IPTC 17073

Explicit Modeling of Hydraulic Fracture Propagation in Fractured Shales

A.A. Savitski, Shell International Exploration and Production Inc.; M. Lin, Shell China Limited; A. Riahi, B.
Damjanac, Itasca Consulting Group Inc.; N.B. Nagel, Itasca Houston, Inc
Copyright 2013, International Petroleum Technology Conference

This paper was prepared for presentation at the International Petroleum Technology Conference held in Beijing, China, 2628 March 2013.

This paper was selected for presentation by an IPTC Programme Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the International Petroleum Technology Conference and are subject to correction by the author(s). The material, as presented, does not necessarily
reflect any position of the International Petroleum Technology Conference, its officers, or members. Papers presented at IPTC are subject to publication review by Sponsor Society
Committees of IPTC. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the International Petroleum Technology
Conference 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 where and by whom the paper was presented. Write Librarian, IPTC, P.O. Box 833836, Richardson, TX 75083-3836, U.S.A., fax +1-972-952-9435


Abstract

Recently a combined application of two technologies, horizontal drilling and multi-stage massive hydraulic fracturing (HF)
has made vast resources of shale gas commercially viable. The HF is a well-established reservoir stimulation technique,
which has been developed over the last half century. There are reliable tools for designing HF in conventional reservoirs, in
which a planar HF is assumed.

On the contrary, in shale gas fracturing, micro-seismic observations have illuminated a complex internal structure resulting
from the interaction of the induced hydraulic fractures with natural fractures. It is widely speculated that the stimulated
natural fractures make a significant contribution to the gas production.

The mechanics of the interaction between multiple fractures during a HF treatment is very complicated. In this paper we
present the results of state-of-the-art modeling of an explicit interaction between a propagating hydraulic fracture and a
statistically generated discrete fracture network. A sensitivity study reveals a number of interesting observations including
importance of initial fracture conductivity for growth of the fracture system, effect of the dilation angle on the generated
conductivity and net pressure and uneven distribution of fracture aperture that is critical for proppant placement.

This work strongly links the production technology and geomechanics and suggests an approach for modeling and designing
HF treatments in unconventional shale gas plays.

Introduction

Natural gas and light oil production from shale formations has become the focus of much attention in the past decade. The
economic success of the development of these unconventional resources owes to a combined application of two key and well-
known technologies: horizontal drilling and multi-stage hydraulic fracturing (King, 2010). The horizontal wellbore sections,
commonly with length ranging from 750 m to over 1600 m, are either cased and cemented or isolated with packers to enable
multi-stage fracture stimulation (Cipolla, 2009). Hydraulic fracturing (HF) is a well-established reservoir stimulation
technique, which has been developed over the last half century. It is not an exaggeration to say that in ultra-tight reservoirs
such as shales the connected reservoir conductivity is created in the process of hydraulic fracturing and that enables any
measurable hydrocarbon production.

The mathematical foundation and numerical tools for design of hydraulic fracturing in conventional (i.e., more permeable)
consolidated formations have been developed and successfully applied over the past few decades (see, e.g., Economides and
Nolte, 2000). One of the fundamental assumptions in conventional HF models is an assumption of the planar geometry of the
induced fracture. It is this assumption that comes under serious scrutiny in shale fracturing. There are different reasons for it
to be questioned. Firstly, the massive HF used in shale stimulation results in a large number of fractures propagating
simultaneously or sequentially. The mechanical interference between these fractures can lead, under certain conditions, to the
non-planar propagation trajectories as demonstrated by the results of numerical modeling (Bunger et al, 2012, and Kresse et
al, 2011) and experimental observations (Bunger et al, 2011). Secondly, non-planar induced fracture geometry may develop
as the result of the interaction with pre-existing fractures and frictional interfaces (Renshaw and Pollard, 1995, Gu et al,
2 IPTC 17073
2011, Weng et al, 2011). In this paper we will only consider the latter possibility an effect of the natural fractures on the
propagation of an induced hydraulic fracture. The focus of our work is on the numerical modeling of HF in fractured shale
formations.

The application of HF in Barnett shale, which sparked the so-called shale gas revolution, brought interesting observations.
The observation that the slick water jobs resulted in very productive wells was attributed to the stimulation of the natural
fractures (e.g., Cipolla et al, 2009a & 2009b; Warpinski et al, 2009). This explanation was further supported by
interpretations of the micro-seismic (MS) data (e.g., Mayerhofer et al, 2006 & 2010). The gel and slick water jobs seem to
produce qualitatively different MS patterns. As seen in Figure 1, the slick water re-fracturing resulted in a much wider special
distribution of the MS events. Unlike in conventional hydraulic fracturing, the slick water MS pattern does not clearly show
the orientation of a single planar hydraulic fracture and leads to the notion of the stimulated reservoir volume that contains
the MS event locations (Mayerhofer et al, 2010). However, the internal structure of this stimulated volume, which is
important for production, is not constrained and remains very uncertain. Hydraulic connectivity of complex fracturing was
confirmed when an injection into a Barnett shale well resulted in temporarily killing five of the surrounding six wells (Fisher
et al., 2004).


Figure 1: MS pattern for cross-linked gel and water HF (Cipolla et al, 2009a).

The concept of complex fracturing (e.g., Cipolla et al, 2009a or Weng et al, 2011) could explain the observed complex MS
patterns. Within this concept the fracture trajectory may deviate from the planar path at the frictional interfaces based on a
physical crossing criterion. Weng et al (2011) show examples of numerical modeling of such complex fracturing. The
generated final fracture system may be consistent with MS observations. What it might contradict is the measured treating
pressure. The fundamental characteristic of HF is that the aperture of the fracture depends not only on the local pressure but
also on the geometry (including size) of the fracture. In the case of unconstrained fracture propagation, it leads to declining
injection pressure under constant injection rate. In the case of complex hydraulic fracture propagation, the fracture is broken
into segments, which will result in smaller apertures and increasing treating pressure. This is not normally observed. While
we are far from claiming that complex fracturing is not happening in real HF treatments, we do stress the importance of
reconciling the predictions based on this conceptual model with the behavior of the net pressure (downhole pressure less the
minimum in-situ stress).

Another conceptual model does not require abandoning the assumption of planar fracture propagation as it shifts the
complexity from the process of fracture propagation to the process of fluid leak-off (see Figure 2). This model will be
further referred to as complex leak-off. Here we make a strong assumption that the induced fracture propagates in a plane
while leak-off of the fracturing fluid into the conductive natural fractures causes their shear and MS events. According to this
model, for the same natural fracture network a wider MS event distribution is expected in a less permeable formation and
when a lower viscosity fluid is injected. That is observed when comparing the MS data for tight sand versus shale gas and for
gel versus slick water treatments. Indeed, in such situations the high pressure front can propagate further away from the
induced hydraulic fracture. The ultra-low matrix permeability is essential for the complex leak-off model and is indeed the
common feature of all shale plays. While the final stimulated fracture networks in two conceptual models may look similar,
IPTC 17073 3
the physical processes leading to this complexity are different. Also, the proppant placement in complex-fracturing and
complex-leak-off models is very different.



Figure 2: Three conceptual models of HF in shales.

This paper reports the qualitative results of a sensitivity study for the complex leak-off model in which the focus is on the
interaction between a planar hydraulic fracture and a statistically generated discrete fracture network (DFN). We have
specifically modeled the effects of DFN connectivity, injection rate, in-situ stress anisotropy, fracture's dilation angle, and
shearing and extension of the natural fractures.

Numerical approach and simulation models

The numerical approach builds on previous work by Nagel et al (2011). 3DEC (Itasca, 2007) is a distinct element code that
explicitly treats a fractured rock mass as an assembly of blocks separated by interfaces. Blocks represent intact and
deformable rock material, while interfaces represent pre-existing discontinuous surfaces such as faults, joints and fractures.
Fluid flow is simulated in a network of flow planes defined by the interfaces between impermeable blocks and is fully
coupled with the mechanical calculations. Prior to building the model with a DFN we have successfully benchmarked this
numerical approach against a simple analytical solution for a constant-height fracture in a homogeneous medium and a
numerical solution for a three-layer system obtained with a commercial HF simulator.

The key part of the numerical model is the DFN. As shown in
Figure 3, the core part of the model containing the DFN is embedded into a larger intact domain. The outer intact domain is
deformable and serves for properly setting the boundary conditions (considering the finite-size model, which represents an
infinite domain). The linear dimensions of the full model are twice as large as those of the core part of the model. In this
study the model core had the dimensions of 2000600400 ft. Also, the DFN model core consists of two homogeneous
layers, each with a height of 200 ft.

Several DFN realizations were constructed statistically according to prescribed parameter distributions. Each layer has two
fracture sets. Greater fracture density is assigned in the upper formation. For the sake of numerical efficiency the orientation
of each fracture set was assumed to be deterministic (i.e., no variability in fracture orientation within a set). Two realizations
were selected for further simulations based on the connectivity of the DFN. Figure 4 shows the fully- and sparsely-connected
fracture systems. The sparsely-connected DFN was further used to build the base case model.

The intact rock is modeled as an elastic impermeable medium. Rock and fracture properties consistent with laboratory
measurements are used. In the base model, the natural fractures are initially conductive with aperture width of 0.1 mm. All
model parameters are based on one of Shells assets. The fluid is injected in the center of the model core at constant rate. The
hydraulic fracture plane is pre-defined to be perpendicular to the direction of the minimum horizontal in-situ stress. The
fracture height is also confined to the height of the DFN model core.

4 IPTC 17073


Figure 3: 3DEC base model with block geometry and zoning.



Figure 4: Full y-connected (top) and sparsely-connected (bottom) 3DEC DFN realizations. Clusters of connected fractures are
colored similarly.

Results: General observations

The results of the simulations for the base case model showed three common and interesting features. Figures 5 and 6
highlight the internal structure of the model.
Figure 5 shows pressure distribution in the induced and natural fractures. The vertical geometry of the fracture can be
inferred from Figure 6.

Firstly, as expected, the aperture of the induced hydraulic fracture is by an order of magnitude greater than the aperture of the
pressurized natural fractures. This is a compound effect of two factors. On one hand, unlike the natural fractures the induced
hydraulic fracture opens against the smaller stress. Although their conductivity increases with increasing pressure, the natural
fractures generally do not even open completely (i.e., the effective normal stress acting on them remains compressive). On
the other hand, the aperture of a completely open fracture depends on its size and the hydraulic fracture is much larger than
any natural fracture. Figure 7 compares the average apertures for induced and natural fractures. Even though we have not
modeled the proppant transport and placement, the consistently small apertures of the natural fractures indicate the slim
chances of placing proppant into them.

IPTC 17073 5


Figure 5: Fluid pressure distribution on a horizontal cross-section cutting through the injection point after 50 minutes of injection
for two DFN realizations.



Figure 6: Distribution of fracture aperture and fluid pressure in the plane of hydraulic fracture propagation.



Figure 7: Evolution of the average aperture for induced and natural fractures during injection for two DFN realizations.

Secondly, the small aperture of the natural fractures is somewhat compensated by their relatively large area compared to the
HF. The definition of the so-called 'stimulated fracture area' should be linked to the residual conductivity under production
conditions. However, the base case model does not allow us to estimate such irreversible effects, which is mainly due to
shear dilation and hysteretic joint stiffness. (The shear dilation was simulated as a special case and the corresponding results
are discussed below). Therefore, we have loosely defined the stimulated fracture area as those fractures with an increase in
either pressure or fracture aperture over specified thresholds. While this definition is not precise, it is fairly good for a
sensitivity study. All results presented in this paper are based on a pressure increase threshold of 1 MPa. Based on this
definition, for the simulated DFN realizations, the stimulated DFN area is several-fold greater than the hydraulic fracture area
(see Figure 8). This result strongly depends on the specific DFN geometry, connectivity and initial fracture conductivity.

6 IPTC 17073


Figure 8: Evolution of the hydraulic fracture area (right) and the ratio of the stimulated DFN area to it (left) based on the pressure
increase threshold of 1 MPa.

The third and last general observation from these simulations is that the distribution of the hydraulic fracture aperture is very
non-uniform, as shown in the upper plot in Figure 6. The geometry and extent of the hydraulic fracture can be inferred from
the contour of the high pressure shown at the bottom of this figure. The apparent mismatch between the contours of pressure
and apertures is clarified by Figure 9. The irregularities of fracture aperture are caused by shearing of the natural fractures
intersected by the hydraulic fracture. This shearing leads to local pinching and depression of the fracture aperture around
the intersection of fractures. This qualitative observation is important for the proppant transport. The HF design tools assume
a smooth aperture distribution, which may not always be the case. The presence of natural fractures may have an adverse
impact on proppant transport.



Figure 9: Close up view of the intersection of a hydraulic fracture with natural fractures. Contour plot shows the aperture
distribution on fracture surface. The black arrows show the local displacements along the hydraulic fracture.

Results of the sensitivity study

Effect of DFN connectivity
DFN is the most uncertain part of the model. Therefore, the primary interest in simulating the sensitivity to DFN
realization is to obtain a better understanding of what output is affected by it. Figure 10 shows the net pressure (injection
pressure less the minimum in-situ stress) and the leak-off ratio (the volume of fluid leaked into DFN divided by the total
injected volume) for fully- and sparsely-connected DFN realizations. While the net pressure is higher for the sparsely-
connected DFN, there is no specific pressure trend that would indicate the type of DFN system present. Contrary to this, the
leak-off ratio curves are qualitatively different. The leak-off ratio for the fully-connected DFN keeps increasing with time,
while for the sparsely-connected DFN it starts to noticeably decline after a certain injection time.

IPTC 17073 7


Figure 10: Comparison of net pressure and leak-off ratio for different DFN realizations.

Effect of injection rate
The injection rate for the base model was 20 bpm (0.053 m
3
/s). For evaluation of injection rate effects on the behavior of
the 3DEC models, the additional rates of 10 bpm and 40 bpm were also considered. We chose to plot the results versus the
injected volume rather than time, as it is an operating parameter. The results shown in Figure 11 indicate that as the injection
rate increased, the net pressure at the injection point expectedly increased as well. The lower injection rates resulted in a
lower fracture aperture and greater area of the stimulated DFN fractures at the cost of the reduced propagation of the HF. The
average aperture of the natural fractures (not shown here) is also lower for the case with smaller injection rate. However, the
impact of the lower rate on the conductivity the HF fracture is much stronger, which results in the higher leak-off ratio.



Figure 11: The results for different injection rates.

Effect of stress anisotropy
The potential of shearing or opening of the natural fractures strongly depends on the in-situ stress and specifically on the
stress anisotropy. In this work the natural fractures are vertical and the stress anisotropy was controlled by the maximum
horizontal stress. Figure 12 compares the results of the simulations for two extreme cases of a normal faulting stress regime
in which the maximum horizontal stress is equal to either minimum horizontal or vertical stress. The larger maximum
horizontal stress resulted in higher injection pressure with a minor difference in trend. However, it is unlikely that such a
trend can be identified in the field data to interpret the stress state. The leak-off ratio is consistently higher for lower
intermediate stress. This is a special case in which the net pressure is high enough to overcome the local in-situ stress and
completely open the natural fractures. The plot of the average DFN aperture is consistent with this interpretation.
8 IPTC 17073
Interestingly, the average aperture of the induced HF is not very sensitive to the stress ratio, despite the difference in the net
pressure.



Figure 12: Effect of the intermediate stress on the net pressure, leak-off ratio and fractures aperture.


Effect of the fracture dilation angle
In the base-case model, the natural fractures do not dilate when sheared. This assumption was made to simplify the
analysis of sensitivity to other parameters. The dilation angle is also a very uncertain parameter.
Figure 13 shows the results for three values of dilation angle. These high values were selected to exaggerate the effect of
dilation. The injection pressure shows diverging trends for dilatant and non-dilatant fractures. The dilation of the natural
fractures strongly increased the average aperture of the stimulated natural fractures, decreased the aperture of the hydraulic
fracture, and, consequently, significantly increased the leak-off ratio. As shown in
Figure 14, the shear dilation may have a significant contribution to the normal opening and, hence, hydraulic conductivity
of the natural fractures.

Effect of the evolution of the initial DFN
In all cases presented so far, the modeling was limited to a so-called static DFN. All fractures were initially conductive
and they could not propagate during the injection. Field observations support that in many shale resources there are relatively
few characterized natural fractures and those observed are often mineralized (Gale et al, 2007; King, 2010). This raises a
question of whether an initial relatively poor fracture system can develop during the stimulation into a more connected and
conductive DFN. We have split this question into two.

First, we evaluated the potential for shearing non-conductive fractures that represent mineralized fractures, or so-called
'planes of weakness'. It is hypothesized that the reopening of these fractures due to stress changes associated with fluid
injection can increase the connectivity of the DFN and enhance the extent of the affected area of the DFN. The dominating
mechanism considered is shear failure.
Figure 15 shows the fracture radius distribution in the base model. Following general geological observation that smaller
fractures are more likely to be mineralized, we have considered several cases in which all fractures below a given threshold
radius were assigned zero conductivity. The fraction of the total area of non-conductive, or glued, fractures that failed
during injection is shown in
Figure 16. The area of DFN glued in these simulations was 17% and 43% of the total area of DFN for the threshold radius of
15 m and 20 m, respectively. For each realization the sensitivity to the shear strength of the fractures was tested. The internal
friction angle of 40 degrees was kept constant but the cohesion varied from 2 MPa to basically zero.

IPTC 17073 9



Figure 13: Effect of dilation angle on net pressure, leak-off ratio, and fracture aperture.



Figure 14: Contribution of dilation to the fracture aperture.

Despite the large area of the original DFN that was glued in the model, only a small fraction of it failed due to fluid injection,
even for the zero-cohesion case. Comparing the results of the models with an equal area of mineralized fractures but different
strength characteristics suggests that the strength of the mineralized fractures was secondary compared to the initial
conductivity. In other terms, high-pressure fluid percolating through the DFN is the major cause of failure; the stress
perturbation alone accounts for only a small fraction of the failure. This conclusion is dependent on the whole set of input
parameters.

The second situation that we investigated was whether the natural fractures would fail and grow driven by the percolating
fluid. The base-case model was modified to allow for fracture propagation due to shear or tensile failure. Software
capabilities limited this extension to the initial planes of the natural fractures, thus making failure more difficult. The latter let
us to treat the results as conservative and qualitative.

The results for three sets of strength parameters are shown in
Figure 17. In the plot legends, C is the cohesion and T is the tensile strength. In the base case both cohesion and tensile
strength were set high enough to prevent any failure. However, as soon as the strength of the intact rock was reduced natural
fractures propagated, which strongly affected the output parameters shown in
Figure 17. In the most extreme case of weak rock, the natural fracture extension can be deduced from the net pressure, while
the leak-off ratio increases by as much as 80%. The assumption of a static DFN leads to a measureable under-estimation of
the stimulated DFN area and over-estimation of the HF aperture.
10 IPTC 17073


Figure 15: Fracture size distribution in the base-case model.



Figure 16: Effect of mineralization of natural fractures.



Figure 17: Effect of the propagation of the natural fractures on net pressure, HF aperture, leak-off ratio and stimulated NF area. In
plot legends C is the cohesion and T is the tensile strength.

IPTC 17073 11
Conclusions

In this work we have numerically studied the effect of hydro-mechanical interaction between a planar hydraulic fracture and
natural fractures. The developed numerical work flow is suitable for studying field cases on a relevant scale, though with
limited fluid properties and without the proppant transport. The developed numerical solution is not a substitute for the HF
design tools, but rather should be used to better understand the subsurface processes. Quantitatively, the leak-off rate
calculations can be used as input to proper HF design.

An extensive sensitivity study yielded the following important results:
While the aperture of the planar hydraulic fracture is much greater than that of the natural fractures, the total area
of the activated (pressurized) natural fractures can be very significant, making them relevant for the production.
In many cases significant leak-off into the DFN can be achieved even if the natural fractures do not completely
open (i.e., remain under compression).
Shearing of natural fractures that cross the hydraulic fracture results in very non-uniform aperture distribution
that can impact the proppant placement.
The DFN connectivity affects the results but does not cause a characteristic response that would allow one to
determine connectivity of the DFN from the field stimulation data.
Injection rate plays a major role in distributing the fluid between the hydraulic fracture and the DFN. Lower
injection rate increases the stimulated fracture network area, provided that there is a connected system of
conductive fractures. This happens at the cost of slower propagation of the main hydraulic fracture and its
reduced aperture, which may lead to a premature screen-out.
For the considered case of a conductive DFN, the lower the horizontal stress anisotropy the more the DFN is
stimulated, because the fractures open against a lower stress. This result is partially pre-determined by the
assumption of no shear dilatancy in the base case model. Indeed, the stress anisotropy would enhance the shear
dilatancy of natural fractures. More sensitivity simulations are recommended to better understand relative
contributions of these two counteracting effects.
While dilation angle is not easy to quantify, its impact on the results can be rather significant. The higher dilation
angle increases the leak-off ratio and suppresses the hydraulic fracture aperture.
For the selected set of input parameters we found that very few non-conductive fractures failed during fluid
injection. This suggests that the stress perturbation is not sufficient to stimulate natural fractures or planes of
weakness, and initial fracture conductivity is critically important.
However, the conductive fractures may further fail in shear or tension, thus expanding the initial DFN and
strongly affecting the results. This was observed for a set of model parameters used in this study.

Acknowledgments
The authors wish to thank Royal Dutch Shell for permission to publish this work and John Dudley for a thorough review of
the manuscript.

References

Bunger, A.P., J effrey, R.G., Kear, J ., and Zhang, X. Experimental investigation of the interaction among closely spaced hydraulic 1.
fractures. Presented at the 45th US Rock Mechanics/Geomechanics Symposium held in San Francisco, California, J une 26-29, 2011.

Bunger, A.P., Zhang, X., and Jeffrey, R.G. Parameters effecting the interaction among closely spaced hydraulic fractures. SPE 2.
J ournal, 17 (1), pp. 292-306, 2012.

Cipolla, C.L., Lolon, E.P., Erdie, J .C., and Tathed, V. Modeling well performance in shale-gas reservoirs. Presented at the SPE/EAGE 3.
Reservoir Characterization and Simulation Conference held in Abu Dhabi, UAE, October 19-21, 2009a.

Cipolla, C.L., Lolon, E.P., and Mayerhofer, M.J . Reservoir modeling and production evaluation in shale-gas reservoirs. Presented at 4.
the International Petroleum Technology Conference held in Doha, Qatar, 7-9 December 2009b.

Economides, M.J . and Nolte, K.G. Reservoir Stimulation. Third edition, Wiley, New York, 2000. 5.

Fisher, M.K., Heinze, J .R., Harris, C.D., McDavidson, B.M., Wright, C.A., and Dunn, K.P. Optimizing horizontal completion 6.
techniques in the Barnett shale using microseismic fracture mapping. Presented at the SPE Annual Technical Conference and
Exhibition held in Houston, Texas, September 16-19, 2004.

Gale, J .F.W., Reed, R.M., and Holder, J . Natural fractures in the Barnett Shale and their importance for hydraulic fracture 7.
treatments. AAPG Bulletin, 91 (4), pp. 603-622, 2007.

12 IPTC 17073
Gu, H., Weng, X., Lund, J ., Mack, M., Ganguly, U., and Suarez-Rivera, R. Hydraulic fracture crossing natural fracture at 8.
nonorthogonal angles: A criterion and its validation. SPE Production and Operations, 27 (1), pp. 20-26, 2011.

Itasca Consulting Group, Inc. 3DEC Three-Dimensional Distinct Element Code, Ver. 4.1. Minneapolis: ICG., 2007. 9.

King, G.E. Thirty years of gas shale fracturing: What have we learned? Presented at the SPE Annual Technical Conference and 10.
Exhibition held in Florence, Italy, September 19-22, 2010.

Kresse, O., Cohen, C., Weng, X., Wu, R., and Gu, H., 2011. Numerical modeling of hydraulic fracturing in naturally fractured 11.
formations. Presented at the 45th US Rock Mechanics/Geomechanics Symposium held in San Francisco, California, J une 26-29, 2011.

Mayerhofer, M.J ., Lolon, E.P., Youngblood, J .E., and Heinze, J .R. Integration of microseismic fracture mapping results with 12.
numerical fracture network production modeling in the Barnett shale. Presented at the SPE Annual Technical Conference and
Exhibition held in San Antonio, Texas, September 24-27, 2006.

Mayerhofer, M.J ., Lolon, E.P., Warpinski, N.R., Cipolla, C., Walser, D., and Rightmire, C.M. What is stimulated reservoir volume? 13.
SPE Production and Operations, 25 (1) pp. 89-98, 2010.

Nagel, N., Gil, I., Sanchez-Nagel, M., and Damjanac, B. Simulating hydraulic fracturing in real fractured rock - overcoming the limits 14.
of pdeudo-3D models. SPE paper 140480, presented at the SPE Hydraulic Fracturing Technology Conference held in the Woodlands,
Texas, J anuary 24-26, 2011.

Renshaw, C.E., and Pollard, D.D. An experimentally verified criterion for propagation across unbounded frictional interfaces in 15.
brittle, linear elastic materials. Int. J . of Rock Mechanics and Mining Sciences & Geomechanical Abstracts, 32 (3), pp. 237-249,
1995.

Warpinski, N.R., Mayerhofer, M.J ., Vincent, M.C., Cipolla, C.L., and Lolon, E.P. Stimulating unconventional reservoirs: maximizing 16.
network growth while optimizing fracture conductivity. J ournal of Canadian Petroleum Technology, 48 (10), pp. 39-51, 2009.

Weng, X., Kresse, O., Cohen, C., Wu, R., and Gu, H. Modeling of hydraulic-fracture-network propagation in a naturally fractured 17.
formation. SPE Production and Operations, 26 (4), pp. 368-380, 2011.