See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/254534807

Predicting Shale Reservoir Response to Stimulation: the Mallory 145 MultiWell

Project

Article · October 2011

DOI: 10.2118/145849-MS

CITATIONS READS
20 252

7 authors, including:

Daniel Moos Alfred Lacazette

independent Geothermal Technologies Inc.
143 PUBLICATIONS   4,243 CITATIONS    45 PUBLICATIONS   595 CITATIONS   

SEE PROFILE SEE PROFILE

George Vassilellis Randall Cade

Repsol Baker Hughes Incorporated
18 PUBLICATIONS   76 CITATIONS    11 PUBLICATIONS   59 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Mapping fractures using passive seismic View project

Sandcontol for waterinjection wells View project

All content following this page was uploaded by Alfred Lacazette on 14 December 2016.

The user has requested enhancement of the downloaded file.

SPE 145849

Predicting Shale Reservoir Response to Stimulation in the Upper Devonian

of West Virginia
Daniel Moos, SPE, G. Vassilellis, SPE, R. Cade, SPE, J. Franquet, Baker Hughes; A. Lacazette 1 , SPE, EQT
Production Company; E. Bourtembourg, G. Daniel, Magnitude SAS

Copyright 2011, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Denver, Colorado, USA, 30 October–2 November 2011.

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
A comprehensive site study carried out in an Upper Devonian shale gas reservoir at a site in southern West Virginia provided
data to test a geomechanical model for stimulation of the Huron formation. Using a model in which natural fractures provide
the primary conduits for production and are the major target for stimulation, and in which stimulation triggers shear slip on
those pre-existing fractures, we were able to predict the shape of the reservoir volume stimulated by injection of high-quality
foam and to match injection flow rates and pressures using a dual porosity dual permeability finite-difference flow simulator
with anisotropic, pressure-sensitive reservoir properties. The resulting calibrated model matched both the relative contribu-
tion of the individual stages measured by production logging and the early-life well production. This suggests that similar
models may in the future provide earlier and better production predictions, guidance for completion and stimulation design,
and recommendations to minimize production decline and maximize well value.

Introduction
A better understanding of the cause of the highly variable shale reservoir response to stimulation (e.g., Ketter et al., 2006) is
required for a number of reasons. In particular, a better understanding of what controls the shape (width, length and height)
and effectiveness (improved access to hydrocarbons and improvement in reservoir flow properties) of the stimulated volume
would help to guide selection of fluids, proppants, flow rates and volumes to maximize stimulation effectiveness. Knowing
the shape and dimensions of the stimulated rock volume would enable frac stage placement and well spacing that maximizes
recovery efficiency. This would also help to optimize well lengths and separations as well as help to determine whether stage
or well spacing should be constant, (a “geometrical” approach) or variable based for example on the variation in rock proper-
ties along the stimulated well. Better understanding of the stimulation process would also help to validate proposed practices
such as the alternating or simultaneous stimulation of adjacent wells or of stages along individual wells.
An improved understanding of the architecture of a stimulated horizontal well will make it possible not only to optimize
stimulation design but also to develop reservoir engineering tools (analytical or numerical) that will predict production per-
formance and match the shape of existing production trends. It would also make it possible to determine whether operational
controls such as surface pressure constraints (known as “choking”) would provide any long term productivity benefits in a
manner that would result in higher value. Another practice that has the potential to add significant value is to re-stimulate
wells that are in decline; decisions on when and how to restore productivity in older wells would benefit from an improved
understanding of the causes and effects of stimulation.
The industry so far has relied on performance analogs to improve understanding of each shale play. One particular ap-
proach is to use data mining (Thompson et al., 2010), however this requires a sufficiently long production history and careful
operational documentation. More important, data mining of known performance does not provide an understanding of the
reasons for performance variations, nor does it provide any information about the extent to which past procedures were op-
timal or how to change those procedures to improve results.
Founded on the idea that productivity enhancement due to stimulation results not just from creation of new hydraulic frac-
tures but also from the effect of the stimulation on pre-existing fractures (joints and small faults), we assembled an integrated
team to carry out a series of comprehensive studies of shale prospects. We looked at surface and downhole seismic, petro-
physical, microseismic, stimulation, and production data to test models based on this idea. We sought to determine the extent

1
Currently with NaturalFractures.com, LLC
2 SPE 145849

to which these data could be utilized in order to predict how reservoir properties change during stimulation, and over what
volume. As part of the effort, we used existing reservoir simulation tools to model the hysteresis of fracture flow properties
that results from microseismically-detectable shear slip, which is increasingly becoming recognized as key to the permanent
enhancement in flow properties and increased access to the reservoir that results from stimulation. A geomechanical model
for how stimulation affects reservoir properties that is based on the concept of critically stressed fractures (see Barton et al.,
1995) provided the unifying theme. Moos et al. (2008) and the references therein provide an overview of the shear stimula-
tion model.
Two projects eventually resulted; more are contemplated. The first was terminated after completion of geomechanical
modeling and analysis of microseismic data, when it became clear that data quality was insufficient to adequately test our
approach. This paper presents the principal results of the second project and provides an overview of the analyses that were
conducted to address the relation of microseismic observations to geomechanics and the insights that were obtained in de-
scribing the architecture of the stimulated rock volume. These insights were used to develop numerical simulation tools to
describe and prognose production performance. For practical publishing purposes, the numerical simulation, the petrophysi-
cal analysis and other observations related to the same data set are discussed separately (see Franquet et al., 2011a;b and Vas-
silellis et al., 2011; other papers are in preparation).

Project Background/history
The analyses and results discussed in this paper are based on data obtained in an Upper Devonian Huron site study which was
initiated by EQT Production Company to study various methods and practices to improve the economics of these assets. EQT
acquired a comprehensive dataset which they made available for analysis and interpretation by the study team. The approx-
imate site location is indicated by the star in the map of the northeastern US shown on the left-hand side of Fig. 1. The trajec-
tories of the project wells and of wells that were previously drilled nearby for which data were available and the locations of
microseismic events triggered by stimulation of the project wells are shown on the right. The overlying topography also
shown on the right is characterized by pronounced through going NW-SE and cross-cutting NE-SW trending topographic
lineaments.

Fig. 1—The Upper Devonian Huron project site is located in hilly terrain in southern West Virginia.

After drilling a vertical pilot well which was extensively logged, a series of four lateral wells were drilled along the ex-
pected NW-SE direction of Shmin, the least horizontal stress. This follows standard practice first carried out in the Barnett
shale of drilling horizontal wells in a direction such that stimulation will produce transverse hydraulic fractures. Fig. 2 shows
the stratigraphy, the wells, and indicates the completion style and stimulation characteristics for each well.
All of these wells were air-drilled within the zones of interest. A comprehensive suite of logs (including density, neutron
porosity, resistivity, and GR), as well as a nearly complete suite of state of the art measurements which included monopole
and crossed dipole acoustic logs, a neutron activation log, and electrical, acoustic, and optical image logs, was acquired in the
vertical pilot. The four lateral wells, three of which were drilled from separate wellheads and one of which was a sidetrack
from the pilot well, were also logged but with a less comprehensive logging suite. The vertical pilot was air-drilled but was
filled with water during acquisition of those logs (e.g. acoustic logs and acoustic image logs) that could not be obtained in an
air-filled well. Figure 3 is a plot of the standard logs that were obtained in the vertical well.
An extensive field-wide data set was collected as part of the project and will be the subject of future publications. Al-
though this publication focuses on the wellbore data collected during the study, it is appropriate to mention these data sets
here. Fracing, shut-in and flowback were observed continuously with a surface microseismic array. Inert gas (Mulkern et. al,
2010) and radioactive frac tracers were used on all fracs. Multiple chemical analyses of produced gases were collected for
the project wells. An extensive pressure monitoring program was carried out throughout the field and included build-ups,
interference testing, and gradient surveys. Core is available from a well drilled approximately 4 miles to the north.
SPE 145849 3

Fig. 2—This figure provides a schematic view at 1:1 (true vertical scale) of the wells which were
drilled during the project and the generalized stratigraphy. The wells are numbered top down from 1 to 4.
The larger numbers on the right correspond to the order in which the wells were stimulated.

Elements of the Upper Devonian Study Discussed in This Paper

The results discussed in this paper used the image and petrophysical (logging) data, the pumping data acquired during injec-
tion and during injection pre-tests, and the downhole microseismic data collected during well stimulations. The approach is
summarized in this section of the paper; each step in the process is discussed in more detail in subsequent sections.
We built a geomechanical model for the site using wellbore image data to detect and characterize wellbore failures (brea-
kouts, drilling-enhanced natural fractures and drilling-induced tensile fractures), core data to calibrate rock strength analyses
(the cores were acquired in an offset well), and logs. Pore pressure and closure stress data were obtained from PIFO (pump-in
fall-off) and pre-frac injection tests at three points in the section, and these were used to constrain smooth profiles of these
parameters. Together, these data were used to constrain the magnitudes of the in situ stresses at specific points along both the
vertical well and the laterals, based on matching the presence or absence and the characteristics of the various drilling-
induced failures. We then used an experimental approach to estimate the vertical variability of the horizontal stresses
throughout the reservoir section. We used these individual stress determinations supplemented and calibrated against ISIP
data collected during the stimulations. The image data were also used to detect and characterize pre-existing natural fractures
and faults.
Microseismic data collected using geophones installed in the #4 well during stimulation of the #1 and #3 wells were com-
prehensively quality-controlled and re-analyzed to improve locations and to determine the event magnitudes and (to the ex-
tent possible) the focal mechanisms (see Godano et al., 2010, for a discussion of the approach). The relative magnitudes of
the horizontal and vertical stresses required to explain the focal mechanisms were evaluated as a means to extrapolate the
well-centric geomechanical model outwards into the reservoir.
The characteristics of the microseismic data were then used to constrain a tensor hysteretic fracture permeability model
calibrated using pressure and injection rate data collected during stimulation of the #1 well. This was added to a commercial
reservoir flow simulator, and this in turn was used to model the shape and permeability of the stimulated rock volume for
selected frac stages. The results were compared to the measured contribution of each stage from a production log that was run
a few months after the well was brought on line, and the resulting full well model was then used to match and predict the
well’s short and long term performance. Vassilellis et al. (2011) discusses in detail the methods used to characterize the sti-
mulated volume and predict early-well-life production.
An advanced petrophysical analysis of the entire sequence was conducted using the full suite of log data (see Jacobi et al.,
2008 and LeCompte et al., 2009 for details of the approach). Advanced analyses were carried out on the acoustic logging data
to characterize the fractures, estimate permeability, and to evaluate the stress sensitivity and static elastic properties and
strength of the various formations (see Franquet et al., 2011).
4 SPE 145849

Fig. 3 shows the set of standard logs that were collected in the vertical pilot well. These standard logs provide sufficient
information to characterize the overall lithology and physical properties of the formations penetrated by this well. As ex-
pected in these well-indurated sedimentary rocks, acoustic velocities are high as is density. Neutron and density logs suggest
a small amount of bound water in clays within the shalier intervals, and two intervals (the B sand and the lower siltstone just
above 4500 feet) show indications of low clay volume and moderate gas-saturated porosity. Subsequent analysis of Stoneley-
wave properties revealed that these intervals had substantially higher permeabilities than the surrounding formations.

B Sand 

Lower siltstone

Fig. 3—This figure shows the well schematic, the lithologies, and the GR, density, neutron,
resistivity, and shear and compressional slowness logs collected in the vertical pilot well.

Natural Fractures Detected Using Image Logs

Three different types of image logs were run in the study wells. These included electrical images of the wellbore, acoustic
images of wellbore shape and reflectivity, and optical images obtained using a video camera. In general, the acoustic image is
better able to delineate natural fractures and breakouts, while the electrical image clearly delineates subtle variations in li-
thology. While natural fractures seen in the electrical image are usually more conductive, i.e., appear in darker colors, some
of the fractures seen in the acoustic image appear to be resistive (lighter in color) in the electrical image, indicating that they
are likely to be mineralized. Fig. 4 shows examples of these images obtained in the same 16-foot section of the vertical well
where cross-cutting fractures could be identified.
SPE 145849 5

Magenta 
sinusoids 
indicate 
fractures 

Fig. 4—Natural fractures detected in a 16-foot section of the acoustic image data obtained in the vertical well are shown on the left.
Many of the same fractures can be detected in the optical image shown in the center. Fractures are more difficult to see on the elec-
trical image and when they are visible often appear to be mineralized. Laminations are visible in all three images, but stand out most
clearly in the electrical images.

Natural fractures were detected and oriented using the acoustic image data in the vertical well. The optical images were
used in the laterals. The fracture orientations were determined by fitting flexible sinusoids to the fracture traces on the un-
wrapped images, as described in Barton, 2008. The orientations of fractures detected in a characteristic section of the vertical
well are shown in Fig. 5 as rose diagrams of the fracture strikes superimposed on a stereographic projection of the poles to
the fractures. The predominant NE-SW striking near-vertical joint set commonly observed in the northeastern United States
is clearly seen in these data. The accompanying NW-SE striking fracture set is not. This is surprising for two reasons. First,
we observed a subset of events triggered by stimulation which have focal mechanisms consistent with slip on near-vertical
NW-SE trending planes. Second, these fractures would explain the NW-SE trending surface topographic features. Sampling
bias due to the fact that the horizontal and vertical wells were both drilled nearly precisely parallel to these features could
explain this, however a few of these features might be expected to be detected in spite of this. Alternatively, it is possible that
their expression in the images is so subtle that they are undetectable, perhaps because they are entirely sealed and/or held
closed by high normal stresses. As discussed below these fractures are perpendicular to SHmax. Regardless, the near-total
absence of these planes in the images in spite of other evidence of their existence is puzzling.

Fig. 5—Orientations of fractures detected in a characteristic section of one well are represented here as a stereographic projection
of the fracture poles, overlain by a rose diagram of fracture strikes. Near-vertical NE-SW trending joints predominate, and some
fractures parallel the near-horizontal bedding. A second set of near-vertical joints trending NW-SE which is commonly seen in the
Marcellus was absent in this interval, however, normal faults with modest offsets striking NE-SW were also detected.

Fracture distribution was not homogeneous. At shallow depth and where lower volumes of organic materials were en-
countered, more natural fractures were detected in the quartz-rich (“brittle”) lithologies. These intervals might be expected to
have higher gas in place. Although the former observation may be a consequence of fractures being harder to detect in less
brittle rocks, fractures were detected immediately above the B sand (see the later discussion of Figure 7) in association with
elevated levels of organic materials, and deeper in the well, more fractures were also detected in the higher GR, more clay
and organic rich intervals. Taken together these are important observations not only because they provide clues to the origin
6 SPE 145849

of the fractures, but also because they suggest that in the lower permeability more clay-rich intervals, higher gas in place is
likely to be associated with a greater number of fractures which can be stimulated to provide access.

Stress-induced Wellbore Failure and the Azimuth and Relative Magnitude of SHmax
The most reliable method for determining in situ stress magnitudes and orientations is through observing the direct effects of
the stresses on the rock in situ (e.g., Zoback et al., 2003). The most obvious of these is the generation of stress-induced com-
pressional (breakout) and tensile (drilling-induced tensile wall fractures) failures of the rock surrounding a previously drilled
wellbore. The presence or absence of these features, and their orientations and characteristics, all provide information about
the in situ state of stress. Stress-induced wellbore failure is most easily detected in wellbore image logs.
Drilling-induced tensile fractures (DITWFs) are common both in the vertical and in the lateral sections of many shale gas
wells (the Barnett shale is typical). This is due not only to the state of in situ stress but also to the effects of overbalance,
cooling, and high weight on bit in these stiff, often quartz-rich rocks. No DITWFs were detected in these wells however, con-
sistent with the fact that they were air drilled, and in such cases these effects are less important or absent. In spite of the fact
that these wells were drilled underbalanced, breakouts (compressive shear failures) were rare. Breakouts detected in the
acoustic images from the vertical well occurred only within the deepest few hundred feet of the logged interval at azimuths of
132 and 312 degrees (+/-8 degrees). Since this well was drilled within a few degrees of vertical, and therefore the breakouts
are expected to form at the azimuth of the minimum horizontal stress, the inferred azimuth of the maximum horizontal stress
(SHmax) is 42+/-8 degrees.
Optical images were obtained in three of the four lateral wells. While the image quality and the ability to determine accu-
rate locations of the features detected in these images was not optimal for a number of reasons, it was possible to detect both
fractures and bedding, as discussed above. Importantly, stress-induced wellbore breakouts were detected at the sides of the
wellbore near the toe of one of the laterals and within the build section of Well #1.

Breakouts 
are visible 
here at the 
peaks and 
troughs of 
fractures 

Fig. 6—Breakouts can be seen clearly in the acoustic image shown on the left, and their presence is affected by fractures crossing
the wellbore, one of which has been highlighted using a magenta sinusoid. Laminations visible in the acoustic image are also readi-
ly apparent in the optical image in the center, and there are some indications of breakout. While horizontal laminations are easily
detected in the electrical image because of their large resistivity contrast, the breakouts visible in the acoustic image would not
have been be detected, nor is there any indication of the presence of the natural fracture.

To understand the implications on stress magnitudes of breakouts occurring on the sides of near-horizontal wells, it is im-
portant to recall that breakouts occur at the position around the well of the smallest far-field stress acting perpendicular to the
wellbore. For a horizontal well drilled in the direction of Shmin the two stresses acting on the well are the vertical stress (Sv),
which acts at the top and bottom of the well, and the maximum horizontal stress (SHmax), which acts on the sides of the
well. The presence of breakouts at the sides of these wells therefore requires that SHmax is less than Sv. The fact that brea-
kouts occur at the sides of Well #1 where it has a deviation of 78 degrees to the NW places further constraints on SHmax as
discussed below.

Computing the Orientation of SHmax Using Crossed Dipole Data

Over the last few decades the advent of acoustic logging tools that generate bending modes in the wellbore has enabled de-
termination of the near-well and far-field orientations of “fast” and “slow” shear velocities of dipole modes in near-vertical
wells. Azimuthal anisotropy can however be caused not just by differences in the horizontal stresses SHmax and Shmin but
also by the intrinsic structural anisotropy of the formations e.g. due to dipping shale beds or aligned steeply dipping com-
pliant fractures. While there are other methods to differentiate between these effects, the most direct way to identify the
source of azimuthal anisotropy is to compare the orientations of compliant structures such as steeply dipping bedding or frac-
tures with the bending mode orientations of the fast and slow dipole modes.
SPE 145849 7

Relatively few intervals were characterized by significant anisotropy in the vertical well, and those intervals that did show
elevated anisotropy also contained steeply dipping aligned fractures striking in the dipole bending mode fast orientation.
Therefore, their anisotropy is almost certainly structurally induced (one of these intervals is shown in Fig. 7 and discussed in
more detail below). Although there is a general trend of fast orientations in this well consistent with the orientation of the in
situ stress derived from breakout analysis (and with the strikes of the predominant joint set), the amount of anisotropy is very
small. There is no evidence in the B sand of differences in velocity as a function of azimuth. Overall these results are indica-
tions either that the formations encountered are relatively stress insensitive, as is certainly the case in the B sand and likely
given the high velocities recorded in most of these rocks, or that the magnitudes of the horizontal stresses are not significant-
ly different.

Advanced Petrophysics and Rock Mechanical Properties Analyses

The neutron activation log recorded in the vertical well was used as the basis for creation of a lithofacies model, as described
in Pemper et al. (2006). The first step in the process was determination of elemental volumes of 18 elements, using in addi-
tion to the pulsed neutron log a gamma log which differentiates the contribution of uranium from that of thorium and potas-
sium. These elemental volumes were then distributed among a characteristic set of minerals encountered in gas shales, result-
ing in a detailed model of the mineral percentages as a function of depth encountered by the vertical well. Ordinarily, NMR
data is used to determine porosity, but this data was not acquired. Also, results are usually calibrated against laboratory mea-
surements and thin section analyses carried out using core from the well; the absence of core was a limiting factor in this
analysis. The mineralogy for the interval surrounding the B sand is shown towards the center of Fig. 7. The B sand shows up
clearly as an interval with higher quartz and lower clay content than the surrounding zones.
Further analysis of the acoustic and petrophysical data allows determination of static physical properties using the LMP
(Log Mechanical Properties) algorithm (see Franquet et al., 2010) which consists of micromechanical modeling of rock
stress-strain curves from logging data. Additional results include determination from the acoustic logs of dynamic elastic
anisotropy, shown in the 6th and 7th tracks in Fig. 7 (see Franquet et al. 2011 for details of the methods used for these deter-
minations). The B sand is essentially isotropic, whereas the surrounding shales are anisotropic—their vertical Young’s mod-
ulus is higher and their vertical Poisson’s ratio is lower than in the horizontal direction. This will result in lower vertical than
horizontal elastic-wave velocities in the shales above and below the B sand. This is important information for analysis of the
microseismic data, because failure to account for elastic-wave velocity anisotropy can result in errors in event locations and
in the focal mechanisms. And, one might expect given the elastic anisotropy that rock strength is also anisotropic.
The different properties of these rocks can be used to estimate how easy it will be to create hydraulic fractures, and how
easily those fractures can propagate, vertically upwards and downwards, over this interval. This analysis is based primarily
on elastic properties. This is not discussed here, nor is an alternative method of obtaining a stress profile which is based on
GR and calibrated against the individual analyses of stress magnitudes is not discussed in this paper but provided similar re-
sults.
The 8th track in Fig. 7 shows hardness determined from elastic properties and mineralogy, which provides qualitative in-
formation to estimate the extent of proppant embedment. The B sand is harder and therefore will be less likely to have this
problem; the surrounding shales are somewhat less hard and so embedment might become a problem there. The reservoir
pressure is very low here and so the pressure drop due to production is not likely to cause large increases in the effective
stress acting on the walls of propped fractures.
Fig. 7 also shows in the 2nd through 4th tracks electrical and acoustic wellbore images and an acoustic anisotropy map de-
rived from the crossed dipole data. The B sand can be seen clearly as a high-resistivity (bright) interval in the statically nor-
malized electrical image shown on the left of these tracks. It appears less finely laminated than the surrounding intervals as
revealed in the adjacent dynamically normalized image. Thus it should not be surprising that its elastic anisotropy is low. The
most obvious feature in the acoustic image is the set of steeply dipping (approximately 80°) dark sinusoidal fracture traces
immediately above the B sand. These are associated with strong azimuthal anisotropy derived from the dipole data which is
shown in the adjacent track and was mentioned above. Given the fact that the fractures strike in the same direction as the fast
azimuth it is most likely that this anisotropy is due to their presence i.e. is structural, not stress-induced. The accompanying
high level of uranium indicated by the green-shaded region in the left-most track suggests that this interval is more organic-
rich than the surrounding intervals, thus perhaps these fractures may be filled with compliant organic material.
A radial profile of shear-wave velocity derived from the acoustic data is shown at the far right. The B sand is shown as a
high velocity (in the left-hand profile) unit which has little change in velocity with radial distance. The surrounding rocks
have variable velocity and their velocity varies somewhat with radial distance from the wellbore. The variation with radius
may reveal a degree of stress sensitivity which is not reflected in the very low azimuthal anisotropy seen in Track 4. Further
study is needed to resolve this apparently conflicting result.
8 SPE 145849

Fig. 7—This figure shows the results of analysis of logs and image data, to determine mineralogy and structural and physical prop-
erties, in a 500-foot vertical interval surrounding the B sand.

Building the Geomechanical Model

A geomechanical model includes knowledge of the magnitudes of the vertical stress and of the two horizontal principal
stresses, the larger of which is usually called SHmax and the smaller of which is usually called Shmin. It is important also to
constrain pore pressure and to identify the orientation of the maximum horizontal stress, and to determine static and dynamic
rock physical properties.
Overburden stress was computed by integrating formation density from the surface downwards. Here, reliable density da-
ta was acquired to within 700 feet of the surface, thus errors introduced during extrapolation of the data upwards were small.
Accurate measurements of pore pressure are extremely difficult to obtain in shale reservoirs due to their very low (often,
below 1 micro Darcy although the B sand and the siltstones have permeabilities which are several orders of magnitude high-
er) permeability. Well test analyses therefore can be used here to obtain approximate estimates of pore pressure based on the
long-term pressure recovery of wells into which fluid has been injected. These pump-in fall-off (PIFO) tests were conducted
prior to stimulating several of the laterals at the test site. The results indicate that the reservoir is significantly underpressured
(locally, the equivalent gradient is approximately 4.3 ppg but this, of course, changes with depth through the reservoir). The
B sand is depleted relative to the surrounding shales, which is not unexpected here due to a productive well drilled through
the B sand near the toe of that well, where the PIFO as conducted.
The least stress, which nearly always acts horizontally, can be measured during fracture stimulation either by conducting
a pre-stimulation minifrac or by monitoring fracture closure after completing each stage. The former is preferred, because
each stage introduces additional stresses on the formation, in particular when proppant is used. Thus, closure stresses are of-
ten observed to increase for each stage. Three values of instantaneous shut-in pressure were provided to us by EQT. These
were obtained from diagnostic minifracs conducted prior to stimulation in the lowermost two wells and in the #1 well, the
same wells for which pore pressure estimates were obtained from PIFO tests. The ISIP’s in the deeper wells were approx-
imately similar and indicated similar least principal stress gradients, however, ISIP was substantially lower in the #1 well.

Summary – Stress State and Natural Fractures

The foregoing analyses provide constraints on the stress state and information on the orientations of pre-existing natural
fractures. The results are summarized in Fig. 8. On the left is a cross-plot of the magnitude of Shmin vs. the magnitude of
SHmax. The magnitude of Sv is indicated by the open circle. The region shaded in color shows the range of stresses consis-
tent with the occurrence of the breakouts detected in the #1 well where it is deviated 78° to the NW, the orientation of Shmin.
In order for breakouts to occur on the sides of this well both Shmin and SHmax must be less than Sv; in contrast to the qua-
litative constraint expressed above this analysis now provides quantitative constraints on the horizontal stress magnitudes.
SPE 145849 9

Because the well is not yet horizontal in this interval, the stress acting on its top and bottom is a function of both the vertical
stress and Shmin. Thus, a lower magnitude of Shmin requires a lower magnitude of SHmax. Regardless, the stress state at
this site must be one in which Sv>SHmax>=Shmin, and the orientation of SHmax must be NE-SW.
We would like to extrapolate these data outwards from the near-wellbore region to create a 3-D model for the site, and
we'd also like to place a lower bound on SHmax, but aside from the fact that breakouts have a constant orientation in the ver-
tical well we do not from these data alone have any further information. In the next section we will discuss briefly how focal
mechanisms of events detected microseismically provide information to help in this process. Before doing so, however, it is
worth considering what the stresses and fracture patterns would suggest regarding the shape of the region affected by stimula-
tion.

Predicting the Shape of the Stimulated Region

There are two end-member models for the growth of a stimulated region in shales. The first model assumes that stress state
alone controls the shape of the stimulated region. This is based on the idea that lateral region growth away from the principal
hydrofrac plane requires creation of hydraulic fractures orthogonal not only to Shmin, as would be expected based on
minimizing the energy required to create and open a new fracture, but also orthogonal to the orientation of SHmax. Bsed on
this model stress states in which the two horizontal stresses are equal and much lower than the vertical stress could have
stimulated regions that grow significantly away from the hydrofrac plane, and cases in which there is a large difference in the
horizontal stresses could not. Based on this model, the relatively small difference between the two horizontal stresses at this
site should enable a broad zone of stimulation.

SHmax 

Shmin 

Fig. 8—On the left is shown the range of horizontal stress magnitudes that are consistent with the occurrence of wellbore breakouts
at the sides of the #1 well just above the B sand, where it is deviated approximately 78° to the NW. Shown on the right,
superimposed on the fracture strike directions shown in red, is a rose diagram of SHmax orientation inferred from the
orientations of breakouts detected in acoustic data in the vertical pilot well (in blue).

On the other hand, if zone growth requires exploiting a pre-existing set of natural fractures, both the stress and the natural
fracture population must be appropriate. A model in which elevated fluid pressure stimulates shear slip on pre-existing frac-
tures, causing a permanent increase in their permeability, requires a distribution of fracture orientations oblique to the orienta-
tions of the principal stresses (Sv, SHmax, and Shmin). The fact that there is a predominant well-developed near vertical joint
set which strikes sub-parallel to SHmax, and that although it is small there is a distinct difference in the two horizontal
stresses suggests that if this model is correct, stimulated zones should be narrow and grow laterally outward along the orien-
tation either of the predominant fractures (if pre-existing fractures control the growth of the stimulated region) or of SHmax
(if hydraulic fractures predominate).
In this paper we relate the shape and size of the region within which microseismic events were detected to that of the re-
gion which was stimulated and hence is expected to be productive. It is important to note however that while we have been
able to match not only stimulation and microseismic data but also early-stage production using this assumption, the presence
of a microseismic event does not necessarily guarantee that fluid has travelled to that point to trigger the event, and thus can
potentially flow back to the well to contribute to production.

Microseismic Results
Microseismic data was collected using a string of geophones placed in the second well from the top. The monitoring well was
cased and cemented to provide good wall contact to maximize signal to noise. A cement bond log showed generally good
cement bond quality throughout the well. The geophones were spaced along the central third of the well, which made loca-
tion of events within the array ends more precise than for events outside of the array. The raw microseismic data were re-
processed, starting by performing QA/QC on the raw data, building a local velocity model based on the acoustic logs, and
10 SPE 145849

calibrating that model to the extent possible using known events (these wells were all, with the exception of the monitor well,
completed using packer and port completions which employed sliding sleeves to access the frac intervals and packers to pro-
vide zonal isolation). The events were then auto-picked, the picks were validated by hand, and the events were then located
using a migration algorithm (for details of that algorithm see Bardainne et al., 2009).
Fig. 9 shows the locations of the events detected during stimulation of the #1 well. Very narrow zones were produced ex-
tending in approximately the direction of the joints and of the orientation of SHmax; height growth was reasonably well con-
trolled. Events located closer to the midpoint of the geophone array (shown in green) along the placement well (represented
by the red curve) tended to form much more compact clusters. This possibly is due to improved lateral location resolution for
those locations which would be expected based on their position relative to the detector array. The same analysis was carried
out for the deeper well (#3) and a more diffuse pattern and fewer events were detected. While it is not clear why the pattern
was different it could be that these events which tended in general to be smaller were less precisely located due to picking
errors.

Fig. 9—The positions of relocated events generated during stimulation of the #1 well are tightly clustered along a NE-SW
trend and fairly well confined vertically. Colors correspond to the stages during which the events were detected;
individual stimulation stages triggered events in more than one stage location along the well.

A subset of the located events which occurred during stimulation of the #1 well had large enough amplitudes and high
enough signal-to-noise ratios to make it possible to determine their focal mechanisms by waveform inversion (e.g., Godano et
al., 2011). However, the narrow aperture of the array required that the focal mechanisms be constrained to be pure double
couple, i.e., no opening or other motion was assumed. This can be a significant problem, although the degree to which it
caused errors in nodal plane orientations or focal mechanisms in this study is unclear.
Fig. 10 shows the event focal mechanisms in side view. These show a fairly wide scatter, but the best-located and con-
strained events, i.e. those that occurred within the lateral span of the detector array, have focal mechanisms which are consis-
tent with predominantly strike-slip motion on near-vertical planes striking sub-parallel to the joints and to the orientation of
SHmax. This implies that the stress state surrounding the MAL-145 wells is predominantly strike-slip, but this is misleading.
Focal mechanisms show the sense of motion of the event, which is largely constrained by the orientation of the plane on
which the event occurred. Inversion of events occurring on a narrow range of fault orientations does not provide a strong
constraint on the driving stress. Because there are few planes optimally oriented for normal faulting, i.e. which strike NE-SW
and have moderate dips, there is little possibility even in this normal faulting environment of large numbers of normal fault-
ing events. As can be seen however in Fig. 8, which compares the strikes of the predominant fracture set and the orientation
of SHmax, there is a slight offset which results in a small component of strike-slip shear on these near-vertical planes. When
these planes slip they will have a strike-slip sense of motion as was observed in the event pattern shown in Fig. 10. Thus,
while it is not possible to rule out a strike-slip stress state, the events are also consistent with the stress state derived from the
analysis discussed above. Thus the preferred model for the site is one in which both SHmax and Shmin are less than Sv and
there is a finite difference between the two horizontal stresses. Evaluation of the predominant sense of motion as a function
of stage number, which should reveal a change if previous stages cause changes in the stress state during later stages, showed
a relatively constant and time-invariant sense of motion.
SPE 145849 11

Fig. 10—A side view of event focal mechanisms for events occurring during stimulation of the #1 well shows for the best-
constrained events (i.e. those that lie between the ends of the array) predominantly strike-slip motion on near-vertical planes strik-
ing sub-parallel to SHmax and the principal strike direction of pre-existing joints.

Predicting Stage Contribution to Production Using Microseismic Activity

A production log run in the #1 well approximately one half year after initiating production showed widely varying contribu-
tions to production from the different stages. The contribution of any stage is likely to be the result of a complex interaction
of factors, which include the conductivity of the induced fracture network, the effective targeting of the gas productive inter-
val, the ability of the hydrocarbons to flow considering the presence of other fluids, and the loss of conductivity with pressure
drop and with the accompanying changes in the effective stresses. What is not well understood in relation to the microseismic
events is (1) whether the size of the cluster associated with each stimulation point represents the extent of stimulation, and (2)
whether the number of events (or the event density) that occurred within that cluster reflects stimulation effectiveness.
Fig. 11 compares the relative production contribution per stage both to the frequency of steep pre-existing fractures which
cross-cut the well within each interval and to the distribution of microseismicity. First, it is clear that there is a general trend
of decreasing contribution with distance from the heel of the well (which is towards the left). This is smaller however than
the variation between adjacent stages. That variation is most strongly correlated with the frequency of natural fractures that
intersect the well within each interval. The greatest number is found within the interval third from the left, whereas the fewest
is found within the interval third from the right. The centermost interval has a high density of fractures and is the third most
productive. There is no correlation whatsoever between the microseismic event frequency or density and the contribution of
the associated stage.
One additional observation that can be taken from the data shown in Fig. 11 is that the microseismic activity does not in-
itiate at the ports (red dots) which are the points of highest pressure during injection. Often, they initiate close to or at the
packers which separate the intervals. In other cases they initiate at locations where pre-existing natural fractures intersect the
well. Both of the latter behaviors are consistent with the mechanical response of the system to elevated fluid pressure. First, it
is well known and has been demonstrated in the oilfield that certain packer designs favor initiation of fractures at or imme-
diately behind the packers. This is not a good thing because it guarantees packer bypass. Second, it can be demonstrated that
the pressure required to create a new hydrofrac by inflating the wellbore can be significantly higher than the pressure required
to open a pre-existing transverse fracture, especially considering that these wells are drilled in the least compressive stress
direction. An observation that is consistent with initiation of breakdown by opening pre-existing fractures is that the trend of
the microseismic cloud is closer to the trend of the pre-existing fractures than it is to the orientation of the maximum stress,
as shown in Fig. 12. Errors in the velocity model which might be expected to result in errors in this trend are less important
when the monitoring well is centered beneath the stimulated well, and if present would be expected to cause the apparent
trajectories to “fan out” from the central trajectory. There is little evidence of either effect near the center of the array where
the trajectories are best defined e.g. for the events indicated in gold in Fig. 12.
12 SPE 145849

Fig. 11—Comparison between relative contribution to production per stage (center), fracture frequency (bottom) and the number of
detected events (top) reveals that the best predictor of stage contribution to production is fracture frequency.

Fig. 12—The orientation of the microseismic clouds, shown as dots, is closer to the orientation of the pre-existing natural fractures,
shown in red, than to the orientation of the principal stress, shown in blue. This suggests that the growth of the stimulated zone is
controlled by pre-existing structures. Notice also the clear alignment of the nodal planes of the strike-slip event which occurs over-
lying the gold microseismic cloud with the predominant fracture trend.

Calibrating a Reservoir Production Model

A key goal of this project was to evaluate the extent to which we could reproduce the creation of a productive volume in
shales using a “shear stimulation” model, to enable earlier and more accurate production prediction. At the present time no
commercial simulator can model this process, although some research simulators have been developed, of which a prime
example is the Japex SHIFT™ simulator (Tezuka et al., 2005). Therefore we had to improvise a workflow that would allow a
finite difference numerical formulation to account for anisotropic fracture permeability enhancement through fluid pressuri-
zation. The following briefly summarizes how the model was designed and calibrated; Vassilellis et al. (2011) provides de-
tails.
SPE 145849 13

The model is initialized with known shale characteristics e.g. matrix porosity and permeability, initial pressure, gas satu-
ration and organic content as described from core and log analysis. It implements an anisotropic dual permeability formula-
tion with initially compliant, low-conductivity natural fractures. A set of permeability-pressure tables is then used to describe
the hysteretic rock behavior that results from shear fracture activation, shown schematically in Fig. 13. The orientations of
the principal flow directions were chosen to correspond to the principal directions of the fracture sets and of bedding, which
in this case approximately corresponded to the principal stress directions which were the same along the entire well.

Figure 13: Example of hysteretic pressure-sensitive permeability model for directional flow

A key constraint on flow properties and the growth of the stimulated volume was the pressure and flow rate during injec-
tion of the known-viscosity fluid. Additional and independent constraints were the pressure at which microseismic activity
initiated, which provided a constraint on fracture strength properties, and the rate of growth of the region within which stimu-
lation occurred, which had to approximately match the growth of the microseismic cloud. The relative magnitudes of the
permeability enhancements in different directions were constrained by the geomechanical analysis and the shape of the mi-
croseismically active volume. The model is set at sufficient spatial and time step resolution to provide an accurate representa-
tion of the injection of fluids in a frac stage (10-20 minutes) and any immediate flowback. The result is an induced permea-
bility 3D map that describes the stimulated rock volume as discrete blocks each with a unique permeability. The stimulated
rock volume is therefore described not as a geometrical shape with identical flow properties throughout, but as a rock body
with variable induced permeability, as shown in Figure 14.

Figure 14: Example predicted induced permeability distribution.

A key additional constraint which is required to match post-stimulation flow properties and pressure sensitivity is the be-
havior during flow-back. Unfortunately, these wells were stimulated continuously starting with the first stage and proceeding
14 SPE 145849

to the last without dropping pressure, and flow-back occurred only after the last stage. Thus we had to use as an alternative
constraint the requirement to match a production log that was acquired 5 months into the life of the well and after the initial
clean-up. Using that data we were able to calibrate the model by adjusting only the properties of the natural fracture network,
which is consistent with our observations that natural fracture frequency was the primary control on relative stage contribu-
tion. Once we were satisfied that the model could reproduce the injection and relative production from each stage, we were
then ready to compile a full well model and predict production performance.
Two measures of how successful we were in modeling production are shown in Figs. 15 and 16. Fig. 15 reveals a very
good match between measured and predicted stage contribution to production at the time the production log was run which
would not have been possible if natural fractures were not important. The second measure of success is shown in Fig. 16
which compares more than 60 simulation runs of production vs. time. The variations in the curves are the maximum allowed
by the constraints placed on the reservoir parameters by the requirement to match stage contribution and stimulation beha-
vior. Without the prior knowledge gained by modeling the stimulation, there would have been no constraints on these what-
soever. If the model had been wrong there would have been no possibility to match the data.

Fig. 15—There is an excellent correlation between predictions of relative stage contribution, shown in red, and the measured contri-
butions one half year after initiation of production, shown in blue.

Fig. 15—This figure shows the effect of varying the remaining parameters in the full well simulator in order to match production.
Blue dots show surface pressure. The drop in production near the end of the recorded time period is due at least in part to the
change in wellhead pressure which was not considered in the production model.

Summary and Conclusions

The above brief overview summarizes the analyses and principal results of a study of stimulation response of the Huron for-
mation carried out at EQT’s Huron test site in southern West Virginia. The overall goal of the study was to test a model for
stimulation of a complex, fractured, mixed lithology (shale, siltstone and gas sand) reservoir in which the principal benefit is
the triggering of shear slip on pre-existing fractures (joints and small faults) which results in a permanent increase in their
conductivity. The results presented here are derived from petrophysical, geomechanical, and microseismic observations of
stimulations carried out in one of the four study wells, which was completed with a packers and ports style completion and
SPE 145849 15

frac’ed with a high-quality (99%) foam. The other wells in the study were also stimulated and data are also available from
those wells, but while nothing in that data conflicts with our results, detailed analysis is outside the scope of this paper.

The principal results of this study are as follows:

1. The site is characterized by very low reservoir pressure, accompanied by low (normal faulting) horizontal stresses;
the maximum horizontal stress is however greater than the minimum horizontal stress and is oriented NE-SW.
2. A well-developed, NE-SW striking near-vertical joint set parallels the orientation of the maximum horizontal stress.
Evidence for the presence included surface morphological features trending NW-SE and the presence in the earth-
quake focal mechanism results of events consistent with strike-slip motion on small NW-SE near-vertical faults but
these were not apparent in our analysis of the wellbore image data.
3. Trends of well-located events triggered by injection were sub-parallel but slightly offset from the orientation of
SHmax and almost identical to the orientation of the NE-SW striking joint set. This suggests that stimulation may
have re-opened pre-existing features rather than creating new hydro-fractures. Consistent with this hypothesis, some
of the event clouds appeared to nucleate at or near the intersections of NE-SW joints with the wellbore.
4. Several of the event clouds initiated at or close to the packers.
5. A production log acquired 5 months after the well was brought on production revealed a considerable variation in the
contribution to production of the various frac stages. Those stages accompanied by the greatest number of events
were not the most productive; the strongest correlation was between production and the number of fractures intersect-
ing the well within the stimulated interval.
6. A model for shear stimulation of pre-existing fractures implemented using a finite-difference dual porosity dual per-
meability reservoir simulator and calibrated using injection data constrained by the size and shape of the region of in-
duced microseismic activity not only matched the relative contribution to production seen in the production log but
also matched the first several months of full-well production data without violating the constraints on the model pa-
rameters required to match the injection data.

A number of conclusions can be derived from the above observations. First, it appears at least to first order that at this site
the size and shape of the microseismic event cloud triggered by injection is in fact related to the actual stimulated volume, but
it is important to note that this does not allow direct prediction of production at this site. The combined observation that
events tend to initiate either near packers or at locations with large numbers of pre-existing fractures suggests that efficient
reservoir access would be achieved taking both of these tendencies into account. For example, stage separation along the
wellbore should perhaps be designed to access an equal number of fractures or fracture clusters per stage rather than an equal
length of wellbore. And, as the microseismic clouds did not overlap nor did the apparent stimulated volumes, additional pro-
duction might be achieved here by reducing stage length. As fractures tend to be the predominant flow paths, care should be
taken to target those which are gas-bearing rather than those which are potential sources of water.
Flow properties of the stimulated reservoir are extremely pressure-sensitive, based on our calibrated model. Thus over-
production might be expected to result in premature production decline; care should be taken to maintain near-wellbore re-
servoir pressure, or, an evaluation should be made of the economic benefit of re-stimulation.
Two recommendations for data acquisition to improve well design are (1) to acquire NMR data to determine effective po-
rosity independent of standard log analyses, and (2) to employ continuous downhole pressure and flow monitoring equipment
both to remove the ambiguity associated with computing down hole pressure from surface measurements and to evaluate
changes in relative stage contribution over the productive life of the well to guide interventions should those be deemed eco-
nomic.
The above results and conclusions were made based on a model which included the fact that the interval through which
the well was drilled is orders of magnitude more permeable than the surrounding more shaley intervals and has significant
porosity. While it is difficult to isolate the contributions of fractures which accessed gas from outside that interval from the
gas that was produced from within it, we believe that much of the production did in fact come through the fractures, based on
the correlation between fractures and production revealed in the production log. Finally, while we were able to develop a
workflow to model injection and production, the approach has several shortcomings, compared to one in which the simulator
incorporates the actual physics controlling the process.

Acknowledgements
We would like to acknowledge EQT for providing us with the data from their Upper Devonian site study and for permission
to publish this paper, and BHI’s internal support to carry out the work.

References
Bardainne, T., E. Gaucher, F. Cerda, and D. Drapeau, 2009, Comparison of picking-based and waveform-based location methods of micro-
seismic events: Application to a fracturing job, SEG Expanded Abstracts 28, 1547-1551
Barton, C., M. D. Zoback, and D. Moos, 1988. Fluid flow along potentially active faults in crystalline rock, Geology, 23(8), 683–686
Barton, C.A., 2008, GMI•Imager™ User’s Manual, Version 5.4, November 2008, ©GeoMechanics International, Houston, TX, 191 pp
 16 SPE 145849

Franquet, J., Wolfe, C., and Rodriguez, E., 2010, Micromechanical Modeling of Rock Stress-Strain Curves from Log Data, paper pre-
sented at the 1st International Conference on Integrated Petroleum Engineering and Geosciences (ICIPEG 2010), a conference of the
World Engineering, Science & Technology Congress ESTCON 2010. Kuala Lumpur, Malaysia, 15-17 June
Franquet, J., et al., 2011a, Integrated Acoustic, Mineralogy, and Geomechanics Characterization of the Huron Shale, Southern West Virgin-
ia, USA, CSUG/SPE-148411, for presentation at CSUG/SPE Unconventional Resources Conference, Calgary, Alberta, Canada, 15 -
17 November
Franquet, J., Patterson, D., and Moos, D., 2011b, Advanced Dipole Borehole Acoustic Processing – Rock Physics and Geomechanics Ap-
plications, for presentation at the 2011 SEG International Exposition and 81st Annual Meeting held in San Antonio, Texas, USA, 18-
23 September
Godano, M., T. Bardainne, M. Regnier and A. Deschamps A., 2011, Moment-Tensor Determination by Nonlinear Inversion of Amplitudes,
Bulletin of the Seismological Society of America, Feb 2011; 101: 366 – 378
Jacobi, D., et al., 2008, Integrated Petrophysical Evaluation of Shale Gas Reservoirs, SPE-114925 presented at the CIPC/SPE Gas Tech-
nology Symposium Joint Conference held in Calgary, Alberta, Canada, 16-19 June
Ketter A.A., J.L. Daniels, J.R. Heinze, and G. Waters, 2006, A Field Study Optimizing Completion Strategies for Fracture Initiation in
Barnett Shale Horizontal Wells, SPE 103232, presented at SPE Annual Technical Conference and Exhibition, San Antonio, Texas,
USA 24-27 September
LeCompte, B., J.A. Franquet, D. Jacobi, 2009, Evaluation of Haynesville Shale Vertical Well Completions with a Mineralogy Based Ap-
proach to Reservoir Geomechanics, SPE-124227 presented at SPE Annual Technical Conference and Exhibition, New Orleans, Loui-
siana, USA, 4–7 October
Moos, D. and C.A. Barton, 2008, Modeling uncertainty in the permeability of stress-sensitive fractures, ARMA 2008, San Francisco, CA,
July 1-3, 2008
Mulkern, M. M. Asadi, and S. McCallum, 2010, Fracture Extent and Zonal Communication Evaluation Using Chemical Gas Tracers: SPE
138877 presented at SPE Eastern Regional Meeting, Morgantown, West Virginia, USA, 12-14 October
Pemper R., et al., 2006, A New Pulse Neutron Sonde for Derivation of Formation Lithology and Mineralogy, SPE-102770 presented at the
2006 SPE Annual Techncial Conference and Exhibition held in San Antonio, Texas, USA, 24-27 September.
Tezuka, K., T. Tamagawa, and K. Watanabe, 2005, Numerical simulation of hydraulic shearing in fractured reservoirs, Presented at the
World Geothermal Congress, Antala, Turkey, 24-29 April
Thompson, J.W., L. Fan, D. Grant, R.B. Martin, K.T. Kanneganti, and G.J. Lindsay, 2010, An Overview of Horizontal Well Completions
in the Haynesville Shale, CSUG/SPE 136875, presented at the 2010 CSUG/SPE Unconventional Resources and International Petro-
leum Conference held in Calgary, Alberta, Canada, 19-21 October
Vassilellis, G., C. Li, R. Seager, and D. Moos, 2010, Investigating the Expected Long-Term Production Performance of Shale Reservoirs,
SPE 138134, presented at the Canadian Unconventional Resources & International Petroleum Conference held in Calgary, Alberta,
Canada, 19–21 October
Vassilellis, G., et al., 2011, Shale Engineering Application: The MAL-145 Project in West Virginia, CSUG/SPE-146912, for presentation
at the 2011 CSUG/SPE Unconventional Resources Conference held in Calgary, Alberta, Canada, 15 - 17 November
Zoback, M.D., Barton, C.A., Brudy, M.O., Castillo, D.A., Grollimund, B.R., Finkbeiner, T., Moos, D.B., Peska, P., Ward, C.D., and Wi-
prut, D.J., 2003, Determination of Stress Orientation and Magnitude in Deep Wells, Intl. J. Rock Mechs, 40, 1049-1076

View publication stats