor-
178
ing of gas injection show highly variable data quality ing can be expected to continue. Increased monitoring
Pressure [psi]
Pressure difference
200.0
180.0
160.0
140.0
120.0
100.0
80.0
60.0
40.0
20.0
0.0
Figure 10.9 Perspective view showing modeled pressure increase contours in relation to microseismicity (black). Injection well (CCS1) is
shown in blue and the VW1 monitoring well in green. (From Lee et al., 2014.)
sensitivity can then provide details of the seismic Bauer, R. A., Carney, M., and Finley, R. J. (2016).
response to injection to better forecast future injection Surface monitoring of microseismicity at the
scenarios at a particular site. Information about gas Decatur, Illinois, CO2 Sequestration Demonstration
Site. International Journal of Greenhouse Gas
leaks and mapping the pressure plume for integra-
Control, 54: 378–388.
tion with complementary monitoring technologies
further adds to the value proposition of microseis- Brune, J. (1970). Tectonic stress and the spectra of shear
waves from earthquakes. Journal of Geophysical
mic monitoring.
Research, 75: 4997–5009.
Daley, T. M., Sharma, S., Dzunic, A., Urosevic, M., Kepic,
References A., and Sherlock, D. (2009). Borehole seismic
Abercrombie, R. E. (1995). Earthquake source scaling monitoring at Otway using the Naylor-1 instrument
relationships from –1 to 5 ML using seismograms String. LBNL-2337E report. Berkeley, CA: Lawrence
recorded at 2.5-km depth. Journal of Geophysical Berkeley National Laboratory.
Research, 100(B12): 24, 015–24, 036. Duncan, P., and Eisner, L. (2010). Reservoir
Aki, K. (1965). Maximum likelihood estimate of b in the characterization using surface microseismic
formula log N = a – bM and its confidence limits. Bulletin monitoring. Geophysics, 75. DOI:10.1190/1.3467760.
of the Earthquake Research Institute, 43: 237–239. Hubbert, M. K., and Rubey, W. W. (1959). Role of fluid
Albaric, J., Oye, V., and D. Kühn. (2014). Microseismic pressure in mechanics of overthrust faulting: 1.
monitoring in carbon capture and storage projects: Mechanics of fluid-filled porous solids and its
Fourth EAGE CO2 geological storage workshop, application to over-thrust faulting. Bulletin of the 179
Stavanger, Norway. Geological Society of America, 70: 115–166.
Shawn Maxwell
Jost, M. L., and Hermann, R. B. (1989). A student’s guide to Oye, V., Aker, E., Daley, T. M., Kuhn, D., Bohloli, B., and
and review of moment tensors. Seismological Research Korneev, V. (2013). Microseismic monitoring and
Letters, 60: 37–57. interpretation of injection data from the In Salah CO2
Kaven, O., Hickman, S. H., McGarr, A. F., and Ellsworth, W. storage site (Krechba), Algeria. Energy Procedia, 37:
L. (2015). Surface monitoring of microseismicity at the 4191–4198.
Decatur, Illinois, CO2 Sequestration Demonstration Prinet, C., Thibeau, S., Lescanne, M., and Monne, J. (2013).
Site. Seismological Research Letters, 86. DOI:10.1785/ Lacq-Rousse CO2 capture and storage demonstration
0220150062. pilot: Lessons learnt from two and a half years
Lee, D.W., Mohamed, F., Will, R., Bauer, R., and Shelander, monitoring. Energy Procedia, 37: 3610–3620.
D. (2014). Integrating mechanical earth models, surface Ramirez, A., and Foxall, W. (2014). Stochastic inversion of
seismic, and microseismic field observations at the InSAR data to assess the probability of
Illinois Basin – Decatur Project. Energy Procedia, 63: pressurepenetration into the lower caprock at In Salah.
3347–3356. International Journal of Greenhouse Gas Control, 27:
Maxwell, S. C. (2014). Microseismic imaging of 42–48.
hydraulic fracturing: Improved engineering of Segall, P., and Lu, S. (2015). Injection-induced seismicity:
unconventional shale reservoirs. SEG Distinguished Poroelastic and earthquake nucleation effects. Journal
Instructor Short Course. DOI:10.1190/ of Geophysical Research: Solid Earth, 120: 5082–5103.
1.9781560803164. DOI:10.1002/2015JB012060.
Maxwell, S. C., and Reynolds, F. (2012). Guidelines for Smith, V., and Jaques, P. (2016). Illinois Basin-Decatur
standard deliverables from microseismic monitoring of Project pre-injection microseismic analysis.
hydraulic fracturing. CSEG Recorder. https://csegrecor International Journal of Greenhouse Gas Control, 54:
der.com/articles/view/guidelines-for-standard- 362–377.
deliverables Soma, N., and Rutledge, J. T. (2013). Relocation of
Maxwell, S. C., Urbancic, T., and McClellan, P. (2003a). microseismicity using reflected waves from singlewell,
Assessing the feasibility of reservoir monitoring using three-component array observations: Application to
induced seismicity. EAGE Abstracts. CO2 injection at the Aneth oil field. International
Maxwell, S. C., Urbancic, T. I., Prince, M., and Demerling, Journal of Greenhouse Gas Control, 19: 74–91.
C. (2003b). Passive imaging of seismic deformation Verdon, J. P., Kendall, J. M., and Maxwell, S. C. (2010). A
associated with steam injection in Western Canada. comparison of passive seismic monitoring of fracture
SPE 84572. stimulation from water and CO2 injection. Geophysics,
Maxwell, S. C., White, D. J., and Fabriol, H. (2004). Passive 75. DOI:abs/10.1190/1.3304825.
seismic imaging of CO2 sequestration at Weyburn. White, D. (2009). Monitoring CO2 storage during EOR
Expanded Abstracts, Society of Exploration at the Weyburn-Midale Field. Leading Edge, July:
Geophysicists. 838–842.
Maxwell, S. C., Du, J., and Shemeta, J. (2008). Passive Will, R., Smith, V., Leetaru, H. E., Freiburg, J. T., and Lee, D.
seismic and surface monitoring of geomechanical W. (2014). Microseismic monitoring, event occurrence,
deformation associated with steam injection. Leading and the relationship to subsurface geology. Energy
Edge, 27: 260–266. Procedia, 63: 4424–4436.
Maxwell, S. C., Rutledge, J., Jones, R., and Fehler, M. (2010). Zoback, M. D., and Gorelick, S. M. (2012). Earthquake
Petroleum reservoir characterization using downhole triggering and large-scale geologic storage of carbon
microseismic monitoring. Geophysics, 75. dioxide. Proceedings of the National Academy of Sciences
DOI:10.1190/1.3477966. of the USA. DOI:10.1073/pnas.1202473109.
180
Chapter
Well Logging
11 Zaki Bassiouni
Introduction hydrocarbons may also be present, as in EOR projects

or CS in depleted hydrocarbon reservoirs.
Carbon dioxide (CO2) has been injected into under-
CO2 properties that can be used for detection and
ground geological reservoirs essentially for two main
quantification are its relatively low density, its non-
purposes: enhanced oil recovery (EOR) and the sto-
conductive nature, and its chemical composition
rage of this greenhouse gas (CGS, CO2 geological
characterized by the presence of carbon and oxygen
storage or geosequestration). In both applications,
and absence of hydrogen.
characterization of movement of CO2 within the
reservoir is critical to ensuring efficient oil recovery
in EOR projects and effective sequestration in CGS Concepts of Open-Hole Logs
projects. Monitoring the progress of injected CO2 All subsurface injection projects start with the dril-
provides insight into the extent of gravity segregation ling of boreholes, which are initially uncased,
and the degree of stratification and other permeability referred to as “open-hole.” Well logs are acquired
distributions and their effect on project management. in open-hole wells to evaluate the geological envir-
In EOR applications, changes of quantitative para- onment and estimate the reservoir potential. These
meters such as volumetric sweep efficiency and oil logs usually represent the base case in monitoring
saturation are critical to recovery optimization. operations schemes.
Several monitoring approaches may be used to Once the well is cased, open-hole logs cannot be
judge the progress of CO2 movement in the reservoir. repeated for monitoring purposes and so cased-hole
These approaches encompass surface measurements, logs are substituted. Understanding the measurement
as in seismic time-lapse techniques, and measure- and analysis of open-hole logs is critical to under-
ments performed in boreholes, such as well logs, standing all subsequent measurements taken in
pressure surveys, and the use of core samples. This cased-holes. Figure 11.1 is a common presentation
chapter is devoted to well logging methods using tools of an open-hole well log usually referred to as “resis-
sensitive to the presence of CO2. In general, monitor- tivity log.” In addition to resistivity curves traced in
ing using well logs consists of log data acquired in the middle track, the log displays several other useful
cased holes. The presence of the steel tubular string curves. The solid curve on the first track is the spon-
and the cement sheath behind it favors the use of taneous potential (SP), measured between an elec-
pulsed neutron and gamma ray tools. In newly drilled trode traveling in the borehole and a grounded
wells, such as infill and observation wells, open-hole electrode. The SP log is used to distinguish between
logs, namely formation resistivity and formation den- shales and permeable formations, in this case sand-
sity tools, are incorporated in the logging program. stones. The SP curve displays maximum voltage read-
The interpretation of the acquired logs can be ings in shales forming a shale baseline. Deflection
qualitative, i.e., limited to the detection of the pre- away from the shale baseline usually occurs at strata
sence of CO2, or quantitative, which involves the boundaries. The deflection reaches a maximum oppo-
estimation of changes in the saturation of fluids site clean or shale-free sand, in this example –100 mV.
present in the formation. At least two fluids will The top and bottom boundaries of the thick sand
always be present, formation water and CO2, as in displayed by the log are at 9975 ft. and 10,035 ft.
the case of aquifer carbon storage (CS). In some cases, respectively. The gross thickness of the sand, i.e., 181
Zaki Bassiouni
Figure 11.1 Example of a “resistivity” log,

courtesy of Schlumberger.
the interval between top and bottom geological salinity determines what well logs can be run and
boundaries is approx. 60 ft. The net, i.e., permeable plays an important role in the quantitative interpreta-
interval, thickness is estimated by subtracting the tion of resistivity logs.
thickness of shale streaks within the sand, if any Because hydrocarbons do not conduct electric
exist, from the gross thickness. current, the formation resistivity is a key hydrocarbon
Theoretical and experimental work demonstrated indicator. However, low-porosity formations and for-
that the SP deflection in clean sand is determined by mations saturated with relatively low-salinity water
the contrast of the resistivity of the formation water also display high resistivity values. To avoid mistaking
Rw , and the resistivity of the water phase of the drilling the latter formations for a hydrocarbon bearing
fluid or mud filtrate Rmf . Using available algorithms, one, a hydrocarbon indicator value Rwa is estimated
Rw , and in turn the formation water salinity, can be using formation resistivity, Rt, and porosity, ϕ, values,
182
estimated from the SP reading. Formation water (Bassiouni, 1994):
Well Logging
this particular example of an unconsolidated sand,

Rwa ¼ Rt ϕ2 ð11:1Þ the sonic curve indicates the hydrocarbon to be gas.
Rwa is also traced on the first track. In a section of The low acoustic velocity of gas results in the rela-
a well where formation water salinity and lithology tively high Dt. Yet, because the response of the SP and
remain the same, the minimum value ðRwa Þmin of the sonic logs support the premise of a uniform porosity
hydrocarbon indicator approximates the true Rw , in throughout the sand, the porosity of the hydrocarbon
this example 0.025 ohm-m at formation temperature. zone is better estimated in the water-bearing zone.
In permeable zones, Rwa values, higher than ðRwa Þmin Using a Dt value of 95 μs/ ft. together with the empiri-
indicate hydrocarbon presence even in low porosity cal values Dtma ¼ 55:5μs=ft:, and Dtf ¼ 189μs=ft:
formations. This is because the calculation of Rwa yields a porosity value of 30% which, in this case, is
makes use of porosity, thus screening out the effect representative of both the hydrocarbon and water
of low porosity. zones.
Based on the response of the resistivity curves From the log of Figure 11.1 the presence of
displayed on the second track, and the calculated a relatively thick permeable sand containing a gas-
Rwa curve, the few feet at the top of the sand display bearing interval at its top was detected. It was also
a zone partially saturated with hydrocarbons. Using possible to quantify the interval’s thickness, average
similar reasoning, the bottom zone of the sand is porosity, and average hydrocarbon saturation.
water bearing. In this analysis, the dashed resistivity The presence of CO2 will result in a similar response
curve is used as it represents deep measurement of the resistivity and sonic logs and can be analyzed in
beyond the zone affected by the borehole and the similar fashion.
mud filtrate invaded zone of the formation. Formation-fluid types are usually determined
The contrast between the resistivity, Rt , in using the type of log displayed on Figure 11.2. This
a hydrocarbon-bearing interval and the resistivity, type of log presentation is usually referred to as “por-
Ro , in a water-bearing interval is inversely propor- osity log.” The dashed curve on the first track of the
tional to the saturation of formation water, Sw . log is a caliper log approximating the diameter of the
According to Archie’s model (Archie, 1942): borehole. The caliper indicates a uniform borehole
1=2 diameter averaging 8.5 inches in the clean sand inter-
Ro vals. Borehole enlargement is shown in shale intervals,
Sw ¼ ; and SHC ¼ 1 Sw ð11:2Þ
Rt which is the result of shale sloughing. The solid curve
where SHC is the hydrocarbon saturation. on the first track is the gamma ray log, which is
Using the Rt value displayed within the hydro- a measure of the natural radioactivity of the forma-
carbon zone (15 ohm-meter or ohm-m) and the Ro tions. Studies of the origin of radioactivity in sedi-
value displayed in the water-bearing zone (0.25 mentary rocks indicate that shale is characterized by
ohm-m), the water saturation in the top of the a relatively high level of gamma ray radioactivity in
sand is 13% and subsequently the hydrocarbon contrast to clean sands. In general, the gamma ray
saturation is 87%. curve correlates reasonably well with the SP curve.
The third track displays the interval transit When formations are irradiated with gamma rays
time Dt, which is the inverse of the acoustic velo- or neutrons, the gamma ray and neutron energies are
city. Dt is a weighted average of the interval transit moderated as they travel from the source of energy to
times of the rock matrix, Dtma , and the formation the detectors of the moderated energy. The degree of
fluids, Dtf : moderation depends on the rock type, e.g., sandstone,
limestone, etc., the fluid type, e.g., gas or liquid, and
Dt ¼ ϕDtf þ ð1 ϕÞDtma ð11:3Þ the amount of fluid that is related to the formation
where ϕ is the formation porosity. Solving for the porosity.
porosity yields The two curves displayed on the second track
are the calculated density porosity ðϕD Þ and neu-
ϕ ¼ ðDt Dtma Þ=ðDtf Dtma Þ ð11:4Þ tron porosity ðϕN Þ values assuming sandstone
The model represented by Eqs. 11.3 and 11.4 was matrix and fresh formation water. The calculated
developed for water-filled formations; however, in porosity is a true value only in water-bearing 183
Zaki Bassiouni
Figure 11.2 Example of a “porosity” log,

courtesy of Schlumberger.
sandstones. Hence, the density porosity ðϕD Þ and empirical transform may be used (Bassiouni,
neutron porosity ðϕN Þ values are close in value in 1994):
liquid-filled formations and are quite different
in gas-bearing formations where ðϕD Þ≫ðϕN Þ, and in ϕ ¼ ½ðϕ2D þ ϕ2N Þ=21=2 ð11:5Þ
shales, where ðϕD Þ≪ðϕN Þ.
These concepts confirm that the hydrocarbon This empirical transform yields an estimated true
present in the upper part of the sand is gas. porosity value of 29%, which is in agreement with the
The porosity curves of Figure 11.2 display average porosity value of the underlying liquid-bearing zone
apparent porosity values of ϕD ¼ 39% and ϕN ¼ where both curves indicate similar porosity values.
10:5% respectively. To determine the true porosity Such agreement is expected in uniform sands, which
184 value, ϕ, of the gas-bearing section, the following is the case in this example log. This value is slightly
Well Logging
different from that estimated using the sonic log

because of the assumed parameters used in the
interpretation.
Analysis of the logs of Figures 11.1 and 11.2 is
relatively straightforward, as the sand of interest is
a clean sand. The log responses in shaly sands are
more complex as they reflect the presence of clay-
minerals. Because clean formations are usually
selected for CS projects to ensure sufficient perme-
ability and in turn adequate injection rates, shaly sand
analysis is not covered here. Readers interested in the
topic are referred to the references, e.g., Archie (1942)
for the coverage of shaly sand analysis.
Concepts of Cased-Hole Resistivity Logs

Open-hole conventional resistivity tools emit a rela-
tively low frequency electromagnetic field. In cased
holes, the presence of a highly conductive steel casing
shunts the emitted field through the casing thereby
rendering the conventional measurement impossible.
However, a relatively new technology led to the design
of a cased-hole resistivity tool. The new tool is basi-
cally a focused current device. A current electrode in
contact with the casing injects current into the bore-
hole environment. Most of the current will flow up
and down the casing steel, the path of least resistance.
A high-frequency current will stay almost entirely
within the steel. The use of a low-frequency current
allows a tiny fraction of the current to leak into the
formation. The higher the leaking current, the lower
the resistivity of the formation. The result is a cased- Figure 11.3 Comparison between open-hole and cased hole
formation resistivity. (From Jiang et al., 2009.)
hole resistivity log that is used to monitor the advance
of injected fluids through the reservoir.
The tool measures the current leaking into the contact between the tool electrodes and the casing.
formation If as the difference of axial currents over The tool may not be able to get good contact over
two adjacent casing segments. The formation resistiv- casing collars. In addition, the measurement itself is
ity Rt is expressed as (Harold et al., 2004) impaired by the presence of corrosion-inhibitor
nonconductive coating.
Rt ¼ KðV=If Þ ð11:6Þ
Figure 11.3 illustrates the potential application of
where V is the casing voltage and K is the tool geo- cased-hole formation resistivity (CHFR) measure-
metric factor, which depends on the tool design. ment in monitoring reservoir performance. This
The leakage current is typically in the order of several example shows significant reduction in resistivity in
milliamperes, V is in 10s of nanovolts, which can be the lower part of the displayed hydrocarbon forma-
measured using new technology. tion, which indicates depletion, i.e., reduction in the
Cased-hole resistivity tools have, however, sev- saturation of the resistive fluid. Equation (11.1) can be
eral limitations related to the measurement environ- used to quantify saturation changes. The depletion
ment. The presence of high-resistivity cement might can be quantified using the ratio of existing saturation
require environmental corrections. Scale accumula- to initial saturation, as seen on the third track of
tion on the casing surface may inhibit efficient Figure 11.3. 185
Zaki Bassiouni
Figure 11.4 Time-lapse induction and sigma logs recorded in Fiberglass casing. (From Gould et al., 1991.)
If observation or monitoring wells are utilized, the neutron, i.e., the loss of kinetic energy. On the other
wells can be cased using high-resistivity fiberglass hand, absorptive reactions result in the disappearance
tubulars. Such casing material allows the propagation of the slow neutron in its free form.
of the electromagnetic field through the casing and Inelastic scattering is a moderating interaction of
into the formation. Resistivity type logs can be fast neutrons possessing relatively high energy. In this
recorded throughout the life of the reservoir, provid- moderating reaction, the neutron collides with the
ing means to monitor changes in water saturation. nucleus. A fraction of the neutron’s kinetic energy
Figure 11.4 shows time-lapse induction logs recorded transfers to the nucleus in the form of internal energy.
in fiberglass casing compared to time-lapse pulsed This in turn results in the excitation of the nucleus.
neutron logs recorded in the same well. Induction The excited nucleus is immediately stabilized by emit-
logs recorded through fiberglass casing provides ting a gamma ray photon. Deexcitation gamma rays,
information that correlates well with that obtained commonly referred to as scattering gamma rays, are
using pulsed neutron tools. Resistivity logs thus characteristic of the nucleus involved in the inelastic
acquired can be integrated with neutron logs for interaction. Spectral analysis of the scattering gamma
quantitative analysis. rays indicates the type and quantity of nuclei present
in the formation.
Inelastic scattering requires that the neutron energy
Concepts of Neutron Logs exceeds the threshold energy needed to excite the
In its basic design, neutron tools consist of a source nucleus. When the neutron energy is moderated to
periodically emitting high-energy neutrons. As they a level below that threshold, elastic scattering takes
propagate, these neutrons interact with the individual place. In this latter interaction, the neutron collides
nuclei of the atoms of the formation. The results of these with a nucleus in a manner similar to that of two
interactions are measured at two detectors within the colliding billiard balls. The neutron scatters off the
tool placed at two different distances from the source. nucleus with less kinetic energy. The energy balance is
There are two types of interactions, moderating transferred to the nucleus in the form of kinetic energy;
186 and absorptive. Moderating interactions, which can its internal energy remains unchanged. Therefore, no
be inelastic or elastic, result in the slowing down of the radiation is associated with elastic scattering.
Well Logging
After sufficient elastic collisions, the neutron is

thermalized, i.e., reaches a low kinetic energy level in
the vicinity of 0.025 eV. The number of collisions
resulting in the thermalization of the neutron depends
on the mass of the nuclei present in the formation.
Because hydrogen nuclei have almost the same mass
as the neutron, fewer collisions with hydrogen atoms
are needed to thermalize a neutron.
Thermal neutrons are captured by a nucleus.
Neutron capture by a nucleus results in a heavier
isotope in an excited state. The excited nucleus
releases the excess energy in the form of gamma
photons, referred to as “gamma ray of capture.”
Energy levels of capture gamma rays are characteristic
of the capturing nucleus. Spectral analysis of capture
gamma rays indicates the types and quantity of nuclei
present in the formation.
Physical parameters associated with the phenom-
ena occurring during the moderation and capture of
neutrons are measured and empirically related to
formation properties. Because hydrogen is responsi-
ble for most of the slowing down of the epithermal
neutrons, i.e., neutrons with an energy level higher
than that of the thermal neutrons. Measuring the
concentration of epithermal neutrons indicates the
hydrogen concentration in the formation, which in
turn is empirically related to porosity because, in Figure 11.5 Example of an early sigma log. (From McGhee et al.,
clean formations, hydrogen exists only in the fluids 1974.)
saturating the pore space.
The rate of decay of thermal neutrons is related to
the capture cross sections, Σ, of the formation consti- and high water salinity environments, unlike Σ, which
tuents. Chlorine has the highest capture cross section, requires relatively high salinity. In addition, the spec-
Σ, of common formation elements; hence measuring tral analysis of the captured gamma ray can be used to
Σ reflects chlorine content and, in turn, water content determine the contents of iron, silicon, and calcium
and, by complement, gas, or oil content. Figure 11.5 is needed for lithology determination. It is also used to
an example of an earlier sigma log. In addition to the determine chlorine and hydrogen contents, which
Σ curve, the raw readings of the near and far detectors allows us to investigate salinity.
are also displayed. The scales used to display these two A schematic of a C/O ratio log is displayed in
curves are different and are selected in a way that the Figure 11.6. The C/O log measurements may be
readings of the near and far detectors will track each obtained by moving the tool at an optimal speed
other in liquid-bearing formations, and separate in continuously or by stationary measurements for
gas-bearing formations. enhanced log quality. To assist with the analysis,
The energy of scattering gamma rays and capture a “C/O envelop” is computed, using the theoretical
gamma rays are, respectively, characteristic of the model presented in the section on water saturation
nucleus involved in the scattering or the capture. and Figures 11.7 and 11.8. The C/O envelop consists
This principle leads to several types of measurements, of two curves (C/O)min computed by setting Sw =1
e.g., the spectral contribution of carbon and oxygen and (C/O)max computed by setting Sw = 0. To perform
allows the determination of carbon/oxygen (C/O) such computation, the lithology and porosity
ratio. The C/O ratio can be determined in both low are usually obtained from open-hole data. The 187
Zaki Bassiouni
C/O Increasing
0.65
So = 0.0 0.5 1.0
C/Omin Low φ
0.60
=1
C/Omax So
LIMESTONE
C/O
Ca (CO3)
0.55
So=0
limestone
High φ
0.50 =1
So
So=0
High φ 0.45 sandstone
Water/Oil Contact 0 10 20 30 40
SANDSTONE POROSITY
Continuous
SiO2
C/O Log
Figure 11.7 C/O interpretation chart constructed for a specific
Low φ matrix. (From North, 1987.)
C/O Stationary
Readings
Figure 11.6 Schematic illustrating the C/O envelop interpretation

technique. (From Smolen, 1996.)
Figure 11.8 Water-flood monitoring case

using C/O log. (From Gilchrist et al., 1983.)
188
Well Logging
variations of the envelop curves reflect lithology and f(C) and f(O) are respectively the total carbon and
porosity variations. The separation between the envel- oxygen contribution from borehole, matrix, and
ope’s two curves at any level scales linearly in terms of formation fluids
water saturation. Bc, Bo = C and O contributions from borehole
Sw, So = Water and oil saturation
Quantitative Analysis Methods α = Atomic concentration of C in matrix
β = Atomic concentration of C in formation fluid
Water Saturation from Sigma Logs
γ = Atomic concentration of O in matrix
The tool measured Σ log is related to the constituents
δ = Atomic concentration of O in formation fluid
of the formation. For a clean, porous formation con-
taining water and hydrocarbon, Σlog is expressed as ϕ = Porosity
(Clavier et al., 1971): Because the model represented by Eq. (11.8) is
independent of water salinity, the C/O log can be
Σlog ¼ ð1 ϕÞΣma þ ϕSw Σw þ ϕð1 Sw ÞΣh ð11:7Þ
applied in any salinity environment including
unknown salinity or mixed salinity, as in the case of
where Σma, Σw, and Σh are the capture cross sec-
CO2 chased by water where the injected water salinity
tion of rock matrix, water, and hydrocarbon, respec-
is different than that of the formation water.
tively. Solving Eq. (11.6) for water saturation, Sw, gives
The C/O model can be solved in several ways. For
ðΣlog Σma Þ ϕðΣh Σma Þ a specific known matrix, Sw can be estimated from
Sw ¼ ð11:8Þ
ϕðΣw Σh Þ a prepared chart similar to that of Figure 11.7.
Families of similar graphs exist for various borehole
The water, rock matrix, and hydrocarbon cross environment and equipment type used.
sections of capture are determined experimentally or The log of Figure 11.8 is from an observation well
deduced from graphical interpretation techniques drilled for reservoir characterization and monitoring
using actual field data. Porosity values are derived purposes. The reservoir is in a water-flood program
from other log measurements. In shaly formations, where the injected water is fresher than the formation
Eq. (11.6) is modified to include the shale contribu- water. Water is injected in the main sand body, which
tion as warranted. lies between 3499 and 3518 ft. It consists of two lobes
The time-lapse technique allows monitoring of separated by a thin, shaly limestone section. Open-hole
water saturation. If only one other fluid is present in logs and cores indicated that the flood front had not yet
addition to the water, e.g., oil in an undersaturated oil reached the well. Cased-hole logs were run more than
reservoir and CO2 in the case of aquifer CS, changes of 6 months after the well had been drilled. Owing to the
the saturation of oil or CO2, the complement of the injection of water of different salinity than the formation
water saturation, can in turn be monitored. water, the salinity at the flood front is unknown. The C/
O log was hence selected for monitoring purposes.
Water Saturation from C/O Log Figure 11.8 displays open-hole logs (SP and
gamma ray). Bulk volume analysis and water satura-
The C/O interpretation model can be written in terms
tion are calculated using the resistivity log. It also
of the matrix, pore space, and borehole contributions,
displays the C/O envelop and C/O stationary mea-
which can be expressed as the product of their respec-
surements. Water saturation values calculated from
tive volume and atomic densities, yielding the follow-
C/O data are shown superimposed on open-hole
ing expression (North, 1987):
water saturation.
C=O ¼ Af ðCÞÞ=f ðOÞ The monitor log clearly indicated water break-
f ðCÞÞ ¼ αð1 φÞ þ βSo ϕ þ Bc ð11:9Þ through in the bottom of the lower sand lobe as the
f ðoÞÞ ¼ γð1 φÞ þ δSw ϕ þ Bo current water saturation is higher than the open-hole
water saturation. This information was integrated
where with other petrophysical and production data to eval-
A = Ratio of average C and O fast neutron uate the flood performance and take necessary steps to
(scattering gamma ray producing) cross sections optimize its performance. Also, should the operator 189
Zaki Bassiouni
elect to follow water flooding with CO2 injection,

a similar monitoring approach can be implemented.
CO2 Saturation from Neutron Porosity Logs

The neutron porosity log measurement reflects the
bulk hydrogen index of the formation. The hydrogen
index of a material is defined as the ratio of the con-
centration of hydrogen atoms per unit volume in the
material to that of pure water at 75°F. The hydrogen
index of water is thus equal to 1. To simplify the
analysis, the hydrogen index of oil is usually assumed
equal to 1 as well.
If the lithology is devoid of hydrogen atoms, e.g.,
clean sandstone or limestone, neutron porosity mea-
surement allows the estimation of the amount of
liquid-filled porosity, ϕt. In a formation partially
saturated with CO2 the log displays an apparent
porosity, ϕa, lower than the total porosity because
the hydrogen index of CO2 is almost nil, actually
0.03. It follows that the CO2 saturation, SCO2, can /
be expressed as
SCO2 ¼ 1–ðϕa =ϕt Þ ð11:10Þ
The decrease of neutron porosity with an increase

of gas saturation is illustrated in the section of the log
of Figure 11.9 showing a series of neutron porosity
logs run to monitor the movement of a gas cap.
Open-hole logs indicated that initially the gas/oil
contact (GOC) was at 9310 ft. The open-hole log
Figure 11.9 Example of monitoring of gas oil contact based on the
recorded on July 11, 1979 displays the true porosity change of gas effect on the neutron porosity. (From Dupree, 1989.)
of the thick uniform sand. The log of November 5,
1983 shows a marked decrease in porosity between
9310 ft. and 9475 ft., the dark shaded area, due to gas reservoirs. The estimation is possible where both
effect. The increase in the gas saturation above the time-lapse cased-hole resistivity and pulsed neutron
GOC situated at 9475 ft. can be calculated using Eq. tools are available. Current water saturation, Sw, is
(11.9). determined from resistivity measurement and the
Another monitoring log was run on June 7, 1986. CO2 saturation, SCO2, can be determined from
The new log indicates that the GOC is now at a depth the neutron porosity log. Hence, the saturation of
of 9545 ft. Geological analysis of the structure indi- the residual oil, So, can be monitored, since
cated that other anomalies at the top of the sand and So ¼ 1 Sw –SCO2 .
between 9410 and 9420 ft. are the result of gas
encroachment under shale layers. The example of
Figure 11.9 mimics the growth of the reservoir volume CO2 Flood Monitoring Examples
occupied by injected CO2.
Monitoring of Timbalier Bay Miscible CO2 Injection
Residual Oil Saturation from Cased- Hole Resistivity and Project
Pulsed Neutron Logs This example’s monitor well was drilled in the Gulf of
The estimation of residual oil saturation is desired in Mexico as part of the Timbalier Bay Miscible CO2
190 CO2 EOR projects and in cases of CGS in depleted oil Injection Project (Moore, 1986; Ruhovets and Wyatt,
Well Logging
Figure 11.10 Sequential monitor logs showing reservoir displacement. (From Moore, 1986.)
1995). The formation is a clean, continuous, high- feasibility of continued CO2 injection can be reached
porosity sandstone. The reservoir was a nearly based on in situ values. The accuracy of this techni-
depleted saturated-oil accumulation produced by que depends on the quality of the data and the repre-
a strong water drive. sentativeness of the models used in the quantitative
After the base logs were run in April, 1984, CO2 evaluation. Nevertheless, reasonable qualitative and
injection was initiated in an up-dip injection well at quantitative monitoring of CO2 flooding can be
the reservoir gas/oil contact. Approximately 6 months achieved.
later, October, 1984, the first monitor logs were run.
Additional monitor logs were run in January, 1985, Monitoring of Saturation Changes for CO2 Sequestration
April, 1985, and August, 1985. Figure 11.10 displays in Aquifers
water saturation calculated from the base and the The Frio Brine Pilot Experiment began in 2002 in
three monitor logs. The first monitor log indicates South Liberty field, Dayton, Texas is discussed in
the arrival of the injected CO2 at the monitoring Sakurai et al. (2006). The target of the CGS test is
well. The consecutive monitor logs show that the a fluvial sandstone of the upper Frio Formation at
water contact is moving downward as additional a depth of 5000 ft. An existing well was used as
CO2 is injected resulting in oil displacement. the observation well and a new injection well was
The analysis made use of available log and core data. drilled 100 ft. away from the observation well, as
Saturation values of water, oil, and CO2 were esti- illustrated in Figure 11.11. A significant amount of
mated using the techniques described in the section data was obtained, including cores and open-hole
“Quantitative Analysis Methods” and Ruhovets and logs. The formation water was sampled using
Wyatt (1995). a wireline formation tester. Results of core and
This example demonstrates that the remaining open-hole log analysis are shown in Figure 11.12.
oil saturation in the CO2 flooded interval can be Core analyses indicated porosity in the range of 191
estimated and a decision as to the economic 20–35%, and the analysis of water samples
Zaki Bassiouni
distribution over the aquifer interval. Figure 11.13

shows the presence of CO2 indicated by the decrease
in Σ. The change of CO2 saturation is displayed in
Figure 11.14. At the end of injection on October 14,
2004 the CO2 saturation exceeded 80% in the vicinity
CO2 Monitoring of the observation well, although these results display
injection well
well
CO2 accumulation and migration only around the
observation well.
References
Archie, G. E. (1942). Electrical resistivity log as an aid in
South Liberty determining some reservoir characteristics.
Salt Dome Transactions of the AIME, 146: 54–61.
Bassiouni, Z. (1994). Theory, measurement,
and interpretation of well logs. SPE Textbook
Series, 4. Richardson, TX: Society of Petroleum
Engineers.
Clavier, C., Hoyle, W. R., and Meunier, D. (1971).
Quantitative interpretation of TDT Logs. Journal of
Petroleum Technology, 23: 743–763.
Dupree, J. H. (1989). Cased-hole nuclear logging
interpretation, Prudhoe Bay, Alaska. Log Analyst,
Injected 162–177.
CO2
Gilchrist, W. A., Jr., Roger, L. T., and Watson, J. T. (1983).
Carbon/oxygen interpretation, a theoretical model.
In SPWLA 24th Symposium, paper FF.
Frio Gould, J., Wackler, J., Quiren, J., and Watson, J. (1991). CO2
Formation monitor logging: East Mallet Unit, Slaughter Field,
Hockley County, Texas. In SPWLA 32nd Annual
Symposium, paper PP.
Harold, B. H., Benimeli, D., Leveques, C., Bebourg, I.,
and Cadenhead, J. (2004). Combinable through-
Figure 11.11 A schematic approximating the subsurface geology
tubing cased hole formation resistivity tool. SPE
surrounding the injection and observation wells. (From Sakurai et al., 90018.
2006.) Jiang L., Tokar, T., and Zyweck, M. (2009). Innovation to
enhanced oil recovery from slim cased hole resistivity
measurement in Gulf of Thailand. In SPWLA 50th
indicated a high salinity of 93 000 ppm. Because of Annual Symposium, paper YY.
the high porosity and formation water salinity, the McGhee, B. F., McGuire, J. A., and Vacca, H. I. (1974).
pulsed neutron tool is optimal for monitoring Examples of dual spacing thermal neutron decay time
saturation changes. The perforated interval, login Texas coast oil & gas reservoirs. SPWLA 15th
shown by the shaded red area, was selected in Annual Symposium, paper R.
the interval displaying best reservoir quality. Moore, J. S. (1986). Design, installation, and early operation
Injection of CO2 started on October 4, 2004 and of the Tambalier Bay Gravity–Stable Miscible CO2
lasted for 11 days. Base logs were recorded in the Injection Project. SPE Production Engineering,
observation well before the start of injection. 369–378, also SPE 14287.
Monitor logs were run at preselected dates afterward North, R. J. (1987). Through-casing reservoir evaluation
to detect the CO2 breakthrough and its subsequent using gamma ray spectroscopy. SPE 16356.
192
Figure 11.12 Results of
core- and open-hole log-
analysis. (From Sakurai et al.,
2006.)
Figure 11.13 Time lapse

decrease of Σ due to presence
of CO2 (shaded red area).
(From Sakurai et al., 2006.)
Zaki Bassiouni
Figure 11.14 Time Lapse

Variations of the saturation of
CO2 (shaded red area). (From
Sakurai et al., 2006.)
Ruhovets, N., and Wyatt, D. F. (1995). Quantitative changes for CO2 CSS; petrophysical support of the
monitoring of gas flooding in oil-bearing reservoirs by Frio Brine pilot experiment. Petrophysics, 47(6):
use of a pulsed neutron tool. SPE 21411. 483–496.
Sakurai, S., Ramakrishnan, T. S., Boyd, A., Mueller, N., Smolen, J. J. (1996). Cased hole and production log
and Hovorka, S. (2006). Monitoring saturation evaluation. Tulsa, OK: PennWell, 108.
194
Part III Case Studies
Chapter
Offshore Storage of CO2 in Norway
12 Eva K. Halland
Introduction without having any negative effects on ongoing or

future petroleum production.
Three projects for offshore storage of carbon dioxide
The Norwegian CO2 Storage Atlas aims to docu-
(CO2) are ongoing in Norway. Chapters 13 by Eiken
ment the distribution and properties of the aquifers
and 18 by Grude and Landrø in this book provide
having a relevant storage potential for CO2 offshore
additional detail. This provides extensive experience
Norway (Halland et al., 2014). Several storage options
with injection and storage of CO2 in geological for-
have been evaluated such as structured aquifers, struc-
mations. Since 1996, approx. 1 Mtonne of CO2 has,
tural closures, dipping aquifers, abandoned fields,
each year, been separated from gas production at the
and the possibility of storing CO2 in combination
Sleipner Vest Field, in the North Sea. The CO2 has
with enhanced oil recovery (EOR). In 2015, the
been stored in the Utsira Formation and, from 2014,
Norwegian government decided to look at the possi-
a further 0.1–0.2 Mtonne/year of CO2 from the
bility of establishing a full-chain carbon capture and
Gudrun Field has been injected into the same
storage (CCS) project in Norway, including capture,
formation.
ship transport, and storage of CO2. Based on the
From the treatment of the production stream from
Norwegian CO2 Storage Atlas, Statoil was appointed
the subsea developed on Snøhvit Field and the liquefied
to further mature three storage sites. The aim was an
natural gas production on Melkøya in the Barents Sea,
injection capacity of 1 Mtonne of CO2 annually with
about 0.7 Mtonne of CO2 is annually safely injected
an opportunity for third-party access at a later stage.
and stored in geological formations. The start of injec-
One geological structure in the Utsira Formation, one
tion was in 2008 and the CO2 was injected into the
saline aquifer in the area east of the Troll gas field and
Tubåen Formation, about 2300 m beneath the seabed.
the third site in the Heimdal depleted gas field have
In early 2010 the field operator realized that there was
now been studied in more detail and all show good
less storage capacity than first expected at the injection
characteristics as CO2 storage sites.
site. In 2011, the site of injection was relocated to the
Stø Formation and a new geological structure is now
under evaluation for further CO2 storage in the area. The Norwegian Continental Shelf
The structures are identified with seismic data. High- Petroleum activity on the Norwegian Continental
quality 3D seismic data that properly image the storage Shelf (NCS) has been ongoing for about 50 years
site, storage complex, and the connecting area are and today takes place in the North Sea, the
required to identify geological formations and struc- Norwegian Sea, and the southern part of the Barents
tures that can be suitable for CO2 storage. Sea (NPD fact pages). These regions also contain
With the recognition of an increased climate chal- several saline aquifers (all aquifers referred to in this
lenge and based on the experience from these storage chapter are saline in nature) at a suitable depth for
projects, the Norwegian Ministry of Petroleum and CO2 storage, and are considered favorable for large-
Energy (MPE) requested a CO2 storage Atlas of scale CO2 sequestration (Fig. 12.1).
Norway. The main objective was to identify storage The NCS belongs to three different geological
capacity for safe and effective long-term storage of provinces. The northern North Sea is a subsiding
CO2, and to facilitate selection of sites that could basin developed above the continental Paleozoic and
be suitable for future CO2 sequestration projects, Mesozoic rift between the UK and Norway. The Basal 195
Eva K. Halland
Figure 12.1 Area status for the Norwegian Continental Shelf (NCS). Depth map of The Base Cretaceous Unconformity and areas evaluated for
CO2 storage. The boundaries to oceanic crust and main volcanic provinces are indicated. (Source: Norwegian Petroleum Directorate, NPD
Facts 2017.)
Cretaceous Unconformity (BCU) marks the end of aquifers (Deegan and Skull, 1977; Vollset and Doré,
the last major rifting event and the top of the cap 1984; Isaksen and Tonstad, 1989).
rock of the Jurassic aquifers. Younger sediments in The Norwegian Sea shelf faces the oceanic crust in
196 the North Sea Basin were sourced from all landmasses the Norwegian-Greenland Sea. The BCU is deeply
surrounding the basin and contain several important buried under the deep waters beyond the shelf slope
Garn Fm (south)
Åre Fm
Garn Fm
Ile Fm
Rogn Fm
Lysing Fm
Egga Fm
Tilje Fm
Conceptual sketch showing location of

aquifers relative to depositional systems
Figure 12.2 Conceptual sketch showing location of aquifers relative to depositional systems. (Source: Norwegian Petroleum Directorate.)
(NPD, 1995). The map in Figure 12.1 does not display store CO2. In the North Sea Basin, the greatest poten-
the BCU at greater burial depths than 5000 m, as these tial capacity for CO2 storage will be in deep saline-
depths are not relevant for CO2 injection. The main water saturated formations (saline aquifers), or in
aquifers suitable for CO2 injection in the Norwegian depleted oil and gas fields.
Sea belong to the Jurassic section. Aquifers may consist of several sedimentary for-
The Barents Sea shelf was filled-in by thick mations and cover large areas. A good understanding
sequences of Upper Paleozoic, Mesozoic, and younger of the geology and sedimentologic development of
rocks following a major carboniferous rifting phase the area is of upmost importance (Figure 12.2).
(NPD, 1996). In the Norwegian sector, petroleum activ- An aquifer is a body of porous and permeable sedi-
ity has been concentrated to the southwestern part, mentary rocks where the water in the pore space is in
which is made up of platforms, shallow basins, and communication throughout (Figure 12.3). All the
highs. The main aquifers consist of Jurassic and identified aquifers on the NCS are saline; most of
Triassic sediments. In the western margin, these sedi- them have salinities in the order of seawater or higher.
ments are too deeply buried to be suitable for CO2 To be suitable for CO2 storage, aquifers need to be at
storage. The studied area was exposed to deep erosion a depth where CO2 can be stored in a supercritical
in the Cenozoic and in the Quaternary. Because reser- state and overlain by impermeable cap rocks acting as
voir properties reflect maximum burial and not the a seal to prevent CO2 migration into other formations
present burial, aquifers with good storage potential are or to sea (Figure 12.4) (NPD, 1996). An extensive
located at relatively shallow depths. seismic data acquisition is used to provide the best
possible basis for mapping the extent of the potential
Geological Storage storage site for CO2. The seismic data is an important
Depending on their specific geological properties, part in order to plan the placement of injection wells
for injecting CO2 in an optimal manner. Finding an 197
several types of geological formations can be used to
Eva K. Halland
Figure 12.3 Relation between

geological formations and aquifers.
(Source: Norwegian Petroleum
Directorate.)
optimal location of injection wells is important to Data Availability

ensure that the CO2 can be stored and that it remains
The Norwegian Petroleum Directorate (NPD) has
in the reservoir in the future.
access to all data collected on the NCS and has
CO2 is held in place in a storage reservoir through
a national responsibility for the data. The NPD’s
one or more of five basic trapping mechanisms: strati-
data, overviews, and analyses make up an important
graphic, structural, residual, solubility, and mineral
information base for the oil and gas activities, which is
trapping. Generally, the initial dominant trapping
also available for the evaluation of CO2 storage.
mechanisms are stratigraphic trapping or structural
The main objective of these Reporting
trapping, or a combination of the two. The lighter
Requirements is to support the efficient exploitation
CO2 will move upwards through the porous rocks
of Norway’s hydrocarbon reserves. Around 50 years
until it reaches the top of the formation, where it is
of petroleum activity, has generated a large quantity of
trapped by an impermeable layer of cap-rock.
data. These data were available for the CO2 storage-
In residual trapping, the CO2 is trapped in the tiny
mapping project. This includes 2D and 3D data and
pores in rocks by the capillary pressure of water. Once
data from exploration and production wells such as
injection stops, water from the surrounding rocks
logs, cuttings, and cores (Figure 12.5). Production
begins to move back into the pore spaces that contain
data and pore pressure data from hydrocarbon fields
CO2. As this happens, the pressure of the added water
and years of dedicated work to establish geological
immobilizes the CO2. This process, which further
play models have all been utilized to define the main
traps the CO2, includes solubility (or dissolution)
aquifers.
trapping (discussed further in Chapter 13 by Eiken,
this volume). Dissolution of the CO2 in the saline
water forms a denser fluid that may sink to the bottom Workflow and Characterization
of the storage formation. Depending on the rock for- An approach for assessing the suitability of the geo-
mation, the dissolved CO2 may react chemically with logical formations for CO2 storage is summed up in
the rock matrix to form stable minerals. Known as a flowchart (Figure 12.6). The intention has been to
mineral trapping, this provides the most secure form develop a systematic way to identify all the aquifers
of storage for the CO2, but it is a slow process and may and to find which aquifers are prospective in terms of
take thousands of years. large-scale storage of CO2.
198
Carbon dioxide: Temperature - pressure diagram

100.000.0
g Line
Meltin
t
1000.0 in
Po
ical
Crit
Pressure (bar)
CO2 Solid CO2 Liquid
100.0
ine
tion L Critical
Satura Point
3
0.5 g/cm
10.0
CO2 Gas
e
Lin Triple Point
tion
ma
bli CO2 Vapour
1.0 Su Sublimation Point
0.1
–100 –90 –80 –70 –60 –50 –40 –30 –20 –10 0 10 20 30 40 50
Temperature (°C)
Figure 12.4 Supercritical fluids behave like gases, in that they can diffuse readily through the pore spaces of solids. However, like liquids, they
take up much less space than gases. Supercritical conditions for CO2 occur at 31.1°C and 7. 38 megapascals (MPa), which occur approx. 800 m
below surface level. This is where the CO2 has both gas and liquid properties and is 500–600 times denser (up to a density of about 700 kg/m3)
than at surface conditions, while remaining more buoyant than formation brine. (Source: Norwegian Petroleum Directorate, CO2CRC.)
In subsequent steps in the workflow, each potential their technical maturity. Some guidelines are devel-
reservoir and seal we identified was evaluated and char- oped to facilitate the characterization. A summary
acterized for its CO2 storage prospectivity. Based on chart for characterization of reservoir, seal, and data
this, the potential storage sites are mapped and the coverage (Figure 12.7) with more detailed checklists is
storage capacity calculated. The evaluation is based on used for both reservoir and sealing properties.
available data in the given areas. This evaluation does Parameters used in the characterization process
not provide an economic assessment of the storage sites. are based on data and experience from the petroleum
Aquifers and structures are characterized in terms activity and the CO2 storage projects on the NCS, and
of capacity, injectivity, and safe storage of CO2. the fact that CO2 should be stored in the supercritical
To complete the characterization, the aquifers are phase to obtain the most efficient and safest storage.
also evaluated according to the data coverage and
199
Eva K. Halland
Figure 12.5 Data availability, seismic, and wells in the North Sea. (Source: Norwegian Petroleum Directorate.)
200
Evaluation process for safe CO2 storage sites
CAP ROCK
Evaluation of data Characterization of

coverage and reservoir/
knowledge injectivity
Characterization of
Stratigraphy
seal
(reservoir and seal)
efficiency
Depth to the top Paleocene

100 m
3200 m
Paleocence sand
Structural Contour interval 200 m Map potential

trapping storage area
Trapping
Estimate
Stratigraphic storage
trapping capacity
Figure 12.6 The workflow for characterization of geological formation for safe CO2 storage. (Source: Norwegian Petroleum Directorate.)
The scores for capacity, injectivity, and seal quality are with larger aquifers. The pressure will increase when
based on evaluation of each aquifer/structure. injecting fluid into a closed or half-open aquifer.
For evaluation of regional aquifers in CO2 storage Pressure increase can, however, be mitigated by pro-
studies, the mineralogical composition and the petro- duction of formation water. The fracture pressure
physical properties of the cap rocks are rarely well depends on the state of stress in the bedrock and is
known. In order to characterize the sealing capacity, typically 10–30% lower than the lithostatic pressure
we have relied mainly on regional pore pressure dis- (Figure 12.8). Fracture gradients were estimated by
tributions and data from leak-off tests combined with comparing pore pressures in overpressure reservoirs
observations of natural gas seeps. with data from leak-off tests.
The storage efficiency factor (Seff) is assessed indi-
Capacity Estimation vidually for each aquifer based on simplified reservoir
Storage capacity depends on several factors, primarily simulation cases. For saline aquifers, the amount of
the reservoir pore volume and the fracturing pressure. CO2 that can be stored can be determined using the
The relation between pressure and injected volume following formula:
depends on the compressibility of the rock and the MCO2 ¼ Vb ∅ n=g ρCO2 Seff
fluids in the reservoir. The solubility of CO2 in the
different phases will also play a part. It is important to where
know if there is communication between multiple MCO2 = mass of CO2
201
reservoirs, or if the reservoirs are in communication Vb = bulk volume
Eva K. Halland
CHARACTERIZATION OF AQUIFERS AND STRUCTURES

Criteria Definitions, comments
3 Large calculated volume, dominant high scores in checklist

Reservoir quality Capacity, communicating volumes
2 Medium - low estimated volume, or low score in some factors
1 Dominant low values, or at least one score close to unacceptable
Injectivity 3 High value for permeability *thickness (k*h)
2 Medium k*h
1 Low k*h
Sealing quality Seal 3 Good sealing shale, dominant high scores in checklist
2 At least one sealing layer with acceptable properties
1 Sealing layer with uncertain properties, low scores in checklist
Fracture of seal 3 Dominant high scores in checklist
2 Insignificant fractures (natural / wells)
1 Low scores in checklist
Other leak risk Wells 3 No previous drilling in the reservoir / safe plugging of wells
2 Wells penetrating seal, no leakage documented
1 Possible leaking wells / needs evaluation
Data coverage Good data coverage Limited data coverage Poor data coverage
Other factors:
How easy / difficult to prepare for monitoring and intervention. The needs for pressure relief. Possible support for EOR projects. Potential for conflicts with future petroleum activity.
Data coverage
Good 3D seismic, wells through the actual aquifer/structure
Limited 2D seismic, 3D seismic in some ares, wells through
equivalent geological formations
Poor 2D seismic or sparse data
Figure 12.7 The summary chart for characterization of reservoir, seal, and data coverage. (Source: Norwegian Petroleum Directorate.)
Figure 12.8 Pressure gradients

obtained from pore pressure data
and leak-off tests in wells from the
Norwegian Sea Shelf and North
Sea at water depths between 250
and 400 m. The fracturing gradient
marks the lower boundary of
measured leak-off pressures and
the upper boundary of measured
pore pressures. The lithostatic
gradient was calculated from
general compaction curves for
shale and sand with a 300-m water
column. The hydrostatic gradient
assumes seawater salinity.
The arrows show how much
pressure can be increased from
hydrostatic pressure before it
reaches the fracture gradient.
(Source: Norwegian Petroleum
Directorate.)
202
∅ = porosity CO2 is contained in the reservoir according to plans

n/g = net to gross ratio and predictions, and second, if there are deviations, to
ρCO2 = density of CO2 at reservoir conditions provide data that can update the reservoir models and
Seff = storage efficiency factor (source: Geocapacity, support eventual mitigation measures. Seismic sur-
2009) veys are most widely used for monitoring of CO2
Seff is calculated as the fraction of stored CO2 geological storage. Careful monitoring of the behavior
relative to the pore volume. of the storage facility is required to establish its safety,
that the injected CO2 is behaving as predicted and is
For a homogeneous reservoir with a permeability
not migrating out of the intended storage site.
of 200 mD and reservoir thickness of 100 m, the
An extensive program to monitor and model the dis-
storage efficiency in a closed system is simulated to
tribution of injected CO2 in the Utsira Formation has
be 0.4–0.8%, with a pressure increase of 50–100 bar.
been undertaken by a number of organizations (and
Discussion of the storage efficiency in the Utsira is
has been partly funded by the European Union).
undertaken in Chapter 13 by Eiken in this volume.
The Utsira monitoring program includes a baseline
The storage potential is estimated for the differ-
3D seismic survey and several time-lapse (4D) seismic
ent types of storage. The storage capacities are pre-
surveys. These surveys provide a good picture of how
sented in a pyramid diagram, where the highest
CO2 moves horizontally and vertically (see Eiken,
level in the pyramid represents the capacity of
Chapter 13 in this volume).
sites that are already used for CO2 storage, while
In exploration wells on the NCS, pressure differ-
the lowest level represents theoretical capacity in
ences across faults and between reservoir forma-
lesser-known aquifers (Figure 12.9). An extensive
tions and reservoir segments are commonly
database has been available for this evaluation.
observed. Such pressure differences give indications
Nevertheless, evaluation of some areas is more
of the sealing properties of cap rocks and faults.
uncertain due to limited seismic coverage and
Based on such observations in the hydrocarbon
a lack of well information. The data coverage is
provinces, combined with a general geological
color-coded to illustrate the data available for each
understanding, one can use the sealing properties
aquifer/structure. Characterization and capacity
in explored areas to predict the properties in less
estimates will obviously be more uncertain where
explored or undrilled areas.
data coverage is poor.
Natural seepage of gas is observed in the hydro-
In summary, the capacity of each aquifer is
carbon provinces in the NCS. Such seepage is
given in the tables as a deterministic volume.
expected from structures and hydrocarbon source
Scores from 1 to 3 indicate the injectivity and seal-
rocks where the pore pressure is close to or exceeds
ing properties (3 is the best mark).
the fracture gradient. Seepage at the sea floor can
The characterization is based on a best estimate of
be recognized by changes in biological activity and
each parameter. Uncertainty is not quantified, but is
by free gas bubbles. Seismically, gas chimneys or
indicated by the color coding for data availability
pipe structures indicate seepage. The seepage rates
and maturity. An example from the Utsira/Skade
at the surface show that the volumes of escaped gas
aquifer is shown in Figure 12.10.
through a shale or clay dominated overburden are
small in a time scale of a few thousand years.
Monitoring Rapid leakage can take place only if open conduits
An important matter when injecting CO2 is to ensure are established to the sea floor. Such conduits
that there is no leakage to other geological formations could be created along wellbores or by reactivation
or to the sea bottom. Evaluation of the sealing capa- of faults or fractures. Established natural seepage
city, faults, and fractures through the seal, in addition systems are also regarded as a risk factor for CO2
to old wells penetrating the seal, can provide impor- injection.
tant information on the sealing quality and the mon- A wide range of monitoring technologies have
itoring challenges. been used by the oil and gas industry to track fluid
Monitoring of injected CO2 in a storage site is movement in the subsurface. These techniques can
important for two main reasons: first, to see that the be adapted to CO2 storage. For example, repeated 203
The maturation pyramid
The evaluation of geological volumes suitable for injecting
Based on injection history and storing C02 can be viewed as a step-wise approxima-
tion, as shown in the maturation pyramid. Data and
experience from over 40 years in the petroleum industry will
Development of injection site contribute in the process of finding storage volumes as high
up as possible in the pyramid.
Step 4 is the phase when C02 is injected in the reservoir.

Injecti Suitable for long term storage Throughout the injection period, the injection history is
on
closely evaluated and the experience gained provides fur-
ther guidance on the reservoir’s ability and capacity to store
CO2.
Exploration
Effecti
ve and Step 3 refers to storage volumes where trap, reservoir
safe s and seal have been mapped and evaluated in terms of
torage
Theoretical volume regulatory and technical criteria to ensure safe and
Cut off effective storage.
criteria
on volu
me/co Step 2 is the storage volume calculated when areas with
nflict o
f intere
st possible conflicts of interest with the petroleum industry
Volum
e calc
ulated
have been removed. Only aquifers and prospects of
on ave reasonable size and quality are evaluated. Evaluation is
rage p
orosity Increased technical
and th based on relevant available data.
icknes
s maturity
Step 1 is the volume calculated on average porosity and
thickness. This is done in a screening phase that identifies
possible aquifers suitable for storage of C02. The theoretical
volume is based on depositional environment, diagenesis,
bulk volume from area and thickness, average porosity,
permeability, and net/gross values.
1 tonne = one metric tonne = 100 kg

1 Mt = one megatonne = 106 tonnes
1 Gt = one gigatonne = 1000 Mt = 109 tonnes
Figure 12.9 The maturation pyramid. (Source: Norwegian Petroleum Directorate.)

Figure 12.10 Evaluation of the Utsira/Skade aquifer. Showing capacities in aquifer and in prospects, reservoir and seal quality, data quality
and maturation. (Source: Norwegian Petroleum Directorate.)
seismic surveying provides images of the subsur- All areas have a significant potential for CO2
face, allowing the behavior of the stored CO2 to be storage, but the three pyramids show that the
mapped and predicted. Other techniques include regions are quite different. Most of the estimated
pressure and temperature monitoring, down-hole capacities are in the green part of the maturation
sensors as well as seabed monitoring. Surface sen- pyramid, defined as exploration phase, with an
sors and satellite imaging will be available for estimated capacity around 70 Gtonne. In the more
onshore storage sites. mature aquifers and structures, a capacity of 1.3
Gtonne is more or less ready for CO2 injection.
The injectivity of the studied aquifers and the seal-
Conclusion/Summary ing properties of their cap rocks are considered
An overview of the results of this study are dis- acceptable or good. The studied areas are located
played in the table and illustrated by maturation in parts of the Norwegian Continental Shelf (NCS)
pyramids for the North Sea, Norwegian Sea, and that are now open for petroleum activity and
southern Barents Sea (Figure 12.11a, b). All the appears to be the most promising areas for CO2
geological formations have been individually storage.
assessed, and grouped into saline aquifers. We hope that this study will provide useful infor-
The aquifers are evaluated with regard to reservoir mation for future exploration for CO2 storage sites.
quality and presence of relevant sealing formations. A team at the Norwegian Petroleum Directorate
Those aquifers that may have a relevant storage has developed the CO2 Storage Atlas: Eva Halland,
potential in terms of depth, capacity, and injectivity Fridtjof Riis, Andreas Bjørnestad, Christian Magnus,
have been considered. Structural maps and thick- Rita Sande Rød, Van T. H. Pham, Inge M.Tappel, Ine
ness maps of the geological formations were used to Tørneng Gjeldvik, Jasminka Mujezinović, Maren
calculate pore volumes. Bjørheim, and Ida Margrete Meling.
205
Eva K. Halland
Aquifer Capacity Gt Injectivity Seal Maturity Data quality

North Sea aquifers
Utsira and Skade Formations 15,8 3 2
Bryne and Sandnes Formations 13,6 2 2/3
Sognefjord Delta East 4,1 3 2/3
Statfjord Group East 3,6 2 3
Gassum Formation 2,9 3 2/3
Farsund Basin 2,3 2 2/3
Johansen and Cook 1,8 2 3
Formations
Fiskebank Formation 1 3 3
Norwegian Sea aquifers
Garn and lle Formations 0,4 3 3
Tilje and Åre Formations 4 2 2/3
Barents Sea aquifers

Realgrunnen Subgroup, 4,8 3 2
Bjarmeland Platform
Realgrunnen Subgroup, 2,5 3 2
Hammerfest Basin
Evaluated prospects
North Sea 0,44
Norwegian Sea 0,17
Barents Sea 0,52
Abandoned fields
North Sea 3
Producing Fields_2050
North Sea 2050 10
North Sea _Troll aquifer 14
Norwegian Sea 1,1
Barents Sea 0,2
Figure 12.11 (a) Storage capacities, maturation, and data quality in the evaluated geological formations, prospects, and hydrocarbon fields
on the NCS. (Source: Norwegian Petroleum Directorate.) (b) Total storage capacities in the different regions of NCS and the maturity of storage
sites. (Source: Norwegian Petroleum Directorate.)
206
Based on injection history
Development of injection site

Based on injection history
Injecti Suitable for long term storage

on
Development of injection site
t Exploration
G t
Effecti
ve an 1 G
d safe 1,
Cut o
storag
e 24 ) Theoretical volume
Theo
T Injectio
n Suitable for long term storage
ff crite t+ s Based on injection history
ria on G ield
v 43 f(
of inte olume/con
rest flict t Exploration
t Effecti G
Volum G ve an t Development of injection site
e calc 4 d safe 15 G
ulated
o Increased technical storag
e 0. 1 Theoretical volume
1. )
and th n average
ickne poros maturity Cut o s
t+
ss ity ff criteri G eld
a on v
4.4 (fi
of inte olume/con Injectio
n Suitable for long term storage
Volum rest flict
e calcu
The North Sea lated
o Increased technical Exploration
t
t
and th n average maturity Effecti G

G
ickne
ss
poros
ity
ve an
d safe 2
storag 0 .0 Theoretical volume
e
)
Cut off
ds .2
G
criteri
el 0
a on v
o
(fi t +
07
lu
0.
me/co
The Norwegian Sea of inte
rest n flict
t
Volum G
e calc 2 Increased technical
ulated
o 7.
and th n average maturity
ickne poros
ss ity
The Barents Sea
Figure 12.11b (Cont.)

Eva K. Halland
NPD (Norwegian Petroleum Directorate). Factpages. http://

References factpages.npd.no
CO2CRC. www.co2crc.au/ NPD (Norwegian Petroleum Directorate). (1995).
Deegan, C. E., and Skull, B. J. (1977). A standard Structural elements of the Norwegian continental shelf.
lithostratigraphic nomenclature for the central and Part II: The Norwegian Sea Region. NPD Bulletin, 8.
northern North Sea. NPD Bulletin, 1, www.npd.no/en/ www.npd.no/no/Publikasjoner/NPD-bulletin/258-
Publications/NPD-bulletins/251-Bulletin-1/ Bulletin8/
Halland, E., Bjørnestad, A., Magnus, C., et al. (2014). NPD (Norwegian Petroleum Directorate). (1996). Geology
CO2 Storage Atlas- Norwegian Continental Shelf. and petroleum resources in the Barents Sea: Norwegian
www.npd.no Petroleum Directorate (NPD), Stavanger.
Isaksen, D., and Tonstad, K. (1989). A revised Cretaceous Vollset, J., and Doré, A. G., eds. (1984). A revised Triassic
and Tertiary lithostratigraphic nomenclature for the and Jurassic lithostratigraphic nomenclature for the
Norwegian North Sea. NPD Bulletin, 5. www.npd.no/e Norwegian North Sea. NPD Bulletin, 3. www.npd.no/e
n/Publications/NPDbulletins/255-Bulletin-5/. n/Publications/NPDbulletins/253-Bulletin-3/
208
Chapter
Twenty Years of Monitoring CO2 Injection
13 at Sleipner
Ola Eiken
Introduction Reservoir
Sleipner is the world’s largest and longest-running The storage reservoir is a saline aquifer of regional
carbon dioxide (CO2) storage project, with injec- extent (e.g., Chadwick et al., 2004a; Zweigel et al.,
tion since September 1996. CO2 is separated from 2004); stretching for more than 400 km north to
natural gas at the production platform (Hansen south and 50–100 km east to west, locally exceeding
et al., 2005) and reinjected into highly porous thicknesses of 300 m in the Sleipner area. The CO2
sands containing saline water of the Utsira Fm., injection point is located beneath a domal feature
through a single near-horizontal injection well at that rises about 10–15 m above the surrounding
a depth of 1012 m below sea level. Water depth is seal topography (Figure 13.1). Significant faulting
80 m, and the top of the storage reservoir is 800 m with a structural origin is absent. The sand bodies
below sea level. About 17 Mtonnes of CO2 had are commonly separated by thin intrareservoir
been injected and stored by the end of 2017. mudstone beds, seen on logs, but barely visible in
The field is operated by Equinor (former name, three-dimensional (3D) seismic images (Figure 13.1;
Statoil). Bitrus et al., 2016). The depositional environment is
The Sleipner project, incentivized by the uncertain, possibly a turbiditic sand (Gregersen
Norwegian state tax on offshore CO2 emissions, et al., 1997).
was developed in a petroleum engineering setting The Utsira Sand characteristically shows a sharp top
(Baklid et al., 1996), with none of today’s carbon and base in both logs and seismic data (Figure 13.1),
capture and storage (CCS) legislation in place. with the proportion of clean sand in the reservoir unit
Geophysical monitoring was not part of the original typically above 70%. The nonsand fraction corresponds
plan, but spectacular four-dimensional (4D) seismic mostly to the thin mudstones (typically about 1 m
images since 1999 have been a driver for understand- thick), which show as peaks on the gamma-ray and
ing the injection and storage process, as well as resistivity logs. Except for one thicker shale near the
for gaining public and authority acceptance. top of the storage reservoir, it is not possible to correlate
The data have been vital for improving reservoir these shale layers between wells, or to identify them in
characterization, developing new analysis techniques, the baseline seismic survey.
cross-disciplinary work, verification, and prediction. Core and cuttings samples show the Utsira Sand
Much of the early monitoring work was undertaken to be mostly fine-grained and largely uncemented.
through EU- and industry-supported research Porosity estimates from core, based on microscopy
projects, summarized in, e.g., Arts et al. (2008), and laboratory experiments, are in the range of
Chadwick et al. (2008), and Chadwick and Eiken 27–42% and regional porosity estimates from geophy-
(2013). More than 300 scientific papers are testament sical logs are in the range of 35–40%. Permeabilities
to the wealth of knowledge that has evolved and are correspondingly high, with measured values (from
been shared, and the great influence of the project. both cores and water-production testing) ranging
Significant contributions have also been shared from from 1 to 8 Darcy.
the operator’s internal work of managing the site. This A significant spread in temperatures is used in
chapter aims to summarize the history and current the literature, with early work based on nearby wells
knowledge. and gravity data (Lindeberg et al., 2000; Nooner et al., 209
Ola Eiken
GR TVD DT
0 1874/1240 1872/1280 1882/1 N 0 150 (m) 240 40
TopUtsira
800
900
Base Correlatable Mudrock
950
1000
Base Utsira
1050
Time shift marker
Figure 13.1 (Left) Map showing traveltime to top Utsira Fm. Red is shallow, blue is deep. (Right) Seismic section and typical well log from
the Utsira Fm. (From Furre and Eiken, 2014.)
16 Figure 13.2 Injected mass and times of

gravity monitor seismic and gravity surveys. (From Alnes, 2015.)
14
12
Accumulated injected CO2
10
(million tons)
seismic monitor
6
gravity base
0
96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
Year
210
Twenty Years of Monitoring CO2 Injection at Sleipner
Table 13.1 Injection gas composition for three samples, in mol%
Component 1998 Sample 2002 Sample 2006 Sample

CO2 98.07 98.38 96.53
CH4 1.23 1.29 1.85
C2H6 0.19 0.17
N2 0.10 1.15
BTX 0.40 0.33 0.30
Source: Equinor.
2007). Water production from the Utsira Sand at the injected at relatively uniform rates of about 0.8–0.9
nearby Volve field (approx. 8 km distant) since 2007 Mtonnes per year (Figure 13.2). There are no com-
yields robust reservoir temperatures, summarized by mercial rate measurements, which results in greater
Alnes et al. (2011). A water temperature of 32.2°C than 10% uncertainty in the volumes injected.
was measured during flow, from a perforation The injection fluid contains 1–2% methane, and smal-
interval of 822–1009 m with unknown inflow profile. ler amounts of nitrogen and heavier hydrocarbons
Projecting these values vertically gives a linearized (Table 13.1).
relationship T(z) = 31.7z + 3.4 (± 0.5) °C, where z is The injection well 15/9-A-16 is highly deviated,
in kilometers below the sea surface. Applying this to with a sail angle of 83° at the injection point.
the Sleipner injection area gives virgin temperatures The wellbore lies beneath the buoyant CO2 plume,
of about 29°C at the reservoir top and 35.5°C at the with no direct contact in the upper part of Utsira Fm,
depth of injection (1012 m). CO2 at these tempera- and hence no increased containment risk. Injectivity
tures and hydrostatic pressure would be supercritical challenges caused by sand influx during the first
and dense. months were remediated by a reperforation and instal-
There is no published formation water composi- lation of a 38 m long gravel pack in August 1997
tion from Utsira Fm. at Sleipner. Properties have been (Hansen et al., 2005).
assumed to be close to seawater (about 35 g/L). The temperature of the CO2 injectant is set at the
Gregersen (1998) reported salinity of 30–43 g/L at last stage of compression, and has mostly been 25°C.
the Grane and Oseberg fields 100 and 240 km north Wellhead pressure has been around 63–66 bar
of Sleipner. Under these conditions, water density is (Figure 13.3). The state of CO2 at the wellhead is at
within 1020–1031 kg/m3. Modeling studies have used the gas/liquid phase boundary, but the ratio has
values of 1020–1021 kg/m3 (Zhu et al., 2015; Williams not been measured (Hansen et al., 2005). The CO2
et al., 2018), which is at the lower end of the range, and passes into supercritical state down the wellbore
would require salinity of about 30 g/L. (Figure 13.4). The 1–2% methane content decreases
The 700 m thick overburden comprises a primary the required gas/liquid ratio slightly compared to pure
caprock comprising a 50–100 m thick mudstone CO2, but otherwise has little effect on the injection.
(Zweigel et al., 2001; Harrington et al., 2010). Above The bottom hole pressure could vary from 90 to
this, prograding sediment wedges are dominantly 140 bar for the same wellhead pressure and tempera-
muddy, but coarsen into a sandier facies upward. ture, dependent on the gas/liquid ratio at the wellhead
The shallower overburden is mostly glaciomarine (Figure 13.5; Lindeberg, 2010; Kiær, 2011).
clays and glacial tills. The calculated bottom hole temperature is 49–51°C
for gas/liquid ratios below 0.8 and 43°C if all the CO2
was in liquid phase at the wellhead. In any case, the
Injection injected CO2 will be warmer than the formation, and
CO2 is separated from the Sleipner Vest gas, since become cooled down in the reservoir (Alnes et al.,
2014 also from the Gudrun field, and has been 2011). 211
Ola Eiken
8.5
8.0
Wellhead pressure [MPa]
7.5
7.0
6.5
6.0
5.5
97 98 99 00 01 02 03 04 05 06 07 08 09 10 11
19 19 19 20 20 20 20 20 20 20 20 20 20 20 20
Figure 13.3 Measured wellhead pressure up until 2011. The increase in 2002–2003 was related to the temperature setting being
increased above 25°C. (From Kiær, 2011.)
900 Figure 13.4 Density of pure CO2 (kg/m3) as

function of pressure and temperature.
60 Wellhead An estimate of the conditions down the
800 injection well is shown in black. Source: Equinor.
70
700
80
Density [kg/m3]
600
Pressure [bar]
90 500
100
X: 48.21 400
Y: 105
P,T in
300
110 formation
P,T at injection
point 200
120 (calculated)
100
20 30 40 50 60 70
Temperature [C]
The injection point was placed 2.5 km NE of the buoyant flow beneath the cap rock, assumed CO2
Sleipner A platform (Figure 13.1), to prevent or delay would flow in a northerly direction. Baklid et al.
backflow to the production wells, where they pene- (1996) envisioned an upward coning plume, as they
trate the storage formation. Exposure to CO2 would drew in Figure 13.6, with lateral spreading underneath
increase the risk of casing corrosion or CO2 leakage the top seal. Detailed flow modeling was not used to
outside the casing. Engineering assessments, based on assist the decision of well placement.
212
52 14 30
51
Temperature at well perforation [ºC]
Pressure at well perforation [MPa]

13
50
Retention time in well [min]

25
49
12
48
20
47 11
46
10 15
45
44
9
43 10
0.0 0.2 0.4 0.6 0.8 1.0
Mass ratio in liquid form at wellhead [kg/kg]
Figure 13.5 Calculated pressure and temperature at the perforation, as function of the gas/liquid ratio at the wellhead, assuming adiabatic
conditions in the well. Retention time for CO2 in the well is shown in green, and varies from 10 to 30 minutes. (From Kiær, 2011.)
Sleipner A
0 Sleipner T
500m
CO2 Injection Well

1000m
CO2
Utsira
Formation
1500m
Sleipner ∅st
Production and Injection Wells
2000m
0 500m 1000m 1500m
2500m
Heimdal Formation
Figure 13.6 Sketch of the CO2 injection with production wells and injection well. Envisioned CO2 distribution after 20 years of injection is
marked in orange. (From Baklid et al., 1996.)
213
Ola Eiken
400
Base line Difference
SW NE SW NE
500
600
700
Two-way-time [ms]
800
9
900
7 8 6
5
1000
1–4
1100
1200
1300
1500 1400
1,500 meters
1500 m 5,000 feet
Figure 13.7 Vertical sections of (left) the 1994 baseline. (Right) The difference between the 2013 and 1994 seismic data. (From Furre et al.,
2017.)
Monitoring CO2 was distributed in nine reflective levels, and

a small amount of CO2 had reached the top seal.
The technology for downhole pressure gauges was
The upper levels (layer 5–9) have been growing
not well established in 1996, when the injection well
since, causing both a densification and expansion of
was drilled. Retrofitting such sensors imposes risks to
the CO2 plume. The seismic survey interval has been
the well integrity, and has not been undertaken.
two to three years (Figure 13.2; Table 13.2), and no big
The overpressure during injection has likely been
surprises have emerged from one survey to the next.
a few bars only, because of the high permeability and
In addition, four gravity surveys, one controlled
the 200 m thick sandy formation above the perfora-
source electromagnetic (CSEM) survey, and one sea-
tion. CO2 is pushed in all directions, including down-
floor imaging survey have provided complementary
ward several tens of meters. Increasing relative
data.
permeability as a result of higher CO2 saturations in
The first data acquisitions were research funded,
the pathway above are thought to have reduced the
and from 2008 on, the operator took over responsi-
overpressure near the perforation after some time.
bility. The seismic monitoring results are now
No injection pressure data are available to test such
required by, and reported to, the Norwegian pollution
a model.
authorities. Equinor has made all 4D seismic data up
At Sleipner, seismic monitoring is by far the most
to 2010 publicly available, together with well logs and
efficient containment monitoring tool, owing to the
depth surfaces.
ability to remotely cover a large area, the strong time-
lapse reflections (Figure 13.7) even for small CO2
saturations, and consequently the ability to detect Seismic
small accumulations of CO2 filled pore space (of All surveys have been acquired using streamers
order 1 m vertically and 10 m laterally). (Table 13.2), covering roughly the same 4 km ×
The overwhelming seismic response has for many 7 km area. The six first had the combined purpose
become a symbol of the Sleipner storage project. of monitoring deeper hydrocarbon targets, which
214 Already in 1999, after three years of injection, the reduced the cost and made some of them possible
Table 13.2 Seismic acquisition parameters
Year/date Comment Shooting Source Flip-flop Shot-point No. of Cable Swath Cable Cable
direction tow depth x-line interval (m) cables sepa- separation length tow
(m) separation ration (m) (m) (m) depth
(m)
8/6–9/10, 1994 Preinjection, N–S 6 50 18.75 5 100 250 3000 8
baseline
10/8–10,1999 First image N–S 6 50 12.5 4 100 200 3600 8
of CO2
9/27–10/1, 2001 N–S 6 50 12.5 6 100 200 3000 8
5/26–6/1, 2002 N–S 6 50 18.75 6 100 300 3600 8
6/13–8/13, 2004 Orthogonal E–W 6 one 18.75 10 37.5 250 4500 8
line direction
6/2–20, 2006 N–S 6 50 8 100 300 3600 8
High- (2D) 3 1 — 3
resolution
lines
5/4–6/15, 2008 N–S 6 50 18.75 9 50 200 3000 8
10/15–17, 2010 Dual sensor N–S 5 37.5 12.5 12 75 375 6000 15
streamer
1/2013 Slanted N–S 7 37.5 12.5 8 100 5100 9
cable
2016 N–S
Ola Eiken
to realize at all. Tow depths of 6 m and 8 m are assuming 200 m thickness of the Utsira sand and
optimized for 2–3 km targets, and cause ghost 37% porosity. The 5–6% number on storage
notches that attenuate frequencies above 60 Hz. efficiency has increased slightly over the years.
The dominant frequency in these images is about Reflection amplitudes have generally been increas-
35 Hz, and temporal thicknesses can be derived ing over the years, giving the plume growth a
with an accuracy of 1–2 ms (Williams and character of simultaneous expansion and densify-
Chadwick, 2012; Furre and Eiken, 2014). High- ing. The growth of each layer, measured in area-
resolution two-dimensional (2D) lines with 3-m tow integrated amplitude, is shown in Figure 13.10.
depths, as well as broadband technologies (Furre and In the early years the shallower layers (6–9) were
Eiken, 2014), have demonstrated the ability to get small, and have grown faster over the last 10 years.
frequencies above 100 Hz reflected back from the Layers 1–4 have, on the contrary, stopped growing,
CO2 plume. and some of them have apparently shrunk. They
The number of cables has increased with technol- have become more challenging to interpret, which
ogy developments, from 4–6 in the first surveys could be caused by inelastic attenuation, transmis-
to 8–12 in the most recent ones (Table 13.2). sion loss, CO2 migration/dissemination or all of
Overlapping swaths have reduced some of the azi- these (Boait et al., 2011). The area-integrated sum
muth and coverage variations at the outer streamers. of seismic amplitudes has been remarkably propor-
The datasets have been processed in a time-lapse tional with injected mass (Eiken et al., 2011).
manner to enhance repeatability, including global Layers 5 and 9 were by 2010 the largest ones.
matching and stacking/migration velocity reduc- The uppermost layer 9 does not suffer from the
tion in the central part of the plume in subsequent “shadow” effects of the lower layers, and is well
surveys. Much of the analysis and interpretation of imaged, as the amplitude maps in Figure 13.11
the plume development has been done on time- show. Its growth clearly follows topographic highs.
migrated full-offset cubes. Normalized root mean After some years of accumulation above the injec-
square (NRMS) repeatability values are 50–60% for tion point, the CO2 has spread rapidly northward in
the full frequency range (Furre and Eiken, 2014), a narrow corridor or channel, and started accumu-
and time shift noise is generally below 1 ms. lating below a structural high approx. 3 km north/
The orthogonal shooting direction in 2004 caused northeast of the injection point. The highest ampli-
variations that cannot be compensated in the pro- tudes are found in the corridor (Figure 13.11;
cessing, causing higher NRMS values of 80% when Chadwick et al., 2009, 2016).
this survey is involved. All surveys and differences Pushdown can be observed as an increasing time
have high signal-to-noise ratios due to the large delay for a horizon in later vintages, and is caused
contrast in acoustic properties between the in situ by lower velocities and longer travel times for the
saline aquifer and the injected CO2, and have seismic waves passing through CO2-saturated rock.
been valuable for understanding the plume The amount of pushdown is increasing down through
development. the plume, with a maximum at the marker horizons
Figure 13.8 shows the time-lapse development in below. It has become increasingly difficult to reliably
reflection response, in a section along the crest of the estimate the pushdown, owing to poorer image qual-
NNE–SSW trending anticline. The bright layers 1–9 ity in the lower part and below the plume in later
(Chadwick et al., 2004b) can be easily recognized in all years. The strong and rapid lateral velocity variations
the surveys. These are interpreted as thin layers of in the CO2 plume complicate the processing and
CO2-filled sand capped by shale strings. Injected distort the poststack time-migrated cubes. Seismic
CO2 thus acts to illuminate the internal reservoir tomography and prestack depth migration could
architecture. potentially handle these variations better, but
The visible plume area has been growing in all attempts have so far shown limited improvements
directions (Figure 13.9), and fastest to the NNE, (Evensen and Landrø, 2010; Rossi et al., 2011;
giving an increasingly elongated shape. The area Romdhane and Querendez, 2014). Broto et al. (2011)
was 3.5 km2 in 2010, which means that 5–6% of were able to extract horizontal velocity slices from
216 the pore volume under this area is filled with CO2, tomographic analysis of the 2006 data that resemble
sw0276
1829/994 1814/1032 1820/1065 1838/1102 1876/1159 1877/1200 1874/1240 1872/1280 1882/1300 1902/1335 1922/13
S N
800
2001–1994
1000
800
2004–1994
1000
Two-way-time [ms below sea surface]
800
2006–1994
800 1000
2008–1994
1000
800
2010–1994
9
8
7 6
5
1000
4 3 2
1
Figure 13.8 Seismic time-lapse difference with base survey for five subsequent repeats. White numbers refer to CO2 layers. The white
dashed ellipses show an area of strong amplitudes in the baseline, which serves as a reference for the time-lapse repeatability. (From Furre
and Eiken, 2014.)
the pushdown maps, with most of the low-velocity a horizon or of cross-correlation between vintages.
area north and east of the injection point. Pushdown The observed pushdown was up to about 50 ms
measure can be the time of either peak amplitude of already in 1999 (Chadwick et al., 2004b), in a narrow
217
Ola Eiken
Figure 13.9 Time-lapse seismic difference reflection amplitude maps, cumulative for all layers. (Rightmost) The rim of the plume for
each year. (Courtesy of Equinor.)
11 Figure 13.10 Layer growth quantified as

layer 9 amplitude * area, up until 2010.
10 layer 8
layer 7
9 layer 6
layer 5
8 layer 4
Amplitude * area
layer 3
7 layer 2
layer 1
6
5
4
3
2
1
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160
Months of injection
218
100
Amplitude relative to maximum %

1 km 90
80
70
60
50
40
30
20
10
1999 2001 2002 2004 2006 2008 2010 0
Figure 13.11 Layer 9 amplitude differences with the baseline survey. (From Kiær et al., 2015b.)
10 ms 1999–1994 2001–1994 2002–1994 2004–1994

0
Chimney
5
10
15
TWT ms below 0 meters
20
1,000 meters 1,000 meters 1,000 meters 1,000 meters

4,000 feet 4,000 feet 4,000 feet 4,000 feet
25
2006–1994 2008–1994 2010–1994 45 000
40 000
30
35 000
Area integrated time shift [m2s]
35
30 000
25 000
40
20 000
15 000
45
10 000
5 000
50
0
1995 2000 2005 2010 2015
55
1,000 meters 1,000 meters 1,000 meters Vintage

4,000 feet 4,000 feet 4,000 feet
55 ms
Figure 13.12 Pushdown maps. White disc denotes injection point. (Courtesy of Equinor.)
area (Figure 13.12). The area-integrated pushdown at all levels, is observed in the center of the plume
has increased regularly from about 11 to 40 × 103 (Figure 13.13). It can be observed only after 3D migra-
m2s between 1999 and 2010 (Chadwick et al., 2004b; tion. Arts et al. (2009) demonstrated that a vertical
Boait et al., 2012). cylinder completely filled with low-velocity material
A seismic “chimney,” circular with diameter such as CO2-saturated rock can cause such seismic 219
60–90 m and with disturbances of the flat reflections features. The center is 90 m to one side of the
Ola Eiken
Figure 13.13 Seismic chimney. (Bottom left) 1994. (Bottom right) 2006. (Courtesy of Equinor).
perforation, which is 2σ on the sideway well position sand core sample flooded with CO2/brine, and this
uncertainty, and it is thus possible that the chimney is can be used to calibrate a rock physics model.
directly above the perforation. It could have been Whether the CO2 is homogeneously or patchily dis-
created by the injection process (Hermanrud et al., tributed, remains an ambiguity that limits the inter-
2009), possibly by the flow of CO2 breaking up the pretation of saturations from measured or inferred
sealing capacity of the thin shales. On the other seismic velocity reductions. Chadwick et al. (2016)
hand, a circular feature was seen at the top Utsira estimated velocities in the topmost CO2 filled layer
Fm. level in the preinjection data (Figure 13.13, in the range of 1350–1430 m/s, consistent with values
lower left), indicating a relation with preexisting geol- derived independently from rock physics.
ogy. In either case, the chimney seems important for Reflection amplitudes generally increase from the
vertical migration of CO2, as both layer amplitudes rim and to the center, as seen for layer 9 in
and pushdown increase toward it. Figure 13.11. The amplitude variations have been
explained by thin-layer tuning of two strong acoustic
contrasts of opposite sign at the top and bottom of
Quantitative Interpretation a layer filled with CO2 (Figure 13.15; Arts et al., 2004b;
From rock physics modeling, Arts et al. (2004a) Chadwick et al., 2004b, 2005). Average layer thickness
expected initial P-wave velocities in the Utsira Fm. of a few meters is to be expected: By 2010, when 12
of 2050 m/s and decreases of 600 m/s or more after Mtonnes had been injected, the cumulative area of the
CO2 flooding. Queisser and Singh (2013a) used simi- nine layers was 12 km2, giving an average CO2 column
lar models, but with lower velocities for fully CO2 height of 1.3 m in each layer, corresponding to a rock
saturated rocks, and a range of curves between the column of about 5 m if 70% saturated. Amplitudes
saturation end points, depending on the way of mix- arising from thicknesses below maximum tuning can
ing of CO2 and water (Figure 13.14). Falcon-Suarez be converted to time thickness, when the wavelet and
220 et al. (2018) measured acoustic properties of an Utsira the maximum amplitudes are known. Further, layer
Figure 13.14 P-wave velocity versus CO2

saturation, for various rock physics models.
Upper patchy bound (triangles), modified upper
patchy bound (green), Brie’s curve (red),
modeled lower bound (blue and gray), lower
homogeneous bound (dotted). The inset shows
the zoomed view of the homogeneous and
modeled lower limits. (From Queisser and Singh,
2013a.)
of topography (Chadwick and Noy, 2010). After about

10 years of injection, layer 5 showed decreasing ampli-
tudes in the central part, interpreted as the temporal
thicknesses exceeding maximum seismic tuning.
With increasing layer thicknesses, time separation of
top and bottom reflections becomes more reliable for
thickness estimates (Furre et al., 2015). Spectral
decomposition analysis can further enhance the
lower resolution limit of the seismic data, utilizing
the properties of the tuning wavelet and spectral
decomposition (Williams and Chadwick, 2012;
White et al., 2013).
Arts et al. (2004a) compared pushdown below
a single layer at the edge of the plume with the
reflection amplitude from that layer, and found on
average a relation resembling the tuning curve.
The pushdown/amplitude ratio is highest at the
Figure 13.15 Amplitude from a CO2-saturated wedge as center of the plume, where a <500 m wide area has
a function of CO2 layer thickness. Different colors represent strong pushdown (Figures 13.9 and 13.12). Some of
increasing CO2 saturation downward. (From Furre et al., 2015.) this could be caused by layers having exceeded the
maximum tuning thickness, and, probably more
important, additional pushdown is caused by non-
thicknesses can be calculated from time thickness and reflective CO2 (Chadwick et al., 2005), for example,
the velocity of CO2 saturated rock. Various estimates in the migration pathways. Ghaderi and Landrø
(Chadwick et al., 2005; Chadwick and Noy, 2010; (2009) developed a method for estimation of thick-
Furre et al., 2015; Kiær, 2015) have given similar ness and velocity changes of CO2 layers for prestack
results, with minimum thin-layer detection level of time-lapse data, and applied the method to a small
about 1 m, and maximum tuning at about 8 m. For part of the plume with one layer only.
layer 9, the thicknesses derived from amplitude tun- Reflective layers with inferred thicknesses can 221
ing fit with a flat CO2–water contact (CWC) and infill account for most of the injected mass, if they are
Ola Eiken
Figure 13.16 The inferred total mass bounds

for 2006 plotted against the patchiness
parameter. (From Bergmann and Chadwick,
2015.)
highly saturated (Chadwick et al., 2004b). The spike of layer 5 to have lower reflectivity than the
additional CO2 needed to explain the pushdown in upper. Kiær et al. (2015a) investigated the amplitudes
the center of the plume must be at lower saturations to close to the rim, and found the rapid decline in
comply with the known injected mass, dependent on amplitude when approaching the rim to be inconsis-
the velocity–saturation relation. Bergmann and tent with the topography, tuning relationship, and
Chadwick (2015) discussed this ambiguity in mass assumption of gravity dominated flow. Introducing
calculations from pushdown from a methodological capillary fringe effect alone did not help, but including
point of view. They introduced the patchiness para- macroscopic fluid flow for low frequencies can give
meter, and found that the injected mass is within the a better match. The highly porous and permeable sand-
upper and lower bound (Figure 13.16). Boait et al. stones can cause wave-induced fluid flow (WIFF) at
(2011) suggested that the CO2 is tending toward uni- seismic frequencies, which may reduce the acoustic
form mixing within the highly saturated layers and velocity for the lowest frequencies (Hofmann, 2006).
patchy mixing elsewhere, which could provide For a thin layer of CO2 capped by a string of shale, this
another constraint on the calculation of the amount could reduce the reflectivity at the lower interface
of CO2 at low saturations. The linear growth of the (CWC) and thus alter the assumed tuning relation.
area-integrated pushdown may indicate that the WIFF is especially significant in the case of patchy
principal characteristics of the velocity saturation saturation, and will also make the pushdown factor
in the CO2 swept regions have remained rather con- frequency dependent (Rubino and Velis, 2011).
stant through the years (Bergmann and Chadwick, Rubino et al. (2011) suggested the effect can modify
2015). The pushdown has, to a limited extent, been the seismic attributes significantly, but this has not yet
brought further from saturation distributions to flow been thoroughly investigated in the quantitative inter-
models. Boait et al. (2012) analyzed the evolution of pretations at Sleipner.
both reflectivity and pushdown for all layers and Several attempts at seismic inversion have been
noted that pushdown has increased in the central made. Stratigraphic inversion (Clochard et al., 2010;
part while amplitudes have dimmed, suggesting Delépine et al., 2011; Labat et al., 2012) showed layers
a continued growth of interlayer CO2. of low acoustic impedance. It is not clear how this
Chadwick et al. (2012) could not discern time shifts compares to layer thicknesses found from amplitude
outside the CO2 plume from noise, which indicates that inversion by the thin-layer tuning assumption.
pressure buildup likely does not exceed 1 bar. The inverted layers are possibly thicker and with less
Borgos et al. (2003) relaxed the thin-layer tuning velocity reduction. Jullum and Kolbjørnsen (2016)
222 assumption and inverted for a pair of opposite polar- developed a framework for Bayesian inversion,
ity reflections of variable size. They found the lower inverted a 2D line for saturation, and found the
Figure 13.17 The P-wave velocity model derived from FWI. The black line corresponds to the projection of the injection well into the seismic
section. The 2D section is located 533 m east of the injection point. (From Debuy et al., 2017b.)
same five CO2 layers as could be seen in qualitative discrimination of layer thickness effects, where con-
interpretation of the seismic data. Queisser and Singh ventional Zoeppritz-based methods usually fail.
(2013a, 2013b) tested full waveform inversion (FWI) Rabben and Ursin (2011) found from AVA inversion
on a 2D line, and related the velocity changes to CO2 of the top sand reflection good results for the P-wave
saturation using a rock physics model. They found impedance, while S-wave and density were overesti-
that applying FWI to real seismic data was challen- mated. Landrø and Zumberge (2017) combined AVO
ging. Possibly they could identify places where the and gravity data to better constrain the density
CO2 saturation had reached maximum. Raknes et al. changes when the multilayered CO2 accumulations
(2015) carried through a 3D elastic FWI study. Ghosh limit the AVO technique. Haffinger et al. (2017)
et al. (2015) inverted poststack data for saturation, used wave-equation–based AVO to calculate injected
and found large uncertainties attached to the rock mass, without uncertainty bounds or comparison
physics modeling. Romdhane and Querendez (2014) with other work, however. Dupuy et al. (2017a)
inverted, and then prestack depth migrated, a 2D test included rock physics models in their AVO workflow,
line and Dupuy et al. (2017b) used FWI (Figure 13.17) and were able to derive high-resolution impedance
and rock physics models to arrive at saturations in the contrasts, but not elastic contrasts.
thin layers of 30–35% for homogeneous mix and 75% No significant changes have been observed in the
for patchy mix, with true mixing between these. seismic response from the overburden, in any survey
Owing to uncertainty in the choice of saturation (Chadwick et al., 2017). How sure can we be that no
model, velocity sensitivity to gas saturation is uncer- CO2 has leaked? Chadwick et al. (2014) estimated
tain, meaning that without additional constraints, detection limits based on data repeatability, the area
CO2 saturation estimates are not conclusive from and thickness of a CO2 accumulation above the reser-
seismic inversions alone. The Sleipner inversion stu- voir, and the probability of identification. As CO2
dies have been method developments, but have not migrates upward, it passes into gas phase and smaller
been vital to reservoir understanding or estimating masses can be detected. Pore volumes of about 3000
CO2 plume properties. More extensive comparisons m3 can be robustly detected at the top of the Utsira
and testing of inversion results against other techni- reservoir, corresponding to about 2100 tonnes, while
ques such as amplitude tuning analysis might add the detection threshold will be around 950 tonnes at
further insight. 590 m and 315 tonnes at 490 m below sea level
It has not been easy to retrieve additional informa- (Chadwick et al., 2014). These numbers are the
tion from amplitude variations with offset (AVO). order of 10−3 of the yearly injection volumes and
Tuning effects need to be disentangled (Neele and sufficient to ensure containment has occurred 223
Arts, 2010). Chadwick et al. (2010) focused on the (Chadwick et al., 2014). Leakage along narrow vertical
Ola Eiken
Figure 13.18 Maps of measured

gravity change, outline of the Ty Fm. in
green and outline of CO2 plume in 2013
in black. (Left) Gravity changes
2009–2002. (Right) Gravity changes
2013–2009. (From Alnes, 2015.)
20 uGal 20 uGal
paths at high rates and with smaller accumulation in The effects of vertical benchmark movements,
the overburden is possible, but unlikely. water influx in the deeper gas reservoirs, and CO2
injection have been simultaneously inverted in several
Gravity attempts based on different routes of data processing
(Nooner et al., 2007; Arts et al., 2008; Alnes et al.,
High-precision gravity monitoring offers direct mea-
2008, 2011; Alnes, 2015; Furre et al., 2017).
surements of mass changes. In 2002, 30 seafloor
The estimated maximum gravity reduction caused
benchmarks were installed in a main profile over
by CO2 injection increased from about 10 μGal in
and beside the CO2 plume, with some further aerial
2002–2005 to more than 20 μGal over the time inter-
coverage (Figure 13.18), and a baseline was acquired
val 2002–2013 (Figure 13.19, right side). For inver-
using the ROVDOG technique, described in
sions, the geometry of the CO2 plume as imaged in 4D
Chapter 7. Subsequent surveys were conducted in
seismic data and the injected amount of CO2, were
2005, 2009, and 2013. As the CO2 plume expanded,
further constraints.
13 additional benchmarks were installed in 2009 and
The estimated average CO2 density show some
2013. Gravity data can give information that is com-
scatter, as discussed in Chapter 7, with the most recent
plementary to the seismic data, with their limited
values of Alnes et al. (2011) and Furre et al. (2017)
ability to quantify CO2 saturations.
based on the longest time span and most advanced
Eroding currents causing seabed scouring has low-
analysis technique being of highest confidence. Alnes
ered some of the benchmarks more than 10 cm into
et al. (2011) arrived at 720 ± 80 kg/m3 for average CO2
the sediments since deployment. After correcting the
density. They used the constraint on rock temperature
gravity for the measured height changes, using
from the Volve field water production well, discussed
a vertical gravity gradient derived from the data,
earlier in this chapter.
a time-lapse gravity increase of nearly 40 μGal was
The gravimetric measurements can be used to
observed at the westernmost station, gradually
estimate an upper bound of the dissolution rate of
decreasing eastwards (Figure 13.18). This is explained
CO2 in the formation water. This is important for
by water inflow to the deeper gas-condensate reser-
the long-term fate of CO2, because brine with CO2
voirs (Ty Fm., outlined in green in Figure 13.18).
dissolved is heavier than pure brine and forms
The water volumes required to explain the gravity
a more stable phase for storage. Dissolved CO2
changes agree with later acquired saturation logs in
is seismically invisible. Alnes et al. (2011) estimated
a nearby well and with 4D seismic data (Alnes et al.,
a dissolution rate between 0 and 1.8% per year.
2008). The stations east of the CO2 plume act as
Hauge and Kolbjørnsen (2015) performed a
references in an area undisturbed by both CO2 injec-
224 Bayesian analysis of the in situ CO2 density and
tion and gas production.
37
10 uGal
Thickness change (m)

Injection
point
Figure 13.19 (Left) Seismic amplitude changes. (Middle) Seismic time delay. (Right) Estimated gravity changes and inverted CO2 thickness,
all from 2002 to 2013. (From Alnes, 2015.)
Figure 13.20 Probability distribution of

dissolution rate, based on the 2002–2009 time
span. (From Hauge and Kolbjørnsen, 2015.)
dissolution, using the same 2002–2005–2009 data Controlled Source ElectroMagnetic

set, with the Ty Fm. and height change effects
removed. They arrived at similar conclusions, Line
as shown in Figure 13.20. After the 2013 data A controlled source electromagnetic (CSEM) line,
were acquired, Alnes (2015) updated the dissolution centered at the long axis of the CO2 plume, was
estimate to less than 2.7% per year, based on acquired in 2008. Twenty-seven seabed receivers
the longer 2002–2013 time span. He also concluded were distributed over 10 km, and the total towline
that the accuracy is limited by uncertainty in length was 30 km. Detectability challenges are related
the subtracted signal from water influx to the Ty to the water depth of 80 m, causing stronger airwave
Fm. and lack of gravity stations over the northern influence than common for most offshore CSEM sur-
part of the plume in the base survey. Alnes (2015) veys, and interference from the network of pipelines
further inverted for CO2 column height, which on the seafloor. Early studies of the data set were
interestingly agrees more with the one derived nonconclusive, but in the most recent analysis, Park
from seismic pushdown than from reflection ampli- et al. (2017) obtained inversion results with more
tudes (Figure 13.19). detailed high-resistivity areas, which they related to 225
Ola Eiken
Figure 13.21 Vertical resistivity (in Ωm)

after 2.5D inversion, superimposed on
seismic data. (From Park et al., 2017.)
the CO2 plume (Figure 13.21). Quantitative numbers Flow Dynamic Interpretations
on, for example, average saturation are still imprecise
Flow modeling can help explain the 4D seismic and
and rely heavily on model assumptions. Eliasson and
understand dynamic mechanisms. Lindeberg et al.
Romdhane (2017) initiated investigations of the value
(2000) found the first monitor survey to be in fair
of combining CSEM inversion and FWI.
agreement with one premonitor model having thin
shale layers as vertical barriers. While early attempts at
Seafloor Monitoring Sleipner were few and scattered, flow modeling studies
The seafloor above the CO2 plume consists of silt at have increased in number and the matches have
82 m water depth, planar and gently dipping about improved over the last decade. Bickle et al. (2007)
1:1000 to the west. Bottom currents are moving the calculated analytically an axisymmetric gravity flow
top sediments around. The seabed was surveyed with for each layer, and obtained a match with seismically
multibeam echo-sounding and side-scan sonar in derived radii and thickness distributions, but had to
2006. In 2011 and 2012, University of Bergen used reduce permeability by an order of magnitude com-
AUV-based side-scan sonar (HISAS 1030) with pared to the expected 1–8 D range. A cause could be
improved 2 × 2 cm resolution and photo mosaic to reduced relative permeability at lower CO2 saturations.
obtain improved seafloor images. None of the above A delay in the feeding into upper layers, and high rates
techniques have indicated any leakage from the of accumulation prior to 1999 for the lower layers, was
Sleipner CO2 injection site (Pedersen, 2011). This included in the model. Chadwick and Noy (2010) had
is not surprising, as we would expect to detect difficulties modeling Darcy flow with a sufficiently
a leakage on 4D seismic data long before it reached rapid northward progression of layer 9 in the channel
the seabed. (Figure 13.11). Permeability needed to be increased to
Bacterial mats have been found on the seafloor at least 10 D – or made anisotropic. Singh et al. (2010)
40 km north of the injection site (Monastersky, compared invasion percolation (capillary flow) and tra-
2013). These are likely fed from seepage, but highly ditional (Darcy flow) simulation, using black oil and
unlikely to be connected to the injection process, compositional fluid descriptions. They concluded that
because of the long distance from the injection the two approaches either propagate the CO2 of layer 9
point, the small pressure buildup in the CO2 too far or too short and circular, compared to the
plume, the lack of 4D seismic signals farther than seismic observations. However, the two methods pro-
4 km north of the injection point, and higher vide upper and lower bounds to the expected plume
volumes of water production at the Volve field behavior. Singh et al. (2010) also introduced the
than CO2 injection at Sleipner, which cause a net Sleipner benchmark model (www.ieaghg.org/networks
suction of mass into the Sleipner-Volve region. /modelling-network), publicly available and based on
Furre et al. (2014) suggested the linear feature is the 4D seismic. A second benchmark model was intro-
related to gas within glacial channels that are pre- duced by Cavanagh and Nazarian (2014) and covers
sent beneath the bacterial mats. the two uppermost layers (8 and 9). Boait et al.
226
Figure 13.22 Observations (leftmost) and models of the extent of layer 9. Observed rim of plume is drawn as black or red line on top of the
models. From left to right: Amplitude observations 2008, Darcy flow model with 2/10 D permeability for 2006 from Zhu et al. (2015), capillary
flow model for 2008 from Cavanagh and Nazarian (2014), Darcy flow model for 2010 with 3/8 D permeability from Williams and Chadwick
(2017). Darcy flow model for 2010 with parameter update and matching from Nilsen et al. (2017).
(2012) analyzed the horizon growth up to 2008, dis- thought, and is likely to further stabilize after
cussed models for vertical migration and dimming injection as a result of dissolution. Cavanagh and
of the lower layers, and concluded that flows are Haszeldine (2014) modeled all layers with capillary
controlled primarily by the spreading of gravity cur- flow using invasion percolation physics, and were able
rents, moderated by topographic infilling. Cavanagh to match the seismic only when they included
(2013) modeled layer 9 with both capillary and fractured shale barriers. The seismic chimney was not
Darcy flow, and found that both can match the considered part of the vertical flow, however. Bandilla
seismic observations. The latter needed to have et al. (2014) compared modeling approaches and found
the overpressure dissipated, however, requiring vertical-equilibrium models able to roughly match layer
a much longer time span for the model run. The 9. However, none of their models were able to accu-
migration simulator was highly sensitive to top seal rately match in detail, suggesting some essential physics,
topography. Dubos-Sallée and Rasolofosaon (2011) e.g., topography of the caprock, was incorrect. Kiær
observed the asymmetric spreading in their seismic (2015) adjusted the topography of the top seal assuming
inversion results, and suggested permeability aniso- gravity dominated flow, based on all seismic vintages up
tropy as the main cause. to 2010. The resulting topography map deviated by
Modeling results have been difficult to match in 5.3 m on average from the time-to-depth conversion
detail with the time-lapse seismic (e.g., Zhang et al., that has been the basis for all the modeling work. This is
2014; Zhu et al., 2015). Some examples are shown in within the mapping uncertainty (Chadwick and Noy,
Figure 13.22. This is due partly to challenges under- 2010).
standing the underlying physics of the CO2 flow, and Boait et al. (2012) found that mechanisms of CO2
partly to uncertainties in geological assumptions such penetration through the mudstone layers needs to be
as the top seal topography and the vertical barriers. better understood, to provide a model for plume
Cavanagh and Nazarian (2014) argued that the plume development. If the chimney consists of broken-up 227
is closer to a stable distribution than previously shales with much increased permeability, it will be
Ola Eiken
a key to correct modeling of the plume growth. Cavanagh et al. (2015) argued that buoyant capillarity
Williams et al. (2018) adjusted permeability and capil- is the dominant process at Sleipner, while Williams
lary entrance pressure to match the observed feeding and Chadwick (2017) revised the parameters going
into each layer. Such a “semipermeable” behavior into Darcy flow modeling, and were able to mimic the
might reflect preferential flow through pervasive rapid northward propagation of the CO2 front in layer
small-scale fractures, as proposed by Cavanagh and 9 with 8 Darcy permeability where the uppermost
Haszeldine (2014). sand has thicker channels. Whereas Williams and
The modeling work has triggered a discussion of Chadwick (2017) were able to obtain gross agreement
what is required of a satisfactory model-to-seismic with viscous modeling, earlier attempts were less suc-
match. Not all details may be required to match, but cessful. Gravity dominated models predict an affinity
rather key features giving the model a stronger predic- to travel too far north too early (Cavanagh 2013; Zhu
tive power. Haukaas et al. (2013) provided a good et al., 2015), and the lack of time dependence limits
match, but gave little detail of the flow mechanism their power of prediction. Geological assumptions,
they included and adjusted, and thus the published such as the number of feeder channels up to the top
work has unknown predictive power. Zhu et al. (2015) layer, permeability, a correct representation of the top
found that local mismatches cannot be eliminated by seal topography, and temperature distribution will all
adjusting parameters other than local topography, but have a strong effect on the final flow pattern, and
resources should not be directed to such efforts. contain some uncertainty (Zhu et al., 2015; Williams
Chadwick and Noy (2015) discussed quality of matches et al., 2018).
using axisymmetric models, and concluded they had
obtained a robust convergence of predictions and
observations. Nilsen et al. (2017) automatically updated Thermal Structure of the CO2 Plume
the model parameters and found a satisfying match. In the reservoir, most of the injected CO2 will be
They also determined the parameters that had the lar- cooled down to the ambient reservoir temperature.
gest influence on the mismatch. Zhang et al. (2017) However, with time a hotter core of the plume will
reported a satisfactory match using a multiphase simu- develop. Expansion of CO2 from the injection point
lator, by adjusting the lateral permeability anisotropy, up to top reservoir at constant enthalpy (i.e., no heat
CH4 in the CO2 stream, and reservoir temperature. exchange with surroundings) would reduce the tem-
They extrapolated the model to 10,000 years, when perature from 49°C to 36.6°C. Such warm CO2 would
both solubility and mineral trapping become have much lower densities than the average 710 kg/m3
important. for CO2 cooled to reservoir temperature; below 500
It is still under debate whether the upscaled flow kg/m3 at the level of the injection point and possibly
process is dominated by viscous or gravity forces. below 400 kg/m3 at the reservoir top (Figure 13.23).
800 Figure 13.23 Density of

Pure CO2 81bar pure CO2 and mixed with 1.5%
1.5%CH4 81bar methane, at 81 bar (green line,
700 Pure CO2 105bar top reservoir) and 105 bar
1.5%CH4 105bar
(blue line, injection level),
respectively. Black line is CO2
600 1.5%CH4 virgin p,T
with methane at virgin
Density [kg/m3]
reservoir conditions, from top

to bottom reservoir.
500
400
300
200
29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49
228 Temperature [ºC]
1.000
ºC
0.7500 14.00
0.5000 10.50
0.2500 7.000
0.0000 3.500
0.0000
(a) (b)
Figure 13.24 2D axisymmetric reservoir model showing the effect of injecting CO2 at 48°C into a cooler reservoir. (a) CO2 saturation in the
reservoir. (b) Thermal anomaly. From Williams and Chadwick (2017).
A rough estimate of the temperature distribution channel, the propagation speed of the front in that
within the CO2 plume can be obtained by assuming channel did not speed up compared to isothermal
the temperature front is sharp, i.e., that the CO2 modeling.
(and the Utsira reservoir) is either at initial reservoir
temperature, or at the high temperature set by the
injected CO2. With a simple assumption of a cylind- Learnings and Remaining Questions
rical high-temperature region spanning the entire CO2 monitoring at Sleipner has been both a technical
height of the CO2 plume, a constant fraction of 7% and an economic success. Important observations are
of the CO2 will be in the high-temperature state that there are no indications of CO2 in the overbur-
(Alnes et al., 2011). den, the CO2 plume grows and densifies steadily and,
Williams et al. (2018) concluded that thermal based on simple extrapolations, CO2 seems unlikely to
effects generally need to be included in flow modeling spill out of the primary structure during the injection
when the injectant stream is significantly different in phase, which may last another 5–15 years based on
temperature to the reservoir, and the CO2 is close to local CO2 sources. Results from this pioneering CCS
the critical point, which is the case at Sleipner. Zhu project have had influence beyond the technical com-
et al. (2015) found the spreading of layer 9 to be munity, on, e.g., legislators and public acceptance. 4D
sensitive to temperature. However, they did not inves- seismic, supported by time-lapse gravimetry, has pro-
tigate the cooling effect likely to take place as the layer ven that the CO2 stays safely in the storage unit.
spreads. Streamer seismic has proved to be a powerful and
Williams and Chadwick (2017) modeled the cost-efficient tool, both for containment (leakage)
detailed temperature structure within the plume in monitoring and for understanding the dynamic
a 2D radial axisymmetric mesh. The thermal imprint mechanisms within the reservoir. As such, the data
of the rising column extends in their model to and interpretations are crucial for building predictive
a radius of around 350 m from the injection well flow models.
(Figure 13.24). In the thin layers, the temperature No good monitoring alternative to seismic
anomaly becomes negligible at distances more than exists. We cannot say we have the luxury of choos-
about one-third of the full layer spread. They further ing from the toolbox. However, technology and
3D modeled the top layer with hot CO2 feeding into it. parameter choices exist within 4D seismic, such as
As it cooled down before reaching the main northerly frequency of repeats, permanent seafloor receivers, 229
Ola Eiken
or high-resolution surveys. Three-component geo- Meeting, June 10–12, 2015. http://ieaghg.org/docs/Ge

phones would have provided more shear wave informa- neral_Docs/8_Mon/6_Gravity_surveys_over_time_at_
tion and permanency would improve repeatability, but SleipnerSEC.pdf
likely at an order of magnitude higher cost than strea- Alnes, H., Eiken, O., and Stenvold, T. (2008). Monitoring gas
mer surveys of opportunity. The survey frequency production and CO2 injection at the Sleipner field using
could in retrospect possibly have been reduced without time-lapse gravimetry. Geophysics, 73: WA155–W161.
much loss of information – easier to say in hindsight Alnes, H., Eiken, O., Nooner, S., Stenvold, T., and
(Kiær et al., 2015b; Nilsen et al., 2017). Zumberge, M. A. (2011). Results from Sleipner gravity
monitoring: Updated density and temperature
In spite of the excellent seismic signal/noise
distribution of the CO2 plume. Energy Procedia, 4,
ratios, quantification in terms of spatial saturation 5504–5511 (10th International Conference on
distribution is difficult, still subject to research, and Greenhouse Gas Control Technologies).
arguably not resolvable. Time-lapse gravimetry pro-
Arts, R., Eiken, O., Chadwick, R. A., Zweigel, P., van der
vides complementary data, because of their ability to Meer, L., and Zinszner, B. (2004a). Monitoring of CO2
detect mass changes, irrespective of small-scale injected at Sleipner using time-lapse seismic data.
homogeneous or patchy saturation distributions. Energy, 29: 1383–1393.
Combined, geometry constraints from seismic and Arts, R., Eiken, O., Chadwick, A., Zweigel, P., Meer, B. v. d.,
mass changes from gravity are more powerful. and Kirby, G. (2004b). Seismic monitoring at the
Improved joint analysis or inversion may be devel- Sleipner underground CO2 storage site (North Sea).
oped, and flow modeling may be included in the In S. J. Baines and R. H. Worden (eds.), Geological
joint solutions. storage of carbon dioxide. Special Publications 233.
The wealth of data has raised new questions, and London: Geological Society, 181–191.
left others yet unanswered, of which some of the more Arts, R., Chadwick, A., Eiken, O., Thibeau, S., and
prominent are Nooner, S. (2008). Ten years’ experience of monitoring
CO2 injection in the Utsira Sand at Sleipner, offshore
• What are the flow mechanisms when CO2 Norway. First Break, 26(January): 65–72.
accumulates at low saturations instead of Arts, R. J., Trani, M., Chadwick, R. A., Eiken, O.,
preferring high-saturation pathways with high Dortland, S., and van der Meer, L. G. H. (2009).
relative permeability? Acoustic and elastic modeling of seismic time-lapse
• What is the geological nature of the main seismic data from the Sleipner CO2 storage operation.
chimney, and the importance for the vertical flow In M. Grobe, J. C. Pashin, and R. L. Dodge (eds.),
Carbon dioxide sequestration in geological media: State
and growth of the CO2 plume? Is there more than
of the science. AAPG Studies in Geology, 59: 391–403.
one significant chimney?
Baklid, A., Korbøl, R., and Owren, G. (1996). Sleipner Vest
• How well can we construct flow models that match CO2 disposal, CO2 injection into a shallow
the seismic amplitudes and pushdown? How well underground aquifer. SPE Annual Technical Conference
can we predict the next seismic survey? and Exhibition, Denver, CO, SPE paper 36600, 1–9.
• What are the uncertainty bounds on flow Bandilla, K. W., Celia, M. A., and Leister, E. (2014). Impact
predictions, dissolution, and mineralization of model complexity on CO2 plume modeling at
decades and centuries ahead? Sleipner. Energy Procedia, 63: 3405–3415.
Recent progress gives hope that further research Bergman, P., and Chadwick, A. (2015). Volumetric bounds
will improve our understanding of the storage process on subsurface fluid substitution using 4D seismic time
in the years to come. In particular, the joint efforts of shifts with an application at Sleipner, North Sea.
geophysicists and reservoir engineers/flow modelers Geophysics, 80(5): B153–B165.
on models that can satisfy all data and reasonable Bickle, M., Chadwick, A., Huppert, H. E., Hallworth, M.,
physical constraints seem fruitful. and Lyle, S. (2007). Modelling carbon dioxide
accumulation at Sleipner: Implications for
underground carbon storage. Earth and Planetary
References Science Letters, 255(1–2): 164–176.
Alnes, H. (2015). Gravity surveys over time at Sleipner. Bitrus, P. R., Iacopini, D., and Bond, C. E. (2016). Defining
Presentation at the 10th IEAGHG Monitoring Network the 3D geometry of thin shale units in the Sleipner
230
reservoir using seismic attributes. Marine and observed CO2 plume behaviour using Sleipner
Petroleum Geology, 78: 405–425. time-lapse seismics. Greenhouse Gas Science
Boait, F., White N., Chadwick A., Noy D., and Bickle M. Technology, 5: 305–322.
(2011). Layer spreading and dimming within the CO2 Chadwick, R. A., Zweigel, P., Gregersen, U., Kirby, G. A.,
plume at the Sleipner Field in the North Sea. Energy Johannessen, P. N., and Holloway, S. (2004a).
Procedia, 4: 3254–3261. Characterisation of a CO2 storage site: The Utsira
Boait, F. C., White, N. J., Bickle, M. J., Chadwick, R. A., Sand, Sleipner, northern North Sea. Energy, 29:
Neufeld, J. A., and Huppert, H. E. (2012). Spatial and 1371–1381.
temporal evolution of injected CO2 at the Sleipner Chadwick R. A., Arts, R., Eiken, O., Kirby, G. A.,
Field, North Sea. Journal of Geophysical Research, 117: Lindeberg, E., and Zweigel, P. (2004b). 4D seismic
B03309. imaging of an injected CO2 plume at the Sleipner Field,
Borgos, H. G., Randen, T., and Sonneland, L. (2003). Super- central North Sea. InR. J. Cartwright, S. A. Stewart,
resolution mapping of thin gas pockets. Extended M. Lappin, and J. R. Underhill (eds.), 3D seismic
Abstract, Society of Exploration Geophysicists Annual technology: Application to the exploration of sedimentary
Meeting. basins. Geological Society, London, Memoirs. London:
Geological Society, 29: 311–320.
Broto, K., Ricarte, P., Jurado, F., Eitenne, G., and Le Bras, C.
(2011). Improving seismic monitoring by 4D prestack Chadwick, R. A., Arts, R., and Eiken, O. (2005). 4D seismic
traveltime tomography: Application to the Sleipner quantification of a growing CO2 plume at Sleipner,
CO2 storage case. Extended Abstract, EAGE 1st North Sea. In A. G. Doré and B. A. Vining (eds.),
Sustainable Earth Sciences Conference. Petroleum Geology: North-West Europe and Global
Perspectives: Proceedings of the 6th Petroleum Geology
Cavanagh, A. (2013). Benchmark calibration and prediction Conference, 1385–1399.
of the Sleipner CO2 plume from 2006 to 2012. Energy
Procedia, 37: 3529–3545. Chadwick, A., Arts, R., Bernstone, C., May, F., Thibeau, S.,
and Zweigel, P. (2008). Best practice for the storage of
Cavanagh, A. J., and Haszeldine, R. S. (2014). The Sleipner
CO2 in saline aquifers: Observations and guidelines
storage site: Capillary flow modelling of a layered CO2 from the SACS and CO2STORE projects. British
plume requires fractured shale barriers within the
Geological Survey Occasional Publication, 14: 1–267.
Utsira Formation. International Journal of Greenhouse
Gas Control, 21: 101–112. Chadwick, R. A., Noy, D., Arts, R., and Eiken, O. (2009).
Latest time-lapse seismic data from Sleipner yield new
Cavanagh, A., and Nazarian, B. (2014). A new and extended insights into CO2 plume development. Energy Procedia,
Sleipner Benchmark model for CO2 storage simulations 1, 2103–2110.
in the Utsira Formation. Energy Procedia, 63:
2831–2835. Chadwick, R. A., Williams, G., Delepine, N., et al. (2010).
Quantitative analysis of time-lapse seismic monitoring
Cavanagh, A. J., Haszeldine, R. S., and Nazarian, B. (2015). data at the Sleipner CO2 storage operation. Leading
The Sleipner CO2 storage site: Using a basin model to Edge, 29, February: 170–177.
understand reservoir simulations of plume dynamics.
First Break, 33(June): 61–68. Chadwick, R. A., Williams, G. A., Williams, J. D. O., and
Noy, D. J. (2012). Measuring pressure performance of
Chadwick, R. A., and Eiken, O. (2013). Offshore CO2
a large saline aquifer during industrial-scale CO2
storage: Sleipner natural gas field beneath the North
injection: The Utsira Sand, Norwegian North Sea.
Sea. In J. Gluyas and S. Mathias (eds.), Geological storage
International Journal of Greenhouse Gas Control, 10:
of carbon dioxide (CO2): Geoscience, technologies,
374–388.
environmental aspects and legal frameworks. Sawston,
UK: Woodhead Publishing, 227–250. Chadwick, R. A., Marchant, B. P., and Williams, G. A.
Chadwick, R. A., and Noy, D. J. (2010). History – matching (2014). CO2 storage monitoring: Leakage detection and
flow simulations and time-lapse seismic data from the measurement in subsurface volumes from 3D seismic
Sleipner CO2 plume. In B. A. Vining and S. C. Pickering data at Sleipner. Energy Procedia, 63: 4224–4239.
(eds.), Petroleum geology: From mature basins to new Chadwick, R. A., Williams, G. A., and White, J. C. (2016).
frontiers. Proceedings of the 7th Petroleum Geology High-resolution imaging and characterization of a CO2
Conference. Petroleum Geology Conferences Ltd. layer at the Sleipner CO2 storage operation, North Sea
London: Geological Society, 1171–1182. using time-lapse seismics. First Break, 34(February):
Chadwick, R. A., and Noy, D. J. (2015). Underground CO2 77–85.
storage: Demonstrating regulatory conformance by Chadwick, R. A., Williams, G. A., and Noy, D. J. (2017). CO2
convergence of history-matched modelled and storage: Setting a simple bound on potential leakage 231
Ola Eiken
through the overburden in the North Sea Basin. Energy Furre, A.-K., Eiken, O., Alnes, H., Vevatne, J. N., and
Procedia, 114: 4411–4423. Kiær, A. F. (2017). 20 years of monitoring
Clochard, V., Delépine, N., Labat, K., and Ricarte, P. (2010). CO2-injection at Sleipner. Energy Procedia, 114:
CO2 plume imaging using pre-stack stratigraphic 3916–3926.
inversion: A case study on the Sleipner field. First Break, Ghaderi, A., and Landrø, M. (2009). Estimation of thickness
28(1): 91–96. and velocity changes of injected carbon dioxide layers
Delepine, K., Clochard, N., Labat, V., and Ricarte, P. (2011). from prestack time-lapse seismic data. Geophysics, 74
Post-stack stratigraphic inversion workflow applied to (2): O17–O28.
carbon dioxide storage: Application to the saline Ghosh, R., Sen, M. K., and Vedanti, N. (2015). Quantitative
aquifer of Sleipner field. Geophysical Prospecting, 59(1): interpretation of CO2 plume from Sleipner (North Sea),
132–144. using post-stack inversion and rock physics modelling.
Dubos-Sallée, N., and Rasolofosaon, P. N. (2011). International Journal of Greenhouse Gas Control, 32:
Estimation of permeability anisotropy using seismic 147–158.
inversion for the CO2 geological storage site of Sleipner Gregersen, U., ed. (1998). Saline aquifer CO2 storage phase
(North Sea). Geophysics, 76(3): WA63–WA69. zero geological survey of Denmark and Greenland.
Dupuy, B., Torres, V. A. C., Ghaderi, A., Querendez, E., and GEUS.
Mezyk, M. (2017a). Constrained AVO for CO2 storage Gregersen, U., Michelsen, O., and Sorensen, J. C. (1997).
monitoring at Sleipner. Energy Procedia, 114: Stratigraphy and facies distribution of the Utsira
3927–3936. Formation and the Pliocene sequences in the northern
Dupuy, B., Romdhane, A., Eliasson, P., Querendez, E., North Sea. Marine and Petroleum Geology, 14: 893–914.
Yan, H., Torres, B., and Ghaderi, A. (2017b). Haffinger, P., Eyvazi, F. J., Doulgeris, P., Steeghs, P.,
Quantitative seismic characterization of CO2 at the Gisolf, D., and Verschuur, E. (2017). Quantitative
Sleipner storage site, North Sea. Interpretation, 5(4): prediction of injected CO2 at Sleipner using
SS23–SS42. wave-equation based AVO. First Break, 35: 65–70.
Eiken, O., Ringrose, P., Hermanrud, C., Nazarian, B., Hansen, H., Eiken, O., and Aasum, T. O. (2005). Tracing
Torp, T. A., and Høier, L. (2011). Lessons learned from the path of carbon dioxide from a gas-condensate
14 years of CCS operations: Sleipner, In Salah and reservoir, through an amine plant and back into
Snøhvit. Energy Procedia, 4: 5541–5548. a subsurface acquifer. Case study: The Sleipner area,
Eliasson, P., and Romdhane, A. (2017). Uncertainty Norwegian North Sea. Aberdeen, UK: SPE Offshore
quantification in waveform-based imaging methods: Europe.
A Sleipner CO2 monitoring study. Energy Procedia, Harrington, J. F., Noy. D. J., Horseman, S. T., Birchall, D. J.,
114: 3905–3915. and Chadwick, R. A. (2010). Laboratory study of gas
Evensen, A. K., and Landrø, M. (2010). Time-lapse and water flow in the Nordland Shale, Sleipner, North
tomographic inversion using a Gaussian Sea. In M. Grobe, J. Pashin and R. Dodge (eds.), Carbon
parameterization of the velocity changes. Geophysics, dioxide sequestration in geological media: State of the
75(4): U29–U38. science. AAPG Studies in Geology, 59: 521–543.
Falcon-Suarez, I., Papageorgiou, G., Chadwick, A., Hauge, V. L., and Kobjørnsen, O. (2015). Bayesian inversion
North, L., Best, A. I., and Chapman, M. (2018). CO2- of gravimetric data and assessment of CO2 dissolution
brine flow-through on an Utsira Sand core sample: in the Utsira Formation. Interpretation, 3(2): sp1–sp10.
Experimental and modelling. Implications for the Haukaas, J., Nickel, M., and Sonneland, L. (2013). Successful
Sleipner storage field. International Journal of 4D history matching of the Sleipner CO2 plume.
Greenhouse Gas Control, 68: 236–246. Extended Abstract, EAGE Conference.
Furre, A.-K., and Eiken, O. (2014). Dual sensor streamer Hermanrud, C., Andersen, T., Eiken, O., et al. (2009).
technology used in Sleipner CO2 injection monitoring. Storage of CO2 in saline aquifers: Lessons learned from
Geophysical Prospecting, 62(5): 1075–1088. 10 years of injection into the Utsira Formation in the
Furre, A. K., Ringrose, P., Cavanagh, A., Janbu, A. D., and Sleipner area. Energy Procedia, 1: 1997–2004.
Hagen, S. (2014). Characterization of a submarine Hofmann, R. (2006). Frequency dependent elastic and
glacial channel and related linear features. Extended inelastic properties of clastic rocks. PhD thesis, Colorado
Abstract, EAGE Near Surface Geoscience. School of Mines.
Furre, A.-K., Kiær, A., and Eiken, O. (2015). CO2-induced Jullum, M., and Kolbjørnsen, O. (2016). A Gaussian-based
seismic time shifts at Sleipner. Interpretation, 3(3): framework for local Bayesian inversion of geophysical
232 SS23–SS35. data to rock properties. Geophysics, 81(3): R75–R87.
Kiær, A. (2011). Trykkutvikling under CO2-lagring. Master’s Pedersen, R. B. (2011). Annual Report 2011. Center for
thesis, NTNU (in Norwegian). Geobiology, University of Bergen.
Kiær, A. F. (2015). Fitting top seal topography and CO2 Queißer, M., and Singh, S. C. (2013a). Full waveform
layer thickness to time-lapse seismic amplitude maps at inversion in the time lapse mode applied to CO2 storage
Sleipner. Interpretation, 3(2): SM47–SM55. at Sleipner. Geophysical Prospecting, 61(3): 537–555.
Kiær, A. F., Eiken, O., and Landrø, M. (2015a). Time lapse Queisser, M., and Singh, S. C., (2013b). Localizing CO2 at
seismic amplitudes close to the rim of Sleipner CO2 Sleipner:Seismic images versus P-wave velocities from
layers. In A. F. Kiær, CO2 fluid flow information from waveform inversion. Geophysics, 78(3): B131–B146.
quantitative time-lapse seismic analysis. PhD thesis, Rabben, T. E., and Ursin, B. (2011). AVA inversion of the
Norwegian University of Science and Technology, 57–72. top Utsira Sand reflection at the Sleipner field.
Kiær, A., Eiken, O., and Landrø, M. (2015b). Calendar time Geophysics, 76(3): C53–C63.
interpolation of amplitude maps from 4D seismic data. Raknes, E. B., Weibull, W., and Arntsen, B. (2015).
Geophysical Prospecting, 64(2): 421–430. Three-dimensional elastic full waveform inversion
Labat, K., Delépine, N., Clochard V., and Ricarte, P. (2012). using seismic data from the Sleipner area. Geophysical
4D joint stratigraphic inversion of prestack seismic Journal International, 202(3): 1877–1894.
data: Application to the CO2 storage reservoir (Utsira Romdhane, A., and Querendez, E. (2014). CO2
sand formation) at Sleipner site. Oil & Gas Science and characterization at the Sleipner field with full waveform
Technology, 67(2): 329–340. inversion: Application to synthetic and real data.
Landrø, M., and Zumberge, M. (2017). Estimating Energy Procedia, 63: 4358–4365.
saturation and density changes caused by CO2 injection Rossi, G., Chadwick, R. A., and Williams, G. A. (2011).
at Sleipner: Using time-lapse seismic amplitude- Traveltime and attenuation tomography of CO2 plume
variation-with-offset and time-lapse gravity. at Sleipner. Extended Abstract, 73rd EAGE Conference
Interpretation, T243–T257. & Exhibition.
Lindeberg, E. (2010). Modelling pressure and temperature Rubino, J. G., and Velis, D. R. (2011). Seismic
profile in a CO2 injection well. Energy Procedia, 4, characterization of thin beds containing patchy carbon
3935–3941. dioxide-brine distributions: A study based on
Lindeberg, E., Zweigel, P., Bergmo, P., Ghaderi A., and numerical simulations. Geophysics, 76: R57–R67.
Lothe A. (2000). Prediction of CO2 distribution pattern Rubino, J. G., Velis, D. R., and Sacchi, M. D. (2011).
improved by geology and reservoir simulation and Numerical analysis of wave-induced fluid flow effects
verified by time lapse seismic. Expanded Abstract, 5th on seismic data: Application to monitoring of CO2 at
GHGT Conference. the Sleipner field. Journal of Geophysical Research, 116:
Monastersky, R. (2013). Seabed scars raise questions over B03 306.
carbon-storage plan. Nature, 504 (December 19/26): Singh, V., Cavanagh, A., Hansen, H., Nazarian, B.,
339–340. Iding, M., and Ringrose, P. (2010). Reservoir modeling
Neele, F. P., and Arts, R. J. (2010). Time-lapse seismic AVP of CO2 plume behavior calibrated against monitoring
analysis on the Sleipner CO2 storage monitoring data data from Sleipner, Norway. SPE, 134891: 1–18.
using CFP processing. Extended Abstract, 72nd EAGE White, J. C., Williams, G. A., and Chadwick, R. A. (2013).
Conference. Thin layer detectability in a growing CO2 plume:
Nilsen, H. M., Krogstad, S., Andersen, O., Allen, R., and Testing the limits of time-lapse seismic resolution.
Lie, K.-A. (2017). Using sensitivities and vertical- Energy Procedia, 37: 4356–4365.
equilibrium models for parameter estimation of CO2 Williams, G., and Chadwick, A. (2012). Quantitative seismic
injection models with application to Sleipner data. analysis of a thin layer of CO2 in the Sleipner injection
Energy Procedia, 114: 3476–3495. plume. Geophysics, 77(6): R245–R256.
Nooner, S. L., Eiken, O., Hermanrud, C., Sasagawa, G. S., Williams, G., and Chadwick, R. A. (2017). An improved
Stenvold, T., and Zumberge, M. A. (2007). Constraints history-match for layer spreading within the Sleipner
on the in situ density of CO2 within the Utsira plume including thermal propagation effects. Energy
formation from time-lapse seafloor gravity Procedia, 114: 2856–2870.
measurements. International Journal of Greenhouse Gas Williams, G. A., Chadwick, R. A., and Vosper, H. (2018).
Control, 1: 198–214. Some thoughts on Darcy-type flow simulation for
Park, J., Sauvin, G., and Vöge, M. (2017). 2.5D inversion and modelling underground CO2 storage based on the
joint interpretation of CSEM data at Sleipner CO2 Sleipner CO2 storage operation. International Journal of 233
storage. Energy Procedia, 114: 3989–3996. Greenhouse Gas Control, 68: 164–175.
Ola Eiken
Zhang, G., Lu, P., and Zhu, C. (2014). Model predictions via Zweigel, P., Hamborg, M., Arts, R., Lothe, A. E., Sylta, O.,
history matching of CO2 plume migration at the and Tommeras, A. (2001). Prediction of migration of
Sleipner Project, Norwegian North Sea. Energy CO2 injected into an underground depository:
Procedia, 63: 3000–3011. Reservoir geology and migration modelling in the
Zhang, G., Lu, P., Ji, X., and Zhu, C. (2017). CO2 Sleipner case (North Sea). In D. Williams, I. Durie,
plume migration and fate at Sleipner, Norway: P. McMullan, C. Paulson, and A. Smith (eds.),
Calibration of numerical models, uncertainty Greenhouse Gas Control Technologies, Proceedings of the
analysis, and reactive transport modelling of CO2 5th International Conference on Greenhouse Gas Control
trapping to 10,000 years. Energy Procedia, 114: Technologies, 360–365.
2880–2895. Zweigel, P., Arts, R., Lothe, A. E., and Lindeberg, E. (2004).
Zhu, C., Zhang, G., Lu, P., Meng, L., and Ji, X. (2015). Reservoir geology of the Utsira Formation at the first
Benchmark modeling of the Sleipner CO2 plume: industrial-scale underground CO2 storage site (Sleipner
Calibration to seismic data for the uppermost layer and area, North Sea). In S. Baines, J. Gale, and R. Worden
model sensitivity analysis. International Journal of (eds.), Geological storage of carbon dioxide for emissions
Greenhouse Gas Control, 43, 233–246. reduction. London: Geological Society, 165–180.
234
Chapter
Case Studies of the Value of 4D,
14 Multicomponent Seismic Monitoring

in CO2 Enhanced Oil Recovery and
Geosequestration
Thomas L. Davis, Scott Wehner, and Trevor Richards
Multicomponent seismic monitoring of carbon diox- diverting the CO2 beyond some pattern boundaries.
ide (CO2) enhanced oil recovery (EOR) projects were The result has been the potential for the bypass of oil
undertaken by the Reservoir Characterization Project reserves, thus requiring additional injectors and pro-
(RCP) at Postle Field, Oklahoma and Delhi Field, ducers to be drilled to improve sweep efficiency.
Louisiana. Both fields are being rejuvenated through To increase the efficiency of a CO2 injection pro-
CO2 flooding. At the time of these studies Scott ject it is essential to monitor where the CO2 is going.
Wehner managed the Postle CO2 EOR project for Owing to the high mobility of CO2 relative to water
Whiting Oil and Gas Company, Trevor Richards and oil, it moves readily into the high-permeability
managed the geophysical reservoir characterization zones and bypasses the low. Bypassing can occur
at Delhi Field for Denbury Resources, and Thomas vertically as well as laterally. Optimizing contact
L. Davis was a codirector of RCP at the Colorado with residual oil that has not been mobilized by pri-
School of Mines. mary production and secondary waterflooding is cri-
At Postle Field miscible CO2 is being injected into tical to achieving economic success. CO2 flooding is
a water-alternating gas (WAG) scheme in a sandstone a viable mechanism for EOR within most conven-
reservoir with an average thickness of 28 ft., at 6100 ft. tional reservoirs while providing an excellent
depth. The flood patterns are a single vertical injection mechanism for associated CO2 geosequestration.
well in the middle of four producing wells; known as
an inverted five-spot pattern. The RCP study area Postle Field
focused on 18 patterns in the northwestern area of Located in the Panhandle of Oklahoma, Postle Field
Postle Field. Tracking the CO2 with time-lapse seis- has approx. 300 million barrels of original oil in place
mic data is challenging owing to the thinness of the (Figure 14.1). The field produces predominantly from
reservoir, the petrophysics of the reservoir and con- the Morrow “A” sandstone of Pennsylvanian age.
fining strata, the depth of investigation, and the oper- A stratigraphic column is shown in Figure 14.2 and
ating practices within the flood. The results are the reservoir properties are listed in Table 14.1.
encouraging in terms of the utilization of time-lapse The portion of the field monitored by RCP was
data to characterize and monitor thin reservoirs with approx. 6 square miles in the Hovey Morrow Unit
very low acoustic impedance contrast. (HMU) (Figure 14.3). Water injection was conducted
At Delhi Field, the flood is an immiscible process initially to build the reservoir pressure to near-
involving the continuous injection of CO2. Reservoir miscible pressures before the injection of CO2.
depth is approx. 3300 ft., with thickness varying
from 100 ft. to an updip wedge (i.e., pinch out).
The reservoir was originally thought to be a blanket Static Reservoir Characterization
sand and, as a result, the flood pattern chosen was Descriptions of cores from four wells in the HMU
originally a line-drive with injection wells updip and study area indicate the presence of four main lithofa-
downdip with producing wells in the middle. Seismic cies. These include a conglomeratic basal lag,
monitoring has identified distributary channels a medium to fine-grain active channel fill, abandoned 235
Updip Valley-Fil Production Figure 14.1 Location map, Postle Field,

Oklahoma.
Oil Fields
Stateline
Sorrento/Mt.Pearl Morrow Trend Gas Fields
Morrow Fields
Postle Field Hugoton

Downap Valley-Fill
Production
Colorado Kansas
Lowstand Shoreline Production
Oklahoma
10 miles
Texas
Postle Field Morrow Type Log

Gamma Ray Depth Resistivity
Generalized Stratigraphic Column
Ochoan
Permian
6100
Upper Morrow
Gaudalupian El Reno
A Sand
Leaonardian Summer /Enid
A1 Sand
Chase
Hugoton Wolfcampian Council Grove
6200
Admire
Wabaunsee A2 Sand
Virgilian Shawnee
Lansing
Missourian Kansas City
Pennsylvanian
Marmaton 6300
Des Moines Cherokee
Atoka 13 Fingers
A Sand B Sand
Upper
A1 Sand
Morrow A2 Sand 6400
B Sand F Sand
Lower Morrow
Lower F Sand
Morrow G Sand
Keys
Chester Chester 6500
issippian
Miss-
Meramac
G Sand
Osage
6600 Keys Sand
Modified from Whiting Petroleum
236
Figure 14.2 Generalized stratigraphic column, Postle Field.
4D Multicomponent Seismic Monitoring
Table 14.1 Reservoir properties, HMU, Postle Field
Postle HMU reservoir properties

• Depth 6100 ft.
• Producing formation Upper Morrow A – Sandstone
• Average thickness 28 ft. (10–70)
• Porosity 17% (10–23)
• Permeability 50 md (20–500)
• Soil 70%
• Reservoir temperature 140°F
• Reservoir pressure initial 1630 psi
• Reservoir pressure current 2300 psi
• Minimum miscibility pressure 2100 psi
• Oil gravity 40 API
• Oil FVF orig 1.28
• Oil FVF current 1.20
• Well spacing 40 Ac (80-Ac 5-spot inj. patterns)
• OOIP Approx. 45 MMBO
FVF, formation volume factor; OOIP, original oil-in-place.
Figure 14.3 Location of the 6.25 square mile, 9-C, time-lapse multicomponent survey in the Hovey Morrow Unit (HMU), Postle Field. Key wells 237
include VSP (circles) and location of cored wells and/or wells with sonic scanner and image logs (yellow diamonds).
channel fill, and overbank deposits. These facies occur

(a)
in a depositional environment associated with braided
100,000
channel deposits filling an incised valley. Provenance
for the sediments is considered to be in relatively close 10,000
proximity due to poor sediment sorting and the angu- 1,000
larity and texture of rock fragments.
Rate
100
Secondary porosity has formed in this reservoir by
dissolution of potassium feldspars. Permeability is 10 Oil, STB/Day
highest in the basal lag and active channel fill facies.

Water, STB/Day
1 Gas, MSCF/Day
Within these facies permeability can vary from 20 to CO2, MSCF/Day
500 mD. As a result, thin braided channels of high 0
00
05
60
65
70
10
80
85
90
75
95
permeability can form conduits for CO2 movement in
20
20
19
19
19
20
19
19
19
19
19
the reservoir. Alignment of wells along the channels, Date
although providing some early oil production, also (b)
leads to early breakthrough of the more mobile CO2. Hovey Morrow Unit
Water alternating gas (WAG) is used to slow the 100,000
displacement in the higher permeability zones and
help provide better vertical conformance; pushing 10,000 2588 BOPD
more CO2 to the less permeable zones. Pattern align-
ment and pattern balancing are important to optimize 1,000
recovery efficiency in the WAG process. The ability to
proactively manage the WAG schemes has an impor-
100 Oil Prod., BOPD
tant bearing on incremental oil recovery and conver- H2O Prod., BWPD
sely, CO2 utilization, as well as its ultimate CO2 Prod., Mscf/D
sequestration. In addition, the ability to dynamically 10
H2O Inj., BWPD
monitor the CO2 WAG not only provides a key to CO2 Inj., Mscf/D
better reservoir management of a costly injectant, but 1

Jan-05 Jan-06 Jan-07 Jan-08 Jan-09 Jan-10 Jan-11 Jan-12
it also allows monitoring for possible losses beyond
the flood pattern or project complex. Figure 14.4 Field revitalization with miscible carbon dioxide
How significant is this project? The HMU was the flooding in the Hovey Morrow Unit, Postle Field.
last of four CO2 EOR expansions within Postle Field;
the first starting in 1996. Prior to expanding into
HMU, elevated reservoir pressures within the offset important impact on the economic success of EOR
acreage (active CO2 WAG operations) were main- projects. Early seismic monitoring should be an inte-
tained by water-curtain well management (i.e., pres- gral part of project design and reservoir management
sure) between the project boundaries. HMU was planning for both operational, as well as geoseques-
a very mature waterflood and had only one remaining tration purposes.
active/producing well prior to beginning CO2 flood
development. Injection began in 2008 and the flood Dynamic Reservoir Characterization
exceeded the waterflood peak production achieved in The baseline seismic program was acquired
1971 within the first year of operation (Figure 14.4). in March 2008 and the first monitoring survey
The HMU CO2 flood is a successful rejuvenation 9 months later in December 2008. Survey parameters
project due in large part to understanding the con- are included in Table 14.2. Several wells were drilled
nectivity of the stacked braided channels. Seismic during 2008 and little drilling went on in 2009 owing
monitoring has played an important role in identify- to an oil price crash carrying over from the latter part
ing the connectivity within the reservoir, which is of 2008. Overall, the surface coupling conditions
a critical part of the EOR reservoir management between the December and March surveys were com-
operation. Establishing the field development plan in parable. Repeatability, as measured by normalized
238 a proactive rather than reactive manner can have an root mean square (NRMS, the square root of the
Table 14.2 Acquisition parameters used for the seismic surveys the vicinity of the northernmost wells, and negligibly
at Postle Field
different elsewhere. An increase in fluid pressure
4-D,9-C Seismic acquisition parameters decreases the acoustic impedance. In addition,
a change in fluid saturation occurs as CO2 mixes
Type survey 4-D, 9-C (time-lapse)
with oil to form a miscible flood bank while predomi-
Subsurface bin size 55 ft. × 55 ft. nantly replacing water in the pore space: a resultant of
Number of receiver 1920 the previous waterflood operation. Laboratory rock
locations physics measurements on cores shows the predicted
Number of source 1920 seismic changes (Figure 14.6).
locations It is critical to understand the fluids in the reser-
Total number of 5760 voir at any given time when planning seismic surveys.
source points Prior to the injection of CO2, water is typically
Type receiver Stationary: 13,200 ft. × 13,200 ft. injected to bring pressure up to miscibility. In most
spread cases, it is far more economical to inject water rather
Instrumentation I/O System IV, 2 ms sample rate, than CO2 to build reservoir pressure. In the case of
45 s record length Postle, with a limited CO2 supply available, it is also
critical to maximize the efficiency of the injection
Receiver array Single 3C digital sensor
volumes. A WAG process to improve conformance
(vectorSeis) within the strata and maintain pressures once the
Source (P-wave) Vertical vibrator: 6–100 Hz linear pressure has reached miscibility, coupled with balan-
sweep, 8 s duration, 4 sweeps cing of the injection and withdrawals within the flood
Source (S-wave) Horizontal vibrator: 4–60 Hz linear patterns, provides a means of managing the efficiency
sweep, 8 s duration, 4 sweeps, of CO2 EOR and associated geosequestration. Two
downdip water injectors (a.k.a. water curtain) on the
one source oriented N–S,
southwest side of the study area were used to provide
one source oriented E–W pressure support and prevent the CO2 from moving
downdip outside the Unit and to maintain the pres-
sure front eastward and northward within the
average of the squares of a series of measurements) reservoir.
after cross-equalization, is approx. 0.2, indicating Overall, the seismic data acquired were very good.
excellent repeatability. The surface is relatively flat agricultural land and
The thickness of the Morrow “A” Sandstone is well source and receiver coupling was good. There was
below the resolution limit of the seismic data, but not much variability in near surface conditions
detectability is dependent on the impedance of the between the two surveys. A vertical seismic profile
sandstone and not just thickness. Even though the (VSP) was acquired in a well in the northern part
Morrow “A” is below seismic resolution, injection of of the survey area. It showed that the Morrow
CO2 can potentially make the reservoir sandstones “A” sandstone is a very weak peak on P-wave seismic,
visible because fluid saturation change and pressure but shows a much stronger peak on the S-wave data
affect the impedance of the sandstones. The term (Figure 14.7). This reflectivity difference is also shown
“time-lapse sands” has been coined for those sand- on the surface seismic data (Figures 14.8–14.11).
stones that become visible by introducing dynamic Scheduling the timing of surveys is critical. A true
changes in the reservoir. baseline was not achieved as injection began in the
The volume of fluids injected in wells between the southern portion of the survey area shortly before the
two surveys is shown in Figure 14.5. In total 5.10 Bscf first survey was acquired. Nevertheless, the hypothesis
of CO2 was injected from the baseline to acquisition of about CO2 increasing the visibility of the sandstones
the first seismic monitor survey. Prior to the baseline was confirmed (Figure 14.9). The high-amplitude area
acquisition, approx. 1.65 Bscf had been injected in the in the south showed good continuity overall and
southernmost wells. The pressure change in the reser- matched the well information regarding presence of
voir between the two surveys was approx. 800 psi in CO2 indicating the potential promise that the flood 239
Fluid Injected between surveys

Water Injection CO2 Injection
(March 2008-December 2008) (March 2008-December 2008)
HMU 10-2 HMU 11-2 HMU 11-4 Water CO2 HMU 10-2 HMU 11-2 HMU 11-4
PUMU 10-3A (Mbbl) (Mcf) HMU 10-3A
HMU 11-1 HMU 12-3 HMU 12-1

300 700000 HMU 11-1 HMU 12-3 HMU 12-1
HMU 11-3 HMU 11-3
HMU 13-2 HMU 13-2
HMU 13-3 HMU 14-1 HMU 15-1 HMU 13-3 HMU HMU 15-1
HMU 14-5 14-1 HMU 14-5
HMU 14-4 HMU 14-4
HMU 13-1 HMU 14-2 HMU 13-1 HMU 14-2
HMU 17-1 HMU

HMU 17-1
HMU 17-5 HMU 17-8 17-1 HMU 18-1 HMU 17-5 HMU 17-8 HMU 17-7 HMU 18-1
HMU 18-4 HMU 18-4
HMU 18-2 HMU 18-2
HMU 17-6
HMU 17-2 HMU 17-4 HMU 17-3 HMU 18-3 150 350000 HMU 17-6
HMU 17-2 HMU 17-4HMU 17-3 HMU 18-3
HMU HMU
HMU 19-1 HMU 21-1 HMU 21-3 HMU 22-1 22-3 HMU 24-1 HMU 24-3 HMU 19-1 HMU 21-1 HMU 21-3 HMU 22-1 HMU 24-1 HMU 24-3
22-3
HMU HMU HMU

HMU 19-2 HMU 21-4 HMU 22-4 HMU 19-2 HMU 21-4 HMU 21-2
21-2 HMU 22-5 HMU 24-4 HMU HMU 24-4 HMU 24-2
22-2 24-2 22-2
HMU HMU HMU

HMU
HMU 27-1 HMU 30-B1 HMU 30-2 HMU 32-1 32-3 35-1 35-7 HMU HMU 27-1 HMU 30-B1 HMU 30-2 HMU 32-1 HMU 35-7
35-3 32-3 HMU 35-1 HMU 35-3
HMU 35-5 HMU 35-5
HMU 31-1C HMU 32-4 HMU 31-1C HMU 35-4

HMU 35-4 HMU 35-2 HMU 32-4 HMU 35-2
HMU 31-2 HMU 36-2 HMU 36-2
HMU 36 HMU 36-2
0 0
2.5 mile 2.5 mile
Figure 14.5 Injected fluid volumes for wells in study area from March through December, 2008.
High Permeability Zone

P-wave Velocity S-wave Velocity
4600 2800
Max Pore Avg Pore Min Pore Max Pore Avg Pore Min Pore
4400 Pressure Pressure Pressure 2600 Pressure Pressure Pressure
S- wave Velocity (m/s)
P-wave Velocity (m/s)
4200 2400
4000 2200
3800 2000
3600 1800
2000 3000 4000 5000 6000 2000 3000 4000 5000 6000
Differential Pressure (psi) Differential Pressure (psi)
Brine Oil Oil-CO2 CO2
Figure 14.6 Laboratory measurements on core under different pressure and fluid saturation conditions.
240
Zero-Offset P-Wave VSP Data Zero-Offset S-Wave VSP Data
Corridor Corridor
(a) 1
CHAN #
60
(b) 1
CHAN #
60 Stack
Stack
2300
0.4
1120 0.3
0.3
Before Deconvolution
0.2 2340
1140 0.2
0.1 0.1
Amplitude
Amplitude
Time (ms)
Time (ms)
1160 2380
0.0 0.0
1180 -0.1 2420 -0.1
-0.2
1200 -0.2
2460 -0.3
-0.3
1220 -0.4
CHAN # Corridor Corridor

(c) 1 60 Stack (d) 1
CHAN #
60 Stack
2300
0.008
1120 0.006
0.006
2340 0.004
After Deconvolution
1140 0.004
0.002
Amplitude
Amplitude
0.002
Time (ms)
Time (ms)
1160 2380
0.000 0.000
1180 -0.002
2420 -0.002
-0.004
1200 -0.004
-0.006 2460
-0.008 -0.006
1220
Figure 14.7 The near offset vertical seismic profile shows the reflectivity of the Morrow “A” sandstone (green arrow) on P- and S-wave data.
The Atoka limestone is shown by a red arrow and is a good seismic marker on both P- and S-wave data.
241
Figure 14.8 A P-wave line from the 3D volume shows the weak reflectivity of the Morrow “A” sandstone on P-wave data.
Figure 14.9 RMS amplitude map from a 10-ms window

RMS Amplitude (P-wave) centered on the Morrow “A” sandstone event on the P-wave data.
Color
Xline Key The southern part of the survey area shows an increased seismic
120
amplitude due to the presence of CO2. By the time of this baseline
6.84
6.26 survey 1.65 Bscf had already been injected in the southernmost
110 5.90 wells. The green dots are oil wells and the red and blue are CO2
5.50 and H2O injection wells, respectively. Well patterns are depicted.
100 5.10
4.74
90 4.45
4.17
80
3.90
3.65
70
3.36
60 3.07
2.81
50 2.59
2.45
40 2.30
2.16
30 2.01
1.83
20 1.54
1.32
10 1.07
0.81
10 20 30 40 50 60 70 80 90 100 110 120 0.52
0.05
Inline
242 Figure 14.10 A 2D line out of the 3D volume showing the Morrow “A” sandstone as a strong seismic peak on the shear wave data. The term
Sh-wave pertains to a shear wave with horizontal particle motion in a plane perpendicular to the source-receiver plane in an isotropic medium.
Figure 14.11 RMS amplitude map of the Morrow

RMS Amplitude (Sh-wave) “A” sandstone over a 20-ms window centered on
Color
Xline
Key the Morrow “A” seismic event.
120
2.50
110 2.31
2.18
100 2.05
1.91
90 1.78
1.67
80 1.59
1.49
70 1.39
1.30
60 1.25
1.20
50 1.15
1.10
40 1.05
1.00
30 0.95
0.88
20 0.78
0.69
10 0.61
0.52
0.32
10 20 30 40 50 60 70 80 90 100 110 120
0.13
Inline
would be successful. By the time of the March survey, area that found 30 ft. of Morrow “A” sandstone very
herein referred to as the baseline, the northern area near the edge of the valley and the boundary of the
had not had CO2 introduced. Unit, thus completing the pattern in that area.
The weak reflectivity of the P-wave data is due to Figures 14.13 and 14.14 show the time-lapse
the compressibility of the sandstones. Compressibility differences for P- and S-waves respectively. The areal
links to porosity, texture, and to the fluids within the extent of these changes aids in mapping permeability
pore space. In the case of the Morrow “A” sandstone, trends in the reservoir and these trends can be then
the high compressibility of the sandstones lowers the entered into the reservoir model and used to improve
seismic velocity and density such that the acoustic history matching in reservoir simulation.
impedance of the sandstones nearly equals that of P-waves are affected by fluid saturation and pres-
overlying Morrow shale rendering the sandstones as sure. A small amount of CO2 drops the P-wave velo-
acoustically invisible to P-wave data. city as shown previously in Figure 14.6. After about
Shear wave data are affected by rigidity and, 10% saturation there is little change in the P-wave
because the sandstones are more rigid than the over- response other than pressure. As this is a WAG injec-
lying shale, they are visible on S-wave data. Fluid tion scheme there is very little response to the flood on
pressures can affect the rigidity and density of the the P-wave data, but more on the S-wave data, as
sandstones and thus fluid pressure and not fluid com- they are only sensitive to pressure. The pressure
position can affect shear waves. front is more diagnostic of where the fluids are mov-
Figure 14.12 shows a seismic-guided isopach map ing than where the CO2 miscible bank is within the
made from well and seismic control. Notice that as the reservoir. By recording both P- and S-waves we can
sandstones thin to the edges of the incised valley, the differentiate between pressure and saturation when
seismic resolution is diminished and, therefore, compared.
weighting functions applied in the geostatistical map- The value of seismic monitoring at Postle Field
ping are diminished. The Unit boundaries also are involves proactive management of the CO2 flood
a constraining factor when it comes to operational through WAG optimization, pattern orientation and
controls in the field. Constraining the CO2 from mov- pattern balancing. This is managed in spite of the
ing out of pattern and outside the boundaries of the absence of a true baseline survey. All of the seismic
Unit is necessary. At Postle Field our study led to the data were acquired, processed, and interpreted for less 243
drilling of a well in the southeast corner of our study than the cost of a single well.
HMU6-14 PUMU8-10
9-3 76
9-5 9-2 6-20
6-14 6-13
HMU9-3
6-21 6-15 9 8-8 38
HMU6-21
28 17
7 PUMU7-5
17
8
13 5 38
10-3 10-4 12-4 RCP Phase XII 3D Survey
HMU10-4
4-M 12-5
10-2 11-2 11-4 HMU12-5
12-6 PUMU9-5
HMU11-4 0 9-5
10-20 3 10-3A
30 44 11
45
12-30 3
1-F 11-1 12-3 12-1 2-M-A 9-12ST PUMU9-3
11-3 HMU11-3
HMU12-1
7 34 19 9-12
20 22 0
P. U. M. U
8 12 13-2 9 10 0 11
53 HMU14-1
14-1 5 15-1 16-2
13-3
46 14-5 2 14 18
39
14-4 13
13-1 14-2
46
HMU13-1
36 HMU14-2
10 Whiting 3D 30
4-A 17-5 17-1 18-1

HMU17-8
17-8 17-7 1-15

5 46 66 18-4
49 35 24 9
18-2
HMU18-1
19
17-6 17-2 17-4 17-3 16
18-3
HMU17-2
HMU17-3
0 75 17 41 21
17 16#1 16 15 14
1-B 19-1 21-1 21-3 22-1 22-3 24-1 24-3 1
HMU24-1
HMU22-1
HMU21-1
16 0 23 54 38 47 14
63 5
19-2 24-5 ST-1

21-4 21-2 22-4 22-2 24-4 24-2
HMU21-2 22-5 HMU24-2
HMU22-2
2424-5
29 27 62 45 67 57 52
H. M. U.
Postle Field - RCP Interp 53
HMU30-B-1
7
Gross 'A' Sand 30-B-1 35-7
30-2 32-1 32-3 35-1 35-6
CONTOURS HMU32-1
HMU35-5 35-3 1
Rafael_CSM_MorrowA_Gross_Thickness_from_PandS_Amp.txt 22 60 59 67 36-5
61 71
35-5
RAFAEL_RCP_MORROW_A_THICKNESS_3-31-11.GRD 55
Contour Interval = 5 3
2 38
HMU37-4
72
31-1C 31-2 32-4 2-B 35-4 35-2 37-2
14
0
5
HMU35-2
36-2 36-3
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
HMU36-4
28 36-4
HMU31-2
34 54 29 69 47 75
25
REMARKS 21 22 72
RMS P Amp, S22 Amp
33-1 34-1 39-3
23
Co-located, Co-krigged 38-1 39-1
41
Figure 14.12 A seismic-guided Morrow “A” isopach map showing the well patterns and the HMU boundary outline. The arrow shows the location of the new-drill well.
Pressure Change P-Wave Impedance

(Simulation) (Seismic Inversion)
1st Monitor-Baseline 1st Monitor-Baseline
10-2 11-2 11-4 4-M 10-2 11-2 11-4 4-M
10-3 10-3
10-3A
1-F 11-1 12-3 1-F 11-1 12-3

12-1 2-M-A 12-1 2-M-A
11-3 11-3
13-2 13-2
14-1 5 15-1 14-1 5 15-1

13-3 13-3
14-5 14-5
14-4 14-4
13-1 14-2 13-1 14-2
4-A 17-5 17-1 17-7 18-1 1-15 4-A 17-5 17-1 18-1 1-15
17-8 17-8 17-7
18-4 18-4
18-2 18-2
17-6 17-2 18-3 17-6 17-2 17-4

17-3 17-3 18-3
16#1 Pressure 16#1

1-B 1-B 19-1
Impedance
19-1 21-1 21-3 22-1 22-3 24-1 24-3 (psi) 21-1 21-3 22-1 22-3 24-1 24-3
Change (%)
7.5
300
19-2 21-4 21-2 22-4 22-2 24-4 19-2 21-4 21-2 22-4 22-2
24-2 24-4 24-2
22-5 22-5 5.0
200
27-1 30-B-1 35-7 27-1 30-B-1 32-1 35-7
30-2 32-1 32-3 35-1 35-6
35-3 100 30-2 32-3 35-1
35-3
35-6 0
704 35-5 704 35-5
2 2
501
0 501 31-1C −5.0
31-1C 32-4 2-B 35-4 35-2 32-4 2-B 35-4 35-2
31-2 36 31-2
100 −7.5
Heris, 2011 Zerpa, 2011
Figure 14.13 P-wave time-lapse map.
Pressure Change S-Wave Impedance

(Simulation) (Seismic Inversion)
1st Monitor- Baseline 1st Monitor- Baseline
10-2 11-2 11-4 4-M 10-2 11-2 11-4 4-M
10-3
10-3A 10-3
10-3A
1-F 11-1 12-3 12-1 1-F 11-1 12-3 12-1

2-M-A 2-M-A
11-3 11-3
13-2 13-2
14-1 5 15-1 14-1 5 15-1

13-3 13-3
14-5 14-5
14-4 14-4
13-1 14-2 13-1 14-2
4-A 17-5 17-1 17-7 18-1 1-15 4-A 17-5 17-1 17-7 18-1 1-15
17-8 17-8
18-4 18-4
18-2 18-2
17-6 17-2 17-4 17-3 17-6 17-2 17-4 17-3 18-3

18-3
16#1 Pressure 16#1

Impedance
1-B 19-1 21-1 21-3 22-1 1-B 19-1 21-1 21-3 22-1
22-3 24-1 24-3 (psi) 22-3 24-1 24-3
Change (%)
300 5
19-2 21-4 21-2 22-4 22-2 19-2 21-4 21-2 22-4 22-2 24-4 24-2
22-5 24-4 24-2 22-5
200 2.5
27-1 30-B-1 30-2 32-1 35-7 27-1 30-B-1 30-2 32-1 35-7
32-3 35-1 35-6
35-3 100 32-3 35-1 35-6
35-3 0
704 35-5 704 35-5
2 0 2 −2.5
501 31-1C 32-4 2-B 35-4 35-2 501 31-1C 32-4 2-B 35-4 35-2
31-2 36 31-2
100 −5
Heris , 2011 Zerpa, 2011

245
Figure 14.14 Shear wave time-lapse map.
WAG Optimization Figure 14.15 Simulation of the process suggests that an

0.25 increasing WAG ratio can increase oil recovery. Conversely,
decreasing the WAG ratio could result in higher associated
sequestration if the total CO2 was not held constant.
0.20
Oil Recovery, fraction
0.15
Continuous CO2
WAG (0.15 :1)
WAG (0.25 :1)
0.10 WAG (0.50 :1)
WAG (0.75 :1)
WAG (1.00 :1)
WAG (1.25 :1)
0.05 WAG (1.50 :1)
WAG (1.75 :1)
WAG (2.00 :1)
0.00
0.00 0.50 1.00 1.50 2.00 2.50
Cumulative Total Injection (HCPV)
Time-lapse seismic data were used to guide the one monitoring technique that can provide early qua-
permeability distribution in the reservoir simulation litative understanding of where CO2 exists in the
model (Heris et al., 2011). The reservoir simulation subsurface. Through proactive reservoir management
was used to evaluate scenarios for prediction of oil and techniques based on monitoring the system response,
CO2 recovery and associated CO2 storage under differ- both EOR economic goals and the associated safe
ent WAG injection schemes. The results are shown in long-term storage of CO2 can be optimized.
Figure 14.15. A 2:1 WAG injection scheme was recom-
mended to the Unit operator. However, WAG optimi- Delhi Field
zation on oil recovery is not necessarily the only focus by
Time-lapse, multicomponent seismic data were
an operator. WAG is more typically optimized on pro-
acquired during the startup of the CO2 flood at
duced gas handling and its associated facility capital
Delhi Field, Louisiana. The purpose of the integrated
infrastructure needs, which were not part of this evalua-
study at Delhi was to identify zones of high residual
tion. Generally, the WAG is modified by “wetting it up”
oil saturation and to monitor possible flow paths in
over years, but the cycle time is short, usually on the
the reservoir. Contacting bypassed oil is critical for
order of weeks, to control the CO2 mobility.
economic EOR.
Geosequestration Study Location

A CO2 project is effectively a “closed loop system.” Delhi Field is located in northeastern Louisiana
From an economical sense, projects are selected with (Figure 14.16). The field was discovered in 1944.
the intent of containing all the injected CO2. The reservoir consists of Upper Cretaceous
Whatever is injected within the reservoir complex (Tuscaloosa) and Lower Cretaceous (Paluxy) sand-
generally stays within the complex. The purpose of stones (Figure 14.17). As of this writing, the majority
monitoring is foremost to manage the fluid flow of the field is undergoing a CO2 flood for enhanced oil
within the system to maximize oil recovery. At the recovery. Depth to the top of the producing interval
same time CO2 stored in association with the EOR averages 3250 ft., porosity averages 25%, and the
process is being geosequestered in this closed loop average permeability is 1400 mD. Permeability in the
system. Nearly 100% of the injected volume is seques- reservoir is extremely variable and is controlled by
tered because what CO2 is produced is captured and depositional facies. Reservoir parameters are listed
repressurized for recycle back into the reservoir com- in Table 14.3.
plex. However, there are downhole situations that The Paluxy is made up of a fluvial deltaic
246 could allow movement beyond the confines of the interval of Early Cretaceous age. The Late
flood patterns and multicomponent seismic offers Cretaceous Tuscaloosa unconformably overlies the
Figure 14.16 Location of Delhi Field.
Paluxy and comprises transgressive marine depos- complexity of fluvial-deltaic environments we know
its overlain by a shallowing upward sequence of this assumed correlation to likely be inaccurate.
barrier bars capped by a fluvial deltaic interval at Also, the potential exists for connectivity between
the top. the Paluxy and the Tuscaloosa owing to overlapping
Gross thickness of the Paluxy varies from 75 to 30 hydraulic units.
ft. in the study area. The Tuscaloosa varies from 100 Static reservoir characterization can unravel the
ft. to a wedge or pinch-out in the updip direction. complexity of the geology to a large degree by careful
Distinguishing the connectivity of reservoir units in analysis of the core, well logs, and the baseline seismic
the Paluxy is the objective of this study. It is difficult survey, all of which was done at Delhi during the RCP
to assess the connectivity based on well control study. The value of monitoring comes in the under-
alone. Figure 14.18 shows assumed continuity of standing of flow unit scale conformance and
sandstone units between wells in cross-section view improved understanding of connectivity, i.e., 247
based entirely on well log correlation. Given the dynamic reservoir characterization.
Seismic Acquisition and Processing seismic surveys encompassing four square miles
each were centered on the CO2 flood startup location
Prior to the monitoring seismic surveys, a pre-CO2
(Figure 14.19). These surveys took place in June of
baseline P-wave seismic survey was acquired in late
2010, six months after the startup of the CO2 flood
2007 and early 2008. Later, two multicomponent
in November 2009, and again in August 2011.
The first few months of 2010 were excessively wet,
preventing acquisition of a survey. A cableless, three-
Table 14.3 Delhi Field reservoir parameters component nodal receiver system was used to acquire
Reservoir parameters the surveys. Dynamite was used as the source after
extensive source testing showed the ineffectiveness of
Reservoir depth approx. 3300 ft. (1 km) surface sources in acquiring data because of ground
Initial reservoir pressure approx. 1500 psia conditions in the area. Multicomponent seismic sur-
Reservoir pressure prior to CO2 flood approx. 1300 psia vey parameters are listed in Table 14.4 and the survey
40° API oil, GOR approx. 400 SCF\STB, Viscosity 0.77 cp outline is shown in Figure 14.20.
Reservoir temperature approx. 135°F The survey was designed to monitor the Paluxy
interval as the initial startup involved six Paluxy injec-
Porosity range: 25–30%
tors. Downdip from the Paluxy injectors, three
Permeability range: 250–3000 md Tuscaloosa injectors were also located in the seismic
Denbury to operate field approx. 1800 psi grid, but only two were operational during the time of
Pressure at injectors approx. 2400 psi our surveys. Each injection well was intended to inject
NW SE
MIDWAY SHALE TERTIARY

CLAYTON CHALK
UPPER CRETACEOUS
MONROE GAS ROCK
TUS
TUSCALOOSA
CAL
OOS
TUS A SA
CAL ND
OO
SA
HOLT-BRYANT ZONE
TUS SAN
CAL D
OO
PA SA
LU S A ND
Major XY
MU
PA RP
Unconformities LU
XY
HY
SA
LA ND
PA NE
NW LU SA
PALUXY
XY ND
LOWER CRETACEOUS
HO
GL LT
EN SA
RO ND
GL SE
SE EN S AN
RO D
SE
UPPER GLEN ROSE
SA
GLEN ROSE ZONE
MENGEL-UPPER
ND
MA
GENERALIZED DIP CROSS SECTION SS
IV
E
AN
HY
DR
IT
E
248 Figure 14.17 Delhi Field produces from Tuscaloosa and Paluxy sandstones, both truncated by unconformities.
Figure 14.18 Well log cross section showing the unconformity trap at Delhi. The Paluxy is highlighted. Above the Paluxy is the Tuscaloosa.
Table 14.4 Multicomponent, time-lapse seismic survey up to 10 MMscf per day. From November 2009 until
parameters
the second survey in August 2011, approx. 24 Bscf of
Acquisition parameters CO2 had been injected into the Paluxy. The minimum
miscibility pressure (MMP) is 2335 psi (Chen et al.,
RCV interval 82.5 ft.
2014) and the goal was to inject up to near miscibility
SRC interval 165 ft. while keeping the reservoir below the fracture gradient,
RCV line interval 495 ft. which at 3300 ft. depth is approx. 2475 psi. Monitoring
SRC line interval 495 ft. the overburden became an important and integral part
Receiver type 3C MEMS of the Delhi monitoring project. S-wave azimuthal
Source type Dyn (1lb @ 30 ft.) anisotropy was used as a tool to monitor stress changes
not only in the reservoir but also the overburden
Recording patch 18 × 108 (~4400 ft. × ~4400 ft. offsets)
(O’Brien and Davis, 2013). Time-lapse time shifts
Sampling rate 2 ms were also studied in the overburden on both the
Listen time 4s P and PS data to ensure overburden integrity through-
Bin size 82.5 ft. × 82.5 ft. out the injection startup where the pressure buildup
Trace density 663,552/sq. mi. (×3) was designed to go up to, but not exceed, 1800 psi in the
Acquisition by Tesla exploration Tuscaloosa–Paluxy zones of interest (Bishop and
Davis, 2014). Operating the immiscible flood at Delhi
250
Figure 14.19 Location of the RCP time-lapse seismic surveys.
Figure 14.20 Survey layout.
requires careful monitoring to maximize reservoir con- Petrophysical Data

tact while at the same time keeping pressures fairly
Log data, calibrated with core data from a new-drill
uniform and well below the fracture gradient.
well in the field, were used to determine reservoir
petrophysical properties. Sixty years of waterflooding
had taken place and yet there is still significant resi-
(a) dual oil in this reservoir. Seismic-assisted analysis of
NW SE
PP data reservoir properties through seismic inversion was
T250 T200 T150 T100 T50 used to build a reservoir model for simulation. This
model was further enhanced though time-lapse stu-
dies to map the permeability structure.
Fluid substitution modeling showed that a small
amount of CO2 would change the acoustic impedance
of the reservoir. Modeling suggested a strong observable
1000
1000
Paluxy response to CO2 injection. Both saturation change and

pressure change would affect the P-wave response.
The S-wave response is only influenced by pressure
change.
(b) PS data Multicomponent Seismic Data

Multicomponent (PP and PS) time-lapse seismic data
were acquired in Delhi Field in June 2010 and
1000
0
August 2011. Data quality was excellent even though
different ground conditions existed at the time of the
two surveys. The PP and PS data have been tied to the
1500
0
subsurface through quality well ties provided by mod-
ern dipole sonic logs (Figure 14.21).
Paluxy
Processing was undertaken to preserve the repeat-
2000
0
ability of the surveys. The time-lapse seismic data
were processed simultaneously. The NRMS was 0.19
for P-wave data and 0.21 for the PS data. The NRMS
2500
0 values indicate excellent repeatability. Generally,

a NRMS of 0.3 or less is considered good repeatability
Figure 14.21 PP and PS data from the 2010 survey.
251
and below 0.2 is excellent.
Amplitude maps Figure 14.22 P-wave RMS amplitude

map of the Paluxy over a 10-ms
Inline Color Key
window.
8587
P monitor 1 8241
7896
25 North 7550
7204
6859
6513
6167
50
5822
5476
5131
4785
75 4439
4094
3748
3402
100 3057
2711
2366
2020
1674
125
1329
983
275 250 225 200 175 150 125 100 75 50 25 637
Xline 292
Interpretation to be a permeability barrier or baffle preventing the CO2

from moving updip and/or a pressure barrier from updip
The Paluxy Formation was picked on the respective
injection. As a result, two downdip water injectors were
seismic volumes. RMS amplitude maps of the Paluxy
completed into the aquifer after our monitoring survey
were generated using windows of time that coincided
showed this downdip movement. The intent of the two
with the Paluxy on both volumes (Figures 14.22 and
water injectors was to provide pressure support to keep
14.23). The maps show that the P-wave data are very
the CO2 from moving into the aquifer. The August 2011
sensitive to fluid saturation in the sandstones of the
survey showed that the curtain was successful in creating
Paluxy. The six injectors were all operational at the
a pressure wall in the producing area of the oil column
time of the first monitoring survey. The PS data are
causing CO2 to preferentially sweep updip
not sensitive to fluid saturation change, but they
(Figure 14.25). A time-lapse pressure map was computed
are sensitive to lithology and pressure. Of interest is
from the near-linear relationship between pressure and
the trend of high amplitudes that are related to the
Vp/Vs. The map shows the effectiveness of the downdip
presence of distributary channels in the Paluxy.
water curtain that was implemented (Figure 14.26).
In addition, there are crevasse splays, clay plugs, and
Monitoring shows lateral migration of the CO2,
low-permeability zones affecting the flood response.
but what about the vertical? Figure 14.27 shows the
The CO2, because of its higher mobility and the fact
vertical migration from the Paluxy up into the
that this is a continuous CO2 injection flood, bypasses
Tuscaloosa between monitors 1 and 2. It is best seen
the lower permeability regions in the reservoir,
on the Vp/Vs time-lapse illustrations. CO2 acts as
degrading the sweep efficiency of the flood.
a tracer in delineating flow units in the reservoir.
A P-wave, time-lapse, difference map between the
A sand-on-sand contact across the unconformity
two surveys, clearly shows that some CO2 from the
allows the upward migration of CO2 into the
southern injectors in the Paluxy was moving downdip
252 Tuscaloosa affecting the lateral sweep in the Paluxy.
into the aquifer (Figure 14.24). The cause was theorized
Amplitude maps Figure 14.23 P–S RMS amplitude

map of the Paluxy over a 20-ms
Inline Color Key window.
8587
PS monitor 1 8241
7896
25 7550
7204
6859
6513
6167
50 5822
5476
5131
4785
75 4439
4094
3748
3402
100 3057
2711
2366
2020
1674
125
1329
983
275 250 225 200 175 150 125 100 75 50 25 637
Xline 292
Base Paluxy RMS Amplitude Difference Figure 14.24 CO2 is clearly moving
downdip from the southern injectors
in June, 2010.
244
234
224
214
204
193
183
173
163
153
143
133
123
112
102
92
82
72
62
52
42
32
21
11
1
253
had been produced through primary and secondary

RMS
recovery. CO2 EOR has rejuvenated an old field and
Top Paluxy 10-ms High seismic monitoring could assist in optimizing and
7500
window centered
predicting the incremental recovery from the field.
7000
1 mile
6500
146-1
6000
Reservoir Modeling and Simulation
5000 5500
RMS_Paluxy_10ms_Cen, UNKNOWN, GP558, AMPLITUDE
158 -4 Downdip mobility occurs on the eastern and southern
164-3
side of the study area, presumably along a distributary
4000 4500
140 -1 channel system. The time-lapse seismic data show
3500
bypass areas and clear evidence where CO2 is moving
1000
upward from the Paluxy into the overlying Tuscaloosa.
2000 2500
Monitoring has helped improve the reservoir character-
WOC ization by defining areas of high transmissibility that can
160-3
1500
be input into a reservoir model and improve reservoir
1000
simulation and associated predictions. Forecasting
500
Low
based on improved reservoir simulation can lead to
proactive reservoir management and greater potential
Figure 14.25 By time of the monitor 2 survey the CO2 had moved for economic success. The prize is forecast to be 10–12%
back updip. incremental recovery of the OOIP (Figure 14.29).
Production History Conclusions

Delhi Field is another example of a successful CO2 Time-lapse multicomponent seismic data have been
EOR project in that, by the time of the second mon- used to monitor CO2 floods at Postle Field, Oklahoma
itoring in 2011, the field was producing 4000 barrels and Delhi Field, Louisiana. It is recognized that the
per day after more than a decade long hiatus of being P-wave seismic response does not significantly change
shut-in (Figure 14.28). At the time of initiation of the after the initial fluid substitution occurs and these data
CO2 flood, 50% of the original oil-in-place (OOIP) show the importance of monitoring the reservoir,
Figure 14.26 Time-lapse multicompo-

4.0 nent pressure monitoring of the change in
pore pressure between the two monitor
3.2 surveys. (From Carvajal et al., 2014.)
Pore pressure change (MPa)
146-1
147-2 2.4
154-2
158-4 154-1 147-1 1.6
164-3
158-2
140-1 0.8
Y
165- 159-2
32 0.0
148-2
176-2
169-5
−0.8
160-1 150-3 −1.6
150-2 −2.4
500 m
X
Carvajal 2013
Producers
CO2 injectors
254 Water injectors

(a) (b)
Monitor1 Monitor2
159-2 159-2
Xline 270 230 190 110 70 30 Xline 270 230 190 110 70 30
25000 25000
900 900
Impedance (ft/s*g/cc)
Impedance ( /s*g/cc)
PP Time (ms)
PP Time (ms)
950 950
20000 20000
1000 1000
B
1050 1050
500 m 500 m
16000 16000
159-2
Xline 270 230 190 110 70 30 159-2
Xline 270 230 190 110 70 30
4 4
900 900
PP Time (ms)
PP Time (ms)
950
950
Vp/Vs
2.6
Vp/Vs
2.6
1000
1000
B
1050
500 m 1050
1.5 500 m
1.5
Figure 14.27 The CO2 is moving from the Paluxy into the overlying Tuscaloosa as evidenced on the time-lapse Vp/Vs response.
BOPD Production history MCFD & BOWD

20,000 100,000 10,000 50,000
18,000 Oil produced 90,000 9,000 45,000

Gas produced 8,000 40,000
16,000 80,000
Water produced 7,000 35,000
14,000 70,000 6,000 30,000
12,000 60,000 5,000 25,000
CO2 injection
10,000 50,000 4,000 20,000
8,000
Prior to 1970 40,000
3,000 15,000
no data 2,000 10,000
6,000 30,000
available 1,000 5,000
4,000 20,000 − −
2008
2008
2008
2008
2008
2008
2009
2010
2010
2010
2010
2010
2010
2011
2011
2011
2011
2,000 10,000
− Time (years)
1944
1948
1953
1957
1962
1966
1971
1976
1980
1985
1989
1994
1999
2003
2008
Figure 14.28 The production history of Delhi Field.
especially in the early startup of the EOR process. sandstone reservoirs. CO2 acts like a tracer to identify
S-waves, on the other hand, exhibit almost linear beha- permeability pathways in a reservoir. CO2 with seis-
vior in response to pressure, making the S-wave data mic monitoring can be used to identify flow units,
valuable for not only lithology prediction but flow paths, bypassed oil, and potential thief zones.
also pressure monitoring. Pressure monitoring is impor- Based on our studies at Postle and Delhi Fields we
tant for the CO2 EOR process and CO2 geosequestration. can say with great certainty that multicomponent
Case studies from Postle and Delhi Fields show seismic monitoring brought value through reservoir 255
that seismic can monitor movement of CO2 in these management modifications.
RCP area oil production rate and cumulative oil production prediction
2012 2016 2020 2024 2028 2032 2036 2040
1.2E+7
6000
Oil cumulative production, [STB]
Oil production rate, [STB/d]

8000000
4000
Historical oil cumulative production, STB
Historical oil production rate, STB/day
Simulated oil cumulative production, STB
4000000
Simulated oil production rate, STB/day
2000
Simulated oil cumulative production prediction, STB
Simulated oil production rate prediction, STB/day

0
0
2012 2016 2020 2024 2028 2032 2036 2040
Date
Figure 14.29 Prediction based on reservoir simulation shows the potential for 10–12% incremental oil recovery in the 4 square mile study
area of the field.
Early seismic monitoring can identify less efficient Carvajal, C., Putri, I., and Davis, T. (2014). Dynamic
reservoir processing, guiding the operator to modify reservoir characterization using 4-D multicomponent
reservoir management practices. A degradation in seismic data and rock physics modelling in Delhi Field,
Louisiana. First Break, 32(2): 63–69.
CO2 efficiency can be found in both flood pattern
management itself, and losses beyond the flood pat- Chen, T., Kazemi, H., and Davis, T. L. (2014). Integration of
tern. Managing the CO2 EOR process in conjunction reservoir simulation and time-lapse seismic in Delhi
Field: A continuous CO2 injection EOR project. SPE
with seismic monitoring brings economic dividends 169049.
and can assist in evaluating the associated CO2
Heris, A. E., Wandler, A., Kazemi, H., and Davis, T. L.
geosequestration potential.
(2011). Quantitative integration of flow simulation and
4-D multicomponent seismic in a CO2 WAG EOR
project. SPE 146960.
References
O’Brien, S., and Davis, T. (2013). Time-lapse shear
Bishop, J. E., and Davis, T. L. (2014). Multicomponent wave splitting analysis to monitor caprock
seismic monitoring of CO2 injection at Delhi Field, integrity at Delhi Field, Louisiana. First Break, 31(5):
Louisiana. First Break, 32(5): 43–48. 75–81.
256
Chapter
Integrated Geophysical Characterization
15 and Monitoring at the Aquistore CO2

Storage Site
Don White
Introduction measurements, emerging technology tests, and finally

by the first time-lapse measurements. A summary of
The Aquistore carbon dioxide (CO2) storage (geose-
the geophysical methods employed and some key
questration) and monitoring site is located near
results are provided in Table 15.1.
Estevan, Saskatchewan, Canada (Figure 15.1).
Aquistore is a commercial-scale storage project that
utilizes a deep saline formation as the storage reser- Background Geology
voir (Worth et al., 2014). As such, it is the world’s first The Aquistore storage site lies within the northern
integrated storage project associated with CO2 cap- Williston Basin, an intracratonic basin in central
ture from an industrial scale coal-fired power plant. North America (Kent and Christopher, 1994). At the
As of October 31, 2016, a total of approx. 100 ktonnes Aquistore site, sedimentary strata have a cumulative
of CO2 had been injected at the site. The Aquistore thickness of approx. 3350 m and comprise
project is conducting research into monitoring meth- a lowermost Cambro-Ordovician clastic sequence
ods for CO2 injection and storage within deep saline overlain by predominantly carbonates and evaporites
aquifers. The project aims to test and demonstrate of Ordovician to Mississippian age. Above these strata
a broad range of geophysical methods for this pur- are Mesozoic age shales, siltstones, and sandstones
pose. To date, geophysical methods have been utilized lying beneath thin Tertiary and Pleistocene clastic
in the initial site characterization, baseline data collec- deposits.
tion, and early CO2 storage monitoring. The final goal The CO2 storage reservoir is located within the
is to assess the relative effectiveness of the various lowermost Paleozoic section immediately above the
monitoring methods, what role they may play in a Precambrian basement. The Deadwood and overlying
storage program, and to integrate them where Winnipeg formations form an approx. 200 m thick
possible. interval with approx. 50% of the interval having
In this chapter, the results from the variety of potential for CO2 injection and storage. These forma-
geophysical methods that have been utilized at the tions have been the target of oil production and waste
Aquistore site are presented. The list comprises water disposal elsewhere in the basin (Brunskill,
various seismic techniques (four-dimensional [4D] 2004). The Deadwood Formation is a sandstone unit
surface, 4D vertical seismic profiles [VSPs], passive), that contains silty-to-shaley interbeds with subordi-
electromagnetic methods (magnetotelluric [MT], nate calcareous layers. The overlying Winnipeg
controlled-source electromagnetic [CSEM]), gravity, Formation includes a lower sandstone unit (Black
surface-deformation measurements (inteferometric Island Member) and an upper shale unit (Icebox
synthetic aperture radar [InSAR], GPS, tilt), and Member) that constitutes the reservoir caprock and
a number of emerging methodologies (borehole-to- primary CO2 storage seal. Regionally, Winnipeg–
surface electromagnetic [BSEM], distributed acoustic Deadwood sandstones have permeabilities of approx.
sensing [DAS], seismic interferometry, downhole 100–1000 mD and porosities of approx. 11–17%
gravimetry) that have been tested. The chapter begins (Vigrass et al., 2007). At the Aquistore site, increased
with the geological characterization of the site using shale content within this interval reduces permeabil-
three-dimensional (3D) seismic data and well logs. ities and porosities to 0.05–300 mD and 5–8%, respec-
This is followed by the various geophysical baseline tively. The Middle Devonian Prairie Formation is an 257
Don White
Figure 15.1 Aquistore Project site map showing the location of the Boundary Dam Power Plant (BD), the CO2 capture facility (CC), and the
CO2 injection well and observation well. Also shown are locations of the surface monitoring stations, short-period geophone arrays (L1 and L2),
area covered by the baseline 3D seismic survey, MT/CSEM sites, and bipole transmitter. Each surface monitoring station includes some
combination of GPS, InSAR reflectors, tilt meters, and broadband seismographs. Inset shows site location in Canada. Note that there is an
additional MT/CSEM sensor located 2.2 km to the southeast as indicated by the lower left symbol outside of the map frame. (© Her Majesty the
Queen in Right of Canada, as represented by the Minister of Natural Resources, 2017.)
approx. 200 m thick evaporite sequence at 2,500 m 2014 at a rate of approx. 2400 tonnes/day, whereas
depth that forms a competent regional aquitard CO2 injection at the Aquistore site was initiated
(Marsh and Love, 2014) and as such represents an in April, 2015. The bulk of the captured CO2 is
excellent regional secondary sealing unit for CO2 sto- delivered to the Weyburn oil field via pipeline for
rage at the site (Khan and Rostron, 2005; Palombi, the purpose of enhanced oil recovery (EOR), with
2008). any surplus CO2 being sent to the Aquistore site for
storage. The Aquistore storage site anticipates
receiving 250–300 ktonnes of CO2 over the first
CO2 Capture and Injection few years of operation at variable rates of up to 800
258 CO2 capture at the nearby SaskPower Boundary tonnes per day. As of October 31, 2016, approx. 100
Dam power plant (Figure 15.1) began in October, ktonnes of CO2 had been injected.
Integrated Characterization and Monitoring
Table 15.1 Summary of geophysical measurements made at the Aquistore site and key results
Method Results
4.0 Characterization and baseline
4.1 3D Surface seismic Seals are intact. Local fault reactivation unlikely. Suitable storage site.
4.2 Surface electromagnetics Regional 1D resistivity structure. Method suited for monitoring shallow CO2
leakage or 100s Mtonnes injection at approx. 3000 m depth.
4.3 Crosswell seismic tomography Detailed image of the reservoir between injection and observation wells.
4.4 Logging Characterization of reservoir properties and CO2 partitioning.
4.5 InSAR, GPS, tilt meters Background surface deformation of 0.2–1.0 cm/year and 0.1–2.0 μrad tilt is
comparable to expected signal levels for 0.5 CO2 Mtonnes/year.
4.6 Surface gravity Background variability of ±20 μGal makes method suited for monitoring
shallow CO2 leakage or 100s Mtonnes injection at 3000 m depth.
5.0 Technology/methodology tests
5.1 Distributed acoustic sensing VSP VSP images obtained are comparable to conventional geophones VSP.
5.2 Borehole gravity No significant borehole gravity density difference with 6500 tonnes injected
150 m away.
5.3 Borehole to surface Measured SNR compared with modeled response suggests approx. 100
electromagnetics ktonnes CO2 detectable.
5.4 Seismic reflection interferometry Promising but requires improvements for time-lapse monitoring at 3000 m
depth.
6.0 Monitoring
6.1 3D surface seismic 200 m wide anomaly in the upper Deadwood unit corresponding to approx.
16 ktonnes CO2. No CO2 observed in Black Island or lower Deadwood where
approx. 4 and 16 ktonnes injected.
6.2 3D VSP Similar to 3D surface results.
6.3 Passive seismic monitoring No significant injection-related seismicity (Mw > –1) detected during first 17
months of CO2 injection, and no smaller magnitude seismicity (Mw > –3)
during the first 8 months.
CO2 is transported to the Aquistore site by pipe- 3D Geological Characterization

line from the capture unit, which is located approx.
3 km to the east (Figure 15.1). As designated by the and Baseline Measurements
provincial regulating agency, CO2 is injected at Design of the CO2 storage and monitoring program
pressures of less than 90% of the estimated forma- for the Aquistore site required determination of how
tion fracturing pressure (gradient of 14.9 kPa/m). large an area would be appropriate for site character-
The initial ambient pressure and temperature ization and subsequent active monitoring during CO2
conditions in the reservoir were approx. 115°C and injection operations. The regional geological architec-
approx. 35 MPa. The temperature of the injected ture near the storage site was well known from large-
CO2 is variable depending on the time of year. scale compilations (e.g., Whittaker et al., 2004 and
Well perforations are located over four depth inter- references therein). Initial characterization of the pro-
vals within the reservoir corresponding to the zones spective CO2 storage site was conducted through
of highest permeability/porosity as determined by interpretation of an existing set of five 2D seismic
well logging. profiles in conjunction with geophysical/geological
259
Don White
logs from nearby wells (RPS Boyd PetroSearch, 2011). continuous and shows no evidence for vertical faulting,
The seismic profiles covered an area of approx. 140 as is also true for the regional evaporitic secondary seal
km2 and showed no evidence for features that would (Prairie Formation) that is greater than 150 m thick and
disqualify the site for storage. A 3D geological model has no salt dissolution features. (4) Above the Prairie
was then constructed that formed the basis for fluid- Formation lie 1500 m of Middle Devonian to Lower
flow simulations using maximum anticipated injec- Cretaceous strata that underlie an additional 1000 m
tion rates for 25 years (e.g., Whittaker and Worth, of Upper Cretaceous and younger sedimentary
2011). The resultant CO2 footprint from these simu- rocks including thick tertiary sealing units: the
lations was approx. 3 km in diameter for injection Watrous Formation (approx. 120 m), Colorado Group
rates of up to 1600 tonnes/day. Based on this result, (>185 m), and Bearpaw Formation. (5) Evidence is
detailed geophysical monitoring was focused on an absent of any vertical faults that extend above the
area of approx. 3 × 3 km. Silurian section (or approx. 2700 m depth). (6)
Specific details of the geology at the storage site Whereas a local subvertical basement fault has been
have been obtained by 3D seismic imaging (White interpreted at the site, it is oriented at an azimuth of
et al., 2016), magnetotellurics (McLeod et al., 2014), 75°–85° relative to the regional maximum horizontal
crosswell tomography, analysis of core samples stress, making reactivation during CO2 injection unli-
(EERC, 2014), well logs, and injection tests from the kely. A flexure observed in overlying Cambrian to
two wells (Rostron et al., 2014). In addition to char- Silurian sedimentary rocks does not appear to be rup-
acterization, these various surveys provide baseline tured or faulted.
measurements (see next section) for the purposes of
time-lapse measurements. Results from each of these Surface-Based Electromagnetic Methods
studies are described in the text that follows.
Surface-based magnetotelluric (MT; e.g., Vozoff, 1991)
and controlled-source electromagnetic (CSEM) meth-
3D Surface Seismic Data ods provide an alternative to seismic methods for
3D seismic data were acquired at the Aquistore site in defining the properties of the host geological structures
2012 to provide the lithostratigraphic framework for and for monitoring underground CO2 storage,
a 30-km2 area (Figure 15.1) around the injection well although with generally lower spatial resolution (e.g.,
(see White et al., 2016 for details). The specific objec- Gasperikova and Hoversten, 2006). Preinjection broad-
tives were (1) to map the Precambrian basement band MT, audio-frequency MT (AMT), and CSEM
structure and overlying sedimentary section; (2) to surveys were conducted at the Aquistore site
delineate the storage reservoir and associated in August, 2013, November, 2014 (MT and AMT
caprock/primary seal; (3) to image the regional con- only), and November, 2015 with the specific objectives
tinuity of the secondary seal (Prairie Formation); and of imaging the host geological structure, defining the
(4) to look for subvertical pathways for potential background EM noise levels and testing the resolution
upward migration of CO2 from the reservoir. and repeatability of natural-source and controlled-
An east–west section from the 3D seismic volume is source EM methods (McLeod et al., 2013; McLeod
shown in Figure 15.2. Interpretation of the 3D data was et al., 2016). The MT and CSEM recording sites
aided by correlation with the Aquistore well logs (Figure 15.1) were located along a profile passing near
(Figure 15.3). Along with consideration of the local the CO2 injection well. The CSEM utilized a 30 Amp,
seismotectonics, this interpretation led White et al. horizontal bipole transmitter that was located approx.
(2016) to conclude that the site has the essential geolo- 3.5 km northeast of the injection well. The transmitter
gical features required for large-scale CO2 storage. produced a rectangular waveform with frequencies
The conclusion was based on the following: (1) between 0.008 and 32 Hz. Additional MT sites were
The site is located in a region of very low natural located off the central part of the profile to provide areal
seismicity with the nearest demonstrably seismogenic coverage.
fault zone located approx. 200 km away. (2) The clastic The observed MT responses are consistent with
reservoir is part of a regionally extensive formation and previous studies within the Williston Basin (e.g.,
is at least 200 m thick at the storage site. (3) The primary Gowan et al., 2009). The 1D resistivity from log-
260 reservoir seal – a 15 m thick shale caprock— is laterally based inversion of the MT data is shown in
Figure 15.2 West-to-east section from the 3D time migrated seismic volume located approx. 100 m north of the injection well.
The approximate location of the injection well is marked by the derrick symbol. Inset (left) shows the tie between the 3D seismic data and the
well-based synthetic seismogram. Seismic horizons identified by correlation with injection well geology are labeled. (Modified from White et al.
[2016]. © Her Majesty the Queen in Right of Canada, as represented by the Minister of Natural Resources, 2017.)
Figure 15.4. McLeod (2016) concluded that (1) MT measurements generally do not have the sensitivity
impedance estimates obtained from the field measure- or resolution for monitoring CO2 at the reservoir
ments were repeatable to within 1–3% under favorable level, but that the CSEM method is suitable for mon-
conditions; (2) nonrepeatability can be attributed itoring large quantities of injected CO2. For example,
mainly to the effects of ambient electromagnetic a 240-m thick, 50 Ωm (CO2-saturated) reservoir inter-
noise rather than difficulties in replicating the val would need to have lateral dimensions of approx.
field work; (3) modeling showed that the MT 3–5 km for the scattered electromagnetic fields to 261
Don White
Figure 15.3 Seismic well tie. From left to right are log-based synthetic seismic trace (gray) superposed on the 3D seismic data at the injection
well location, tops of a subset of geological units, and logs of compressional wave velocity (Vp), acoustic impedance (AI), and natural gamma
ray. The nonlinear depth scale is determined from the two-way travel times using the velocities from the well-tie process. (Modified from White
et al. [2016]. © Her Majesty the Queen in Right of Canada, as represented by the Minister of Natural Resources, 2017.)
produce detectable phase anomalies of 0.8˚–1.9˚. This image (Figure 15.5) provides meter-scale definition of
suggests that the surface CSEM methodology is better the reservoir but is limited to the 150-m-wide 2D zone
suited for monitoring for CO2 leakage into shallower, between the injection and monitoring wells. In a more
more conductive geological formations. complex reservoir setting, details on this scale might be
useful in defining reservoir heterogeneity, but in the
Aquistore setting the primary utility of crosswell ima-
Crosswell Seismic Tomography ging is in monitoring details of how the CO2 plume
The highest resolution seismic images of the reservoir initially spreads. Although saturation logs (e.g.,
are provided by crosswell tomography (e.g., Sato et al., Reservoir Saturation Tool [RST] or Pulsed Neutron
2011). At the Aquistore site, the tomographic velocity Decay [PND] logs) in the observation well are capable
262
10000
a) b)
1000
100
10
0
1
Figure 15.4 (a) Resistivity log from
0 0
Till observation well with geological units
Bearpaw (reproduced from McCleod, 2016).
200
(b) Resistivity obtained by 1D iterative
linearized inversion of MT data using the
400
base model as the starting model.
Belly River
(Modified from Figure 8.7 of McLeod
600
(2016). © Her Majesty the Queen in Right
800 of Canada, as represented by the Minister
of Natural Resources, 2017.)
Colorado
1000 1000
Manville
1200 Vanguard
1400 Jurassic
1600
Depth (m)
Watrous
1800
2000 2000
Bakken
Devonian
2200 Duperow
2400 Manitoba
Dawson Bay
Prairie Evaporite
2600
Interlake
2800
3000 Yeoman 3000

Icebox
Black Island
3200 Deadwood
Precambrian
3400
Base model
Recovered model
10–1 101 103
Resistivity (Ωm)
of determining CO2 arrival and time-lapse saturation deployed in the injection well and a vibratory source
profiles, crosswell seismic tomography and reflection in the observation well. Source and receiver positions
imaging are potentially capable of providing a CO2 were occupied at 1.5-m intervals over the depth range of
saturation cross-section between the wells. Other down- approx. 2890–3375 m. At each source location, the
hole methods (gravity, EM) are also capable of provid- source was operated eight times with a 2.6-second linear
ing this type of information. Thus, the main objective in upsweep of 100–800 Hz. The resultant eight records
acquiring the crosswell seismic tomogram was to pro- were stacked to increase signal-to-noise ratio (SNR).
vide a baseline survey for further time-lapse imaging. Multiple passes of the source over the entire depth
The baseline crosswell seismic survey was acquired range (2889–3374 m) were required for each
in March, 2013 (White et al., 2014). The crosswell data 28.5-m interval covered by the receiver array between
were recorded using a 20-level hydrophone array 2953 and 3372 m.
263
Don White
Figure 15.5 Composite 2D seismic reflection/tomographic velocity image for the zone between injection and observation wells. Horizontal
exaggeration is approx. 1.5:1.0. Porosity log and perforated intervals (blue bars) are overlain. Also shown (from left to right) for the reservoir
interval are a log-based synthetic seismogram, Vp (black) and density (red) logs with a Vp(z) curve (blue) derived from the velocity image
superposed, the gamma log, and a lithologic log. (© Her Majesty the Queen in Right of Canada, as represented by the Minister of Natural
Resources, 2017.)
Tube wave energy was strong in the raw corre- acquiring these logs was to characterize rocks within
lated data and had to be removed during processing. the storage complex and to determine baseline
3D anisotropic travel-time tomography (e.g., properties prior to the start of CO2 injection. The log
Washbourne et al., 2002) was applied using travel suite comprised resistivity, gamma ray, dipole sonic,
time picks for both first breaks and reflection events. density, neutron porosity, elemental and natural
The starting structural model used for tomography gamma ray spectroscopy, and nuclear magnetic reso-
was derived from the formation tops in the two nance. Baseline CO2 saturation (pulsed neutron) logs
wells. Reflection events could be recognized and were acquired over the reservoir interval. In situ stres-
were picked on amplitude versus angle (AVA) gath- ses were partially characterized by measurement of
ers. A complementary reflection image was obtained reservoir formation pressures and two mini-frac tests.
by processing the upcoming reflection events. Temperature–depth profiles were measured continu-
ously using the fiber optic distributed temperature
Logging sensing (DTS) systems attached to the tubing string
in the injection well and behind casing in the observa-
An extensive suite of geophysical well logs was
tion well.
acquired in both the injection and observation
Figure 15.5 shows examples of the logs over the
wells before and after casing and cementation of
264 reservoir interval. As can be seen, the injection well
the wells (Rostron et al., 2014). The purpose of
perforation intervals are located within the better Model-based tilt estimates exceed 10 μrad, which
quality sand units. Downhole CO2 flow measure- is significantly greater than background tilt noise
ments from a spinner survey conducted shortly after (Earth tide and precipitation effects) of 0.1–2.0 μrad
the start of injection indicated that approx. 10%, 45%, measured at the site (Wang, 2015). These estimated
and 45% of the injected CO2 was flowing respectively signal levels are small but detectable and are based on
into the Winnipeg, upper Deadwood, and lower CO2 and pressure communication being restricted to
Deadwood formations. the reservoir zone. If this is not the case (e.g., CO2
migrates or pressure communicates with shallower
InSAR, GPS, Tilt Meters depths) then greater ground deformation can be
expected.
Ground deformation monitoring has proven to be
a powerful tool for identifying nonconformance in
large-scale CO2 storage projects (e.g., Rutqvist et al., Gravity
2010; Vasco et al., 2010). A total of nine surface- Time-lapse gravity measurements have been utilized
deformation monitoring sites (Figure 15.1) covering for monitoring water-table levels (Liard et al., 2011),
a total area of approx. 8 km2 were deployed at the oil field water injection (Brady et al., 2006), and CO2
Aquistore site and comprise some combination of injection monitoring (Sherlock et al., 2006; Alnes
GPS receivers, InSAR reflectors, and tilt meters. et al., 2008). The first objective of surface-based grav-
These stations were deployed in 2012–2013 providing ity measurements at the Aquistore site was to deter-
two to three years of preinjection baseline data. mine the maximum background seasonal gravity
From June 2012 to October 2014, space-borne variations. Typically, these variations can be on the
differential interferometric synthetic aperture radar order of ±10 μGal and are generally attributed to
(DInSAR) monitoring allowed measurement of the changes in ground water levels (Liard et al., 2011).
surface deformation field associated with back- The second objective was to monitor for gravity
ground natural and anthropogenic processes changes associated with the injection of CO2.
(Samsonov et al., 2015; Czarnogorska et al., 2016). Anticipated CO2- injection-induced gravity changes
Vertical and horizontal ground deformation was are expected to be on the order of a few μGal for
measured with local maximum rates of ±1.0 and Mtonne-scale CO2 quantities at the reservoir level,
±0.5 cm/year and with precision of 0.3 and 0.2 cm/ and thus are generally below the noise level.
year (2ơ), respectively. Areas of high background However, in the case of upward migration of CO2
elevation changes are associated with processes into the overburden, the associated gravity changes
such as erosion, groundwater withdrawal and (10’s of μGal; e.g., White, 2012: p.176) can signifi-
recharge, mining operations, and land reclamation. cantly exceed background noise levels.
Average vertical velocities determined from GPS Gravity measurements were made at the Aquistore
monitoring are –2.0 mm/year (Craymer et al., site in fall 2013, spring 2014, fall 2014, and fall 2015.
2015), consistent with the regional background velo- Up to 11 sites were visited in each of the surveys. All
city predicted from a Canada-wide velocity solution measurements were made with an A10 absolute gravi-
(Craymer et al., 2011). The GPS and InSAR veloci- meter with estimated uncertainties of approx. 5 μGal.
ties are generally in good agreement. The mean magnitude of measured gravity differences
Simple poroelastic analytic modeling (Mathias for 10 Aquistore sites from fall 2013 to spring 2014
et al., 2009; Xu et al., 2012a) was conducted for the was 18 μGal. This is significantly larger than gravity
Aquistore site by Samsonov et al. (2015), who esti- changes of 0.34 μGal/year due to regional subsidence
mated that the maximum expected vertical deforma- (–0.2 cm/year) or maximum differences of ±3 μGal/
tion will be centered on the injection well and will be year due to local vertical ground motions of ±1.0 cm/
less than 1.6 cm for CO2 injection at a rate of 0.5 year (see previous section). The measured gravity
Mtonnes/year. Vertical deformation in close proxi- variability at the Aquistore sites may result from
mity to the injection well exceeded 0.5 cm as early local “transient” mass transfer signals associated
as one year after the start of injection, and vertical with changes in total water storage, where 10 μGal
deformation exceeding 1.0 cm is predicted to dis- would be equivalent to a 24 cm thick layer of water.
tances of approx. 6000 m after 25 years of injection. The variability in water table levels from local 265
Don White
monitoring wells for May 2013, October 2013, DAS acquisition. The fiber optic cable in the observa-
and June 2014 is 18 cm (Klappstein, pers. comm., tion well (Figure 15.1) extends from the surface to
2016). For 20–40% porosity, this would result in grav- a depth of approx. 2700 m. As was subsequently
ity difference of only 2–4 μGal. Thus, changes in the demonstrated at the Aquistore site, comparable DAS
water table alone are insufficient to explain the data can be acquired using either single or multimode
observed gravity differences. In regard to CO2 mon- fibers, as reported in Harris et al. (2015, 2016) and
itoring, related gravity changes would exceed back- Miller et al. (2016).
ground noise only if large volumes of CO2 migrate to A direct comparison of VSP data acquired using
depths of less than 1000 m. For example, 1.5 Mtonnes DAS and clamped geophones was conducted in
of CO2 at 600 m depth would produce a gravity anom- 2013 as part of the baseline survey prior to any
aly of 60 μGal (White, 2012). CO2 injection. Figure 15.6a compares raw data for
DAS and geophones. The same seismic arrivals
Technology/Methodology Tests (direct wave and reflections) can be seen in both
data but the geophone data clearly have a higher
Distributed Acoustic Sensing Data SNR; 45 dB as compared to 15 dB for the DAS
data (Harris et al., 2017). Furthermore, the band-
Acquisition width (for SNR >1) of the geophone data (approx.
Early successful results of VSP recording using dis- 5–150 Hz) is greater than that for the DAS data
tributed acoustic sensing (DAS; e.g., Mesteyer et al., (approx. 0–80 Hz). Figure 15.7 compares 2D
2011) provided the incentive to include a casing- depth-migrated images obtained for both DAS and
conveyed fiber optic cable in the designs for the geophone data. Despite the lower SNR, the DAS
injection and observation wells that were drilled at data produce images that are comparable to the
the Aquistore site in 2012. A multimode fiber cable geophone data and, because of the greater deploy-
was included in the original well design, but a single- ment interval for the DAS, provide imaging cap-
mode fiber cable was added for the sole purpose of ability at much shallower depths.
Figure 15.6 (a) Raw DAS seismic data and vertical-component geophone data for a single shot gather in 2013. The shot point is located
470 m from the observation well. Geophone data are substituted for DAS data in the depth range of 1455–2295 m. There are 963 single-mode
DAS traces spaced at 2-m intervals and 57 vertical-component geophone traces spaced at 15-m intervals. (b) A comparison of RAW DAS data
266 for shots at the same shot location in 2013 and 2016 surveys (SP4608-3800). (© Her Majesty the Queen in Right of Canada, as represented by the
Minister of Natural Resources, 2017.)
0 Figure 15.7 Eight-fold 2D-

migrated VSP images for the 2013
DAS and vertical-component
geophone data superposed on
500 a slice from the migrated 3D surface
data (green/blue/red) along an east–
Lower Colorado west line. Geological formations are
1000 Mannville labeled. The geophone image shows
the same area as the DAS image, but
has been shifted left to allow
Watrous comparison. (Modified from Harris
DAS
1500 Poplar et al., 2016. © Her Majesty the Queen
Depth (m)
in Right of Canada, as represented by

the Minister of Natural Resources,
2017.)
2000 Bakken 1 Shale
Birdbear
Prairie Evaporite
2500
Winnipegosis
3000
Winnipeg
Reservoir Deadwood
Precambrian
The first monitor VSP was acquired in gravity by making a measurement at a surface station
February 2016, again using both DAS and downhole occupied by an A5X absolute gravity meter. Tool depth
geophones. The improvement in the DAS acquisition positions are accurate to 1–2 cm, resulting in depth-
technology from 2013 to 2016 is clearly visible in related gravity errors of approx. 1–2 μGal.
Figure 15.6b and 15.6c, where the SNR is much In the absence of preinjection baseline measure-
improved. The results of the time-lapse imaging from ments, it is not possible to consider time-lapse gravity
this survey are discussed in the “4D VSP” section. changes for the August 2015 survey. As a substitute,
In addition, a 35-day short-term test of passive seis- Black et al. (2016) compared borehole gravity (BHG)
mic monitoring was conducted using DAS in densities calculated for the postinjection survey with
April 2015. The purpose of this experiment was to assess open hole γγ log densities acquired prior to any CO2
the utility of DAS as a potential alternative to downhole injection. The density differences (Figure 15.8) repre-
geophone systems for microseismic monitoring. sent a combination of mass changes away from the
wellbore and near-wellbore fluid invasion-related
changes. At the time of the survey, PND logs showed
Downhole Gravity no CO2 saturation at the observation well, and thus the
Borehole gravity measurements for the purposes of CO2 presence of any CO2 would contribute to mass changes
monitoring have been investigated in a number of stu- only away from the wellbore. Brine and supercritical
dies to date (e.g., Sherlock et al., 2006; Gasperikova and CO2 have a density difference of 520 kg/m3 at reservoir
Hoversten, 2008; Dodds et al., 2015). Borehole gravity P, T conditions and thus zones where CO2 replaces
measurements were made in the Aquistore observation brine will result in a local mass deficit. The BHG-γγ
well during 2014–2015. After several attempts, data density differences (Figure 15.8b) range from –50 to 40
were successfully acquired with a prototype BluecapTM kg/m3. Zones where CO2 replaces brine away from the
borehole gravity meter in August 2015 with resolution wellbore should produce a negative density difference
of less than 1 μGal and a repeatability of 10 μGal (Black that would reach a maximum of –46 kg/m3 when the
et al., 2016). Notably, this gravimeter was operating CO2 arrived at the observation well. The positive den-
under high-temperature conditions (115°C). sity differences provide a measure of non-CO2-related
The survey occurred four months after the start of variability in the BHG-γγ density differences, as only
CO2 injection when a total of approx. 6500 tonnes had negative differences would be expected due to the pre- 267
been injected. The borehole survey was tied to absolute sence of CO2. This indicates that the uncertainties are
Don White
Aquistore site (e.g., Daley et al., 2015). Early plans

to deploy casing-conveyed borehole electrodes were
abandoned owing to complexities of deploying casing
at the reservoir depth (>3200 m). As an alternative,
a borehole-to-surface electromagnetic (BSEM) method
was chosen for testing that utilizes surface-based source
and receivers (Hibbs, 2015). The high conductivity steel
well casing is utilized to channel electric current from
a surface source to the deep reservoir. The purpose of
the survey was to test the suitability of this methodology
for time-lapse CO2 monitoring – specifically, whether
adequate SNR could be achieved to allow detection of
the very small signals that would be expected due to
CO2 injection.
A BSEM survey was conducted in August
to September of 2013 (Hibbs, 2015) prior to the start
of CO2 injection. Capacitively coupled electric-field
sensors were deployed along two 1000-m lines emanat-
ing radially from the location of the CO2 injection well
(Figure 15.9), each with 13–15 sensor locations. A 20
A source transmitted a square wave with fundamen-
tal frequency of 0.5 Hz for at least one hour for each
receiver position. Hibbs (2015) used noise measure-
ments made during the survey to assess whether the
surface measurements would have enough sensitiv-
ity to detect a CO2 saturated zone within the reser-
voir. The measured noise was compared with the DC
Figure 15.8 (a) Comparison of preinjection open-hole γγ density response at the surface predicted by modeling
logs (RHOZ, RHOZ blocked) and the borehole gravity (BHG) density
values (IDEN) for the observation well. (b) Density differences (method of Schenkel and Morrison, 1994) for
between BHG density values for the postinjection gravity survey and a model that included a 150-m radius, 30 m thick
the open hole γγ densities for (RHOZ blocked) logs acquired prior to CO2 saturated zone. A uniform background con-
any CO2 injection. Perforated intervals are indicated by the black
bars. (c) Interval Bouguer gravity measurements (μGal) for a density ductivity of approx. 0.4 S/m (2.6 Ωm) was used
of 2620.8 kg/m3. Both (a) and (b) are modified from Black et al. (2016). consistent with the conductivity at the reservoir
depth (3200–3400 m) from well logs. The DC
comparable to the signal (maximum of –46 kg/m3) we response for a 20 A source predicted a difference of
are looking for. Also, a comparison of the injection 0.1–0.2% which, after adjustment based on skin
intervals with zones of negative density difference depth, predicts that a change of approx. 0.01% will
shows no correlation, suggesting that the gravity effect be observed at the surface owing to the presence of
of 6500 tonnes of CO2 injected 150 m away from the CO2 in the reservoir. Accounting for the very high
observation well was too small to detect. Further vali- SNR achieved with the survey geometry, a SNR ≥ 10
dation of gravity (or density) differences with the is achieved out to 240 m offset from the injection
approach of CO2 will have to wait for the first monitor well and a detectable signal (SNR ≥ 2) continues out
survey, which will use the 2015 survey as a baseline. to approx. 600 m. Notably, the recorded signal was
reduced by a factor of 10 or less only over the dis-
tance of 80–1140 m from the injection well. This
Controlled-Source Borehole to Surface EM suggests that this technique might be applied over
Survey a much broader area if higher SNR can be achieved
A variety of controlled-source borehole-based EM by increasing the transmitter current and/or extend-
268 methods were considered for application at the ing the signal averaging time.
Figure 15.9 Plan view of the

BSEM survey configuration
comprising six electric dipole
sources, each of which used
the injection well casing as
one electrode in combination
with a far-offset electrode
(blue dots). Two arrays of
receivers (dashed red lines).
(© Her Majesty the Queen in
Right of Canada, as repre-
sented by the Minister of
Natural Resources, 2017.)
Seismic Interferometry The passive 2D seismic image (Figure 15.10) bears

moderate resemblance to the 2D section extracted
Feasibility studies of seismic interferometry for CO2
from a controlled-source 3D seismic cube. The
storage monitoring are limited (e.g., Boullenger et al.,
ambient noise image and well-based synthetic seismo-
2015; Gassenmeier et al., 2015; Xu et al., 2012b). Passive
grams show moderate correlation values of approx.
seismic reflection interferometry has been conducted at
0.5–0.65 in the two-way time range of 0.8–1.5
the Aquistore CO2 storage site to test its ability to image
s. The quality of the ambient noise image is limited
the subsurface to the depth of the storage reservoir
by the unidirectional and weak nature of the local
(3200 m) and to assess its potential for time-lapse
noise source, which is most likely the local power
imaging (Cheraghi et al., 2017). Data from two passive
plant and nearby associated industrial activity, all
monitoring lines (L1 and L2 in Figure 15.1) were ana-
located approx. 3 km to the east. The best ambient
lyzed for 23 days of recording in June 2014 and 13 days
noise images were achieved by maximizing the dura-
in February 2015. Virtual shot gathers retrieved for
tion of the utilized noise panels. Significant differ-
noise panels of up to one hour in length were subjected
ences observed in passive images – at times
to a common midpoint (CMP) processing sequence to
separated by eight months – indicate that substantial
obtain 2D reflection images. The best image obtained
improvement in the repeatability of the images will be
using this approach is shown in Figure 15.10 where it is
required before this methodology can be used for
compared with a colocated controlled-source image
time-lapse imaging at the Aquistore site. Cheraghi
from Roach et al. (2015) and a log-based synthetic 269
et al. (2017) suggest that possible improvements
seismogram.
Don White
Figure 15.10 (a) 1D log-based synthetic seismogram superposed on L1 ambient noise image (from Cheraghi et al., 2017). (b) Synthetic
seismogram generated from a wavelet extracted from the image shown in (a) convolved with borehole log data from the Aquistore site.
The superposed wiggle trace in (a) is the same trace shown in (b). (c) Cross-line slice from the 3D active-source migrated seismic cube (from
Roach et al., 2015) at location of L1. The passive and controlled-source images have been filtered to 30Hz and the shape of the amplitude
spectra were generally matched within the passband. The active-source images shown in (c) were generally devoid of frequencies lower than
10 Hz prior to filtering. (d) Velocity log used to generate the synthetic traces in (b). (Figure is modified from Cheraghi et al. [2017]. © Her Majesty
the Queen in Right of Canada, as represented by the Minister of Natural Resources, 2017.)
might be achieved by using the areal array of geo- dynamite seismic surveys in March 2012,
phones of the Aquistore permanent array for 3D May 2013, and November 2013 (referred to as
imaging and by strategic selection of time periods Baseline, Monitor 1, and Monitor 2 surveys).
for analysis to better match the spectral characteristics The objective of acquiring these data was to assess
of the predominant noise sources. the data repeatability and overall performance of the
permanent array. Comparing the raw data from the
Monitoring Results first two of these surveys (Baseline and Monitor 1)
with a conventional high-resolution 3D Vibroseis
3D Time-Lapse Surface Seismic Data survey, White et al. (2015) concluded that (1) a 6-
A permanent sparse areal array of buried geophones to 7-dB increase in SNR is achieved for the buried
(Figure 15.11) deployed at the Aquistore CO2 site has geophones relative to surface-deployed geophones
been used for 4D seismic monitoring (White et al., and an additional 20-dB increase is observed for
2015). The use of permanent buried geophones was single dynamite shots compared to a Vibroseis
designed to improve the repeatability of the time- source; (2) the permanent array data has excellent
lapse data by reducing the effects of near-surface repeatability with a mean normalized root mean
variability, ensuring consistent receiver coupling, square (NRMS) value of 57% for the raw baseline –
eliminating intersurvey positioning errors and monitor difference; (3) shot gathers have a higher
increasing the SNR. The sparse permanent array also NRMS variance (18%) than do receiver gathers (7%);
economizes monitoring by minimizing mobilization (4) raw data repeatability is improved by a factor of 3
and deployment costs, allowing use of the geophones over comparable surface geophone data acquired at
for multiple purposes (e.g., controlled-source surveys a nearby location. The use of a permanent sparse
and passive monitoring) and accommodating flexible buried array demonstrably achieved reduced ambi-
on-demand surveys. ent noise levels and increased data repeatability,
270 Prior to the start of CO2 injection, the permanent both of which are keys to successful 4D seismic
array was used to acquire a total of three 3D monitoring.
Figure 15.11 Aquistore permanent

geophone array. The positions of
sensors forming the 2D permanent
array are shown along with the shot
points used for the 3D time-lapse
surface and VSP surveys. Also shown
are the continuous-recording passive
monitoring lines (L1 and L2), and the
locations of the surface broadband
seismograph stations. (© Her Majesty
the Queen in Right of Canada, as
represented by the Minister of Natural
Resources, 2017.)
The first post-CO2-injection 3D time-lapse reduced repeatability (labeled as N in panel c).

seismic survey (Monitor 3) was conducted at the As summarized in Roach et al. (2017), the following
Aquistore CO2 storage site in February, 2016 observations can be made in respect to Figure 15.12:
(Roach et al., 2017). By this date, 36 ktonnes of CO2 (1) No significant NRMS changes occur within the
had been injected within the reservoir. Monitor 3 Black Island unit. (2) A 200-m-wide zone in the
data were processed using the same “4D-friendly immediate vicinity of the injection well occurs within
simultaneous” processing flow used by Roach et al. the upper Deadwood unit, where NRMS values range
(2015) for the previous surveys. Excellent repeatabil- from 10% up to 25%. (3) In the lower Deadwood unit,
ity was achieved among all surveys with global no significant NRMS changes are observed near the
NRMS (GNRMS) values of 113–119% for the raw injection well. Large-amplitude NRMS changes at dis-
prestack data relative to the baseline data. GNRMS tance from the injection well (labeled N) are zones of
values decreased during processing to values of reduced reliability and thus are discounted as not
approx. 10% for all of the monitor final cross-equal- being significant. In summary, there is clear evidence
ized migrated data volumes. from the time-lapse seismic for CO2 within the upper
Figure 15.12 shows Monitor 3 NRMS difference Deadwood zone, whereas evidence is absent for CO2
maps for three levels of the reservoir. Inspection of residing in either the Black Island or lower Deadwood
corresponding pre-CO2-injection monitoring maps intervals. The latter could be the result of only small
(included in Roach et al., 2017) indicates that the quantities of CO2 residing within these zones,
NRMS maps are most reliable for the upper two a distribution of CO2 that is below the seismic detection
reservoir units (Black Island and upper Deadwood) threshold, and/or the depth-dependent sensitivity and
whereas the lower Deadwood map has some zones of reliability of the data (Roach et al., 2017). 271
Don White
a) Figure 15.12 NRMS amplitude difference values determined within

5441100 the (a) Black Island Formation, (b) upper Deadwood Formation, and (c)
lower Deadwood Formation. Shown are the mean values within a 10-ms
window of the 10-ms NRMS after 3 × 3 smoothing has been applied.
The injection (INJ) and observation (OBS) wells are labeled in a). A 600 m
radius circle is shown for scale. “N” labels in (c) identify anomalies that
are due to time-lapse noise at this level. (Modified from Roach et al.,
5440200 2017.) The high NRMS values toward the edges of the images are an
expected result of the survey geometry and represent an increase in
OBS time-lapse noise due primarily to lower data fold toward the periphery
of the area. Northing and easting coordinates (in metres) are labelled on
the figure axes. (© Her Majesty the Queen in Right of Canada, as
5439300 INJ represented by the Minister of Natural Resources, 2017.)
5438400
639000 639900 640800 641700

b)
5441100
5440200
5439300
5438400
639000 639900 640800 641700

c)
5441100
5440200
5439300
5438400
639000 639900 640800 641700

6 8 10 12 14 16 18 20%
272
As the cumulative amount of injected CO2 O’Brien et al., 2004). VSP also allows the ability to
increases, the associated time-lapse seismic changes directly tie reflections to geological horizons at the
will more generally exceed detection thresholds and wellbore, to verify velocity–depth profiles using
thus should provide a more definitive picture of how zero-offset shot direct arrivals, to estimate the down-
the CO2 is distributed within and among the various going wavelet, and reduce ghosting effects (Hardage,
reservoir units. Further assessment of the existing time- 2000; Kuzmiski et al., 2009).
lapse results can be made by considering the partition- Results for the VSP-Monitor 1 survey (Figure
ing of the 36 ktonnes of CO2 according to the spinner 15.13) are compared to the corresponding 3D surface
log results (see the section “Logging”). The time-lapse results for a horizon within the lower part of the
seismic results can be interpreted in the following man- reservoir (upper Deadwood Formation). Near the
ner. CO2 is undetected in the Black Island unit where center of the images in the vicinity of the CO2 injec-
approx. 4 ktonnes (10% of total) of CO2 has been tion well, a clear NRMS amplitude difference is
injected. This amount is clearly below the seismic detec- observed in both of the images where the amplitude
tion threshold. In contrast, the pronounced time-lapse difference is well above the background noise level.
anomaly observed in the upper Deadwood unit corre- This zone likely represents the CO2 plume where the
sponds to approx. 16 ktonnes (45% of total) of CO2. product of CO2 saturation and plume thickness
Minimum CO2 plume thickness estimates for this zone achieve the threshold level required to produce seis-
based on 1D seismic modeling are 4–10 m for assumed mic amplitude differences. 3D seismic modeling of
CO2 saturations of 50–100% (Roach et al., 2017). CO2 flow simulation results for 27 ktonnes of CO2
The lower Deadwood unit has similarly received injected have been conducted by Harris et al. (2016).
approx. 16 ktonnes of CO2, but no significant time- The seismic modeling results show that a 10 m thick
lapse difference is observed. The absence of an observed zone in the upper Deadwood with CO2 saturations of
time-lapse difference in the lower Deadwood interval is 50% or more produces NRMS values exceeding 1.0 in
likely due to a combination of lower seismic repeatabil- this zone of the corresponding simulated seismic
ity in this zone and CO2 injection occurring over volume. Note that the amplitude of the NRMS anom-
a larger depth interval. aly is much higher in the VSP image than the surface-
seismic image owing to the higher resolution of the
4D Vertical Seismic Profile VSP. However, the repeatability of the VSP is lower
than the surface seismic largely due to the lower SNR
3D vertical seismic profile (VSP) data were acquired
of the 2013 DAS system (cf. Figure 15.6).
at the same times as the November 2013 and
The seismically imaged CO2 plume is asymmetric
the February 2016 3D surface surveys (Harris et al.,
about the injection well and extends 200 m north-
2016, 2017). The number of dynamite shots was
ward. The observed asymmetry in the CO2 plume at
increased for the VSP surveys to 680 as compared
this level contrasts with early flow simulations
to 260 for the surface-only 3D surveys (Figure 15.11).
(Jiang et al., 2016) and will require adjustment of the
For the VSP data, the November 2013 survey con-
permeability characteristics of the existing geological
stitutes the Baseline, and the February 2016 survey is
model. The VSP image suggests that the CO2 plume
the Monitor 1 survey. These correspond to the
may reach the observation well (in contrast to the
Monitor 2 and Monitor 3 3D surface seismic surveys
surface image) within the upper Deadwood interval,
of Roach et al. (2017). The VSP survey parameters
but with lower saturation based on the strength of
were designed to reliably image a subsurface CO2
the anomaly. PND logs acquired at the time of
plume residing within a radius of approx. 500 m
the February monitor survey suggest that a thin zone
from the injection well. The VSP data were acquired
of CO2 had reached the observation well within the
with a DAS system using a casing-conveyed fiber
upper Deadwood interval.
optic cable permanently cemented in place during
well completion (Harris et al., 2016). 4D VSP mon-
itoring has the same general objectives as the 4D Passive Seismic Monitoring
surface seismic but it is designed to provide higher Passive seismic monitoring has been ongoing at the
resolution near-borehole images of the reservoir Aquistore site since August, 2012. This provides
zone during the initial stages of injection (e.g., a record of almost three years of local background 273
Don White
0.7 0.2
0.6
0.175
0.5
0.15
0.4 Observation Well Observation Well

nRMS
Injection Well Injection Well 0.125

0.3
0.1
0.2
0.1 0.075
0 500 m 0 500 m
0 0.05
Figure 15.13 Comparison of NRMS difference maps (fractional rather than percentage) for (a) 3D VSP and (b) surface 3D for a horizon within
the upper Deadwood formation. (Modified from Harris et al., 2017.) The dashed circle (for scale) has a radius of 150 m. The high NRMS
amplitude differences observed toward the edges of the VSP image represent an increase in time-lapse noise toward the periphery of the
imaging area. (© Her Majesty the Queen in Right of Canada, as represented by the Minister of Natural Resources, 2017.)
seismicity in advance of CO2 injection that started semicontinuously at internal temperatures of 115°C.
in April, 2015 and continuous monitoring since then Near-surface orientation shots (1 kg dynamite) were
(Stork et al., 2016; White et al., 2017). Passive mon- detonated to determine the orientation of the geo-
itoring has been conducted using a number of differ- phones in the well. The downhole monitoring was
ent systems each tailored for distinct purposes. First, intended to extend the magnitude threshold for
three telemetering broadband surface seismographs event detection to Mw > –3 so that a complete record
(Guralp CMG-40 T 0.1–50 Hz seismometers) are of microseismic activity could be achieved for com-
deployed within approx. 1 km of the injection well parison with geomechanical modeling and reservoir
(Figure 15.11). These stations are designed to provide response to injection.
near real-time event detection capability for either Event detection methods including direct inspec-
microseismic activity (Mw > 0–1) or potentially larger tion, application of short-term average/long-term
magnitude induced seismic events. Second, are two average (STA/LTA) detection algorithms, and multi-
orthogonal lines (L1 and L2 in Figure 15.11, 2.5 km channel stacking methods have variably been applied
each) comprising 65 10-Hz geophones (1C and 3C in to the data from each of the monitoring systems item-
Figure 15.11) buried at 20 m depth. Data from these ized earlier. Although many teleseismic, regional, and
stations are retrieved once per month. This continu- local industrial seismic events have been detected
ously monitoring array is designed to provide during four years of recording, no significant injec-
a complete record of local seismicity for magnitudes tion-related seismicity (Mw > –1) was detected during
of Mw > –1. These data are amenable to multichannel the first 17 months of CO2 injection, nor smaller
data enhancement methods improving the overall magnitude seismicity (Mw > –3) during the first eight
sensitivity of the monitoring system. Third, months when the downhole array was active. This
a downhole geophone array was deployed during the result is in stark contrast to some other CO2 injection
first eight months of CO2 injection (Nixon et al., projects where swarms of microseismic events are
2017). Five 3C (15-Hz) clamping geophones were detected (e.g., Decatur; Kaven et al., 2015), but is not
deployed in the observation well (Obs in unlike the low levels of seismicity associated with injec-
Figure 15.11) over a depth of 2850–2910 m (that is tion activities at the nearby Weyburn EOR field
274 approx. 250 m above the reservoir) and operated (Verdon et al., 2016).
the ability to detect or image a zone containing less

Conclusions than approx. 16 ktonnes in the reservoir at greater
A broad spectrum of geophysical methods has been than 3000 m depth and to distinguish the interval
employed at the Aquistore CO2 storage site for the within which this CO2 resides.
purposes of site characterization, baseline definition, The nonseismic surface measurement techniques
and monitoring. 3D seismic data interpreted in con- employed at the Aquistore site have resolving power
junction with well logs have been used to define the that is generally inversely proportional to the depth of
architecture of the CO2 storage container and assess the resident CO2. That is, the deeper the CO2, the larger
its suitability for long-term storage. These data show the injected quantity required to produce a discernible
that over an area of 30 km2 the reservoir zone and signal at the surface. This limits the effectiveness of
caprock are generally continuous and devoid of major these methods for plume monitoring at the Aquistore
structures. A basement fault is identified with site where the reservoir (at 3150 m depth) is very deep
a flexure in the overlying Cambrian to Silurian strata, as compared to most CO2 storage sites. Very large
including the reservoir and caprock. These units do quantities of CO2 within the reservoir are required for
not appear to be ruptured or faulted. Evidence is the plume to be monitored using either surface-based
absent of any vertical faults that extend above the electromagnetic methods or gravity based on a com-
Silurian section (or approx. 2700 m depth) where parison of modeling results and background noise
the regional evaporitic secondary seal (Prairie levels measured at the Aquistore site. For example,
Formation at 2500 m) is at least 150 m thick and has CSEM modeling of a CO2-saturated, 240 m thick 50
no salt dissolution features. The subvertical basement Ωm reservoir interval estimates that the plume
fault is oriented nearly orthogonal to the regional would have to be approx. 3–5 km in diameter (or
maximum horizontal stress making reactivation dur- 200–600 Mtonnes for 5% porosity) to produce
ing CO2 injection unlikely. Naturally occurring seis- a detectable response. The low sensitivity of these
micity in the area is very low. The nearest seismogenic methods makes them better suited to leakage detec-
fault zone of any significance is located approx. tion of CO2 that has migrated to shallower depths.
200 km away. For example, 1.5 Mtonnes of CO2 at 600 m depth
Baseline measurements made for many of the would produce a gravity anomaly of approx. 60 μGal
geophysical techniques have identified the natural which significantly exceeds the ±20 μGal background
background noise levels at the site. These are com- noise signal at the Aquistore site.
pared with modeled CO2-related signals to assess Background surface deformation measured prior
whether CO2-produced signals will exceed or be to the injection of CO2 includes regional vertical sub-
distinguished from background time-lapse noise. sidence of –0.2 cm/year, local vertical deformation of
This is considered in the context of either outlining up to 1.0 ± 0.3 cm/year, and background tilt noise of
the CO2 plume within the reservoir or detecting CO2 up to 2.0 μrad. These values compare with modeled
that may have migrated to shallower levels. In all maximum vertical deformation of < 2 cm and tilt of
cases, to varying degrees, the ability of the various approx. 10 μrad for CO2 injection at a rate of approx.
geophysical monitoring methods to detect CO2 is 0.5 Mtonnes/year. The maximum annual injection
dependent on the amount and depth of the CO2. rate at the Aquistore site achieved in the first 20
Time-lapse seismic methods are shown to provide months of operation was approx. 0.1 Mtonnes/year,
the most detailed maps of the CO2 plume within the and thus any associated surface uplift due to deep
reservoir. This is due to the inherent resolving power injection during this period is likely below the detec-
of the technique and the good to excellent repeatabil- tion threshold.
ity at the Aquistore site. Repeatability of 3D surface New or developing geophysical methods have
seismic was excellent (NRMS of approx. 0.10) and was been tested at the Aquistore site with mixed results.
good (approx. 0.30) for 3D DAS-based VSP. The 3D DAS-based VSP surveys have proven to be suitable for
DAS-VSP repeatability is expected to improve over time-lapse imaging of the CO2 plume. As compared to
time, as the technology has improved significantly VSP acquired using conventional wire-line geo-
since the baseline acquisition in 2013. The first time- phones, DAS provided higher spatial sampling and
lapse seismic monitoring results have demonstrated a larger recording aperture, but generally lower SNR. 275
Don White
However, the achievable quality of DAS data should No injection-related seismicity (Mw > –1) was detected
improve significantly for future surveys as a result of during the first 17 months of CO2 injection using
continued technological improvements since 2013 a surface based array, nor was any smaller magnitude
that significantly increase SNR and broaden the fre- seismicity (Mw > –3) during the first eight months
quency bandwidth. Downhole borehole gravity mea- when a downhole array was operating within 250 m
surements demonstrated resolution of less than 1 of the reservoir. This result is consistent with the low
μGal and a repeatability of 10 μGal. Although these levels of seismicity associated with injection activities at
values are comparable to surface-based gravity mea- the nearby Weyburn EOR field.
surements, the proximity of the measurements to the
CO2-associated density anomalies and large distance Acknowledgments
from near-surface seasonal effects increases both the
The seismic interpretation presented here was based
repeatability and the expected CO2-related signal.
partly on an internal report provided by RPS Boyd
Controlled-source borehole-to-surface EM data
Petrosearch. The Aquistore CO2 Storage Project is
acquired at the Aquistore site compared to the mod-
managed by the Petroleum Technology Research
eled surface DC response for a 30 m thick, 150 m
Centre, Regina, Canada. Funding for this research was
radius CO2 plume in the reservoir suggest that detec-
provided in part by Natural Resources Canada,
tion is possible. This provides a detection level that
Sustainable Development Technology Canada,
exceeds that of surface-based EM methods by two to
Saskatchewan Go Green Fund, Enbridge, SaskPower,
three orders of magnitude. Seismic reflection inter-
Schlumberger Carbon Services, Korea National Oil
ferometry applied to passive data from the Aquistore
Corporation, SaskEnergy, and Research Institute of
site produced images of the subsurface having mod-
Innovation Technology for the Earth. Contributions
erate correlation values (approx. 0.5–0.65) at depths
to the work described in this chapter were made
of 900–2200 m (0.8–1.5 s) with well-based synthetic
by M. Craymer, S. Samsonov, J. Craven, J. Silliker,
seismograms. However, the image quality was poor at
B. Roberts, M. Czarnogorska, J. Henton, I. Ferguson,
the reservoir depth (3200 m), indicating that substan-
J. McCleod, K. Harris, L. Roach, C. Samson,
tial improvement in the repeatability of the images is
S. Cheraghi, A. Stork, D. Schmitt, C. Nixon, K. Worth,
required before this methodology can be used for
A. Leniuk, T. Daley, M. Robertson, D. Miller,
time-lapse imaging.
B. Freifeld, J. Cocker, I. Marsden, R. Chalaturnyk,
The first post-CO2 injection 3D time-lapse seismic
G. Zambrano, C. Hawkes, B. Rostron, and
monitoring data were acquired in February, 2016.
G. Klappstein.
At this time, a total of 36 ktonnes of CO2 had been
injected. Zones of CO2 within the upper Deadwood
Formation were imaged in both the 3D surface and
VSP time-lapse data volumes. It is estimated that the
References
imaged plume corresponds to approx. 16 ktonnes of Alnes, H., Eiken, O., and Stenvold, T. (2008). Monitoring
CO2 within this zone. The seismically imaged CO2 gas production and CO2 injection at the Sleipner field
using time-lapse gravimetry. Geophysics, 73(6):
plume (representing a minimum size) extends for
155–161.
a distance of approx. 200 m from the injection well
and is consistent with a relatively thin zone (4–10 m) Bachu, S., and Hitchon, B. (1996). Regional-scale flow of
formation waters in the Williston Basin. AAPG Bulletin,
for assumed CO2 saturations of 50–100%. The imaged
80(2): 248–264.
plume is asymmetric about the injection well with
apparent preferential flow up-dip to the NNW and Black, A., Hare, J., and MacQueen, J. (2016). Borehole
gravity monitoring in the Aquistore Boundary Dam
subparallel with structural fabric in the underlying CO2 sequestration well. Expanded Abstract, Society of
Precambrian basement. This is in contrast to initial Exploration Geophysicists.
reservoir simulation models, which are characterized
Boullenger, B., Verdel, A., Paap, B., Thorbecke, J., and
by CO2 distributions that generally show axial sym- Draganov, D. (2015). Studying CO2 storage with
metry about the injection well. During four years ambient-noise seismic interferometry: A combined
of passive seismic recording at the Aquistore site, numerical feasibility study and field-data example for
276 no significant local seismicity has been detected. Ketzin, Germany. Geophysics, 80(1): Q1–Q13.
Brady, J. L., Hare, J. L., Ferguson, J. F., et al. (2006). Results Gassenmeier, M., Sens-Schönfelder, C., Delatre, M., and
of the world’s first 4D microgravity surveillance of Korn, M. (2015). Monitoring of environmental
a waterflood – Prudhoe Bay, Alaska. SPE Paper 101762, influences on seismic velocity at the geological storage
2006 SPE Annual Technical Conference and Exhibition. site for CO2 in Ketzin (Germany) with ambient seismic
Brunskill, B. (2004). CO2 disposal potential in the deep noise. Geophysical Journal International, 200: 524–533.
subsurface of southeast Saskatchewan. Internal report, Gowan, E. J., Ferguson, I. J., Jones, A. G., and Craven, J. A.
prepared for the University of Regina and (2009). Geoelectric structure of the northeastern
Saskatchewan Power Corporation. Williston basin and underlying Precambrian
Cheraghi, S., White, D. J., Draganov, D., Bellefleur, G., lithosphere. Canadian Journal of Earth Sciences, 46:
Craven, J. A., and Roberts, B. (2017). Passive seismic 441–464.
reflection interferometry: A case study from the Hardage, B. A. (2000). Vertical seismic profiling: Principles.
Aquistore CO2 storage site, Saskatchewan, Canada. Oxford: Elsevier Science.
Geophysics, 82(3): B79–B93. Harris, K., White, D., Samson, C., Daley, T., and Miller, D.
Craymer, M., Henton, J., Piraszewski, M., and Lapelle, E. (2015). Evaluation of distributed acoustic sensing for
(2011). An updated velocity field for Canada. Eos 3D time-lapse VSP monitoring of the Aquistore CO2
Transactions, AGU, 92(51), Fall Meeting Supplement, storage site, GeoConvention, May 4–8, Calgary,
Abstract G21A-0793. Expanded Abstract.
Craymer, M., White, D., Piraszewksi, M., Zhao, Y., Harris, K., White, D., Melanson, D., Samson, C. and Daley, T.
Henton, J., Silliker, J., and Samsonov, S. (2015). First (2016). Feasibility of time-lapse VSP monitoring at the
results of geodetic deformation monitoring after Aquistore CO2 storage site using a distributed acoustic
commencement of CO2 injection at the Aquistore sensing system. International Journal of Greenhouse Gas
underground CO2 storage site. Eos Transactions, Control, 50: 248–260. http://dx.doi.org/10.1016/j
American Geophysical Fall Meeting, Paper G33A-1132. .ijggc.2016.04.016.
Czarnogorska, M., Samsonov, S., and White, D. (2016). Harris, K., White, D., Samson, C. (2017). 4D VSP
Airborne and spaceborne remote sensing for Aquistore monitoring at the Aquistore CO2 storage site.
carbon capture and storage site characterization. Geophysics , 82(6), M81–M96.
Canadian Journal of Remote Sensing, 42(3): 274–290. Hibbs, A. D. (2015). Test of a new BSEM configuration at
DOI:10.1080/07038992.2016.1171131. Aquistore, and its application to mapping injected CO2.
Daley, T. M., Smithy, J. T., Beyer, J. H., and LaBrecque, D. In K. F. Gerdes (ed.), Carbon dioxide capture for storage
(2015). Borehole EM monitoring at Aquistore with in deep geological formations: Results from the CO2
a downhole source. In K. F. Gerdes (ed.), Carbon Capture Project, Vol. 4. Thatcham, Berks, UK: CPL
dioxide capture for storage in deep geological formations: Press, 759–776.
Results from the CO2 Capture Project, Vol. 4. Thatcham, Jiang, T., Pekot, L. J., Jin, L., et al. (2016). Numerical
Berks, UK: CPL Press, 733–758. modelling of the Aquistore CO2 Project: GHGT-13
Dodds, K., Krahenbuhl, R., Reitz, A., Li, Y., and Hovorka, S. Proceedings. Energy Procedia, 114.
(2015). Evaluating time-lapse borehole gravity for CO2 Kaven, J. O., Hickman, S. H., McGarr, A. F., and
plume detection at SECARB Cranfield. In K. F. Gerdes Ellsworth, W. L. (2015). Surface monitoring of
(ed.), Carbon dioxide capture for storage in deep microseismicity at the Decatur, Illinois, CO2
geological formations: Results from the CO2 Capture sequestration demonstration site. Seismological
Project, Vol. 4. Thatcham, Berks, UK: CPL Press, Research Letters, 86(4): 1–6. DOI:10.1785é0220150062.
651–664.
Kent, D. M., and Christopher, J. E. (1994). Geological
Energy and Environmental Research Center. (2014). Geologic history of the Williston Basin and Sweetgrass River
modeling and simulation report for the Aquistore Arch. In G. D. Mossop and I. Shetsen (comp.),
Project: University of North Dakota & U.S. Department Geological atlas of the Western Canada sedimentary
of Energy National Energy Technology Laboratory. basin. Canadian Society of Petroleum Geologists and
Gasperikova, E., and Hoversten, G. M. (2006). A feasibility Alberta Research Council, 421–430.
study of nonseismic geophysical methods for Khan, D. K., and Rostron, B. J. (2005). Regional
monitoring geologic CO2 sequestration. Leading Edge, hydrogeological investigation around the IEAWeyburn
25, 1282–1288. DOI:10.1190/1.2360621. CO2 Monitoring and Storage Project Site.
Gasperikova E., and Hoversten, G. M. (2008). Gravity In E. S. Rubin, D. W. Keith, and C. F. Gilboy (eds.),
monitoring of CO2 movement during sequestration: Proceedings of the 7th International Conference on
Model studies. Geophysics, 73(6): WA105–WA112. Greenhouse Gas Control Technologies, September 5–9,
277
Don White
2004, Vancouver, Canada, Vol. 1: Peer Reviewed Papers O’Brien, J., Kilbride, F., and Lim, F. (2004). Time-lapse VSP
and Overviews, 741–750. reservoir monitoring. Geophysics, 23(11), 1178–1184.
Kuzmiski, L., Charters, B., and Galbraith, M. (2009). Palombi, D. D. (2008). Regional hydrogeological
Processing considerations for 3D VSP: CSEG Recorder, characterization of the northeastern margin in the
34(4). https://csegrecorder.com/articles/view/processing- Williston Basin. M.Sc. thesis, University of Alberta,
considerations-for-3d-vsp Edmonton, Alberta.
Liard, J., Huang, J., Silliker, J., Jobin, D., Wand, S. and Roach, L. A. N., White, D. J., and Roberts, B. (2015).
Doherty, A. (2011). Detecting groundwater storage Assessment of 4D seismic repeatability and CO2
change using micro-gravity survey in Waterloo detection limits using a sparse permanent land array at
Moraine, Proceeding. Geohydro. the Aquistore CO2 storage site. Geophysics, 80(2):
Marsh, A., and Love, M. (2014). Middle Devonian prairie WA1–WA13.
evaporite: Isopach map, map 132. In Regional Roach, L. A. N., White, D. J., Roberts, B., and Angus, D.
Stratigraphic Framework of the Phanerozoic in (2017). Initial 4D seismic results after CO2 injection
Saskatchewan, Saskatchewan Phanerozoic Fluids and start-up at the Aquistore storage site. Geophysics, 82(3).
Petroleum Systems Project, Saskatchewan Ministry of http://dx.doi.org/10.1190/geo2016-0488.1
the Economy, Saskatchewan Geological Survey, Open
Rostron, B., White, D., Hawkes, C., and Chalaturnyk, R.
File 2014–1, set of 156 maps.
(2014). Characterization of the Aquistore CO2 Project
Mathias, S., Hardisty, P., Trudell, M., and Zimmerman, R. storage site, Saskatchewan, Canada. In 12th
(2009). Approximate solutions for pressure buildup International Conference on Greenhouse Gas Control
during CO2injection in brine aquifers. Transport in Technologies, GHGT-12. Energy Procedia, 63:
Porous Media, 79: 265–284. 2977–2984, DOI:10.1016/j.egypro.2014.11.320.
McLeod, J. (2016). Magnetotelluric and controlled-source RPS Boyd PetroSearch. (2011). Interpretation of 2D seismic
electromagnetic pre-injection study of Aquistore CO2 data for the Aquistore Project near Estevan,
sequestration site, near Estevan, Saskatchewan, Saskatchewan: Internal Report for Petroleum
Canada. M.Sc. thesis, University of Manitoba. Technology Research Centre, Regina, 54.
McLeod, J., Craven, J. A., Ferguson, I. J., Roberts, B. J., Rutqvist, J., Vasco, D., and Myer, L. (2010). Coupled
Bancroft, B., and Liveda, T (2014). Overview of the 2013 reservoir geomechanical analysis of CO2 injection and
baseline magnetotelluric and controlled-source ground deformation at In Salah, Algeria. International
electromagnetic geophysical study of CO2 Journal of Greenhouse Gas Control, 4(2): 225–230.
sequestration at the Aquistore site near Estevan,
Saskatchewan. Geological Survey of Canada, Open File Samsonov, S., Czarnogorska, M., and White, D. (2015).
7617. DOI:10.4095/293921. Satellite interferometry for high-precision detection of
ground deformation at a carbon dioxide storage site.
McLeod, J., Craven, J. A., Ferguson, I. J., and Roberts, B. J. International Journal of Greenhouse Gas Control, 42:
(2016). Overview of the 2013 and 2014 baseline 188–199, DOI:10.1016/j.ijggc.2015.07.034.
magnetotelluric and controlled-source electromagnetic
studies of CO2 sequestration at the Aquistore site near Sato, K., Mito, S., Horie, T., et al. (2011). Monitoring and
Estevan, Saskatchewan. Geological Survey of Canada, simulation studies for assessing macro- and meso-scale
Open File 8101. migration of CO2 sequestered in an onshore aquifer:
Experiences from the Nagaoka pilot site, Japan.
Mestayer, J., Cox, B., Wills, P. (2011). Field trials of International Journal of Greenhouse Gas Control, 5:
distributed acousticsensing for geophysical 125–137, DOI:10.1016/j.ijggc.2010.03.003.
monitoring. In 81st Annual International Meeting of
the Society of Exploration Geophysics, Expanded Schenkel, C. J., and Morrison, H. F. (1994). Electrical
Abstract, 4253–4257. resistivity measurement through metal casing.
Geophysics, 59(7): 1072–1082.
Miller, D. E., Daley, T. M., White, D., et al. (2016).
Simultaneous acquisition of distributed acoustic Sherlock, D. A., Toomey, A., Hoversten, M.,
sensing VSP with multi-mode and singlemode fibre Gasperikova, E., and Dodds, K. (2006). Gravity
optic cables and 3 C-geophones at the Aquistore CO2 monitoring of CO2 storage in a depleted gas field:
storage site. CSEG Recorder, 28–33. A sensitivity study. Exploration Geophysics, 37: 37–43.
Nixon, C. G., Schmitt, D. R., Kofman, R. S., et al. (2017). Stork, A. L, Nixon, C., Schmitt, D. R., White, D. J.,
Experiences in deep downhole digital micro-seismic Kendall, J.-M., and Worth, K. (2016). The seismic
monitoring near 3 km at the PTRC Aquistore CO2 response at the Aquistore CO2 injection project,
Sequestration Project, 2017 Geoconvention, Expanded Saskatchewan, Canada. Seismological Research Letters,
278 Abstract. 87(2B): 477. DOI:10.1785/0220160046.
Streich, R. (2016). Controlled-source electromagnetic White, D., Roach, L. A. N., Roberts, B., and Daley, T. M.
approaches for hydrocarbon exploration and (2014). Initial results from seismic monitoring at the
monitoring on land. Surveys in Geophysics, 37: 47–80. Aquistore CO2 storage site, Saskatchewan, Canada.
Vasco, D., Rucci, A., Ferretti, A., et al. (2010). Satellite-based Energy Procedia, 63: 4418–4423. DOI:10.1016/j.
measurements of surface deformation revealfluid flow egypro.2014.11.477.
associated with the geological storage of carbon dioxide. White, D. J., Roach, L. A. N., and Roberts, B. (2015). Time-
Geophysics Research Letters, 37(3): L03303, 1–5. lapse seismic performance of a sparse permanent array:
Vasco, D. W., Ferretti, A., Rucci, A., et al. (2016). Geodetic Experience from the Aquistore CO2 storage site.
monitoring of the geological storage of greenhouse gas Geophysics, 80(2): WA35–WA48.
emissions. Submitted to AGU books. White, D. J., Hawkes, C. D., and Rostron, B. J. (2016).
Verdon, J. P., Kendall, J.-M., Horleston, A. C., and Stork, A. Geological characterization of the Aquistore CO2
(2016). Subsurface fluid injection and induced storage site from 3D seismic data. International Journal
seismicity in southeast Saskatchewan. International of Greenhouse Gas Control, 54(1), 330–344.
Journal of Greenhouse Gas Control (in press). White, D., Harris, K., Roach, L., et al. (2017). Monitoring
Vigrass, L., Jessop, A., and Brunskill, B. (2007). Regina results after 36 ktonnes of deep CO2 injection at the
Geothermal Project. In Summary of Investigations 2007, Aquistore CO2 storage site, Saskatchewan, Canada.
Vol. 1, Saskatchewan Geological Survey, Saskatchewan In 13th International Conference on Greenhouse
Industry Resources, Misc. Rep. 2007–4.1, CD-ROM, Gas Control Technologies, GHGT-13. Energy
Paper A-2. Procedia, 114.
Vozoff, K. (1991). The magnetotelluric method. Whittaker, S., and Worth, K. (2011). Aquistore: A fully
In M. N. Nabighian, (ed.), Electromagnetic methods in integrated demonstration of the capture,
applied geophysics, Vol. 2: Applications. Tulsa, OK: transportation and geologic storage of CO2. Energy
Society of Exploration Geophysicists, 641–711. Procedia, 4: 5607–5614.
Wang, Y. (2015). Design, deployment, performance and Whittaker, S., Rostron, B., Khan, D., et al. (2004). Theme 1:
baseline data assessment of surface tiltmeter array Geological characterization. In M. Wilson and
technology in Aquistore geologic CO2 storage project. M. M. Monea (eds.), IEA GHG Weyburn CO2 Monitoring
Sc. thesis, University of Alberta. and Storage Project summary report 2000–2004, from
the Proceedings of the 7th International Conference on
Washbourne, J. K., Rector, J. W., and Bube, K. P. (2002). Greenhouse Gas Control Technologies. Regina:
Crosswell traveltime tomography in three dimensions. Petroleum Technology Research Centre, 1–72.
Geophysics, 67(3) (May–June): 853–871, DOI 10.1190/
1.1484529. Worth, K., White, D., Chalaturnyk, R., et al. (2014).
Aquistore Project measurement, monitoring and
White, D. (2012). Geophysical monitoring. In B. Hitchon verification: From concept to CO2 injection. In 12th
(ed.), Best practices for validating CO2 geological storage. International Conference on Greenhouse Gas Control
Sherwood Park, AB: Geoscience Publishing, 155–210. Technologies, GHGT-12. Energy Procedia, 63:
White, D. J. (2013a). Seismic characterization and 3202–3208. DOI:10.1016/j.egypro.2014.11.345.
time-lapse imaging during seven years of CO2 flood in Xu, Z., Fang, Y., Scheibe, T., and Bonneville, A. (2012a).
the Weyburn Field, Saskatchewan, Canada. A fluid pressure and deformation analysis for
International Journal of Greenhouse Gas Control, 16S: geological sequestration of carbon dioxide.
S78–S94. Computational Geosciences, 46: 31–37.
White, D. J. (2013b). Toward quantitative CO2 storage Xu, Z., Juhlin, C., Gudmundsson, O., et al. (2012b).
estimates from time-lapse 3D seismic travel times: Reconstruction of subsurface structure from
An example from the IEA GHG Weyburn–Midale CO2 ambient seismic noise: An example from Ketzin,
monitoring and storage project. International Journal of Germany. Geophysical Journal International, 189:
Greenhouse Gas Control, 16S: S95–S102. 1085–1102.
279
Chapter
Development and Analysis of a Geostatic
16 Model for Shallow CO2 Injection at the Field

Research Station, Southern Alberta, Canada
Donald C. Lawton, Jessica Dongas, Kirk Osadetz, Amin Saeedfar, and
Marie Macquet
Introduction At the FRS, small volumes (<600 tonnes/year)

will be injected into the Basal Belly River Sandstone
Storage of carbon dioxide (CO2) in geological forma-
Formation at a depth of 296 m below the ground
tions has become increasingly well recognized as
surface, designed to simulate CO2 leakage from
a key, viable long-term method to reduce emissions
a deeper, large-scale CO2 storage site into the over-
of CO2 into the atmosphere. Carbon capture and
burden. Currently Alberta legislation and regula-
storage (CCS) is a strategy that is being demonstrated
tions permit geological sequestration of CO2 only
in western Canada by the Quest CCS Project devel-
at depths greater than 1000 m below the ground
oped by Shell and partners (Rock et al., 2015) and the
surface, where CO2 is a supercritical fluid. In the
Aquistore Project in Saskatchewan (Rostron, 2014).
unlikely event that CO2 migrates above the storage
The fate of an injected CO2 volume is ultimately site-
complex due to a failure of containment and con-
specific, and is determined by the dominant trapping
formance, it will be most easily detected at shal-
mechanisms present. The geochemical and geome-
lower depths in a gas phase, prior to its contact
chanical properties of the reservoir and seal intervals
with the groundwater protection zone. To address
govern how the reactive CO2 volume will behave.
these challenges, the FRS has been established to
The Containment and Monitoring Institute
undertake research into the efficacy and evaluation
(CaMI) of CMC Research Institutes Inc., in conjunc-
of monitoring technologies at shallow to intermedi-
tion with the University of Calgary, has developed
ate depth in a realistic field setting. The facility will
a comprehensive Field Research Station (FRS) in south-
be used to test new measurement, monitoring, and
ern Alberta, Canada (Figure 16.1), to facilitate and accel-
verification (MMV) technologies as they are devel-
erate research and development leading to improved
oped and commercialized (e.g., fiber optic devices,
understandings and technologies for geological contain-
slim wells, new analytical instruments for air and
ment and secure storage of CO2. Of particular impor-
water analyses) with an emphasis on integrated
tance is the ability to detect vertical migration of CO2
geophysical, geochemical, and geodetic surveys as
out of the storage complex and a change in phase of the
well as new approaches to the integration of
CO2 from liquid to gas as it enters the shallow subsur-
volume-based geophysical data sets (e.g., 3D seismic
face. Key goals for the FRS program are to develop and
volumes) with high-resolution point measurements
refine monitoring technologies to determine the detec-
(wells). This chapter provides an overview of the
tion threshold of gas-phase CO2 in the subsurface and to
general development of a geostatic model for CO2
develop geophysical and other monitoring technologies
injection at the FRS and outcomes of reservoir
for optimizing CO2 enhanced oil recovery (CO2-EOR)
simulation of CO2 injection using the physical prop-
and to de-risk CCS in general. New approaches and
erties of the geomodel as input. A particular focus
innovative technologies need to be developed for
of the chapter is on numerical modeling of the
sampling, measurement, and monitoring methodolo-
time-lapse seismic response of CO2 injection over
gies in order to provide comprehensive models of the
five years of injection, followed by five years of
subsurface and to identify early loss of containment or
postclosure seismic monitoring.
280 conformance of the injected CO2.
Development and Analysis of Shallow CO2 Injection
Figure 16.1 Map of Alberta, Canada showing

the location of the Containment and Monitoring
Institute Field Research Station (FRS).
Project Contribution site with surface access provided through a 10-year

lease from Cenovus Energy (now transfered to
This chapter presents a case history demonstrating
Torxen). The site is currently in its final construction
a methodological workflow to incorporate integrated
stage and both the initial injection well (10-22-17-
data sets to characterize the reservoir and caprock
16W4) and two 350 m deep monitoring wells and
sequence for a CO2 injection and monitoring experi-
three water wells all less than 90 m deep have been
ment. The work provides baseline characterization
completed. Research activities at the FRS will be under-
and automatic workflows that have been established
taken for at least 10 years.
to enable the static model to be updated and refined
Building a geomodel in a petroleum field that has
on acquisition of newer data sets during the CO2
been previously explored and produced can be highly
injection program. The simulation and modeling
advantageous in providing existing data sets to under-
results will also provide insight into the anticipated
stand the geological and geophysical attributes of
plume behavior, and to guide the implementation of
a reservoir. Such is the case for the QUEST project
monitoring technologies and sampling strategy to be
by Shell and partners in central Alberta, and also for
used for the going monitoring program at the FRS.
enhanced oil recovery (EOR) projects such as the
Teapot Dome project in the United States, and the
Site Selection and Characterization Weyburn-Midale project in southern Saskatchewan,
The FRS has been constructed in Newell County, Canada. A desktop study known as the Wabamun
southwest of Brooks, Alberta, Canada on a 200-hectare Area Sequestration CO2 Project (WASP), located 281
D. C. Lawton, J. Dongas, K. Osadetz, A. Saeedfar, and M. Macquet
west of Edmonton in Central Alberta, was conducted The Cretaceous strata unconformably overly an
in a similar way to this project. In that project, eroded carbonate and evaporate-dominated Paleozoic
a geomodel was built based on 96 wells with existing succession, approx. 1200 m thick, that itself lies uncon-
geophysical logs, seismic data sets, and limited core formably over Precambrian basement rocks (Wright
measurements that penetrate the reservoir in et al., 1994). Regional seismic data indicate minor local
Paleozoic carbonate Nisku Formation (Eisinger and structure due to Paleozoic salt dissolution in the east-
Jensen, 2009). The stratigraphic framework and 3D ernmost part of the study area, but the Phanerozoic
model grid was defined by the wireline logs and seis- sedimentary successions at the CO2 injection site and
mic data. Using petrophysical models of porosity and within the anticipated extent of the CO2 plume are
permeability, the model grid was then populated and locally undeformed. Petroleum is produced locally
upscaled using geostatistical algorithms to define the from Lower Cretaceous sandstones that occur below
spatial distribution of the subsurface properties. Some the horizons of interest for the CO2 injection and
other examples of reservoir characterization studies monitoring program. The first injection well
that have been undertaken for CO2 injection and/or (Countess 10-22-17-16W4) was drilled to a depth of
storage projects include the Sleipner project offshore 550 m to characterize the overburden and the under-
Norway (Chadwick et al., 2004), potential sites in burden within the FRS. This well has subsequently
Korea described by Kim et al. (2014), and in the been completed as an injection well in the BBRS.
Gippsland Basin offshore Australia (Gibson-Poole The three wells drilled at the FRS are completed in
et al., 2008). a typical Upper Cretaceous succession for southeast-
ern Alberta (Glombick, 2010a,b, 2011a,b, 2014a,b;
Glombick and Mumpy, 2014a,b). The site lies at the
Geological Setting eastern edge of the Bow City coalfield north of Bow
The FRS is located in the Countess Oil Field between River and the Countess 10-22 well lies just east of the
the Bow and Red Deer Rivers, in a mixed grassland erosional edge of the Bearpaw Shale and encounters
region on the Alberta Plains (Figure 16.1). This the Lethbridge Coal Zone and Dinosaur Park Fm. at
glaciated landscape is characterized by Laurentide the bedrock surface. The eroded top of Dinosaur Park
(continental) till-covered plains of low topographic Fm. is unconformably overlain by a granitic granule
relief and internal drainage between incised river val- lag at the base of Holocene glacial fluvial sediments
leys where Upper Cretaceous bedrock strata crop out and soils. Shales of the Bearpaw Fm. outcrop on the
(Fenton et al., 1994). The FRS is located in the Interior north side of the Bow River about 5 km south of the
Platform Structural Province, a nondiastrophic FRS. The Dinosaur Park Fm. crops out in Dinosaur
region of predominantly undeformed, unconfor- Provincial Park in the Red Deer River valley northeast
mity-bounded Phanerozoic stratigraphic successions of Brooks.
that are in places locally deformed through dissolu- Belly River Group strata were deposited gradation-
tion of Paleozoic evaporates in underlying strata, dif- ally and conformably on top of the underlying Pakowki
ferential sedimentary compaction, and rare faults in mudstones in shoreface marine and nonmarine
the Precambrian basement. depositional environments (Figure 16.2). The BBRS
At the FRS the shallow stratigraphy is composed injection zone is at a depth of 295 m below ground
of approx. 1000 m of interbedded sandstone and shale- surface and it is a 7-m-thick, upper shoreface, fine to
dominated strata of Cretaceous age, deposited during medium-grained sandstone of poorly to well sorted,
successive regressive and transgressive sequences, asso- angular to subangular grains respectively, at the base
ciated with Laramide orogenic foreland progradation of the Foremost Fm. (Hamblin and Abrahamson,
into the Cretaceous Interior seaway. The stratigraphic 1996). They found significant presence of diagenetic
succession of the Upper Cretaceous strata is shown in clays and calcite cements that reduce reservoir perme-
Figure 16.2. The storage complex lies within the Belly ability and that could provide some local barriers to
River Group, with the injection zone being at the base vertical and lateral flow.
of this unit, in the Brosseau Member of the Foremost The overlying sealing succession of the storage
Formation, also known as the Basal Belly River complex is provided by interbedded heterolithic stra-
282 Sandstone (BBRS). tification of meter-scale interbeds of mudstone, fine-
Figure 16.2 Stratigraphic succession in the CMCRI Countess #1 Well (10-22-17-16W4, kb 784.5 m).
grained sandstone, and uncleated coals within the (Beaton, 2003) suggest that the superior Foremost
McKay coal zone that directly overlies the injection Fm. overlying the injection area has very low perme-
zone (Figure 16.2). Additional seals are provided by ability and that it will act as an effective top seal
the stratigraphically higher Taber Coal zone, of simi- generally, but especially at the FRS, where a 1 m
larly interbedded mudstones, fine sandstones, and thick mudstone bed directly overlies the injection
uncleated coals. Together, the potential top seal has zone in all the wells.
a combined average net thickness of 225 m in the Occasionally natural gas, predominantly methane,
study area. Groundwater and petroleum studies occurs in shallow aquifers, associated with near- 283
complex that were input into a regional geomodel and

416000 420000 424000
subsequently used in the reservoir simulator to model
the CO2 injection program and development of the
CO2 plume.
Two 3D seismic reflection volumes were available
for the study. A 30-km2 regional survey was provided
5592000
5592000
courtesy of Cenovus Energy, and had been acquired in
1997. Figure 16.4 shows an interpreted south–north
10-22-17-16W4
seismic section extracted from this processed seismic
5 km
volume. The shallow reflectors, including the BBRS
event, are reasonably flat, but deeper horizons (approx.
5588000
5588000
750 ms) show evidence of structural collapse due to
dissolution of evaporites, although these structures do
5 km not appear to propagate up-section to the level of the
storage complex. In order to obtain higher resolution
seismic data, particularly for the shallow part of the
5584000
5584000
section, a 1 km × 1 km multicomponent seismic survey

was collected in 2014 by CMC Research Institutes as
part of baseline studies for the FRS. A map showing the
416000 420000 424000 acquisition layout is shown in Figure 16.5, which also
0 1000 2000 3000 4000 5000m includes the layout of wells and other surface monitor-
1:83101 ing technologies that have been deployed.
An orthogonal geometry was used for the seismic pro-
Figure 16.3 Surface map showing locations of wells and seismic gram, with shot lines oriented east–west and receiver
surveys used to develop the geomodel. The red square depicts the lines oriented north–south. The shot and receiver line
5 km × 5 km area of the initial geomodel and the dashed square
depicts the detailed 1 km × 1 km geomodel that was extracted for intervals were nominally 100 m, decreasing to 50 m
reservoir simulation and time-lapse analysis. Colors of the wells are within the inner 500 m × 500 m survey area, in order to
based on weighting distance from the 10-22 well. ensure optimal imaging of the storage complex at
300 m depth. Shot and receiver intervals were both
10 m, yielding a 5 m × 5 m bin size. The sources used
surface coal beds. Some of the coal zones have much
for the survey were two Envirovibes sweeping from 10
higher hydraulic conductivities compared to the clay-
to 150 Hz over 16 s. For development of the geomodel,
rich, fine-grained sandstones and mudstones within
primarily the PP volume was used and a view of the
which they are interbedded. The base of groundwater
processed data cube is displayed in Figure 16.6.
protection occurs at 220 m below ground level and is
In 2015, CMC Research Institutes drilled the
above the base of the surface casing in the injection
Countess 10-22 well and acquired a comprehensive
well, within the overlying seals of the storage complex.
log suite including a dipole sonic log, and also cored
the storage complex. Following the core and log ana-
Geomodel Development lysis and the interpretation of the regional and newly
For the development of the geomodel, legacy wells acquired seismic data, a 5 km × 5 km geomodel was
and an available seismic volume within a 13 km × developed within the area shown in Figure 16.3, in
11 km area around the FRS were examined initially. order to place greater weighting of the petrophysical
This area includes 88 wells and a 3D seismic volume, data on wells local to the FRS. Detailed reservoir and
as shown in Figure 16.3. Most of the legacy wells geophysical modeling was subsequently undertaken
include wireline log suites with gamma ray, caliper, in a 1 km × 1 km area centered on the 10-22 well.
spontaneous potential, compressional sonic, shallow-
deep resistivity/induction, bulk density, density-por-
osity, and neutron-porosity well logs. These well logs Structural and Stratigraphic Framework
were interpreted using geostatistical methods to Within the 5 km × 5 km surface area of the geomodel,
284
obtain general petrophysical parameters of the storage the local structural and stratigraphic framework
Composite line 15 Composite line 13 XLine 77

XLine 12 Composite line 15 Composite line 20 XLine 67 XLine 77 Composite line 12
IL 28 28 28 28 28 28 28 28 28 28
XLine 153
15 29 44 58 72 86 100 115 129 143
XL
−100
−200
BBRS Unit
Pakowki Fm
−300
Milk River Fm
Colorado Gp
−400
Medicine Hat Mb
−500
Second White Specks Fm
Base Fish Scales Unit
−600
Bow Island Fm
−700 Mannville Gp
−800
−900
−1000
Figure 16.4 South–north seismic section from the regional P-wave seismic volume, showing key horizons interpreted and also the well-ties.
The short well is the 10-22 well drilled in the center of the 1 km × 1 km detailed geomodel. The extent of the seismic survey is shown in
Figure 16.3.
determined from well data was confirmed and data were available for this project; thus the TDR was
improved through incorporating the interpretation developed through the completion of a sonic calibra-
of the seismic data sets. A miss-tie step was required tion step. The velocity model for the 2014 local 3D
to remove a 5 ms difference in seismic horizon inter- volume utilized the TDR produced from the 10-22
pretation between the regional and new 1 km × 1 km well tie, and the velocity model for the 1997 regional
seismic volumes. Horizon tops were then verified with volume utilized the TDRs produced from the adjacent
the depth conversion of the two 3D seismic PP well ties. Each model was run separately to calculate
volumes through log-derived velocity modeling. the interval velocities for each subsurface horizon
The two multilayered models were computed for zone below the seismic reflection datum of 800 m
each seismic volume, within which velocity was trea- above sea level.
ted velocity as a linear function of depth over discrete The final framework that defines the stratigraphy
intervals. The interval velocities of each subsurface in the FRS study region combines the formation tops
zone were calculated from the time–depth relation- and interpretations of the seismic horizons.
ship (TDR), a direct output from completing the The weighting regime was tailored to consider a 70%
seismic-to-well tie. Three seismic-to-well ties were weighting on the 1997 regional seismic volume and
completed, using the compressional sonic log, bulk 30% on the 2014 local seismic volume, where the
density log, and zero-phase extracted statistical wave- formation top interpretation guided interpolation 285
lets at each respective well location. No check-shot between well locations. Isopach and isochron maps
1000
Access road
&
y
Surface seismic r ra tic
a p
grid T eo
800 ER ibr
Injector well f
Observation well
Groundwater
well 600
Seismometers
1 km
Classroom
Classroom
Instrument 400
shed Shack
200
3C-3D seismic array
0
0 200 400 600 800 1000
1 km
Figure 16.5 Layout of the Containment and Monitoring Institute Field Research Station, showing the seismic acquisition grid and locations of
wells and monitoring equipment installations.
5589000 420800 Figure 16.6 Processed PP volume from the

5589200 420600
5589400 420400 420200 2014 seismic survey. Base of chair display
5589800
5589600 420000 illustrates the time slice at the top of the Basal
Belly River Formation.
0
0
0.2
0.2
0.4
0.4
Time (s)
0.6
0.6
0.8
0.8
1
1 5589000
420800 5589200
)
420600
(m
5589400
ing
Ea 420400
rth
stin
g( 5589600
No
m) 420200
286 420000 5589800
Table 16.1 Formation tops, two-way seismic travel times, and characterizing the BBRS. The lack of separation in
formation thicknesses
the shallow to deep induction logs (AF10-90) in
Formation TWT (ms) Depth (m) Thickness both the Foremost Fm. and BBRS are generally indi-
top (m) cative of relatively low permeability. Routine core
Foremost — 143 152
analysis and tight rock analysis provided sample-
Fm. based porosity and permeability results that permitted
a well log-to-core calibration of inferred rock proper-
BBRS 245 295 7
member
ties in the reservoir and seal intervals. Once calibrated
and modeled, the BBRS averaged an effective porosity
Pakowki 250 302 61
of 11% and intrinsic permeability of between 0.5 mD
Fm.
and 2 mD.
Milk River 298 363 76 The ELAN mineralogical-derived data was una-
Fm.
vailable at the offset wells, and thus porosity was
Colorado 353 439 43 calculated using a volume of shale (v-shale) model,
Group. where effective porosity is defined as the total porosity
Medicine 373 482 217 less the clay-bound fluid content. Intrinsic permeabil-
Hat Fm. ity is an isotropic property and was calculated using
Second 500 699 88 the free-fluid Timur–Coates equations (Timur, 1968;
White Coates and Dumanoir, 1974). Average effective por-
Speckled osity and intrinsic permeability values of the
Shale Fm. Foremost Fm. could not be provided for the entire
Base Fish 560 787 40 152-m-thick unit owing to heterogeneity; however,
Scales Fm., well logs and core measurements confirm low perme-
Bow Island 600 827 112 ability values of 1.5E-4 mD for the shale caprock
Fm. immediately above the BBRS.
Mannville 658 939
Group
Facies Modeling
The coal zones within the Foremost Fm. are of specific
interest, as they are estimated to be one of the main
were used to determine subsurface formation thick- laterally continuous sealing units above the BBRS.
nesses, defined by the depth or reflection time to the The wireline log character of the coals demonstrated
top of each formation, respectively; the general rela- artificially high neutron and low-density responses,
tionships between seismic reflection travel times and resulting in skewed values of effective porosity and
formation tops are listed in Table 16.1. intrinsic permeability. To our knowledge of the local
FRS study region, the coals are known to be water-
Petrophysical Analysis saturated and have adsorbed volumes of CO2, result-
Two main reservoir properties used to characterize ing in limited relative permeability. No visible cleating
the BBRS and Foremost Fm. were effective porosity systems were observed in the core slabs or routine
and intrinsic permeability. The wireline well logs of core analysis completed.
Countess 10-22 were studied using Schlumberger’s To constrain further the petrophysical calculations,
elemental log analysis (ELAN) that provided calcu- a facies model with a simple lithology legend made to
lated depth profiles of effective porosity, permeability, mimic the ELAN mineralogical-derived data was devel-
and rock fraction composition. The wireline logs in oped to isolate the coal zones from the sandstone, silty-
Countess 10-22 are displayed in Figure 16.7. sand, and shale formations. In the geomodel, where
The sandstone reservoir demonstrates an approxi- a coal interval was defined, low values of porosity and
mate 45 API signature on the gamma ray log, with permeability were assigned (0.03 and 0.001 mD respec-
trailing neutron– (NPSS) and density–porosity tively, from core analysis results). The lithological inter-
(DPSS) logs indicative of the mild clay content pretations are shown in Figure 16.7. 287
COUNTESS 10-22
MD (m) 6 CAL AF90 RHOZ IIite QE VCL BF Helium Porosity Nitrogen Permeability
in 16 0.2 ohm.m 2001.65 g/cc 2.65 BWater_QE 0 % 0.7 0 % 0.3 0 % 40 0.01 mD 100
FACIES
BS AF60 NPSS Quartz_QE PHIE P-Decay Permeability
6 in 16 0.2 ohm.m 200 0.6 p.u. 0 Coal_QE 0 p.u. 0.4 0.01 mD 100
GR AF30 DPSS Calcite_QE PIGN Klinkenberg Permeability
0 API 150 0.2 ohm.m 200 0.6 p.u. 0 UWater_QE 0 p.u. 0.4 0.01 mD 100
AF20 UIWater_QE PHIE (model) K_INT
0.2 ohm.m 200 0 p.u. 0.4 0.01 mD 100
AF10 Porosity - fill KINT_GEO
0.2 ohm.m 200 0.01 mD 100
RXOZ K_INT (model)
0.2 ohm.m 200 0.01 mD 100
Permeability - fill
277.9
Coal
280
282
Silty Sand
284
Foremost Fm
286
Coal
288
290
292
294
Silty Sand
296
Basal BBRS
298
Sand
300
302
Pakowki Fm
304
306
Sand
308
Figure 16.7 Interpreted well logs through the storage complex from the 10-22 well. Dots within the BBRS interval for the porosity and
permeability columns are calibrated values from core measurements.
Geomodel Upscaling permeability well logs to populate the geomodel.

The reservoir properties in each well were upscaled
The geomodel volume is defined by the number of
using an arithmetic mean, and a Gaussian Random
cells in the Cartesian grid, and the layer thicknesses.
Function Simulation (GRFS) algorithm was chosen to
The structural framework defined by the 3D model
populate the full 3D model given the correlation statis-
horizons resulted in 12 subsurface intervals. For the
tics. It is considered to be a conditional and stationary
upscaling, the cell lateral dimensions were 25 m ×
simulation algorithm, incorporating both kriging and
25 m and the cell thickness varied within the model,
unconditional simulation. The conditional portion of
from 0.5 m in the storage complex and under-burden,
the algorithm honors the input data and is able to
to 5 m in the remaining parts of the model.
model the expected variability in property distribution,
The dimensions of the geomodel volume are 197 ×
whereas the unconditional simulation portion incorpo-
198 × 1557, resulting in a total of just over 60 million
rates a degree of randomness with each realization, as it
cells.
does not replicate the data’s mean, variance, or semi-
Variogram analysis was used in conjunction with
288 the calculated effective porosity and intrinsic
variogram outputs.
In summary, the procedure for building the 3D • Using a GFRS algorithm to populate the upscaled
static geomodel included the following: properties into the 3D model grid. Considered as
a conditional simulation algorithm, the function
• Import, management, and quality control of all
honors the input data as well as inserting a degree
data into project (wireline logs, 3D seismic
of randomness in each model realization.
volumes, well locations, deviation surveys, core
measurements).
• Definition of structural framework through
interpretation completed on wireline logs and 3D
Storage Capacity of the BBRS
seismic volumes. Depth to formation tops for the Storage capacity estimates were computed through
upper 200 m of the section (i.e., within the zone of a P10-50-90 framework for the 1 km2 detailed geo-
surface casing) was extrapolated from geological model volume, involving the Foremost Fm., BBRS,
bedrock maps. and underlying Pakowki Fm. These three cases define
• Reservoir characterization through log analysis the conservative, typical, and optimistic estimates
and petrophysical analysis to define the two main of the total pore volume within the BBRS.
reservoir parameters, namely effective porosity Corresponding permeability volumes were computed
and intrinsic permeability. Calculations were by kriging each 10th, 50th, 90th percentile of total
calibrated from core measurements with samples pore volume. Figure 16.8 displays the P50 case for
taken from reservoir and caprock intervals. effective porosity for the 1 km2 area extracted from
the 5 km × 5 km geomodel, used as input for dynamic
• Statistical analyses of the petrophysical data for
effective porosity and intrinsic permeability, modeling and fluid flow simulations.
including normal score transformation and
variograms to define the optimum distribution Reservoir Simulation
and range. Reservoir simulations were undertaken to provide
• Upscaling the model using the following initial predictions of the CO2 plume behavior in the
constraints: reservoir and the caprock strata, based on the pet-
• Gridding the 3D model using a vertical pillar rophysical properties of the geomodel. Simulations
method. The lack of any significant geological were undertaken using the GEM simulator of CMG
structure eliminated the need to incorporate (Computer Modelling Group Ltd.) software. GEM
faults into the model. is an advanced equation-of-state compositional and
Defining the stratigraphy of the model from geomechanical reservoir simulator. Perforations in
•
the well tops and seismic interpretation. In the the BBRS formation were placed over a vertical
final model, there were 13 model horizons that thickness of 7 m in the Countess 10–22 well,
bind the top and bottom of each model zone, located at the center of the geomodel. The reservoir
with a total of 12 model intervals. simulation undertaken was for CO2 injection over
a five-year period into the BBRS with a further five-
• Layering the grid cells within each interval from
year monitoring period to follow. The simulations
the bottom upwards, thus eliminating any
involved testing multiple continuous injection sce-
topographic character that may propagate into
narios on the static geomodel. Given the relatively
the underlying zone layering. Grid cells were
low permeability of the reservoir, it was desired to
assigned a height of 0.5 m in the Foremost Fm.,
inject at the maximum allowable bottom-hole pres-
BBRS, as well as the underlying Pakowki Fm.
sure. A water injectivity test yielded a fracture clo-
Grid cells in the remaining model zones were
sure pressure of 6.62 MPa that matches the
assigned a height of 5 m. Each layer within
lithostatic pressure at the BBRS depth of 300 m,
a model zone is evenly distributed between the
based on analysis of the density log. For regulatory
top and bottom of a model horizon.
reasons, the maximum allowable injection pressure
• Locally within the model, upscaling effective must be less than 90% of the fracture gradient, so
porosity and intrinsic permeability well logs at in this project a maximum bottom-hole pressure of
well locations into the appropriate cell using an 5.8 MPa was input into the simulator. Considering
arithmetic mean method. this bottom-hole pressure, a minimum reservoir 289
Figure 16.8 Final 1 km x 1 km detailed geomodel illustrating effective porosity.
temperature of 20°C is necessary to ensure that the permeability coal layers to be represented in the
CO2 will remain in the gas phase. The injection Foremost Fm. From the static geomodel, the grid
simulation presented in this chapter used these was clipped vertically to include only the BBRS and
injection parameters. Foremost formations, as well as the underlying
Pakowki Fm. seals. We did not set no-flow boundaries
Conditioning Data to the static geomodel in order to properly predict the
lateral distribution and migration of the injected CO2
The initial 25-km2 simulation grid consisted of more
plume within the geomodel, without edge effects.
than 60 million grid cells, requiring some further
upscaling to reduce computational run times.
A tartan grid configuration aided in decreasing the Simulation Parameters
total number of cells, as well as preserving the detailed Most parameters governing the pixel-based statistical
cell dimensions along the Countess 10-22 well path. simulation were taken directly from data from the
The resultant grid is defined by (124 × 124 × 73 cells), Countess 10-22 well. For the flow parameters, the
with the cell size in the X- and Y-direction set at 8 m estimated intrinsic permeability realizations are
and the vertical cell thickness remaining at 0.5 m. assumed to be isotropic laterally and the ratio of ver-
In the overlying and underlying intervals, the cell tical to horizontal intrinsic permeability (kv/kh) was
height increased to 51 m above and below the storage assumed to be 0.1, thus resulting in primarily horizon-
290 complex. Some layer discretization enabled the low tal flow of the CO2 in the BBRS reservoir. This ratio of
a) b)
280 280
290 290
300 300
0.5
310 310
0.4
CO2 saturation
320 320
0.3 350 400 450 500 550 600 650 350 400 450 500 550 600 650
0.2 c) d)
280 280
0.1
0 290 290
300 300
310 310
320 320
350 400 450 500 550 600 650 350 400 450 500 550 600 650
Figure 16.9 CO2 saturation during and after the injection program. (a) After one year of injection. (b) After five years of injection. (c) After
one year postclosure. (d) After five years postclosure. The top and bottom of the BBRS are shown by the white lines.
vertical to horizontal permeability is commonly used the gaseous CO2, the kv/kh ratio and local variations in
in the Alberta Basin (Schlumberger, pers. commun.). permeability in the model that were derived from
In the absence of core analysis data, the relative per- upscaling of the well-log data.
meability of gaseous CO2 and water denoting the The results illustrated in Figure 16.9 show that the
irreducible water saturation was estimated using the injected gaseous CO2 is contained within the BBRS.
Brooks–Corey approximation (Brooks and Corey, The plume has a radius of approx. 110 m after five years
1964) and CO2-brine effects (Bachu and Bennion, of injection, extending to a radius of about 140 m after
2008). Using capillary pressure data (Li and Horne, five years of postinjection. The maximum saturation
2006), the minimum and critical water saturation were achieved, local to the 10-22 injector well, is close to 0.5.
assumed to be 0.5. The dynamic model assumes 100% Interestingly, the simulations show no upward migra-
initial water saturation and does not consider geo- tion of the CO2 above the BBRS due to the thin shale
chemical or compositional phase changes, although seal (Figure 16.7), but do demonstrate minor vertical
the dynamic model does consider CO2 gas dissolution migration downward into the Pakowki Fm. Within the
into the formation water in the pore spaces. first five-year injection phase, the plume migrates
downward about 3 m into the Pakowki Fm.
Results The plume shows no additional vertical migration
above or below the BBRS during the five-year postclo-
For the geomodel developed and the injection para-
sure period.
meters described earlier, the reservoir simulation
Through the duration of the injection phase, an
determined an injected CO2 mass of 266 tonnes/
induced pressure plume develops in the BBRS reser-
year, yielding a total of 1,330 tonnes of CO2 injected
voir with maximum pressure reaching the maximum
over the five-year injection program. Example sec-
bottom-hole pressure proximal to the well
tions illustrating CO2 saturation and changes in pres-
(Figure 16.10). The simulations show the excess pres-
sure in the storage complex are displayed in
sure quickly dissipates after one-year postinjection.
Figures 16.9 and 16.10, respectively. Realizations
The pressure almost equilibrates to the baseline pres-
after one year and five years of injection are shown,
sure environment after five years postinjection
as well as one year and five years postinjection. The
CO2 plume shape is driven mainly by the buoyancy of
(Figure 16.10d). 291
a) b)
250 250
Pressure change relative to hydrostatic pressure (MPa)
Depth (m)
Depth (m)
300 300
2.5
2
350 350
1.5 300 350 400 450 500 550 600 650 700 300 350 400 450 500 550 600 650 700
E-W direction (m) E-W direction (m)
1
c) d)
0.5 250 250
0
Depth (m)
Depth (m)
300 300
350 350
300 350 400 450 500 550 600 650 700 300 350 400 450 500 550 600 650 700
Figure 16.10 Pressure changes relative to baseline reservoir pressure due to gas-phase CO2 injection. (a) after one year of injection. (b) After
five years of injection. (c) After one year postclosure. (d) After five years postclosure.
Time-Lapse Numerical Seismic As expected, the shape of the property anomalies is

highly correlated with the shape of the CO2 saturation
Modeling plume in the reservoir over the same time interval.
Before starting the CO2 injection, we modeled the fea- The average decrease in P-wave velocity throughout
sibility of seismic monitoring of the development of the the volume of the plume is approx. 6%, although this
CO2 plume using synthetic data. Initially, Gassmann can reach 38% near the base of the reservoir. This
fluid substitution (Gassmann, 1951) was used to gen- large velocity change corresponds to a high-calcite/
erate the different elastic property models (Vp, Vs, and low-clay content layer at a depth of 306 m
density). This was followed by forward seismic model- (Figure 16.7), which increases the matrix bulk mod-
ing using a finite difference approach. For this part, the ulus. The S-wave velocity increases only slightly
geomodel needed to be rescaled to have uniform depth (around 0.3%) due to an average decrease in forma-
sampling. We used a horizontal sampling interval of tion density of approx. 0.6%.
5 m and a vertical sampling interval of 1 m.
Seismic Modeling
Gassmann Fluid Substitution Numerical seismic data were computed using
Gassmann fluid substitution modeling was underta- TIGERTM, which is a 3D finite-difference modeling
ken using the approach outlined by Smith et al. software (from SINTEF Petroleum Research).
(2003). The bulk modulus of the porous rock frame, We used a simplified version of the acquisition survey
the mineral matrix and the reservoir fluids were first shown in Figure 16.5 (36 sources and 561 receivers).
computed from the dipole sonic well-log data (for Vp The data time sampling interval was 4 ms and we used
and Vs), density-log data, and ELAN analysis for a Ricker wavelet with 40-Hz dominant frequency as
mineral composition analysis, yielding the matrix the source. Data were then processed using Vista
bulk modulus. The results of the Gassmann fluid processing software through a standard processing
substitution are shown in Figure 16.11, as the varia- flow including deconvolution, normal-moveout cor-
tions of the elastic parameters between the baseline rection, common midpoint stack, and poststack
292
model and the model after five years of injection. migration.
a) CO2 saturation b) VP variation (%)

280 280 0
0.4 −5
290 290 −10
Depth (m)
Depth (m)
0.3 −15
300 300 −20
0.2
−25
310 310 −30
0.1
−35
320 0 320
400 450 500 550 600 650 400 450 500 550 600 650
c) VS variation (%) d) Density variation (%)

280 280 0
0.8
290 290 −0.5
Depth (m)
Depth (m)
0.6
300 300 −1
0.4
310
0.2
310 −1.5
320 0 320
400 450 500 550 600 650 400 450 500 550 600 650
Figure 16.11 2D sections from the geomodel showing the variation of elastic parameters after five years of CO2 injection, for the bottom
hole pressure set to 5.8 MPa, reservoir temperature of 20°C. (a) Rescaled CO2 saturation. (b) P-wave velocity change (%). (c) S-wave velocity
change (%). (d) Density change (%).
Results seen at the bottom of the reservoir (Figure 16.11a).

The time slices in Figure 16.13 are were taken at a two-
Figure 16.12 shows a perspective view of in-line and
way reflection time of 0.21 s. Considering a mean
cross-line sections from the processed 3D numerical
overburden velocity of 2800 m/s, the corresponding
seismic volumes from the baseline survey and the
depth is 295 m, which is the depth to the top of the
monitor survey after five years of CO2 injection.
BBRS reservoir.
The two data sets appear to be very similar, but the
time-lapse response is visible after the baseline
volume is subtracted from the monitor volumes. Discussion and Conclusions
Figure 16.13 shows amplitude difference displays of This chapter has outlined the approach and metho-
vertical sections and time slices after one year and five dology used to predict the behavior of CO2 injection
years of injection. The maximum lateral extent of the into the Basal Belly River Sandstone at a depth of
amplitude anomalies are shown by the black lines in 296 m below the ground surface. This shoreface
Figure 16.13 and for both time periods, the lateral sandstone has effective porosity of 11% with intrin-
extent of the reflectivity anomaly in the time slices sic permeability of 0.5–2 mD, and is overlain by
correspond well to the lateral dimensions of the CO2 complexly interbedded shale-silt sequences with lat-
plume on the vertical sections. Despite the small erally extensive coal seams. The project is designed
amount of CO2 injected (266 tonnes), the plume to simulate leakage of CO2 from a deep sequestra-
radius is 55 m after one year of injection. This tion site into formations at shallow to intermediate
plume should be detectable in the seismic data during depths, and to determine the detection threshold
the first year of injection. The seismic anomaly is of the CO2 for various monitoring technologies.
generated by the full thickness of the CO2 plume In this chapter we examined the 4D P-wave
and not just by the strong P-wave velocity contrast seismic response during the injection program. 293
a) baseline b) 5 years of injection
0 0
100 100
200 200
time (ms)
time (ms)
300 300
400 400
500 500
600 600
650 650
N-S 550 650 N-S 550 650
dir 550 dir 550
ec 450 450 ( m ) ec 450 450 ( m )
tio tion tio n
n(
m)
350 350 direc n( 350 350 directio
E-W m) E-W
Figure 16.12 Processed 3D seismic volumes. (a) Baseline volume. (b) Monitor volume after five years of injection.
Multicomponent seismic data were recorded during compositional changes, such as chemical dissolution
the acquisition program but the converted-wave or precipitation of minerals.
(PS) data has not yet been interpreted. Well log Changes in the reservoir petrophysical properties
and core data from the Countess 10-22 well as well after CO2 injection were computed using Gassmann’s
as petrophysical data from surrounding wells were equations and numerical time-lapse seismic modeling
used to characterize the BBRS reservoir and caprock of the updated reservoir properties indicates that seis-
strata. Reservoir simulations were then undertaken mic data should be able to detect the CO2 plume after
to predict the extent of the gas-phase CO2 plume one year of injection, despite the small amount of CO2
after five years of injection. With the relatively low (266 tonnes) and the limited radius of the plume
permeability of the reservoir sandstones (0.5–2.0 (55 m). Actual injection of CO2 into the BBRS is
mD) and the limit on the injection pressure to 5.8 scheduled to begin in mid-2018 and monitoring
MPa at the reservoir resulted in a predicted injection results will be history matched against actual reservoir
rate of 266 tonnes/year, this reaching a total volume and caprock performance and monitoring surveys
of 1330 tonnes of CO2 injected over the planned conducted.
program period of five years. At this injection rate,
it is predicted that the CO2 plume will reach
a diameter of approx. 220 m after five years of Acknowledgments
injection, and a diameter of 280 m five years after We thank CMC Research Institutes, Inc. for per-
cessation of injection. The results demonstrate that mission to publish this chapter and for access to
both the pressure and CO2 plumes should remain the well and seismic data. Funding for the con-
within the 1 km × 1 km footprint of the Field struction of the FRS was provided by the Province
Research Station. Cumulatively, injection rates of Alberta (Department of Environment and Parks)
appear to gradually increase over time and reach and from the Government of Canada (Western
a plateau, caused by relative permeability hysteresis. Economic Diversification), with strong support
This is an effect where the relative permeability from the University of Calgary. Financial support
(CO2-water) increases over time as a result of the from Shell Global Solutions and Klohn Crippen
decreasing water saturation (Bachu and Bennion, Berger is also gratefully acknowledged. Support
2008). Two other factors contributing to this effect for J. Dongas was provided by the industrial spon-
include gas compressibility and CO2 dissolution into sors of the Consortium for Research in Elastic
the BBRS formation water (salinity is approx. 8000 Wave Exploration Seismology (CREWES) at the
ppm). Note that the simulated scenarios do not take University of Calgary, and the Natural Sciences
into account further complexity of reservoir and Engineering Council of Canada (NSERC)
294
Figure 16.13 Vertical and time slice seismic amplitude difference displays from the 3D volumes. (a) Vertical section, difference between t =
1 year and the baseline. (c) Vertical section, difference between t = 5 years and the baseline. On both sections, the lateral extent of the inferred
CO2 plume is denoted by the vertical black lines. (b) Time slice at 0.21 s of the difference volume between t = 1 year and the baseline. (d) Time
slice at 0.21 s of difference volume between t = 5 years and the baseline. Interpreted lateral extent of the CO2 plume is indicated by the black
outline.
through grant CRDP 379744–08. Support for Beaton, A. (2003). Production potential of coalbed methane
M. Macquet was provided by the University of resources in Alberta. Alberta Energy and Utilities Board,
Calgary Eyes High Postdoctoral Fellowship pro- EUB/AGS Earth Sciences Report 2003–03.
gram. Schlumberger Carbon Services is thanked Brooks, R. H., and Corey, T. (1964). Hydraulic properties of
for Project Management during the development porous media. Colorado State University, Hydrology
and construction of the Field Research Station, Papers, 3, 1–24.
including drilling and logging of wells, and also Chadwick, R. A., Zweigel, P., Gregerson, U., Kirby, G.A.,
for providing access to PETREL software. Holloway, S., and Johannessen., P. N. (2004).
Geological reservoir characterization of a CO2 storage
The Computer Modelling Group (CMG) is thanked
site: The Utsira Sand, Sleipner, northern North Sea.
for providing GEM simulation software, and Energy, 29: 1371–1381.
SINTEF is acknowledged for providing access to
Coates, G. R., and Dumanoir, J. L. (1974). A new approach
TIGER seismic modelling software. IHS Energy
to improved log-derived permeability. Log Analyst,
Canada provided the wireline data used in the January–February: 17.
construction and characterization of the static geo-
Eisinger, C. L., and Jensen, J. L. (2009). Data integration,
model. Cenovus Energy has kindly provided access petrophysics, and geomodelling: Wabamun Area CO2
to the site for a period of 10 years. We also thank Sequestration Project (WASP). Energy and
Dr. Malcolm Wilson and Dr. Thomas Davis for Environmental Systems Group, Institute for
constructive reviews of this contribution. Sustainable Energy, Environment and Economy,
University of Calgary.
Fenton M. M., Schreiner, B. T., Nielsen, E., and
References Pawlowicz, J. G. (1994). Quaternary geology of the
Bachu, S., and Bennion, B. (2008). Effects of in-situ Western Plains. In G. D. Mossop and I. Shetsen (comp.),
conditions on relative permeability characteristics of Geological Atlas of the Western Canada Sedimentary
CO2-brine systems. Environmental Geology, 54(8): Basin. Canadian Society of Petroleum Geologists and
1707–1722. Alberta Research Council, Calgary, 413–420.
295
Gassmann, F. (1951). Über die Elastizität poröser Medien. Glombick, P. M., and Mumpy, A. J. (2014b). Subsurface
Viertel. Naturforschenden Gesellschaft Zürich, 96: 1–23. stratigraphic picks for the Milk River ‘shoulder’,
Gibson-Poole, C. M., Svendsen, L., Underschultz, L., et al. Alberta Plains: Including tops for the Milk River Fm.
(2008). Site characterization of a basin-scale CO2 and Alderson Mbr. of the Lea Park Fm: Alberta Energy
geological storage system: Gippsland Basin, southeast Regulator Alberta Geological Survey, Open File Report
Australia. Environmental Geology, 54: 1583–1606. 2013–17.
Glombick, P. M. (2010a). Top of the Belly River Gp. in the Hamblin, A. P., and Abrahamson, B. W. (1996) Stratigraphic
Alberta Plains: Subsurface stratigraphic picks and architecture of “Basal Belly River” cycles, Foremost
modelled surface. Alberta Energy Regulator Alberta Formation, Belly River Group, subsurface of southern
Geological Survey, Open File Report 2010–10. Alberta and southwestern Saskatchewan. Bulletin of
Canadian Petroleum Geology, 44(4): 654–673.
Glombick, P. M. (2010b). Top of the Belly River Gp. in
the Alberta Plains: Subsurface stratigraphic picks and Ah-Ram Kim, Gye-Chun Cho, and Tae-Hyuk Kwon.
modelled surface. Alberta Energy Regulator (2014). Site characterization and geotechnical aspects
Alberta Geological Survey Digital InFm. Series DIG on geological storage of CO2 in Korea. Geosciences
2010–0022. Journal, 18(2): 167–179.
Glombick, P. M. (2011a). Subsurface stratigraphic picks for Li, K., and Horne, R. N. (2006). Comparison of methods to
the top of the Oldman Fm. (Base of Dinosaur Park calculate relative permeability from capillary pressure
Fm.), Alberta Plains. Alberta Energy Regulator Alberta in consolidated water-wet porous media. American
Geological Survey, Open File Report 2011–13. Geophysical Union, Water Resources Research, 42:
W06405. DOI:1029/2005WR004482.
Glombick, P. M. (2011b). Subsurface stratigraphic picks for
the top of the Oldman Fm. (Base of Dinosaur Park Rock, L., Brydie, J., Jones, D., Jones, J.-P., Perkins, E., and
Fm.), Alberta Plains. Alberta Energy Regulator Alberta Taylor, E. (2015). Methodology to assess groundwater
Geological Survey, Digital InFm. Series DIG 2011–0006 quality during CO2 injection at the Quest CCS project.
(tabular data, tab-delimited format, to accompany Geoconvention 2015, Geoscience New Horizons.
Open File Report 2011–13. Rostron, B., White, D., Hawkes, C., and Chalaturnyk, R.
Glombick, P. M. (2014a). Subsurface stratigraphic picks for (2014). Characterization of the Aquistore CO2 project
the Belly River Gp./Lea Park Fm. Transition in storage site, Saskatchewan, Canada. Energy Procedia,
East-Central Alberta. Alberta Energy Regulator 63: 2977–2984.
Alberta Geological Survey, Digital InFm. Series DIG Smith, T., Sondergeld, C., and Rai, C. (2003). Gassmann
2013–0031. fluid substitutions: A tutorial. Geophysics, 68, 430–440.
Glombick, P. M. (2014b). Subsurface stratigraphic Timur, A. (1968). An investigation of permeability,
picks for the top of the Foremost Fm. (Belly River porosity, and residual water saturation relationship for
Gp.), Alberta Plains. Alberta Energy Regulator sandstone reservoirs. Log Analyst, 9(4): 8.
Alberta Geological Survey, Digital InFm. Series DIG Wright, G. N., McMechan, M. E., Potter, D. E. G., and
2013–0030. Holter, M. E. (1994). Structure and architecture of the
Glombick, P. M., and Mumpy, A. J. (2014a). Subsurface Western Canada Sedimentary Basin. In G. D. Mossop
stratigraphic picks for the top of the Milk River and I. Shetsen (comp), Geological atlas of the Western
‘shoulder’, Alberta Plains. Alberta Energy Regulator Canada Sedimentary Basin. Canadian Society of
Alberta Geological Survey, Digital InFm. Series DIG Petroleum Geologists and Alberta Research Council,
2013–0025. Calgary, 25–40.
296
Chapter
Seismic and Electrical Resistivity
17 Tomography 3D Monitoring at the Ketzin

Pilot Storage Site in Germany
Christopher Juhlin, Stefan Lüth, Monika Ivandic, and Peter Bergmann
Summary and Gale, 2004), pilot-scale projects, designed for the

injection of small CO2 quantities (100 ktonnes or
Geophysical monitoring is essential for any CO2 sto-
less), are important laboratories, because they
rage site. At the Ketzin CO2 pilot storage site, seismic
provide well-controlled conditions and allow testing
and electrical resistivity tomography (ERT) measure-
various technologies at relatively low cost.
ments were integral components of the monitoring
The Ketzin project, located west of Berlin (Figure
program. The operational phase (injection) at the site
17.1), is one of these pilot sites. It was initiated in 2004
lasted between June 2008 and August 2013. During this
by the GFZ German Research Centre for Geosciences
time a total of about 67 ktons of CO2 was injected into
in collaboration with European industry and academic
a saline formation in the depth range of 630–650 m.
partners as the first European onshore storage project
For monitoring purposes, four 3D seismic surveys have
(Martens et al., 2014). The Ketzin project was designed
been acquired to date, one baseline, two repeats during
to cover all life-cycle phases of a storage project, from
the injection period, and one postinjection survey. ERT
the assessment to the postclosure and posttransfer
was performed using both crosshole ERT measure-
phases. The operational phase (injection) lasted
ments and surface-downhole ERT measurements as
between June 2008 and August 2013. The Ketzin site
the main geometries. Amplitude and time-shift analy-
was permitted according to German Mining Law.
sis of the seismic data allow high-resolution tracking of
Therefore, the transfer of liability for this site will
the gaseous CO2 plume to be made at Ketzin.
follow German Mining Law, but the research and
The preferred plume migration is in the northwesterly
development activities at Ketzin will continue in
direction, nearly parallel to the anticlinal structure of
order to close the full life cycle of a storage site accord-
the site. Heterogeneity in the reservoir formation is
ing to the EU CCS-Directive (Martens et al., 2014).
interpreted to control the plume migration. ERT time
The main components of the Ketzin site infra-
lapse analysis shows a significant decrease in resistivity
structure are the injection well (CO2 Ktzi 201/2007,
in the vicinity of the borehole due to CO2 injection.
abbreviated as Ktzi 201) and four monitoring wells.
Even these data indicate a predominantly northwes-
The injection well and two monitoring wells (Ktzi
terly migration of the CO2 plume. Comparison of the
200, Ktzi 202) were drilled in 2007 and reached
geophysical data to fluid flow modeling of the injection
depths of about 750–800 m. In two further drilling
process shows that the site is behaving in conformity to
campaigns, a shallow monitoring well, P300, and
expectations. The amount of gaseous CO2 imaged is in
a deep monitoring well, Ktzi 203, were drilled, reach-
line with the modeling given that there is a high dis-
ing depths of 446 m and 701 m, respectively. The
solution rate and that there is significant lateral spread-
injection facility, including tanks, pumps, and heating
ing of the CO2 in thin layers that have a thickness less
system, was installed in 2008 and dismantled after
than the seismic detection limit. Based on the geophy-
injection closure in 2013.
sical data it is concluded that no measureable amounts
CO2 injection at Ketzin lasted from June 30, 2008
of CO2 have leaked out of the storage reservoir.
until August 29, 2013. A total mass of 67 ktonnes of
CO2 was injected into sandstone layers of the Triassic
Introduction Stuttgart Formation at a depth of 630–650 m. Typical
Although industrial-scale CO2 storage projects have injection rates were between 1,400 and 3,250 kg CO2/
297
been operational for many years (e.g., Sleipner: Torp hour. Figure 17.2 shows the cumulative mass of CO2
C. Juhlin, S. Lüth, M. Ivandic, and P. Bergmann
354000 356000 358000

6° 8° 10° 12° 14° 16° 18°
56° 0 500 1,000 N
Baltic Sea
Meters
North Sea
55°
NWGB 54°
Ug Ktzi 163/69
5820000
5820000
NEGB Poland PT 53° 550
Germany
Berlin
Ketzin 52°
Potsdam
Europe 51°
200 km
5818000
5818000
CO2 injection site 0
60
650
700
750
Ketzin
5816000
5816000
354000 356000 358000
Figure 17.1 (Left) Map of the European Permian Rotliegend Basin with the location of the Ketzin pilot site (black star). NWGB, Northwest
German Basin; NEGB, Northeast German Basin; PT Polish Trough. (Right) Interpreted depth of the top Stuttgart Formation (in meters below
ground level) and mapped faults of the Central Graben Fault Zone (CGFZ; Juhlin et al., 2007). The bold star marks the location of the Ketzin pilot
site. (Figures modified from Norden and Frykman, 2013.)
injected over time. Pressure was monitored at 550 m storage efficiency and the stabilization of the storage
depth in order to ensure that the injection process did complex:
not cause unsafe elevated reservoir pressure. Mainly,
• Back-production of 240 tonnes of CO2 and 55 m³
food grade quality CO2 was injected (>99.9% purity),
of brine
delivered by road tankers. From May to June 2011,
• Injection of 2900 tonnes of brine
1.5 ktonnes of CO2 captured from the Vattenfall
oxyfuel pilot plant Schwarze Pumpe was used. This The CO2 and brine back-production experiment was
power plant CO2 had a slightly lower degree of purity performed in October 2014 (Martens et al., 2015).
(>99.7%) than the food grade CO2 used in the main The purpose of this experiment was to investigate
experiment. Between March and July 2013, the injec- reservoir and well behavior during back-production
tion temperature was incrementally decreased from and to test the performance of monitoring methods
40°C (approximate reservoir temperature) to 10°C in for detecting the propagation of atmospheric CO2 in
order to test the feasibility of CO2 injection without the case of leakage through a well. The back-produc-
energy-intensive heating (Möller et al., 2014). At the tion test was accompanied by an extensive sequence of
end of the injection phase, in July and August 2013, ERT monitoring surveys.
the coinjection of CO2 and N2 was tested in order to The brine injection experiment took place
demonstrate the technical feasibility of continuous between October 2015 and January 2016 (Möller et al.,
injection of impure CO2 (Martens et al., 2014). 2016). The purpose of the brine injection was to trigger
After the end of the injection phase, two experi- accelerated imbibition of the reservoir rocks with
ments were performed in order to investigate the brine close to the injection well. By ERT monitoring
298 of this process key data for the quantification of
Seismic and ERT 3D Monitoring
Figure 17.2 Cumulative mass of CO2 injected over time (green curve) and pressure monitored at 550 m depth in the injection well (blue
curve). Injection shut-in phases are characterized by fast reservoir pressure reduction and field tests (cold injection, back production, and brine
injection) triggered strong pressure perturbations. (Figure modified from Wipki et al., Energy Procedia 97, 2016.)
residual gas saturation was collected. Furthermore, Bergmann et al. (2016) paper with more recent
this experiment was performed as a test of brine monitoring results and provide an overview of the
injection for leakage remediation in the case of a most important monitoring results.
damaged leaking well.
Further site operations after the termination of the
field experiments comprise the abandonment of the Geological Setting
remaining wells (well Ktzi 202 was abandoned in Förster et al. (2006) give an extensive overview of the
2015) and dismantling the remaining surface infra- geological setting of the Ketzin site and we briefly
structure at the site, executed in 2017. summarize the main points here. Flow of the Upper
Several papers have been published concerning Permian Zechstein Salt in the northeast German Basin
the geophysical monitoring of the Ketzin site. began in the Triassic and initiated the development of
Ivanova et al. (2012) reported on results from the the Roskow–Ketzin anticline (Kossow et al., 2000) that
first seismic repeat survey, Ivandic et al. (2015) from the Ketzin site is located on. From the later part of the
the second repeat survey, and Huang et al. (2017) Triassic into the Cretaceous, salt movement continued
from the third repeat survey (the first postclosure in phases, with the two main ones being at 140 Ma and
survey). Huang et al. (2015) presented results inte- 106 Ma. The former resulted in uplift and erosion of the
grating dynamic flow simulations with seismic mon- Jurassic successions and the latter one in erosion of
itoring observations. Schmidt-Hattenberger et al. Lower Cretaceous deposits (Förster et al., 2006).
(2014) reviewed the extensive ERT monitoring pro- During the Upper Cretaceous there was no deposition
gram that has been ongoing at Ketzin. Bergmann and at the site, as it was part of a structural high
Chadwick (2015) estimated volumetric bounds on the (Stackebrandt and Lippstreu, 2002). Therefore,
amount of CO2 that is observable in the time lapse Oligocene sediments unconformably overlie the
seismic surveys. A recent comprehensive overview of Jurassic ones (Förster et al., 2006). The present day
the Ketzin project was presented by Bergmann et al. strike of the anticline axis is in the NNE–SSW direction,
(2016). In this contribution we complement the with flanks dipping at about 15 degrees. 299
Stratigraphy Lithostratigraphy Gamma P−Sonic Density Syn symmetric Inline 1165 Syn anti−symmetric
g/cc
API m/s
0 100 200 3000 1
6000 2 3
0 0.0
Quaternary
100 Paleogene Rupelian

0.1
200
Pliensbachian
0.2
Jurassic
Sinemurian
300
Measured Depth (m)
Hettangian 0.3
Time (s)
400
Exter Fm.
0.4
500
Arnstadt Fm.
Weser Fm. (K2 Horizon)

0.5
Triassic
600 Weser Fm
.
Stuttgart Fm.
0.6
700 Lithology
Sandstone
Siltstone
Grabfeld Fm.
Mudstone
Anhydrite
800
0.7
Figure 17.3 Stratigraphy of the Ktzi200/2007 based on Prevedel et al. (2008) with the natural gamma, P-wave sonic, and density logs. The natural
gamma log is shaded by lithology as shown in Sopher et al. (2014). Zero-offset synthetic seismograms were generated using the velocity and density
logs after averaging over 1.5-m intervals with a 40-Hz symmetric Ricker wavelet as input (left seismic panel) and an antisymmetric first-derivative
Gaussian wavelet (right seismic panel). The migrated seismic data (middle seismic panel) are from Inline 1165 just to the west of the well.
The actual injection site is located on the southern 10–20 m thick anhydrite layer (the K2 reflector) is
flank of the eastern dome of the anticline, where sand- present within the Weser formation and generates
stones, shales, and siltstones are the predominant lithol- a strong seismic reflection (Juhlin et al., 2007). Above
ogies of the Triassic–Jurassic succession (Figure 17.3). the K2 reflector is the Exter Formation that consists of
CO2 was injected into the saline water bearing reservoir shallow marine deposits, forming the top of the Triassic
rocks of the Upper Triassic Stuttgart Formation at units. Natural gas was stored in the overlying Jurassic
630–650 m depth. This heterogeneous c. 80 m thick sandstones just to the north of the Ketzin site until
formation consists of both high-quality sandy channel the year 2000 (Lange, 1966). The Tertiary Rupelton
reservoir facies and poor-quality clayey flood plain clay comprises the caprock for the former gas storage,
deposits (Förster et al., 2006). The Weser and Arnstadt above which lie Quaternary sediments that constitute
Formations, containing mainly fine grain marls and present-day freshwater aquifers (Figure 17.3).
mudstones (Förster et al., 2006), overlie the Stuttgart The K2 reflector generates the most prominent
300 Formation and constitute the caprock. An approx. seismic response based on logging data from the
Ktzi 200 well (Figure 17.3). This is due to the large an overlapping template scheme with 5 receiver lines
increase in the velocity and density of the anhydrite containing 48 active channels in each template. Five
layer compared to the surrounding clastic rocks. receiver lines made up one swath. In each template,
Excellent correlation is observed between the syn- 200 nominal source points were activated using an
thetic seismic response and an inline from the 3D accelerated weight drop. After all sources had been
seismic volume (Juhlin et al., 2007), regardless of activated within a template, the receiver locations
whether a symmetric or antisymmetric wavelet is were shifted by half a template when moving within
used as input. A good correlation between the syn- a swath. A similar overlapping scheme was used
thetic and real data is also seen at the Top Arnstadt between the source points of adjacent swaths. This
Formation and Top Exter Formation at this location. overlap of source points and recording stations from
The Top Stuttgart Formation generates a reflection template to template yielded an even nominal fold of
about 40 ms below the K2 reflection and represents 25 for the survey area. However, owing to logistical
the top of the CO2 injection reservoir. The impedance restrictions such as roads, villages, and infrastructure,
contrast between the top of the reservoir and the the number of actual source points in each template
overlying caprock is negative. Injected CO2 that varied.
remains in a gaseous state will reduce the P-wave The first two repeat 3D seismic surveys were
velocity and density of the reservoir further and, acquired in the autumns of 2009 (Ivanova et al.,
therefore, the amplitude of reflected waves from the 2012) and 2012 (Ivandic et al., 2015) during the injec-
reservoir portions containing CO2 should be tion phase (Figure 17.4a). The cumulative amount of
enhanced compared to those with no CO2. injected CO2 at the times of the two surveys was
approximately 23 ktonnes and 61 ktonnes, respec-
tively. The third repeat survey was conducted in the
Data Acquisition postinjection phase, two years after a total amount of
Seismic Data 67 ktonnes of CO2 had been injected.
The recording equipment and acquisition para-
Various seismic monitoring methods, such as 2D and
meters used in the repeat surveys were identical to
3D surface seismic surveys (Bergmann et al., 2012;
those used in the baseline (Juhlin et al., 2007) to
Ivanova et al., 2012, Ivandic et al., 2012), crosswell
ensure high repeatability of the data sets. Table 17.1
surveys (Zhang et al., 2012), and moving source profil-
summarizes the acquisition parameters used for the
ing (Yang et al., 2010), have been applied at the Ketzin
3D surface seismic surveys.
CO2 pilot site to monitor the CO2 plume behavior
within the target reservoir at different scales. Among
them, the time-lapse 3D surface seismic method has ERT Data
been an essential tool for mapping the lateral extent of Aside from seismic measurements, ERT was per-
the spreading gaseous plume and tracking its move- formed in order to monitor the injected CO2
ment. Between autumn 2005 and autumn 2015 four (Kiessling et al., 2010). The use of ERT is motivated
full 3D seismic surveys were performed at the Ketzin by the significant change in rock resistivity that occurs
site (Figure 17.4a). The first 3D seismic survey cover- when electrically well conductive brine in the pore
ing an area of about 12 km2 around the injection site space is substituted by poorly conductive CO2 (e.g.,
was acquired in summer/autumn 2005 prior to the Natatsuka et al., 2010). Crosshole ERT measurements
injection. Main objectives of the survey were (1) to and surface-downhole ERT measurements are the two
verify earlier geological interpretations of the structure main geometries that have been applied at Ketzin
based on vintage 2D seismic and borehole data; (2) to (Kiessling et al., 2010). The technical layout primarily
provide, if possible, an understanding of the structural focused on the crosshole measurements and com-
geometry for flow pathways within the reservoir; (3) prised a vertical electrical resistivity array (VERA),
a baseline for later evaluation of the time evolution of consisting of 45 downhole electrodes deployed in the
rock properties as CO2 is injected into the reservoir; three wells Ktzi200, Ktzi201, and Ktzi202 (Schmidt-
and (4) detailed subsurface images near the injection Hattenberger et al., 2011, 2012). Each well was
borehole for planning of the drilling operations (Juhlin equipped with 15 electrodes, arranged with a vertical
et al., 2007). The 3D seismic data were acquired using spacing of 10 m. To suppress bypassing electrical 301
5822
10:5
5821 10:4
10:3
10:2 9:5
10:1 9:4
9:3
9:2 8:5
5820 8:4
9:1
8:3
7:5
8:2
8:1 7:4
7:1
7:1 6:5
7:1 6:4
UTM Northing (km)
5819 6:3
5:5
6:2
5:4
6:1
5:3
6:0 4:5
5:2
5:1 4:4
5818 5:0 4:3
3:5
4:2
3:4
4:1
3:3
3:2 2:5
3:1 2:4
2:3
5817 1:5
2:2
2:1 1:4
1:3
1:2
1:1
5816
5815
3353 3354 3355 3356 3357 3358 3359
UTM Easting (km)
Figure 17.4 (a) Template system used in the 3D seismic data acquisition and outlines of the four 3D surveys conducted at the Ketzin site.
Black polygon outlines the baseline survey area, whereas red, blue, and green polygons enclose areas of the first, second, and third repeat
survey, respectively. Red and yellow dots show the locations of the injection well and three observation wells, respectively; (b) actual
common depth point fold for the baseline 3D survey with inlines and crosslines marked by blue dashed lines. White dot marks the location
of the injection well.
currents through the metallic well casings, the casings and two electrodes for voltage registration). This
were covered with an insulation (Kiessling et al., setup is varied over the range of available electrodes.
2010). At Ketzin, bipole–bipole, dipole–dipole, and dipole–
In ERT data acquisition, each reading involves dipole across three boreholes were the setups applied
302 four electrodes (two electrodes for current injection, for crosshole measurements (Schmidt-Hattenberger
5822
0
145
125
120
0
115
0
140
0
5821
110
0
105
0
0
145
100
0
135
0
0
0
5820 140 130
0
0
135 125
0
UTM Northing (km)
5819 0
130 120
0
0
125 115
0
5818
0
120 110
0
0
115 105
0
5817
0
110 100
0
125
Inlin
120
0
105
115
5816 s
0
line
es
ss
110
Cro
0
105
0
100
100
Fold
0
0 2 4 6 8 10 12 14 16 18 20 22 24 26
5815
3353 3354 3355 3356 3357 3358 3359
UTM Easting (km)
Figure 17.4 (cont.)
et al., 2016). The limit for current injection was 2.5 injection site (Figure 17.5). In order to optimize the
Ampere, which was limited by the current transmis- signal-to-noise (S/N) ratio, each current injection was
sion capability of the cables that connect the electrode performed over a period of 45–60 minutes with
to the acquisition unit at the surface. A ZT-30 (Zonge a maximum power of up to 10 kilowatts using
Engineering) was used for current transmission and a Scintrex TSQ-4 power source.
a GPD32 unit (Zonge Engineering) was used for vol- While the surface-downhole ERT measurements
tage registration. comprise seven surveys in total (two baseline measure-
The surface-downhole ERT measurements ments before CO2 injection and five repeat measure-
included current injections at the surface and voltage ments during ongoing injection), the crosshole ERT
registrations by means of the VERA electrodes measurements were repeated with much higher fre-
(Bergmann et al., 2012). For this purpose, current quency. That is, initial crosshole ERT measurements
injections were performed at 16 surface locations, were performed on a daily basis. Once the CO2 arrived
which were concentrically arranged around the at the first observation well (i.e., approx. 15 days after 303
Table 17.1 Template acquisition parameters for the 3D surveys beginning injection), the repetition rate was reduced
at Ketzin
to two measurements per week. After the arrival of the
Parameter Value CO2 at the second observation well (March 21, 2009;
Martens et al., 2011; Zimmer et al., 2011), the repeti-
Receiver line spacing /number 96 m/5
tion rate was further reduced to one measurement per
Receiver station spacing /channels 24 m/48 week (Schmidt-Hattenberger et al., 2016).
Source line spacing /number 48 m/12
Source point spacing 24 m or 72 m Time-Lapse Results
Common depth point bin size 12 m × 12 m
Nominal fold 25
Seismic Data
Almost identical processing steps as used for the base-
Geophones 28 Hz single
line survey were applied to the three repeat data sets in
Sampling rate 1 ms order to optimize repeatability. Owing to the different
Record length 3s weather and ground conditions during the surveys,
only static corrections had to be recalculated sepa-
Seismic source: 240 kg accel. weight drop, 8 hits per source point.
rately for each repeat data set (Kashubin et al., 2011;
Bergmann et al., 2014b). Poststack cross-calibration
further minimized the time-lapse noise and enhanced
the time-lapse reservoir signature (Ivanova et al.,
a) Surface-downhole ERT b) Crosshole ERT

CO2
Depth (m)
Ktzi201 Ktzi200 Ktzi202
550
570
Etzin
590
590
5820
610 Weser Fm
0
57
UTM Northing (km)
630
5819
650
0
61 67
0
0 670 Stuttgart Fm
5818 69
0
630 71
650 73
0 690
Neu 0
Falkenrehde
75
0
5817 77 710
0
79
81
0
730 Grabfeld Fm
Ketzin
830
5816 750
3354 3355 3356 3357 3358
Well annulus: Lithology:
UTM Easting (km) Well casing: Cement Mudstone
Blank casing Drill mud Sandstone
Insulated casing XC-polymer Anhydrite
VERA electrode Openhole Sandstone (cemented)
Figure 17.5 (a) Acquisition of the surface-downhole ERT surveys. Gray points indicate electrode pairs used for current injection in the surface-
downhole ERT experiments. Isolines indicate depth of the storage formation (Top Stuttgart Formation). (b) Schematic of the VERA installation
showing the downhole electrodes, well completions, and surrounding lithology (after Förster et al. 2010) at the Ketzin site. (Figure modified
after Bergmann et al., 2017)
304
5819.0
2009 2012 2015
5818.5
UTM Northing (km)
600 600 600
5818.0
650 650 650

5817.5
700 700 700

5817.0
3354.5 3355.0 3355.5 3354.5 3355.0 3355.5 3354.5 3355.0 3355.5
UTM Easting (km) UTM Easting (km) UTM Easting (km)
Diff
0.0 0.2 0.4 0.6 0.8 1.0
Figure 17.6 Normalized amplitude differences (baseline minus repeat) at the reservoir level for the first (left), second (middle), and third
(right) repeat survey. White dot marks the locations of the injection well. Isolines indicate the elevation of top of the Stuttgart Formation.
2012; Ivandic et al., 2015; Huang et al., 2017). The data migration of the plume. The observed amplitude
processing workflow is described in detail in Juhlin anomaly was similar in shape to the one observed at
et al. (2007). the time of the first repeat survey, but larger by
The 3D surveys imaged a sequence of clear reflec- approx. 150 m in the N–S direction and 200 m in
tions from approx. 150 ms down to 900 ms two-way- the W–E direction. Furthermore, it was much stron-
time. Reduced image quality of near-surface struc- ger, with the highest amplitude nearly centered at the
tures is observed only where the fold is lower injection well. The thickness of the plume had also
(Figure 17.4b). The stronger reflections in the target increased, varying from 10 m to 30 m (Ivandic et al.,
reservoir area observed around the injection site in 2015). The latest survey, conducted in 2015 during the
the three repeat data sets were interpreted to be due to postinjection phase, revealed a decrease in the inten-
the increased impedance contrast between the sity and extent of the amplitude anomaly in both the
caprock and the reservoir sandstone caused by CO2 horizontal and vertical directions. The most signifi-
injection. However, to image the lateral distribution cant decrease in the amplitude anomaly occurs east
of amplitude variations related to the injected CO2, and south of the injection site. The amplitude
the seismic amplitudes of the baseline and repeats decrease in this area may be due to lower S/N ratio
were normalized to the peak amplitude of the K2 in the 2015 survey in this area, preferential movement
reflection and subtracted. Figure 17.6 shows maps of of the brine, greater lateral flow into thin layers, or
the normalized amplitude differences at the top of the a combination of factors (Huang et al., 2016).
Stuttgart Formation from the repeat data sets. Figure 17.7 displays vertical cross sections
At the time of the first repeat survey, a pronounced extracted from the three repeat volumes along inline
amplitude anomaly with a lateral extent of 300–400 m 1166 and crossline 1100, both crossing close to the
and a thickness of 5–20 m, was situated at the top of injection site. They show a prominent time-lapse sig-
the reservoir near the injection well. The gaseous nature at approx. 42 ms below the K2 reflector, which
plume had a westerly trending tendency, revealing corresponds to the approximate depth of the top
the heterogeneous nature of the reservoir caused by Stuttgart Formation. The growth of the amplitude
the complex sedimentary history in a fluvial environ- anomaly in the reservoir during the injection phase
ment (Ivanova et al., 2012). Results from the second can be observed in both the inline and crossline direc- 305
3D repeat survey showed further growth and tion. However, as mentioned earlier, at the time of the
(a) Ktzi|201 (b) Ktzi|201

W V E S V N
0.2
Inline 1166 Crossline 1100
0.3
0.4
Time (s)
0.5
0.6
0.7
2009−2005 2009−2005
0.8
1050 1100 1150 1150 1200 1250
Crosslines Inlines
(c) Ktzi|201 (d) Ktzi|201
W V E S V N
0.2
0.3
0.4
Time (s)
0.5
0.6
0.7
2012−2005 2012−2005
0.8
1050 1100 1150 1150 1200 1250
Crosslines Inlines
(e) (f)
Ktzi|201 Ktzi|201
W V E S V N
0.2
0.3
0.4
Time (s)
0.5
0.6
0.7
2015−2005 2015−2005
0.8
1050 1100 1150 1150 1200 1250
Crosslines Inlines
Figure 17.7 Vertical cross sections of time-lapse amplitude differences along inline 1166 (left) and crossline 1100 (right). From top to
bottom: first repeat, second repeat, and first postinjection results. The red line represents the K2 horizon. The location of the injection well
is marked by the black vertical line (Huang et al., 2017).
first postinjection survey, the range and intensity of thinning of the gaseous CO2 component into layers
the amplitude anomaly within the reservoir has thinner than the seismic detection limit. Nevertheless,
decreased and the push-down effect at deeper levels as dissolution is a crucial trapping mechanism in the
is weaker compared to the second repeat survey. This early postclosure phase, and spreading of the CO2
has been interpreted to be mainly due to CO2 dissolu- through rather thin layers indicates a pressure relaxa-
306 tion in the formation brine and lateral spreading of tion in the reservoir, both processes demonstrate
the gaseous CO2 (Huang et al., 2017), both leading to ongoing stabilization of the CO2 plume at the Ketzin
site (Huang et al., 2017). No systematic changes in the bearing reservoir sandstones have similar ranges of
seismic amplitudes within the overburden were electrical resistivity. However, after CO2 injection
detected by the three monitoring surveys, indicating starts the models display an increase in resistivity at
that CO2 is being contained within the storage the depth of the injection. For the later stages of CO2
reservoir. injection, weakening of the resistivity signature is
observed that coincides with a reduction in the peak
Electrical Resistivity Tomography CO2 injection rates from 3.25 tonnes/hour to 1.5
tonnes/hour (Figure 17.8c). In general, the crosshole
Processing of the ERT data consists of two steps. First,
ERT models provide a better resolution than the sur-
preprocessing of the acquired voltage time series into
face-downhole ERT models (not shown in
resistivity values is required. For both measurement
Figure 17.8). In this context it is important to note
setups the voltage time series cover various cycles of
that the imaged resistivity model generally suffers
the input DC signal (the number of signal cycles is
from a rapid loss in resolution with increasing dis-
chosen according to the S/N ratio conditions as well as
tance from the wells. Although this implies that the
to the number of four-electrode combinations that
time-lapse electrical resistivity signature cannot be
can be realized). This sequence is then stacked into
compared in detail to the corresponding time-lapse
time series of a single cycle, which undergoes despik-
seismic signature in terms of lateral details, it is in
ing, filtering, and phase corrections (Labitzke et al.,
agreement about a preferential migration trend in the
2012). On the basis of the peak and trough voltage
northwesterly direction (Figure 17.9).
amplitudes in the stacked time series, and the geo-
metric arrangement of the involved electrodes, appar-
ent resistivity values are determined, which represent Discussion
weighted averages of the true resistivities in the Resolution
volume between the electrodes. In addition, resistivity
error values are estimated in order to characterize Seismic Resolution
the quality of the individual readings (Schmidt- As indicated by the time-lapse analysis of the three
Hattenberger et al., 2016). repeat surveys, the amount of gaseous CO2 obser-
In the second step of the ERT processing, the vable in the Stuttgart Formation has decreased sig-
apparent resistivities and estimated errors are used nificantly from the second repeat time to the third
to image the resistivity distribution in the particular repeat time even though more CO2 had been
subsurface region under investigation. This imaging is injected. This decrease has been attributed to two
based on an iterative update loop (e.g., Ellis and main processes, dissolution of the CO2 into the
Oldenburg, 1994), in which a numerical resistivity saline water and lateral migration (spreading of
model, as a representation of the true subsurface, is the CO2 as a thin layer at the top of the reservoir).
used to generate synthetic resistivity data. In each The former process alone is estimated to account
iteration, the synthetic resistivity data set is compared for about 30% of the CO2 being dissolved into the
against the measured resistivity data set, and the formation water. The influence of the latter process
synthetic resistivity model is updated until a sufficient on the observations is highly dependent on the
match between the two data sets is achieved. In seismic acquisition parameters used (source fre-
the case of Ketzin, this imaging was based on finite- quency and strength, geometry repeatability, near
element forward modeling (Rücker et al., 2006) and a surface changes, and ambient noise) and then how
Gauss–Newton inverse procedure for model updating thin the CO2 layer is. Under perfect conditions any
(Günther et al., 2006). The data errors mentioned changes in the subsurface can be detected by time-
earlier are herein used in order to weight the input lapse surveys.
resistivity data and give preference to those with In the Ketzin case, source frequency and strength
higher quality. were very similar for all surveys, so variations in the
Figure 17.8a shows sections of the resistivity mod- source characteristics are not expected to have a great
els obtained from crosshole ERT. For the preinjection influence on the time-lapse results. Likewise, there
stage, the resistivity models show rather weak internal was good repeatability on the source and receiver
contrasts, because the reservoir mudstones and brine- locations, also implying that this factor probably is 307
Figure 17.8 (a) Examples of crosshole ERT inversion results. (b) CO2 saturation assessed from the resistivity change in the crosshole ERT
results for variable regularization parameter. The regularization parameter is a smoothness constraint applied in the processing. Red line shows
the regularization parameter that fits CO2 saturations inferred from neutron-gamma logging (after Baumann et al., 2014, indicated by red
markers) best. Text boxes show the regularization parameters that fit the logging results best. (c) CO2 injection rate and reservoir pressure for
the period of CO2 injection (after Möller et al., 2012; Martens et al., 2014.) (Figure modified after Bergmann et al., 2017.)
not influencing the results significantly. It has been layer thickness decreases below the tuning thickness,
shown that the near surface conditions have varied the amplitude on the difference section will decrease.
during the different surveys due to weather (Kashubin When ambient noise is present this difference in
et al., 2011). These variations influenced the velocity amplitude will be masked by the noise.
in the uppermost loose sediments causing different In order to test how ambient noise may affect the
static corrections having to be applied for each survey. capacity to map thin CO2 layers in time-lapse seismic
However, after the static corrections, the same proces- surveys, synthetic data were generated based on the
sing sequence could be used and the final normalized logs from the Ktzi 200 borehole (Figure 17.3). The K2
root mean square values away from the injection area reflector and reservoir were modeled using the values
are quite low or can be explained by low fold. This listed in Table 17.2 and the same convolutional seis-
leaves ambient noise and CO2 layer thickness as two mic modeling routine was applied as in Figure 17.3
factors that can affect the seismic signature of the using the antisymmetric wavelet. A comparison of
308 gaseous CO2 plume. The two are related since, as the the seismic response using the model in Table 17.2
Table 17.2 Parameters used for the seismic modeling in Figure 17.10
Rock type Top (m) Bottom (m) Velocity (m/s) Density (g/cc)
K2 anhydrite 555 575 5500 2.75
Reservoir (saline) 630–650 650 2750 2.25
Reservoir (CO2) 630 630.5–650 2200 2.20
Background 3300 2.40
Figure 17.9 (a) Time-lapse electrical resistivity anomaly and (b) time-lapse seismic anomaly along the Near Top Stuttgart horizon. Subfigure
(a) shows the resistivity ratio, obtained from the constrained inversion, of the third repeat surface-downhole ERT survey and the second
baseline surface-downhole ERT survey. Subfigure (b) shows the amplitude difference of the first repeat 3D seismic survey and the baseline
3D seismic survey after Ivanova et al. (2012). Both methods image a northwesterly trend for the CO2 migration. (Figure modified after
Bergmann et al., 2014a.)
with no CO2 in the reservoir and the response using due to thin bed tuning generating constructive inter-
the log data as input shows that the simplified model ference at the larger thicknesses.
gives a reasonably similar response as the logs Kashubin et al. (2011) analyzed the S/N ratios on
(Figures. 17.10a and 17.10b). Given that the simpli- source gathers at Ketzin on the baseline and
fied model seems adequate, a thin CO2 layer was the second repeat surveys by comparing the ampli-
gradually introduced into the model and the seismic tudes at the reservoir level with those before the first
difference response calculated, starting with arrivals in the offset range 300–600 m. They found
a thickness of 0.5 m at the top of the reservoir and that the S/N ratio was generally around 1 throughout
increasing to 20 m (filling the entire reservoir). most of the survey area. If the noise is Gaussian then
On the difference section (Figure 17.10c) it is stacking would increase the S/N ratio to 5 for the
observed how the increased thickness increases the processed data with a fold of 25. However, the noise
amplitude and delays the arrival of the main peak. is probably not Gaussian and the fold is lower in the
Note also that at 20 m thickness the amplitude dif- vicinity of the injection site, so a lower S/N ratio may
ference is about the same as the amplitude of the K2 be expected on many parts of the processed data.
reflection at a value of about 0.4. (the source wavelet To test the influence of noise on the difference section
had a maximum amplitude of 1). This is greater than (Figure 17.10c) low-pass (40 Hz) Gaussian noise was
the actual reflection coefficient of any interface and is generated using Seismic Unix (www.cwp.mines.edu) 309
0.4 0.4 0.4

(d) S/N=2.5 (e) S/N=5 (f) S/N=10
Time (s)
0.5 0.5 0.5
0.6 0.6 0.6

0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Thickness (m) Thickness (m) Thickness (m)
0.4 0.4 0.4

(a) Log input (b) Model input (c) Noise free
Time (s)
0.5 0.5 0.5
0.6 0.6 0.6

0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Trace Trace Thickness (m)
Figure 17.10 Analysis of the influence of CO2 layer thickness and S/N ratio on the seismic response of the reservoir. Modeling was done
using a first derivative Gaussian wavelet with a dominant frequency of about 40 Hz. (a) Log data from Figure 17.3 used as input to the
convolutional model. (b) Values in Table 17.2 used as input. (c) Seismic response on the difference section for CO2 layer thickness increasing
from 0.5 to 20 m. (d) Same as (c), but with noise added at 2.5 S/N. (e) Same as (c), but with noise added at 5 S/N. (f) Same as (c), but with
noise added at 10 S/N.
at S/N ratio levels of 2.5, 5, and 10 and added to the injection. Note that this assumption may not be valid
difference section (Figures 17.10d to 17.10f). The S/N based on the fluid flow modeling and the seismic
ratio is defined based on the maximum amplitude in observations (Huang et al., 2015). However, the
the section before adding it. For a high S/N ratio value assumption allows us to convert CO2 saturation levels
of 10, CO2 layers as thin as 2–3 m can potentially be from relative resistivity changes by means of an
identified. For S/N ratio values in the range of 2.5–5 Archie equation, as presented by Kummerow and
the modeling indicates that CO2 layer thickness needs Spangenberg (2011) for the Ketzin reservoir sand-
to be at least in the the range of 5–8 m to be detected, stone. A second assumption, which is made in this
with the higher value corresponding to an S/N ratio of approach, is that the electrical conductivity contribu-
2.5. These values are roughly consistent with the seis- ted by the clay is small compared to the conductivity
mic observations and will be discussed later in the of the brine. Based on the revision of the petrophysical
mass estimate section. model, Bergmann et al. (2017) suggest that this
assumption is valid for CO2 saturation levels of up
Electrical Resistivity Tomography to 70%. Converting resistivity changes into CO2
The following quantitative interpretation is based on saturation levels on the basis of this petrophysical
the assumption that the change in resistivity, which is model leads to the CO2 saturation levels in
displayed in Figure 17.8, is dominantly caused by the Figure 17.8b. These levels have been further calibrated
substitution of brine with CO2. In other words, dis- by thickness-weighted averages of the reservoir CO2
solution and mineral alteration are assumed to play saturations reported by Baumann et al. (2014) on the
310 minor roles, at least throughout the early years of CO2 basis of repeated neutron-gamma loggings. It is
obvious that, apart from the spreading CO2 distribu- acquisition geometries, a comparison of the baseline
tion, other processes are also of relevance for the and repeat AVO volumes reveals good repeatability for
estimated CO2 saturation levels. Rapid saturation the three repeat seismic data sets (Figure 17.11).
increases can, for instance, in several cases be corre- A prominent decrease in the intercept values is
lated with short-term pressure changes and injection observed only at the top of the reservoir around
changes. Generally, the estimated CO2 saturations can the injection well in all three repeat data sets
be seen to be in the range of 40–60% for most of the (Figure 17.11). This correlates well with the spatial
CO2 injection period. variations of the amplitude response at the top of the
ERT methods generally have more limited resolu- reservoir displayed in Figure 17.6. Based on the AVO/
tion compared to that of seismic reflection methods. AVA modeling by Ivanova et al. (2013) and the mod-
Without the use of the VERA electrodes, i.e., solely eling conducted in this study, the observed anomaly is
surface-based measurements, CO2 monitoring in the interpreted to be caused by increased CO2 saturation.
reservoir would in fact not be feasible by means of The same modeling studies also show that both ele-
ERT. Bergmann et al. (2012, 2016) estimate that the vated reservoir pressure and increased CO2 saturation
vertical resolution of the surface-downhole ERT would result in somewhat higher gradient values.
results corresponds to approx. 2–3 electrode spacings However, petrophysical experiments on core samples
(i.e., approx. 20–30 m). Because of the denser data from the Ketzin reservoir indicate that the pressure
coverage and diversity of electrode configurations that increase measured during the injection phase could
can be realized in the crosshole ERT measurements, only have had a rather minor impact on the seismic
the corresponding results are interpreted to have amplitudes. Moreover, because the injection was tem-
a resolution of approx. 5 m in the near-wellbore porarily stopped during the acquisition of the second
region. repeat survey and the third survey was conducted dur-
ing the postinjection phase, the slight increase in gra-
Amplitude versus Offset dient values observed in those data are most likely due
to CO2 saturation.
Amplitude versus offset (AVO), together with results
of petrophysical experiments on core samples from
the Ketzin reservoir, are used to compare the repeat Patchy versus Homogeneous Saturation
seismic data sets with the baseline prior to CO2 The 4D seismic response considered earlier is an
injection in order to investigate the AVO response expression of changes in seismic wave velocity and
from the sandstone reservoir during the injection bulk density that is due to the displacement of brine
and postinjection phases and obtain better insight by CO2 in the pore space. While the density, and
into the time-lapse signature observed in the seismic thereby the density change, can be simply determined
data sets. from the densities and volume fractions of the con-
First, AVO modeling was done to determine stituents (i.e., rock minerals, brine, and CO2), the
whether an AVO anomaly is to be expected and, if same approach is not valid for seismic wave velocity.
so, which type. The input model for the baseline Seismic wave velocity is depending on the density and
survey was based on well logging data and lithologi- the rock’s bulk modulus (K), the latter of which spe-
cal information (Kazemeini et al., 2010). cifies the rock’s volume change when being exposed to
The thickness of the reservoir was assumed to be uniaxial force. We refer to the dependency of seismic
20 m. After applying Gassmann fluid substitution velocity and CO2 saturation here as a velocity–satura-
in the reservoir by replacing brine with a range of tion relationship, which is typically described by
CO2 saturation levels, synthetic seismic traces were means of the Gassmann model (Gassmann, 1951).
generated for both the baseline and repeat cases. This model assumes that the bulk modulus of the
The wavelet used for the synthetic trace was two-phase pore fluid can be averaged from the bulk
extracted from the seismic data. moduli of brine and CO2. This means that the mix-
For the analysis of the real data sets, supergathers ture of CO2 and brine is treated as a single homo-
were generated by averaging over 4 common depth geneous fluid, which is referred to as uniform
points (CDPs) to improve the S/N ratio. In spite of saturation (Mavko et al., 2003). For the Ketzin
the varying weather conditions and variations in the case, the solid line in Figure 17.12 shows the 311
5820.0
5819.5
5819.0
UTM Northing (km)
5818.5
5818.0
5817.5
5817.0
2009 2009
5816.5
3354 3355 3356 3357 3354 3355 3356 3357
5820.0
5819.5
5819.0
UTM Northing (km)
5818.5
5818.0
5817.5
5817.0
2012 2012
5816.5
3354 3355 3356 3357 3354 3355 3356 3357
5820.0
5819.5
5819.0
UTM Northing (km)
5818.5
5818.0
5817.5
5817.0
2015 2015
5816.5
3354 3355 3356 3357 3354 3355 3356 3357
UTM Easting (km) UTM Easting (km)
−0.2 −0.1 0.0 0.1 0.2 −0.2 −0.1 0.0 0.1 0.2
Intercept Gradient
Figure 17.11 Intercept (left column) and gradient (right column) values at the top of the reservoir of the three repeat volumes. The injection
well is marked by the white dot.
velocity–saturation relationship by Kazemeini et al. a significant decrease of the compressional wave

312 (2010) computed on the basis of the Gassmann velocity (Vp) for CO2 saturations of approx. up to
model. This velocity–saturation relationship shows 5%, which is explained by the fact that the bulk
not obeyed. The Gassmann model can then be con-

sidered to be valid only locally, i.e., in regions that
comprise sufficiently homogeneous fluid mixing and
permeability conditions. This case is referred to as
the patchy saturation model (Dvorkin and Nur,
1998; Eid et al., 2015), and the corresponding bulk
modulus has to be computed from a weighted aver-
age that is also taking the rock’s shear modulus into
consideration (Hill, 1963; Berryman and Milton,
1991). The dashed line in Figure 17.12 shows the
corresponding velocity–saturation relationship by
Kazemeini et al. (2010) computed on the basis of
the patchy saturation model. In comparison to the
Figure 17.12 Compressional wave velocity (Vp) as functions of uniform saturation case, it shows a linear relation-
CO2 saturation for uniform and patchy saturation for the Ketzin ship for the velocity–saturation relationship. In fact,
reservoir model. (Modified after Kazemeini et al., 2010).
both models constitute the upper and lower velocity
bounds for the range of possible velocity–saturation
modulus of the gaseous CO2 at Ketzin is very small relationships (Mavko et al., 2003). On the basis of
compared to that of brine (KCO2 = 0.00832 GPa, laboratory measurements of Ketzin core samples,
Kbrine = 3.68 GPa for Ketzin reservoir conditions). Kummerow and Spangenberg (2011) observed that
For larger CO2 saturations, a gradual rebound in ultrasonic velocities show trends from both models
the compressional wave velocity is expected, which for different CO2 saturation ranges (Bergmann et al.,
is due to the progressive decrease in density 2016). Based on these laboratory measurements,
(Whitman and Towle, 1992). Kummerow and Spangenberg (2011) report velocity
Aside from the assumption of uniform saturation, decreases from approx. 3 km/s to approx. 2.77 km/s.
the Gassmann model is also based on the assumption
that the rock is homogeneous and isotropic, and the Mass Estimates and Conformity of Observed
rock pores are hydraulically well connected. These
assumptions are, however, in the context of CO2 sto- and Simulated Reservoir Behavior
rage prevalently not met. This is conditioned by the Estimating the quantity of stored CO2 that has been
fact that a region, once it has been swept, has an detected by geophysical observations is an important
increased effective permeability to flow, which leads task for assuring containment and/or quantifying any
to spatial variations in fluid mobility and preferential potential leakage (i.e., unexpected migration of CO2
migration paths. In most cases, reservoir rocks are out of the storage complex). For a large scale storage
further characterized by spatial variations in wettabil- site, Chadwick et al. (2005) demonstrated that it is
ity, permeability, or shaliness, which can cause finger- possible to quantify the mass of CO2 stored with some
type CO2 migration (Asveth, 2009). The occurrence of uncertainty remaining. Pilot scale storage sites typi-
nonuniform CO2 migration has been observed in cally contain small CO2 quantities that are in the same
a wide range of experimental studies (e.g., Shi et al., order of magnitude as would be potential accumula-
2011; Alemu et al., 2013; Nakagawa et al., 2013), tions of CO2 leaking/migrating out of the storage com-
which indicate that (1) nonuniform CO2 migration plex. A reliable quantification of leakage is a
is most likely the relevant case and (2) that the degree prerequisite for determining appropriate remediation
of saturation heterogeneity is scale dependent. This measures. Comprehensive geophysical monitoring and
scale dependency has direct implications on the rela- petrophysical investigations at the Ketzin pilot site pro-
tionship of seismic velocity and CO2 saturation, as vide the data for an integrated mass estimation of the
a passing seismic wave causes local pore-pressure CO2 in the storage formation that has been imaged by
differences. Given the circumstance that heterogene- time-lapse seismic monitoring.
ities in the fluid mixing and porosity variations ham- The integrated mass estimation aims at attributing
per the equilibration of wave-induced pore-pressure the mass of CO2 to the normalized amplitude differ-
313
differences, the assumption of uniform saturation is ences detected by the three seismic monitoring
CO2 were estimated to be present for the first

and second repeat surveys, respectively (Figure 17.13;
Ivanova et al., 2012; Ivandic et al., 2015). This is to be
compared to 22–25 kilotonnes of actually injected
CO2 by the first repeat survey and 61 kilotonnes of
actually injected CO2 by the second repeat survey.
For the second and third repeat surveys, i.e., the
surveys before and after CO2 injection ended,
Huang et al. (2017) estimated the amount of CO2 to
lie between 55.7 and 10.8 ktonnes (Fig. 17.13), which
is a significant reduction that cannot be explained
completely by reservoir stabilization processes such
as dissolution and CO2 propagation in very thin
layers (fingering). Another effect adding uncertainty
to the mass estimation may be the potential varia-
bility of CO2 density in the reservoir.
The integrative approach suffers from a number of
uncertainties that are discussed in the aforementioned
publications. Saturation and porosity distributions
are assumed to be represented by average values that
may be incorrect for large areas of the reservoir, and
the petrophysical model, derived from rock core
experiments, may be incorrect for reservoir scale sur-
(after Huang et al., 2017). face seismic observations (see also discussion in the
preceding section on homogeneous vs. patch satura-
Figure 17.13 Estimated CO2 mass for each common depth point
bin for the first repeat survey (after Ivanova et al., 2012), second tion). Also, an incorrect assumption of initial seismic
repeat survey (after Ivandic et al., 2015), and third repeat survey. Vp velocity in the reservoir before CO2 injection (full
(after Huang et al., 2017) brine saturation) may result in a significant uncer-
tainty of the mass estimation.
surveys acquired in the years 2009, 2012, and 2015 Huang et al. (2017) further addressed the uncer-
(Ivanova et al., 2012; Ivandic et al., 2015; Huang et al., tainties in the velocity–saturation relationship by
2017). The integrated mass estimation used the follow- revisiting the previous quantification studies by
ing results from geophysical and petrophysical investi- means of an alternative approach. This is of particular
gations: Normalized difference amplitudes from 3D relevance because the petrophysical measurements,
repeat seismic surveys, time shifts of seismic reflections from which the velocity–saturation relationship is
from below the reservoir, saturation profiles at well interpreted, were performed only on a limited num-
locations from Pulsed Neutron Gamma (PNG) logging ber of core samples (Kummerow and Spangenberg,
measurements, and seismic velocity changes due to 2011). In contrast to the approach presented in
CO2 replacing brine observed on core samples from Ivanova et al. (2012) and Ivandic et al. (2015), which
the storage formation. Difference amplitudes and is based on time-lapse amplitudes and 4D time shifts,
saturation profiles were used to estimate the lateral the alternative quantification approach provides
distribution of average CO2 saturation within the CO2 bounds on the amount of CO2 present based on 4D
plume. Furthermore, time shifts and CO2-related velo- time shifts only (Bergmann and Chadwick, 2015).
city changes were used to determine the thickness of the Huang et al. (2017) compared the CO2 mass bounds
CO2 plume. Using average values for reservoir porosity to the true amount of CO2 injected in order to validate
and CO2 density, the total mass of gaseous CO2 was alternative velocity–saturation relationships that have
then computed by summing up inferred CO2 masses not been considered in the previous interpretations.
from each image point in the seismic data. Based on this Thereby, they showed that the set of possible velocity–
314 approach, 23.83 kilotonnes and 52.51 kilotonnes of saturation relationships is not confined to patchy
fluid mixing models, but contains also moderate observations were in good conformity with the simu-
Gassmann-type models. lated plume footprint for a minimum thickness of
Both quantification approaches have demon- 5–7 m. Equivalent observations were made for the
strated that uncertainties are significant and, there- similarity index comparing plume shapes from seismic
fore, only a rough estimate of the amount of gaseous data and simulations where best results were achieved
CO2 imaged by time-lapse seismic measurements can for the central part of the simulated CO2 plume with
be provided. Mass estimations alone are therefore of a minimum thickness of 5 m. This case study showed
rather limited use for demonstrating full contain- that several performance parameters are available pro-
ment. They can be used for quantifying leakage by viding stable quantitative measures for the conformity
providing reasonable minimum to maximum esti- (or nonconformity) of observed and simulated CO2
mates of the detected CO2 mass. plume behavior. However, for an implementation of
An important requirement for a safe long-term this approach of quantitative assessment, reference
storage site operation, in particular in the postclosure data from real large scale sites as well as from generic
and pretransfer phase of the site, is demonstrating modeling studies are needed as a basis for translating
conformity between observed (from monitoring) quantitative assessment results into a catalog of criteria
and predicted (from reservoir simulations) reservoir for the decision making whether observed site behavior
behavior. This requirement is defined in carbon cap- is within the expectations or not.
ture and storage regulations, but to date no specific
instructions have been made available indicating Conclusions
under which conditions conformity between observed
and predicted behavior is achieved. An important Both seismic and electrical resistivity tomography
monitoring parameter assuring long-term stability of (ERT) measurements have been employed at the
the storage complex is pressure, and matching simu- Ketzin site for monitoring the injected CO2. In this
lated pressure history to observational data can be paper we have focused on the 3D time-lapse seismic
regarded as a valid approach to demonstrate confor- surveys and the crosshole ERT measurements.
mity. However, at various storage sites, pressure The seismic measurements allow the lateral extent of
measurements may not be available, or may be avail- the gaseous CO2 plume to be mapped in high spatial
able at only a few locations. The availability of resolution, while the crosshole ERT measurements
seismic monitoring data and full reservoir scale allow the CO2 distribution in the near wellbore
simulations have allowed an evaluation of the volume to be mapped in high temporal resolution.
performance conformity for the Ketzin storage site Both data sets show a predominantly northwesterly
involving the full CO2 plume extent to have been migration of the CO2 plume, interpreted as due to the
made. For the quantitative assessment of conformity heterogeneity of the reservoir. Comparison of the
between seismic observations and reservoir simula- geophysical data to fluid flow modeling of the injec-
tions, different performance parameters can be used tion process shows that the site is behaving in con-
describing geometrical properties of the CO2 plume formity to expectations. The amount of gaseous CO2
such as plume footprint area, plume volume, and lat- imaged is in line with the modeling given that there is
eral migration distance (Chadwick and Noy, 2015). a high dissolution rate (30%) and that there is signifi-
A direct comparison of the actual plume shapes can cant lateral spreading of the CO2 in thin layers that
be quantified using the “similarity index” (Lüth et al., have a thickness less than the seismic detection limit.
2015). Applying these performance parameters to the Based on the geophysical data, it is concluded that no
seismic data and modeling results for Ketzin revealed significant amounts of CO2 have leaked out of the
high quantitative mismatches between observations storage reservoir.
and simulation. For example, in terms of the plume
footprint area, seismic monitoring was able to detect Acknowledgments
only between 30% and 85% of the footprint area pre- The authors gratefully acknowledge the funding for
dicted by history-matched simulations (Lüth et al., the Ketzin project received from the European
2015). Comparing simulation results with a range of Commission (6th and 7th Framework Program),
minimum thickness thresholds showed that the seismic two German ministries (the Federal Ministry of 315
Economics and Technology and the Federal Ministry Berryman, J., and Milton, G. (1991). Exact results for
of Education and Research), and industry since 2004. generalized Gassmann’s equations in composite porous
The ongoing R&D activities are funded within the media with two constituents. Geophysics, 56:
1950–1960.
project COMPLETE by the Federal Ministry of
Education and Research. Further funding is received Chadwick, R. A., and Noy, D. J. (2015). Underground CO2
by VGS, RWE, Vattenfall, Statoil, OMV, and the storage: demonstrating regulatory conformance by
convergence of history-matched modeled and observed
Norwegian CLIMIT program. CO2 plume behavior using Sleipner time-lapse
seismics. Greenhouse Gas Science and Technology, 5:
305–322. DOI:10.1002/ghg.1488.
References Chadwick, R. A., Arts, R., and Eiken, O. (2005). 4D seismic
Alemu, B. L., Aker, E., Soldal, M., Johnsen, Ø., and quantification of a growing CO2 plume at Sleipner,
Aagaard, P. (2013). Effect of sub-core scale North Sea. Geological Society, London, Petroleum
heterogeneities on acoustic and electrical properties of Geology Conference series, 6, 1385-1399. https://doi
a reservoir rock: A CO2 flooding experiment of brine .org/10.1144/0061385
saturated sandstone in a computed tomography
Dvorkin, J., and Nur, A. (1998). Acoustic signatures of
scanner. Geophysical Prospecting, 61(1): 235–250.
patchy saturation. International Journal of Solids and
Asveth, P. (2009). Exploration rock physics. In K. Bjørlykke Structures, 35(34): 4803–4810.
(ed.), Petroleum geoscience: From sedimentary
Eid, R., Ziolkowski, A., Naylor, M., and Pickup, G. (2015).
environments to rock physics, Berlin/Heidelberg:
Seismic monitoring of CO2 plume growth, evolution
Springer-Verlag.
and migration in a heterogeneous reservoir: Role,
Baumann, G., Henninges, J., and Lucia, M. D. (2014). impact and importance of patchy saturation.
Monitoring of saturation changes and salt precipitation International Journal of Greenhouse Gas Control, 43:
during CO2 injection using pulsed neutron-gamma 70–81.
logging at the Ketzin pilot site. International Journal of
Ellis, R., and Oldenburg, D. (1994). Applied geophysical
Greenhouse Gas Control, 28: 134–146.
inversion. Geophysical Journal International, 116: 5–11.
Bergmann, P., and Chadwick, A. (2015). Volumetric
Förster, A., Norden, B., Zinck-Jørgensen, K., et al. (2006).
bounds on subsurface fluid substitution using 4D
Baseline characterization of the CO2 SINK geological
seismic time shifts with an application at Sleipner,
storage site at Ketzin, Germany. Environmental
North Sea. Geophysics, 80: B153–B165.
Geosciences, 13: 145–161.
Bergmann, P., Schmidt-Hattenberger, C., Kiessling, D.,
Förster A., Schöner R., Förster H., et al. (2010). Reservoir
et al. (2012). Surface-downhole electrical resistivity
characterization of a CO2 storage aquifer: The upper
tomography applied to monitoring of CO2 storage at
triassic Stuttgart Formation in the Northeast German
Ketzin, Germany. Geophysics, 77: B253–B267.
Basin. Marine and Petroleum Geology, 27(10): 2156–2172.
Bergmann, P., Ivandic, M., Norden, B., et al. (2014a).
Gassmann, F. (1951). Über die Elastizitgät poröser Medien.
Combination of seismic reflection and constrained
Vierteljahresschrift der Naturforschenden Gesellschaft
resistivity inversion with an application to 4D imaging
Zürich, 96: 1–24.
of the CO2 storage site, Ketzin, Germany. Geophysics,
79(2): B37–B50. Günther, T., Rücker, C., and Spitzer, K. (2006). Three-
dimensional modelling and inversion of DC resistivity
Bergmann, P., Kashubin, A., Ivandic, M., Lüth, S., and
data incorporating topography II: Inversion.
Juhlin, C. (2014b). Time-lapse difference static
Geophysical Journal International, 166: 506–517.
correction using prestack crosscorrelations: 4D seismic
image enhancement case from Ketzin. Geophysics, 79: Hill, R. (1963). Elastic properties of reinforced solids: Some
B243–B252. theoretical principles. Journal of the Mechanics and
Physics of Solids, 11: 357–372.
Bergmann, P., Diersch, M., Götz, J., et al. (2016). Review on
geophysical monitoring of CO2 injection at Ketzin, Huang F., Juhlin, C., Kempka, T., Norden, B., and Zhang, F.
Germany. Journal of Petroleum Science and Engineering, (2015). Modeling 3D time-lapse seismic response
139: 112–136. induced by CO2 by integrating borehole and 3D seismic
data: A case study at the Ketzin pilot site, Germany.
Bergmann, P., Schmidt-Hattenberger, C., Labitzke, T., et al.
International Journal of Greenhouse Gas Control, 36:
(2017). Fluid injection monitoring using electrical
66–77.
resistivity tomography: Five years of CO2 injection at
Ketzin, Germany. Geophysical Prospecting, 65(3): Huang, F., Bergmann, P., Juhlin, C., et al. (2017). The first
316 859–875. post-injection seismic monitor Survey at the Ketzin
pilot CO2 storage site: Results from time-lapse analysis. Lange, W. (1966). Geologisch-lagerstättenphysikalische und
Geophysical Prospecting, 66(1): 62–84. förder-technische Fragen bei der Erkundung des
Ivandic, M., Yang, C., Lüth, S., Cosma, C., and Juhlin, C. Untergrundspeichers Ketzin. Zeitschrift für
(2012). Time-lapse analysis of sparse 3D seismic data Angewandte Geologie, 12(1): 27–34.
from the CO2 storage pilot site at Ketzin, Germany. Martens, S., Liebscher, A., Möller, F., Würdemann, H.,
Journal of Applied Geophysics, 84: 14–28. Schilling, F., and Kühn, M. (2011). Progress report on
Ivandic, M., Juhlin, C., Lüth, S., et al. (2015). Geophysical the first European on-shore CO2 storage site at Ketzin
monitoring at the Ketzin pilot site for CO2 storage: New (Germany): Second year of injection. Energy Procedia,
insights into the plume evolution. International Journal 4: 3246–3253.
of Greenhouse Gas Control, 32: 90–105. Martens, S., Möller, F., Streibel, M., Liebscher, A., and the
Ivanova, A., Kashubin, A., Juhojuntti, N., et al. (2012). Ketzin Group. (2014). Completion of five years of safe
Monitoring and volumetric estimation of injected CO2 CO2 injection and transition to the post-closure phase
using 4D seismic, petrophysical data, core at the Ketzin pilot site. Energy Procedia, 59: 190–197.
measurements and well logging: a case study at Ketzin, DOI:10.1016/j.egypro.2014.10.366.
Germany. Geophysical Prospecting, 60(5): 957–973, Martens, S., Kempka, T., Liebscher, A., et al. (2015). Field
http://dx.doi.org/10.1111/j.1365–2478.2012.01045.x experiment on CO2 back-production at the Ketzin pilot
Juhlin, C., Giese, R., Zinck-Jørgensen, K., et al. (2007). 3D site. Energy Procedia, 76: 519–527. DOI:10.1016/j.
baseline seismics at Ketzin, Germany: The CO2 SINK egypro.2015.07.902.
project. Geophysics, 72: B121–B132. Mavko, G., Mukerji, T., and Dvorkin, J. (2003). The rock
Kashubin, A., Juhlin, C., Malehmir, A., Lüth, S., Ivanova, A., physics handbook: Tools for seismic analysis of porous
and Juhojuntti, N. (2011). A footprint of rainfall on media. Cambridge: Cambridge University Press.
land seismic data repeatability at the CO2 storage pilot Möller, F., Liebscher, A., Martens, S., Schmidt-
site, Ketzin, Germany. In 81st Annual International Hattenberger, C., and Streibel, M. (2014). Injection of
Meeting, Expanded Abstracts, Society of Exploration CO2 at ambient temperature conditions: Pressure
Geophysicists. and temperature results of the “cold injection”
Kazemeini, S. H., Juhlin, C., and Fomel, S. (2010). experiment at the Ketzin pilot site. Energy Procedia,
Monitoring CO2 response on surface seismic data: 63: 6289–6297.
A rock physics and seismic modeling feasibility study at Möller, F., Liebscher, A., and Schmidt-Hattenberger, C.
the CO2 sequestration site, Ketzin, Germany. Journal of (2016). Report on the dataset of the Brine Injection at
Applied Geophysics, 71(4): 109–124. the CO2 Storage Pilot Site Ketzin, Germany: Scientific
Kiessling, D., Schmidt-Hattenberger, C., Schütt, H., et al. Technical Report STR; 16/05, Potsdam: GFZ German
and the CO2SINK Group. (2010). Geoelectrical Research Centre for Geosciences. DOI:10.2312/GFZ.
methods for monitoring geological CO2 storage: First b103-16059.
results from cross-hole and surface-downhole Nakagawa, S., Kneafsey, T. J., Daley, T. M., Freifeld, B. M.,
measurements from the CO2SINK test site at Ketzin and Rees, E. V. (2013). Laboratory seismic monitoring
(Germany). International Journal of Greenhouse Gas of supercritical CO2 flooding in sandstone cores using
Control, 4: 816–826. DOI:10.1016/j.ijggc.2012.05.001. the Split Hopkinson Resonant Bar technique with
Kossow, D., Krawczyk, C., McCann, T., Strecker, M., and concurrent x-ray computed tomography imaging.
Negendank, J. F. W. (2000). Style and evolution of salt Geophysical Prospecting, 61(2): 254–269.
pillows and related structures in the northern part of Natatsuka, Y., Xue, Z., Garcia, H., and Matsuoka, T. (2010).
the Northeast German Basin. International Journal of Experimental study on CO2 monitoring and
Earth Science, 89: 652–664. quantification of stored CO2 in saline formations using
Kummerow, J., and Spangenberg, E. (2011). Experimental resistivity measurements. International Journal of
evaluation of the impact of the interactions of CO2– Greenhouse Gas Control, 4(2): 209–216.
SO2, brine, and reservoir rock on petrophysical Norden, B., and Frykman, P. (2013). Geological modelling
properties: A case study from the Ketzin test site, of the Triassic Stuttgart Formation at the Ketzin CO2
Germany. Geochemistry Geophysics Geosystems, 12: storage site, Germany. International Journal of
Q05–Q010. Greenhouse Gas Control, 19. DOI:10.1016/j.
Labitzke, T., Bergmann, P., Kiessling, D., and Schmidt- ijggc.2013.04.019.
Hattenberger, C. (2012). 3D Surface-downhole electrical Prevedel, B., Wohlgemuth, L., Henninges, J., Krüger, K.,
resistivity tomography data sets of the Ketzin CO2 Norden, B., and Förster, A. (2008). The CO2SINK
storage pilot from the CO2 SINK project phase. GFZ boreholes for geological storage testing. Scientific
Scientific Technical Report, 10(5) (available online). Drilling, 6: 32–37.
317
Rücker, C., Günther, T., and Spitzer, K. (2006). Three- storage site, Germany. Journal of Applied Geophysics,
dimensional modelling and inversion of DC resistivity 107: 171–186.
data incorporating topography I: Modelling. Stackebrandt, W., and Lippstreu, L. (2002). Zur
Geophysical Journal International, 166(2): 495–505. geologischen Entwicklung Brandenburgs.
Schmidt-Hattenberger, C., Bergmann, P., Kießling, D., In W. Stackebrandt and V. Manhenke (eds.), Atlas zur
Krüger, K., Rücker, C., and Schütt, H. (2011). Geologie von Brandenburg im Maßstab 1:1,000 000.
Application of a vertical electrical resistivity array Kleinmachnow, Germany: Landesamt für
(VERA) for monitoring CO2 migration at the Ketzin Geowissenschaften und Rohstoffe Brandenburg, 13–18.
site: First performance evaluation. Energy Procedia, 4: Torp, T., and Gale, J. (2004). Demonstrating storage of CO2
3363–3370. in geological reservoirs: The Sleipner and SACS
Schmidt-Hattenberger, C., Bergmann, P., Labitzke, T., et al. projects. Energy, 29(9–10): 1361–1369.
(2012). A modular geoelectrical monitoring system as Whitman, W., and Towle, G. (1992). The influence of elastic
part of the surveillance concept in CO2 storage projects. and density properties on the behavior of the
Energy Procedia, 23: 400–407. Gassmann relation. Log Analyst, 33(6): 500–506.
Schmidt-Hattenberger, C., Bergmann, P., Labitzke, T., and Wipki, M., Ivanova, A., Liebscher, A., et al. (2016).
Wagner, F. (2014). CO2 migration monitoring by Monitoring Concept for CO2 storage at the Ketzin pilot
means of electrical resistivity tomography (ERT): site, Germany: Post-injection continuation towards
Review on five years of operation of a permanent ERT transfer of liability. Energy Procedia, 97: 348–355.
system at the Ketzin pilot site. Energy Procedia, 63:
4366–4373. Yang, C., Juhlin, C., Enescu, N., Cosma, C., and Lüth, S.
(2010). Moving source pro-file data processing,
Schmidt-Hattenberger, C., Bergmann, P., Labitzke, T., modelling and comparison with 3D surface seismic
Wagner, F., and Rippe, D. (2016). Permanent crosshole data at the CO2SINK project site, Ketzin, Germany.
electrical resistivity tomography (ERT) as an Near Surface Geophysics, 8: 601–610.
established method for the long-term CO2 monitoring
at the Ketzin pilot site. International Journal of Zhang, F., Juhlin, C., Cosma, C., Tryggvason, A., and
Greenhouse Gas Control, 52: 432–448. Pratt, R. G. (2012). Cross-well seismic waveform
tomography for monitoring CO2 injection: A case study
Shi, J., Xue, Z., and Durucan, S. (2011). Supercritical CO2 from the Ketzin Site, Germany. Geophysical Journal
core flooding and imbibition in Tako sandstone: International, 189: 629–646, http://dx.doi.org/10.1111/
Influence of sub-core scale heterogeneity. International j.1365-246X.2012.05375.x
Journal of Greenhouse Gas Control, 5(1): 75–87.
Zimmer, M., Erzinger, J., and, Kujawa, C. (2011). The gas
Sopher, D., Juhlin, C., Huang, F., Ivandic, M., and Lüth, S. membrane sensor (GMS): A new method for gas
(2014). Quantitative assessment of seismic source measurements in deep boreholes applied at the CO2
performance: Feasibility of small and affordable seismic SINK site. International Journal of Greenhouse Gas
sources for long term monitoring at the Ketzin CO2 Control, 5(4): 995–1001.
318
Chapter
Time-Lapse Seismic Analysis of the CO2
18 Injection into the Tubåen Formation at

Snøhvit
Sissel Grude and Martin Landrø
Introduction the effect of gravity, will lead to vertical segregation of

the fluids, and a thin layer of CO2 will form at the top
At the Snøhvit field (Figure 18.1) in the Norwegian
of the formation (Nordbotten et al., 2005).
sector of the Barents Sea, the produced gas contains
A modified version of the time-lapse pressure–
between 5% and 8% carbon dioxide (CO2). The CO2 is
saturation inversion introduced by Landrø (2001) is
separated from the sales gas at the Melkøya LNG
applied to separate the pressure and saturation
plant, piped back and, in the period between 2008
effects caused by CO2 injection at the Snøhvit
and 2011, the CO2 was injected into the saline
field. A first-order approximation is used here to
Tubåen Formation (Fm.). This injection into the
describe the pressure effect instead of using
Tubåen Fm. was monitored by repeated 3D seismic
Landrø’s second-order approximation. The inverted
with data sets acquired in 2003 and 2009 in addition
pressure and saturation are then used to predict
to downhole pressure measurements. The faults and
expected time shifts that are compared to conven-
depositional environment limit the volume of the
tional time shifts estimated directly from the time-
Tubåen Fm., which can be regarded as semiclosed.
lapse seismic data. These can be seen as two inde-
The limited formation volume and lack of near well
pendent measurements of the time shifts and used
injectivity led to a pore pressure buildup and low
as a quality check of the inverted pressure and
injection rates. The injection into the Tubåen Fm.
saturation results.
was stopped in 2011. Today, CO2 is injected into the
CO2 injection laboratory experiments on two core
Stø Fm. at the Snøhvit field. In total, 1.6 Mtonnes of
plugs from the Tubåen Fm. were performed to evalu-
CO2 was injected into the Tubåen Fm. The main
ate the effect of salt precipitation on injectivity.
purpose of monitoring the injected CO2 was to
Analysis of bottom-hole-pressure (BHP) was per-
demonstrate safe storage and detect possible leakage.
formed to gain information about flow regimes acting
Monitoring will also give valuable information
in the reservoir. The CO2 saturation inverted from
regarding CO2 migration and reservoir pressure
time-lapse seismic was compared to an analytical
changes.
expression of CO2 plume saturation and thickness
CO2 migration in a saline aquifer is governed by
for viscous dominated flow (Nordbotten and Celia,
viscous, capillary, and gravitational fluid forces at an
2006).
early stage of injection (Lindeberg and Wessel-Berg,
1997). The dominant flow regime is site specific and
depends, among other factors, on injection rate, aqui-
The Snøhvit Field and the Tubåen
fer volume, reservoir injectivity, reservoir depth, for- Formation
mation pressure, and temperature. CO2 must The Tubåen Fm. is approx. 110 m thick and 2500 m
overcome the capillary entry pressure to migrate wide close to the injection well, compartmentalized by
into brine-saturated pores (Bryant et al., 2008). a series of E–W trending faults (Figure 18.2).
Viscous force is pressure driven and is dependent on The formation can be vertically subdivided into four
the injection overpressure (Class et al., 2009). sandstone units, Tubåen 1–4, separated by inter-
Supercritical CO2 is less dense and less viscous than bedded shale units that potentially act as barriers to
the resident formation brine. The mobility contrast vertical flow. Overall net to gross of the formation is
between supercritical CO2 and brine, combined with 0.7. Gamma ray, porosity, density, Vp, and Vs well logs
319
Figure 18.1 Location of the Snøhvit field, north of Finnmark, offshore Norway.
from the injection well are displayed in Figure 18.3. uppermost Tubåen 4 unit at a later stage. The fourth
The base of the reservoir is located at 2670 m true perforation interval did not improve the injectivity
vertical depth (TVD), corresponding to 2800 m mea- (Hansen et al., 2011). A production logging tool was
sured depth, as the well is slightly deviated. run in 2011. This showed that 80% of the injected CO2
The deepest 10–15 m thick Tubåen 1 sandstone unit migrated into the Tubåen 1 unit and the remaining
is interpreted to be distributary channels oriented in a 20% into the Tubåen 2 and 3 units (Hansen et al.,
north–south direction (Hansen et al., 2013) with por- 2013).
osity up to 20% and permeability up to 12 000 mD; the
lateral extent of the good sandstones is uncertain
(Eiken et al., 2011). Quartz cement reaches 20% in Fluid-Pressure Discrimination
parts of the formation. Initial reservoir temperature Landrø (2001) exploits the time-lapse near and far
was 95°C, brine salinity 14%, overburden pressure 56 offset seismic stacks as independent measurements,
MPa, and initial pore pressure 29 MPa, which and uses these to invert for fluid and pressure changes
increased to 35 MPa by the time the monitor seismic in the seismic amplitude. The method is based on the
survey was acquired. Smith and Gidlow (1987) linearized amplitude versus
CO2 was originally injected through three perfora- offset (AVO) equation for P-wave reflectivity as
tion intervals covering the Tubåen 1–3 sandstone a function of incidence angle, which estimates the
units (marked by the three red bars in Figure 18.3), reflectivity at a boundary surface in a two-layer
320 with an additional perforation interval covering the model. The change in P-wave reflectivity, ΔR (θ),
Time-Lapse Seismic Analysis of CO2 Injection
caused by pressure and fluid changes in the reservoir,

Time (ms)
1900 4D Difference 09-03 can be found by subtracting the reflectivity before any
injection (base) from the reflectivity after injection
(monitor), given the assumption that the changes in
P-wave velocity (α), S-wave velocity (β), and density
2100
(ρ) occur only in the reservoir layer. These measure-
ments, for pressure (ΔRP) and saturation (ΔRF),
change respectively:
0 500 1000 1500 2000m
2300 1 ΔρF ΔαF ΔαF
1:20480 ΔR ðθÞ ≈
F
þ tan2 θ ð18:1Þ
2 ρ α 2α
1900 Base survey 03
1 ΔαP 4β2 ΔβP ΔαP

ΔRP ðθÞ ≈ þ 2 sin 2 θ tan2 θ
2 α α β 2α
2100
ð18:2Þ
This combines to
2300 0 500 1000 1500 2000m

1:20480 Injection zone 1 ΔρF ΔαF ΔαP 4β2 ΔβP
ΔRðθÞ ≈ þ þ þ 2
2 ρ α α α β
Figure 18.2 Cross section through the reservoir zone. F
The uppermost picture shows time-lapse difference between the 1 Δα Δα P
seismic surveys. The lowermost picture is from the 2003 survey.

sin2 θ þ tan2 θ ð18:3Þ
2 α α
The top and base of the reservoir zone are shown by the black lines
and the injection well is the blue vertical line. The injection zone is ΔρF, ΔαF, ΔαP, and ΔβP are parameter contrasts
marked by the red arrow. from fluid and pressure changes in layer 1. ΔαP =
αPPost inj – αPPre inj. α = (α1 + α2)/2 etc. Subscripts 1
Figure 18.3 Gamma log (GR), porosity, bulk

density (RHOB), P-wave- and S-wave velocity
2.68 logs from the injection well. Vertical axis shows
measured depth in kilometers; the well is slightly
deviated. The reservoir zone is located between
2.68 and 2.8 km. Depth, Vp, Vs, and density of
2.7
brine substituted core measurements are
plotted as orange circles. Perforated intervals are
marked by red vertical lines.
2.72
Depth (km MD)
2.74
2.76
2.78
2.8
0 50 100150 0 0.1 0.2 0.3 2.2 2.4 2.6 2.8 3 4 5 1.5 2 2.5 3 3.5
GR Porosity RHOB (g/cc) Vp (km/s) Vs (km/s)
321
and 2 denote the two layers in the model. It is assumed expressed in terms of pressure and saturation changes
that the shear modulus is constant for fluid substitu- is obtained:
tion and that the density change from pressure 1
changes is negligible (see Landrø [2001] for further ΔRinj ðθÞ ≈ ðkρ ΔS þ kα ΔS þ lα ΔPÞ
2
details). The reflectivity between the base of the reser- !
voir and the underlying strata is estimated. This gives 1 4β2 lβ ΔP
ðkα ΔS þ lα ΔPÞ sin2 θ ð18:7Þ
an opposite sign on the reflectivity compared to the 2 α2
Landrø original method where the reflectivity
between the top of the reservoir and the caprock was With AVO intercept and gradient terms:
evaluated. Relative changes in P-wave velocity, S-wave
1
velocity, and density, due to pressure and saturation ΔR0 ≈ ðkρ ΔS þ kα ΔS þ lα ΔPÞ ð18:8Þ
2
changes, are determined from rock physics.
The relationship between saturation changes and P- !
wave velocity is estimated from Gassmann (1951) 1 4β2 lβ ΔP
ΔG ≈ ðkα ΔS þ lα ΔPÞ ð18:9Þ
fluid substitution. Landrø (2001) made a linear fit to 2 α2
the velocity–saturation curve that is appropriate when
the fluids have similar compressibilities. Meadows Relative change in two-way travel time induced by
(2001) expanded the velocity–saturation curve to variation in seismic velocity can be expressed as
a parabolic form, which more accurately represents
a range of fluid compressibilities such as for CO2 and ΔT Δα
≈ ≈ ðkα ΔS þ lα ΔPÞ ð18:10Þ
water, but introduces nonlinearity into the inversion. T α
In our method, Landrø’s original straight-line
approximation is used. Dry core measurements are (Landrø, 2002). T denotes the layer thickness in milli-
used to find a relationship between the pressure seconds and ΔT the change in layer thickness caused
changes and velocities. Landrø (2001) used a second- by injection, also in milliseconds. Reservoir thickness
order approximation to describe the observed pres- is assumed constant, i.e., no expansion due to the
sure–velocity dependency. Effective pressure in the injection. Inversion-predicted time shift is used to
Tubåen Fm. is 27 MPa, meaning that the relevant validate the inverted pressure and saturation 3D
section of the pressure–velocity curve is almost linear, cubes, by comparison with the time shift measured
Figure 18.6. CO2 injection increases the pore pressure from conventional cross-correlation.
but reduces the effective pressure to 21 MPa according The Top Fuglen Fm., a strong reflector situated
to well measurements. We find that a straight-line above the Tubåen Fm., is used to calibrate the reflec-
pressure approximation is still valid in this case, tivity due to the absence of coherent reflectors in the
which removes the second-order pressure term in reservoir zone. A maximum error of 13% is found
Landrø’s method. Expressions for the relative changes between calibrated reflectivity on the seismic data
in velocities and density from saturation and pressure and estimated reflectivity from the Smith and
changes can be written as Gidlow (1987) approximation after thin layer tuning
correction (Lin and Phair, 1993) (Figure 18.4). A thin
Δα layer tuning effect (Lin and Phair, 1993) is applied to
≈ kα ΔS þ lα ΔP ð18:4Þ
α the 10–15 m thick layer. Frequency matched near
(1–17°) and far (34–51°) offset seismic stacks are
Δβ used as input to invert for pressure and saturation
≈ kβ ΔS þ lβ ΔP ð18:5Þ
β effects. There will always be a tradeoff using far offset
data; it makes the measurements more independent,
Δρ but violates the assumptions behind the reflectivity
≈ kρ ΔS ð18:6Þ
ρ measured from the Smith and Gidlow (1987) approx-
imation of small angles (up to 30°). Also, for angles
By inserting Eqs. (18.4)–(18.6) into Eq. (18.3) and above 45°, the tan2(θ)-term in the Smith and Gidlow
assuming tan2 (θ) ≈ sin2 (θ) (small-angle approxima- (1987) approximation, (Eq. (18.3)) will deviate from
322 tion) an expression for the change in reflectivity the sin2(θ) term and cause erroneous time-lapse AVO
Figure 18.5 Relative changes in P-wave velocity from CO2 saturation

Figure 18.4 P-wave reflectivity between the Tubåen Fm. and change estimated from Gassmann fluid substitution. The red curve is
underlying unit from the Zoeppritz equation (black) and the Smith CO2 distributed in large patches in the pores. Blue curve is partial
and Gidlow approximation (red). Dashed lines is estimated reflec- patchy CO2 distribution according to Brie’s equation with exponent e =
tivity after thin-layer tuning correction. The tuning thickness is 13 ms. 3. The green curve is uniform CO2 distribution in the pores.
Calibrated reflectivity from near, mid and far seismic stacks is marked
by black circles.
available measurements from Snøhvit that might
trends. The error introduced by applying the Smith indicate whether the saturation is patchy or uniform.
and Gidlow (1987) approximation to the Zoeppritz The choice of Brie, e = 3, is based on geological knowl-
equation increases with offset and is at a maximum of edge (heterogeneities at several scales), and the results
from the Nagaoka data set. It should be stressed,
8% for the far offset seismic stack (Figure 18.4).
however, that the patchiness in the Nagaoka data set
is observed for frequencies well above seismic fre-
Rock Physics Input quencies. The second order approximation to the
Relative change in velocity caused by saturation saturation change, as introduced by Meadows
changes is estimated from Gassmann fluid substitu- (2001), would better describe the observed variation
tion (Gassmann, 1951). This model assumes that the in the relative velocity for uniform fluid distribution.
bulk modulus is sufficient to characterize the pore The straight line approximation introduces an error
fluid. Assessing this number is not straightforward of 0.5% to the data assuming a CO2 saturation of 0.3
in systems in which one has a mixture of fluids, for patchy fluid distribution in the pore space. This
which is the case when CO2 is injected into a saline error increases to 5% for partially patchy (Brie equa-
aquifer (Sen and Dvorkin, 2011). Effective bulk mod- tion) fluid distribution and 50% for uniform fluid
ulus of fluids distributed in large patches (patchy distribution, where the velocity–saturation curve
saturation) taking residual water saturation into becomes highly nonlinear and a straight line approx-
account can be described from Mavko et al. (2009) imation is no longer sufficient to fit the data.
and Sengupta (2000). Uniform fluid distribution in The pressure impact on the velocity is estimated
the pores is estimated from the harmonic average of from an average of three dry core measurements taken
the fluid bulk moduli. The difference between the from the Tubåen Fm. Depths, Vp, Vs, and density of
bounds can be large. Studies undertaken by Konishi these, after brine substitution, is plotted in orange in
et al. (2008) and Caspari et al. (2011) indicate the well logs in Figure 18.3. Effective pressure before
a partially patchy fluid distribution for the Nagaoka injection (27 MPa) is reduced to 21 MPa by the
data set where CO2 was injected into a saline aquifer, increase in pore pressure (Figure 18.6, upper figure).
measured at log frequencies (10–100 kHz). In our The reduction in pressure is comparable to a pressure
study the patchy and uniform fluid distributions are effect with a Hertz–Mindlin exponent of (1/12)
used as upper and lower bounds for the fluid bulk for both the P and S velocities (Figure 18.6, lower
moduli, and compared to a partial patchy fluid satura- diagram). The Hertz–Mindlin model (Mindlin,
tion described from Brie’s equation (Brie et al., 1995) 1949) describes the effective elastic properties of
using an exponent e ¼ 3 (Figure 18.5). There are no a precompacted granular rock of 36% porosity and
323
six grain-to-grain contacts. The Hertz–Mindlin model

gives a P- and S-wave velocity proportional to the
different stress raised to a power of (1/6). Hertz–
Mindlin is known not to fit real pressure data well
(Saul and Lumley, 2013), and is here used only to
describe the grain arrangement, which is an indication
of the stiffness of the rock. Duffaut and Landrø (2007)
showed that by increasing the number of grain-to-grain
contacts in the Hertz–Mindlin model they were able to
represent the rock consolidation. Field examples from
the Gullfaks and the Statfjord fields showed that Hertz–
Mindlin exponents of (1/10) and (1/15) correlated well
with dry core measurements. The higher amount of
grain-to-grain contacts at the Statfjord field was due
to more poorly sorted sandstone.
Results of Fluid-Pressure
Discrimination Process
Figure 18.6 (Upper) P-wave- (red) and S-wave (blue) velocity as Time-lapse difference between the base and monitor
a function of effective pressure estimated from dry core measure- surveys observed in Figure 18.2 is a combined effect
ments in the Tubåen Fm. (Lower) relative change in velocity from
the dry core measurements compared to the relative change in of both pressure and CO2 saturation increase.
velocity estimated with a Mindlin exponent of 1/12. The P- wave and The pressure effect dominates the time-lapse seismic
S-wave show approximately the same pressure sensitivity, and is except in the near well area where a fluid effect is
assumed similar. Effective pressure for base and monitor survey is
indicated in the figures. also observed (Figure 18.7). The calibration of the
reflectivity was performed on the bottom reservoir
Time (ms) ΔP(MPa)

20.00
2100
10.00
0.00
2200 −10.00
−20.00
0 500 1000 1500 2000m
−30.00
2300
ΔS(0−1)
2100 0.40
0.0
2200
−0.40
0 500 1000 1500 2000m
2300
Figure 18.7 Cross section through the reservoir zone for the inverted pressure measured in MPa (upper) and the inverted saturation
324 measured in fraction of 1 (lowermost). Negative ΔS is caused by sidelobe effects. Top and base reservoir zone are shown by the black lines and
the injection well is the blue vertical line. A maximum pressure of 15 MPa is measured and CO2 saturation of 0.22 for partially patchy fluid
distribution is measured for the lowermost horizon close to the well.
and is hence the horizon showing representative parts of the Tubåen Fm. are limited to the near well
numerical values. The pressure increase is strong in area as compared to the lowermost unit, where the
the near well area (up to 15 MPa) and decreases with effect reaches the faults. The variation between the
distance from the well. A continuous pressure buildup sandstone units is an indication of minor vertical
(4–5 MPa) terminates against the faults for the low- flow between the units. The formation was perforated
ermost sandstone unit. A CO2 saturation of 0.22 is in three different intervals (red vertical lines in
observed in the near well area for an intermediate Figure 18.3), and the observed anomalies show differ-
fluid distribution according to Brie’s equation with ent behavior for the layers.
an exponent e = 3. The anomalies for the uppermost Scaling of the input data would have an impact on
the magnitude of the saturation and pressure. Here it
is reasonable to assume some uncertainty in the scal-
dP (MPa) ing, as the predicted pressure increase from the BHP
3.50
data is maximum 10 MPa and the estimated value of
3.00
15 MPa exceeds that.
2.50
Root mean square (RMS) seismic amplitudes mea-
2.00
sured within the Tubåen Fm. (Figure 18.8) indicate
1.50
that the fluvial channels may restrict horizontal pres-
1.00
sure distribution. The sandstone channel(s) from
0.50
southeast to northwest act as high-permeability path-
0.00
ways for pressure, which terminate against the faults.
An amplitude response is observed in the western
corner. This response may be caused by pressure
propagation westward inside the reservoir block and
into the neighboring block, where the fault fades into
a ramp and is, therefore, not limiting the migration of
pressure any longer, or it may be an artifact.
1 000 m
Amplitude maps from the Fuglen Fm. above the reser-
voir exclude acquisition nonrepeatability as an effect
causing this anomaly. Changing the fluid distribution
Figure 18.8 RMS amplitude measured in the Tubåen Fm. on the in the pores, from uniform to patchy mixing, leads to
inverted pressure 3D cube. The CO2 injection well is shown by the
black circle, where a maximum pressure of 9 MPa is measured. scaling of the saturation effect, with no change in the
The north direction is indicated by the black arrow. lateral extent (Figure 18.9). RMS amplitude maps (for
Patchy Brie e=3 Uniform

dS (0−1) dS (0−1) dS (0−1)
0.22 0.11 0.06
0.20 0.10
0.05
0.16 0.08 0.04
0.06 0.03
0.12
0.04 0.02
0.08
0.02 0.01
0.04
0.00 0.00
0.00
1 000m
Figure 18.9 RMS amplitude measured in the Tubåen Fm. on the inverted saturation 3D cube, for fluids distributed in large patches (left),
according to the Brie theory with Brie e = 3 (center) and uniform fluid distribution (right) as described above the figures. The CO2 injection well
is shown by the black circle, where a maximum saturation of 0.37 (Patchy), 0.17 (Brie e = 3), and 0.10 (Uniform) measured in the well. The north 325
direction is indicated by the black arrow.
dP (MPa) dS (0−1) Figure 18.10 RMS amplitude mea-

6.00 0.15 sured in the Tubåen Fm. on the inverted
0.14 pressure (left) and saturation (right) 3D
5.00 0.12
4.00 cubes based on near and mid seismic
0.10
0.08
stack (uppermost cubes) and near and
3.00
far seismic stack (lowermost cubes). The
2.00 0.06
0.04
CO2 injection well is shown by the black
1.00
0.02
circle. The north direction is indicated
0.00 0.00
by the black arrow. The red circle
focuses on the amplitude effect in the
neighbor fault block to the injection
block, where the noise is far less for the
near-far seismic stacks (lowermost fig-
ures). The leakage between the inverted
pressure and saturation can be seen for
the near-mid seismic stacks (uppermost
dP (MPa) dS (0−1) figures).
3.50 0.11
0.10
3.00
2.50 0.08
2.00 0.06
1.50
0.04
1.00
0.50 0.02
0.00 0.00
1 000 m
the entire reservoir thickness) of the estimated pres- make offset inversion more stable. The reader should
sure and saturation change are expected to show the especially note the amplitude effect observed in the
extent of the fluid and pressure distribution. However, southwest corner on both the pressure and saturation
the actual numerical values are then reduced, since the amplitudes for the near and mid seismic stack. This
effects are spread over the entire reservoir thickness. anomaly is absent on the saturation inverted from the
The estimated pressure increase of 15 MPa in the well near and far offset seismic stacks (lower Figure 18.10).
is reduced to 9 MPa when RMS amplitudes are esti- A comparison of time shift estimated from cross-
mated for the entire reservoir thickness. Estimated correlation and inverted from the pressure and
CO2 saturation of 0.22 is reduced to 0.17 for RMS saturation can be seen in Figure 18.11. A partially
amplitudes with a partially patchy fluid distribution in patchy fluid distribution according to Brie e = 3 is
the pore space. Background RMS seismic amplitudes used as an input. The time shifts show a good correla-
measured above the Tubåen Fm. show a mean noise tion, both in lateral extent and in magnitude. Based on
level of SCO2 = 0.04 and a corresponding noise level of this, one can assume that the saturation and/or the
1.5 MPa in pressure change. pressure changes occurred in the lowermost layer.
Figure 18.10 shows pressure and saturation Using the RMS values and not assuming that peak
changes using near and mid seismic stacks (18–34°) values are valid for the interval gives a reasonable
as input to the inversion instead of near and far quantitative fit. This fit also indicates a fluid distribu-
seismic stacks; the remaining workflow is identical. tion somewhere between pure uniform and large
Clear leakage between the pressure and saturation 3D patches. An anomaly is observed in the neighboring
cubes can be seen. The leakage shows the importance reservoir segment (lower western corner) on the
of having a large separation between the seismic inverted pressure that is not visible on the time-shift
326 stacks to treat them as separate measurements and map. This indicates that the anomaly observed on the
ms
2.50
2.00
1.50
1.00
0.50
0.00
1 000 m
Figure 18.11 Estimated time shift from the inverted saturation and pressure (left) compared to time shift estimated from cross-correlation
(right). A cross-correlation window of 200 ms covering the reservoir zone is applied. The CO2 injection well is located in the middle of the
reservoir block where the maximum time-shift value is present. The north direction is indicated by the black arrows.
inverted pressure (marked by the red ring in error of a maximum of 8% at far offset seismic
Figure 18.10) is either caused by a very thin pressur- stack as compared to the Zoeppritz equation.
ized layer or that it is an artifact. A mean value is chosen to represent the entire
span of the near (1–17°) and far offset (34–51°)
Uncertainties seismic stacks. Other weighting would lead to
Uncertainties in the inverted parameters will depend on different zero-offset reflectivity and gradient
the uncertainty in the input parameters (Landrø, 2002). reflectivity 3D cubes.
A detailed quantitative study of the uncertainties is not • Linear fit to the velocity–saturation curve:
performed here, but the following qualitative uncer- The linear fit to the velocity–saturation curve as
tainties of some of the input parameters are listed: shown in Figure 18.5 is valid for a fluid distribution
• Reservoir thickness, T: The thickness of the between the uniform and patchy bounds up to SCO2
reservoir experiencing the time shift caused by = 0.3. No unique CO2 saturation can be found from
injection may not be proportional to the reservoir the relative change in velocity for a uniform fluid
thickness. Intervening shale units act as barriers to distribution in the pore space.
vertical flow so the CO2 may not have reached the • Uncertainties in zero-offset reflectivity and
top of the reservoir. The pressure front terminates gradient reflectivity are not evaluated here.
against impermeable shale units. The top reservoir In general, the zero-offset reflectivity is expected to
is impermeable, and considered as the maximum be fairly stable (approx. 2 × NRMS), while the
thickness. gradient reflectivity will be more uncertain (approx.
• Time shift, ΔT: Different cross-correlation 5 × NRMS) (David Lumley, pers. commun.).
windows result in different time shifts. The average Comparison between the fluid effect in Figure 18.9
and standard deviation from 11 different cross- and Figure 18.10 shows the importance of a wide
correlation windows is used in this study. separation between the near and far offset seismic
• Seismic 3D cubes for ΔR0 and ΔG: A small-angle stacks to have two independent measurements, and
approximation is used in the calibration of the reduce the leakage between the inverted saturation
input zero-offset reflectivity and gradient 3D and the pressure. The far offset seismic stack violates
cubes. The small angle approximation gives an the assumption of θ below 30° in the Smith and 327
Gidlow (1987) approximation. The deviation between Synthetic brine of similar salinity as the formation
the Smith and Gidlow approximation and the brine (14%), and with the main salt components,
Zoeppritz equation is 3–8% for the far offset seismic was then injected into the sample. The confining pres-
stack (Figure 18.4). sure (up to 56 MPa) and pore pressure (up to 28 MPa)
Pressure and fluid changes are expected to dom- were increased simultaneously to keep a fixed differ-
inate the time-lapse signature and are the only effects ential pressure (ΔP) of 28 MPa. Supercritical CO2 was
that are evaluated in this study. Other changes caused then injected at a rate of approx. 1 ml/min, until no
by injection, such as thermal fracturing, may also more brine was produced at the outlet. The chosen
influence the time-lapse signature. The injected CO2 injection rate was based on the capacity of the labora-
has a temperature of approx. 26°C, while the tempera- tory equipment, and does not represent the reservoir
ture in the formation is 95°C. The injected CO2 will condition. Finally, an unloading sequence was
cause a significant cooling of the rock that may lead to applied, and measurements of porosity and perme-
thermal fracturing, followed by a thermal expansion ability on CO2 dry samples at the residual water
of the CO2 plume as it heats up. saturation were performed. Reservoir temperature in
the Tubåen Fm. at Snøhvit is 95°C. The laboratory
Reservoir Injectivity experiment was first described by Grude et al. (2013a)
and is rephrased here for completeness.
Reservoir injectivity characterizes the ease with which
Precipitated salt reduced the porosity and per-
fluid can be injected into a geological formation with-
meability for both the horizontal (Q-274h) and
out pressure building up (Benson et al., 2005).
vertical (Q-274v) core plugs. The impact is greatest
The injectivity is primarily controlled by the perme-
in the vertical core plug (Figure 18.12). Preinjection
ability and available pore volume. Permeabilities
pressure sensitivity of the P-wave velocity is more
above 100 mD are recommended to ensure sufficient
significant than the post injection, especially for the
access to the pore space (Cooper, 2009). The perme-
horizontal core plug with pressure below 10 MPa
ability around the wellbore may be different from the
(Figure 18.12). A cycle of loading–unloading of the
formation permeability, due to “skin effects” caused
confining pressure was performed prior to injection
by drilling, pore clogging with drilling mud and che-
to reduce the impact of nonelastic deformation
mical reactions. A positive skin represents a zone of
during loading and unloading cycles on the vertical
reduced permeability around the wellbore, while
core plug. The reduction in porosity due to preci-
a negative skin represents increased permeability
pitated salt and grain rearrangement will lead to
(Hurst et al., 1969). Salt precipitation may clog pore
increased P-wave velocity. A porosity–velocity rela-
throats and reduce the porosity and permeability.
tionship from well log measurements was used
This effect is probably limited to the dry-out front,
to correct the laboratory measured velocity
a small radius around the injection well (Pruess
(Figure 18.12, magenta triangles). Porosity correc-
and Müller, 2009; Bacci et al., 2011). Weekly injection
tions make the pre- and postvelocity measurements
of a 90:10 mixture of methyl ethyl glycol and water
comparable in pressure sensitivity at ΔP above 20
(MEG) was implemented in the Tubåen Fm. to miti-
MPa (Grude et al., 2013b).
gate salt precipitation (Hansen et al., 2013).
Relative change in P-wave velocity (δVp/Vp) due to
pressure, saturation, and salt precipitation in these
Laboratory Measurements two experiments is shown in Figure 18.13 and
Supercritical CO2 was injected into two brine filled Table 18.1. Differential pressure was 27 MPa before
core plugs from the Tubåen Fm. to understand how CO2 injection. This was reduced to 17–21 MPa in the
precipitated salt could reduce the porosity and perme- time between the base and monitor seismic surveys.
ability, and impact the P-wave velocity. Experimental Observations from this laboratory data show that the
parameters were chosen to mimic the reservoir con- P-wave velocity is reduced by the reduced differential
ditions in the Tubåen Fm., given the capacity of the pressure (marked ΔVPressure in Figure 18.13). Fluid
laboratory equipment. The dry samples were pre- substitution from brine to CO2 also reduces the
heated to 60°C, before the loading pressure curve P-wave velocity (marked ΔVFluid in Figure 18.13).
was acquired, up to 28 MPa confining pressure. Salt precipitation is not affecting the P-wave velocity
328
5000 Figure 18.12 P-wave

velocities as a function of
differential pressure (ΔP)
estimated on horizontal (left)
4500
and vertical (right) plugs from
Q-274. The filled diamonds
are loading of the dry plug;
4000 the open diamonds are
unloading of the CO2-dried
Vp (m/s)
plug. Pink triangles are

3500 unloading measurements
corrected for porosity
Loading Loading reduction. Red circles are
3000 Brine brine-filled and blue are CO2-
φpre = 18.0 φpre = 18.5 Brine
−7% filled. Porosity and perme-
−5%
φpost = 17.1 CO2 φpost = 17.2 CO2 ability measured pre- and
postinjection are given in the
2500 kpre = 312 kpre = 263 figure insets (Grude et al.,
Unloading Unloading
−7% −17% 2013b).
kpost = 290 kpost = 216 Φ corr
Φ corr
0 10 20 30 0 10 20 30
Differential pressure (MPa)
Horizontal Vertical Figure 18.13 P-wave velocities as

4600 a function of differential pressure (ΔP)
estimated on horizontal (left) and vertical
(right) plugs from Q-274. Symbols have
ΔVSalt precip the same meaning as in Figure 18.12.
P-wave velocity variation caused by
4400 ΔVFluid
pressure, saturation, and salt precipita-
ΔVPressure tion is shown in the figure. (Modified
ΔVSalt precip
Vp (m/s)
from Grude et al., 2013b.)

ΔPInjection
4200
ΔVFluid
Loading
Brine ΔVPressure
4000 CO2
ΔPInjection
Unloading
20 22 24 26 28 20 22 24 26 28
ΔP (MPa)
in the horizontal core plug. It increases the P-wave A continuous rise in pressure is observed, with
velocity in the vertical core plug (marked ΔVSalt precip minor reduction in periods of shut-in. A rapid pres-
in Figure 18.13). The directional variations in P-wave sure increase occurs after 5000 hours of injection.
velocity caused by salt precipitation may be due to This pressure increase is interpreted as reduced injec-
uncertainties in the laboratory experiment or hetero- tivity caused by salt precipitation. The injectivity was
geneities at core scale. partly improved by weekly injection of MEG (Hansen
et al., 2013).
Analysis of fall-off data can give information
Fall-Off Pressure Measurements about flow regimes acting in the reservoir that can
A downhole pressure gauge installed 800 m above the be associated with reservoir geometry and flow
Tubåen Fm. monitored pressure change caused by barriers. Several points should be considered for
CO2 injection (Hansen et al., 2013). The pressure the analysis of fall-off tests. As the shut-in time
measurements can be seen in Figure 18.14. is much less than the injection time, the
329
Table 18.1 Relative change in P-wave velocities (δVp/Vp) Different flow regimes can be identified by char-
measured on core plugs in laboratory
acteristic patterns on a log-log plot of logarithmic
Range (m) (δVp/Vp) pressure derivative versus shut-in time, i.e., (dP/dln
(Δt)). Pressure derivatives are calculated from the fall-
Pressure, ΔP 2000 (to faults) –0.022a
off data using a differentiation algorithm. In general,
Saturation, ΔS 200–300 –0.028a the following flow regimes will occur:
Salt Maximum 10 0.033 (vertical
• Early time unit slope line (Wellbore storage)
precipitation, Δφ plug)
• Zero slope line (Radial flow regime)
a
Average of horizontal and vertical core plug. • Half slope line (Linear flow regime)
• Late time unit slope line (pseudo-steady-state flow
regime)
Wellbore storage affects the pressure immediately
Miller–Dyes–Hutchinson (MDH) method of analyz- after shut-in and is mainly governed by wellbore
ing fall-off data can be used (Bourdarot, 1998; hydraulics. The flow regime dominating the pressure
Johnson and Lopez, 2003). That is, the time function response after wellbore storage depends on the size of
for further analysis is considered to be the shut-in the reservoir and potential flow barriers. Radial flow
time. The pressure analysis technique should suffice toward the well is most likely to occur for open homo-
because of the relatively small pressure changes seen geneous reservoirs without flow barriers. Linear flow
during the pressure fall-off test. Furthermore, CO2 is indicates a semiclosed reservoir where faults, sand-
injected in the supercritical state, which makes stone channels, or other geological heterogeneities
liquid well testing analysis applicable to CO2 fall-off channelize the flow.
testing. A stable injection rate and long shut-in Pseudo-steady state flow indicates a closed system,
period are desirable to obtain high-quality pressure where the response from the well has encountered all
data for analysis. Four shut-in periods lasting for the boundaries around the well (Johnson and Lopez,
more than 250 hours were chosen based on these 2003). Boundary dominated flow regime is a combina-
criteria (Figure 18.14). The first shut-in period tion of linear and pseudo-steady-state flow regimes.
occurred during the high-pressure period after Coinciding pressure and pressure derivative
approx. 5000 hours of injection. Hence, the data are curves with close to unity slope were observed on
valuable in order to analyze the impact of precipi- logarithmic plots initially for all the shut-in periods.
tated salt on the injectivity and pressure behavior. This is interpreted as pressure measurements
Figure 18.14 (Upper) Pressure measured

800 m above the injection interval and
extrapolated to bottom-hole-pressure (BHP).
(Lower) Injection rate (Qinj) measured in the well.
The rapid increase in BHP observed after approx.
5000 hours of injection is likely caused by
reduced injectivity from precipitated salt. Four
shut-in periods are used in fall-off analysis. Base
survey was acquired prior to injection start. Time
of monitor seismic survey is marked.
330
affect the pressure measurements up to 30 hours after

101
shut-in. Considering the CO2 velocity in the reservoir
100
this corresponds to approx. 0.7 m and can be an
ΔP (bar)
indication of the dry-out radius where salt precipita-

10−1 tion reduced the injectivity.
Hansen et al. (2013) published an analysis of the
10−2 pressure fall-off data from the 3 months long shut
d(ΔP) /d(Δt) (bar/hr)
100.009 down in 2009 (corresponding to shut-in 3 in

Figure 18.14) from a production logging tool (PLT)
and the down-hole pressure gauge. Based on the PLT
falloff data, Hansen et al. (2013) found that a flow
100.004 barrier at about 3000 m from the well must be added,
100 101 102 103
in order to match the permanent gauge pressure
Δt (hrs) history and falloffs. The active and connected reser-
voir to the well then becomes rectangular in shape
Figure 18.15 Log-log plot of fall-off pressure (upper) and with no flow boundaries at 110 m, 3000 m, and
derivative of the pressure (lower, dΔP/dΔt) versus time for the four 110 m distance from the well. The shallow location
shut-in periods. ΔP and Δt corresponds to pressure and time
since shut-in. The colors correspond to the shut-in periods in of the downhole pressure gauge generated chal-
Figure 18.14. lenges for the fall-off analysis, and reservoir beha-
vior was recognized after 100 hour (Hansen et al.,
2013). The analysis performed here is based on the
early injection time, and the results should be trea-
dominated by wellbore effects. The wellbore storage ted with care owing to the location of the pressure
period is interpreted to gradually transform into gauge. Still, the shallow location of the pressure
boundary-dominated flow. A deviation from this gen- gauge is expected to influence all the shut-in periods
eral trend can be observed for the first shut-in period. equally, and valuable information can be gained by
This deviation is evident in a derivative plot (dΔP/ comparing these.
dΔt) (Figure 18.15). A unit slope line is observed on
pressure and pressure derivative plots for shut-in per-
iods 2–4 (Figure 18.15). This can be interpreted as
CO2 Saturation and Thickness
a pseudo-steady-state flow regime, where the pressure Prediction from Analytical Expression
boundaries (channel boundaries and faults) are felt This study focuses on the 0.4 Mtonnes of CO2 injected
immediately. The slope is steeper for the first shut- into the Tubåen 1 sandstone unit prior to the mon-
in during the initial 30 hours. This indicates a closed itoring survey in September 2009. Table 18.2 sum-
reservoir with even lower injectivity initially that marizes the reservoir parameters of the Tubåen 1
slows down the pressure propagation, and may be unit at the Snøhvit field, including the estimated
caused by flow from the formation behind the dry- mobility ratio and gravity factor. CO2 and brine
out front where the salt has reduced the injectivity. fluid properties are given in Table 18.3.
The slope declines approx. 30 hours after shut-in. Capillary entry pressure for two core plugs from
The steeper slope initially is not present for the the Tubåen 3 sandstone unit is 0.01 MPa and the
shut-in periods after MEG injection, which indi- capillary pressure is 1 MPa at residual brine saturation
cates that the low injectivity was resolved by injec- measured in the laboratory for these samples. The
tion of MEG. capillary pressure is low compared to the injection
The extent of the CO2 front between base and overpressure of a maximum of 6–10 MPa. No mea-
monitor seismic surveys (approx. 18 000 hours of surement of capillary pressure was performed on core
injection) is interpreted to be approx. 420 m in the plugs from the Tubåen 1 sandstone unit, but the
channel direction (Grude et al., 2013c). This length of capillary pressure is most likely negligible based on
the anomaly corresponds to horizontal CO2 migra- the measurements in the Tubåen 3 unit and the aver-
tion of approx. 0.023 m/hour. The low-permeability age permeability of 750 mD in the Tubåen 1 unit.
zone surrounding the injection well was interpreted to 331
Laboratory measurements of the vertical and
Table 18.2 Average reservoir properties, mobility ratio, and gravity factor for the Tubåen 1 sandstone unit
Thickness h 14 m
Porosity φ 0.19 fraction
Permeability k 750 mD
Residual brine saturation Sbr 0.13a fraction
Relative permeability of CO2 at Sbr kr,CO2 0.7b fraction
Injected volume Q(Mtonnes/s)× t(s) 0.4 Mtonnes
Mobility ratio λ (λ CO2/λ Brine) 5.65
Gravity factor Γ 0.6
a
Average of horizontal and vertical core, estimated from Sengupta (2000), taking the porosity and permeability into consideration.
b
No laboratory measurement available from the Tubåen 1 unit. The value is assumed from comparison with laboratory measurements of
the relative permeability on three composite cores from the Tubåen 3 sandstone unit at 2735 m measured depth (MD) and published
literature (Bennion and Bachu, 2008).
Table 18.3 Fluid properties estimated at reservoir pressure of contributes to the fluid flow (Eq. 14 in Nordbotten
27 MPa, temperature of 90°C, and brine salinity of 14%. et al. [2005]):
CO2 density, ρCO2 663 kg/m3 2πΔρgλb kB2
3 Γ¼ ð18:11Þ
Brine density ρBrine 1090 kg/m Qwell
CO2 viscosity μCO2 0.055 mPa s
where Δρ = ρbrine – ρCO2 is the difference in density
Brine viscosity μBrine 0.444 mPa s
between brine and CO2; g is the gravitational con-
The CO2 density is estimated from the Span and Wagner (1996) stant, λb = kr, b/μb is the mobility of brine in the
equation of state. CO2 viscosity is estimated from Scalabrin et al. formation, which depends on the relative permeabil-
(2006). Brine density is estimated from Batzle and Wang (1992).
Brine viscosity is estimated from Kestin et al. (1978).
ity (kr,b) and viscosity of the brine (μb ). The saturation
in the brine phase is 1 due to the assumptions of
immiscible fluids; this makes kr,b = 1 (Okwen et al.,
2010). B is the thickness of the formation, k is the
vertical permeability, and Qwell is the injection rate
horizontal permeability were performed on core plugs (Nordbotten et al., 2005). Nordbotten et al. (2005) and
from the sandstone units. These showed that the per- Nordbotten and Celia (2006) found that a lower cutoff
meability was approximately similar in both direc- value of 1 for the gravity factor is suitable to illustrate
tions. Gravitational forces are expected to be small when gravity forces exert an influence on the CO2
in this 10–15 m thick sandstone unit with vertical flow. Okwen et al. (2010) revised the cutoff value to
permeability of 750 mD. It is reasonable to assume 0.5, to within a 10% error. Based on this equation, one
that viscous forces govern the CO2 flow in the Tubåen observes that the influence of gravity increases with
1 sandstone unit. the thickness of the formation and vertical permeabil-
Nordbotten et al. (2005) and Nordbotten and ity, and decreases with increasing brine viscosity and
Celia (2006) developed an analytical expression that injection rate. The gravity factor is estimated to be
describes the shape of a CO2 plume when the capillary approx. 0.6, using Eq. (18.11), in the Tubåen 1 sand-
pressure is neglected. In this model, the CO2 is stone unit. This value is at the lower cutoff for gravity
injected into a porous formation that is confined influenced flow and it is reasonable to neglect this
above and below by impermeable rock formations. effect as CO2 migration is expected to be viscous
The expression is viscous dominated, with influence dominated.
from gravity. Their dimensionless gravity factor, Γ, Supercritical CO2 is less dense and less viscous
332 can be used to identify when the gravity force than the formation brine and therefore injected CO2
will migrate upward and along the overlying shale constant saturation within the plume. The CO2
owing to the density difference (Nordbotten and migrates along the top of Tubåen 1 sandstone unit
Celia, 2006). The mobility ratio between CO2 and and leads to a reduction in effective CO2 saturation
brine is defined as within the Tubåen 1 sandstone unit. According to this
λCO2 kr;CO2 μBrine model, the Tubåen 1 sandstone unit is filled with CO2
¼ ð18:12Þ to approx. 40 m from the injection well, before the
λBrine kr;Brine μCO2
CO2 layer thins due to vertical segregation. Maximum
where λCO2 and λBrine are the mobilities of the CO2 radius of the CO2 layer is approx. 750 m. The CO2
and brine respectively. Using estimates of the relative saturation inside the plume is assumed to be uni-
permeability and viscosities of the two fluids formly distributed, and equal to 0.87. The residual
(Table 18.3), a mobility ratio of 5.65 is derived in the brine saturation (Sbr) is estimated to Sbr = 0.13 for
Tubåen Fm. The mobility ratio is strongly dependent the Tubåen 1 sandstone unit. The reduction in effec-
on the viscosity for the Tubåen 1 sandstone unit, as tive CO2 saturation is due to the thinning of the layer.
the CO2 has a much lower viscosity than the forma-
tion brine at the reservoir pressure of 27 MPa, tem- Comparing Analytical Solution and
perature of 95°C, and brine salinity of 14%. A mobility Time-Lapse Seismic Data
ratio of 5.65 is relatively low compared to typical
A CO2 saturation (SCO2) of 0.87 is estimated within
mobility ratios of 5–20 for CO2 displacing brine
the plume from the analytical expression. This is
(Okwen et al., 2010).
much higher than the CO2 saturation of 0.22 esti-
The thickness and effective saturation of the CO2
mated close to the well from the time-lapse seismic
layer (Figure 18.16) are calculated from the similarity
fluid-pressure inversion assuming an intermediate
solutions described in Nordbotten and Celia (2006),
fluid distribution in the pore space. This discre-
with Eq. (18.11) as input. The reader is referred to
pancy between the analytical solution and time-
Nordbotten and Celia (2006) for further details.
lapse seismic data is most likely caused by the
The negligible capillary pressure will lead to a
linear approximation between the CO2 saturation
CO2 plume thickness (m)

0.86 14
SCO2 effective
0.43 7
0 0
0 100 200 300 400 500 600 700 800
CO2 plume radius (m)
100m
Analytical 2480
Below time lapse noise level
2470
Xline
X-Section 1
X-Section 2 2460
X-Section 3
X-Section 4 2450
3130 3120 3110
Inline
Figure 18.16 Effective CO2 saturation and CO2 plume thickness from the analytical solution (black) and four time-lapse seismic lines (blue,
green, gray, and red). The input parameters for the analytical solution are given in Tables 18.2 and 18.3. The maximum CO2 saturation
corresponds to 0.87 (1 – residual brine saturation). The time-lapse noise level is marked in the figure. The location of the seismic lines is shown 333
in the figure below. The north direction is indicated by the arrow.
and relative change in velocity caused by saturation CO2 plume than in the SE and NE directions. It is
(Figure 18.5). The linear approximation is valid up likely to assume that the CO2 saturation is close to
to approx. SCO2 = 0.1 for uniform fluid distribution the residual water saturation, and constant within
in the pore space and SCO2 = 0.3 for an intermedi- the plume from comparison between the analytical
ate fluid distribution in the pore space. The spatial expression and the AVO inverted seismic data.
extent of the inverted CO2 (Figures 18.7 and 18.9)
is still valid for higher saturations, but no unique
CO2 saturations can be found between approx. SCO2 Discussion
= 0.3–0.87 for an intermediate fluid distribution in Static input parameters for Vp/Vs, porosity, and pres-
the pore space. sure sensitivity are used in the estimate of pressure
The lack of a unique solution gives the possibility and fluid variations caused by CO2 injection in the
of much higher saturations than originally found in Tubåen Fm. Given the strongly heterogeneous nature
Grude et al. (2013c). Comparison of the CO2 layer of the reservoir, additional information on the varia-
thickness and effective CO2 saturation obtained from tion in reservoir microstructure will give valuable
the analytical expression and time-lapse seismic data input to the inversion. Averaging based on the prop-
can be made by assuming that the saturation obtained erties in the near well area of the lowermost injection
from the analytical expression is true, and scale the interval will introduce an error if the reservoir proper-
SCO2 from the time-lapse seismic according to this. ties change away from the well. Studies have shown
Comparison of the shape of the layer with distance that small uncertainties in the reservoir parameters,
from the injection well can then indicate if the such as porosity, can lead to large uncertainties in the
assumption of viscous dominated flow is valid, and inverted pressure and saturation (Trani et al., 2011;
if the CO2 saturation within the layer is close to the Shahraeeni, 2012). Still, the inverted time shift and
residual water saturation, as assumed in the analytical time shift estimated from cross-correlation reveal
expression. a striking correlation that is an indication that the
Comparison of the CO2 layer thickness and major changes occur in the lowermost reservoir
effective saturation from the analytical expression section.
and the time-lapse seismic is presented in The pressure sensitivity on velocity is expected to
Figure 18.16. Four cross sections are chosen for be low from dry core measurements, the depth of the
comparison, two in the channel direction (gray formation (2670 m), and the high degree of cementa-
and red) and two perpendicular to the channel tion. Still, a major pressure buildup is observed that
direction (blue and green). Location of the seismic extends to the faults; this is most likely due to pressur-
line is indicated in the figure. Background RMS ized water (Hansen et al., 2013). No pressure changes
seismic amplitudes measured above the Tubåen in the neighboring segments indicate that the faults
Fm. correspond to approx. 18% of the maximum are sealing, with pressure propagation westward into
saturation. This relates to SCO2 effective = 0.16 for the neighboring reservoir block where the fault fades
the analytical expression. A CO2 layer radius above into a ramp.
400 m is not detectable in the seismic data, this area Trani et al. (2011) showed that introducing
is shaded blue in Figure 18.16. The decrease in CO2 the second-order term for the saturation effect does
layer thickness and effective saturation inverted not improve the uncertainty in the gradient term.
from time-lapse seismic follow the same trend as The validity of combining time shift measured over
the analytical solution with distance from the injec- the reservoir thickness and the localized amplitude at
tion well, but decreases more rapidly. There are the base of the reservoir is questionable for reservoirs
many uncertainties in the input parameters that where the saturation and/or the pressure change
can explain the deviation. For instance, the mobility occur in the entire reservoir zone. This error is
ratio is strongly dependent on relative permeability. expected to be small in the Tubåen Fm., where most
A lower relative permeability would lead to a lower of the saturation and pressure changes occur in the
mobility ratio, and a more piston-like displacement deeper parts of the reservoir.
of the fluid front. The CO2 saturation in NW and The permeability reduction caused by salt preci-
334 SW directions can better describe the radius of the pitation was estimated to be –7% (horizontal core
plug) and –17% (vertical core plug) from laboratory not constant. However, a constant injection rate is
experiments. This reduction in permeability is not assumed in the model, based on the total amount of
sufficient to explain the near well plugging on the CO2 injected into the Tubåen 1 sandstone unit
pressure data at an early stage of injection between acquisition of the seismic surveys. Despite
(Figure 18.15). There may be several causes for this all the deviation from the assumptions made in the
discrepancy. Comparing laboratory experiments with analytical expression, there is a good agreement in
fluid flow at reservoir scale is not straightforward, and the shape of the CO2 layer radius found from the
the results from the laboratory measurements can at analytical solution and the time-lapse seismic
best indicate the effects observed at a larger scale. measurements.
The permeability in the Tubåen Fm. varies signifi- The viscosity ratio between brine and CO2 is
cantly. Heterogeneities in the reservoir and clogging approx. 8 under the pressure and temperature
of important pore throats may have a major impact on regimes of the Tubåen Fm. The CO2 is less viscous
the permeability that is not observable at core scale. than the brine, and hence has the highest mobility.
Also, the interplay among viscous, capillary, and grav- Based on this significant viscous fingering effects are
ity forces in the reservoir fluid flow cannot be repro- expected to occur and the effective CO2 saturation of
duced at laboratory scale. the formation is low. The viscous fingers are most
The analysis is limited to early injection times, likely below the resolution of the seismic data, and
when the primary trapping mechanisms for CO2 are are difficult to detect.
stratigraphic and structural trapping. CO2 dissolution
into the formation brine, evaporation of brine into the
dry CO2, capillary trapping, and mineral trapping Conclusions
mechanisms are expected to dominate at a later Pressure dominates the time-lapse signature with the
stage (Bachu et al., 2007; IPCC, 2005). The fraction exception of areas near the well, where both a pressure
of CO2 dissolution into formation water and brine and fluid effect is estimated. Varying the fluid distri-
evaporation into the dry CO2 is expected to be small at bution in the pore space leads to scaling of the fluid
an early stage of injection (Nordbotten and Celia, effect, not changes in the spatial distribution.
2006). Estimated pressure close to the well is 15 MPa for
Laboratory measurements of capillary pressure the lowermost horizon; this is reduced to approx.
on core samples from the Tubåen 3 sandstone unit 4–5 MPa close to the faults. Estimated fluid saturation
indicate that capillary pressure can be ignored. These varies from 0.06 to 0.22 close to the well for the low-
are samples taken at one location in a heterogeneous ermost horizon, depending on the fluid distribution
reservoir. Local heterogeneity may introduce capil- in the pores (uniform to patchy). There is a good
lary barriers that cannot be ignored and will influ- correlation between observed time shifts from cross
ence the reservoir fluid flow. The derivation of the correlation and time shifts estimated from inverted
analytical solution makes numerous assumptions pressure and saturation. This fit indicates a fluid dis-
and simplifications, but concedes that they are not tribution somewhere between pure uniform and large
particularly restrictive and are approximately valid patches.
for most plausible injection scenarios. In this study, Results from the laboratory measurements
a radial CO2 distribution around the injection well showed that precipitated salt reduced the porosity
has been assumed. The assumption is inaccurate in and permeability for both of the core plugs. The
this fluvial environment, where an elliptical shape P-wave velocity increased in the vertical core plug
better describes the CO2 layer distribution. In addi- (Q-274v), while it was not affected by salt precipita-
tion, CO2 at a temperature of 26°C is injected at tion for the horizontal core plug (Q-274h). Analysis of
Snøhvit, and heated to 95°C by the formation. This bottom-hole-pressure measurements (BHP) from
violates the assumption of constant temperature in periods of shut-in indicates a low-permeability zone
the analytical expression. Furthermore, CO2 injec- surrounding the well at early stage of injection.
tion in the Tubåen Fm. occasionally halted due to The low-permeability zone may be caused by precipi-
operational challenges at the Melkøya LNG plant tated salt reducing the injectivity. MEG was injected
(Hansen et al., 2013). Hence, the injection rate was weekly to mitigate salt precipitation. 335
The distribution and saturation of CO2 in the References

Tubåen 1 sandstone unit have been estimated from
an analytical expression. The analytical expression Bacci, G., Korre, A., and Durucan, S. (2011).
An experimental and numerical investigation into the
assumes that capillary pressure can be ignored, and
impact of dissolution/precipitation mechanisms on
that the fluid flow is viscous dominated. Furthermore, CO2 injectivity in the wellbore and far field regions.
it supposes the CO2 distribution inside the plume is International Journal of Greenhouse Gas Control, 5:
uniform and equal to 1 – Sbr = 0.87, while the reduc- 579–588.
tion in effective CO2 saturation is due to the thinning Bachu, S., Bonijoly, D., Bradshaw, J., et al. (2007). CO2
of the layer. The influence of gravity has been shown storage capacity estimation: Methodology and gaps.
to be negligible on the reservoir CO2 flow. Analytical International Journal of Greenhouse Gas Control, 1:
results suggest CO2 fills the entire sandstone unit up 430–443.
to approx. 50 m away from the injection well before Batzle, M., and Wang, Z. (1992). Seismic properties of pore
the plume is reduced to a thin wedge that propagates fluids. Geophyics, 57: 1396–1408.
along the top of the formation. The maximum radius Bennion, D. B., and Bachu, S. (2008). Drainage and
of the CO2 layer from the analytic expression is 750 m, imbibition relative permeability relationships for
with 400 m likely to be observable above the time- supercritical CO2/brine and H2S/brine systems in
lapse noise level. intergranular sandstone, carbonate, shale, and
CO2 layer thickness and effective CO2 saturation anhydrite rocks. SPE Reservoir Evaluation &
Engineering, 11: 487–496.
obtained from the analytical expression and time-
lapse seismic data are in good agreement. This is an Benson, S. M., Cook, P., Anderson, J., et al. (2005).
indication of viscous dominated CO2 flow and that Underground geological storage. IPCC Special Report
on Carbon Dioxide Capture and Storage, Chapter 5:
SCO2 is close to the residual water saturation. The Intergovernmental Panel on Climate Change.
influence of gravity is negligible on the reservoir
Bourdarot, G. (1998). Well resting: Interpretation methods.
CO2 flow. The maximum observable CO2-layer thick-
Paris: Technip Publications.
ness inverted from time-lapse seismic is on average
405 m in the channel direction, and 273 m perpendi- Brie, A., Pampuri, F., Marsala, A. F., and Meazza, O. (1995).
Shear sonic interpretation in gas-bearing sands. SPE,
cular to the channel direction.
30595: 701–710.
Bryant, S. L., Lakshminarasimhan, S., and Pope, G. A.
Acknowledgments (2008). Buoyancy-dominated multiphase flow and its
effect on geological sequestration of CO2. SPE Journal,
We want to acknowledge Statoil and their Snøhvit 13: 447–454.
license partners Petoro, Total E&P Norge, GDF
Caspari, E., Müller, T. M., and Gurevich, B. (2011). Time-
SUEZ E&P Norge, and RWE Dea Norge for permis- lapse sonic logs reveal patchy CO2 saturation in-situ.
sion to use their data. The Norwegian Research Geophysical Research Letters, 38: L13301.
Council is acknowledged for support to the BIGCCS
Class, H., Ebigbo, A., Helmig, R., et al. (2009). A benchmark
center for environmentally friendly Energy Research study on problems related to CO2 storage in geologic
(FME) and to the ROSE consortium at NTNU. Jack formations. Computational Geosciences, 13: 409–434.
Dvorkin, Gary Mavko, Anthony Clark, Tiziana
Cooper, C. (2009). A technical basis for carbon dioxide
Vanorio, and the rest of the SRB group at Stanford storage. Energy Procedia, 1: 1727–1733.
University are acknowledged for their input and host-
Duffaut, K., and Landrø, M. (2007). Vp/Vs ratio versus
ing Sissel Grude during the research exchange to differential stress and rock consolidation:
Stanford. The laboratory measurements were done A comparison between rock models and time-lapse
at Stanford. Bård Osdal, Olav Hansen, Philip AVO data. Geophysics, 72: C81–C94.
Ringrose, Ola Eiken, Ashkan Jahanbani, Hossein Eiken, O., Ringrose, P., Hermanrud, C., Nazarian, B.,
Mehdi Zadeh, Jan Martin Nordbotten, Erik Torp, T. A., and Høier, L. (2011). Lessons learned from
Lindeberg, Jon Kleppe, James C. White, Rune 14 years of CCS operations: Sleipner, In Salah and
Martin Holt, Per Avseth, Ole Torsæter, Alexey Snøhvit. Energy Procedia, 4: 5541–5548.
Stovas, and Børge Arntsen are all acknowledged for Gassmann, F. (1951). Uber die Elastizitat poroser Medien.
336 direct input and discussions that made this chapter Vierteljahrsschrift der Naturforschenden Gesselschaft,
possible. 96: 1–23.
Grude, S., Clark, A., Vanorio, T., and Landrø, M. CO2. Energy Conversion and Management, 38
(2013a). Changes in the rock properties and (Supplement), S229–S234.
injectivity due to salt precipitation on the Snøhvit Mavko, G., Mukerji, T., and Dvorkin, J. (2009). The rock
CO2 injection site. Trondheim CCS Conference, physics handbook: Tools for seismic analysis of porous
June 4–6, Trondheim. media. Cambridge: Cambridge University Press.
Grude, S., Dvorkin, J., Clark, A., Vanorio, T., and Meadows, M. A. (2001). Enhancements to Landro’s method
Landrø, M. (2013b). Pressure effects caused by CO2 for separating time-lapse pressure and saturation
injection in the Snøhvit Field. First Break, 31: 3. changes. In 71th Annual International Meeting, Society
Grude, S., Landrø, M., and Osdal, B. (2013c). Time-lapse of Exploration Geophysicists, Expanded Abstracts,
pressure–saturation discrimination for CO2 storage at 1652–1655.
the Snøhvit field. International Journal of Greenhouse Mindlin, R. D. (1949). Compliance of elastic bodies in
Gas Control, 19: 369–378. contact. ASME Journal of Applied Mechanics, 16:
Hansen, O., Eiken, O., Østmo, S., Johansen, R. I., and 259–268.
Smith, A. (2011). Monitoring CO2 injection into Nordbotten, J. M., and Celia, M. A. (2006). Similarity
a fluvial brine-filled sandstone formation at the Snøhvit solutions for fluid injection into confined aquifers.
field, Barents Sea. In 81th Annual International Journal of Fluid Mechanics, 561: 20.
Meeting. Expanded Abstracts, Society of Exploration
Geophysicists, 4092–4096. Nordbotten, J., Celia, M., and Bachu, S. (2005).
Injection and storage of CO2 in deep saline
Hansen, O., Gilding, D., Nazarian, B., et al. (2013). aquifers: Analytical solution for CO2 plume
Snøhvit: The history of injecting and storing 1 Mt CO2 evolution during injection. Transport in Porous
in the fluvial Tubåen Fm. Energy Procedia, 37: Media, 58: 339–360.
3565–3573.
Okwen, R. T., Stewart, M. T., and Cunningham, J. A. (2010).
Hurst, W., Clark, J. D., and Brauer, B. (1969). The skin effect Analytical solution for estimating storage efficiency of
in producing wells. Journal of Petroleum Technology, 21: 7. geologic sequestration of CO2. International Journal of
IPCC. (2005). Special report on CO2 capture and storage. Greenhouse Gas Control, 4: 102–107.
Cambridge: Cambridge University Press. Pruess, K., and Müller, N. (2009). Formation dry-out from
Johnson, K., and Lopez, S. (2003). The nuts and bolts of CO2 injection into saline aquifers. 1.Effects of solids
falloff testing. Washington, DC: United States precipitation and their mitigation. Water Resources
Environmental Protection Agency. Research, 45: W03402.
Kestin, J., Khalifa, H. E., Abe, Y., Grimes, C. E., Saul, M. J., and Lumley, D. E. (2013). A new velocity–
Sookiazian, H., and Wakeham, W. A. (1978). Effect pressure–compaction model for uncemented
of pressure on the viscosity of aqueous sodium sediments. Geophysical Journal International.
chloride solutions in the temperature range 20–150 DOI:10.1093.
degree C. Journal of Chemical & Engineering Data, Scalabrin, G., Marchi, P., Finezzo, F., and Span, R. (2006).
23: 328–336. A reference multiparameter thermal conductivity
Konishi, C., Azuma, H., Nobuoska, D., Xue, Z., and equation for carbon dioxide with an optimized
Watanabe, J. (2008). Estimation of CO2 saturation functional form. Journal of Physical and Chemical
considering patchy saturation at Nagaoka. In 70th Reference Data, 35: 1549–1575.
Conference & Exhibition, EAGE, Extended Abstract, Sen, A., and Dvorkin, J. (2011). Fluid substitution in
I018. gas/water systems: Revisiting patchy saturation.
Landrø, M. (2001). Discrimination between pressure and In 81th Annual International Meeting, Society of
fluid saturation changes from time-lapse seismic data. Exploration Geophysicists, Expanded Abstracts,
Geophysics, 66: 836–844. 2161–2165.
Landrø, M. (2002). Uncertainties in quantitative time-lapse Sengupta, M. (2000). Integrating rock physics and
seismic analysis. Geophysical Prospecting, 50: 1–12. flow simulation to reduce uncertainties in seismic
Lin, T. L., and Phair, R. (1993). AVO tuning. In 63th reservoir monitoring. PhD thesis, Stanford
Annual International Meeting, Society of University.
Exploration Geophysicists, Expanded Abstracts, Shahraeeni, M. S. (2012). Effect of lithological uncertainty
727–730. on the timelapse pressure-saturation inversion. In 74th
Lindeberg, E., and Wessel-Berg, D. (1997). Vertical Conference & Exhibition Incorporating SPE EUROPEC
convection in an aquifer column under a gas cap of EAGE, Extended Abstract, Y045.
337
Smith, G., and Gidlow, P. M. (1987). Weighted stacking for 800 MPa. Journal of Physical and Chemical Reference
rock property estimation and detection of gas. Data, 25: 1509–1596.
Geophysical Prospecting, 35: 993–1014. Trani, M., Arts, R., Leeuwenburgh, O., and Brouwer, J.
Span, R., and Wagner, W. (1996). A new equation of state (2011). Estimation of changes in saturation and
for carbon dioxide covering the fluid region from the pressure from 4D seismic AVO and time-shift
triple-point temperature to 1100 K at pressures up to analysis. Geophysics, 76: C1–C17.
338
Chapter
Illinois Basin–Decatur Project
19 Robert A. Bauer, Robert Will, Sallie E. Greenberg, and Steven G. Whittaker
Introduction activity, repeat Reservoir Saturation Tool (RST)1 log-

ging measurements, and repeat surface and borehole
The Illinois Basin–Decatur Project (IBDP) is an inte-
seismic surveys have been used to monitor the devel-
grated bioenergy carbon capture and geological storage
opment of pressure and CO2 saturation throughout
(BECCS) project conducted at Archer Daniels Midland
the project. In addition to subsurface monitoring and
Company’s (ADM) corn processing plant in Decatur,
characterization, the IBDP MVA program included
Illinois, USA (Figure 19.1). Over three years approx.
extensive soil flux, atmospheric monitoring, and
1000 tonnes/day of carbon dioxide (CO2), obtained
shallow groundwater monitoring activities. A unique
from ethanol production at the ADM plant, was com-
challenge related to this site is that it is at a very active
pressed, dehydrated, sent along a 1.9-km pipeline, and
industrial site that impacts significantly on monitor-
injected into the Mt. Simon Sandstone 2.14 km deep in
ing activities and strategy. This chapter will focus
the Illinois Basin. Injection took place
on the geophysical aspects of site characterization
from November 17, 2011 to November 26, 2014, with
and monitoring conducted at the IBDP from 2007 to
999 215 tonnes of supercritical CO2 injected and geo-
2017 (Figure 19.2).
logically stored (Greenberg et al., 2017).
The IBDP is led by the Illinois State Geological
Survey (ISGS) through the Midwest Geological Geological Setting and Site
Sequestration Consortium (MGSC) that includes Characterization
partners Schlumberger Carbon Services, Trimeric Geological site characterization at the IBDP is based
Corporation, ADM, and other research organizations. on examination of more than 250 m of whole core,
The MGSC is one of seven regional partnerships geophysical well logs, two-dimensional (2D) and
funded by the U.S. Department of Energy–National three-dimensional (3D) seismic, and a range of tests
Energy Technology Laboratory (DOE–NETL) to and analyses (Freiburg et al., 2014). The primary
build capacity and gain experience in CCS through injection target of the IBDP is the Cambrian Mt.
regional characterization, pilot studies, and demon- Simon Sandstone, an extensive formation that under-
stration projects. In addition to the injection opera- lies much of the Midwestern United States, which is
tions, the MGSC has deployed a full range of surface also used regionally for geological storage of natural
and subsurface monitoring verification and account- gas. The Mt. Simon reaches its maximum thickness of
ing (MVA) technologies throughout each of the pro- approx. 790 m near the IBDP site, and is divided
ject phases (preinjection, injection, and postinjection) lithostratigraphically into Lower, Middle, and Upper
to evaluate their effectiveness for use at long-term, sections. These major sections are further divided into
large-scale CCS projects, including geophysical units based on depositional facies interpreted using
methods. core examination and downhole geophysical logging
The IBDP site includes one injection well (CCS1), results (Figure 19.3), and include fluvial braided
one deep monitoring well (VW1), one dedicated river, floodplain, and alluvial plain deposits; eolian
geophysical well (GM1), and a variety of near (windblown deposit) sandsheet, dune, and interdune
surface monitoring wells and equipment. For MVA, deposits; and shallow marine deposits. The injection
a combination of continuous pressure measurements of CO2 took place within a single lithological unit of
from the injection and multilevel monitoring well,
continuous passive monitoring of microseismic 1
339
Mark of Schlumberger.
R. A. Bauer, R. Will, S. E. Greenberg, and S. G. Whittaker
Figure 19.1 Map of the IBDP site with the location of

wells. The CCS1, VW1, and GM1 wells are infrastructure
for IBDP. CCS2, VW2, and GM2 are similar well
VW2
configurations installed for the Illinois Industrial Carbon
Capture and Storage (ICCS) Project.
GM2
CCS2
~800 m
VW1
GM1
CCS1
NORTH
the Lower Mt. Simon Sandstone between 2129 m and siltstone. Together the Mt. Simon and Eau Claire
2138 m, where porosity and permeability range form the storage complex at the IBDP.
18–25% and 40–380 mD, respectively; the reservoir Pre-Mt. Simon strata (informally known as the
interval overall has an average porosity of about 20% Argenta Formation) are in disconformable contact
and an average permeability near 200 mD. with the lower Mt. Simon and, although they are
The Mt. Simon is overlain by the Eau Claire lithologically similar to the Mt. Simon, they have
Formation, a 150 m thick impermeable package that distinct depositional character and are generally
hydraulically isolates the Mt. Simon from overlying well cemented and compacted throughout, result-
strata (Palkovic, 2015). A 70-m shale unit in the lower ing in significantly lower porosity and permeability
part of the Eau Claire forms a highly effective seal to (Figure 19.3). The Argenta deposits comprise the
vertical fluid movement, and the upper portion of the basal beds in the sedimentary succession at IBDP,
340 Eau Claire is dense limestone with thin stringers of and are in sharp, nonconformable contact with the
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Injection and
Groundwater Drilling Begins Compression Monitoring Well Startup
Monitoring Injection Well Facility Mechanical Construction Operation
Begins CCS#1 Completion and Completion
Pre-Injection: Design Pre-Injection: Infrastructure and Facility Construction Injection Post Injection
Project
and Engineering
Activities
Seismic Initial 2D 3D VSP 3D Surface 3D VSP 3D VSP 3D VSP 3D VSP
Monitoring Surface Seismic Baseline 1 Seismic Baseline 2 Monitor 1 Monitor 2 Monitor 3 Monitor 4
Survey extended for ICCS
Activities
3D Surface 3D VSP 2D Surface 3D Surface
Seismic Baseline 2 Seismic Survey Seismic Monitor 1
Baseline 1 Monitor 1 IBDP Final
(ICCS initial)
Continual Microseismic Monitoring
2D Surface Seismic 3D VSP 3D Surface Seismic
Figure 19.2 Geophysical program timeline.

Figure 19.3 Cross section of the IBDP study site. CCS1–VW1 cross section showing the Precambrian crystalline basement, Argenta Formation
(Pre-Mt. Simon), Mt. Simon Sandstone, and part of the Eau Claire Formation.
Precambrian basement. The Argenta can contain

angular clasts of underlying basement rhyolite, and Subsurface Configuration
regional seismic data indicates more than 500 m topo- The subsurface components of the IBDP geophysical
graphic relief may exist on the Precambrian basement site characterization and monitoring network are
surface prior to Phanerozoic deposition (Leetaru and deployed in three multipurpose wells; the injector
McBride, 2009). At the IBDP site, the Precambrian CCS1 (Figure 19.5), the deep verification and monitor-
crystalline basement is layered rhyolite over grano- ing well VW1 (Figure 19.5), and the geophysical mon-
diorite over granite, as shown by sidewall cores and itoring well GM1 (Figure 19.6). In addition to its
geophysical logs (Bauer et al., 2016). The basement primary function of injecting supercritical CO2 into
rock can be highly fractured showing mineralization, the Lower Mt. Simon, CCS1 hosts downhole pressure
slight alterations along the cracks, or a combination of and temperature gauges, primary elements of the IBDP
both (Haimson and Doe, 1983). passive seismic monitoring array, and was an access for
The bedrock strata in the region of the IBDP site repeat geophysical logging, all of which are key compo-
exhibit a slight dip of approx. 1° to the southeast, and nents of the MVA program. The VW1 well, although
the closest structures to the site (Figure 19.4) are not hosting seismic measurements, provided 11 pres-
342 minor anticlines with small structural closure about sure measurement and fluid sampling ports throughout
40 km south and 30 km north. the geological column extending from the Precambrian
0 60 mi
0 100 km
MI
h
rc
lA
n ta
MN
ti ne
Wis
n
SD co
ns
con
a
Tr
sin
WI MI
Arc
h
MICHIGAN
BASIN
IA h
Ar c
NE Ka
iver
ch
FOREST CITY nka
Ar
kee
pi R
y
BASIN Arc
la
h
nd
ip
Fi
siss
OH
ch
IL IBDP & IN
Mis
ri Ar
ICCS
Cin c
sou
ILLINOIS
inna
Mis
KS BASIN
t i Ar
NE
WV
ch
MO Ozark
Dome KY
New Madrid Rough Creek VA
Rift System N
Graben SI
BA
Pa N
if t
sc il l e s IA
R
ola in
edge of Mississippi
Ar c s hv n ta ACH
OK Embayment of Gulf h Na om
e ou L
M PPA NC
Coastal Plain ot TN D A
l fo
ee
ian
ARKOMA BASIN AR R
ch
la
Ouachita SC
pa
Mountains Ap
ic B c hita
BLACK WARRIOR
elt
BASIN
a
Oro on - Ou
TX AL GA
margin of
gen
orogenic belt
r at h
LA
MS
Ma
Figure 19.4 Regional fault map. The map shows regional faults with the location of the Illinois Basin (pink shaded area) and the IBDP site
(blue star).
2
The VW1 well was originally completed with a Westbay
to the Underground Source of Drinking Water pressure/temperature and fluid sampling system. At the
(USDW).2 In addition, VW1 also was accessed time of this writing, the VW1 well has been recompleted
for repeat geophysical logging. The GM1 well, with a different monitoring system not described herein 343
originally intended to facilitate repeat 3D vertical (section “Pressure Monitoring”).
(a) (b)
Injection Well In-Zone Monitor Well
CCS1 VW1
0 ft
0 ft
Shallow Water Zones
Shallow Water Zones
(61 m) 200 ft thickness 200 ft (61 m)
Surface (115 m) 377 ft Surface

(108 m) 355 ft
8", 54.5#/ft, J-55 to
20", 94#, BTC, 26" bit
length 353 ft (108 m) 377 ft (115 m)
Intermediate Csg Intermediate Csg

13 ³⁄8", 59.5# & 66.2#, J55, 9 5⁄8", 40.5#/ft,
BTC, 17 ½" bit N-80 to 5,322 ft (1,622 m)
length 5,339 ft (1,627 m)
(636 m) 2,088 ft
New Albany Shale (636 m) 2,088 ft
(675 m) 2,214 ft thickness 126 ft (38 m) New Albany Shale
(675 m) 2,214 ft thickness 126 ft (38 m)
(796 m) 2,611 ft
Maquoketa Shale (796 m) 2,611 ft
(859 m) 2,817 ft thickness 206 ft (63 m) Maquoketa Shale
Oilfield Tbg
2 / ", 6.5#/ft,
6.5# J-55
from 4,745 ft to
(997 m) 3,270 ft surface (1,446 m)
St. Peter Sandstone
(1,055 m) 3,477 ft thickness 207 ft (63 m)
Production Csg
9 5⁄8", 40#, N80, 12 ¼" bit
length 5,272 ft (1,607 m)
Tubing
4 ½", 12.6#, 13CRL80
length 6,363 ft (1,939 m)
Geophone #3 @ 4,925 ft
Reactive Element packers

with perforation below.
(1,538 m) 5,047 ft
BH gauges (green) and BH
(1,627 m) 5,339 ft (1,538 m) 5,047 ft sliding sleeve (red).
Eau Claire main
thickness 498 ft (152 m)
Eau Claire main
Geophone #2 @ 5,743 ft (1,690 m) 5,545 ft
Reactive Element
packers.
Packer
Geophone #1 @ 6,137 ft
BH gauges (green) and BH
Mt. Simon Sandstone sliding sleeve.
Pressure/Temperature thickness 1,620 ft (494 m)
Gauge Mandrel @ 6,325 ft Mt. Simon Sandstone
(1,940 m) 6,365 ft thickness 1,620 ft (494 m)
Production Csg
9 5⁄8", 47#, 13CRL80, 12 ¼" bit Longstring
length 1,947 ft (593 m) 5 ½", 17#/ft, CR-13 from
7,260−5,047 ft;
5 ½", 17 #/ft, N-80 casing
from 5,047 ft to surface
(2,149 m) 7,051 ft Mechanical packer with BH
(2,206 m) 7,236 ft gauge and perforation
Precambrian
below.
(2,184 m) 7,164 ft
(2,217 m) 7,272 ft Precambrian
Figure 19.5 Schematic of the injection well CCS1 and deep

monitor well VW1.
344
Geophone Well
GM1
0 ft
Shallow Water Zones
(107 m) 350 ft Surface

9 5 8" 40 lbpf, LT&C
length 350 ft (107 m)
Tubing
3 ½", 9.3 lbpf EUE 8RD-L-80
to 3,496 ft (1,066 m)
Cement circulated to surface
8 ½" open hole to

a depth of 3,500 ft (1,067 m)
(636 m) 2,088 ft
New Albany Shale
(796 m) 2,611 ft
Maquoketa Shale
(997 m) 3,270 ft
St. Peter Sandstone
(1,066 m) 3,496 ft
(1,067 m) 3,500 ft
Figure 19.6 Schematic of geophysical monitoring well GM1.
seismic profile (VSP) data acquisition, is also and roles of each well will be elaborated on in
a vital component of the passive monitoring the following sections on respective geophysical
network. The specific equipment descriptions methods.
345
Geophysical Methods and Data nearby wells, were used to reduce project risk asso-
ciated with reservoir thickness, potential faulting, and
Acquisition Programs storage capacity at the site. As a result of positive
Overview preliminary assessment of the area, an additional
four 2D seismic data lines were acquired at the site
The IBDP geophysical program has provided valu-
in 2009 to obtain additional geological detail and
able direct and indirect measurements of key elastic
supplement pre-drill planning. The seismic data
and petrophysical properties for characterization of
acquisition program for site characterization is
the injection zone, seal, and underlying formations.
described in Table 19.1. New data further verified
A systematic geophysical site characterization
the absence of significant basement paleotopographic
effort started with acquisition of 2D seismic data
highs at the site and a very thick Mt. Simon section.
to support pre-drill feasibility assessment. The site
These data also supported estimates of local regional
characterization was progressively refined through
trends during site characterization and development
drilling and geophysical logging of the injection and
of the 3D reservoir model (Leetaru and Freiburg,
monitoring wells, and acquisition of surface and
2014).
borehole three-dimensional (3D) seismic surveys.
The program entailed systematic improvement of 3D Surface Seismic Data
spatial coverage and measurement resolution as the
The IBDP site characterization and monitoring pro-
project progressed. Site characterization also
gram involved three separate surface 3D seismic data
included establishment of the background micro-
acquisition efforts (Table 19.1). The first 3D survey
seismic activity through 18 months preinjection
conducted in 2010 was designed for detailed site char-
passive monitoring.
acterization over the anticipated CCS1 plume area.
The IBDP site characterization and monitoring
In 2011, this survey was extended to accommodate
program includes a combination of permanent and
potential additional plume monitoring needs. This
retrievable borehole and surface measurement tech-
extended survey (2011) also served as the baseline
niques to acquire data both continuously and periodi-
survey for subsequent IBDP time-lapse monitoring
cally. Various geophysical methods are required to
work. The third survey, conducted in 2015, was after
estimate intrinsic properties of the storage complex
completion of injection of CO2 and this survey serves
for site characterization and to monitor transient
as the first time-lapse monitor survey for the IBDP
reservoir state (pressure, saturation, stress) for MVA
plume.
purposes. The field configuration and methodology
Acquisition of the 3D surveys at the IBDP pre-
have evolved throughout the project to achieve the
sented many evolving challenges. The proximity of
best solutions to multiple technical challenges under
heavy industrial activity and transportation infra-
ever changing operational and commercial conditions
structure resulted in high levels of anthropogenic
and constraints. The following sections describe each
noise. New construction at the industrial site meant
element of the site characterization and monitoring
that each successive survey encountered additional
system. Other chapters in this book (particularly
obstructions to access and related increase in noise.
Chapter 2) provide more detail on the fundamentals
The requirement to acquire surveys during the winter
of the methods used at the IBDP.
season owing to both agricultural activity and desir-
able ground coupling conditions further compounded
Active Seismic Methods the difficulties. The resulting progressive diminish-
ment of achievable surface access, the accompanying
2D Surface Seismic Data progressive increase in levels of noise contamination
Before commencement of the IBDP geophysical data due to surface industrial activity, and an electrical
acquisition campaign, preexisting regional 2D seismic power infrastructure onsite created challenges to seis-
data were the primary source of structural informa- mic data acquisition. These acquisition challenges
tion for pre-drill site characterization. Two 2D seis- were managed through a combination of acquisition
mic profiles (ADM1 and ADM2) were acquired along design, equipment, and parameter selection. For
346 the east–west and north–south roads that pass by the example, the 2D seismic data was used to determine
ADM plant in the fall of 2007. These data, along with the potential frequency bandwidth and resolution of
Table 19.1 Seismic data acquisition program components
Date Data Purpose and comments

2007 Review of existing Increased understanding of regional geology
regional geophysical Relatively poor quality
data
Not site specific for project
October 2007 Two 2D seismic Refined understanding of project site and regional geology
profiles (IA17 and
IA 27)
December 2009 Four high-resolution Validated regional dip from previous interpretations
2D seismic profiles Provided confirmation that no resolvable faults were present
Contributed to the understanding of local geology
January 2010 High-resolution 3D Provided detailed structural and stratigraphic characterization
surface seismic survey Used to derive rock properties for the Eau Claire and Mt. Simon
January 2011 High-resolution 3D Merged with 2010 survey to increase the seismic data coverage for
surface seismic survey structural and stratigraphic characterization; includes the ICCS
Project site
January 2015 High-resolution 3D Served as postinjection time-lapse monitor survey for the IBDP and
surface seismic survey baseline survey for the ICCS Project
the new 3D seismic data set, which showed that it was can build a more robust low-frequency model less
feasible to acquire a high fold 3D surface seismic biased by sparse well-log data.
survey over the site despite the access limitations. Although under ideal circumstances exact
However, some of the pre-survey source and receiver repeatability is desired in all aspects of time-lapse
positions were lost once the acquisition crew started seismic surveying, unavoidable events result in var-
work in the field because of ground conditions and iations in recording equipment between surveys.
changes in infrastructure related to the industrial site. By maintaining consistent high-density single-
The acquisition crew worked with the survey design receiver locations, together with specially designed
team to adapt the survey where surface access was an vibrator sweeps, both signal and noise modes pro-
issue to ensure high fold coverage was maintained. vided the fidelity required for noise reduction in
As a result, the final post-survey fold coverage was data processing. The data acquisition parameters,
very similar to the ideal pre-survey fold coverage. which were maintained throughout all 3D surveys,
Acquisition design at the IBDP is discussed in detail are shown in Table 19.2. The source and receiver
by Couëslan et al. (2009). locations used for the 2015 time-lapse monitor
The Schlumberger Q-Land point-receiver land survey are shown in Figure 19.7a. All of these
seismic system was used at the IBDP, which incor- source and receiver locations are redundant with
porates a pattern of high-density, high-resolution those occupied in the 2011 survey. The fold of
18-Hz geophone accelerometers. This acquisition coverage on the processing bin size achieved in
configuration was intended to give the maximum the 2015 survey within the offset limited range of
amount of flexibility in attenuating the high levels 2135 m corresponding to target depth is illustrated
of noise contamination at this industrial site, while in Figure 19.7b. The uniformity of common depth
maintaining high-frequency bandwidth during pro- point (CDP) coverage in light of the numerous
cessing. The 27 216 kg vibrator trucks used surface obstructions demonstrates the power of
a maximum displacement sweep design to generate high-density source and receiver effort.
energetic low frequencies from 4 to 100 Hz to max-
imize the low-frequency content in the dataset. Borehole Seismic: Vertical Seismic Profiles
Seismic data with frequencies between 2 and 10 Hz Time-lapse 3D vertical seismic profiles (VSP) were an
347
are particularly useful for inversion analysis, as they important component of the MVA plan for the IBDP.
Table 19.2 3D surface seismic data acquisition parameters source point locations. This was particularly proble-
Recording parameters matic at the IBDP site where permitting issues
resulted in restricted access to major sections of
Recording system Variable
the surface for the first three monitor surveys.
Recording format SEG-D, IEEE The resulting reduction in source points resulted
Record length 5s in reduction of image quality for these surveys and
Sample rate 2 ms also in low-resolution time-lapse signals from these
Recording filter (Hi-Cut) 100 Hz monitor surveys. Monitor 4 overcame some of the
Recording filter 4 Hz restrictions to surface access and Figures 19.8a and
(Low-Cut) 8b show source and receiver location analysis for the
Baseline 2-Monitor 4 (B2-M4) survey set. Green
Nominal fold 60
polygons in Figure 19.8a delineate the shotpoints
Receiver type GAC-C
not accessible during M1, M2, and M3 surveys.
Line geometry Detector interval = 10 ft. This, together with greater quantity of injected
Source Parameters CO2, is believed to be a major factor in the enhanced
Source Vibroseis stability of time-lapse analysis for the B2–M4 survey
Source type AHV IH 67 000 lbs. pair (Section 5.2.1).
Shotpoint interval 80 ft.
Sweep number 4 Passive (Microseismic) Monitoring
Sweep start frequency 4 Hz System Overview
The IBDP microseismic monitoring network is
Time-lapse VSPs were intended to provide informa- comprised of multiple subsurface arrays deployed in
tion on CO2 plume development, demonstrate three separate wells. The subsurface data acquisition
containment of the CO2 in the storage formation, network components are described in Table 19.4.
and provide data to verify and update models and The arrays in CCS1 and GM1 were calibrated using
simulations over the life of the project (Couëslan drilling noise and perforation shots in the borehole of
et al., 2013). The VSP array was permanently cemen- VW1, which Smith and Jaques (2016) describe in
ted into GM1 – the dedicated geophysics monitoring detail.
well located approx. 60 m northwest of the injection Data from all subsurface arrays are integrated and
well. The array consisted of 31 levels of 3C phones real-time data acquisition functions are performed
deployed at depths ranging from approx. 50 m to using an integrated data acquisition system located
1050 m below ground surface. at the field office. In addition to the subsurface equip-
Six 3D VSP surveys were acquired at the site: two ment, ISGS installed five surface seismometer stations
baseline surveys and four monitoring surveys (Figure 19.9). The raw microseismic data stream from
(Table 19.3). Although borehole deployment of the the subsurface array is recorded in 10-s SEG2 files.
receiver array, combined with the much-reduced The rate of data sampling during the preinjection
energy source footprint, resulted in fewer access pro- period was every 2 ms (500 samples per second).
blems than encountered during surface seismic acqui- The data set composition is described in terms of
sition, the relative contribution of each VSP source date ranges, number of files, number of triggers, and
point to the final image is much greater in VSP sur- events identified. The number of triggers represents
veying. As such, surface access was a major challenge the total number of possible events detected by the
to achieving repeatability between successive VSP system, and so includes many false triggers. Given the
surveys. The VSP acquisition conditions, source para- sensitivity of the system, false triggers are caused by
meters, and quantity of CO2 injected for each survey any combination of transient electrical glitches and
are summarized in Table 19.3. well-related activity such as pipeline maintenance.
Given the use of a permanent receiver array, the A false trigger will not have a P- and S-wave signature
greatest remaining challenge to successful time- common to microseismicity or place an event in the
348 lapse imaging is in achieving repeatability of energy subsurface.
Figure 19.7 Seismic source and receiver locations with fold coverage. (a) 2015 monitor survey source and receiver postplot with surface obstructions (shaded red) and
(b) resulting fold of coverage on processing bins and limited to 2135 m.
Table 19.3 Schedule of time-lapse VSP surveys and injected CO2 quantity.
Survey Date Ground Vibrator Repeated shots Amount of

conditions sweep (Hz) (relative to B2) injected CO2
(tonnes)
Baseline 1 (B1) Jan. 27–30, 2010 Wet 2–100 — Preinjection
Baseline 2 (B2) Apr. 12–14, 2011 Dry 8–120 — Preinjection
Monitor 1 (M1) Feb. 11–12, 2012 Frozen/Dry 8–120 467 74 000
Monitor 2 (M2) Apr. 4–5, 2013 Damp 8–120 385 433 000
Monitor 3 (M3) Feb. 3–5, 2014 Frozen 8–120 378 730 000
Monitor 4 (M4) Jan. 15–17, 2015 Frozen 8–120 458 approx. 1 000 000
Geophone Arrays microseismic monitoring. However, given the reduc-

Injection Well (CCS1) tion of observations in CCS1, as mentioned earlier,
In the original monitoring plan, the passive seismic the significance of observations from the GM1 array
monitoring equipment in CCS1 was anticipated to for microseismic characterization greatly increased.
provide the majority of the observations for micro- Data recording in GM1 also commenced in May 2010.
seismic location and characterization (Figure 19.5). Data from a selected 28-level (86-channel) subset of the
A specialized deployment mechanism was used to 31-level array are recorded to the 96-channel DAUs,
decouple the sensors from the injection tubing string, along with data from CCS1.
thereby reducing flow-induced noise. As originally Geophones Orientation
designed, the four sensor levels in CCS1 were set at
Locating microseismic events requires precise knowl-
subsurface depths spanning the interval from the Eau
edge of the position of the geophones in x, y, and
Claire to Unit A of the Mt. Simon (deepest part of the
z directions, and the orientation of the geophones
formation), which was intended to provide reasonable
within that position.
signal fidelity and location accuracy throughout the
Tool orientation calculations were performed on
primary seal and injection intervals. Damage to two
the various IBDP subsurface arrays at different times
sensors during deployment resulted in only the two
during the monitoring project as new subsurface
sensors located nearest the injection zone being
components were added and energy sources became
usable. The eight channels of data from these two
available because of field activities. Because the array
sensors are recorded via the 96-channel DAU located
in CCS1 provides the primary observations constrain-
at the nearby field office. Preinjection data acquisition
ing the azimuth used in calculating microseismic
began in May 2010.
event locations, verifying accurate orientation of
Geophysical Monitoring Well (GM1) these geophones was attempted during all orientation
The dedicated geophysical monitoring well, GM1, is efforts.
1067 m deep and contains a 31-level array of 3-com-
ponent tools permanently cemented into place (out- Surface Passive Monitoring Stations
side the casing) during construction (Figure 19.6). In 2013, a network of five 3-component seismometers
The geophones are approx. 500 m shallower than were installed in shallow, near-surface vaults on ADM
those in CCS1, and 29 of the geophones are located property and away from buildings (Figure 19.9). Four
between 623 and 1049 m depth with the remaining of the surface seismic stations were deployed in
two at 41 and 108 m (Will et al., 2016a). approximately the four cardinal directions from the
The primary objective of the GM1 array was to CCS1 well at radial distances of 1066.8 to 1828.8 m.
obtain high repeatability time-lapse 3D VSP surveys The fifth station is located outside the National
for plume monitoring. The main well construction Sequestration Education Center building at Richland
and array design considerations focused on VSP ima- Community College (labeled as Seis3 RCC in
350
ging objectives, with secondary objectives to support Figure 19.9) where real-time data monitoring is
Figure 19.8 VSP source and receiver locations. (a) Source locations from B2 and M4 surveys. Orange dots are B2 and Blue are M4 shot points. Green boxes show the shots that were not
available in M1, M2, and M3 surveys. (b) Extracted collocated sources, survey B2coM4. The color of circles indicated the relative distance from the corresponding source location from M4
survey.
Table 19.4 Microseismic system components
Well Configuration Recording

CCS1 2-level array of a 4-component WellWatcher PS3* passive 96-channel data acquisition unit
seismic sensing system with 4-Geospace 2400 (DAU) located at the field office
geophones in tetrahedral configuration (two levels near CCS1
functioning)
GM1 A 31-level array of 3 Geospace 2400 geophones in 96-channel DAU located at the
orthogonal configuration field office near CCS1
VW2: Operated by 5-level array of a 3-component multilevel downhole 24-channel DAU located at the
the ICCS Project seismic array, Sercel array installed temporarily (Sept. 26, wellhead
2013 and removed Feb. 25, 2015)
processing, the data have been used for independent

Seis5 North
verification of larger events.
Nonseismic Measurements
Seis3 RCC
Geophysical Logging
A comprehensive suite of geophysical logs were
Seis1 West acquired in boreholes CCS1, VW1, and GM1, along
with physical borehole and core tests in CCS1 and
VW1. In addition to measurements required for geo-
logical characterization, the logging program was
designed to support advanced geophysical and geome-
Injector
chanical integration workflows. Elastic properties,
Seis2 East
mineral fractions, and fluid saturations were measured
or computed for seismic inversion and time-lapse
integration. For geomechanical characterization and
microseismic analysis, in situ stress and direction were
documented by observation of tensile-induced drilling
fractures and breakouts as shown by the Formation
NORTH
MicroImager (FMI)3 fullbore formation microimager
Seis4 South
and a minifrac, injection and step rate tests. Some of
the specialized well logs, tests, and resulting measured
Figure 19.9 Location of five surface seismometers at the IBDP site. and calculated properties used for geophysical integra-
tion at the IBDP are shown in Table 19.5.
performed for this station. This near-surface network
was intended to be a backup for the downhole micro- Repeat Pulse Neutron Logs (RST)
seismic systems for large magnitude events. RST logging at IBDP was performed for operational
The instruments in the near-surface system will con- and verification monitoring; CCS1, VW1, and GM1
tinue to record on-scale if a seismic event, large are logged at least annually to verify the mechanical
enough to saturate the downhole microseismic sys- integrity of each well. The logs are also used to moni-
tems, were to occur near the IBDP site. If a large event tor the development of the CO2 plume adjacent to
were to occur, the sensors in the downhole systems CCS1 and VW1 and to constrain the dynamic reser-
would be able to detect the event and determine its 3D voir modeling. Two baseline surveys were acquired in
location, although they may not accurately measure CCS1 and VW1; one shortly after each well was drilled
352 the event magnitude. Though the near-surface data
3
are not integrated with downhole data in event Mark of Schlumberger.
Table 19.5 Petrophysical logs run in boreholes with measured

and calculated properties. Results
Petrophysical logging Measured and Surface Seismic Data
tests calculated properties
Imaging for Site Characterization
Spontaneous potential Density The main processing objectives for the 2010 and 2011
Neutron porosity Clay content 3D seismic surveys focused on site characterization.
Resistivity Static Young’s modulus Processing flows were designed for
Microresistivity imaging Static Poisson’s ratio • Merging the 2011 extension with the 2010 survey
Sonic velocities Static shear modulus • Producing a data set suitable for acoustic and
Elemental capture Biot’s coefficient elastic impedance inversion
spectroscopy • Producing a volume suitable for structural
Natural gamma ray Frictional angle interpretation
Magnetic resonance Tensile strength • Imaging any faults or fractures in the zone of
Thermal neutron decay Unconfined interest (0.8–1.1 s)
compressive strength The main data processing challenge for the sur-
Fullbore formation Overburden stress face 3D data set was the significant amount of ran-
microimager* Minimum horizontal dom noise due to coupling issues and from
stress anthropogenic activity, mainly roads throughout
Maximum horizontal the acquisition area. Unfiltered prestack migrated
stress gathers in Figure 19.10a illustrate the extent of
Azimuth minimum noise contamination. The same gathers after
horizontal stress a series of premigration filtering processes designed
Tensile-induced drilling
to maintain frequency and phase integrity of coher-
fractures and breakouts
ent signals are shown in Figure 19.10b. The corre-
Mineral fractions
Fluid saturations sponding stacked sections in which imaging is
greatly improved throughout the vertical section
* Mark of Schlumberger. are shown in Figures 19.11a and 11b.
and a second some months after to give the drilling

Time-Lapse Processing
fluids a chance to migrate away from the wellbores. Data Processing
The 2011 and 2015 surveys were reprocessed for four-
Pressure Monitoring dimensional (4D) time-lapse detection purposes.
VW1 was originally completed with a As discussed, although it is desired to maintain identical
multilevel Westbay monitoring and sampling com- acquisition parameters for time-lapse surveys, many
pletion system optimized specifically for CCS appli- factors affect acquisition such as changes of surface
cations and depth of operation. The Westbay system accessibility, random noise patterns, and changes in
completion continuously monitored pressure and ground conditions. To reveal the true 4D difference
temperature using the Modular Subsurface Data caused by CO2 injection in seismic datasets, it is impor-
Acquisition System (MOSDAX™, Westbay tant to remove other inferences as much as possible.
Instruments, North Vancouver, British Columbia) Both the baseline and monitor datasets at IBDP were
modular subsurface data acquisition system probes processed (coprocessed) using the same processing
at 11 different depths in the well. Each sampling zone sequence (Figure 19.12). For several key steps, such as
was isolated from the other ports through redundant refraction tomography statics solution, residual reflec-
packers (Figure 19.5). Fluid samples were also taken tion statics solution, and deconvolution, tests were run
from each port at discrete time intervals for geo- for both a joint solution and a standalone solution.
chemical analysis when the pressure and tempera- Also, throughout processing, 4D quality control (QC)
ture probes are pulled from the well. procedures were conducted to ensure non 4D 353
Figure 19.10 Prestack migrated 2011 3D seismic gather. Prestack migrated seismic gathers from before (a) and after (b) multichannel
filtering.
354 Figure 19.11 Stack 2011 3D seismic gather. Stack before (a) and after (b) multichannel filtering from gathers corresponding to Figure 19.10.
Figure 19.12 4D surface seismic data

Survey 2011 seismic Survey 2015 seismic
processing flow.
Refraction Tomo Refraction Tomo
Pre-decon noise Pre-decon noise

attenuation attenuation
Joint robust surface

consistent Decon
Post-decon noise Post-decon noise

attenuation attenuation
Joint residual statics

solution-1
Nominal acquisition
geometry matching
Vel. pick (on 2011) &

Convert to OVT Joint residual statics Convert to OVT
domain & gap filling solution-2 domain & gap filling
PSTM; Radon Mig Vel from Survey Mig Vel. pick & PSTM
2011 applied to 2015 Radon
Stack, Footprint Stack Vel from Survey Stack Vel. Pick, Stack
removal 2011 applied to 2015 Footprint removal
4D attributes extraction
NRM disp. field derived
inferences were reduced as processing progressed. reservoir acoustic impedance and/or causes changes in
Because of the processing workflow, very subtle changes reflector timing beneath the reservoir. Pore pressure
in migrated data (Figure 19.13) can be identified that changes may also cause gas dissolution or exsolution,
are the result of time-lapse attribute analysis. producing a large change in acoustic impedance, or
cause stress changes above, and possibly below, the
Time-Lapse Attribute Extraction reservoir; producing a change in velocity causes time
Two time windows were defined for 4D attribute shifts between time-lapse volumes below the reservoir.
extraction at IBDP (Figure 19.14). It is assumed that These changes to the storage system could be
strata within the time Window-1 (400–700 ms) above detected through examination of characteristics of seis-
the caprock should not be affected by CO2 injection. mic data, and 4D attributes extracted from seismic stack
Strata within the time window-2 (900–1300 ms) that volumes. Time-lapse attributes such as reliability
includes the CO2 injection zone could be impacted in (Figure 19.15) and normalized root mean square
several ways. (NRMS) (Figure 19.16) were extracted from different
Replacement of one fluid by another with different time windows and were compared to detect any 4D
density/properties changes the velocity and density changes caused by CO2 injection. In both cases, the
of the reservoir interval, resulting in modification of attributes from window 1 (above the injection zone)
reservoir acoustic impedance and/or causes small indicate good time-lapse repeatability while those from
changes in reflector timing beneath the reservoir. window 2 (including the injection zone) show very
Pore pressure changes also modify the seismic veloci- similar time-lapse differences, which are attributed to 355
ties within the reservoir, resulting in modification of fluid replacement.
Figure 19.13 Subtle changes in migrated data. Zoom-in proximity of CO2 injection spot (yellow bar indicates the injection well path). Notice
the very subtle changes in amplitude and waveform in seismic.
Figure 19.14 Survey 2011, 4D attribute extraction window definition. Window-1, above the CO2 injection zone; Window-2, including the
injection zone.
An additional time-lapse interpretation tool is the expect the NRM attribute to show anomalies below
nonrigid matching (NRM) attribute. An NRM uses the plume where arrival times have been impacted by
a nonrigid matching method time- or depth- to align the overlying fluids. The NRM computed using the
two cubes of seismic data by generating a 3D displace- 2011 baseline survey as the reference volume is shown
ment field that compares the input data to a reference in Figure 19.17. An NRM anomaly in the vicinity of
356 volume. Owing to the fluid effect on velocity we the injection well is clearly evident.
Figure 19.15 Attribute predictability comparison. Red indicates high similarity. Blue indicates low similarity.
Figure 19.16 Attribute normalized root mean square comparison. Red indicates large difference in amplitude. Blue indicates small difference
in amplitude.
3D VSP for seismic signal propagation. In addition, there was

significant 60-Hz noise from nearby electrical sources,
Vertical Component VSP Analysis such as power lines and an electrical substation,
Baseline Surveys recorded in the data although this noise was attenu-
The first baseline 3D VSP was acquired ated through processing. The acquisition footprint of
in February 2009, after the 3D surface seismic survey, the baseline 3D VSP also suffers from the same chal-
lenges posed by surface infrastructure obstructions to 357
but while ground conditions were damp and not ideal
Figure 19.17 3D NRM displacement field zoomed in near the CCS1 injection zone.
energy source points. Holes in the acquisition foot- both data sets to further reduce the differences
print surface seismic data can be compensated by between the data sets related to noise and acquisition
adding in-fill source and receiver locations; however, artifacts while maintaining the differences related to
3D VSP data is not as flexible given the geometry of CO2 injection.
the ray paths. As a result, there are identifiable arti-
facts in the 3D VSP images related to holes in their Time-Lapse Attributes
acquisition footprint. In 2011, a second baseline 3D Analysis of time-lapse VSP data utilizes many of the
VSP survey was acquired under dry ground condi- same tools as are used for analysis of surface time-
tions, and an isolation transformer with a Faraday lapse data. Results from the Monitor 1 and Monitor 2
shield was used to reduce the 60-Hz noise. surveys were ambiguous, likely from a combination of
acquisition conditions and limited quantities of CO2
Monitor Surveys injected. However, the Monitor 3 (Figure 19.20) and
Because of the favorable acquisition conditions of Monitor 4 (Figure 19.21) surveys yield coherent time-
Baseline 2, all monitor surveys were collocated and lapse anomalies when coprocessed with Baseline 2.
coprocessed relative to Baseline 2 through a vertical NRMS maps of the overburden interval show no
component processing workflow (Figure 19.18). definite trend indicating good repeatability, whereas
Preprocessing included receiver selection, 60-Hz in the injection interval the highest NRMS values are
notch filter, cross-equalization based trace-by-trace in the southeast near where injection well CCS1 is
amplitude scaling, linear radon transform for wave- located (Figures 19.20 and 19.21).
field separation, deterministic shot-by-shot reference
trace wave-shape deconvolution. Collocated energy Microseismic Monitoring
source records (15 m tolerance) were cross-equalized
to allow for residual matching of the amplitude and Microseismicity Pre-CO2 Injection
phase spectra, and then processed for wavefield Preinjection microseismic activity was monitored by
separation and VSP deconvolution. The same 1D the arrays in CCS1 and GM1 from May 1, 2010 until
anisotropic model that was built using zero offset start of injection on November 15, 2011, during which
VSP and 3D VSP travel times from the baseline survey they recorded more than 68,000 triggered events.
was used for imaging both the datasets (Figure 19.19). Most events were related to surface noise, drilling of
358
After migration, nonrigid matching was applied to the VW1 borehole, well activities, VW1 perforation
Figure 19.18 Adopted

Delete top 2 De-noise and processing flow of single-axis
Raw Z 60 Hz
and noisy Amplitude 3D VSP processing.
data notch
receivers scaling
Linear Radon
wave-field
separation
Down-going Up-going
VSP shot-by-
shot reference
deconvolution
De-convolved De-convolved
Down-going Up-going
359
Figure 19.19 Migrated images B2 and M4.
Figure 19.20 Time-lapse NRMS attribute comparison of B2-M3. (a) B2-M3 NRMS attribute in the overburden zone. (b) B2-M3 NRMS in the reservoir.
Figure 19.21 Time-lapse NRMS attribute comparison of B2-M4. (a) B2-M4 NRMS attribute in the overburden zone. (b) B2-M4 NRMS in the reservoir zone showing a distinct time-lapse
anomaly in the vicinity of the injection well.
shots, and many distant events assumed to be mine a slip of a fraction of a tenth of a millimeter on a larger
blasts because of distance and time of day (8:00 to plane of tens of meters.
17:00). Twenty-one distant natural earthquakes (12 in The range of magnitudes of events is nearly iden-
the United States Geological Survey catalog) and eight tical in each major geological unit in which events were
local microseismic events were recorded, which located: the Precambrian had the highest magnitudes
appear to be unrelated to well activity and presumed of 1.16 and 1.147, whereas magnitude highs in the mid-
to be background events. These eight background to lower 0.8s, were recorded in the Argenta and Mt.
events were located in the Mt. Simon and Argenta; Simon, respectively. The lowest moment magnitudes
six were located within 457 m of CCS1 and two were detected in the Precambrian were –2.02 and in the
located 1646–2408 m from CCS1. The magnitude Cambrian sandstones –2.13. The very low magnitudes
range for the eight background events was –2.16 to – of microseismic events, their locations and estimated
1.52 (average of –1.83), with a similar range for events slip plane size, indicate there is little to no risk that
related to drilling and well activity of –2.6 to –1.6 induced seismicity at IBDP could cause fault slippage
(Smith and Jaques, 2016). through the caprock and compromise the seal.
About eighty percent of the microseismic events are
Microseismicity during CO2 Injection in the Precambrian crystalline basement, and the
Locatable microseismic events started one month after remainder in the Mt. Simon and Argenta Formations.
the start of injection, with two events in December 2011 The first locatable events were about 600 m from CCS1
and 13 in January 2012. During injection, 4848 events and as events continued at this distance, new ones
were located with an average of four per day having an started about 300 m away. Around each of these dis-
average magnitude of –0.8. Ninety-four percent of tances, separate elongated clusters of events began to
events were less than magnitude 0 but five events were emerge. Eventually 18 clusters of microseismic events
above 1 ranging from 1.01 to 1.17. (The amount of were identified at IBDP, with no correlation existing
energy released represented by a –1 magnitude is between time of cluster development and distance from
about that of the typical inch long, pencil-thin fire- the injection well (Figure 19.22). For example, clusters
cracker.) Using Zoback and Gorelick’s (2012) magni- are numbered in order of appearance, and Cluster 1
tude versus slip surface size and displacement diagram, averages 548 m from CCS1, Cluster 2 is nearly half that
a magnitude –1 represents a slip of less than distance (335 m), Cluster 3 averages 457 m, and Cluster
a millimeter on a plane of a couple of meters in size to 4 averages 823 m from CCS1. Later clusters also
Figure 19.22 Locations of microseismic

clusters.
362
developed out of sequence with increasing distance preexisting natural fractures that are favorably oriented
from the injection well. for slip (Pearson 1981; Rutledge and Phillips, 2003;
Clusters at IBDP continued to add microseismic Shapiro et al., 2006). Baig et al. (2012) observed that
events during injection and transient shut-in periods, with hydraulic fracturing of some horizontal wells
and most developed as an elongated pattern oriented events occurred in two main trends: early events were
SW–NE. Likely factors other than pore pressure roughly 30° from SHmax and late events were approxi-
changes influenced the sequence of cluster develop- mately parallel to SHmax. Yang et al. (2013) found similar
ment, such as the angle of preexisting planes of weak- trends for hydraulic fracturing in the Bakken formation
ness or defects in relation to maximum horizontal where detected microseismic events from many stages
stress direction. Preexisting weak planes or defects trend approx. 30° from the direction of SHmax. This
close to 30° from the maximum horizontal stress relation appears to be present at IBDP through the
(SHmax) direction are optimally oriented in the direc- sequence of cluster development, with Cluster 1 at 28°
tion expected for strike-slip movements. from SHmax, Cluster 2 at 11°, Cluster 3 at 9°, and Cluster
Evidence related to microseismic events induced by 4 at 3° (Figure 19.23). The strike of each plane was
hydraulic fracturing suggest they are triggered along developed by a best-fit plane through the 3D cloud of
1172700
Cluster 4 - N 65 E
3 degrees from SHmax
2,715 ft from CCS1
1172300
1171900
1171500
Cluster 3 - N 59 E
Cluster 1 & 11 - N 40 E 1,520 ft from CCS1
1171100
1,793 ft from CCS1 Cluster 2 - N 57 E
1,161 ft from CCS1
1170700
1170300
on
re ssi
N 68 E mp
l Co
z o nta
1169900 r i x
Ho SHma
um
xim
Ma
CCS1 - Injection Well
1169500
340700 341100 341500 341900 342300 342700 343100 343500 343900
363
Figure 19.23 Clusters 1–4 directions in relation to maximum horizontal in situ stress and average distance from injection well CCS1. Axes are
northing and eastings in feet. (From Bauer et al., 2016.)
3.0
b = −1.102
2.5
2.0
Log Frequency
1.5
1.0
0.5
0.0
−2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5
Magnitude (bin 10ths)
Figure 19.24 Gutenberg–Richter plot of all the microseismicity during the three years of injection. (From Bauer et al., 2016.)
events. The strike of the planes in relation to the in situ 10

SHmax fits the first motion analysis indicating right lat-
eral strike-slip motions on many of the clusters
(Couëslan et al., 2014). These planes are interpreted to
be the reactivation of preexisting features as shown by 1
the Gutenberg–Richter (Gutenberg and Richter, 1956)
P/Sh
plots for all the microseismicity, which show b-values

close to 1 (Figure 19.24). Will et al. (2016b) present an
0.1
in-depth analysis of the data integration and reservoir
response resulting in microseismicity. Negative
Positive
The total event population was separated into sub- Mixed
sets on the basis of spatial clustering and the fault plane Model
0.01
solution (FPS) analysis was performed on individual −80 −60 −40 −20 0 20 40 60 80
event subsets. Figure 19.25 shows the FPS result for Azimuth from N45E (Deg)
one of the most developed and coherent event clusters
(#4), which exhibits a very good fit to the strike-slip Figure 19.25 FPS for a selected event cluster (#4) using a strike-slip
model for an azimuth of N45E. Figure 19.26 shows the assumption and azimuth.
FPS superimposed on event clusters.
containment. However, because of intrinsic charac-
Monitoring Strategy and Discussion teristics such as spatial coverage, resolution, and phe-
nomenological ambiguity, in order to realize its full
Integrated Modeling value, seismic data must be integrated with other
Each component of the IBDP geophysical program direct measurement types such as borehole logs,
has made a unique contribution to the understanding core, and well tests. Therefore, the IBDP site charac-
of key aspects of site characterization such as geologi- terization and monitoring program has included an
cal structure, stratigraphy, and reservoir properties extensive integrated geologic modeling effort.
364
dictating well injectivity, storage capacity, and The various contributions of active seismic methods
Figure 19.26 FPS results for all clusters.
Figure 19.27 Contribution of seismic data to

the static geologic model.
to the IBDP geologic model through interpretation of properties through integration with petrophysical
stratigraphic horizons, acoustic impedance and seis- well logs are shown in Figure 19.27.
mic porosity inversion, time-depth domain conver- Geophysical methods have also been instrumental
sion through integration with acoustic and density in monitoring transient (both natural and injection
well logs, and colocated cokriging of reservoir induced) changes within the storage unit as part of the 365
MVA program. Active and passive seismic measure- state as provided by fluid flow and geomechanical mod-
ments, such as time-lapse surveys and passive micro- els. Fluid flow and geomechanical models were part of
seismic monitoring, provide valuable indirect early stage IBDP characterization and have been
measurements of hydrodynamic and mechanical state updated as the project evolves.
within the reservoir. However, while these measure- The following sections describe the pivotal role of
ments provide valuable information about the timing the IBDP integrated model in interpretation of time-
and location of changes in reservoir state, each on its lapse seismic and passive monitoring data for under-
own is insufficient to yield an unambiguous physical standing of plume development and microseismic
characterization of the responsible phenomena. mechanisms. The geologic and transient process mod-
Characterizing these phenomena requires an under- els have been periodically updated to incorporate new
standing of local transient hydrodynamic and stress characterization and calibration data (Table 19.6).
Table 19.6 IBDP integrated model development history
Structure/stratigraphy/ Hydrodynamic properties Use

discrete features
2008 • Layer cake stratigraphy • Uniform zonal porosity • Site characterization
Preliminary defined by well tops from and permeability • Basis for initial reservoir
analog well 60 miles • Assigned using logs from simulation model
away analog well 60 miles away
• No discrete features
2010 Update • Layer cake stratigraphy: • Stochastic zonal porosity • Updated site
well tops from CCS1 and permeability characterization
• No discrete features • Conditioned to CCS1 well • Basis for initial reservoir
logs simulation plume
predictions
2011 Update • Stratigraphy: 2010 3D • Stochastic zonal porosity • Update site
seismic survey and well and permeability characterization
top control from CCS1 • Conditioned to CCS1 and • Basis for final Class VI
and VW1 VW1 well logs and 2010 permit reservoir
• No discrete features seismic inversion products simulation area of review
• Basis for preliminary finite
element model (FEM)
2013 Update • Stratigraphy: 2011 • Stochastic zonal porosity • Update site
extended 3D seismic and permeability characterization
survey and well top • Conditioned to CCS1, VW1, • Basis for updated FEM
control from CCS1, VW1, and VW2 well logs and and preliminary
and VW2 2011 seismic inversion microseismic prediction
• Provisional fault products research
interpretation
• Mechanical features
inferred from
microseismic data
2016 Update • Stratigraphy: 2015 • Updated stochastic zonal • Update site
seismic survey was porosity and permeability characterization for
completed to be a time- based on the additional reservoir simulations
lapse survey for CCS1 updated structure
injection and baseline • Conditioned to CCS2,
survey for CCS2 injection CCS1, VW1, and VW2 well
366 with a 400-m extension logs and 2011 seismic
to the north inversion products
Contributions include stratigraphic horizons, seis- Gassmann fluid substitution model and used to
mic porosity inversion, and colocated–cokriged reser- forward model attribute differences corresponding
voir properties. to field survey times.
Before time-lapse integration, the calibrated reser-
Mapping the CO2 Plume voir simulation model predicted an approximately
radial plume centered on the injection well (Senel
Injection of CO2 produces variations in elastic prop-
et al. 2014). This simulation was calibrated to 15
erties in the reservoir that change the time-lapse seis-
months of injection and monitoring well pressure
mic response; these changes are often subtle and are
and verified with pulse neutron logs. Upon extraction
nonunique with respect to saturation and pressure
of time-lapse attributes (Figures 19.15–19.17) from
effects. At the IBDP, a “simulation-to-seismic” work-
the 2011–2015 surface seismic survey pair,
flow has been used for model based integration of
a significant discrepancy was noted in the shape and
time-lapse seismic data and direct measurements of
extent of the current plume prediction and time-lapse
reservoir state using the geological and reservoir
observation. While differences are expected owing to
simulation models developed for site characterization
vertical plume geometry and seismic detection limits,
and plume predictions.
the significant deviation in aerial geometry (in parti-
The reservoir model was calibrated to injector
cular the truncation of northward plume migration)
and multilevel monitoring well pressures to com-
indicated fundamental deficiencies in the simulation
pute the isotropic effective elastic properties for the
model. Examining the 2011 survey through seismic
reservoir at the times of seismic surveys.
attribute and inversion analysis revealed that features
Thermodynamic properties for brine were com-
exist in the baseline seismic data suggestive of
puted using Batzle–Wang equations (Batzle and
enhanced reservoir quality and a possible flow barrier
Wang, 1992), and CO2 properties were computed
having geometrical characteristics consistent with the
using new analytical expressions in the National
time-lapse anomaly. These features were not incorpo-
Institute of Standards and Technology Reference
rated in the original model due to use of a global
Fluid Thermodynamic and Transport Properties
Database (NIST–REFPROP) software. An average interpolation scheme. In an updated model, the reser-
of Hashin–Shtrikman (H–S) bounds were used to voir quality was modified and a hypothetical flow
calculate the bulk and shear moduli of solid rock barrier was inserted corresponding to the features
identified in Figure 19.28. Figure 19.29 shows the
using rock compositions derived from elemental
resulting simulated plume and forward modeled
analysis on available logs. Effective elastic proper-
time-lapse acoustic impedance change along with
ties of a fluid-filled formation were computed using
Reflection Intensity Sweetness Envelope
Fault? Fault?
Fault?
Outline of NRM Anomaly Outline of NRM Anomaly Outline of NRM Anomaly
Figure 19.28 Seismic attributes used as the basis for modification of the reservoir model for time-lapse seismic history matching. (From
Will et al., 2017.) 367
Figure 19.29 Plume simulation and forward-modeled acoustic impedance change. (Left) Aerial view of simulated 2015 monitor CO2 plume
and outline of 2001–2015 surface seismic time-lapse NRM attribute (from Figure 19.17). (Right) Aerial view of forward modeled time-lapse
acoustic impedance change and outline of 2001–2015 surface seismic time-lapse NRM attribute. (From Will et al., 2017.)
the outline of the NRM anomaly (from Figure 19.17) during and after injection. Sophisticated noise attenua-
in which discrepancies have been greatly reduced. tion and imaging data processing techniques help to
Qualitative analysis of this new result suggests overcome many of the challenges encountered in typical
a detectability limit of approx. 3–4% change in acous- operational scenarios, providing input to structural
tic impedance. Similar efforts are being deployed interpretation, reservoir property estimation, and mon-
toward addressing saturation inversion. itoring. Geophysical monitoring and analysis of the
IBDP will continue throughout the postinjection and
Summary postinjection site care compliance monitoring phase
(expected end date 2020).
Integrating geophysical and geological data is necessary
to develop robust geological, fluid flow, and geomecha-
nical models, and to provide essential support for feasi- References
bility assessment, operational planning, and regulatory Baig, A. M., Urbancic, T., and Viegas, G. (2012). Do hydraulic
compliance. The IBDP demonstrates practical applica- fractures induce events large enough to be felt on
tion of existing acquisition and analytical geophysical surface? CSEG Recorder, 10(October 2012): 40–46.
methods to achieve the goals of a large-scale CO2 sto- Batzle, M., and Wang, Z. (1992). Seismic properties of pore
rage demonstration project. Seismic geophysical meth- fluids. Geophysics, 57: 1396–1408.
ods comprise an extensive set of tools that may be used Bauer, R. A., Carney, M., and Finley, R. J. (2016). Overview
to fulfill many site characterization and monitoring of microseismic response to CO2 injection into the Mt.
objectives for regulatory compliance or scientific inves- Simon saline reservoir at the Illinois Basin-Decatur
tigations. Data acquisition techniques may be tailored Project. International Journal of Greenhouse Gas
for the operational environment and the specific ima- Control, 54(1): 378–388.
ging objective throughout the life of a CO2 storage Couëslan, M. L., Leetaru, H. E., Brice, T., Leaney, W. S., and
368 project from pre-drill site selection to plume monitoring McBride, J. H. (2009). Designing a seismic program for
an industrial CCS site: Trials and tribulations. Energy Rutledge, J. T., and Phillips, W. S. (2003). Hydraulic
Procedia, 1(1): 2193–2200. stimulation of natural fractures as revealed by
Couëslan, M. L., Ali, S., Campbell, A., et al. (2013). induced mircroearthquakes, Carthage Cotton
Monitoring CO2 injection for carbon capture and Valley gas field, east Texas. Geophysics, 86(2):
storage using time-lapse 3D VSPs. Leading Edge, 441–452.
32(10): 1268–1276. Senel, O., Will, R. and Butsch, R. J. (2014). Integrated
Couëslan, M. L., Butsch, R., Will, R., and Locke, R. A. II reservoir modeling at the Illinois Basin–Decatur
(2014). Integrated reservoir monitoring at the Illinois Project. Greenhouse Gases: Science and Technology,
Basin-Decatur Project. Energy Procedia, 63: 2836–2847. 4(5): 662–684.
Freiburg, J. T., Morse, D. G., Leetaru, H. E., Hoss, R. P., and Shapiro, S. A., Dinske, C., and Rothert, E. (2006).
Yan, Q. (2014). A depositional and diagenetic Hydraulic-fracturing controlled dynamics of
characterization of the Mt. Simon Sandstone at the microseismic clouds. Geophysical Research Letters, 33:
Illinois Basin-Decatur Project carbon capture and L14312.
storage site, Decatur, Illinois, USA, Illinois State Smith, V., and Jaques, P. (2016). Illinois Basin–Decatur
Geological Survey Circular 583. Project pre-injection microseismic analysis.
Greenberg, S. G., Bauer, R., Will, R., et al. (2017). Geologic International Journal of Greenhouse Gas Control, 54(1):
carbon storage at a one million tonne demonstration 362–377.
project: Lessons learned from the Illinois Basin– Will, R., El-Kaseeh, G., Jaques, P., Carney, M.,
Decatur Project. Energy Procedia, 114: 5529–5539. Greenberg, S., and Finley, R. (2016a). Microseismic
Gutenberg, B., and Richter, C. F. (1956). Magnitude and data acquisition, processing, and event
energy of earthquakes. Annals of Geophysics, 9(1): 1–15. characterization at the Illinois Basin–Decatur Project.
International Journal of Greenhouse Gas Control,
Haimson, B. Z., and Doe, T. W. (1983). State of stress, 54(1): 404–420.
permeability, and fractures in the Precambrian granite
of Northern Illinois. Journal of Geophysical Research, Will, R., Smith, V., Lee, D., and Senel, O. (2016b). Data
88(B9): 7355–7371. integration, reservoir response, and application.
International Journal of Greenhouse Gas Control, 54(1):
Leetaru, H. E., and Freiburg, J. T. (2014). Litho-facies and 389–403.
reservoir characterization of the Mt Simon Sandstone
at the Illinois Basin–Decatur Project. Greenhouse Gases: Will, R., El-Kaseeh, G., Leetaru, H., Greenberg, S.,
Science and Technology, 4(5): 580–595. Zaluski, W., and Lee S.-Y. (2017). Quantitative
integration of time lapse seismic data for reservoir
Leetaru, H. E., and McBride, J. H. (2009). Reservoir simulator calibration: Illinois Basin–Decatur Project.
uncertainty, Precambrian topography, and carbon Carbon Capture, Utilization & Storage Conference,
sequestration in the Mt. Simon Sandstone, Illinois Chicago, IL, April 10–13.
Basin. Environmental Geosciences, 16(4): 235–243.
Yang, Y., Zoback, M., Simon, M., and Dohmen, T. (2013).
Palkovic, M. (2015). Depositional characterization of the Eau An integrated geomechanical and microseismic study
Claire Formation at the Illinois Basin–Decatur Project: of multi-well hydraulic fracture stimulation in the
Facies, mineralogy and geochemistry. M.S. thesis, Bakken Formation. SPE paper 168778 presented at the
University of Illinois at Urbana-Champaign. Unconventional Resources Technology Conference,
Pearson, C. (1981). The relationship between Denver, CO, August 12–14.
microseismicity and high pore pressures during Zoback, M. D., and Gorelick, S. M. (2012). Earthquake
hydraulic stimulation experiments in low permeability triggering and large-scale geologic storage of carbon
granitic rocks. Journal of Geophysical Research, 86: dioxide. Proceedings of the National Academy of Sciences
7855–7864. of the USA, 109(26): 10,164–10,168.
369
Part IV Summary
Chapter
What Next?
20 Thomas L. Davis, Martin Landrø, and Malcolm Wilson
Perhaps the most significant international event the field. Indeed, pipeline specifications preclude high
recently that affects the capture and use of carbon concentrations of CO2 in natural gas streams, so the
dioxide (CO2), and hence the topic of this book, was produced gas must be processed first to remove most
the United Nations Framework Convention on of the CO2 entrained in the gas stream. Instead of
Climate Change (UNFCCC) Paris Agreement. This venting, this CO2 is compressed and injected into the
Agreement was adopted in December 2015 and came expansive Utsira Formation.
into force less than a year later in November 2016. The Norwegian government has now embarked on
The basis of this agreement was to work cooperatively a more challenging project compared to the Sleipner
to restrict global temperature rise to between 1.5°C case: to capture CO2 from onshore high emission CO2-
and 2°C above pre-1850 levels. It is clear from groups sources such as a waste recycling plant and a cement
like the Intergovernmental Panel on Climate Change plant. The idea is to transport the captured CO2 via
(IPCC) and the International Energy Agency (IEA) ship to a subsurface storage site offshore. This is
that meeting such aggressive targets will require the a costly project, which will not be possible without
large-scale deployment of CO2 Capture, Transport a willingness to spend government money for this
and Geological Storage or Geosequestration (both as purpose. Presently, Equinor (formerly Statoil) has
storage in saline aquifers and in mature oil fields as been given the operator responsibility for this project
enhanced oil recovery [EOR]). Such technology will and has teamed up with Shell and Total as partners.
be applicable to large point sources of emissions such The final investment decision is scheduled for 2019,
as electrical generation facilities, refineries, cement with the Front-End Engineering Design (FEED) study
plants, and gas processing facilities. in progress, and the future will show if this project can
The idea of putting CO2 back into the subsurface be realized or not.
has now matured for decades. However, the progress Norway is not the only country to take leadership
is slow, mainly because of cost issues related to the with projects that improve our understanding and
capture process. One hundred and twenty years after provide the technological platforms for the continued
the famous publication of Svante Arrhenius in development of the processes mentioned in the text
Philosophical Magazine, where he suggested that the that follows. Here we see developments in Canada
increase in atmospheric CO2 could alter the Earth’s (Weyburn and Quest) and in the United States
surface temperature, there has probably been more (several Regional Sequestration Partnerships). Both
spent on research to determine the veracity of his of the Canadian projects are fully integrated with
finding than on remediation. It is always a good idea capture plants; Boundary Dam (a coal-fired electrical
to look back in order to predict the future. From the generation facility) is linked to Weyburn, an EOR
technical and scientific point of view, there is no project, and Quest (a capture unit on a refinery/
doubt that underground storage of CO2 is feasible upgrader) is linked to a saline aquifer storage facility.
and that we, with a high degree of confidence, can Excess CO2 produced from Boundary Dam can be
monitor and ensure that the CO2 is stored safely. routed to a saline aquifer geosequestration site to
The Sleipner project clearly shows that a tax level of prevent any venting to the atmosphere.
30–50 USD per tonne CO2 is sufficient to make such Geophysical methods related to safe storage of
a project economically feasible, given that the CO2 is CO2 underground are crucial, both in ensuring that
captured directly from the methane gas produced at the CO2 is actually stored where we want and also in 371
Thomas L. Davis, Martin Landrø, and Malcolm Wilson
detecting potential leakage or pressure buildups as we are to assess the use of CO2 in shallow horizons for
early as possible. The rapid development of cheaper increasing oil recovery.
geophysical listening equipment is critical in this There are interesting developments related to
respect. Of particular interest here is Distributed other geophysical methods, such as gravimetric and
Acquisition Systems (DAS) technology, enabling the electromagnetic methods (Chapters 7 by Eiken and 9
use of cheap fiber optic cables to be used as receivers. by Gasperikova and Commer, this volume). Gravity is
Currently, most of the testing for this technology has sensitive for density variations in the subsurface and
been in wells (see Chapters 15 by White and 16 by time-lapse gravity has now been tested and proved for
Lawton et al., this volume). So far, the best response several fields. It is attractive both as a standalone
has been achieved when the fiber measures seismic monitoring tool and as a complement to seismic
waves traveling in the same direction as the fiber-optic methods. In contrast to the seismic method, this
cable. There are solutions being tested where the cable method is sensitive to only one geophysical para-
has a helical shape to mitigate the weak signal in the meter: density. Time-lapse electromagnetic methods
perpendicular direction. We think that in the future it are sensitive to subsurface saturation changes, and are
might be possible to monitor CO2 storage by deploy- still under development. Again, there is a significant
ing two-dimensional (2D) fiber optic cables at the future potential for this method, both as a standalone
seabed or the land surface for monitoring purposes. tool and used in combination with other geophysical
Such systems will be significantly cheaper than the methods.
Permanent Reservoir Monitoring (PRM; for example, The In Salah CO2 storage project demonstrated
at Aquistore in Chapter 15 by White, this volume) that satellite altimetry offers a nice opportunity to
solutions that are presently used for geosequestation monitor onshore injection projects. Surface move-
sites and hydrocarbon reservoirs. Horizontal fiber ments were measured with a precision of millimeters.
optic cables at the surface will probably record Combined with geomechanical modeling such mea-
refracted waves and diving waves sufficiently accu- surements give us reliable and very useful monitoring
rately to be exploited for monitoring purposes. The data. This technique is being tested in a more challen-
time-lapse refraction method has been tested for sev- ging environment at Aquistore (Chapter 15 by White,
eral examples (North Sea and onshore fields in North this volume). Such data are also used for monitoring
America) so far, demonstrating that this method has of landslides and movements of large rock volumes
a potential to detect and map shallow gas migration or and faults in mountains.
CO2 migration through the subsurface. This method EOR has generally taken the lead in the develop-
is particularly sensitive to thin accumulations of CO2 ment of geophysical techniques because of the eco-
with large horizontal extension. It is therefore very nomic return to the producer. This has led to most of
likely that we will see a rapid development related to the advances over the last number of decades. Many
cheaper and more cost-effective methods for monitor- CO2 EOR projects involve water alternating gas
ing the shallow subsurface. (WAG) injection which is very difficult to monitor
In the last two decades there has been a tremen- unless one looks specifically at stress change and not
dous increase within the field of near surface geophy- saturation change alone. Detection of stress change
sics (see Chapters 16 by Lawton et al. and 17 by Juhlin versus fluid saturation change has been largely
et al., this volume). Conferences and workshops dedi- ignored except in some cases in the Norwegian
cated to this topic have attracted a large number of North Sea, where time-lapse amplitude versus offset
participants in recent years. This interest is encoura- (AVO) and time-shift studies have been conducted to
ging and it is very likely that this will attract vendors separate the effects of stress change from saturation
and new companies developing geophysical instru- change or in the case of multicomponent seismic
ments and methods dedicated to near surface map- surveying conducted by the Colorado School of
ping and monitoring. The near-surface work will be Mines. The opportunity exists to enhance the value
an essential component for the demonstration of safe of seismic information by designing surveys specifi-
storage in the deeper subsurface and early warning of cally to include wide-azimuth multicomponent
any leaks that may occur in either geosequestration permanent reservoir monitoring (PRM) to monitor
372 sites or EOR projects. Indeed, this will be essential if stress and saturation change. In doing so, we foresee
What Next?
the opportunity to record data that will enable more has been strengthened by the Paris agreement.
accurate quantitative interpretation. Already we are However, political actions related to implementation
seeing the value of improved data interpretation in of tax systems for CO2 emissions are not really in
determining stress directions and stress changes in place yet (although many countries and subnational
the subsurface as a means of guiding re-fracking jurisdictions have taken some leadership in this
operations. In the future, this will guide the applica- regard, following Norway’s example). We think this
tion of CO2 as an enhanced recovery agent and for will come, and that this will be an enabler for CCS
geosequestration in fractured formations. technology in the future, and hence also for geophy-
Although the number of storage projects world- sical monitoring of the CCS process.
wide is not increasing rapidly at present, we think that It is hoped that the reader will find the chapters in
the number of projects will, however, increase in the this book to be a guide to the current state-of-the-art
years to come. After all, carbon capture and storage and that it will provide a sense of where mainstream
(CCS) offers a possibility to store relatively large geophysics will be headed in the future as it relates to
volumes of CO2. The political willingness to do so geosequestration.
373
Index
4D attributes, 355 bulk modulus, 14, 18, 20, 22, 41, 64, 72, cross-correlation, 217, 322, 326, 327,
accelerometers, 171 73, 74, 75, 136, 292, 311, 313, 323 334
acoustic properties, 74, 216, 220 buoyant, 7, 199, 211, 212, 228 crosswell tomography, 260, 262
Ampere’s law, 159
amplitude versus angle (AVA), 40, 134, caliper log, 183 damage zone, 104, 105, 106, 107
153, 223, 233, 264, 311 capillary entry pressure, 319 Darcy flow, 226, 227, 228
amplitude versus offset (AVO), 27, 28, capillary flow, 226, 227 Delhi Field, 235, 246, 248, 251, 254,
35, 40, 41, 43, 53, 60, 61, 66, 67, 68, capillary trapping, 335 255, 256
129, 134, 135, 136, 138, 139, 141, 144, carbon capture and storage, 4, 280 depleted gas field, 123, 132, 175, 195,
145, 148, 149, 152, 152, 223, 232, 311, cased-hole logs, 181 278
320, 322, 334, 336, 337, 338, 372 CCS legislation, 209 depleted reservoirs, 114, 178
anisotropic, 14, 29, 61, 83, 84, 85, 134, chimney, 87, 219, 220, 227, 230 differential interferograms, 109
172, 226, 264, 358 clastic reservoir, 24, 260 differential pressure, 328
anomalies, 26, 35, 36, 38, 39, 122, 135, CO2 plume, 7, 12, 16, 17, 27, 28, 29, 32, dipole array, 158
154, 165, 166, 190, 262, 272, 276, 292, 34, 37, 39, 40, 47, 54, 62, 64, 65, 66, displacements, 8, 45, 94, 101
293, 325, 356, 358 81, 123, 126, 127, 128, 130, 131, 136, dissolution, 7, 24, 25, 27, 40, 74, 77,
aperture, 28, 40, 72, 101, 102, 103, 104, 152, 158, 159, 161, 165, 172, 173, 174, 123, 128, 129, 130, 131, 155, 198, 224,
106, 107, 113, 171, 275 211, 214, 216, 222, 223, 224, 225, 226, 225, 227, 230, 232, 238, 260, 275, 282,
apparent resistivity, 157, 158, 161, 307 228, 229, 230, 231, 232, 233, 234, 262, 284, 291, 294, 297, 306, 307, 310, 314,
aquifer, 5, 26, 27, 28, 42, 45, 52, 62, 63, 273, 275, 276, 277, 282, 284, 289, 290, 315, 335, 336, 355
64, 67, 68, 81, 82, 103, 112, 123, 124, 291, 292, 293, 294, 295, 297, 301, 306, Distributed Acquisitions Systems
125, 131, 152, 166, 181, 189, 192, 195, 308, 314, 315, 316, 319, 328, 332, 333, (DAS), 73, 257, 266, 267, 273, 275,
197, 201, 203, 209, 216, 230, 231, 232, 334, 348, 352, 368 372
252, 278, 316, 319, 323, 337, 371 CO2 saturation, 15, 20, 22, 23, 26, 39, downhole fluid analysis, 62
Aquistore, v, 26, 33, 53, 107, 109, 110, 40, 41, 42, 45, 51, 55, 61, 65, 73, 74, downhole monitoring, 172, 173, 175,
111, 112, 114, 257, 258, 259, 260, 262, 75, 76, 77, 78, 79, 80, 82, 123, 125, 177, 274
265, 266, 267, 268, 269, 270, 271, 273, 136, 137, 138, 139, 140, 141, 144, 147,
275, 276, 277, 278, 279, 280, 296, 372 149, 151, 155, 156, 159, 161, 165, 190, earthquake, 13, 14, 43, 45, 46, 47, 96,
Archie’s law, 79, 155 192, 221, 223, 229, 263, 264, 267, 273, 112, 170, 172, 180
Arrhenius, ix, 371 291, 292, 293, 308, 310, 311, 313, 314, effective pressure, 322, 323, 324
Arrival-time inversion, 172 319, 323, 324, 326, 327, 333, 334, 335, elastic, 14, 15, 18, 19, 22, 26, 29, 30, 33,
attenuation, 18, 29, 74, 75, 76, 81, 82, 336, 339 40, 41, 42, 43, 53, 57, 58, 60, 71, 72,
83, 84, 172, 216, 368 CO2 storage atlas, 195 73, 76, 77, 82, 83, 84, 85, 122, 152,
azimuthally anisotropic media, 83 compaction, 15, 26, 28, 32, 37, 52, 53, 152, 186, 187, 223, 232, 233, 292, 293,
100, 127, 132, 202, 282, 337 318, 323, 328, 346, 353, 367
compressibility, 13, 18, 20, 21, 22, 23, electrical resistance tomography, 72,
background microseismic activity, 38, 72, 73, 84, 120, 201, 243, 294 79, 166
346 conductivity, 57, 58, 60, 61, 62, 64, 73, electrical resistivity, 57, 62, 64, 66, 68,
Backus averaging, 136 78, 79, 80, 156, 161, 268, 310 78, 79, 80, 154, 155, 161, 165, 166,
Bayesian inversion, 131, 222, 232 containment risk, 211 297, 299, 301, 307, 309, 315, 316, 317,
benchmarks, 93, 118, 121, 127, 224 continuous active-source seismic 318
borehole gravity, 118, 259, 267 monitoring, 64 electrical resistivity tomography
borehole-to-surface electromagnetic continuous monitoring, 65, 66, 126, (ERT), v, xii, 62, 66, 72, 79, 80, 160,
(BSEM), 257, 268, 269, 277 274 161, 162, 163, 164, 165, 167, 297, 298,
Bouguer, 121, 268 controlled source electromagnetic, 299, 301, 302, 303, 304, 307, 308, 309,
braided channel, 238 159 311, 315, 318
Brie, 221, 323, 325, 326, 336 critical point, 20, 27, 118, 136, 155, electrode, 156, 157, 160, 162, 163, 167,
Brune model, 169 229 181, 185, 269, 303, 304, 307, 311 375
Index
electromagnetic, xii, 59, 81, 94, 154, hydraulic fracture, 169, 174, 369 mobility, 80, 235, 246, 252, 254, 313,
158, 166, 167, 278, 279 hydrology monitoring, 115 319, 331, 332, 333, 334, 335
enhanced oil recovery (EOR), ix, 5, 7, Mohr circle, 168, 169
11, 20, 52, 57, 71, 100, 125, 169, 181, Illinois Basin–Decatur, 176, 177, 178, moment magnitude, 170, 176, 177
195, 235, 246, 258, 280, 281 180, 368, 369 multicomponent seismic data, 83, 84,
error ellipsoid, 172 immiscible, 136, 235, 250, 332 85, 87, 152, 246, 254, 256
evaporitic, 260, 275 impermeable, 12, 64, 197, 198, 327,
332, 340 natural seepage, 203
fall-off data, 329, 330, 331 In Salah, 8, 28, 63, 69, 94, 101, 103, 104, neutron logs, 186, 367
Faraday’s law, 158, 159 111, 112, 113, 121, 122, 129, 131, 132, neutron porosity, 183, 190, 264
fiberglass casing, 186 175, 176, 180, 232, 278, 336 nodal receiver, 248
fiber optic cable, 266, 273 induced seismicity, 60, 64, 168, 170, nonuniqueness, 15, 28, 102, 111,
flow simulation, 29, 32, 135, 178, 233, 171, 172, 175, 178, 180, 279, 362 163
256, 273, 337 inelastic, 168, 186, 216, 232 normalized root mean square (NRMS),
fluid substitution, 72, 73, 74, 75, 77, 78, injection point, 27, 29, 31, 42, 46, 119, 26, 31, 33, 34, 35, 36, 47, 88, 89, 216,
230, 251, 254, 292, 311, 316, 322, 323, 134, 136, 137, 138, 140, 141, 142, 145, 238, 251, 308, 327, 355, 357, 358, 360,
328, 337, 367 144, 147, 148, 150, 150, 174, 176, 209, 361
formation water salinity, 182, 183, 192 211, 212, 216, 217, 219, 223, 226, 228 Norwegian Petroleum Directorate
four-component seismic, 83 injection wells, 27, 28, 38, 39, 93, 103, (NPD), 195, 196, 197, 198, 208
fracture slip, 169 111, 118, 197, 235, 243
fractured reservoir, 84 injectivity, 8, 12, 27, 52, 103, 199, 201, ocean acidification, 2
full waveform inversion, 43, 52, 233 203, 205, 289, 319, 320, 328, 329, 330, ocean tides, 117
331, 335, 336, 337, 364 offset, 40, 42, 60, 61, 91, 132, 135, 136,
gamma ray log, 183, 287 interferogram, 95, 96, 97, 98 138, 139, 140, 141, 142, 142, 144, 145,
Gassmann equation, 22, 72, 74 interferometric synthetic aperture 144, 146, 148, 151, 159, 160, 169, 177,
Gaussian noise, 124, 309 radar (InSAR), 8, 13, 28, 45, 60, 72, 216, 223, 233, 238, 243, 268, 269, 273,
geochemical, 12, 15, 18, 24, 25, 26, 27, 93, 94, 95, 96, 97, 99, 100, 101, 102, 287, 300, 309, 311, 320, 322, 326, 327,
28, 29, 40, 56, 58, 61, 62, 63, 65, 66, 103, 107, 111, 112, 113, 114, 122, 175, 347
67, 71, 73, 74, 77, 78, 80, 82, 103, 107, 180, 257, 258, 259, 265 Ohm’s law, 156, 157
165, 280, 291, 353 Intergovernmental Panel on Climate open-hole logs, 181, 189, 191
geodesy, 93 Change (IPCC), ii, 1, 2, 4, 5, 7, 8, 11, Otway, 22, 23, 29, 31, 32, 34, 35, 37, 38,
geodetic techniques, 93 55, 68, 71, 81, 335, 336, 337, 371 46, 47, 48, 49, 50, 51, 52, 53, 68, 71,
geomechanical effects, 7, 26, 27 isopach map, 243, 244 80, 175, 179
geophone, 34, 35, 47, 83, 88, 171, 258,
266, 267, 270, 271, 274, 347 Keeling curve, 1 passive microseismic monitoring, 168,
geophysics, ix, x, 12, 65, 71, 93, 158, Ketzin, v, 66, 80, 166, 167, 276, 277, 178, 366
167, 348, 372, 373 279, 297, 298, 299, 300, 301, 302, 304, patchy saturation, 23
geopressure, 85 306, 307, 309, 310, 311, 313, 315, 316, perforated interval, 192
geosequestration, i, ii, v, 53, 54, 169, 317, 318 Permanent Reservoir Monitoring
181, 235, 246, 371 kriging, 288, 289, 365 (PRM) systems, 87
geothermal, 12, 15, 20, 27, 115, 118, permeability, 5, 12, 58, 61, 62, 73, 85,
119, 126, 128, 132, 169, 178 lateral velocity variations, 172 86, 93, 101, 105, 107, 114, 120, 126,
GPS monitoring, 265 leakage detection, 55, 88, 231, 275 159, 174, 181, 185, 203, 214, 226, 227,
gradiometry, 118 liquid natural gas (LNG) production, 228, 230, 232, 235, 238, 243, 246, 251,
gravity monitoring, 67, 72, 115, 118, 195 252, 255, 259, 273, 282, 283, 287, 288,
121, 123, 125, 126, 130, 132, 152, 224, lithology, 58, 61, 67, 88, 183, 187, 190, 289, 290, 291, 293, 295, 296, 313, 320,
230 252, 255, 287, 300, 304 325, 328, 329, 331, 332, 333, 334, 335,
greenhouse effect, ix 340, 366, 369
greenhouse gas emissions, 2, 3, 4, 107 magnetic fields, 154 petrophysical, 60, 61, 62, 64, 67, 189,
groundwater, 59, 60, 64, 118, 283 magnetic permeability, 159 201, 251, 282, 284, 287, 289, 294, 310,
guided-wave, 159 magnetotelluric, 260, 279 311, 313, 314, 317, 346, 365
Gutenberg–Richter law, 168, 170 marine 4C acquisition, 87 phase shifts, 94
microgravity, 64, 67, 115, 129, 131, 277 Poisson’s ratio, 83, 84
Hashin–Shtrikman, 18, 367 microseismic events, 46, 47, 169, 170, pore pressure, 15, 18, 19, 26, 27, 39, 41,
helical, 372 174, 175, 176, 274, 350, 362, 363 42, 59, 60, 72, 125, 136, 168, 178, 198,
Hertz–Mindlin, 19, 323 microseismicity, 7 201, 202, 203, 255, 319, 320, 322, 323,
heterolithic stratification, 282 mineral, 12, 19, 27, 77, 156, 198, 228, 328, 363
history matching, 125 292, 310, 335, 352 porous media, 53, 71, 72, 80, 160, 296,
376 horizontal electric dipole, 159, 160 mitigation, 1, 2, 3, 4, 8, 55, 203, 337 316, 317
Index
porous rock, 13, 18, 126, 292 seafloor monitoring, 226 S-wave anisotropy, 86
Postle Field, 235, 236, 237, 238, 239, seal, 26, 42, 54, 58, 62, 64, 93, 103, 107, synthetic aperture radar, 93, 94, 103,
243, 254 168, 172, 197, 199, 201, 202, 203, 205, 109, 111, 265
Precambrian basement, 257, 260, 276, 209, 212, 214, 227, 228, 233, 257, 260, synthetic seismograms, 269, 276,
282, 342 275, 280, 283, 287, 291, 340, 346, 350, 300
pressure buildup, 7, 93, 111, 121, 226, 362
250, 319, 334 sealing efficiency, 12 tilt meters, 63, 258, 259, 265
pressure front, 47, 55, 239, 243, 327 seismic imaging, 15 time-lapse gravity, 64, 67, 115, 122,
pressure gauges, 214 seismic inversion, 16 133, 265
pressure relaxation, 306 seismic moments, 178 time-lapse seismic, 8, 13, 14, 15, 18, 22,
pressure–saturation inversion, 319 seismic reflection interferometry, 269, 24, 26, 31, 33, 35, 36, 37, 38, 39, 41,
prestack inversion, 134 277 42, 45, 134, 135, 140, 173, 233, 273,
principal stress, 168 seismometer, 171, 175, 176, 177, 348 275, 280, 315, 333, 334, 335, 336,
PS-converted data, 87 sensor drift, 117, 118 367
public assurance, 54, 55, 64, 65, 66 shear modulus, 14, 18, 41, 72, 73, 169, time-lapse seismology, 13, 14
pushdown, 216, 219 313, 322 tomographic image, 39, 40, 160
P-wave, 19, 22, 23, 27, 38, 39, 40, shear slipping, 169 tracer, 61, 62, 165, 252, 255
42, 45, 72, 73, 75, 76, 77, 78, 85, 87, side-scan sonar, 226 transgressive, 247, 282
88, 90, 91, 134, 136, 137, 138, 139, signal-to-noise ratio, 97, 173, 263, 303 tube wave, 264
140, 144, 151, 153, 220, 223, 233, 239, skin effects, 328 tuning thickness, 221, 308, 323
243, 248, 251, 255, 285, 292, 293, 300, Sleipner, v, ix, x, 7, 8, 10, 20, 23, 29, 31, turbiditic sand, 209
301, 320, 321, 323, 324, 328, 329, 330, 38, 41, 42, 64, 66, 71, 80, 103, 119,
335 121, 126, 127, 130, 131, 132, 134, 135, uncertainty analysis, 15, 52, 234
136, 138, 139, 144, 146, 149, 151, 151, uncertainty bounds, 223, 230
quantitative analysis, 3, 71, 72, 186 153, 195, 209, 211, 212, 214, 222, 223, unconsolidated sand, 13, 18, 23, 29, 30,
quantitative interpretation, 74, 79, 85, 226, 228, 229, 230, 231, 232, 233, 234, 183
182, 310, 373 276, 282, 295, 297, 316, 318, 336, 371 uniform mixing, 222
Smith and Gidlow, 320, 322, 323, 328 Utsira Formation, 132, 135, 136, 137,
radar sensor, 94, 95 solution instability, 163 151, 195, 233
radioactivity, 183 sonic logs, 51, 172, 183, 251, 336
regularization, 100, 102, 103, 163, 308 source intensity, 168 variogram analysis, 288
regulatory compliance, 55 Span and Wagner, 136, 332 vertical magnetic dipole, 159
relative spring meter, 116 Spatial clustering, 173 vertical seismic profile (VSP), 34, 39,
Reservoir Characterization Project spectral decomposition, 221 52, 59, 61, 64, 67, 78, 81, 88, 89, 92,
(RCP), i, ii, 85, 86, 235, 247, 250 spontaneous potential, 59, 181, 284 237, 239, 257, 259, 266, 267, 271, 273,
reservoir simulation, 132, 133, 201, static corrections, 304, 308 274, 275, 276, 277, 278, 345, 347, 348,
233, 243, 246, 254, 256, 276, 280, 284, steady-state flow, 330, 331 350, 351, 357, 358, 359
289, 291, 366, 367 storage efficiency, 5 vibrator trucks, 347
residual, 7, 16, 20, 23, 36, 40, 117, storage integrity, 7 virtual shot gathers, 269
118, 119, 190, 198, 235, 246, 251, 296, stratigraphic, 12, 58, 173, 175, 198, 232, Voigt–Reuss, 19
299, 323, 328, 331, 333, 334, 336, 353, 235, 236, 282, 284, 335, 347, 365,
358 367 warping algorithms, 83
Reuss, 18, 19, 23, 73, 75 stratigraphic layers, 173 water alternating gas (WAG), 235, 238,
reverse time migration, 85 strike-slip mechanism, 177 239, 243, 246, 256, 372
rim, 127, 218, 220, 222, 227, 233 structural, 12, 58, 61, 62, 103, 154, 155, water influx, 128, 224, 225
risk assessment, 10, 55, 56, 57, 161 161, 195, 198, 209, 216, 264, 276, 284, wave-induced fluid flow, 222, 233
rock and fluid physics, 14, 18, 26, 40 288, 289, 299, 301, 335, 342, 346, 347, wedge, 221, 235, 247, 336
353, 368 wellhead, 63, 120, 211, 212, 213
saline aquifers, ix, 10, 12, 18, 20, 80, subsidence, 98, 121, 130, 131, 132 well logging, 61, 63, 71, 76, 77, 79, 82,
132, 166, 178, 195, 197, 201, 205, 231, subsurface changes, 15, 54, 56, 166 181, 259, 311, 317
232, 257, 337, 371 supercritical, 19, 119, 199, 318, 319, Weyburn, ii, ix, 7, 8, 10, 11, 21, 23, 24,
salt precipitation, 77, 316, 319, 328, 328, 332, 336 33, 38, 39, 52, 103, 114, 175, 180, 258,
329, 331, 334, 335, 337 swaths, 45, 216, 301 274, 276, 279, 281, 371
satellite, 8, 13, 28, 45, 63, 94, 95, 96, 98, S-wave, 14, 22, 25, 27, 28, 38, 39, 40, 41, Wood’s law, 20
99, 102, 103, 104, 107, 111, 113, 118, 77, 78, 83, 86, 87, 88, 134, 135, 152,
175, 205, 372 223, 239, 243, 251, 255, 292, 293, 321, Zoeppritz equation, 139, 323, 327,
Schlumberger array, 157 324, 348 328
377

Geophysics

Uploaded by

Copyright:

Available Formats

Geophysics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Geophysics

Uploaded by

Copyright:

Available Formats

Geophysics and

Thomas L. Davis is Professor Emeritus at the Colorado School of Mines and

Martin Landrø is Professor at the Norwegian University of Science and

Malcolm Wilson has worked within government, academia, and the

Cambridge University Press is part of the University of Cambridge.

Part I Introduction 10 Microseismic Imaging of CO2 Injection 168

2 The Role of Geophysics in Carbon Capture and

18 Time-Lapse Seismic Analysis of the CO2 Part IV Summary

B magnetic ﬁeld ERT electrical resistance tomography

1 Capture and Storage in Mitigation

Cumulative reductions by sector and technology

Transport Fuel switching

Industry Energy efficiency

Table 1.2 IEA GHG CO2 monitoring selection tool

1.6 Practical Experience in various geological formations within a wide range

Introduction a potential future global market in CO2 sequestration

Baseline Monitor M-B

Red – Softening of Reservoir, caused by either:

Baseline Monitor M-B

Blue – Hardening of Reservoir, caused by either:

imaging is the process of using seismic data to image

3000 Voigt Avg.

Bulk Modulus Difference [MPa]

Bulk Modulus Difference [MPa]

Density Difference [kg/m3]

Temperature [0C] Temperature [0C] Temperature [0C]

800 800 –200

the seismic response, and large dry-frame compressi- Patchy Saturation

Top Pyrenees (gas)

Pyrenees Sands oil-water-contact

200 T0 T1-T0 T2-T0 200

receiver positioning diﬀerences between surveys; nat-

where RMS is the standard deﬁnition of “root-mean-

5734000 658000 SSTVD

Time-Lapse Seismic Imaging

4D processing flow, Similar 3D processing flows, Similar 3D processing flows,

750 750 750 750 750 750

Real 4D fluid flow anomaly

CRC 2 CRC 2 CRC 2

Baseline Parallel Simultaneous

Figure 2.19 Map view of time-lapse seismic

amplitude diﬀerence maps are primarily sensitive to

matrix (e.g., dissolution of minerals and grain

Figure 2.22 Time-lapse seismic

At GOC1295.5m TVDSS (+/–1m) At 1–3m below GOC

VANCOGH-1 ST1 VANCOGH-1 ST1

At 1301.5m TVDSS (+/–1m) At OOWC 1307.7m TVDSS (2–4m below)

VANCOGH-1 ST1 VANCOGH-1 ST1

1.550 1.600 1.650 1.700 1.750 1.800

False Northing [m]

Time [s] –1300

funding through its industry members and research

Pre-injection CO2 injection

Pre-injection CO2 injection

Resources, Energy and Tourism (DRET),

3 Why and How to Assess the Subsurface Changes

Why Monitor? characterized, along with the technology used, are

Possible site Probable site Proven site

Data Injection &

Geophysical Monitoring It is important to understand the physical para-