Originally Published As

Originally published as:
Zang, A., Oye, V., Jousset, P., Deichmann, N., Gritto, R., McGarr, A., Majer, E., Bruhn, D. (2014): Analysis of induced
seismicity in geothermal reservoirs – An overview. - Geothermics, 52, p. 6-21.
DOI: http://doi.org/10.1016/j.geothermics.2014.06.005
Geothermics 52 (2014) 6‐21
doi: 10.1016/j.geothermics.2014.06.005
Analysis of induced seismicity in geothermal reservoirs – An overview
Arno Zanga,∗, Volker Oyeb, Philippe Joussetc, Nicholas Deichmannd, Roland Grittoe,
Art McGarrf, Ernest Majerg, David Bruhnc
a German Research Center for Geosciences (GFZ), Section 2.6 Seismic Hazard and Stress Field, Telegrafenberg, 14473 Potsdam, Germany
b NORSAR, P.O. Box 53, N‐2027 Kjeller, Norway
c GFZ, International Center for Geothermal Research, 14473 Potsdam, Germany
d Swiss Seismological Service, ETH Zürich, Sonneggstrasse 5, CH‐8092 Zürich, Switzerland
e Array Information Technology, 2020 Cedar Street, Berkeley, CA 94709, USA
f U.S. Geological Survey, Menlo Park, CA 94025, USA
g Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
ARTICLE INFO ABSTRACT
Article history: In this overview we report results of analysing induced seismicity in geothermal reservoirs
Received 2 November 2012 in various tectonic settings within the framework of the European Geothermal Engineering
Received in revised form 23 May 2014 Integrating Mitigation of Induced Seismicity in Reservoirs (GEISER) project. In the
Accepted 16 June 2014 reconnaissance phase of a field, the subsurface fault mapping, in situ stress and the seismic
Available online 7 July 2014 network are of primary interest in order to help assess the geothermal resource. The
hypocentres of the observed seismic events (seismic cloud) are dependent on the design of
the installed network, the used velocity model and the applied location technique. During
Keywords: the stimulation phase, the attention is turned to reservoir hydraulics (e.g., fluid pressure,
Fluid‐induced seismicity injection volume) and its relation to larger magnitude seismic events, their source
Key reservoir parameters characteristics and occurrence in space and time. A change in isotropic components of the
Enhanced geothermal systems full waveform moment tensor is observed for events close to the injection well (tensile
Larger magnitude events character) as compared to events further away from the injection well (shear character).
Maximum observed magnitude Tensile events coincide with high Gutenberg‐Richter ‐values and low Brune stress drop
Crustal stress values. The stress regime in the reservoir controls the direction of the fracture growth at
depth, as indicated by the extent of the seismic cloud detected. Stress magnitudes are
important in multiple stimulation of wells, where little or no seismicity is observed until the
previous maximum stress level is exceeded (Kaiser Effect). Prior to drilling, obtaining a 3D
‐wave ( ) and ‐wave velocity ( ) model down to reservoir depth is recommended. In
the stimulation phase, we recommend to monitor and to locate seismicity with high
precision (decametre) in real‐time and to perform local 4D tomography for velocity ratio
( / ). During exploitation, one should use observed and model induced seismicity to
forward estimate seismic hazard so that field operators are in a position to adjust well
hydraulics (rate and volume of the fluid injected) when induced events start to occur far
away from the boundary of the seismic cloud.
1. Introduction production at hydrocarbon reservoirs (e.g., Suckale, 2009) and

the disposal of wastewater in the subsurface (e.g., Healy et al.,
Over geological time scales tectonic forces build up stress 1968). Induced seismicity related to deep geothermal applica‐
in the Earth’s crust (Zang and Stephansson, 2010). Apart from tions has been studied for many years, e.g. as part of the Fen‐
crustal creep, earthquakes are the primary brittle processes ton Hill, New Mexico hot dry rock experiment (HDR, Pearson,
that partially relax stresses in the crust. Earthquakes of tecton‐ 1981; Bame and Fehler, 1986; Ferrazzini et al., 1990), but
ic or volcanic origin occur when stresses on pre‐existing progress to understand this phenomenon slowed down soon
planes of weakness exceed their strength (e.g., Scholz, 1990). after. Recently, induced seismicity in geothermal applications
Induced earthquakes are similar, but occur when either stress (geothermal seismicity) has attracted public interest due to
or strength are perturbed by natural or anthropogenic influ‐ issues of seismic risk (e.g., Giardini, 2009).
ences (e.g., Trifu, 2002). Naturally induced earthquakes have The goal in engineered or enhanced geothermal reservoir
been associated with heavy rainfalls (e.g., Grünthal, 2014, operations is to locate or safely create permeability through
induced seismicity category 7) or with transient surface waves e.g. fractures in high‐temperature rock such that water and
from major tectonic earthquakes (e.g., Jousset and Rohmer, steam can circulate through these pathways to efficiently
2012). Man‐made induced seismicity has been observed in transfer heat to the surface. While in the initial HDR concept,
numerous settings (e.g., Trifu, 2010). Among them are im‐ inclined injection and production wells are connected via
poundment of lakes behind dams giving rise to reservoir‐ discrete, parallel hydraulic tensile fractures (engineered geo‐
triggered seismicity (e.g., Talwani and Acree, 1985), mining‐ thermal system), in enhanced geothermal systems (EGS) the
induced seismicity (e.g., Mendecki, 1997), the injection and natural rock joint network permeability is increased by mas‐
sive water injection in extended open‐hole sections (Jung,
2013). Tester et al. (2006) defined EGS as engineered reser‐
voirs designed to economically extract heat from low permea‐
* Corresponding author. Tel.: +49 331 288 1325; fax: +49 331 288 1127. bility formations. In places where fractures are not naturally
E‐mail address: zang@gfz‐potsdam.de (A. Zang) occurring, new ones need to be created or existing ones reac‐
6
A. Zang et al. / Geothermics 52 (2014) 6‐21
tivated. Most EGS projects require drilling of boreholes to

reach adequate temperatures at depth and pumping of fluids
at high pressures to enhance permeability through hydraulic
fracturing (mode I crack), hydro‐shear (mode II crack), com‐
bination of both (mixed modes), or acidizing. In this article,
geothermal seismicity (sensu stricto) is associated with the
development and operation of EGS systems.
Hydraulic fracturing has become a standard technology in
geothermal stimulation (Bromley, 2010) and has been for
decades in oil and gas applications (Maxwell et al., 2010). One
of the challenges in EGS is to increase permeability in a way
that the heat transfer is optimally efficient, which is a funda‐
mental difference to the challenges in unconventional reser‐
voirs (enhanced recovery). Pumping too much water, or too
fast, or in a critically stressed fault may create excessive per‐
meability, which may decrease the efficiency of the heat ex‐
changer and also may induce larger magnitude events thereby
increasing seismic hazard.
Induced seismicity in geothermal settings has been docu‐
mented in areas such as Indonesia (Silitonga et al., 2005; Mulya‐
di, 2010), the Philippines (Bromley et al., 1986), Japan (Nagano
et al., 1994), Kenya (Simiyu, 1999), North and South America
(Henderson et al., 2002; Yamabe and Hamza, 1996), Australia
(Baisch et al., 2006) and New Zealand (Hunt and Latter, 1982)
for over 40 years. In Europe, an early description of industrial
exploitation of geothermal resources was published by Batch‐
elor and Garnish (1990). Recently, Evans et al. (2012) compiled a
survey of induced seismicity responses to fluid injection in
European geothermal and CO2 reservoirs.
Annually, thousands of seismic events are generated dur‐
ing exploitation of geothermal fields but in most cases these
events are below local magnitude 2, and below the de‐
tection threshold of communities (e.g., Evans et al., 2012).
Geothermal sites in the Rhine Graben near Basel (Deichmann
and Giardini, 2009), Landau (Grünthal, 2014) and Soultz‐sous‐
Forêts (Dorbath et al., 2009), however, have experienced
2.5 events due to EGS activities (Table 1). Although this
seismicity has been short lived it has attracted public concern
due to its proximity to populated areas (Kraft et al., 2009). At
other sites, geothermal seismicity is entirely of low magni‐
tudes and then may drop below the detection threshold, pos‐
sibly due to geologic conditions, e.g. high attenuation of seis‐
mic waves caused by overlying sediment formations. One
example of low magnitude seismicity is the German site Groß
Schönebeck. Even massive hydraulic stimulations performed
by injection of 13,000 m3 of water into the well Gt GrSk 4/05
produced only 80 seismic events with moment magnitudes in
the range of −1.8 to −1.0 (Kwiatek et al., 2010). Another exam‐
ple of low seismicity rate is the Bouillante geothermal field
(French West Indies), where no earthquakes could be related
to the exploitation of the geothermal field (Sanjuan et al.,
2010). A possible explanation could be the absence of reinjec‐
tion. That is, the small pore pressure decrease (a few bars)
tends to strengthen rather than to weaken the rock mass and
thereby reduces the likelihood of induced seismicity. Evans
et al. (2012) investigated 41 European case histories of in‐
duced seismic responses to fluid injection. In 25 cases, the
injection was done into sedimentary rocks from which seven
experiments were associated with felt seismicity. In eight out
of 25 experiments with sedimentary rocks, injection occurred ervoir rocks can respond with tensile failure (hydraulic frac‐
into or close to faults from which only one site produced felt tures; Hubbert and Willis, 1957), or with shear failure of pre‐
seismicity, i.e. Unterhaching (Megies and Wassermann, 2014). existing joint sets (dilatant shear; Hubbert and Rubey, 1959).
In addition to the hazard of induced seismicity, it is also im‐ In the oil industry, the relationship between seismic events
portant to understand why geothermal seismicity is generated and hydraulic fractures has been studied extensively (Fix et al.,
at some sites and is evidently absent at others. 1989; Zhu et al., 1996; Dyer et al., 1999; Shapiro and Dinske,
The actual underlying physical mechanism of shear dilata‐ 2007; Fischer et al., 2008). In geothermal activities, no matter
tion, due to injection of large fluid volumes at high pressure, if the HDR, EGS, or multi‐fracturing concept is used the inter‐
however, is not yet fully understood. Depending on rock prop‐ action of tensile and shear fractures may complicate the seis‐
erties, injection pressure, fluid volume and temperature, res‐
7
mic footprint of a reservoir to be developed (Jung, 2013).

This overview article is structured according to the devel‐
opment of a geothermal field. In Section 2, the seismic re‐
sponse to fluid injected is evaluated. For this, the seismic net‐
work (borehole versus surface stations), velocity model and
structural geology aspects (subsurface faults) are discussed in
relation to hydraulic parameters, rock type and in situ stress
regime. In Section 3, one of the most challenging aspects in the
analysis of EGS seismicity is addressed, understanding the
mechanisms leading to larger magnitude events (LME). Here,
we discuss LME in relation to pore fluid pressure, injected
volume and in the context of fault and reservoir size. Note that
Grünthal (2014) replaces the term LME by SEECo, seismic
events of economic concern. In Section 4, the maximum ob‐
served magnitude of EGS seismicity is compared to the maxi‐
mum expected magnitude estimated from fault‐scaling rela‐
tionship, deterministic fracture models, probabilistic ap‐
proaches and empirical laws.
Figure 1. Observed maximum magnitude of seismic events in geo‐
thermal operations (squares), wastewater disposal wells (triangles),
2. Seismic response to fluid injection: network design, hydraulic fracturing (circles), and fluid injection in the KTB scientific
velocity model, hydraulic parameter and reservoir stress well (stars) as functions of volume of injected fluid. Numbers by sym‐
regime bols correspond to the order in which data are listed in Table 2.
Before a geothermal system can be exploited, it has to be

assessed. During this reconnaissance phase several geophysi‐ less fluid volume while smaller maximum magnitudes are
cal techniques can be used. In this section we focus on the observed. Lower‐bound values come from KTB stimulations
analysis of induced seismicity as a tool to probe key parame‐ (Fig. 1, stars #6 and #12) and the Groß Schönebeck geother‐
ters of the geothermal reservoir. The goal is to improve the mal site (Fig. 1, square #7). In Fig. 1, the dashed line (from
technique of EGS, first by investigating the role of induced McGarr, 2014) appears to define an upper bound to the data.
seismicity as an instrument to image fluid pathways generated A widely observed feature in the spatiotemporal distribu‐
by hydraulic stimulation treatments, and second, by address‐ tion of seismicity is a gradual migration from the vicinity of
ing the consequences of hydraulic treatments for potential the borehole to distances farther from the borehole as fluid
seismic hazard (Bruhn et al., 2011). We summarize and inter‐ injection is progressing. In many cases as injection proceeds
pret results from the analyses of various geothermal sites, seismicity continues throughout the volume rather than in an
located in different tectonic settings (Table 1). All of these expanding ring. This has implications on the rate of permeabil‐
sites have been investigated within the framework of the ity creation. The ability to track details in the spatiotemporal
European Union project GEISER. In Europe, locations include distribution of seismicity and its source parameters, however,
the Lower Rhine Graben site in Soultz‐sous‐Forêts, France depends primarily on the quality and the configuration of the
(Evans et al., 2005), the Upper Rhine Graben site in Basel, seismic network, on the quality of the velocity model and on
Switzerland (Häring et al., 2008), Icelandic geothermal test the applied algorithms to invert for event locations.
sites (Jousset et al., 2010, 2011), Latera (Italy) and Groß The total extension of the seismic network and the config‐
Schönebeck in the North German Basin (Kwiatek et al., 2010). uration of the seismic stations influence the completeness and
Non‐European sites investigated include Berlin, El Salvador accuracy of the event locations. Networks that include single
(Bommer et al., 2006; Kwiatek et al., 2014), The Geysers in borehole sensors or strings of deep borehole three‐component
California, USA (Oppenheimer, 1986; Majer et al., 2007), sensors can detect and locate more events at a lower magni‐
Cooper Basin (Asanuma et al., 2005; Baisch et al., 2006) and tude threshold than networks that consist solely of surface
Paralana (Hasting et al., 2011; Albaric et al., 2014), both in stations. This is primarily due to the shorter distances be‐
Australia and Bouillante in Guadeloupe (Sanjuan et al., 2010; tween the sensors and the event locations, and to avoiding the
Calcagno et al., 2012). Two reference sites were included, heterogeneous and attenuating near surface layers, as well as
where seismicity was induced by other than geothermal caus‐ to the improved signal‐to‐noise ratio from being below the
es. At the KTB deep drilling site, injection of 200 m3 of brine in surface. This results in reduced amplitude losses from geo‐
a stable continental region produced low magnitude seismicity metrical spreading and intrinsic and scattering attenuation, as
(Zoback and Harjes, 1997; Baisch and Harjes, 2003), and in The observed when comparing data from deep borehole with
Netherlands (NL) the induced seismicity is related to gas pro‐ surface seismic data (Oye et al., 2004). Near‐surface low‐Q
duction (van Eck et al., 2006; Dost and Haak, 2007; van Wees sedimentary layers (see Table 1, Groß Schönebeck), are par‐
et al., 2014). ticularly detrimental to the recorded signal quality, especially
The analysis of seismic data from various EGS stimulations for ‐wave arrivals (Oye et al., 2010). Strong impedance con‐
shows that a clear relationship exists between fluid injection trasts between the surface and the injection level can add
and seismic response represented by the maximum observed complexity in the wave‐form data, such as ‐to‐ wave con‐
magnitude at each site. Fig. 1 displays the observed maximum verted phases (e.g., observed in Basel by Deichmann et al.
magnitude of seismic events as a function of injected fluid at (2014) and in Paralana by Albaric et al. (2014)). Such interfac‐
EGS sites (Table 2) in comparison to data from wastewater es complicate the waveforms and thereby automatic data
disposal wells, hydraulic fracturing operations and scientific processing (phase arrival picking and identification), focal
drilling projects (McGarr, 2014). It is apparent that the maxi‐ mechanism determination (Deichmann et al., 2014) and full
mum observed magnitude increases with increasing volume of moment tensor inversion (Zhao et al., 2014). In general, the
injected fluid. The comparison between wastewater disposal shorter the travel paths the simpler the wave‐forms, which
wells (Fig. 1, triangles from McGarr, 2014) and geothermal results in better constrained source inver‐
operations (Fig. 1, squares) indicates that the latter involve
8
9
sion due to Green’s functions that reveal less complexity.

In the case of EGS, economic reasons often prevent exten‐
sive downhole monitoring systems and additional surface
stations or a combination of deep and shallow borehole sta‐
tions might be considered. An additional important factor is
the location of the planned EGS site, as the influence of cultural
noise varies significantly between densely populated areas, as
in the case of Basel, and remote places with insignificant noise
sources such as the Australian bushlands in the environs of
Paralana. Depending on the noise conditions, different depths
for borehole sensors might be advisable to achieve an opti‐
mized ratio of noise reduction and costs. E.g., Majer et al.
(2008) found a 60 dB improvement in signal‐noise ratio at
depth over ca. 152 m. In any case, we recommend inclusion of
a surface network to enhance the azimuthal coverage required
Figure 2. Sketch of potential seismic networks to monitor seismicity
for the determination of source mechanism and to ensure that
related to a hydraulic stimulation at depth. Schematic waveform
the seismic events are within the footprint of the network. For examples are shown for four different scenarios: closely placed down‐
example the seismic network in Basel consists of six deep hole sensors show the highest quality and wave onsets, whereas
borehole stations with one three‐component sensor deployed for surface seismic sensors the observed amplitudes may suffer from
in each borehole, in addition to an extensive surface seismic attenuation, high‐impedance contrasts that introduce secondary or
network. Kraft and Deichmann (2014) investigated how to converted phase arrivals and higher noise levels. (For interpretation
improve the catalogue of about 3500 events using cross‐ of the references to colour in this figure legend, the reader is referred
correlation methods and high‐quality downhole data. They to the web version of the article.)
also compare estimated focal mechanisms from the downhole
network with focal mechanisms based on data from the sur‐ depth (Tables 1 and 2), do not exhibit a relationship between
face network. Of the 3500 events recorded and located by the the number of recorded events or the maximum magnitude
borehole network, less than 200 were detected by the surface and reservoir depth (Evans et al., 2012). However, the depth of
stations. The minimum magnitude and the magnitude of com‐ the reservoir needs to be taken into account when it is linked
pleteness of the surface data were 0.9 and 1.5, in contrast to the injection pressure and to the crustal stress state. In Fig.
to 0.1 and 0.8 for the borehole data. 3, hydraulic reservoir parameters are displayed in combina‐
A critical factor for successful location of microseismic tion with the tectonic stress regime. As such, the maximum
events is the quality of the ‐wave and ‐wave velocity mod‐ inflow of the reservoir is plotted as a function of the maximum
els. Shear‐wave velocities are often extrapolated from simple wellhead pressure for geothermal sites (squares), waste dis‐
‐wave velocity models (Castagna et al., 1985), resulting in posal wells (triangles) and the KTB scientific drilling site
high uncertainties in the event locations. Albaric et al. (2014) (stars). For example, the Soultz‐sous‐Forêts EGS site (Fig. 3,
discussed the importance of obtaining a good understanding square #15) is characterized by high inflow and low wellhead
of the local structural geology before the main stimulation pressure in a normal faulting regime. On the other hand, the
phase. In their paper about hydraulic stimulation at the Para‐ Cooper Basin EGS site (Fig. 3, square #1) shows low inflow
lana EGS in Australia, the authors took advantage of several 2D and high wellhead pressure values in a thrust faulting regime.
seismic lines and additional information from borehole logs to Wastewater disposal wells are characterized by low maximum
construct 3D velocity models for ‐and ‐wave velocities, inflow values (Fig. 3, triangles). Like in Fig. 1, the Groß
which were subsequently used to locate the seismic events. Schönebeck datum (square #7) is sitting outside the boundary
The knowledge of the velocity models remains important of most of the data indicated by the solid line.
during source parameter determination, derivations of Pre‐stimulation seismic monitoring might be useful to
Green’s functions for moment tensor calculation and temporal identify buried faults and before installation of borehole seis‐
changes of the velocity models due to EGS operations (Fig. 2). mometers in observation wells. Well‐logging and well testing
Positive temporal correlation between injected fluid volume is recommended to gather information on the seismic velocity
and / ‐ratio was observed at The Geysers by Gritto and distributions and on the in situ stress conditions, which may
Jarpe (2014). Calò et al. (2011) interpreted anomalies provide better constraints to extrapolate to actual reservoir
observed during stationary injection conditions at Soultz‐ stress conditions.
sous‐Forêts as being caused by effective stress variations
linked to fluid diffusion. Fast changes of wave velocities asso‐
ciated with flow rate changes were interpreted by aseismic 3. Source mechanism of larger magnitude events (LME)
motions within the reservoir. and their occurrence in space and time in the stimula‐
In some cases, for instance, the Bouillante geothermal tion phase
field, Guadeloupe, French Antilles, little seismicity is associat‐
ed with the geothermal exploitation and therefore seismicity In the stimulation phase of EGS, high pressure fluid is
cannot be used for reservoir characterization (Sanjuan et al., pumped into the rock formation to generate fracture networks
2010). Recently, ambient seismic noise analysis has been needed for heat exchange. Hydraulic parameters measured
successfully applied to image velocity structures even in the include temperature of injected water, fluid inflow rate and
absence of seismicity (e.g., Clarke et al., 2013). This technique wellhead pressure. We introduce the scalar quantity injectivity
can be used to retrieve information about structural features (I), which is defined by the ratio of maximum inflow rate to
of the reservoir, which can subsequently be used to highlight maximum wellhead pressure to characterize the reservoir.
relative velocity changes and also to identify possible regions Note that in oil and gas exploration, the injectivity index (II) is
of aseismic slip. It is important to note that such data need to obtained from field tests upon completion of wells, and its
be acquired before the first stimulation phase, so that baseline computation requires a more sophisticated approach. In gen‐
data are available for later comparison. Modelling should be eral, injectivity describes the capacity of a well or formation to
used to optimize the network geometry for this method. accommodate pumped‐in fluid. Injectivity tests are conducted
Today’s EGS reservoirs, developed between 2 and 5 km to establish the rate and pressure at which fluids can be
10
Figure 3. Hydraulic parameters in the reservoir displayed together Figure 4. Observed maximum magnitude of induced seismic events as
with the tectonic stress regime. Maximum inflow is plotted as a func‐ a function of injectivity of the reservoir (ratio of maximum injection
tion of maximum wellhead pressure for fluid‐injection field experi‐ rate and maximum wellhead pressure) computed from Table 2 data.
ments from Fig. 1. Symbols are colour coded according to stress re‐ Line indicates upper boundary trend with one exception, the Rocky
gime prevailing at target depth (red = normal‐, purple = strike‐slip‐, Mountain Arsenal wastewater disposal (#27). Lower limit datum
blue = thrust faulting). Insets indicate elongation of the seismic cloud represents the Groß Schönebeck geothermal site (#7).
in the direction of maximum principal stress. (For interpretation of
the references to colour in this figure legend, the reader is referred to
the web version of the article.) pathway for fluids, preventing efficient heat exchange. A better
physical understanding of the reservoir‐ and seismicity gener‐
ation process is needed in order to develop techniques to
pumped into the treatment target without fracking the for‐ reduce the probability of occurrence of LME.
mation. Computing the injectivity index II is the most common It was first proposed by McGarr (1976) that the total mo‐
way of analysing performance of injection wells. The computa‐ ment of an induced seismic cloud is proportional to the vol‐
tion includes injection rate, injection pressure corrected for ume of injected fluid, see Fig. 1. However, the actual amount of
bottom hole flow conditions, and far‐field reservoir pressure seismicity (or the proportionality factor) can vary substantial‐
(e.g., Economides and Saputelli, 2005). In the following, the ly between different reservoirs for similar injected fluid vol‐
simple injectivity definition I is used instead of the injectivity umes. Using a completely different approach of pressure diffu‐
index II because of lack of data. In Fig. 4, reservoir injectivity I sion theory, Shapiro et al. (2007, 2010) obtained a similar
is plotted as a function of maximum observed seismic magni‐ result in which the total number of induced events is propor‐
tude. Apart from the data collected at the Rocky Mountain tional to the injected fluid volume. In addition, they describe
Arsenal wastewater disposal site (Fig. 4, #27), disposal wells the proportionality factor, the seismogenic index, as a function
are characterized by lower injectivities ( 1.5) compared to of measurable seismological quantities and rock properties.
most EGS sites. This seems counter intuitive to what one Combining these considerations with the assumption that
would desire in a wastewater disposal well but may be ex‐ seismicity always follows a Gutenberg–Richter‐type magni‐
plained by the fact that for EGS sites the injectivity computa‐ tude distribution leads to a probabilistic estimate of the max‐
tion is conceptually problematic. This is because one has to imum expectable magnitude.
correct for dynamic (fluid‐driven) fractures, which is required The conclusion is that the probability of a LME generally
in EGS stimulation. In our simple approach, EGS sites (Fig. 4, increases with injected volume, even though there are signi‐
squares) indicate a wide range of injectivity values from 0.32 ficant regional differences. For this reason, we shall differenti‐
(Fig. 4, #4, Paralana) to 6.25 (Fig. 4, #5, Rosmanowes) with ate between long‐term injection operations (e.g., wastewater
LME ranging from −1 to 3.7 (Fig. 4, #7 and #1). For compari‐ disposal (triangles in Figs. 1, 3 and 4) or CO2 disposal) where
son, the observed maximum seismic magnitudes in large (net) volumes may accumulate over time, and the short‐
wastewater disposal range from 3.3 (Fig. 4, #20) to 5.7 (Fig. 4, term stimulation operations at the beginning of an EGS project
#30). The solid line is an upper boundary to most of the data, (squares in Figs. 1, 3 and 4). Most of the LME reported in the
except for the wastewater disposal at Rocky Mountain Arsenal literature have occurred either after long‐term fluid injection
(Fig. 4, #27). The Groß Schönebeck datum (Fig. 4, #7) indi‐ (Ake et al., 2005; Frohlich et al., 2011, McGarr, 2014), or as a
cates the lower limit. result of reservoir impoundment (Gupta, 2002, 2011), both of
One major topic addressed by a number of papers is the which can bring pre‐existing fractures in the shallow crust
characterization of LME, coupled with the investigation of closer to failure. Of the short‐term stimulation activities, EGS
strategies to assess and, ideally, mitigate seismic hazard asso‐ stimulations have generally shown a much higher propensity
ciated with stimulation operations. As seismicity is induced in to produce LME, compared, e.g., to hydraulic fracturing in the
the process of enhancing or creating the permeability, it is oil‐ and gas industry. Shapiro et al. (2010) find significantly
important not to induce or to trigger LME, which may not only higher seismogenic indices for geothermal stimulations in
cause damage at the surface, but also might lower the efficien‐ crystalline rocks than for comparable operations in sedimen‐
cy of the geothermal system through the creation of high‐ tary formations. Similarly, Evans et al. (2012) suggested that
permeability pathways. If a LME occurs, it may create a master injection into sedimentary rocks tends to be less seismogenic
11
than into crystalline rocks. Also from laboratory testing it is

known that acoustic emission activity and source mechanisms
are very different for e.g., sandstones (Lockner and Byerlee,
1972; Zang et al., 1996) as compared to granite (Lockner and
Byerlee, 1992; Zang et al., 2000). This is caused by different
fracture mechanisms operating during fluid injection into
different rock types as discussed in Section 5.
Large or damaging earthquakes tend to occur on devel‐
oped or active fault systems. In other words, large earth‐
quakes are unlikely to occur in absence of a fault large or long
enough to release considerable energy. Evans et al. (2012)
compare the maximum magnitude observed at each project
with the regional seismic hazard expressed in terms of the
long‐term probability of exceeding threshold peak ground
acceleration (PGA), and they speculate that fluid injection in
areas with lower natural seismicity may have a lower risk of
triggering LME. However, a causal relationship if any between
the two properties (maximum magnitude and PGA threshold Figure 5. Nearly North–South trending depth‐cross‐section through
level) remains unclear and requires a more detailed geo‐ the seismic cloud induced during the stimulation of the Paralana EGS
mechanical examination of each project (see Section 4). (modified from Albaric et al., 2014). The symbols are color‐coded
Table 1 lists GEISER related sites with the largest LME in according to the time of occurrence of each event and their size is
decreasing order of event magnitude. Some of these cases, proportional to the magnitude ( . . ). The numbered
stars correspond to the four LME with in the range 2.4–2.5. (For
mainly Basel (Switzerland), Soultz‐sous‐Forêts (France), Para‐ interpretation of the references to colour in this figure legend, the
lana (Australia) and Berlin (El Salvador), were studied in reader is referred to the web version of the article.)
detail as part of the GEISER programme. In addition to magni‐
tudes (moment magnitudes wherever available), the time and
location of recorded LME at these sites are listed in the text stimulated in June 2000, GPK3 in May 2003 and GPK4 in Sep‐
below Table 1. As stated before, it has been frequently ob‐ tember 2004 and February 2005. Events with magnitudes as
served that LME occur after shut‐in and at larger distances high as 2.5 were recorded during the stimulation of both GPK2
from the injection well (Baisch et al., 2010). and GPK3, and in each case, the strongest event occurred after
shut in – a magnitude 2.6 event some 10 days after shut‐in of
3.1 Berlin, El Salvador GPK2 and two magnitude 2.9 and 2.7 events three‐and‐a‐half
and five days after shut‐in of GPK3 (Dorbath et al., 2009). The
Kwiatek et al. (2014) present a case study on the geother‐ stimulation of GPK4 occurred in three phases – the first two as
mal project in Berlin, El Salvador, where a 3.6 LME conventional freshwater injections and the third with the
occurred two weeks after shut‐in of the injection in the central addition of acid. A total of four events with magnitudes greater
part of the reservoir ca. 2 km away from the injection well than 2 were induced by these activities: during the first phase,
using relocated data. They found that the seismic activity a magnitude 2.3 event occurred a few hours after injection
induced by multiple injections into well TR8A did not reach was stopped; during the second phase, a magnitude 2.7 event
distances longer than 500 m from the injection point. They do occurred at the height of the injection; during the third phase,
not observe any sequence of events migrating further away a magnitude 2.2 event occurred during stimulation and a
towards the nucleation of the LME. However, the multiple magnitude 2.3 event occurred about eight days later, when
injection operations at TR8A can be responsible for changing both the well‐head pressure and the seismic activity had de‐
dominant fluid migration pathways, and indirectly may be clined to almost normal levels again. Dorbath et al. (2009)
responsible for the occurrence of LME in the Berlin geo‐ showed that the relative size distribution of earthquakes, as
thermal field. They also found a dependence of stress drop on expressed by the ‐value of the Gutenberg–Richter relation,
distance from the injection point, somewhat similar to what was significantly different for the events induced during the
was found by Goertz‐Allmann et al. (2011) for Basel, but not stimulation of GPK2 and GPK3: whereas the average ‐value
nearly as well defined. was around 1.23 at GPK2, it was only 0.94 at GPK3. The differ‐
ence seems to be linked to the observation that the seismic
3.2 Paralana, Australia cloud induced at GPK2 is diffuse and structureless, whereas at
GPK3 the hypocentres align along major faults that can be
In Paralana, Australia, during a stimulation in July 2011,
traced back to the wells, where they were identified in the
four LME with 2.4‐2.5 occurred during periods of mas‐
borehole logs. Furthermore, Dorbath et al. (2009) reported
sive injection. As shown by Albaric et al. (2014), the hypocen‐
that the two 2003 events with magnitude 2.9 and 2.7 were
tres of the first three of these events were located at the pe‐
located on two of these major faults.
riphery of the extent of the seismic cloud at that time, whereas
the fourth occurred closer to the well, in a zone which had
3.4 Basel, Switzerland
experienced activity earlier during the stimulation process
(Fig. 5). One of the most intensively studied EGS stimulations that
triggered several LME causing damage at the surface is the
3.3 Soultz‐sous‐Forêts, France project in Basel, Switzerland. In Basel the largest magnitude
event ( 3.4) occurred a few hours after shut‐in but short‐
The EGS at Soultz‐sous‐Forêts, France, has a long history of
ly before the well was opened, while three additional events
stimulation and circulation through several wells at different
with 3 occurred one to two months later, long after the
depths. A short summary of the activities can be found in
well‐head pressure had dropped to near‐normal levels. The
Evans et al. (2012). A stimulation in 1993 at depths of 3–3.5
hypocentres of these earthquakes were located at the periph‐
km produced only minor seismic activity, with a maximum
ery of the seismic cloud. Several studies have observed that a
magnitude of 1.9. Subsequently, three additional wells, GPK2,
large percentage of the seismicity induced by the Basel EGS
GPK3 and GPK4, were drilled to depths of 4.5–5 km. GPK2 was
12
tion to earthquake phenomena, Scholz (1968) highlighted that

the ‐value of magnitude frequency distributions (grain‐scale
cracks or earthquakes) strongly depends on the state of stress.
Recent laboratory measurements and also regional ‐value
estimates have shown that ‐values are inversely proportional
to the differential stress and thus may qualitatively be used as
stress indicators at depth, where generally no direct meas‐
urements are available (Amitrano, 2003; Schorlemmer et al.,
2005). Assuming that earthquakes occurring in regions of high
differential stress or on faults with large shear stresses, on
average tend to have higher stress drops, one should expect a
systematic spatial variability of earthquake stress drops simi‐
lar to the variability observed for the ‐values. In fact, the
stress drops determined by Goertz‐Allmann et al. (2011) for
the induced seismicity of Basel vary over a wide range and
tend to increase systematically with distance from the injec‐
tion point and, thus, with time.
Zhao et al. (2014) analysed the 19 largest events that oc‐
curred during and after the injection sequence and deter‐
mined moment tensors from full waveforms. That analysis
resulted in similar double‐couple components as found by
Figure 6. The seismic cloud induced by the stimulation of the Basel
Deichmann and Ernst (2009). In addition, it revealed signifi‐
EGS in 2006 and 2007. The hypocentres are colour coded according to
the ‐values calculated for the rock volume in which they occurred cant isotropic components during the early injection phase,
(modified from Bachmann et al., 2012).While values range from whereas most of the events in the later stage are dominated by
0.8 to 3.5, the colour bar is limited from 1 to 2 for a clearer visibility. double‐couple components. Their result implies sizeable vol‐
The stars correspond to the four LME with (1: 2006/12/08, ume changes caused by large pore pressures at the early times
. ; 2: 2007/01/06, . ; 3: 2007/01/16, . ; 4: close to the injection, consistent with the pore pressure esti‐
2007/02/02, . ).(For interpretation of the references to color mates of Terakawa et al. (2012). Isotropic moment tensor
in this figure legend, the reader is referred to the web version of the components may relate to the occurrence of long‐period
article.)
events (Bame and Fehler, 1986; Ferrazzini et al., 1990; Jousset
et al., 2010). The locations of the events with high isotropic
occurred in clusters of similar events, which implies that their components also coincide with previously found regions of
hypocentres are located close to each other and that their focal high ‐values (Bachmann et al., 2012) and low stress drops
mechanisms are nearly identical (e.g., Deichmann et al., 2014 (Goertz‐Allmann et al., 2011).
and references therein). In other words, it seems that earth‐ All of these observations provide a consistent picture in
quakes within these clusters of similar earthquakes occur on the near‐ and far‐field of the stimulation well. Early during the
the same faults. Results of high‐precision relative locations of stimulation phase and close to the injection well, pore pres‐
the events within such clusters show that their hypocentres sures are high (near‐field). Here many small events, often on
define near‐planar structures that coincide with one of the faults with low applied shear stress, are induced, consistent
nodal planes of their focal mechanisms. Detailed analyses of with the high ‐values, low stress drops and significant volu‐
the sequences associated with the larger magnitude events metric components of the focal mechanisms. As the injected
( 2) induced during the stimulation of the Basel EGS water migrates away from the injection point and the pore
showed that the activated faults have dimensions of the order pressure decreases (far‐field), the volumetric component of
of several 100 m and are often oriented obliquely to the over‐ the focal mechanism becomes insignificant and the shear
all trend of the microseismic cloud (Deichmann et al., 2014). stress necessary to trigger an event must increase. This is
These results reveal a complex internal structure of the flow reflected in higher stress drops and in lower ‐values. The
paths in the rock volume stimulated by the water injection and latter implies a higher probability for the occurrence of the
imply that geomechanical models consisting of a single often‐observed larger magnitude events at the periphery of
through‐going structure are not realistic. the stimulated volume and during the later stages of the
One of the key discoveries of the research performed on stimulation.
the seismicity induced by the Basel EGS concerns the relative
size distribution of the induced earthquakes. The results of a
high‐resolution ‐value mapping by Bachmann et al. (2012) 4. Maximum observed and expected seismic magnitude
are summarized in Fig. 6, which shows the highly‐systematic
spatial heterogeneity of the ‐values associated with the seis‐ In this section, it is important to distinguish between the
micity induced by the Basel EGS. Unusually high ‐values maximum observed seismic magnitude ( ) at a given
shown in red are found near the injection point and earlier in geothermal site (see Tables 1 and 2), and the maximum ex‐
the sequence; further out, ‐values tend to be closer to the pected (or possible) magnitude ( ) as typically used in
normal tectonic average of around 1.0. A small ‐value is probabilistic seismic hazard assessment, e.g. McGuire (2004)
equivalent to a higher probability for the occurrence of larger‐ and Kijko (2004). While represents the maximum mag‐
magnitude events and is consistent with the observation that nitude of an induced seismic event during the development of
the largest induced seismic events often – but not always – a geothermal reservoir (exploration, stimulation, circulation),
occur after shut‐in and at the periphery of the stimulated rock stands for the very rare seismic event happening only
volume. Indeed, as shown in Fig. 6, the hypocentres of the four every thousand or ten thousand years, and is difficult to be
largest events are located in regions of low ‐values. estimated based on observed data (Holschneider et al., 2014).
The size distribution of fractures in rock is a power func‐ This is because the observation periods in geothermal reser‐
tion of fracture size and the exponent of this relation scales voirs (Table 1, decades) are much too short to sample .
inversely with stress. Based on Mogi (1962) laboratory acous‐ Therefore, using instead of to assess the seismic
tic emission studies on heterogeneous materials and its rela‐
13
hazard of a given site will result in an underestimation (lower Increasing the area of fluid pressure will increase during
limit value only). Many efforts have been made to anticipate stimulation with increasing injection time. Consequently (as‐
the size and rate of occurrence of earthquakes. In the structur‐ suming constant injection pressure), the observed seismicity
al geological approach, the maximum magnitude is inferred migrates away from the injection well.
from the largest potentially active fault in the geothermal A dynamic fracture‐growth approach is the discrete frac‐
reservoir (e.g., Majer et al., 2007). Assigning a maximum ture network model by Yoon et al. (2014) (see Fig. 8). In this
earthquake magnitude to a given fault is based on empirical hydromechanical coupled model of fluid injection in a natural‐
relations between magnitude and fault parameters such as ly‐fractured reservoir, fractures are present before stimula‐
length, width and displacement, as discussed by Wyss (1979) tion (Fig. 8a, lineaments), and new fractures can be formed
and Wells and Coppersmith (1994). indicated by induced seismic events during (Fig. 8b, grey dots)
In Fig. 7, an earthquake fault relationship, after Leonard and after stimulation treatment (Fig. 8c, red dots). The fractur‐
(2010), is shown modified for maximum observed magnitudes ing process operates in mixed‐mode. This means that pre‐
at geothermal sites (green stars) and tectonic earthquakes existing fractures can behave as tensile cracks (mode I) or
(red stars). Formulas used to prepare this figure and refer‐ hydroshears (mode II), but also new fractures can form from
ences are listed in Table 3. Earthquake magnitude, fault size existing ones, e.g. as wing cracks (mode I) from existing shear
and rupture energy increase from left to right (Fig. 7). Differ‐ crack tips (mode II). The goal is to identify the size and overall
ent rows indicate peak ground velocity (PGV), peak ground orientation distribution of existing and newly created frac‐
acceleration (PGA) and their relation to macroseismic intensi‐ tures during EGS stimulation and operation, and their relation
ty and moment magnitude, and their scaling with fault‐related to the three principal stresses in order to optimize reservoir
parameters like slip ( ), rupture length ( ) and rupture ener‐ productivity and to reduce induced seismicity at the same
gy. For a geothermal reservoir, the problem is to downscale time. Dynamic fracture distributions allow computation of
these empirical relations to the much smaller size of fractures seismic moment tensors and radiated seismic energy for com‐
expected to be created or reactivated, about 50–500 m in parison with hydraulic energy injected into the EGS system
length (Section 3 and Fig. 7, green stars). This downscaling can (Zang et al., 2013). Seismic catalogues for various stimulation
be performed by a number of ways and should take into ac‐ scenarios in dynamic fracture growth models can be used for
count non‐linearity that may exist in the scaling laws (Douglas hazard assessment of specific sites and stimulation treatments
and Jousset, 2011). Although typical reservoir faults may not (Hakimhashemi et al., 2013, 2014).
be detectable by 3D seismic reflection they might be able to In approach (2), the probability of occurrence of an earth‐
generate seismicity with local magnitudes between 3 and 4, quake with a chosen is derived from the magnitude‐
depending on the stress drop (Evans et al., 2012). frequency distribution of earthquakes, the Gutenberg–Richter
Apart from this structural geological approach, there are law. This requires an earthquake catalogue that spans a wide
other methods to estimate the maximum magnitude: (1) The range of magnitudes. In addition, a truncation of the Guten‐
deterministic approach, (2) the probabilistic approach, and (3) berg–Richter relation and thus the definition of the locally
the empirical approach. In the deterministic approach, the valid value generally need additional assumptions about
generation and propagation of fluid‐filled fractures when a the tectonics and rock mechanical processes that cause the
geothermal reservoir is simulated must take into account earthquakes. To this end, Shapiro et al. (2010) modified the
geometry, rock properties and in situ stresses. In this ap‐ Gutenberg–Richter law by introducing a tectonic potential,
proach, fracture distributions can be prescribed (Bruel, 2007; and computed a characteristic scalar quantity of reservoir
Baisch et al., 2010; Wassing et al., 2014), or dynamic crack fluid stimulation, the seismogenic index (cf. Section 3). The
growth models can be used (Hazzard and Young, 2004). For seismogenic index may vary between different wells of the
example, in Baisch et al. (2010), crack distributions are pre‐ same field or even within one well for different stimulation
scribed and slip occurs on critically stressed slider‐spring scenarios. In order to guide reservoir operations, we recom‐
patches. This model consists of a subvertical fault zone (3 km mend computing indices to quantify seismic reservoir behav‐
× 3 km in size) with constant pore pressure (48 MPa), which is iour as a function of time, in particular in the stimulation and
subdivided into smaller slip patches (20 m × 20 m in size) and production phase of a reservoir.
intersected by a well at a depth of 4.8 km (Soultz‐sous‐Forêts Hakimhashemi et al. (2013) used a hybrid approach for
scenario). By analysing hydraulic overpressure distributions computing induced seismicity. They started from a geo‐
inside the fault zone as a function of radial distance to the mechanical model (approach 1) and applied probabilistic
injection well, Baisch et al. (2010) identified the slip patches of techniques to compute seismic hazard (approach 2). To our
LME with zones of increasing overpressurization after shut‐in. knowledge this was the first attempt to link approaches (1)
In this model, therefore, LME were explained by non‐ and (2). Time‐dependent Gutenberg–Richter ‐value and ‐
stationary fluid pressure conditions during EGS injections. value curves computed from synthetic seismic catalogues of
After pumping stops, the pressure diffusion continues and the stimulated reservoirs allow estimation of maximum hourly
outer boundary of the reservoir, depicted by the seismic cloud, occurrence rates of induced seismic events for a given reser‐
becomes subject to higher in situ pressures than before. This voir and fixed stimulation strategy. As opposed to the tradi‐
may cause a larger spatial area concurrently to become over‐ tional probabilistic seismic hazard assessment approach
critically stressed, which tends to result in larger magnitude (PSHA), this method is called forward induced seismic hazard
events. Since moment magnitude is proportional to the loga‐ assessment (FISHA).
rithm of the source area times the average slip, can be Gischig and Wiemer (2013) used a 2D flow model with
estimated from the maximum size of the induced seismic non‐linear pore pressure diffusion in combination with a
cloud as a function of time from the start of the injection (See stochastic seed model. The transient pressure field was used
approach (3) below). This is based, however, on the assump‐ to trigger seismicity at randomly distributed seed points. A
tion that the rupture surface of the largest event is limited by differential stress normal distribution was assigned at each
the size of the stimulated rock volume at a particular time, and seed point, which is a potential seismic hypocentre. As in
estimated in this way is highly sensitive to the adopted previous models, Mohr–Coulomb failure was assumed as well
source model and assumed static stress drop. According to the as an inverse relationship between stress drop and ‐value
model by Baisch et al. (2010), the maximum magnitude is (Scholz, 1968). Random seismic magnitudes were assigned
controlled by the area over which fluid pressure increase from ‐values corresponding to stress drop values at seed
brings the state of stress of the rock mass close to failure. points. These are the main assumptions of Gischig and Wiemer
14
Figure 7. Earthquake fault scaling relationship. Rows indicate ground motion parameters, macro‐seismic intensity (EMS) and earthquake mo‐
ment magnitude in relation to fault properties ( , ) and energy of the rupture process. Selected EGS seismicity (green stars) and tectonic earth‐
quakes (red stars) are shown in depth section. The boundary between intensity IV (dark green) and intensity V (light red) is indicated by a
change in colour. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.)
(2013), who generated synthetic seismic catalogues to repro‐ alone. The hybrid approach by Hakimhashemi et al. (2013)
duce the Basel induced seismicity data. combining deterministic, coupled hydromechanical reservoir
At this stage, we rely on the synthetic induced seismicity models with forward induced seismic hazard assessment
catalogues from Yoon et al. (2013, 2014) because their hydro‐ (FISHA) seems promising.
mechanically coupled model approach using dynamic fracture McGarr (2014) argued that maximum seismic moments, or
growth and interaction principles (Fig. 8) captures basic phys‐ moment magnitudes, for a given fluid injection activity can be
ical principles of the natural fracturing process operating in estimated based on five plausible assumptions. (1) The for‐
situ during the development of a geothermal field, and in addi‐ mation is either seismogenic or there is hydraulic communica‐
tion reproduce main features of observed seismic catalogues tion between the injection interval and seismogenic regions of
monitored in EGS stimulation (post shut‐in larger magnitude the crust. The seismogenic regions, which are often Precambri‐
events, development of induced seismic cloud in the direction an crystalline basement formations, contain numerous pre‐
of maximum principal stress, magnitude‐frequency distribu‐ existing faults, some of which are well oriented for failure in the
tions of events for different injection scenarios). ambient state of stress (e.g., Townend and Zoback, 2000). (2)
In the empirical approach (3), the size of the activated geo‐ Before injection, faults in the vicinity of the injection wells that
thermal reservoir is determined by the extent of the hypocen‐ are well oriented for slip in the ambient stress field are stressed
tre distribution of the induced events in hot‐dry‐rock projects to within a seismic stress drop of failure. (3) The rock mass is
– the so‐called seismic cloud (Fehler et al., 2001). In a critically fully saturated before injection begins. (4) The seismic response
stressed crust, however, the maximum possible event released to injection is a Gutenberg and Richter (1954) earthquake distri‐
during fluid injection into the reservoir cannot be captured bution log , where is moment‐magnitude. (5) The
with either a deterministic or empirical approach using the induced earthquakes are localized to the region where the crust
moment release – volumetric expansion hypothesis by McGarr has been weakened due to fluid injection (Hubbert and Rubey,
(1976). Instead one needs to take into account crack interac‐ 1959). This last assumption seems plausible, but it is, nonethe‐
tion processes either by prescribed shear avalanche (Baisch less, difficult to prove that it always applies.
et al., 2010), or mixed‐mode crack growth models (Yoon et al., If these assumptions all apply, then it is straightforward to
2013), in order to account for larger, dynamic moment release show that the maximum seismic moment and the maximum
than predicted by the fluid volume injected (seismic cloud) moment magnitude depend almost entirely on the total vol‐
alone. In addition, in such critically stressed environments, the ume of fluid injected up to the time of the occurrence of the
influence of remote triggering is much higher (Jousset and largest event (McGarr, 2014).
Rohmer, 2012). If no seismicity is observed, approaches (2) For most projects the maximum magnitude earthquake is
and (3) are inapplicable, and ambient seismic vibration stud‐ observed to be much lower than estimated on the basis of
ies need to be taken into account. injected fluid volume (see Fig. 1). For instance, the rock mass
In summary, induced seismic hazard assessment should may respond to injection by deforming aseismically because of
not rely on either deterministic or probabilistic methods its material properties.
Table 3.
List of parameters, formulas and references used to compute earthquake fault scaling relationship shown in Fig. 7.
Parameter Unit Formula Reference
Peak ground velocity, PGV cm/s 5.11 2.35 log Faenza et al. (2010)
Peak ground acceleration, PGA cm/s2 1.68 2.58 log
EMS Intensity, 0.667 0.30 log 0.10 Grünthal et al. (2009)
Displacement, cm log 0.833 log 1.07 Leonard (2010)
Fault length, km 1.67 log 4.32
EMS, European Macroseismic Scale; moment magnitude; epicentral intensity; focal depth of earthquake.
15
5. Discussion
Baisch and Vörös (2010) compared seismic signatures of

three crystalline geothermal reservoirs, Basel, Cooper Basin
and Soultz‐sous‐Forêts. The deep seismicity at Soultz, whether
viewed as a planar structure (Charlety et al., 2007) or volu‐
metric cloud (Michelet and Töksöz, 2007) is influenced by the
existence of a larger tectonic fault in the Rhine Graben. This
fault perturbs the in situ stress field with maximum horizontal
stress striking N170°E (Table 1, Cornet et al., 2007). Induced
seismicity in the Cooper Basin reservoir is also associated with
a larger fault and driven by the stress field (Baisch et al.,
2009). The stress regime in the Cooper Basin, however, is
compressional, with maximum horizontal stress oriented E–W
(Table 1). In this respect, induced seismicity at both sites is
quite similar when accounting for the rotation of stresses
(Baisch and Vörös, 2010). Since fractures propagate in the
direction of maximum compressive stress, the seismic cloud at
Soultz is oriented sub‐vertical while the cloud in the Cooper
Basin is oriented sub‐horizontal. This relationship is shown
schematically in Fig. 3 (inset, ellipses with arrows) for normal
faulting (NF) and thrust faulting (TF) regime. Even if this is
based on a simple fracture mechanics principle, Fig. 3 can
serve as a first approach in estimating key parameters for the
understanding of induced seismicity such as inflow, reservoir
pressure and in situ stress. We rate in situ stress the second
most important parameter in EGS development after the pres‐
ence of heat. One should follow International Society for Rock
Mechanics standards for determining in situ stress at a given
(geothermal) site, Stephansson and Zang (2012).
Consistent with observations at Cooper Basin and Soultz,
post‐injection seismicity at Basel occurs at the outer boundary
of the zone of previous seismic activity (Häring et al., 2008).
The field Kaiser Effect (Baisch and Harjes, 2003; absence of
seismicity until the stress level of previous stimulations is
exceeded) is a characteristic feature of geothermal sites locat‐
ed in hot granitic rock masses. In GEISER, this effect has also
been documented in meta‐greywacke at Berlin, El Salvador
(Kwiatek et al., 2014). In sedimentary basins, the Kaiser Effect
has not been reported. The effect, however, has been dis‐
cussed in the context of the collapse of a salt cavity (Lebert
et al., 2011).
In Fig. 9, two end‐member cases of geothermal reservoir
stimulation are shown, one for fluid injection into sedimentary
rock (Fig. 9a and c), and one for injection into crystalline rock
(Fig. 9b and d). The undisturbed rock under primary, in situ
state of stress is shown in Fig. 9a and b. The future location of
injection wells is shown by dashed circles. The perturbed
stress state with borehole and fluid‐stimulated rock mass
(blue colour) is shown in Fig. 9c and d. In sedimentary rocks,
isolated porosity (Fig. 9a and c, dots) can guide hydraulic
fractures in the direction of maximum stress.
In the stimulation phase, mechanisms like pore collapse
and grain crushing can lead to a more localized damage zone
around the fluid‐driven master tensile fracture (Fig. 9c). As a
consequence, apart from mini‐shear halos, the hydraulic frac‐
ture in sedimentary rock is dominated by implosion (pore
collapse) and explosion sources (tensile cracks). In contrast,
Fig. 9b and d illustrate the corresponding damage zone exten‐
Figure 8. Discrete fracture network model with pore fluid flow algo‐ sion in hydraulic fracturing of crystalline rock. In Fig. 9b, the
rithm (modified from Yoon et al., 2014). (a) Naturally fractured reser‐ naturally fractured (saturated) crystalline host rock is shown
voir with fluid injection point in the centre. (b) Stimulated reservoir before high pressure fluid injection. In Fig. 9d, the natural
with pre‐shut‐in induced seismic events (grey dots). (c) Post‐shut‐in fracture network around the injection well allows for fast fluid
events (red dots). In (b) and (c), the 0.1 MPa pore pressure contour is
transport into neighbouring fractures. Hydro‐shear events
shown by blue solid line, and size of dots scales with seismic moment
magnitude from mixed‐mode dynamic fracture growth. The maximum along pre‐existing, natural fractures in combination with ten‐
horizontal stress is oriented in the North–South direction. (For inter‐ sile hydraulic opening at the very end of fractures (tips) may
pretation of the references to color in this figure legend, the reader is be the dominating mechanism in driving the overall hydraulic
referred to the web version of the article.) fracture in the direction of maximum stress. Branching of
fractures is common if the energy required to generate new
16
and mechanisms operating in situ, as evidenced by the differ‐

ence in the extension of the acoustic (seismic) cloud and the
actual fracture pattern observed in granite cores in the labora‐
tory (Zang et al., 2000). Although basic aspects of generating
fluid‐induced seismicity are understood, there are, nonethe‐
less, gaps in our understanding of the key reservoir parame‐
ters that control the rate and magnitudes of the seismic
events. All parameters may not be monitored, and triggering
mechanisms may be of many different origins, including re‐
mote triggering (Jousset and Rohmer, 2012). Consequently,
physics‐based models are needed in a forward approach to
analyse occurrence rate and maximum magnitude of induced
seismic events in the development of a geothermal field.
Could seismogenic permeability (Talwani et al., 2007) be
the key to understanding the link between aseismic/relaxed
and seismic/locked patches within the reservoir? Microseis‐
micity results from a variety of causes. The spatiotemporal
Figure 9. Synoptic picture of complex fracture modes in geothermal evolution of the triggering front depends on different parame‐
reservoirs. Sedimentary or weak volcanic rock (left) and crystalline
ters, including stress, injection rate, diffusivity and rock prop‐
rock (right) are displayed in an undisturbed stress state (upper panel,
dashed circles indicate location of future borehole), and after fluid erties (e.g., friction coefficient). For example, an elliptical
stimulation in a perturbed stress state (lower panel, with borehole). shape of the induced seismic cloud can be caused by the ani‐
Horizontal section of reservoir at target depth is shown in an aniso‐ sotropy of hydraulic diffusivity (Shapiro et al., 1997, 1999), or
tropic stress field with fluid pathways (blue). (For interpretation of by an anisotropic stress tensor (Schoenball et al., 2010). Due to
the references to colour in this figure legend, the reader is referred to poroelastic effects, it is impossible to distinguish between the
the web version of the article.) two types of anisotropy from geothermal seismicity alone. In
addition, Yoon et al. (2014) using hydromechanical coupled
tensile fractures is lower or comparable to the energy re‐ discrete fracture network models demonstrated that also the
quired to reactivate existing, natural fractures. Depending on injection scenario affects the shape of the seismic cloud. A
the initial fracture distribution around the well, the damage clear understanding of the physical processes operating at the
zone around the hydraulic fracture in crystalline rock may be border between seismically inactive and seismically active
much larger compared to the one in sedimentary rock. In the zones (asperity, hydraulic barrier) needs to be developed. On a
near field, however, injection into both rock types may suffer longer than stimulation perspective (production), processes
from the effect of an increasing number of tensile cracks due like heat transport coupling and geochemical effects may
to elevated pore pressure discussed in Section 3. dominate induced seismicity in hydrothermal systems.
Therefore, EGS development in weak volcanic rock (Fig. 9c
and Table 1, e.g., Berlin El Salvador, Groß Schönebeck) may be
quite different from permeability enhancement in crystalline 6. Conclusions
rock (Fig. 9d and Table 1, e.g., Basel, Cooper Basin, Soultz‐
sous‐Forêts). While in hard crystalline rock, dilatant shear Based on the analysis of seismicity associated with the de‐
may be the dominating mode II failure mechanism of the geo‐ velopment of geothermal sites compared to wastewater dis‐
thermal reservoir, in weak volcanic or sedimentary rock, sub‐ posal and KTB ultra‐deep fluid‐injection experiments into the
critically stressed mode I fractures may chiefly be generated Earth’s crust, a synoptic picture of the damage zone evolution
instead. This does not imply the absence of shear cracking in a in fluid‐driven fractures during stimulation of enhanced geo‐
sedimentary environment, but apart from accompanying thermal systems (EGS) is developed and used to understand
propagating master‐tensile fractures, dilatant shear may be of the relation between hydromechanical reservoir parameters
minor importance. This in turn can lead to a more persistent (injectivity, fluid volume), the state of stress at depth, and the
fracture growth path in sedimentary rock (Fig. 9c) where occurrence of larger magnitude seismic events (LME) in space
pores are interconnected by tensile crack growth. In contrast, and time during the stimulation phase of EGS sites. We draw
in granitic reservoirs high‐pressure fluid injection leads to a ten conclusions.
broader damage zone because the fluid can travel easily into
pre‐existing and newly‐created fracture networks. Close to the (1) The maximum observed seismic magnitude increases
stimulation well (near‐field), tensile crack formation can dom‐ with the volume of the fluid injected into the Earth’s
inate while in the far‐field hydro‐shears may predominate. crust. Compared to waste water disposal wells, geother‐
The overall direction of fracture path in both rock types, how‐ mal operations (in particular EGS stimulations) involve
ever, is parallel to the maximum compressive stress (Fig. 9c less fluid volume resulting in smaller maximum magni‐
and d). The width of the fracture damage zone is expected to tudes. In crystalline reservoirs with multiple stimulation
be smaller for sedimentary than crystalline rock. If so, the wells, seismicity is absent until the stress level of previ‐
seismic cloud in a sedimentary environment would be nar‐ ous stimulations is exceeded (Kaiser Effect). Previous
rower because more localized damage in sedimentary rock stimulations change the near‐well stress field for subse‐
produces less seismicity. quent stimulations. Stress mapping while injecting fluid
Based on this synoptic picture of complex fracture modes is strongly recommended.
in geothermal reservoir stimulation, we recommend using (2) A seismic network consisting of both, borehole and sur‐
high resolution source analyses of geothermal seismicity to face stations, allows increasing the network sensitivity
determine what fracture mechanisms are operating in situ. towards lower event magnitudes and improves the mag‐
Because the distribution of the seismic cloud in the reservoir nitude of completeness. E.g. at the Basel EGS, the mini‐
strongly depends on the type and geometry of the seismic mum magnitude and the magnitude of completeness of
network, velocity model and location method used, its size surface data were 0.9 and 1.5, in contrast to 0.1
cannot be related one‐to‐one to the actual fracture location and 0.8 for the borehole data. Including borehole data in
the processing and interpretation of seismic lines im‐
17
proves the depth resolution of the models. High quality crystalline rocks (e.g., Basel, Cooper Basin, Soutz‐sous‐
of the velocity model is vital, because it determines the Forets). While in hard rock dilatant shear may be the
accuracy of hypocentre locations, source parameters and dominating failure mechanism (mode II) in EGS stimula‐
their temporal changes in EGS operations. tion, in sedimentary formations sub‐critically stressed
(3) Background seismicity should be monitored prior to any tensile cracks (mode I) may be chiefly generated instead.
stimulation activity to obtain a baseline to evaluate The width of the fluid‐driven damage zone in naturally
changes in the seismicity rate during stimulation. In the fractured crystalline rock is expected to be wider than
absence of induced seismicity, ambient seismic vibra‐ that for sedimentary formations. If so, the seismic cloud
tions can help in determining structural features and induced by EGS stimulation should be narrower in weak
temporal changes in reservoir properties. Such data compared to hard rocks (under the assumption of the
need to be acquired before the first stimulation, such same seismic resolution).
that baseline data are available for comparison at a later (10) Concerning seismic hazard potential at EGS sites, several
stage. approaches were discussed. In the structural geology
(4) Local geological structures and seismic velocities should approach, the maximum magnitude is inferred from the
be mapped during the first stimulation phase of a geo‐ largest potentially active fault in the geothermal reser‐
thermal reservoir. Fracture mapping is recommended voir using an earthquake‐fault relationship. Determinis‐
down to the reservoir depth. In the stimulation phase, tic fracture model approaches use prescribed slip patch‐
new fractures are created. Careful seismic monitoring is es in slider spring models, or dynamic mixed‐mode frac‐
needed to maintain control of this permeability‐ ture models. Both models allow reproducing LME in the
enhancing process. Continuous monitoring of induced post shut‐in phase of EGS stimulations. As opposed to the
seismicity is required from the beginning of the stimula‐ traditional, probabilistic seismic hazard assessment
tion experiment to detect runaway fracturing, also along (PSHA) approach, we recommend to apply the hybrid
buried faults. technique FISHA (forward induced seismic hazard as‐
(5) Due to the nature of EGS sites (engineered fracture net‐ sessment) combining deterministic fracture models with
works), these systems are characterized by higher injec‐ probabilistic forward computations.
tion rates (up to 100 l/s) compared to most wastewater
disposal wells (injection rate < 20 l/s). Reservoir injec‐
tivity (ratio of maximum inflow to maximum wellhead Acknowledgements
pressure) indicates low values ( 1.5) for waste water
disposal sites and values up to 6.3 at EGS sites where This work was supported by the European Union funded
creation of new fractures is desired. project GEISER (Geothermal Engineering Integrating Mitiga‐
(6) During the EGS stimulation phase, larger magnitude tion of Induced Seismicity in Reservoirs, Grant no.: 241321‐2).
events (LME) occur often after shut‐in at greater dis‐ We acknowledge the permission of Stefan Baisch (Q‐con Geo‐
tances from the injection well. The probability of a LME thermal Reservoir Engineering, Bad Bergzabern, Germany) to
increases with injection volume, even though there are use field data from the Basel SERIANEX study. We thank
geologic differences. We differentiate between long‐term Corinne Bachmann and Julie Albaric for providing one figure
injection operations (reservoir impoundment, waste wa‐ from Basel and Paralana site stimulation, respectively.
ter disposal) generating relatively larger LME compared
to short‐term fluid injection. Of the short‐term injections,
EGS stimulations have in general shown a higher pro‐ References
pensity to produce LME, compared to hydraulic fractur‐
ing in oil and gas operations. Ake, J., Mahrer, K., Connell, D.O., Block, L., 2005. Deep‐injection and
(7) Early during the stimulation phase and close to the injec‐ closely monitored induced seismicity at Paradox Valley, Colorado.
tion well, pore pressures are generally high (near‐field). Bull. Seismol. Soc. Am. 95, 664–683.
In this situation, many small events are induced, often on Albaric, J., Oye, V., Langet, N., Hasting, M., Lecomte, I., Messeiller, M.,
2014. Monitoring of induced seismicity during the first geothermal
faults with low applied shear stress, consistent with high
reservoir stimulation at Paralana, Australia. Geothermics 52, 120–131.
‐values, low stress drops and significant volumetric Amitrano, D., 2003. Brittle–ductile transition and associated seismici‐
components of the focal mechanisms. As the injected wa‐ ty: experimental and numerical studies and relationship with the b
ter migrates away from the injection point and the pore value. J. Geophys. Res. 108 (B1), 2044, http://dx.doi.org/10.1029/
pressure decreases (far‐field), the volumetric component 2001JB000680.
of the focal mechanism becomes less significant and the Asanuma, H., Nozaki, H., Niitsuma, H., Wyborn, D., 2005. Interpretation
shear stress necessary to trigger events increases. This is of microseismic events with larger magnitude collected at Cooper
reflected in higher stress drops and in lower ‐values. Basin, Australia. Geotherm. Resour. Council Trans. 29, 87–91.
Bachmann, C., Wiemer, S., Goertz‐Allmann, B., Woessner, J., 2012.
The latter implies a higher probability for the occurrence
Influence of pore pressure on the size distribution of induced earth‐
of LME at the periphery of the stimulated volume and quakes. Geophys. Res. Lett. 39, 9, http://dx.doi.org/10.1029/
during the later stages of the stimulation. 2012GL051480.
(8) Activities to optimize reservoir performance cause tem‐ Baisch, S., Harjes, H.‐P., 2003. A model for fluid‐injection‐induced
poral changes of physical rock properties such as cooling seismicity at the KTB, Germany. Geophys. J. Int. 152, 160–170.
of rock as well as dissolution and precipitation processes Baisch, S., Vörös, R., 2009. Induced seismicity, Technical Report SERI‐
after long term circulation. Velocity ratios ( / ) are ANEX, AP3000, Appendix 2., pp. 56–59.
indicative of fluid path ways and pore‐fluid aggregate in Baisch, S., Vörös, R., 2010. Reservoir induced seismicity: where, when,
why and how strong? In: Proceedings of World Geothermal Con‐
the fractured rock mass. For the analysis of seismicity
gress 2010, Bali, Indonesia, 25–29 April 2010.
and the local seismic velocity, 4D velocity tomography is Baisch, S., Weidler, R., Vörös, R., Wyborn, D., DeGraaf, L., 2006. Induced
needed. Since significant portions of aseismic slip cannot seismicity during stimulation of a geothermal HFR reservoir in the
be ruled out at EGS sites, seismic monitoring should be Cooper Basin (Australia). Bull. Seismol. Soc. Am. 96, 2242–2256.
conducted in concert with tilt metre measurements. Baisch, S., Vörös, R., Weidler, R., 2009. Investigation of fault mecha‐
(9) EGS development in weak volcanic or sedimentary rocks nisms during geothermal reservoir stimulation experiments in
(e.g., Berlin, El Salvador, Groß Schönebeck) may be very Cooper Basin, Australia. Bull. Seismol. Soc. Am. 99 (1), 148–158.
different from the permeability enhancement process in Baisch, S., Vörös, R., Rothert, E., Stang, H., Jung, R., Schellschmidt, R.,
18
2010. A numerical model for fluid injection induced seismicity at 82, 504–508, http://dx.doi.org/10.1785/gssrl.82.4.504.
Soultz‐sous‐Forêts. Int. J. Rock Mech. Mining Sci. 47, 405–413. Dyer, B.C., Jones, R.H., Cowles, J.F., Barkved, O., Folstad, P.G., March
Bame, D., Fehler, M., 1986. Observations of long period earthquakes 1999. Microseismic survey of a North Sea reservoir. World Oil, 74–
accompanying hydraulic fracturing. Geophys. Res. Lett. 13 (1), 149– 78.
152. Economides, M.J., Saputelli, L., 2005. Chapter 6.2. Production optimiza‐
Batchelor, A.S., Garnish, J.D., 1990. The industrial exploitation of geo‐ tion. In: Amadei, C. (Ed.), Encyclopedia of Hydrocarbons. Vol. 1 Ex‐
thermal resources in Europe. Tectonophysics 178, 269–276. ploration, Production and Transport. Fondata da Giovanni Treccani
Bommer, J.J., Oates, S.J., Cepeda, M., Lindholm, C., Bird, J., Torres, R., S.p.A, Rome, Italy, pp. 725–760.
Marroquin, G., Rivas, J., 2006. Control of hazard due to seismicity Evans, L., Moriya, H., Niitsuma, H., Jones, R.H., Phillips, W.S., Genter, A.,
induced by a hot fractured rock geothermal project. Eng. Geol. 83, Sausse, J., Jung, R., Baria, R., 2005. Microseismicity and permeability
287–306. enhancement of the hydrologic structures during massive fluid in‐
Bromley, C.J., 2010. Promoting beneficial environmental effects and jections into granite at 3 km depth at Soultz HDR site. Geophys. J. Int.
improving long‐term utilization strategies through IEA‐GIA collabo‐ 160, 388–412.
ration. In: Proceedings of World Geothermal Congress 2010, Bali, Evans, K.F., Zappone, A., Kraft, T., Deichmann, N., Moia, F., 2012. A
Indonesia, 25–29 April 2010. survey of the induced seismic responses to fluid injection in geo‐
Bromley, C.J., Pearson, C.F., Rigor Jr., D.M., PNOC‐EDC, 1986. Micro‐ thermal and CO 2 reservoirs in Europe. Geothermics 41, 30–54.
earthquakes at the Puhagan geothermal field, Philippines – a case of Faenza, L., Michelini, A., 2010. Regression analysis of MCS intensity
induced seismicity. J. Volcanol. Geotherm. Res. 31 (3–4), 293–311. and ground motion parameters in Italy and its application in
Bruel, D., 2007. Using the migration of the induced seismicity as a ShakeMap. Geophys. J. Int. 180, 1138–1152.
constraint for fractured Hot Dry Rock reservoir modeling. Int. J. Fehler, M., Jupe, A., Asanuma, H., 2001. More than cloud: new tech‐
Rock Mech. Mining Sci. 44, 1106–1117. niques for characterizing reservoir structure using induced seismici‐
Bruhn, D., Huenges, E., Ágústsson, K., Zang, A., Kwiatek, G., Rachez, X., ty. Lead. Edge 20, 324–328.
Wiemer, S., van Wees, J.‐D., Calcagno, P., Kohl, T., Dorbath, C., de Na‐ Ferrazzini, V., Chouet, B., Fehler, M., Aki, K., 1990. Quantitative analysis
tale, G., Oye, V., 2011. Geothermal engineering integrating mitigation of long‐period events recorded during hydrofracturing experiments
of induced seismicity in reservoirs – The European GEISER Project. at Fenton Hill, New Mexico. J. Geophys. Res. 95, 21871–21884.
Geotherm. Res. Council Trans. 35, 1623–1626. Fischer, T., Hainzl, S., Eisner, L., Shapiro, S.A., Le Calvez, J., 2008. Micro‐
Calcagno, P., Bouchot, V., Thinon, I., Bourgine, B., 2012. A new 3D fault seismic signatures of hydraulic fracture growth in sediment for‐
model of the Bouillante geothermal province combining onshore mations: observations and modelling. J. Geophys. Res. 113, B02307,
and offshore structural knowledge (French West Indies). Tectono‐ http://dx.doi.org/10.1029/2007JB005070.
physics 526–529, 185–195. Fix, J.E., Adair, R.G., Fisher, T., Mahrer, K., Mulcahy, C., Myers, B., Swan‐
Calò, M., Dorbath, C., Cornet, F., Cuenot, N., 2011. Large scale aseismic son, J., Woerpel, J.C., 1989. Development of Microseismic Methods to
motion identified through 4D P wave tomography. Geophys. J. Int. Determine Hydraulic Fracture Dimensions, Gas Technology Institute
186, 1295–1314. Technical Report No. 89‐0116.
Calò, M., Dorbath, C., Frogneux, M., 2014. Injection tests at the EGS Frohlich, C., Hayward, C., Stump, B., Potter, E., 2011. The Dallas–Fort
reservoir of Soultz‐sous‐Forêts, Seismic response of the GPK4 stimu‐ Worth earthquake sequence: October 2008 through May 2009. Bull.
lations. Geothermics 52, 50–58. Seismol. Soc. Am. 101 (1), 327–340, http://dx.doi.org/10.1785/
Castagna, J.P., Batzle, M.L., Eastwood, R.L., 1985. Relationships be‐ 0120100131.
tween compressional‐wave and shear‐wave velocities in clastic sili‐ Frohlich, C., Ellsworth, W., Brown, W.A., Brunt, M., Luetgert, J., Mac‐
cate rocks. Geophysics 50, 571–581. Donald, T., Walter, S., 2014. The 17 May 2012 M4.8 earthquake near
Charlety, J., Cunot, N., Dorbath, L., Haessler, H., Frogneux, M., 2007. Tipson, east Texas: an event possibly triggered by fluid‐injection. J.
Large earthquakes during hydraulic stimulations at the geothermal Geophys. Res. Solid Earth 119, 581–593, http://dx.doi.
site of Soultz‐sous‐Forêts. Int. J. Rock Mech. Mining Sci. 44, 1091– org/10.1002/2013JB010755.
1105. Giardini, D., 2009. Geothermal quake risk must be faced. Nature 426,
Clarke, D., Brenguier, F., Froger, J.‐L., Shapiro, N.M., Peltier, A., Stau‐ 848–849.
dacher, T., 2013. Timing of a large volcanic flank movement at Piton Gischig, V.S., Wiemer, S., 2013. A stochastic model for induced seismic‐
de la Fournaise volcano using noise‐based seismic monitoring and ity based on non‐linear enhancement. Geophys. J. Int.,
ground deformation measurements. Geophys. J. Int., http://dx.doi. http://dx.doi.org/10.1093/gji/ggt164.
org/10. 1093/gji/ggt/276. Goertz‐Allmann, B.P., Goertz, A., Wiemer, S., 2011. Stress drop varia‐
Cornet, F., Berard, T., Bourouis, S., 2007. How close to failure is a tions of induced earthquakes at the Basel geothermal site. Geophys.
granitic rock mass at 5 km depth? Int. J. Rock Mech. Mining Sci. 44, Res. Lett. 38, L09308, http://dx.doi.org/10.1029/ 2011GL047498.
47–66. Gritto, R., Jarpe, S.P., 2014. Temporal variations of VP/VS ratio at The
Cuenot, N., Dorbath, C., Dorbath, L., 2008. Analysis of the microseis‐ Geysers geothermal field, USA. Geothermics 52, 112–119.
micity induced by fluid injections at the EGS site of Soultz‐sous‐ Grünthal, G., 2014. Induced seismicity related to geothermal projects
Forêts (Alsace, France): implications for the characterization of the versus natural tectonic earthquakes and other types of induced
geothermal reservoir properties. Pure Appl. Geophys. 165, 707–828. seismic events in Central Europe. Geothermics 52, 22–35.
de Pater, C.J., Baisch, S., 2011. Geomechanical Study of Bowland Shale Grünthal, G., Stromeyer, D., Wahlström, R., 2009. Harmonization check
Seismicity. Cudrilla Resources Ltd, pp. 57, www.cuadrillaresources. of Mw within the central, northern, and northwestern European
com/wp‐content/uploads/2012/02/Geomechanical‐Study‐of‐ earthquake catalogue (CENEC). J. Seismol. 13, 613–632.
Bowland‐Shale‐Seismicity_02‐11‐11.pdf Gupta, H.K., 2002. A review of recent studies of triggered earthquakes
Deichmann, N., Ernst, J., 2009. Earthquake focal mechanisms of the by artificial water reservoirs with special emphasis on earthquakes
induced seismicity in the 2006 and 2007 below Basel (Switzerland). in Koyna, India. Earth Sci. Rev. 58 (3–4), 279–310.
Swiss J. Geosci. 102, 457–466. Gupta, H.K., 2011. Artificial water reservoir triggered earthquakes. In:
Deichmann, N., Giardini, D., 2009. Earthquakes induced by the stimula‐ Gupta, H.K. (Ed.), Encyclopedia of Solid Earth Geophysics, vol. I.
tion of enhanced geothermal system below Basel (Switzerland). Springer Science and Business Media B.V., Dordrecht, The Nether‐
Seismol. Res. Lett. 80, 784–798. lands, pp. 15–24.
Deichmann, N., Kraft, T., Evans, K., 2014. Mapping of faults activated by Gutenberg, B., Richter, C., 1954. Seismicity in the Earth, 2nd ed. Prince‐
the stimulation of the Basel enhanced geothermal system. Geother‐ ton University Press, Princeton, NJ, pp. 16–25.
mics 52, 74–83. Hakimhashemi, A.H., Yoon, J.‐S., Heidbach, O., Zang, A., Grünthal, G.,
Dorbath, L., Cuenot, N., Genter, A., Frogneux, M., 2009. Seismic re‐ 2013. FISHA – Forward induced seismic hazard assessment applica‐
sponse of the fractured and faulted granite to massive water injec‐ tion to synthetic seismicity catalogue generated by hydraulic stimu‐
tion at 5 km depth at Soultz‐sous‐Forêts (France). Geophys. J. Int., lation modeling. In: Proceedings of Thirty‐Eight Workshop on Geo‐
http://dx.doi.org/10.1111/j.1365‐246X.2009.04030.x. thermal Reservoir Engineering Stanford University, Stanford, CA,
Dost, B., Haak, H.W., 2007. Natural and induced seismicity. In: Wong, STP‐TR‐198, pp. 455–461.
Th0E., Batjes, D.A.J., de Jager, J. (Eds.), Geology of the Netherlands. Hakimhashemi, A.H., Schoenball, M., Heidbach, O., Zang, A., Grünthal,
Royal Netherlands Academy of Arts and Sciences, pp. 223–239. G., 2014. Forward modelling of seismicity rate changes in georeser‐
Douglas, J., Jousset, P., 2011. Modeling the difference in ground‐motion voirs with a hybrid geomechanical–statistical prototype model. Ge‐
magnitude‐scaling in small and large earthquakes. Seismic Res. Lett. othermics 52, 185–194.
19
Häring, M.O., Schanz, U., Ladner, F., Dyer, B.C., 2008. Characterization analysis of seismicity induced at Berlin geothermal field, El Salva‐
of Basel 1 enhanced geothermal system. Geothermics 37, 469–495. dor. Geothermics 52, 98–111.
Hasting, M., Albaric, J., Oye, V., Reid, R., Messeiller, M., Llanos, E., Malin, Lebert, F., Bernardie, S., Mainsant, G., 2011. Hydroacoustic monitoring
P., Shalev, E., Hogg, M., Alvarez, M., Miller, A., Walter, C., Boese, C., of a salt cavity: an analysis of precursory events of the collapse. Nat.
Voss, N., 2011. Real‐time induced seismicity monitoring during Hazards Earth Syst. Sci. 11, 2663–2675, http://dx.doi.org/10.5194/
wellbore stimulation at Paralana‐2, South Australia. In: Australian nhess‐11‐2663‐2011.
Geothermal Energy Conference 2011. Leonard, M., 2010. Earthquake fault scaling: self‐consistent relating of
Hazzard, J.F., Young, R.P., 2004. Dynamic modelling of induced seis‐ rupture length, width, average displacement, and moment release.
micity. Int. J. Rock Mech. Mining Sci. 41, 1365–1376. Bull. Seismol. Soc. Am. 100, 1971–1988.
Healy, J.H., Rubey, W.W., Griggs, D.T., Raleigh, C.B., 1968. The Denver Lockner, D., Byerlee, J.D., 1972. Hydrofracure in Weber sandstone at
earthquakes. Science 161 (3848), 1301–1310. high confining pressure and differential stress. J. Geophys. Res. 82,
Henderson, J.R., Barton, D.J., Foulger, G.R., 2002. Fractal clustering of 2018–2026.
induced seismicity in The Geysers geothermal area, California. Geo‐ Lockner, D., Byerlee, J.D., 1992. Fault growth and acoustic emissions in
phys. J. Int. 139 (2), 317–324. Herrmann, R.B., Park, S.‐K., Wang, C.‐ confined granite. Appl. Mech. Rev. 45 (3 (part 2)), S165–S173.
Y., 1981. The Denver earthquakes of 1967–1968. Bull. Seismol. Soc. Majer, E.L., Baria, R., Stark, M., Oates, S., Bommer, J., Smith, B., Asanu‐
Am. 161, 1301–1310. ma, H., 2007. Induced seismicity associated with enhanced geo‐
Holschneider, M., Zöller, G., Clements, R., Schorlemmer, D., 2014. Can thermal systems. Geothermics 36, 185–222.
we test for the maximum possible earthquake magnitude? J. Ge‐ Majer, E.L., Queen, J., Daley, T., Long, R., 2008. Proceedings of Annual
ophys. Res. Solid Earth 119, 2019–2028, http://dx.doi.org/10. Meeting of Society of Exploration Geophysicists, Houston, TX.
1002/2013JB010319. Maxwell, S.C., Rutledge, J., Jones, R., Fehler, M., 2010. Petroleum reser‐
Horton, S., 2012. Disposal of hydrofracking waste fluid by injection voir characterization using downhole microseismic monitoring. Ge‐
into subsurface aquifers triggers earthquake swarm in central Ar‐ ophysics 75, http://dx.doi.org/10.1190/1.3477966, 75A129–
kansas with potential for damaging earthquake. Seismol. Res. Lett. 75A137.
83, 250–260. McGarr, A., 1976. Seismic moments and volume change. J. Geophys.
Hsieh, P.A., Bredehoft, J.D., 1981. A reservoir analysis of the Denver Res. 81 (8), 1487–1494.
earthquakes: a case of induced seismicity. J. Geophys. Res. 86, 903– McGarr, A., 2014. Maximum magnitude earthquakes by induced fluid
920. injection. J. Geophys. Res. Solid Earth, http://dx.doi.org/10.1002/
Hubbert, K.M., Rubey, W.W., 1959. Role of fluid pressure in mechanics 2013JB010597.
of overthrust faulting: parts I and II. Geol. Soc. Am. Bull. 70, 115– McGuire, R.K., 2004. Seismic Hazard and Risk Analysis, vol. MNO‐10.
205. Earthquake Engineering Research Institute, Oakland, CA, USA, pp.
Hubbert, K.M., Willis, D.G., 1957. Mechanics of hydraulic fracturing. 221.
Petrol. Trans. AIME T.P. 4597 210, 153–166. Megies, T., Wassermann, J., 2014. Microseismicity observed at a non‐
Huenges, E., Trautwein, U., Legarth, B., Zimmermann, G., 2006. Fluid pressure‐stimulated geothermal power plant. Geothermics 52, 36–49.
pressure variation in a sedimentary geothermal reservoir in the Mendecki, A.J., 1997. Seismic Monitoring in Mines. Chapman and Hal,
North German Basin: case study Groß‐Schönebeck. Pure Appl. Ge‐ London, UK, pp. 280.
ophys. 163, 2141–2152. Meremonte, M.E., Lahr, J.C., Frankel, A.D., Dewey, J.W., Crone, A.J.,
Hunt, T.M., Latter, J.H., 1982. A survey of seismic activity near Overturf, D.E., Carver, D.L., Bice, W.T., 2002. Investigation of an
Wairakei geothermal field, New Zealand. J. Volcanol. Geotherm. Res. earthquake swarm near Trinidad, Colorado, August–October 2001,
14 (3–4), 319–334. Open‐File Report 02‐0073, U.S. Geological Survey., pp. 32.
Jost, M.L., Busselberg, T., Jost, O., Harjes, H.‐P., 1998. Source parame‐ Michelet, S., Töksöz, M.N., 2007. Fracture mapping in the Soultz‐sous‐
ters of injection‐induced microearthquakes at 9 km depth at the Forêts geothermal field using microearthquake locations. J. Geophys.
KTB deep drilling site, Germany. Bull. Seismol. Soc. Am. 88, 815–832. Res. 112, B07315, http://dx.doi.org/10.1029/2006JB00.4442.
Jousset, P., Rohmer, J., 2012. Evidence for remotely triggered micro‐ Mogi, K., 1962. Magnitude frequency relation for elastic shocks ac‐
earthquakes during salt cavern collapse. Geophys. J. Int. 191, 207– companying fractures of various materials and some related prob‐
223, http://dx.doi.org/10.1111/j.1365‐246X.2012.05598.x. lems in earthquakes. Bull. Earthquake Res. Inst. 40, 831–853.
Jousset, P., Haberland, C., Bauer, K., Arnason, K., Weber, M., Fabriol, H., Mulyadi, 2010. Case study: hydraulic fracturing experiment in the
2010. Seismic tomography and long‐period earthquakes observa‐ Wayang Windu Geothermal field. In: Proceedings World Geothermal
tion and modelling at the Hengill geothermal system. In: Proceed‐ Congress, Bali, Indonesia, 25–29 April 2010.
ings World Geothermal Congress, Bali, Indonesia, 25–29 April 2010. Nagano, K., Moriya, H., Asanuma, H., Sato, M., Niitsuma, H., Kaieda, H.,
Jousset, P., Haberland, C., Bauer, K., Arnason, K., 2011. Hengill geo‐ 1994. Down‐hole acoustic emission measurement of hydraulic frac‐
thermal volcanic complex (Iceland) characterized by integrated ge‐ turing in Ogachi HDR model field. J. Geotherm. Res. Soc. Jpn. 16, 85–
ophysical observations. Geothermics 40, 1–24. 108 (in Japanese).
Jung, R., 2013. EGS–Goodbye of back to the future. In: Bunger, A.P., Nicholson, C., Wesson, R.L., 1990. Earthquake hazard associated with
McLennon, J., Jeffrey, R. (Eds.), Earth Planetary Sciences, Geology deep well injection: A report to the U.S. Environmental Protection
and Geophysics, Effective and Sustainable Hydraulic Fracturing., Agency, U.S. Geological Survey Bulletin 1951. United States Govern‐
http://dx.doi.org/10.5772/56458, ISBN 978‐953‐51‐1137‐5. mental Printing Office, Washington, DC, pp. 74.
Keranen, K.M., Savage, H.M., Abers, G.A., Cochran, E.S., 2013. Potential‐ Oppenheimer, D.C., 1986. Extensional tectonics at The Geysers geo‐
ly induced earthquakes in Oklahoma, USA: links between thermal area, California. J. Geophys. Res. 91, 11463–11476.
wastewater injection and the 2011 Mw 5.7 earthquake sequence. Oye, V., Chavarria, J.A., Malin, P.E., 2004. Determining SAFOD area
Geology, http://dx.doi.org/10.1130/G34045.1. microearthquake locations solely with Pilot Hole seismic array data.
Kijko, A., 2004. Estimation of the maximum earthquake magnitude, Geophys. Res. Lett. 31, L12S10, http://dx.doi.org/10.1029/
Mmax. Pure Appl. Geophys. 161, 1655–1681. 2003GL019403.
Kim, W.‐Y., 2013. Induced seismicity associated with fluid injection Oye, V., Gharti, H.N., Kühn, D., Braathen, A., 2010. Microseismic moni‐
into a deep well in Youngstown, Ohio. J. Geophys. Res. Solid Earth toring of fluid injection at the Longyearbyen CO2‐Lab, Svalbard. In:
118, 3506–3518. Ritter, J., Oth, A. (Eds.), Cahiers du Centre Européen de Géody‐
Kraft, T., Deichmann, N., 2014. High‐precision relocation and focal namique et de Séismologie, Proceedings of the Workshop, “Induced
mechanism of the injection induced seismicity at the Basel EGS. Ge‐ Seismicity”, November 15–17, 2010, Luxembourg, p. 141, ISBN 978‐
othermics 52, 59–73. 2‐91989‐709‐4.
Kraft, T., Mai, P.M., Wiemer, S., Deichmann, N., Ripperger, J., Kästli, P., Pearson, C., 1981. The relationship between microseismicity and high
Bachmann, C., Fäh, D., Wössner, J., Giardini, D., 2009. Enhanced geo‐ pore pressures during hydraulic stimulation experiments in low
thermal systems: mitigation risk in urban areas. EOS Trans. Am. Ge‐ permeability granitic rocks. J. Geophys. Res. 86, 7855–7864.
ophys. Union 90, 273–274. Pine, J.R., Batchelor, A.S., 1984. Downward migration of shearing in
Kwiatek, G., Bohnhoff, M., Dresen, G., Schulze, A., Schulte, T., Zimmer‐ jointed rock during hydraulic injections. Int. J. Rock Mech. Mining
mann, G., Huenges, E., 2010. Microseismicity induced during fluid‐ Sci. Geomech. Abst. 21 (5), 249–263.
injection: a case study from the geothermal site at Groß Schönebeck, Rutqvist, J., Oldenburg, C.M., Dobson, P.F., 2010. Predicting the spatial
North German Basin. Acta Geophys. 58, 995–1020. extend of injection‐induced zones of enhanced permeability at the
Kwiatek, G., Bulut, F., Bohnhoff, M., Dresen, G., 2014. High‐resolution Northwest Geysers EGS demonstration project. In: 44th US Rock
20
Mechanics Symposium and 5th U.S.‐Canada Rock Mechanics Sympo‐ Geology 28, 399–402.
sium, Salt Lake City, UT June 27–30. American Rock Mechanics As‐ Trifu, C.I. (Ed.), 2002. The mechanism of induced seismicity. Pure and
sociation, pp. 10–502. Applied Geophysics 159(1–3), Topical Issue. Birkhäuser, Basel, p.
Sanjuan, B., Jousset, P., Pajot, G., Debeglia, N., de Michele, M., Brach, M., 617.
Dupont, F., Braibant, G., Lasne, E., Duré, F., 2010. Monitoring of the Trifu, C.I. (Ed.), 2010. Monitoring induced seismicity. Pure Applied
Bouillante geothermal exploitation (Guadeloupe, French West In‐ Geophysics, Topical Issue. Birkhäuser, Basel.
dies) and the impact on its immediate environment. In: Proc. World Trubitt, T., Walker, A.B., Browitt, C.W.A., 1987. Perceptible hydrofrac‐
Geothermal Congress 2010, Bali, Indonesia, 25–29 April 2010. ture seismic events caused by the Hot‐Dry‐Rock geothermal project
Schoenball, M., Müller, T.M., Müller, B.I.R., Heidbach, O., 2010. Fluid‐ in Cornwall. In: Global Seismology Report Number 339, British Geo‐
induced micro‐seismicity in pre‐stressed rock masses. Geophys. J. logical Survey, Edinburgh, UK, p. 23.
Int. 180, 813–819. van Eck, T., Goutbeek, F., Haak, H., Dost, B., 2006. Seismic hazard due
Scholz, C.H., 1968. The frequency–magnitude relation of microfractur‐ to small‐magnitude, shallow‐source, induced earthquakes in The
ing in rock and its relation to earthquakes. Bull. Seismol. Soc. Am. 58 Netherlands. Eng. Geol. 87, 105–121.
(1), 399–415. van Eijis, R.M.H.E., Mulders, F.M.M., Nepveu, M., Kenter, C.J., Scheffers,
Scholz, C.H., 1990. The Mechanics of Earthquakes and Faulting. Cam‐ B.C., 2006. Correlation between hydrocarbon reservoir properties
bridge University Press, Cambridge, UK, pp. 439. and induced seismicity in the Netherlands. Eng. Geol. 84, 99–111.
Schorlemmer, D., Wiemer, S., Wyss, M., 2005. Variations in earth‐ van Wees, J.‐D., Buijze, L., van Thienen‐Visser, K., Nepveu, M., Wassing,
quake‐size distribution across different stress regimes. Nature 437 B., Orlic, B., Fokker, P., 2014. Natural stress and fault controls in in‐
(7058), 539–542. duced seismicity: what can we learn from gas depletion in The
Shapiro, S.A., Dinske, C., 2007. Violation of the Kaiser effect by hydrau‐ Netherlands? Geothermics 52, 206–219.
lic‐fracturing‐related microseismicity. J. Geophys. Eng. 4, 378–383. Wassing, B., Orlic, B., Fokker, P., Thienen‐Visser, K., van Wees, J.‐D.,
Shapiro, S.A., Huenges, E., Borm, G., 1997. Estimating the crust perme‐ Buijze, L., 2014. Coupled continuum modelling of fracture reactiva‐
ability from fluid‐injection‐induced seismic emission at the KTB site. tion and induced seismicity during enhanced geothermal opera‐
Geophys. J. Int. 131, F15–F18. tions. Geothermics 52, 153–164.
Shapiro, S.A., Audigane, P., Royer, J.‐J., 1999. Large‐scale permeability Wells, D.L., Coppersmith, K.J., 1994. New empirical relationships
tensor of rocks from induced microseismicity. Geophys. J. Int. 137, among magnitude, rupture length, rupture width, rupture area and
207–213. surface displacement. Bull. Seismol. Soc. Am. 84, 974–1002.
Shapiro, S.A., Dinske, C., Kummerow, J., 2007. Probability of a given‐ Wyss, M., 1979. Estimating maximum expectable magnitude of earth‐
magnitude earthquake induced by a fluid injection. Geophys. Res. quakes from fault dimensions. Geology 7, 336–340.
Lett. 34 (L22314), http://dx.doi.org/10.1029/2007GL031615. Yamabe, T.H., Hamza, V.M., 1996. Geothermal investigations in a area
Shapiro, S.A., Dinske, C., Langenbruch, C., Wenzel, F., 2010. Seismogen‐ of induced seismic activity, Northern Sao Paulo State, Brasil. Tecto‐
ic index and magnitude probability of earthquakes induced during nophysics 253, 209–225.
reservoir stimulations. Lead. Edge March 2010, 304–309. Yoon, J.‐S., Zang, A., Stephansson, O., 2013. Simulation of multi‐stage
Silitonga, T.H., Siahaan, E.E., Suroso, 2005. A Poisson’s ratio distribu‐ hydraulic fracturing and induced seismicity in naturally fractured
tion from Wadati diagram as indicator of fracturing of Lahendong reservoir using PFC2D. In: Zhu, Detournay, Hart, Nelson (Eds.), Con‐
geothermal field, North Sulawesi, Indonesia, WGC 2005, Antalya, tinuum and Distinct Element Numerical Modeling in Geomechanics.
Turkey, 24–29 April. Itasca International Inc., Minneapolis, ISBN 978‐0‐9767577‐3‐3,
Simiyu, S.M., 1999. Induced seismicity during interference tests at OW‐ paper 06‐02.
719, Kenya. J. Geotherm. 28, 785–802. Yoon, J.‐S., Zang, A., Stephansson, O., 2014. Numerical investigation on
Stephansson, O., Zang, A., 2012. ISRM suggested methods for rock optimized stimulation of intact and naturally fractured deep geo‐
stress estimation – Part 5: Establishing a model for the in‐situ stress thermal reservoirs using hydro‐mechanical coupled discrete parti‐
at a given site. Rock Mech. Rock Eng. 45, 955–969. cles joints model. Geothermics 52, 165–184.
Suckale, J., 2009. Induced seismicity in hydrocarbon fields. Adv. Ge‐ Zang, A., Stephansson, O., 2010. Stress Field of the Earth’s Crust.
ophys. 51, 55–106. Springer Science + Business Media B.V., Dordrecht, pp. 322.
Talwani, P., Acree, A., 1985. Pore pressure diffusion and the mecha‐ Zang, A., Wagner, F.C., Dresen, G., 1996. Acoustic emission, microstruc‐
nism of reservoir induced seismicity. Pure Appl. Geophys. 122, 947– ture, and damage model of dry and wet sandstone stressed to fail‐
965. ure. J. Geophys. Res. 101 (B8), 17507–17521.
Talwani, P., Chen, L., Gahalaut, K., 2007. Seismogenic permeability, ks. Zang, A., Wagner, F.C., Stanchits, S., Janssen, C., Dresen, G., 2000. Frac‐
J. Geophys. Res. 112, B07309, http://dx.doi.org/10.1029/ ture process zone in granite. J. Geophys. Res. 105 (B10), 23651–
2006JB004665. 23661.
Terakawa, T., Miller, S.A., Deichmann, N., 2012. High fluid pressure and Zang, A., Yoon, J.‐S., Stephansson, O., Heidbach, O., 2013. Fatigue hy‐
triggered earthquakes in the enhanced geothermal system in Basel, draulic fracturing by cyclic reservoir treatment enhances permeabil‐
Switzerland. J. Geophys. Res. 117, B07305, http://dx.doi.org/ ity and reduces induced seismicity. Geophys. J. Int. 195, 1282–1287.
10.1029/2011JB008980. Zhao, P., Oye, V., Kühn, D., Cesca, S., 2014. Double‐couple and moment
Tester, J.W., Anderson, B., Batchelor, A., Blackwell, D., DiPippo, R., tensor solutions of seismicity induced by hydraulic stimulation of
Drake, E., Garnish, J., Livesay, B., Moore, M.C., Nichols, K., Petty, S., the geothermal site in Basel, Switzerland. Geothermics 52, 74–83.
Toksoz, N., Veatch, R., Augustine, C., Baria, R., Murphy, E., Negraru, Zhu, X., Gibson, J., Ravindran, N., Zinno, R., Sixta, D., 1996. Seismic
P., Richards, M., 2006. The Future of Geothermal Energy: Impact of imaging of hydraulic fractures in Carthage tight sands: a pilot study.
Enhanced Geothermal Systems (EGS) on the United States in the Lead. Edge 15, 218–224.
21st Century. Mass. Institute of Technology, Cambridge, MA, USA, Zoback, M.D., Harjes, H.‐P., 1997. Injection‐induced earthquakes and
pp. 209. crustal stress at 9 km depth at the KTB deep drilling site, Germany. J.
Townend, J., Zoback, M.D., 2000. How faulting keeps the crust strong. Geophys. Res. 102 (B8), 18477–18491.
21

Originally Published As

Uploaded by

Copyright:

Available Formats

Originally Published As

Uploaded by

Document Information

Original Title

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

Originally Published As

Uploaded by

Copyright:

Available Formats

Originally published as:

Analysis of induced seismicity in geothermal reservoirs – An overview

ARTICLE INFO ABSTRACT

1. Introduction production at hydrocarbon reservoirs (e.g., Suckale, 2009) and

tivated. Most EGS projects require drilling of boreholes to

mic footprint of a reservoir to be developed (Jung, 2013).

Before a geothermal system can be exploited, it has to be

sion due to Green’s functions that reveal less complexity.

than into crystalline rocks. Also from laboratory testing it is

tion to earthquake phenomena, Scholz (1968) highlighted that

Baisch and Vörös (2010) compared seismic signatures of

and mechanisms operating in situ, as evidenced by the differ‐

You might also like