The role of preexisting fractures and faults during multistagehydraulic fracturing in the Bakken Formation

Yi Yang1 and Mark D. Zoback1

Abstract

We performed an integrated study of multistage hydraulic fracture stimulation of two parallel horizontalwells in the Bakken Formation in the Williston Basin, North Dakota. There are three distinct parts of this study:development of a geomechanical model for the study area, interpretation of multiarray downhole recordings ofmicroseismic events, and interpretation of hydraulic fracturing data in a geomechanical context. We estimatedthe current stress state to be characterized by an NF/SS regime, with SHmax oriented approximately N45°E. Themicroseismic events were recorded in six vertical observation wells during hydraulic fracturing of parallel wellsX and Z with three unusual aspects. First, rather than occurring in proximity to the stages being pressurized,many of the events occurred along the length of well Y, a parallel well located between wells X and Z that hadbeen in production for approximately 2.5 years at the time X and Z were stimulated. Second, relatively fewfracturing stages were associated with an elongated cloud of events trending in the direction of SHmax aswas commonly observed during hydraulic fracturing. Instead, the microseismic events in a few stages appearedto trend approximately N75°E, approximately 30° from the direction of SHmax. Earthquake focal plane mech-anisms confirmed slip on faults with this orientation. Finally, the microseismic events were clustered at twodistinct depths: one near the depth of the well being pressurized in the Middle Bakken Formation and the otherapproximately 800 ft above in the Mission Canyon Formation. We proposed that steeply dipping N75°E strikingfaults with a combination of normal and strike-slip movement were being stimulated during hydraulic fracturingand provided conduits for pore pressure to be transmitted to the overlaying formations. We tested a simplegeomechanical analysis to illustrate how this occurred in the context of the stress field, pore pressure, anddepletion in the vicinity of well Y.

IntroductionThe Mississippian-Devonian Bakken Formation is a

restricted, shallow-water, mixed carbonate-clastic se-quence deposited over most of the deep part of the Wil-liston Basin (Gerhard et al., 1990). Using horizontaldrilling and multistage hydraulic fracturing led to thesuccessful development of the Elm Coulee (Montana)and Parshall (North Dakota) fields and demonstratedthe great potential of the Bakken Formation. Despitethe generally successful exploitation of the Bakken For-mation, questions remain about how to optimize hy-draulic fracturing and the importance of preexistingfractures and faults as fluid pathways in the reservoir.

The study area consists of three horizontal produc-tion wells (X, Y, and Z) and six vertical monitoring wells(A-F) (Figure 1). The three approximately 10;000‐ ft-long horizontal wells are located in the Middle Bakken.The well spacing is approximately 500 ft. Our study fo-cuses on stimulation of wells X and Z. The middle well,well Y, had been hydraulically fractured previously and

was in production for about 2.5 years prior to stimula-tion of wells X and Z.

Well-log data including gamma ray, electrical resis-tivity, density, and P- and S-wave sonic velocities areavailable from geophysical logs in vertical wells A, B,D, E, and F. Well logs from vertical well A are shownin Figure 2. The total thickness of the Bakken Forma-tion is approximately 140 ft in the study area, with thetop of the reservoir at a depth of approximately10,000 ft. Wells X, Y, and Z were drilled in the MiddleBakken. As shown in Figure 2, the Upper, Middle,and Lower Bakken members can be easily distin-guished from gamma-ray logs. The high gamma rayand high resistivity indicate the oil-saturated, organic-rich shale layer in the Upper and Lower Bakken, andthe low gamma ray and low resistivity in the Middlemember is an indication of the target formation com-prised principally of a dolomitic siltstone. The Upperand Lower Bakken shales are also characterized bylower density (approximately 2.3 g∕cm3) and lower

1Stanford University, Department of Geophysics, Stanford, California, USA. E-mail: alecyang@stanford.edu; zoback@stanford.edu.Manuscript received by the Editor 7 October 2013; revised manuscript received 15 December 2013; published online 21 May 2014; corrected

version published online 4 June 2014.This paper appears in Interpretation, Vol. 2, No. 3 (August 2014); p. SG25–SG39, 20 FIGS., 3 TABLES.http://dx.doi.org/10.1190/INT-2013-0158.1. © 2014 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved.

t

Special section: Microseismic monitoring

Interpretation / August 2014 SG25Interpretation / August 2014 SG25

VP and VS values, which can be attributed to lithology(clay- and organic-rich shale) and fluid-saturationdifferences.

In the sections below, we first report the develop-ment of a geomechanical model that includes knowl-edge of the magnitude and orientation of principalstresses, the existence and orientation of natural frac-tures and faults, and the mechanical properties of theformations being produced. Knowledge of the stressfield can be used to better understand microseismicevents and fracture networks resulting from hydraulicfracture stimulation (Fehler et al., 1987; Phillips et al.,1998; Rutledge et al., 1998; Maxwell et al., 2002). There-fore, linking observations of microseismic data withgeologic and geophysical information can help us bet-ter understand the geomechanical properties of the

Bakken Formation and the role of preexisting fracturesand faults on the effectiveness of hydraulic fracturingstimulation in this reservoir.

Geomechanics of the study areaTo build a quantitative geomechanical model, we fol-

low the general procedure outlined in Zoback (2007) toconstrain the orientation and magnitudes of the threeprincipal stresses in the reservoir. The magnitude ofvertical principal stress was determined from theweight of the overburden, whereas estimates of porepressure (Pp) were available from diagnostic fractureinjection test (DFIT) data, which provide instantaneousshut-in pressure (ISIP) to constrain the magnitude ofShmin. Physical properties were determined by commer-cial laboratory tests on Bakken core samples from well

A. The values are listed in Table 1.Unfortunately, there are no image

logs at this site to determine the orien-tation of the maximum compressivestress from the orientation of compres-sive and/or drilling-induced tensile well-bore failures. Previous studies close tothe study area report the orientationof SHmax to be approximately N45-55°E, based on the DITFs observed inFormation MicroImager (FMI) logs ob-tained from three horizontal wells (Ol-sen et al., 2009; Sturm and Gomez,2009). Using the physical property mea-surements from the core samples, aswell as estimates of SV , Shmin, and Ppfrom this study, we used the methodol-ogy outlined in Peška and Zoback (1995)to analyze the occurrence of tensile frac-tures in the wells studied by Sturm andGomez (2009) as a function of SHmax ori-entation and magnitude (see Appen-dix A). This modeling indicates thatwhen SHmax is oriented toward N45-55°E, transverse tensile fractures thatare quite similar to those reported bySturm and Gomez (2009) will occur.However, the magnitude of SHmax isnot well constrained by the modeling.The stress state could either be a normalfaulting or a strike-slip faulting regime.

Microseismic events duringhydraulic stimulation

As shown in Figure 1, microseismicevents during hydraulic fracturing ofwells X and Z were monitored with ap-proximately 1900-ft-long, 40-level, 3Cgeophone arrays in six vertical obser-vation wells (A-F). The spacing betweenseismometers in each array was 49.2 ft.Hydraulic fracturing was first performedalong well X from toe to heel and

Figure 1. Entire microseismic events and the well geometry in this study, in-cluding three horizontal wells (X, Y, and Z), and six vertical observation wells(A-F). Events are colored by stages. (a) Map view with the arrows representsstimulation sequence along wells X and Z, (b) north–south cross view with geo-logic formations and geophone locations along each vertical observation well,and (c) histogram of number of events for each stage.

2 4 6 8

−200

0

200

400

600

800

1000

VP (km/s)

Rel

ativ

e de

pth

(ft)

1 2 3 4

VS (km/s)

0 500 1000

Gamma ray (GAPI)

2 2.5 3

Density (g/cm3)

0 2000 4000 6000

Resistivity (ohm-m)

Mission Canyon

Lodgepole

Upper BakkenMiddle BakkenLower Bakken

Three forks

Figure 2. Petrophysical log collected from well A showing the lithologic profileacross the Bakken Formation.

SG26 Interpretation / August 2014

consisted of 29 stages, followed by another 38 stagesalong well Z. For both wells, different hydraulic fractur-ing methods were used along the lengths of the wells:Stages 1–18 (well X) and stages 31–49 (well Z) used aball-activated sliding sleeve with swell-packers; stages19–29 (well X) and stages 50–67 (well Z) used pump-down perforation guns with bridge plugs. Both stimula-tions used a hybrid fluid system (slickwater, linear gel,and crosslinked gel), sand, and ceramic proppants. Thecumulative fluid volumes injected during stimulation inwells X and Z were approximately 72 and 84 Mbbls, re-spectively. Note that the middle well Y was hydraulicfractured approximately 2.5 years prior to the fractur-ing along wells X and Z in this study; however, it was notmicroseismically monitored at the time of fracturing.

A total of 6499 treatment-induced microseismicevents were detected and located by the contractor(Figure 1a and 1b). Event histograms of each stage (Fig-ure 1c) show that the number of events in several stages(i.e., stages 2–4, stages 50–53) is much higher than inother stages. Events from six selected stages (i.e.,stages 2–5, 43, and 50) were reprocessed by a secondcontractor. Both initial and reprocessed hypocenter lo-cations show unusual characteristics.

Events along wellbore YThe first unusual observation of the microseismicity

is that many of the events occur along well Y duringpressurization of stages in wells X and Z. Figure 3 dem-onstrates the spatial-temporal evolution of events instage 4. A few events occurred near the perforationlocation in well X as expected at the beginning ofinjection; however, after approximately20 min, the great majority of events oc-curred along the length of well Y, thou-sands of feet away from the perforationlocation near the toe of well X.

The pressure records of wells X and Yduring stage 4 show that the occurrenceof microseismic events along well Y wasassociated with increased pressure inwell Y starting about 20 min after pres-surization of well X (Figure 3d). In thebeginning of stage 4 when the eventsclustered close to the perforation loca-tion, the pressure in well Y remainedconstant at approximately 2213 psidownhole pressure. When many eventsbegan to occur thousands of feet south-ward along well Y, the fluid pressure be-gan to rapidly increase in well Y. Thesudden increase in pressure in well Yindicates that fluid pressure from pres-surization of well X was being observedin well Y, presumably via fluid pathwaysthrough preexisting fractures and faults.Dohmen et al. (2013) argued that the mi-croseismic events along well Y are dis-tributed throughout the depletion zone

surrounding this original production well. As explainedbelow, our analysis suggests that microseismic eventsalong well Y do not necessarily depend on priordepletion. In other words, it does not matter if the res-ervoir was at initial or depleted conditions, these micro-seismic events would occur during hydraulic fracturingas long as there are preexisting fracture and faults inthe area.

Events cluster at multiple depthsAs illustrated in Figure 1, another unusual observa-

tion is that the microseismic events occur at two dis-tinct depths: one clustered within or near the Middle

Table 1. List of parameters values of geomechanicalproperties and sources at Middle Bakken.

ParametersValues at Middle

Bakken Source

Sv gradient 1.05 psi∕ft Integration of densitylog

Shmin gradient 0.79 − 0.85 psi∕ft DFIT data

Pp gradient 0.66 − 0.7 psi∕ft DFIT data

Young’smodulus

3.77 � 106 psi Rock mechanics test

Poisson’s ratio 0.21–0.31 Rock mechanics test

UCS 16 � 103 psi Rock mechanics test

Internalfriction

0.8 Rock mechanics test

Sliding friction 0.6 Common estimate

A

B

C

D

E

F

Y ZX

Bakken events

Pressure (psi)

0 10Time since injection in stage 4 (min)

2000

4000

6000

8000

10000

P in well Y

injection P in well X

12000

0 10 20

20 30 40 50 60 70 80 90 100

30 40 50 60 70 80 90 100

Colorbar

a) b)

c)

d)

0

500

1000

1500

2000

2500

3000

C A B D FE

Cross view

Rel

ativ

e de

pth

(ft)

Stage 4

Stage 4

Pre

ssur

e (p

si)

Time since injection in stage 4 (min)

0 1000 ft

N

N

Figure 3. (a) Chronicle progression of microseismic events and correspondingpressure recording of stage 4 during well X stimulation. The triangle symbolsrepresent perforation location. (a) Map view of microseismic events, (b)north–south cross view of microseismic events, (c) colormap of events chronicledistribution in (a) and (b), and (d) injection pressure of X (red line) and down-hole pressure record of Y (black line) of stage 4.

Interpretation / August 2014 SG27

Bakken Formation and the other clustered at a muchshallower depth approximately 800 ft above the welldepth in the Mission Canyon (MC) Formation. In fact,there are more microseismic events located in the MCFormation than in the Bakken. Figure 4 shows that thestages associated with the largest number of events arethe stages with many events in the MC Formation. Inparticular, see stages 2, 3, and 5 near the toe of well

X and stages 50–52 near the middle of well Z. Moreover,whereas the events in the MC Formation occurred injust a few stages of well X (stages 2–5 and stage 18),events in the MC Formation are associated with morethan half of the stages of well Z.

Interestingly, the average surface treatment pressureduring the stages associated with large number of shal-low events does not seem to be different from the pres-

sure in other stages. For example, stageswith a larger number of microseismicevents in the MC Formation are not as-sociated with higher treatment pres-sures (with exception of stage 18),whereas two stages with relatively hightreatment pressure (i.e., stages 17 and47) are associated with very few events.We investigated the correlation of mi-croseismic data with other variablessuch as the maximum/average treat-ment pressure and the volume of fluidand proppant injected, and no correla-tion was found. Hence, these shallowevents seem to be more likely the resultof a vertical hydraulic connection asso-ciated with preexisting fractures andfaults rather than anomalous hydraulicfracturing procedures.

Figure 5 illustrates a selected shallowevent from stage 50 recorded by 3C geo-phones in well B. The three componentsof each geophone within the entire arrayare plotted together, where the bottomof the array is approximately 50 ft abovethe top of the Bakken Formation. ClearP- and S-wave arrivals are observed withsimple moveout. The waves arrivaltimes clearly indicate that the event islocated in the MC Formation. The mini-mum P-wave arrival time is observed

approximately 800 ft above the well depth.

Events trend approximately 30° with respectto SHmax

During hydraulic fracture stimulation, one expectsMode I (opening mode) hydraulic fractures to openin a plane perpendicular to the least principal compres-sive stress, or equivalently, parallel to the direction ofSHmax in a strike-slip or normal faulting environment.As the microseismic events are associated with slipon preexisting faults surrounding the hydraulic frac-tures when the well is being pressurized, the microseis-mic events are generally expected to form an elongatedcloud extending in the SHmax direction from the perfo-rations (e.g., Maxwell et al., 2002; Fisher et al., 2004;Moos et al., 2011). As described below, the great major-ity of events located in this study are not consistent withthis pattern. Instead of trending approximately N50°E inthe direction of SHmax as expected, the dominant trendof hypocenters is approximately 30° from SHmax. A total

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

−400

−200

0

200

400

600

800

10001200

−400−200

0200400600800

10001200

30 313233 3435 3637 38 3940 414243 4445 464748 49 505152 53 5455 5657 585960 6162 636465 66 67

Wel

l XW

ell Z

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 290

500

1000

1500

Stage number

Eve

nts

num

ber

45005000

5500

6000

6500

Ave

rage

inje

ctio

n

pres

sure

(ps

i)

Events number versus average injection pressure

Dep

th r

elat

ive

to B

akke

n (f

t)E

vent

s nu

mbe

rD

epth

rel

ativ

e to

Bak

ken

(ft)

30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 670

500

1000

1500

Stage number

45005000

55006000

6500

Ave

rage

inje

ctio

n

pres

sure

(ps

i)

Events number versus average injection pressure

Lodgepole

Bakken

MC

Lodgepole

Bakken

MC

Figure 4. Relative depth (with respect to the Bakken Formation) histograms ofmicroseismic events by stages for wells X and Z, as well as the correspondingaverage surface treatment pressure of each stage.

H2

Z

H1

P-wave S-wave

1900

ft19

00ft

Figure 5. Waveforms of a large event located at MC in stage50, recorded by 3C (H1, H2, and Z) geophones in array B. Thearray ranges from approximately 8000 to 9960 ft, and the bot-tom of the array is approximately 50 ft above the top of theUpper Bakken Formation.

SG28 Interpretation / August 2014

of approximately 1500microseismic events from stages50 to 52 are shown in Figure 6 and illustrate a consistenttrend approximately 30° from the SHmax direction (theblue solid line). This approximately 30° offset with re-spect to SHmax indicates that these microseismic eventsare probably occurring along a preexist-ing natural fault at the orientation ex-pected for strike-slip faults. Therefore,it suggests that well-oriented, preexist-ing strike-slip faults exist in the area,and the fluid injection during hydraulicstimulation of stages 50–52 activatedslip on the faults and caused the approx-imately 1500 microseismic events.

To constrain the source mechanismof these microseismic events, we con-structed earthquake focal plane mecha-nisms for selected microseismic eventsin stage 50 by using the first motion ofP-wave observed on the borehole geo-phones (Aki and Richards, 2002). Toconstruct the focal plane mechanisms,the waves were assumed to propagatealong a linear path between the sourceand the receiver. Because the distancebetween the microseismic hypocentersand geophones is very small (<2000 ft)and the P-wave velocity between the hy-pocenters and seismometers is nearlyconstant (Figure 2), this assumption isquite reasonable. Second, for the shal-low events, the events and geophonesare located within a relatively homo-geneous limestone formation. The sim-plicity of the waveforms shown inFigure 5 indicates the simple path be-tween the source and receiver and thelack of converted phases.

The 3C seismograms are rotated fromthe geographical coordinates (H1, H2,and Z) to radial and transverse direc-tions. Because the two horizontal com-ponents (H1 and H2) are rotated inthe horizontal plane, it requires addi-tional controlled source to calibrate geo-phone rotations. Because the horizontalorientation of the geophones is notavailable for this analysis, instead of ori-enting the vertical and one horizontalcomponent in the same plane with thewave propagation direction, we directlyrotate the vertical component (Z) to thewave propagation direction (Figure 7a).Therefore, only direct P-wave arrivalcan be used for construction of the focalplane mechanisms. After rotation, thedirect P-waves would propagate in thedirection from the source to the receiverwith the first motion up for compression

and down for dilation. Figure 7b and 7c illustrates anexample of seismogram recordings before and afterrotation.

To find the best-fitting focal planes with the limitedreceiver azimuth coverage, we use the P-wave radiation

A

B

C

D

E

F

Y ZX

)nimooz(weivssorCeventsnekkaB

Strike-slip fault trend ~±30° with

Shmin

~30˚

Stages 50-52

0 1000 ft

N

~30˚

0

a) b)

c)

500

1000

1500

2000

2500

3000

A B D F

Rel

ativ

e de

pth

(ft)

0 1000 ft

N

SHmax

SHmax

SHmax

Figure 6. The distribution of 1497 microseismic events of stages 50–52 during Zstimulation. (a) Map view of the microseismic events, the fitted orientation of apotential fault (red solid line), the current N50°E SHmax orientation (blue arrowline), and the orientation of theoretical conjugate fault (red dashed line), (b)zoom in view of north–south cross section of the microseismic events, and(c) idealized relationship between conjugate sets of strike-slip fault and principalstresses (map view).

Source

Receiver

ZH1

H2

Radial

Transverse 1

Transverse 2

Wave propagation direction

ZH

2H

1

Rad

ial

Tra

nsve

rse

1T

rans

vers

e 2

a)

b) c) noitatorretfAnoitatorerofeB

û=(ux,uy,uz)

(x1,y1,z1)

(x2,y2,z2)

−0.25

0

0.25

−0.25

0

0.25

−0.25

0

0.25

−0.25

0

0.25

−0.25

0

0.25

−0.25

0

0.25

Figure 7. (a) Schematic illustration of seismogram rotation, (b) seismogram re-corded by a 3C geophone before rotation, and (c) seismogram of the 3C geo-phone after rotation.

Interpretation / August 2014 SG29

pattern in the homogeneous far-field topredict the polarization (Aki and Ri-chards, 2002) and compare the pre-dicted results with the observations.Because the microseismic events ofstages 50–52 follow a consistent trendof N75°E, we assume this is the orienta-tion of the slipping fault and then predictthe P-wave first motions by varying dipand rake values. Figure 8 shows an ex-ample of the predicted P-wave first mo-tions with strike/dip/rake as 75°∕65°∕–74° (a) and the observed P-wave firstmotions (b), based on the same sourceand receiver locations for a selected mi-croseismic event. Although there aresome minor inconsistencies betweenthe observation and prediction, theoverall fit is quite good. We define themisfit as the percentage of the numberof inconsistency over the total numberof observed polarizations and apply agrid search methodology to calculatethe minimum misfit as a function ofdip and rake. Figure 9 shows a contourmap of misfit for varying dip and rakefor a fault striking N75°E. It is clearlyseen that although there is no uniquesolution due to the limited geophonecoverage, there are multiple solutionsthat can equally fit the observationwithin a limited range of rake and dipvalues (Figure 8b). Clearly, the radiatedseismic wavefield implies a steeply dip-ping fault (>65°), with high rake values

(−70° to −90°), suggesting normal faulting with a smallstrike-slip component.

Figure 10 shows six events with hypocenters spa-tially spread out along the interpreted fault in Figure 6.Note that the selected six events occur in the MC For-mation during pressurization of stage 50. These eventshave a high signal-to-noise ratio, and the signals are wellrecorded bymost of the geophones along arrays A, B, D,and F. The distribution of focal mechanisms (Fig-ure 10c) also suggests the existence of approximatelyN75°E trend fault plane in the area. Dip angles are be-tween 65° and 88°. The slip on the N75°E steeply dip-ping plane in these events is normal and strike-slip,which indicates magnitudes for three principal stressesof Sv ∼ SHmax > Shmin.

To compare and contrast the microseismic eventsoccurring in the Bakken and MC, Figure 11 showsthe locations of the events only during stage 50. Itcan be clearly seen that the events from the Bakkenand MC follow a similar N75°E trend. The limitedP-wave polarization of an event from the Bakken For-mation (Figure 11c) which is consistent with focalmechanism of events from the MC Formation, suggeststhat the events at both depths have the similar source

Predicted P first motion and fitting plane Observed P first motion and fitting plane

up:compression down:dilation

a) b)

up:compression down:dilation

Figure 8. Radiation pattern of P-wave first motions for the selected event andcorresponding beach-ball solutions using recording from arrays A, B, D, and F.(a) Predicted P-wave first motions of a certain focal mechanism and correspond-ing beach ball and (b) observed P-wave first motions and best-fit focal planesolutions.

Dip (°)

Rak

e (°

)

60 65 70 75 80 85 90−90

−85

−80

−75

−70

−65

−60

−55

−50

10

15

20

25

30

35

40

45

50misfit %Strike = N75°E

Figure 9. Contour map of misfit between predicted and ob-served P-wave first motions for various dip and rake.

SG30 Interpretation / August 2014

mechanism. This suggests that the steeply dipping fault,determined from focal plane solutions on MC events,appears to connect the deep Bakken Formation andserves as fluid conduit during hydraulic stimulationin stage 50 and causes the events in the MC Formation.Therefore, a steeply dipping, approximately N75°E-trending fault appears to connect the Bakken andMC Formations, providing a fluid path-way during hydraulic fracturing in theBakken Formation. An Ant Track™ im-age of 3D seismic in the study area sug-gests faults at the depth of MC with asimilar trend (M. Simon, personal com-munication, 2013).

Events near the toe of well XEvents from stages 2–5 near the toe

of well X were reprocessed by a secondcontractor (Figure 12). Although in gen-eral, the reprocessed event locationscaptured the characteristics we ob-served in Figure 1 (see Appendix B),there are some differences worth not-ing. First, the reprocessed events atthe toe of well X (i.e., stage 2) now showa strong lineation in the direction ofN75°E that is consistent with what wehave observed in stages 50–52 (Fig-ure 6). Second, many reprocessedevents are located in the Lodgepole For-mation, displaying a connection be-tween the Bakken and MC. Originally,few events were located at this depth.Altogether, the reprocessed eventsseem to outline a steeply dipping planeconnecting the MC and Bakken Forma-tions, similar to the fault conduit we in-terpreted for stages 50–52. Interestingly,as the hydraulic fracturing was proceed-ing from stages 2–5, the hypocenters oc-cur progressively downward from theMC to the Bakken (Figure 12).

Spatial-temporal plots of theseevents, including depth and horizontalposition versus time, are shown in Fig-ure 13 to examine the temporal growthof the microseismic clouds associatedwith stages 2–5. The downward progres-sion of the events in these four stages isfairly continuous although there is aclustering of events at specific depthsduring each stage (Figure 13b). Thehorizontal positions of the events alongthe N75°E trend are plotted with respectto injection location of stage 2 (zeroalong the y-axis in Figure 13c). Thestages show different characteristicsof growth along the N75°E trend. Micro-seismic events in stages 2, 4, and 5 are

principally to the southwest of the well (especially stage2), whereas events in stage 3 are concentrated nearlydirectly above the well. The strong N75°E lineationof hypocenters seen in stage 2 appears to be associatedwith slip along a prominent fault. Events from stages 4and 5 appear to be caused by slip along many smallerfractures clustered closer to the well, above in the

A

B

C

D

E

F

Y ZX

Map view

0

a) b)

c)

500

1000

1500

2000

2500

3000

C A B D F

E

Cross view

Rel

ativ

e de

pth

(ft)

1000 ft0

Stage 50

B

D

Zoom in view

Stage 50

N

N

Figure 10. (a) Map view of selected events for focal mechanism analysis forstage 50, (b) cross view of selected events, and (c) focal plane solutions ofsix selected events.

A

B

C

D

E

F

Y ZX

Bakken events

0

500

a) b)

c)

1000

1500

2000

2500

3000

C A B D FE

Cross view

Rel

ativ

e de

pth

(ft)

up: compression

Middle BakkenMission Canyon

Bakken event P- first mostion

0 1000 ft

N

N

down: dilation

Figure 11. (a and b) Map view and cross view of events from the Bakken For-mation and MC Formation during stage 50 and (c) P-wave polarization of anevent from the Bakken Formation.

Interpretation / August 2014 SG31

Lodgepole Formation and below in the Three ForksFormation (Figure 12c).

A striking observation of the pumping history duringstage 5 (Figure 13a) further supports the argument thathydraulic stimulation is inducing microseismicity bychanneling flow through a network of preexistingfaults. During stage 5, the fluid injection was shut downafter approximately 20 min due to the broken blender atthe surface, with the pressure dropping from 4100 to2800 psi (Figure 13a). Because the pressure after thepause of injection was far below the fracture gradient,the distribution of microseismic events in stage 5 wastriggered by relatively low pressure in preexisting net-work of fractures and faults without the presence of alarge hydraulic fracture. Recall that during stages 4 and5, microseismic events were also seen thousands offeet southward along well Y (Figure 3). Thus, the spatialdistribution of events and the occurrence of eventsduring stage 5 at pressures lower than the fracture gra-dient support the hypothesis of fluid channeling alongfaults near well X and length of well Y during stage 4stimulation.

Fault slip analysisThe hypocenter distributions of microseismic events

and the focal mechanisms suggest the existence of

steeply dipping faults in the area that connect the deepBakken Formation with the shallower MC Formation.To test whether the pore pressure perturbation duringhydraulic fracture stimulation at the Bakken Formationis likely to cause the slip on the faults in the local stressfield, we assess the proximity of each nodal plane in thefocal mechanisms shown in Figure 10 to evaluate thepotential for shear failure in the local stress field inthe context of the Mohr-Coulomb failure criterion.We use a stress field in which Sv ∼ SHmax > Shminin the model because the focal plane mechanisms indi-cate normal faulting with a small component of strike-slip faulting. We use geomechanical properties such asPoisson’s ratio and the Biot coefficient of the MiddleBakken (Table 2) because they were measured directlyfrom core samples collected from well A and were notavailable for the MC Formation.

We test two scenarios with different stress paths.First, fluid is injected into the reservoir in its initialstate, ignoring any depletion effects from productionin well Y (Figure 14a and 14b). The second scenario at-tempts to account for the depletion effects on the stressfield prior to hydraulic fracturing of wells X and Z. Thestress path associated with this scenario is based on ap-proximately 4450 psi depletion in well Y as illustratedby the pressure data shown in Figure 3 (Figure 14c

and 14d). We assume that the fault planeorientations shown in Figure 8 are rep-resentative of the study area and plotthem in the stereonets and Mohr circles.The background color of the stereonetsrepresents the excess pressure neededto cause slip for a coefficient of frictionof 0.6 for poles of the fault planes. TheMohr circles illustrate the effectiveshear and normal stresses on the planesfor the various cases considered.

For the first scenario without deple-tion, the initial reservoir pressure inthe Middle Bakken is approximately6700 psi. Note that for the estimated ini-tial stress field and pore pressure, nei-ther the steeply dipping, N75°E fault,or its auxiliary plane would be expectedto slip prior to stimulation (Figure 14a).During hydraulic fracturing, a pore pres-sure perturbation of only approximately330 psi would be sufficient to causethe well-orientated N75°E fault to slip(Figure 14b). Because the averagepumping pressure during hydraulic frac-turing is approximately 1300 psi abovethe initial reservoir pressure, approxi-mately 330 psi pore pressure increasecould be easily achieved during stimu-lation.

For the second scenario, the reser-voir is depleted prior to the stimulationdue to the previous production in well

Figure 12. Reprocessed events locations at stages 2–5 near the toe of well X. (a)Map view of the events, (b) cross view of the events, and (c) 3D view (looking tothe direction of S75°W).

SG32 Interpretation / August 2014

Y . With approximately 4450 psi depletion from the ini-tial reservoir state, the Mohr circle moves to the rightwith increasing circle size, due to the relative increasein the vertical effective stress with respect to the hori-zontal effective stresses (Figure 14c). The stress stateafter depletion indicates that slip is not expected onthe approximately N75°E-trending fault. As shown inFigure 14d, a pore pressure increaseof approximately 1600 psi is required toinduce slip on approximately N75°E-trending faults. As the reservoir pres-sure is only approximately 2200 psiafter depletion, an increase in pore pres-sure of approximately 1600 psi couldalso be easily achieved during hydraulicfracturing because the pumping pres-sure is close to 8000 psi. As discussedin Appendix C, Dohmen et al. (2013)carry out a similar analysis for the zoneof depletion around well Y and esti-mated that a pore pressure perturbationapproximately 1200 psi was required toinduce shear failure. Details of our re-spective analyses, including the sensitiv-ity of the results to physical properties(i.e., Biot coefficient, Poisson’s ratio),are discussed in Appendix C.

Figure 15 summarizes our interpreta-tion of events at stages 2–5 and 50–52.We propose that relatively large andsmall preexisting fractures and faultsare present in the vicinity of wells X,Y, and Z and significantly affect the hy-draulic fracture stimulation. Relativelylarge scale faults serve as fluid pressureconduits and transmit injected fluid toother formations, resulting in the out-of-zone seismicity. Small fractures alsotransmit injected fluid from well X towell Y and cause the microseismicevents along the length of well Y.

ConclusionsIn this study, we integrated geome-

chanics with microseismic data tounderstand the role of natural fracturesand faults during multistage hydraulicfracturing stimulation in the MiddleBakken Formation. Analysis of the mi-croseismic locations, focal plane mech-anisms, and fault modeling points to theimportance of preexisting faults as con-duits for pressure and fluid transmissionduring hydraulic fracturing. In severallocations, the existence of relativelylarge-scale, steeply dipping faults inthe area transmit pressure and fluidfrom the Bakken Formation upward tothe MC. Slip would not be expected

on these preexisting faults without the perturbationduring multistage hydraulic fracturing. Overall, weattribute the spatial pattern of hypocenters to be con-trolled by the distribution of preexisting fracturesand faults. We believe that the unusual microseismicpatterns observed in this study result from fluid chan-neling dominated by preexisting fractures and faults,

0

a)

b)

c)

d)

2000

4000

6000

8000

10000

Pre

ssur

e (p

si)

0

20

40

60

Inje

ctio

n ra

te (

bpm

)

Rel

ativ

e de

pth

(ft)

Pos

ition

alo

ng N

75E

t

rend

(ft)

Time (hour)

0

50

100

150

200

250

Num

ber

of e

vent

s

0

500

1000

0 1 2 3 4 5 6 7 8 9 10 11 12 130 1 2 3 4 5 6 7 8 9 10 11 12 13

Stage2Stage3Stage4Stage5

0

1000

–1000

500

–500

Stage 2 Stage 3 Stage 4 Stage 5

Injection PInjection Rate

Injection shut down

Figure 13. Spatial temporal relations of microseismic events at stages 2–5. (a)Injection records of stages 2–5, (b) events locations along vertical direction withrespect to the depth of the Middle Bakken, (c) events positions along the N75°Wtrend with respect to the perforation location of stage 2, and (d) histogram ofevent numbers.

Table 2. Inputs for Mohr circle analysis with different stress paths.

(a) (b) (c) (d)

Sv (psi) 10,605 10,605 10,605 10,605

SHmax (psi) 10,302 10,302 7819.3 7819.3

Shmin (psi) 8080 8080 5597.3 5597.3

Pp (psi) 6666 6996 2213 3813

ΔP (psi) 0 330 –4453 1600

Biot coefficient (α) 0.8 0.8 0.8 0.8

Poisson’s ratio 0.25 0.25 0.25 0.25

Friction coefficient (μ) 0.6 0.6 0.6 0.6

SHmax azimuth N50ºE

Strike N75ºE

Dip angle/dip direction 69º/N165ºE

Interpretation / August 2014 SG33

and that the depletion near well Y pro-moted the fluid pressure transmissionprocess from wells X and Z toward wellY during hydraulic fracture stimulation.

AcknowledgmentsWe thank the Hess Corporation for

providing the data to carry out the re-search and permission to publish thisstudy. We thank B. Hughes for providingsoftware used in geomechanical analy-sis. We also thank T. Dohmen andM. Simon for the valuable discussions.Financial support was provided by Stan-ford Rock Physics and Borehole Geo-physics (SRB) Industrial Consortiumand Hess Corporation.

Appendix A

The SHMAX orientation andmagnitude

FMI image logs provide a continuousunwrapped image of the borehole walland include information about the distri-bution of natural fractures, as well asthe distribution of wellbore failures,such as breakouts and drilling-inducedtensile fractures (DITFs). Boreholebreakouts are compressive failures onthe walls of the borehole that occurwhen the maximum resolved compres-sion exceeds the rock failure strength.DITFs are tensile failures on the well-bore wall that occur when the minimumresolved effective stress becomes nega-tive and the wall fails in tension. The ef-fective stresses at the well face of avertical wellbore in a linear isotropic

elastic medium are described by the Kirsch equations:

σθθ ¼ SHmax þ Shmin − 2ðSHmax þ ShminÞ cos 2θ − 2Pp

− ΔP − σΔT ; (A-1)

σzz ¼ SV − 2ðSHmax − ShminÞ cos 2θ − Pp − σΔT ; (A-2)

and

σrr ¼ ΔP; (A-3)

where ΔP is the difference between the pore pressureand the mudweight, σΔT is the thermal stress from thetemperature contrast between the reservoir and thedrilling fluids, and θ ¼ 0 is the direction of SHmax (Zo-back et al., 2003).

However, when the borehole axis is not aligned withone of the principal stress axes, the direction of maxi-mum tension becomes oblique relative to the boreholeaxis, resulting in an inclined DITF (Figure A-1). To

Figure 14. Stability analysis of the fault planes in the Middle Bakken. (a) Aprestimulation scenario with no pressure perturbation, (b) a poststimulationscenario from initial state with pressure perturbation of 330 psi, (c) a depletedscenario with pore pressure depletion of 4453 psi, and (d) a poststimulationscenario from depleted state with pressure perturbation of 1600 psi. The back-ground color of the stereonets is the pore pressure elevation necessary to ini-tiate slip for a coefficient of friction of 0.6. The white circle is the pole of theactivated fault, and the black tadpole is the slip motion of the activated plane.Mohr circles include the representative fracture/fault and its auxiliary plane.The activated planes in response to pore pressure elevation are shown inred in the Mohr circle.

35 4 252-50

Bakken

Lodgepole

Three forks

MC

Fault Preexisting fractures N

Figure 15. Schematic model to demonstrate the effects ofpreexisting fractures and faults during hydraulic fracturestimulation at stages 2–5 and 50–52.

SG34 Interpretation / August 2014

calculate the stress concentration around the borehole,we need to transform the principal stresses (i.e., S1, S2,S3) into borehole Cartesian coordinates, and this is de-scribed in detail by Peška and Zoback (1995). Aftercoordinate transformation and adding the effect of Ppto obtain effective stress, the perturbed stress fieldaround a misaligned borehole at the borehole wall isgiven by Hiramatsu and Oka (1966),

σzz ¼ σ33 − 2υðσ11 − σ22Þ cos 2θ − 2υσ12 sin 2θ; (A-4)

σθθ ¼ σ11 þ σ22 − 2ðσ11 − σ22Þ cos 2θ − 4σ12 sin 2θ

− Pm þ Pp; (A-5)

and

τθz ¼ 2ðσ23 cos θ − σ13 sin θÞ; (A-6)

where θ is the angle around the borehole wall in thecylindrical coordinate system and ν is the static Pois-son’s ratio of the medium. Considering the principalstresses at the borehole wall, the magnitude and orien-tation of the tangential stress become (Peška and Zo-back, 1995):

σt maxðθÞ ¼12ðσzz þ σθθ þ

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðσzz − σθθÞ2 þ 4τ2θz

qÞ; (A-7)

σt minðθÞ ¼12ðσzz þ σθθ −

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðσzz − σθθÞ2 þ 4τ2θz

qÞ; (A-8)

and

ωðθÞ ¼ 12arctan

�2τθz

σzz − σθθ

�; (A-9)

where ω is the inclination of the maximum tangentialprincipal stress from the borehole axis, which deter-mines the inclination of the DITFs relative to the bore-hole axis. DITFs will occur around the wellbore wherethe most tensional stress becomes negative, assumingthe tensile strength of a rock is essentially zero.

In our study, we first analyze the three FMI logs ob-tained from three horizontal wells described by Sturmand Gomez (2009) that are close to our study area.Although the quality of the FMI logs are low quality,we are able to find that DITFs occur at a few depthranges at the top and bottom of the wellbore and thatmost of them intersect the wellbore axis with a high an-gle (transverse). Figure A-2a displays an exampleof DITFs with high inclination angle (approximately50°–60°) located at the bottom of the wellbore. We thenuse the physical property measurements from the coresamples, as well as estimates of Sv, Shmin, and Pp tomodel the DITFs occurrence along this east–west-trending horizontal well. Modeling is performed usingBSFO module of the software SFIB by GeoMechanicsInternational. An example of matching results is shown

in Figure A-2b, with an SHmax orientation of N50°E andmagnitude of 1.02 psi∕ft. Although the DITFs modeledmatch the observation with a high inclination angle atthe top and bottom of the wellbore, the inputs for SHmaxorientation and magnitude are not unique becausemany other combinations will also predict similarDITFs. Therefore, we follow the methodology outlinedby Peška and Zoback (1995) and predict the positionand inclination angle of DITFs, with a wide range values

Inclination ω

Sv not parallel to borehole axis

Position θ around the borehole

0˚ 180˚ 360˚

ω

θ

Figure A-1. Schematics of occurrence of DITFs in an imagelog when borehole axis is not aligned with principal stress(modified from Zoback, 2007).

Figure A-2. Observed and predicted DITFs along an east–west-oriented horizontal wellbore. (a) DITFs occur at topand bottom of the wellbore with high inclination angle, an ex-ample from FMI logs of well Nesson State 44X-36H that werereported by Sturm and Gomez (2009) and (b) predicted DITFswith inputs listed in Table 1 and SHmax oriented N50°E with amagnitude of 1.02 psi∕ft.

Interpretation / August 2014 SG35

of SHmax orientation and magnitude. Figure A-3 com-bines many information including presence/absenceof DITFs (color versus white), the inclination angle(ω) of DITFs with respect to the borehole axis (colormap), and their position around the borehole. It isclearly seen that when the magnitude of SHmax is enor-mously high (i.e., upper left corner of the plot), trans-

verse DITFs will be predicted with SHmax orientedbetween N0° and 40°E, but they will occur near the sideof the wellbore instead of top and bottom. To be con-sistent with the observation, with approximately50°–60° inclined fractures at the top and bottom ofthe wellbore, SHmax should orient approximatelyN45°–55°E as reported by Sturm and Gomez (2009).However, the magnitude of SHmax cannot be furtherconstrained, and stress state suggests either a normalfaulting or a strike-slip faulting regime.

Appendix B

Microseismic events uncertaintiesThe microseismic events shown in this study (except

for events in Figures 12 and 13) were initially monitoredand processed by the first contractor, and they deter-mined the microseismic event location by using P-wavearrival times (i.e., moveout) at the geophones and con-verted them to source location using a Kirchhoff diffrac-tion migration and stacking method in a 3D grid. The3D-velocity model applied was upgraded from wellboresonic log data using the seismic signals from the perfo-ration shot events, which accounts for horizontal andvertical anisotropy. Perforation calibration of the 84known perforation locations showed an averaged loca-tion error approximately 69 ft, which corresponds tothe average location error of the microseismic events.The second contractor reprocessed six selected stages(stages 2–5, 43, and 50) and used P- and S-pick timesand the 3C hodogram information plus forward model-ing on a grid-search-based velocity model to arrive at alocation that has low residual time error.

Figure B-1 compares the initial andreprocessed event locations for the re-processed stages. Obviously, the reproc-essed event locations also exhibit theunique patterns similar to what we ob-served on initial locations. For example,the events in stage 50 follow the N75°Etrend, and many events in stage 4 dis-tribute thousands of feet southwardalong the middle well. Moreover, eventsare located at the Bakken Formationand MC Formation. The consistency be-tween the initial and reprocessed loca-tions suggests that these microseismicevent locations were credible, and theobserved event patterns indeed re-flected the reservoir responses duringhydraulic fracture stimulation.

Appendix C

Fault slip analysis of events alongwell Y

As mentioned above, prior to hy-draulic fracturing of wells X and Z, themiddle well Y has been in productionfor approximately 2.5 years. Therefore,

Figure A-3. Predicted occurrence of DITFs with varyingcombinations of SHmax orientations and magnitudes, coloredby inclination angle ω to the wellbore axis; the higher ω (hotcolor) represents transverse DITFs, and the lower ω (coolcolor) represents axial DITFs.

A

B

C

D

E

F

Y ZX

A

B

C

D

E

F

Y ZX

0

1000

a) b) c)

d)

2000

3000

C A B D F E

0

1000

2000

3000

C A B D F E

Stage2 Stage3 Stage4 Stage5 Stage43 Stage50

Reprocessed location Initial location Reprocessed location

Initial location

N N

N

N

Rel

ativ

e de

pth

(ft)

Rel

ativ

e de

pth

(ft)

0 1000 (ft)

Figure B-1. Initial and reprocessed microseismic event locations at stages 2–5,43, and 50. (a and b) Map view of reprocessed and initial locations’, (c andd) north–south cross view of reprocessed and initial locations.

SG36 Interpretation / August 2014

the reservoir adjacent to well Y was depleted and thelocal stress field was affected. The change of horizontalstress magnitude resulting from depletion is estimatedby poreoelastic theory and described by

ΔSHor ¼ αð1 − 2υÞð1 − υÞ ΔPp (C-1)

(Engelder and Fischer, 1994). The α is the Biot coeffi-cient and ν is the Poisson’s ratio.

While the Mohr circles shown in Figure 14 corre-spond to the input values from Table 2, it is obvious thatthe depletion effects on the horizontal stress are con-trolled by the Biot coefficient and Poisson’s ratio(see equation C-1). Laboratory measurements of sixcore samples obtained from the Middle Bakken in wellA show that the Poisson’s ratio ν ranges from 0.21 to0.31. The Biot coefficient α is defined by

α ¼ 1 −Kb

Kg(C-2)

(Nur and Byerlee, 1971), where Kb is drained bulkmodulus of the rock and Kg is the bulk modulus ofthe rock’s individual solid grains. We measured thevolumetric strain of the two core samples at the hydro-static condition, using the core samples from the MiddleBakken in well A. Our results suggestedthat Kb ranges from 7.85 to 9.53 GPawhen the confining pressure changesfrom 20 to 30 MPa. The mineralogy ofthe sample measured by XRD is listedin Table C-1. Using the Hill’s average(Mavko et al., 2009) of 18.5% clay,52.5% calcium, and 29% quartz, Kg is ap-proximately 41.15 GPa. Therefore, α inthe Middle Bakken ranges from 0.78to 0.81.

Figure C-1 shows the dependence ofspecific parameters in the fault slipanalysis. With a coefficient of friction of0.6 and 4453 psi depletion, Figure C-1aillustrates the pore pressure perturba-tion required to make the existing frac-tures/fault unstable, with varying α andν. Clearly, the results are more sensitiveto α than ν. This is reasonable because αdetermines the pore pressure effects,thus, directly affecting the position andsize of the Mohr circles. Figure C-1bshows the similar analysis with a con-stant ν ¼ 0.25, and varying depletionand α. As expected, depletion and αare affecting the results. For the initialreservoir with minimal depletion, shearfailure would occur in the Bakken witha pore pressure perturbation <500 psi.However, when the reservoir is depletedapproximately 4500 psi due to prior pro-

duction, the prediction becomes sensitive to α and re-quires higher pore pressure perturbation when α islower. On the other hand, with a constant α, shear fail-ure is more easily achieved for a reservoir with lessdepletion.

Dohmen et al. (2013) investigate the importance ofdepletion effects and the pore pressure perturbationrequired to cause shear failure in the vicinity of wellY. Although the result of our analysis is generally con-sistent with theirs, our respective analyses differ in sev-eral ways. First, we estimate Shmin from directmeasurements and then estimate the change of Shminassociated with depletion effects. In contrast, they esti-mated Shmin using a general empirical relation based onbilateral-constraint (the horizontal stress is related tothe vertical stress by a term ν∕ð1 − νÞ). As the bilat-eral-constraint is of questionable applicability for pre-dicting (Zoback [2007], pp. 292–295), we argue thatusing measured values of Shmin is preferable. Second,we use physical properties based on lab measurementson the Middle Bakken samples, rather than estimatedvalues. Third, our analysis uses the specific faultorientations determined from the seismicity trendsand focal plane mechanisms. Finally, based on the focalplane mechanisms, we use an initial stress state withSHmax slightly smaller than Sv whereas Dohmen et al.(2013) assume a stress state representing a pure

Figure C-1. Sensitivity analysis of required pore pressure elevation for shearslip occurrence along the representative fracture/fault (see Table 2), with coef-ficient of friction of 0.6. (a) Varying Biot coefficient and Poisson’s ratio, with4453 psi reservoir depletion. (b) Varying Biot coefficient and reservoir depletion,with constant Poisson’s ratio of 0.25.

Table C-1. Mineralogy of core samples from the Middle Bakken andtheir bulk modulus.

Quartz(Mavko et al., 2009)

Calcite(Mavko et al., 2009)

Clay(Vanorio et al., 2003)

K (GPa) 37 70.2 12

% 0.29 0.525 0.185

Interpretation / August 2014 SG37

normal faulting stress regime. The results presented inFigure C-1 demonstrate that our analysis indicates thatmicroseismic events along the well Y would have oc-curred with or without depletion.

ReferencesAki, K., and P. G. Richards, 2002, Quantitative seismology:

Theory and methods: University Science Books.Dohmen, T., J. Zhang, C. Li, J. P. Blangy, K. M. Simon, D. N.

Valleau, J. D. Ewles, S. Morton, and S. Checkles, 2013,A new surveillance method for delineation ofdepletion using microseismic and its application to de-velopment of unconventional reservoirs: Presented atSPE Annual Technical Conference and Exhibition,SPE 166274.

Engelder, T., and M. P. Fischer, 1994, Influence of poroe-lastic behavior on the magnitude of minimum horizontalstress, Sh in overpressured parts of sedimentary basins:Geology, 22, 949–952, doi: 10.1130/0091-7613(1994)022<0949:IOPBOT>2.3.CO;2.

Fehler, M., L. House, and H. Kaieda, 1987, Determiningplanes along which earthquakes occur: Method and ap-plication to earthquakes accompanying hydraulic frac-turing: Journal of Geophysical Research, 92, 9407–9414,doi: 10.1029/JB092iB09p09407.

Fisher, M., J. Heinze, C. Harris, C. Wright, and K. Dunn,2004, Optimizing horizontal completion techniques inthe Barnett shale using microseismic fracture mapping:Presented at SPE Annual Technical Conference and Ex-hibition, SPE 90051.

Gerhard, L. C., S. B. Anderson, and D. W. Fischer, 1990,Petroleum geology of the Williston Basin: Part II: Se-lected analog interior cratonic basins, in M. W.Leighton, D. R. Kolata, D. F. Oltz, and J. J. Eidel,eds., Interior cratonic basins: AAPG Memoir 51, 507–559.

Hiramatsu, Y., and Y. Oka, 1966, Determination of the ten-sile strength of rock by a compression test of an irregu-lar test piece: International Journal of Rock Mechanicsand Mining Sciences & Geomechanics Abstracts, 3, 89–90, doi: 10.1016/0148-9062(66)90002-7.

Mavko, G., T. Mukerji, and J. Dvorkin, 2009, The rock phys-ics handbook: Tools for seismic analysis of porous me-dia: Cambridge University Press.

Maxwell, S., T. Urbancic, N. Steinsberger, and R. Zinno,2002, Microseismic imaging of hydraulic fracturecomplexity in the Barnett shale: Presented at SPEAnnual Technical Conference and Exhibition, SPE77440.

Moos, D., G. Vassilellis, R. Cade, J. Franquet, A. Lacazette,E. Bourtenbourg, and G. Daniel, 2011, Predicting shalereservoir response to stimulation in the Upper Devon-ian of West Virginia: Presented at SPE Annual TechnicalConference and Exhibition, SPE-145849.

Nur, A., and J.D. Byerlee, 1971, An exact effective stresslaw for elastic deformation of rock with fluids: Journal

of Geophysical Research, 76, 6414–6419, doi: 10.1029/JB076i026p06414.

Olsen, T., E. Gomez, D. McCrady, G. Forrest, A. Perakis,and P. Kaufman, 2009, Stimulation results and com-pletion implications from the consortium multiwellproject in the North Dakota Bakken shale: Presentedat SPE Annual Technical Conference and Exhibition,SPE 124686.

Peška, P., and M.D. Zoback, 1995, Compressive and tensilefailure of inclined well bores and determination of insitu stress and rock strength: Journal of GeophysicalResearch: Solid Earth, 100, 12791–12811, doi: 10.1029/95JB00319.

Phillips, W., J. Rutledge, T. Fairbanks, T. Gardner, M.Miller, and B. Schuessler, 1998, Reservoir fracture map-ping using microearthquakes: Austin chalk, Giddingsfield, TX and 76 field, Clinton Co., KY: SPE ReservoirEvaluation & Engineering, 1, no. 2, 114–121, doi: 10.2118/36651-PA.

Rutledge, J.T., W.S. Phillips, and B.K. Schuessler, 1998,Reservoir characterization using oil-production-in-duced microseismicity, Clinton County, Kentucky: Tec-tonophysics, 289, 129–152, doi: 10.1016/S0040-1951(97)00312-0.

Sturm, S., and E. Gomez, 2009, Role of natural fracturing inproduction from the Bakken Formation, Williston basinNorth Dakota: Presented at AAPG Annual Conventionand Exhibition, 50199.

Vanorio, T., M. Prasad, and A. Nur, 2003, Elastic propertiesof dry clay mineral aggregates, suspensions and sand-stones: Geophysical Journal International, 155, 319–326.

Zoback, M., C. Barton, M. Brudy, D. Castillo, T. Finkbeiner,B. Grollimund, D. Moos, P. Peska, C. Ward, and D. Wi-prut, 2003, Determination of stress orientation and mag-nitude in deep wells: International Journal of RockMechanics and Mining Sciences & Geomechanics Ab-stracts, 40, 1049–1076, doi: 10.1016/j.ijrmms.2003.07.001.

Zoback, M. D., 2007, Reservoir geomechanics: CambridgeUniversity Press.

Yi (Alec) Yang received a B.S.in geology from Peking University,China, and an M.S. in petroleumengineering from Stanford University,which is where he is pursuing a Ph.D.in geophysics. His Ph.D. research fo-cuses on unconventional reservoir de-velopment, particularly on using the

geomechanics and microseismic to evaluate and optimizemultistage hydraulic fracturing stimulation. His researchinterests also include lab study of the mechanical andtransport properties of unconventional reservoir rocksand their application to reservoir development. He cur-rently works as a petrophysicist for Shell.

SG38 Interpretation / August 2014

Mark D. Zoback is the Benjamin M.Page Professor of Geophysics at Stan-ford University. He conducts researchon in situ stress, fault mechanics,and reservoir geomechanics with anemphasis on shale gas, tight gas,and tight oil production. He is theauthor of Reservoir Geomechanics,

published in 2007 by Cambridge University Press. He isthe author/coauthor of more than 300 technical papersand holds five patents. He served on the National Academyof Engineering committee investigating the DeepwaterHorizon accident and the Secretary of Energy’s committeeon shale gas development and environmental protection.He is currently serving on the National Academy of Scien-ces Advisory Board on drilling in the Gulf of Mexico.

Interpretation / August 2014 SG39