SPE-200021-MS

Learning from Behavioral Frac Maps: A Montney Case Study in Integration

of Modern Microseismic and Production Data Analyses

Peyman Moradi, Doug Angus, and Katie Jeziorski, ESG Solutions; Amanda Greig and Michael de Groot, Encana

Copyright 2020, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Canada Unconventional Resources Conference originally scheduled to be held in Calgary, Alberta, Canada, 18 –
19 March 2020. Due to COVID-19 the physical event was postponed until 29 September – 2 October 2020 and was changed to a virtual event. The official proceedings
were published online on 24 September 2020.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Aimed at sharing the unconventional wisdom gained from a hydraulic fracturing monitoring case study
in the Montney tight gas play, the work showcases the ability of 4D modeling of collective behaviors of
microseismic events to chase the frac fluid and navigate the spatiotemporal fracture evolution. Moreover,
microseismicity-derived deformation fields are integrated with volumetric estimates made by rate transient
analysis to calibrate spatially-constrained SRV models. Through the case study, we give evidence of
fracture containment, evaluate the role of natural fractures and the use of diverting agents, estimate
cluster efficiencies, conduct analytical well spacing optimization, model productivity decline induced
by communication frac-hits from offsets, and provide contributing fracture dimensions and numerical
production forecasts. To support the interpretations, we supplement the work by the results of 3D physics-
based analytical modeling and multi-phase numerical simulations, and the findings are then validated using
two extensive datasets: production profiles acquired by fiber optic DAS, and reservoir fluid fingerprints
extracted from mud logs. Besides describing the evolution of seismicity during the treatment, the applied
integrated fracture mapping process gives a more reliable and unique SRV structure that streamlines forward
modeling and simulations in unconventional reservoirs as well as contributes to solving inverse problems
more mechanistically.

Introduction
Tight and shale plays are not easy to deal with as the complex features of the host, together with the
significant complexity in fracture growth (Warpinski et al. 1993, Fisher et al. 2002), make it elusive and
difficult to optimally design the stimulation and reliably forecast the production. For the unconventional
industry to better control the fracturing fluid delivery, the key would be to learn from understimulation/hit-
and-misses via access to rigorous models of the induced fracture network. A realistic fracture description
enables answering some of the most challenging questions on treatment optimization: What is the optimum
spacing of wells, stages, and clusters? Which diverging agent, perforation strategy, proppant pad design,
frac order and timing, and innovative technologies promote fracture uniformity and reservoir contact? Shall
we skip the faulted intervals? How to maximize energy conservation in the near-wellbore area? How to
2 SPE-200021-MS

take advantage of stress shadows? Applying the widely-used geomechanical models for predicting fracture
growth is highly uncertain and does not yield consistent results, as due to myriad factors, including but
not limited to, multi-scale heterogeneous reservoir properties (Deglint et al. 2017, Ross and Bustin 2009,
Li and Sheng 2017), layering and structural/geological lineaments and faults (Zoback et al. 2012, Wu and
Olson 2014), and induced and pre-existing stress non-uniformities (Morales et al. 2016, Manchanda et al.
2018, Agrawal and Sharma 2018), the fracture lateral and vertical growth is non-uniform and may change
dramatically. In this regard, microseismic monitoring provides a unique opportunity to listen to the rockmass
response to the treatment and deterministically visualize fracture and/or constrain stochastic fracture models
(Cornette et al. 2012, Neuhaus et al. 2012, Delorme et al. 2013, Pankaj et al. 2018).

Trifecta of Fracture Growth

Shear failures, however, enlarge the microseismic event clouds and make them bigger and more complex
than the conductive, contributing fracture network, carrying information on dry, wet, propped and
unpropped fractures. In other words, the microseismic-derived fracture lengths and heights (hereafter
referred to as Type I) are often orders of magnitude larger than the contributing fracture dimensions since
they rely on the individual event locations possibly hundreds of feet away from the well, many of which
with negligible apertures and/or no hydraulic connection to the well triggered by the far-field stress changes.
More information can be extracted from the cloud of microseismicity by clustering of events in time
and space to define the relationship of the events (and event clusters) to each other, which is beyond
the capabilities of conventional static interpretations. In this regard, Dynamic Parameter Analysis (DPA)
clusters events in a 4D space to model the collective behaviors and characterize different event populations
(Baig et al. 2017, Urbancic et al. 2017, Ben- Zion 2008, Nizkous et al. 2020). Therefore, the individual event
characteristics, together with the observed collective behaviors (each supported by a threshold, critical mass
of individual events) reveal the fracture pattern at each location. Since DPA enables chasing the inelastic
deformation associated with a high level of strain-softening and screens out the stress-driven seismicity, it
gives estimates of wet fracture dimensions (Type II). The wet fractures, induced by frac fluid, remain open
as long as the fluid pressure dominates the closure stress, and are responsible for: 1) pressure interference
between wells, 2) fracture interferences, 3) stage-to-stage and well-to-well interactions with open unpropped
fractures (Manchanda et al. 2014), and more importantly 4) diverting the frac fluid to distant areas from the
wellbore, leading to the loss of energy and inefficient usage of fluid and proppant. Therefore, knowing the
wet fractures is essential to evaluate the fracture uniformity and stimulation efficiency, unravel the effect of
stress shadowing, and optimize the treatment design (e.g., treatment order and timing). This zone shrinks
soon after the pumping pressure drops and/or the well is put into production. Type II fracture dimensions
describe the boundaries of fluid-driven seismicity and provide the upper limits of the Stimulated Reservoir
Volume (SRV).
As depicted by Fig.1, the wet fractures are divided into unpropped and propped (fully or partially) entities.
A propped fracture does not necessarily manifest itself in microseismic analysis, because the placement
process is meant to reach the permeable fractures and not to create new ones. Therefore, the propped
dimensions (Type III) are somewhat challenging to define, and a secondary source of information is needed
to distinguish different entities of wet fractures. Conventionally, one or more filters are applied to different
seismic attributes (e.g., moment magnitude and inter-event distance) to shrink the event cloud and reach the
propped fracture pattern. Alternatively, using the proppant-loading approach, fractures are ordered based
on distance from wellbore and then filled by proppant. However, the outcome of both types of approaches is
directly impacted by the 1) assumptions made regarding the fracture width and proppant placement pattern,
2) processing artifacts, 3) varying fracture density (which is not necessarily correlated to productivity), and
4) coverage/positioning of monitoring arrays. For instance, applying the proppant-loading approach gives
larger dimensions in areas with a limited number of events (by filling all of them with proppant), while
underestimates the fracture length and height in stages with higher microseismicity. Furthermore, relying on
SPE-200021-MS 3

the characteristics of individual events to explicitly model the fracture network, irrespective of their nature
and the population they belong to, could be misleading because their communication in recursive feedback
with the cloud determines the role they play in the treatment and production processes. Considering the fact
that the proppant-laden fluid is not as mobile as the clean fluid and does not travel far away from the well
(e.g., gravity, friction, and natural fractures decrease the frac fluid carrying capacity), the propped fracture
dimensions are potentially very close to the contributing fracture dimensions that connect the reservoir to the
well via continuous flow pathways. Therefore, the contributing fracture volume obtained from the modern
production data analysis could provide a valuable criterion to refine the SRV boundaries and distinguish
the contributing (propped/partially propped) from non-contributing wet fractures (Moradi and Angus 2019,
2020). In particular, the volume of Type III fractures can be inferred from the Rate Transient Analysis
(RTA), and its integration with the DPA-derived intricate spatial deformation fields would enable to calibrate
the fracture model, assuring that the final model of the SRV meets and leverages the findings of both
microseismic and production data analyses. Type III dimensions achieved through the integration of RTA
and DPA are explainable in the context of the fluid flow in propped and partially propped fractures. It
should be noted that, since the contributing fracture may endure a variety of secondary phenomena, such
as proppant embedment or interference frac-hits, the Type III dimensions are the maximum limits of the
contributing fracture.

Figure 1—Schematic demonstration of the evolution of wet fractures. The volume of the propped fracture is
potentially very close to the contributing fracture that is predictable by RTA, therefore the integration of RTA
and DPA allows to distribute the RTA-derived volume through the area with the highest level of deformation,
indicated by DPA, to give estimates of the spatial structure of the contributing (most-likely propped) fracture.

Workflow
The work summarizes the findings of the DPA-RTA integration in a case study in the Montney formation.
Beginning with multiple attributes and by applying a distribution of thresholds, we map out the relationships
4 SPE-200021-MS

among the collective behaviors and fracing parameters, track down the complexity and quality of fracture
network, and more importantly prioritize fractures according to the local strain-softening experienced by the
rock (deformation degree). Using four DPA behavioral indices, we describe fracture evolution and discretize
the event cloud into multiple populations potentially associated with different source mechanisms and
different levels of conductivity, deformation, and anisotropy. Subsequently, the DPA results are integrated
with modern production data analysis (flowing material balance and linear flow analysis) to model the
SRV structure and propagate the permeability. The work is then followed by validation of the results and
3D analytical and numerical modeling to history-match and adjust the reservoir parameters. The overall
workflow designed for the project is depicted in Fig.2.

Figure 2—Modeling pipeline designed for the project. The integrated SRV
modeling provides the main input required for analytical and numerical modeling.

Background
As depicted in Fig.3, the case study entails hydraulic fracturing of 4 stacked Multi-Frac Horizontal Wells
(MFHWs), hereafter referred to as A, B, C, and D, targeting the lower Montney in British Columbia. The
treatment starts with zipper-frac of A and C wells and is followed by zipper-frac of B and D pair. The
approximate site location is near the Dawson Creek area depicted on the right-hand side of Fig.4. The four
lateral wells are drilled along the expected direction of Shmin, the minimum horizontal stress, as such the
stimulation produces transverse hydraulic fractures.

Figure 3—Trajectories of the wells under consideration targeting the Lower Montney.
SPE-200021-MS 5

Figure 4—Montney unconventional outline (Modified NEBC 2012). The

case study targets the lower Montney near the Dawson Creek BC.

Geological Setting
The thickness of the Montney formation, as a stratigraphical unit of Lower Triassic age in the Western
Canadian Basin in British Columbia and Alberta, varies between 100 and 300 m. The developments in
the Montney started in the 1950s, mostly focused on the conventional sandstone and dolomite reservoirs.
Owing to the advances in fracing, the development of the Montney's unconventional siltstones started in
2005, and to date the number of MFHWs in the Montney exceeds 1400, contributing to more than 40
percent of Canada's natural gas production in 2018. In the Montney, as the generated hydrocarbons are in-
situ (or in very close proximity to the source rocks), the overall fluid system is reversed; water above the
oil, and oil above gas (increasingly drier with depth). When compared to the oil-producing area, the gas
condensate producing zones have larger pore throat sizes, a wider range of pore size distribution, lower
porosities, and higher permeabilities (Iwuoha 2018, Akai 2018). The present-day regional stress regime
of the Montney shale is compressional strike-slip with the regional maximum horizontal stress direction
approximately N40E (with local variations near faults). The upper Montney spans from the mid-Montney
to the base of Doig phosphate. The wells in the upper Montney often land closer to the Doing boundary
rather than the mid-Montney because the Doing usually acts as a fracture barrier and makes fractures grow
downward. The mid-Montney approximates the Smithian–Spathian boundary and can be identified on well
logs as a sharp increase in gamma-ray. Fractures initiated in the mid Montney may grow upward into the
upper Montney (Sekine et al. 2016). According to the petrophysical interpretations available for an appraisal
well within the area of the case study, there is a major difference between the Paleo and Archie porosities
in the mid Montney that indicates a sharp increase in TOC and corresponds to behavioral changes in water
saturation and gamma logs, possibly associated with major variations in the rock mechanical properties as
well (a potential fracture barrier). The middle and lower Montney are separated via the Dienerian-Smithian
boundary. Dividing the lower Montney to three lower, middle and upper units, the case study targets the
6 SPE-200021-MS

middle unit. Although usually remain constrained to the lower unit, due to the absence of a strong fracture
barrier that can impede the fracture height growth, fractures might extensively grow upward into the upper
units in the lower and middle Montney. According to the publicly available models within the target area,
the lower Montney shows higher gamma values and lower porosities, a CGR between 40-70 bbl/MMSCF,
and a pore pressure gradient between 10-14 kPa/m (Fig.5).

Figure 5—3D models of gamma and porosity in the Montney near the target spot (extracted from the
model made by Alberta Energy Regulator in 2018 based on 6513 gamma-ray and 534 porosity logs)

Initial Conditions
Using depth histograms of the located microseismic events, a 95-m-thick zone is considered as the target
vertical window of the wells under consideration extended from the mid-upper unit to the mid-lower unit.
The depth-dependent values of mechanical and petrophysical properties are used to conduct the integrated
SRV modeling and numerical simulations, while for the analytical modeling purposes, we use the average
porosity and saturation of the target window (3.6 and 10 %, respectively). To achieve maximum consistency
between saturation functions and core measurements, the generalized Corey correlation is adjusted using
5 and 30 % connate water and critical gas saturation, respectively, based on water, gas and oil saturation
exponents of 3, 2.5, and 2.5, respectively. Using DSI log data, the average rock density, shear and Young
moduli and the Poisson ratio are equal to 2620 kg/m3, 23.75e9 kg/ms2, 34e9 kg/ms2, and 0.214, respectively.
The average pore pressure gradient is 0.619 psi/ft, the initial reservoir pressure is equal to 5400 psi, and the
saturation pressure and oil density are 3600 psi and 60 API, respectively.
The gamma logs in the laterals are almost uniform along the well length and show more radiation in the
A and B wells, which may imply more silty rock in the southern wells (C and D). Therefore, rock nearby
the A well is more ductile and is expected to create thinner fracture regions with short-length branches. In
the mud gas data, the balance (C1-2/C3-5) is always smaller than 100 and higher than the wetness (C2-5/C1-5)
which indicates a wet gas to gas condensate fluid system. The PR 1978 model matches the CCE and DL
test points accurately with some minor modifications in the characteristics of the plus fraction.
The DFIT results in three of the wells under consideration show high Net Fracture Pressures (NFPs),
and together with the ISIP values that are comparable with the predicted overburden pressure, raise the
possibility of creation of complex fracture network with both vertical and horizontal fractures (Fig.6). High
NFPs in such reservoirs with a high fabric complexity (natural fractures) and elevated stress indicates
the creation of a complex fracture pattern (rather than parallel bi-wing vertical fractures), where a high
level of interactions between the hydraulic fractures and rock fabric is likely to occur (Potocki 2012).
When compared to the coupled pressure/stress analysis, the results show uncoupled and compressional
deformations (weak coupling of changes in pore pressure and minimum horizontal stress), which can
SPE-200021-MS 7

be translated into a high confinement level, leading to a high possibility/magnitude of losing unpropped
fractures (and proppant embedment).

Figure 6—DFIT diagnostic plots in the A, B and C wells demonstrating NFP complexity leading to a complex fracture
pattern (left) and an uncoupled stress and pore pressure due to tectonic stress (right). As depicted on the left for a
sample stage of the C well, the induced fracture pattern is rather complex with both horizontal and vertical fractures.

Image Logs
In unconventional reservoirs the hydraulic fracture can interact with pre-existing natural fractures and
stimulate shear slip on them. Although they can dehydrate the frac fluid and divert the energy to remote
areas from the wellbore, the natural fractures can also contribute to generating multiple complex fracture
sets and increase the contributing fracture surface area. Generally, if operators image the natural fractures
prior to the completion, they can adjust the treatment design accordingly. The oil-based micro-resistivity
electrical images of fractures and breakouts in a 150-m section of the D well are depicted in Fig.7, where
the natural fractures appear in darker colors. The azimuths are consistent with the Shmax direction. Notably,
the fact that there is a well-developed network that strikes sub-parallel to Shmax confirms that the induced
fractures should grow laterally, outward along the orientation of either the natural fractures, if re-activating
them, or of the Shmax, if inducing new ones. The natural fracture distribution is not uniform along the well
length, mostly due to heterogeneity in brittleness. Thus, investigating the microseismicity nearby the area
with a lower natural fracture density can help further reveal the role of natural fractures in controlling the
hydraulic fracture growth within the targeted units.
8 SPE-200021-MS

Figure 7—Electrical image of natural fractures and laminations detected in a 150-m section
of the D well. Orientations of fractures are represented by rose diagrams of fracture strikes.

Production data
The production data of the parent well (adjacent to the C well), together with almost two years of production
data of the four wells under consideration, are depicted in Fig.8 and Fig.9. The post-shut-in sharp increase
in WGR of the parent is associated with frac-hits from the child wells under consideration. Following a
continuous reduction in WGR, the parent recovers its productivity, potentially indicating that the observed
parent-child frac-hit is a communication frac-hit rather than fracture interference. In fact, in this project
shutting down the parent well before the treatment of the child wells has successfully minimized the parent-
child fracture interference. The productivity of the D well appears to be higher than the B and both perform
better than the A and C wells. The B and D wells are located between (and treated after) the C and A laterals.
The average WGRs of the wells are comparable and around 10 STB/MMSCF. Based on the McKetta and
Wehe plot about 0.8 bbl/MMSCF of the produced water is the gas water content. The increase in WGR (and
pressure) of the A well in Oct 2017 is possibly associated with frac-hit from an offset well and results in
a reduction in its productivity, which needs to be taken into account through the history-matching process.
Based on the cumulative production plots of the child and parent wells, the initial producing CGR is rather
constant and equal to 44 STB/MMSCF and considering the fact that the instantaneous producing CGR's of
the wells (and the DFIT estimates of initial pressure) are comparable, the initial reservoir pressure appears to
be distributed uniformly. The concave downward trend in the cumulative plots of the child wells represents
production in saturated condition where due to the condensate drop-out, oil remains in the reservoir and
the producing CGR decreases steadily. The cumulative CGR in the A well is affected by the higher-than-
expected condensate rates (unloading) in an eight-month period. It should be noted that, due to the flowback
of treatment fluids, the conversion from surface pressure data to BHP is subject to considerable uncertainty
during the first weeks of production.
SPE-200021-MS 9

Figure 8—Plots of gas production rate (left) and the logarithm of WGR (right) of the parent and four child wells under
consideration. Based on the trends in WGR (also confirmed by BHP), the parent and A wells both are possibly impacted
by frac-hits through their production lives, leading to a reduction, either permanent or temporary, in well productivity.

Figure 9—Plots of cumulative condensate production versus cumulative gas production (left) and bottom hole
pressure (right) of the parent and four child wells under consideration. The initial CGRs (also the instantaneous CGRs)
of the child wells are comparable which confirms that the initial reservoir pressure is distributed rather uniformly.

Two sample log-log FRI plots in the B and C wells are depicted in Fig.10, where the Pressure Normalized
Rate (PNR) is plotted versus the material balance time. The slope .25 indicates the early bilinear flow
through fractures which is followed by a linear flow through SRV (slope .5). Before the linear flow through
the Non-stimulated Regions Between Fractures (NSRBF), there is a pseudo steady state flow (BDF) which
helps us find the SRV volume. Two logarithmic and square root time derivatives also confirm the presence of
both bilinear and linear flow periods, where the bilinear flow would indicate that the fracture conductivity is
a finite value. Notably, in the Cartesian plot of RNP versus square root time of the A well, there is evidence of
a drop in the well productivity which might be caused by a frac-hit-induced change in the well productivity.
10 SPE-200021-MS

Figure 10—Log-log plot of PNR vs. material balance time in the B (top) and C wells (bottom). The bilinear
flow through fractures is followed by a linear flow through SRV, which in turn is followed by the PSS flow.

Microseismic Monitoring
Fig.11 illustrates the location of the microseismic events detected during the stimulation of the case study
via acquiring space- and time-variability of shear failures. A total number of 53000 events are located for
the four wells under consideration. The applied treatment design can be summarized as follows: 10000 ft
lateral length, 1000 lb/ft proppant, 15 bbl/ft fluid, 80 ft cluster spacing, 330 ft stage spacing, and 25 stages
per well, except the A well with 24 stages. The microseismic data are sampled at 4000 Hz using one 24-
level Whip-array® deployed in the D well to monitor the zipper treatment of the two leading C and A wells,
followed by one 24-level vertical array deployed in the C well to monitor the zipper treatment of the B and
D laterals. Due to the array configuration, only events with a high moment magnitude are recorded for the
toe-side stages of the D and B wells. The seismicity in the two inner wells appears to be effectively shielded
by the pressurized zones created by the outer wells resulting in the creation of taller well-controlled fracture
zones extending in the direction of the natural fractures and of the Shmax. The temporal pattern of events
is consistent with the diffusion of fracture branches around the early-time wet fractures that are mostly
associated with a high degree of deformation. The microseismic event clouds created around the wells are
overlapping and cover a large volume of the reservoir, where the entire space between the four wells has
been highly involved in the treatment. This potentially indicates different types of phenomena with either
positive or negative impacts: 1) well-to-well stress interference that enables/complicates fracture steering,
2) dehydrating frac fluid either via natural fractures or open unpropped fractures of the leading stages, 3)
creation of contributing, complex fracture branches that may promote the lateral and longitudinal uniformity
of the SRV, and 4) fracture overlaps which can directly affect the long-term productivity of the wells via
stimulating production competition. Assuming a comparable treatment design for the wells, the propagation
of energy in a smaller reservoir volume would indicate a higher density of fracture branches in the near-
wellbore area. Hence, considering the fact that the B and D wells develop slightly smaller event clouds,
SPE-200021-MS 11

the SRV in both of them is evidently larger than the A and C wells, and as it will be shown through the
production data analysis, those develop larger contributing fracture surface area.

Figure 11—Depth (top) and map (middle) views of event locations colored by well ID, and map
view of event locations colored by event date serial number (bottom). The treatment starts
with the toe-side stages of the C and A and ends with the heel-side stages of the D and B.

Fig.12 shows the magnitude of events plotted as a function of their source-receiver distance in the A
and C. The stages closer to the array tend to form more compact clusters of events, possibly due to the
improved resolution in locations closer to the detectors. In other words, the low magnitude events can be
detected closer to the array, whereas detectability drops off in distances away from the array. Here, through
the collective behavior analysis, a representative dataset with a minimum moment magnitude of −2.1 is
used to minimize the bias. As depicted by Fig.13-right, there are two peaks in the time histogram of events,
which respectively are associated with the pad and proppant stages. The induced seismicity is concentrated
in the upper part of the target formation (Fig.13-left). The upward event propagation suggests that a high
amount of energy has traveled into the upper unit. The upward fracture dilation is minimum for the B well,
and the downward growth is lowest for the C well. It should be noted that the recorded response is not solely
the result of shearing in newly created tensile fractures (hybrid failures) but also includes dry seismicity
together with the shearing and reactivation of the pre-existing natural fractures. The pre-existing natural
12 SPE-200021-MS

structures (faults and fractures) accessible by the growing induced fracture might dilate via experiencing
either inflation or shear failure, where the latter may lead to low-perm fractures with poor connectivity
(depending on the dilation angle and roughness). As confirmed by the DFIT, the compressional zone (the
non-contributing wet fracture) forms a large portion of the event cloud and must be distinguished from the
SRV.

Figure 12—Moment magnitude of events plotted as a function of their source-receiver

distance for the C and A wells. To minimize the bias, through the collective behavior
analysis the events with a moment magnitude smaller than −2.1 are screened out.

Figure 13—Depth (left) and elapsed time (right) histograms of the events recorded for the four wells under consideration.
The majority of the events are located between the mid-basal to mid-upper units happening during the pumping period.

Modeling Collective Behavior

The DPA entails collective behavior analysis of microseismic events and characterizes each event cluster
using four behavioral indices; Stress Index (SI), Plasticity Index (PI), Diffusion Index (DI), and Ternary
Index (TI). A high-PI cluster would indicate a cluster of events with a good level of deformation and rock
susceptibility that meets all the criteria required for an effective collective behavior: enough number of
events with proximity in both time and space domains and uniformity in attributes that concentrate the
energy and create fractures with a reasonable fracture growth rate. In instances where the degree of fracture
complexity drops (a widespread energy propagation pattern) the SI increases, and where the diffusion of
seismicity is not explainable in the context of fluid-driven activities, the DI is a large number (e.g., events
happening instantaneously). The TI as a linear combination of the three DPA indices is then used to prioritize
and sort the event clusters; a high-TI behavior translates into a high-PI, low-SI and low-DI activity and
indicates an irreversible anelastic deformation which possesses all the characteristics required to develop
a wet tensile fracture. When the number of high-quality, calibrated, and on-scale waveforms qualified for
Seismic Moment Tensor Inversion (SMTI) is not enough, the DPA provides estimates of failure planes,
giving the fracture propagation direction at each event location. In this section we discuss a series of results
generated by DPA, together with their application in evaluating the stimulation efficiency.
SPE-200021-MS 13

Temporal Ternary Diagrams

Since both the treatment parameters and the stress field are time-dependent and dynamic, some valuable
information can be obtained by modeling temporal dynamics of event clusters. In this regard, the DPA
ternary diagrams can be used to describe the time variability of collective behaviors, explainable in the
context of 1) the applied treatment design, 2) the recognized features of the host formation, and more
importantly 3) the underlying physics of the process. For instance, Fig.14 illustrates the normalized DPA
indices on a ternary diagram for four sample stages of the wells under consideration (hereafter called A1
to A4). The early-time clusters are depicted in large, blue symbols, while a red, small star denotes a late-
time event cluster. Normally, and as depicted for stages A2 and A4, with the start of stage treatment, the
high-PI event clusters tend to occur at increasing distance from the well and can be considered as the main
high-perm wet fractures. During the proppant placement, the density of high-DI, high-SI clusters tends to
increase further from the well as well as localized behind the extending wet fractures nearby the wellbore.
However, as the result of adding diverting agents to the frac fluid in stages A1 and A3, two rather different
late-time trends are observable. The late-time deformation pattern in A1 moves back towards the high-PI
area, which in this case indicates effective diverting and potentially creating new high-TI fractures, while in
A3, the high-DI late-time seismicity is intensified which would create a wider seismicity active region that
does not necessarily contribute to the fracture uniformity and complexity in the near-wellbore area (lower
stimulation efficiency).

Figure 14—DPA ternary diagrams of four sample stages of the case study. A different fracture temporal
evolution is observed in the A1 and A3 stages, showing the effective use of diverting agents in the A1.

Behavioral Fracture Maps

Although a temporal ternary diagram helps describe the behavior change over time, elaborating the spatial
location and connectivity of fracture in 3D is of greater interest and can be achieved by constructing
behavioral fracture maps. For instance, Fig.15 shows the results of DPA in the C well using depth and
14 SPE-200021-MS

map views of locations of cluster centroids and estimates of fracture planes. Along the majority of the well
length, fractures do not necessarily remain constrained to the top of the target unit and grow upward into
the immediate upper unit, except for stage 19 and 20 with a different fracturing regime probably due to a
strong effect of weak surfaces.

Figure 15—Well C event locations (a: map view, e: depth view), estimates of shear vectors colored by
stage (b: map view, f: depth view), estimates of fracture propagation directions colored by stage (c: map
view, g: depth view), and DPA cluster centroids sized by TI and colored by PI (d: map view, h: depth view).

The applied diverter has successfully contributed to increased fracture density in the near wellbore area.
However, in a limited number of stages, adding diverting agents and applying multiple proppant sweeps
have possibly contributed to the creation of a thinner fracture zone elongating towards the A well. For
instance, injecting clean fluid after the first proppant stage might help the fluid easily reach the fracture tips
and further extend the fracture length, potentially reducing hydraulic fracture nucleation in the critically-
stressed zone nearby the propped fractures. Similarly, the diverting agent may create partial blockage and
contribute to the increased off-target microseismicity. The fracture extension towards the A well via high-
PI, high-TI pathways may impact both the SRV volume and the contributing fracture surface area in the
A well; the farther the energy travels, the larger is the NSRBF. The mechanical interference caused by the
extensive lateral fracture growth might also impact the cluster efficiency and fracture uniformity in the
lagging well (Fig.16).
SPE-200021-MS 15

Figure 16—Estimated shear vectors in the two zipper frac wells of C (leading) and A(lagging). The
extensive lateral fracture growth in the C well might impact the stimulation efficiency in the lagging well.

The C well is the closest well to the parent, however, its fractures develop almost symmetric and there
is no evidence of fracture dilation towards the parent, which supports the idea that within the target area
the reservoir pressure profile is almost uniform, and the initial pressure is comparable between the four
wells under consideration. Furthermore, it confirms that shutting down the parent well before the treatment
of the child wells contributes to minimizing the parent-child fracture interference. The estimated fracture
propagation directions are consistent with the regional maximum horizontal stress direction (N40E) and
comparable with the azimuth of natural fractures imaged in the D well.
The heatmaps (Fig.17) of DPA indices are also generated to better study the spatial structure of the
stimulated zones. The DI and SI maps confirm the presence of a high-DI, high SI interval at the heel-side
which is created via elongated event clusters with large inter-event distances. The DPA in the two inner wells
shows a higher level of fracture complexity and connectivity in the near wellbore area with a more uniform
deformation pattern. As depicted in Fig.18 for four stages of the B well, the fracture propagation directions
confirm the creation of dual bi-wing fractures with an average distance of 75 m (260 ft) with horizontal and
vertical fracture branches in between. In instances in which more than two fractures are created, due to the
strong stress concentration, the inner fractures may experience only a limited dilation.
16 SPE-200021-MS

Figure 17—Well C heatmaps of PI (a), DI (b), SI (C) and TI (d). The heatmaps
help reveal the presence of non-stimulated zones between fractures (if any).

Figure 18—Map view of the estimated shear vectors (a) and map view (b) and depth view
(c) of the estimated fracture planes in four stages of the B well denoting the development
of two bi-wing planar fractures with horizontal and vertical fracture branches in between.

Stimulation Evaluation
Deformation Degree (DD) as another behavioral parameter is strain rate over stress and can be seen as
the resistance of flow of coseismic anelastic deformation. Based on the Shannon's entropy of the weighted
average of DD along the length of each stage, it is possible to derive an indicator to describe the fracture
uniformity, hereafter called the stimulation efficiency. The stimulation efficiency is also followed by an
alphabet indicating the fracture pattern, where H, T, C, E, and U, respectively, denote heel-ward, toe-
ward, center-squeezed, edge-ward, and uniform fracture patterns (Fig.19). The DD-derived stimulation
efficiencies are quantitative indicators of deformation uniformity (reservoir contact) and are affected by
myriad factors such as stage/cluster spacing, formation height (Li and Zhang 2017), fracture interference,
proppant pad design, and fracturing method. The advantage of the DPA-based stimulation efficiencies is
that unlike the in-well sensing techniques, such as fiber optics, that use the injectivity profile along the stage
length to estimate the cluster efficiency, it relies on the induced fracture pattern within the rock mass in
order to determine the fracture uniformity. Thus, it takes into account the coalescence of fractures.
SPE-200021-MS 17

Figure 19—Schematic of how entropy plots of weighted DD enable predicting the stimulation efficiencies for a
sample stage with 10 clusters. The stimulation efficiency describes the fracture pattern and quantifies its uniformity.

Predicting the stimulation efficiencies for six stages of the D well with variable stage/cluster spacings
and number of shots (Fig.20) demonstrates that although increasing stage and cluster spacing (or decreasing
the number of shots) appears to reduce the stress interference, there is an upper limit in which, for instance,
with further increasing the stage length, the non-uniformity increases, shrinking the reservoir contact
and resulting in a smaller SRV. For instance, in stage 19, increasing the stage spacing has resulted in
the development of two bi-wing fractures approximately 145 m apart from each other, leaving a wide
non-stimulated zone between them. The opposite is also true: a small stage length does not guarantee a
higher uniformity (e.g., stage 22) since it potentially intensifies the cluster-to-cluster interactions (stress
interference). The maximum efficiency is achieved at a stage length of about 110 m. Stages 21, 23 and
24 with nearly similar stage lengths and different cluster spacings create fractures with comparable cluster
efficiencies.
18 SPE-200021-MS

Figure 20—DPA clusters colored by stage and sized by TI (a), estimated shear vectors colored by stage (b), and
stage configurations and efficiencies (c) in six stages of the D well. The DPA-based stimulation efficiency describes
the uniformity in deformation degree along the stage length inversely weighted by the distance from the well.

Depleted Zone Microseismicity

In parent-child interactions, depending on parent distance and age (stress contrast), the fracturing fluid might
significantly divert towards the sink. Plotting the distance of centroids of DPA clusters (or events) from the
wellbore versus the square root time (Shapiro's plot) allows one to examine the creation of an apparent,
secondary diffusion front and investigate if the presence of a depleted zone impacts the fracture growth
(via transmitting strong pressure signals rapidly over considerable distances) (Dohmen et al. 2013). In other
words, when the fracture propagation front touches a depleted zone, the pressure front accelerates through
its low-pressure fluid-saturated pores and because of the higher pressure difference it induces failures
with a high deformation degree and creates a secondary fracturing front. The fractures created within the
drainage volume of parents, either propped or unpropped, most likely do not contribute to the productivity
of child wells, and often negatively impact the stimulation efficiency. Here, based on Shapiro's plots and
the behavioral fracture maps, the fractures induced during the treatment of the wells under consideration do
not appear to be interacting with the parent. As discussed, the frac-hit observed in the parent well is mainly
a communication frac-hit (rather than an interference frac-hit) and is caused by water invading its drainage
volume, complicating the multi-phase fluid flow regime through its SRV which temporarily hinders its
productivity (lowering the gas effective permeability).

Energy Radiation Pattern

Seismic rupture is controlled by the energy balance, and during each microearthquake three types of energy
(wave, frictional, and fracture) are involved. The latter two are called the non-radiated energy. The ratio of
the non-radiated (or radiated) energy to the total energy budget can be used to infer valuable information
regarding the fracturing process (Kanamori et al. 2001). In this regard, Beeler et al. (2003) based on the
expected relations between apparent stress and static stress release and using a standard energy balance
introduced a simple relationship between stress overshoot and the Savage-Wood seismic efficiency. The
overshoot itself is determined using the sliding and residual stresses, and it is a positive value if the static
stress drop is larger than the dynamic stress drop (Fig.21). A large stress overshoot may correspond to an
increase in the strain rate, a reduced rupture velocity, a lower seismic efficiency, a decreased slip speed,
SPE-200021-MS 19

and more energy consumption in deforming the medium (Walter and Brune 1993, Zecevic and Eaton 2017,
Beeler et al. 2012), and thus potentially denotes a mostly tensile or tensile-equivalent rupture. In extreme
dynamic weakening (low sliding stress), more energy is radiated, the shear resistance drops dramatically,
and the dynamic stress drop becomes comparable or even exceeds the static stress drop. This is often
characterized by an overshoot smaller than 0.2. Abrupt arresting of the rupture front may lead to rebuilding/
self-healing of the final stress through the stress undershoot (negative overshoot).

Figure 21—Schematic of the shearing stress evolution on a rupture plane during weakening, slipping and arresting phases.

When interacting with weak surfaces (natural fractures or interfaces), both the stress drop rate and the
energy budget components depend on the surface characteristics as well as the rock mechanical properties.
For undamaged rocks the non-radiated energy is expended for frictional and cohesive components, whereas
for fractured rocks it might be predominantly frictional (if reactivation is dominant).
The on-target fracture growth at the heel-side of the C well (particularly, stages 19 and 20), showed by
light color on the heatmap of overshoot in Fig.22, represents that a higher portion of the seismic energy
has been radiated and may suggest 1) the occurrence of unstable frictional slip on existing interfaces (or
fractures), 2) a higher fracture growth rate, 3) a reduction in the strain rate, 4) less energy consumption
in deforming the medium (more shear than tensile), and 5) ease to propagate the seismicity. In fact, for
the fracture tip to be arrested at the interface, the overburden stress must be close to Shmin, which is the
case in the target formation. Furthermore, the estimated fracture planes demonstrate fracture containment
and tell us that the fractures in this area are more horizontal and tilted than vertical, and mostly remain
constrained to the top of the target unit (shear displacement of weak interfaces). This indicates that there is
a weak interface that can blunt the fracture growth and initiate interface fractures. The same low PI, low-
overshoot, on-target activities nearby the top horizon of the target unit is common between all the four
wells under consideration within approximately the same location (the same two to three stages). Looking
at the density of imaged fractures in the D well (Fig.23) reveals a lower density of natural fractures in these
stages and indicates that the absence of natural fractures is a possible reason behind the intensified fracture
containment. Although there is another region with a low density of natural fracture at the toe-side, but as
depicted in Fig.23, in that location the well lands much closer to the top horizon supporting additional stress
to dominate the interface effects.
20 SPE-200021-MS

Figure 22—Overshoot heatmap in the C well (top) reveals an area with a different energy radiation pattern at the
heel-side. The estimated fracture planes (bottom) show that in this area fractures do not grow upward and remain
constrained to the top of the target unit which can be explained in the context of shear displacement of weak interfaces.

Figure 23—(top) Density of natural and drilling-induced fractures in the D well, showing two intervals with a lower
density of natural fractures. (bottom) Depth view of the C well together with the top horizon of the target unit.

Calibrating Reservoir Model

In this part, the production performance of the contributing fractures, given by the RTA, is coupled with
the detailed description of the fracture network, given by the DPA, to map the induced fracture and decide
which part of it is associated with the conductive and connected fracture. The modeling workflow entails
applying Four Straight-Line (4SL) techniques (two RTA plots and two microseismicity plots) to adjust the
SRV boundaries.
The linear flow analysis and the Flowing Material Balance (FMB) are the two widely-used RTA plots to
characterize MFHWs in unconventional reservoirs. The FMB (hereafter referred to as SL1) uses the well
SPE-200021-MS 21

production history and the reservoir fluid properties to determine the original hydrocarbon-in-place based
on the concept of BDF. Supported by a series of log-log and semi-log FRI plots, the SL1-estimated fluid-
in-place, which corresponds to the end of the linear flow on the linear flow analysis plot, belongs to the
contacted reservoir volume which in MFHWs is most likely associated with the SRV. Through the linear
flow analysis (SL2), the early-time production data indicates linear flow and can be diagnosed by a straight-
line on the graph of normalized pressure versus square root of material balance time or material balance
pseudo-time. The slope of the line gives the linear flow parameter (A√k), and the intercept captures the
near-wellbore effects (e.g., skin, fracture conductivity). In our previous work (Moradi and Angus 2020), in
a backward scheme, we conducted a series of numerical simulations in gas and oil wells with known SRVs
and used the RTA characterization plots to estimate the SRV volume. The results confirmed that using the
two FMB and the linear flow analysis plots shows a good sensitivity to the SRV volume, however, to avoid
misinterpretation of the first linear flow period it was recommended to use sub-daily production data at least
for the first two to eight weeks of production.
The SL3 is the Shapiro's plot (diffusivity plot), where assuming that in a poroelastic porous medium
the triggering front of the frac fluid-induced seismicity develops as a diffusive process, the plot is applied
to adjust the fracture hydraulic diffusivity. As discussed, the DPA ternary plot (SL4) describes the quality
of the collective behavior of event clusters using four behavioral indices. The fourth index (TI) is a linear
combination of the other three and represents a straight-line on the ternary diagram.
In particular, through the main SRV modeling loop, a volume matching process is accomplished to
distribute the RTA-derived SRV volume through areas with the highest degree of deformation:
1. The SRV volume is predicted using the SL1 (FMB)
2. A√k comes from the SL2 (linear flow analysis)
3. The triggering front is adjusted by the SL3 (Shapiro's plot)
4. The RTA-derived SRV volume is used as an additional limit to find the best TI on the ternary diagram
and extract the spatial SRV structure (volume-matching)
5. In case there are mute stages with not enough events (long distance from monitoring arrays or higher
rock rigidity), an iterative process is conducted: in step 4 when the smallest geometry that contains
all the high-TI clusters is determined, the average fracture dimensions are assigned to the mute stages
and then the total volume is predicted and compared with the FMB-derived volume, if they do not
match, the TI limit is updated
6. Based on the acquired SRV structure, the fracture area is determined and the SRV permeability is
propagated (Fig.24).
22 SPE-200021-MS

Figure 24—Two RTA and two microseismicity plots are integrated to

refine the spatial structure of the contributing fracture in the C well.

The integrated process ensures that the refined SRV boundaries honor the DPA-derived information
(contain fractures that are highly conductive, connective and hydraulically connected to well), and at the
same time the SRV model meets the RTA-derived criteria (see Moradi and Angus 2020 for more detail).
According to the results, the average height of the fractures that most likely contribute to the SRV of the
case study is about 206 ft (60 m), typically touched early during the stimulation. At this point the event
clusters associated with the SRV are designated, so the SRV model can then be constructed in different
formats. If there is enough microseismic data for all stages along the well length, the discrete fracture
network model can be extracted and used for direct numerical simulations through unstructured grids or
embedded structured systems. Or if the deterministic fracture does not cover all the stages, the location of
clusters can be used to set spatial limits on stochastic fracture models. However, the most computationally-
efficient approach would be to construct one or multiple Enhanced Fracture Regions (EFRs) on each stage
and then use Cartesian grids to conduct numerical simulations. Notably, constructing EFRs not only reduces
numerical complexity and results in an affordable computational cost, but also contributes to minimizing the
bias that can arise when fractures are modeled explicitly (e.g., correlating microseismicity to productivity
via constraining fracture surface area to fracture density). Finally, in order to apply analytical modeling, the
average dimensions can be used. In all these scenarios, the SRV volume and the linear flow parameter are
properly conserved, contributing to resolving the non-uniqueness of RTA solutions and providing a reliable
dataset for advanced fracture studies.

Validation of SRV Model

The integration of RTA and DPA has led us to an SRV volume which is not evenly distributed between
the stages. In other words, due to a variable stimulation efficiency, the fracture dimensions change from
stage to stage. Hence, in this section, we use the findings of fiber optics and an extensive mud gas dataset
to validate the SRV dimensions achieved for each stage and also to confirm that the contributing fractures
do not necessarily remain constrained to the top of the target unit (Fig.25).
SPE-200021-MS 23

Figure 25—(a) depth view of DPA clusters colored and sized by TI, (b) depth view of EFRs intersecting
the top horizon of the target unit, (c) map view of DPA clusters colored and sized by TI located in
the upper unit, and (d) map view of DPA clusters colored and sized by TI located in the target unit.

Production Profiling
In the D well, the available Distributed Acoustic Sensing (DAS) strain data gives information on the
contribution of each perforation cluster (and stage) to the treatment as well as to the production. The latter
entails passive DAS data acquired during two weeks of production, giving the contribution of each stage to
the well productivity. Fig.26 compares the DAS results with three other stage-contribution profiles acquired
using three different techniques:1) the DAS-based treatment monitoring that gives the mass of successfully-
placed proppant at each stage, 2) the DPA analysis of microseismic data that gives an estimate of deformation
level created by wet fractures at each stage (Type II), and 3) the DPA-RTA integration that refines the SRV
boundaries and gives an estimate of the contributing fracture at each stage (Type III).
24 SPE-200021-MS

Figure 26—Production profile acquired by DAS in well sensing in two weeks of production in the D well
(gray bars), together with the proppant mass and stage contribution profiles acquired by DPA and RTA-
DPA techniques. The DPA-RTA-derived production profile represents excellent agreement with the DAS
measurements. The gray arrow represents an increasing trend in the productivity of stages from 18 to 10, while
as denoted by the red arrow the proppant mass injected does not follow the same trend. Below the horizontal
black line, there is a limited number of microseismic events (increased distance from the monitoring arrays).

A series of conclusions concerning the capabilities of different techniques can be drawn:

1. The DAS indicates a widely varying production contribution profile along the well length.
2. A higher amount of injected (placed) proppant does not necessarily guarantee a better stage production
(stages 7, 9 and 20) and vice versa (stage 4).
3. The inferred trends from the placed proppant profile might be quite different than the productivity
of the fracture network (the arrows). This undermines the reliability of the conventional proppant-
loading approach to model the SRV.
4. There is an agreement between the DPA estimates of the deformation level at each stage and the
DAS-derived contribution of that stage to the production. However, because the Type II dimensions
describe wet fractures, those may overestimate the stage productivity. For instance, wet fractures may
include faraway plastic deformations that are created due to the transfer of frac fluid to distant areas
via induced and/or natural fracture corridors (intensified by stress interference). While the off-target
fractures are not propped and/or not connected to the well via a continuous propped path, hence most
likely do not contribute to the well productivity (stage 22, 20, 15).
SPE-200021-MS 25

5. DPA-RTA-derived share of each stage to the SRV volume (and to the fracture surface area) is
comparable with its DAS-based contribution to the well productivity, demonstrating that refining the
SRV boundaries by integrating RTA and microseismic analysis can successfully replicate the fracture
network behavior along the well length.
6. Due to its broad spatial coverage, the collective behavior analysis of microseismic event clouds gives
realistic insights into the cluster efficiencies, and its integration with volumetric estimates of RTA
provides a more realistic model of the contributing fracture network.

Compositional Fingerprinting
Producing and mud gas data of several thousand wells in the Montney formation has confirmed the
existence of at least two hydrodynamic units with different fluid systems (Euzen et al. 2019). This enabled
Clarkson et al. (2016) to use gas compositions obtained while drilling, together with the flowback gas
data, to successfully define a compositional marker and adjust fracture heights in a Montney case study.
As discussed, in the area of the case study, the lower Montney is more shaly, tighter and less permeable
than the upper units, and the hydrocarbon is most likely present as in-situ generated gas, increasingly drier
with depth. The available mud gas data entails minimal spatial averaging, therefore, the analysis of mud
logging data through the elbows of the four wells under consideration enables fingerprinting the fluid facies
with depth, hence one can track down the contributing reservoir height by comparing the producing gas
composition with the mud gas-derived patterns.
Since the tuned contributing fracture dimensions showed the occurrence of over-penetrating fractures,
the specific objective of this section is to compare the separator gas composition of the A well with the
fingerprints of the targeted formation and its upper unit. As depicted by Fig.27, the Haworth & Whittaker
wetness and the Pixler ratios suggest two distinct behaviors for the upper and lower Montney; increasing
wetness with depth in the upper Montney and decreasing density and wetness with depth in the Lower
Montney. It should be noted that the mud gas is fractionated and leaner than the reservoir gas, thus it
systematically shows dryer composition than the reservoir fluid. Besides, the effect of ROP on the mud gas
must be eliminated. Here, the mud logging data through laterals are used to extract the relationship between
ROP and fluid composition and remove the effect of ROP on the gas chromatograph results (Akihisa et al.
2018). Then we apply two extraction efficiency coefficients (tuned based on the data presented by Mole
2016) to correct the fractioning effect. The corrected C1/C2 and C1/C3 of the A well in the target unit are,
respectively, equal to 12.9 and 32.57, while those achieved from the separator gas sample are, respectively,
equal to 8.56 and 19.1. Therefore, the producing gas is wetter-than-expected which endorses the fact that
the A well fractures penetrate into the upper unit with lower C1/C2 ratios.
26 SPE-200021-MS

Figure 27—Mud gas data of the four wells under consideration can be used to confirm that the
contributing fracture produces from both the target and its upper unit in the Lower Montney.

Single-well Analysis
Analytical Modeling
In our previous work (Moradi and Angus 2019) we used behavioral fracture maps to demonstrate that the
induced fracture branches behave like a natural fracture network and can be modeled by covering them
with cuboids of an enhanced permeability (EFRs). By dividing the reservoir into nine linear flow regions,
we proposed a 3D analytical model, called the 9-grain model. The model was derived from the linearized
pressure diffusion equation by applying the Laplace transformation, as one of the most widely-used
technique to solve the fluid flow problems in porous media. To take into account the variable stimulation
efficiency, the model enables defining a variable fracture width. Furthermore, in order to deal with pressure-
and time-dependent well productivity variations, the model uses the superposition theory and allows to
modify the permeability of the reservoir, SRV and/or fractures. By validating the model with numerical
simulations, we highlighted the fact that in unconventional reservoirs with limited knowns (and few reliable
measurements) to support the numerically-added complexities, an analytical single-phase model would be a
better starting point than multi-phase multi-physics numerical models. In other words, numerical cases that
are constructed by adding more physics to analytical models (with step-by-step progress), can not be easily
replaced by other solutions because those reflect, meet and are empowered by the fundamental physics of
fluid flow identified by RTA characterization plots and validated by physics-based analytical models.
Following to the SRV modeling, the average fracture heights and lengths are used to construct and history-
match the analytical models of the wells. Fig.28 demonstrates the history-matched analytical model of the
C well. The single-phase model used to obtain the pressure transient response matches the well production
history with an SRV permeability multiplier in Aug 2018, which is consistent with the FRI findings. The
first five mute stages are synthesized as there is limited microseismic information available for them. The
matrix permeability is achieved equal to 4e-6 md (gas permeability), consistent with the results of the other
three wells.
SPE-200021-MS 27

Figure 28—Schematic of the C well single-phase physics-based analytical model and its history matching
results. The analytical models allows us to define a time- and/or pressure-dependent permeability.

Analytical Well spacing Optimization

To conduct an analytical well spacing optimization study, a 1-mile section is considered, and the following
steps are taken,
1. Using the history-matched model properties and the fracture dimensions achieved through the
integration, a new well is added to the section.
2. The well is put to production at BHP=800 psi for two 10-year and 30-year periods.
3. A new well is added to the section (via a random assignment of the acquired fracture dimensions to
the newly added well) and the no-flow boundaries are updated.
4. Step 2 is repeated and the total production from the section and from the wells are predicted.
5. Step 3 is repeated for multiple times, where well-to-well fracture overlaps decrease the EFRs length.
6. The optimum number of wells per section can then be achieved by applying a limit on either the
production reduction per well or the production enhancement per section.
Applying the optimization methodology to the case study confirms that the fracture half-length and
the matrix permeability are among the most important parameters that impact the optimum number of
wells per section. Furthermore, the time-period specified to predict the EUR greatly impacts the optimal
design; the optimum design for a 30-year period is not necessarily a good design in a 10-year period. For
instance, applying a 5% limit on the recovery decline per well, the optimum well spacing based on 10-
year calculations is achieved 20 percent less than the 30-year study. The analytical production forecasts
made for each scenario are then used for economic analysis to decide the optimal design which maximizes
productivity and net present value (NPV).
28 SPE-200021-MS

Numerical Modeling
To be able to history-match the multi-phase flow rates and adjust the absolute permeabilities and saturation
functions, the single-well numerical models are then constructed (Fig.29). The EUR estimates made using
the single-well numerical models for 10, 20 and 30 years of production are comparable (±2% difference) to
the predictions made by the analytical models of the wells. The D well appears to be the most productive
well between the four wells investigated (with the largest fracture surface area and SRV volume). According
to the results, the upward fracture growth may result in the partial depletion of the reservoir in the vertical
direction (particularly in the lower part of the target unit). However, the stages that penetrate into the upper
unit appear to be more productive than those with height-contained fractures. Furthermore, the fracture
permeability shows a significant impact on the pressure decline rate. In unconventional gas wells, due to
the high capillary pressure, the trapped liquid in the rock matrix nearby the fracture face hardly moves and/
or the fluid might be trapped within the fracture itself, which is often exacerbated by low drawdown, low
matrix permeability, and low gas rates.

Figure 29—Schematic of the SRV model (left) and the history matching results (right) in C well.
The top surfaces of the lower and middle units are depicted in grey and brown, respectively.

As discussed, the effective well length (as a function of the cluster efficiency) must be taken into account
in both analytical and numerical models. In unconventional wells, the SRV makes the reservoir and that's
why besides the fracture height and length, the cluster efficiency plays an important role in maximizing the
recovery. For instance, adding 10% to the EFR width in the C well, results in a 5-percent increase in the
10-year gas recovery at BHP=800 psi.

Empirical Modeling
Due to its ease of use, the Decline Curve Analysis (DCA) is still one of the widely-used techniques
to evaluate the long-term productivity of wells and sections. Here, in order to better adjust the decline
exponents within the study area, we compared the findings of the analytical and numerical results with the
predictions made using various DCA models, including the three modified hyperbolic, Duong (2010), and
logistic growth models.
According to the results, the Duong decline model provides more accurate reserve estimates. Joshi and
Lee (2013) suggest that to improve the accuracy of the Duong model, the line of best fit on q vs. t (a, m) plot
must pass through the origin (q∞=0). Based on our analysis, the SRV non-uniformity modifies the duration
of the linear flow period and elongates the transition from the linear flow to the BDF. Furthermore, we found
out that two more modifications help promote the predictive capability of the Duong model:
SPE-200021-MS 29

1. In q vs. t (a, m) plot, a ‘linear extrapolation through the origin’ (relying on the late-time trend) gives
more conservative and accurate estimates than building a ‘linear regression model passing through
the origin’. The extrapolation gives the best, latest estimate of the well productivity, which depending
on the available data may only include SRV, or both SRV and NSRBF, or a combination of SRV,
NSRBF, and the reservoir matrix.
2. A terminal decline rate between 5 to 7 percent improves the accuracy in long-term forecasting by
taking into account the transient-to-BDF (exponential) transition.

Multi-well Analysis
As depicted in Fig.30, the presence of NSRBF is more frequent in the C and A wells (the outer wells
that are completed earlier). The C well develops more fracture complexity than the A, possibly associated
with the observed variation in the rock brittleness in that direction. The A well shows no strong SRV
overlaps with its adjacent well. This is also confirmed by a communication frac-hit experienced by the
well during its production life that has not noticeably impacted the others. The highest level of deformation
and concentration of the energy is associated with the B and D wells, resulting in the development of a
complex high-TI fracture network between them. As it is identifiable on the TI map, the C is connected
to the D via toe-side and heel-side stages, while the stages in the middle show a lower level of overlap
between the two wells. The multi-well heatmap also reveals the presence of multiple areas associated with
a lower deformation degree common between two or more wells, potentially due to the presence of local
geological/structural lineaments. Notably, at the toe-side stages there is lack of microseismic monitoring
coverage (increased distance from the monitoring arrays).

Figure 30—DPA cluster centroids sized by TI and colored by PI (top) and the TI heatmap (bottom) including the A to D wells.
30 SPE-200021-MS

Based on the fracture dimensions and the model parameters acquired using the single-well models, a
sector model is constructed (Fig.31). The sector model by incorporating the DPA-RTA-derived fracture
dimensions, allows reservoir specialists to make more informed decisions about future productivity
restoration operations (e.g., re-fracturing or infill drilling). When compared to the single-well models, after
30 years of production at BHP=800 psi, all the wells produce marginally less gas and condensate, because the
no-flow boundaries between the wells (and stages) are not as uniform as the single-well models. However,
in short-term (10 years) the total production from both the B and D wells is improved by about 5 % which
confirms that the successfully created complex fracture between the two wells improves the productivity
of both (Fig.32). When compared to the C and A wells, in long term (30 years), the total production from
both the B and D wells shows less deviation from the single-well models, which confirms that the no-flow
boundaries between the C and D and between the A and B are closer to the outer wells (A and C wells).
As depicted in the horizontal cross-section of the pressure decline pattern after ten years (Fig.31 left), the
SRV spatial structure directly influences the sweep efficiency.

Figure 31—(left) horizontal cross-section of the pressure decline pattern after ten years of production confirms that
minimizing the non-stimulated zone between the stages is as important as improving the fracture length and height
since both directly promote the reservoir contact. (right) predicted pressure profile through the model in 2017 shows that
the SRV spatial structure directly influences the drainage volume dynamics in both horizontal and vertical directions.

Figure 32—Estimated gas flow streamlines in 2023; the complex fracture

created between B and D improves the productivity of both wells.

Summary
We applied collective behavior analysis to refine the event cloud and unravel the fracture behavior during
the treatment in a case study in the Montney Shale. It was demonstrated that the reactivation of natural
fractures has influenced the treatment by contributing to the creation of a complex fracture network in
both horizontal and vertical directions (between wells and stages). Most of the fractures were symmetric or
SPE-200021-MS 31

slightly asymmetric, and not impacted by the parent well (either vertical or horizontal). Fracture interference
was observed in several instances, to a large extent between the successively completed stages. We showed
that the entropy plots of the deformation degree give informed guidance for the optimization of treatment
design. The stage length appeared to be the most important factor in maximizing the reservoir contact by
having a direct impact on the fracture uniformity. It was shown that behavioral fracture maps give the
opportunity to evaluate the efficiency of innovative technologies such as diverting particles. The DPA-RTA
integration was then used to model the SRV and construct deterministic models of the contributing fracture
network. The modeled SRV streamlined the empirical, analytical and numerical modeling of the wells. The
analytical modeling was applied to adjust the matrix permeability, confirm the capability of the SRV model
in reproducing the production histories, and describe any productivity change and/or skin effects over time.
More complexity was then added, and single-well numerical models were constructed to model multi-phase
fluid flow, adjust the absolute and relative permeabilities, and make production forecasts. Finally, a sector
model was built to investigate how the drainage volumes (and SRVs) interact over time. The DAS fiber-optic
results and mud logging data were used to demonstrate the reliability of the DPA-RTA study undertaken.

Conclusions
The literature is full of case studies that divert attention to parent-child interference, while this work tried
to give other sensitive features of fracture growth, from child-child interference to fracture containment
and to stress shadowing. In light of the collective behavior analysis, a greater understanding was achieved
regarding the down-spacing potential, the efficiency of diverting agents, the role of natural fractures, and
the effects of stage/cluster spacing on fracture uniformity. The outcome emphasizes that integrating RTA
and advanced microseismic analysis provides many valuable insights regarding the induced fracture pattern,
thus it can be used to evaluate and improve the design of hydraulic fracturing and to constrain dynamic
reservoir models.

Nomenclature

WOR Water Oil Ratio STB/STB

WGR Water Gas Ratio bbl/MMSCF

GOR Gas Oil Ratio MMSCF/STB

Stimulated Reservoir
SRV -
Volume

Dynamic Parameter
DPA -
Analysis

PI Plasticity Index -

SI Stress Index -

DI Diffusion Index

DD Deformation Degree -

TI Ternary Index -

Flow Regime
FRI -
Identification

Enhanced Fracture
EFR -
Region

NFP Net Fracture Pressure psi (kPa)

32 SPE-200021-MS

Boundary Dominated
BDF -
Flow

DSI Dipole Sonic Imager -

CGR Condensate to Gas Ratio bbl/MMSCF

Instantaneous Shut-In
ISIP psi (kPa)
Pressure

Non-Stimulated Region
NSRBF -
Between Fractures

Rate-Normalized
RNP
Pressure

References
Agrawal, S., Sharma, M. M. (2018). Impact of pore pressure depletion on stress reorientation and its implications on the
growth of child well fractures. Unconventional Resources Technology Conference, pp. 31-42.
Akihisa, K., Knapp, L. J., Sekine, K., Akai, T., Uchida, S., Wood, J. M., Sanei, H. (2018). Integrating mud gas and cuttings
analyses to understand local CGR variation in the Montney tight gas reservoir. International Journal of Coal Geology,
197, 42-52.
Baig, A. M., Urbancic, T., Bosman, K., Ardakani, E., Thompson, J. M. (2017). Characterizing reservoir behavior with
cluster-based microseismic analysis. Unconventional Resources Technology Conference, pp. 3606-3617.
Beeler, N. M., Wong, T. F., Hickman, S. H. (2003). On the expected relationships among apparent stress, static stress drop,
effective shear fracture energy, and efficiency. Bulletin of the Seismological Society of America, 93(3), 1381-1389.
Beeler, N., Kilgore, B., McGarr, A., Fletcher, J., Evans, J., Baker, S. R. (2012). Observed source parameters for dynamic
rupture with non-uniform initial stress and relatively high fracture energy. Journal of Structural Geology, 38, 77-89.
Ben- Zion, Y. (2008). Collective behavior of earthquakes and faults: Continuum-discrete transitions, progressive
evolutionary changes, and different dynamic regimes. Reviews of Geophysics, 46 (4).
Clarkson, C. R., Ghaderi, S. M., Kanfar, M. S., Iwuoha, C. S., Pedersen, P. K., Nightingale, M., Mayer, B. (2016).
Estimation of fracture height growth in layered tight/shale gas reservoirs using flowback gas rates and compositions–
Part II: Field application in a liquid-rich tight reservoir. Journal of Natural Gas Science and Engineering, 36,
1031-1049.
Cornette, B. M., Telker, C., De La Pena, A. (2012). Refining discrete fracture networks with surface microseismic
mechanism inversion and mechanism-driven event location. SPE Hydraulic Fracturing Technology Conference.
Deglint, H. J., Ghanizadeh, A., DeBuhr, C., Clarkson, C. R., Wood, J. M. (2017). Comparison of micro-and macro-
wettability measurements for unconventional reservoirs: the devil is in the detail. Unconventional Resources
Technology Conference, pp. 2544-2552
Delorme, M., Daniel, J. M., Kada-Kloucha, C., Khvoenkova, N., Schueller, S., Souque, C. (2013). An efficient model to
simulate reservoir stimulation and induced microseismic events on 3D discrete fracture network for unconventional
reservoirs. Unconventional Resources Technology Conference, pp. 1433-1442.
Dohmen, T., Zhang, J., Li, C., Blangy, J. P., Simon, K. M., Valleau, D. N., Checkles, S. (2013). A new surveillance method
for delineation of depletion using microseismic, and its application to development of unconventional reservoirs. SPE
Annual Technical Conference and Exhibition.
Duong, A. N. (2010). An Unconventional Rate Decline Approach for Tight and Fracture-Dominated Gas Wells. Society
of Petroleum Engineers.
Fisher, M. K., Wright, C. A., Davidson, B. M., Goodwin, A. K., Fielder, E. O., Buckler, W. S., Steinsberger, N. P.
(2002). Integrating fracture mapping technologies to optimize stimulations in the Barnett Shale. SPE annual technical
conference and exhibition.
Ichikawa, M., Kurosawa, I., Uchida, S., Kato, A., Ito, Y., Takagi, S., Hara, S. (2019). Case study of hydraulic fracture
monitoring using low-frequency components of DAS data. SEG Technical Program, pp. 948-952.
Joshi, K., Lee, W. J. (2013). Comparison of various deterministic forecasting techniques in shale gas reservoirs. SPE
Hydraulic Fracturing Technology Conference.
Kanamori, H., Teisseyre, R., Majewski, E. (2001). Energy budget of earthquakes and seismic efficiency. Earthquake
Thermodynamics and Phase Transformations in the Earth's Interior, 76, 293-305.
Li, L., Sheng, J. J. (2017). Nanopore confinement effects on phase behavior and capillary pressure in a Wolfcamp shale
reservoir. Journal of the Taiwan Institute of Chemical Engineers, 78, 317-328.
SPE-200021-MS 33

Li, Y., Zhang, T. (2017). 14. Interaction of multiple hydraulic fractures from horizontal wells in tight gas shale reservoirs.
Revista de la Facultad de Ingeniería, 32 (10).
Manchanda, R., Bhardwaj, P., Hwang, J., Sharma, M. M. (2018). Parent-child fracture interference: explanation and
mitigation of child well underperformance. SPE Hydraulic Fracturing Technology Conference and Exhibition.
Manchanda, R., Sharma, M. M., Holzhauser, S. (2014). Time-dependent fracture-interference effects in pad wells. SPE
Production & Operations, 29(04), 274-287.
Moradi, P. M., Angus, D. (2019). Modeling frac-hits using dynamic microseismicity-constrained enhanced fracture
regions. Unconventional Resources Technology Conference, pp. 5129-5154.
Moradi, P., Angus, D. (2020). Unraveling Contributing Fractures Based upon Flowing Material Balance and Induced
Seismicity Analysis. In SPE/AAPG/SEG Unconventional Resources Technology Conference.
Morales, A., Zhang, K., Gakhar, K., Marongiu Porcu, M., Lee, D., Shan, D., Acock, A. (2016). Advanced modeling of
interwell fracturing interference: An Eagle Ford shale oil study-refracturing. SPE Hydraulic Fracturing Technology
Conference.
Neuhaus, C. W., Williams-Stroud, S. C., Remington, C., Barker, W., Blair, K., Neshyba, G., McCay, T. (2012). Integrated
microseismic monitoring for field optimization in the Marcellus Shale-A case study. SPE Canadian Unconventional
Resources Conference.
Nizkous, I, Moradi, P., Angus, D. Bosman, K. (2020) Coupling Microseismic and Production Data Analysis to Optimize
Well Spacing in Unconventional Reservoirs, SPE Unconventional Resources Conference
Pankaj, P., Shukla, P., Kavousi, P., Carr, T. (2018). Integrated well interference modeling reveals optimized well completion
and spacing in the Marcellus shale. SPE Annual Technical Conference and Exhibition.
Potocki, D. J. (2012). Understanding induced fracture complexity in different geological settings using DFIT net fracture
pressure. SPE Canadian Unconventional Resources Conference.
Ross, D. J., Bustin, R. M. (2009). The importance of shale composition and pore structure upon gas storage potential of
shale gas reservoirs. Marine and petroleum Geology, 26(6), 916-927.
Sekine, K., Singh, C., Hara, S. (2016). Fracture optimization study in the Montney tight sand play, Journal of the Japanese
Association for Petroleum Technology, 81 (6), 479-488.
Urbancic, T., Baig, A., Viegas, G., Thompson, J. M., Anderson, D., Rice, C., Martin, L. (2017). Effective constraint of
RTA models utilizing microseismicity derived flow attributes. Unconventional Resources Technology Conference, pp.
2411-2426.
Walter, W. R., Brune, J. N. (1993). Spectra of seismic radiation from a tensile crack. Journal of Geophysical Research:
Solid Earth, 98(B3), 4449-4459.
Warpinski, N. R., Lorenz, J. C., Branagan, P. T., Myal, F. R., Gall, B. L. (1993). Examination of a cored hydraulic fracture
in a deep gas well (includes associated papers 26302 and 26946). SPE Production & Facilities, 8(03), 150-158.
Wu, K., Olson, J. (2014). Mechanics analysis of interaction between hydraulic and natural fractures in shale reservoirs.
Unconventional Resources Technology Conference, pp. 1824-1841.
Zecevic, M., Eaton, D. W. (2017). Differentiating between shear and tensile events using spectral parameters.
Geoconvention.
Zoback, M. D., Kohli, A., Das, I., Mcclure, M. W. (2012). The importance of slow slip on faults during hydraulic fracturing
stimulation of shale gas reservoirs. SPE Americas Unconventional Resources Conference.