Azimuthal variations in attributes for inducedfracture detection in the Barnett Shale

Amanda Trumbo1 and Jamie P. Rich2

Abstract

Wide azimuth seismic data acquisition has become standard in many plays where azimuthal characteri-zation of fractures or stress is important for fully understanding reservoir behavior. Usually, the desire is toimage preexisting fracture networks, but, in artificially stimulated reservoirs, it is also desirable to under-stand the fracture state after completion. To characterize the induced fractures in a heavily drilled area ofthe Barnett Shale, a wide azimuth survey was acquired and processed to generate four separate azimuthalstacks. Attributes exploiting variations in amplitude and waveform were independently calculated for eachazimuthal stack. Azimuthal variations in the attributes were then quantified through ellipse fitting. The attrib-utes with the most statistical significance in azimuthal variability were those related to frequency such as thewavelet dominant frequency. These attributes were consistent with each other but show a wide distributionover azimuths across the survey. In addition, random distribution of orientations within the Barnett shalecompared to the over and underlying Marble Falls and Viola Limestones indicate a difference in orientationfor the induced as opposed to preexisting fractures. With the ability to map the induced fractures, it becomesfeasible to take an intelligent approach to planning infill locations to avoid the previously fractured areas.

IntroductionEconomic production from the Barnett Shale re-

quires hydraulic stimulation to create pathways for fluidflow. Much of the early development in the Barnett tookplace with vertical or horizontal wells without a goodunderstanding of optimal placement and spacing. Thishas resulted in the current situation where there isbelieved to be tremendous potential for continual devel-opment of bypassed pay. With the possibility for sus-tained low natural gas prices in the future, there is astrong desire to take a more intelligent approach toplacing these infill wells than was used for primary de-velopment. This study looks at an attempt to image pre-viously stimulated areas to maximize the developmentof bypassed pay zones. It is desirable to avoid previ-ously stimulated areas due to potential interferencewith preexisting induced fractures during stimulationas well as minimizing exposure to depleted zones. En-gineering data alone cannot predict the stimulated areafor existing wells. Microseismic data exist for only asmall subset of the original wells, and where it does ex-ist, there is a questionable connection between micro-seismic events and effectively propped and producedvolume. For these reasons, we turned to surface seis-

mic data, which has been extensively exploited in theBarnett for hazard avoidance and natural fracture char-acterization (Rich and Ammerman, 2010). The goal inthis study was to exploit a 3D seismic study which wasshot after primary development for imaging of inducedfractures. Although it is common practice to use seis-mic data and its derivative products (attributes, velocityanisotropy, amplitude variation with azimuth) to mapnatural fractures (Schoenberg and Sayers, 1995; Lynn,2004; Chopra and Marfurt, 2007), little has been re-ported on mapping induced fractures. Wide azimuthsurface seismic data is often used to characterize azi-muthal velocity anisotropy which is related to the strikeof fractures or maximum stress direction which deter-mines how the reservoir responds to stimulation (Richand Ammerman, 2010). Due to variations in the orien-tation and intensity of fractures, attributes calculatedon azimuthal volumes can also provide insight intothe spatial variability of the fractures. Because theseseismic experiments can detect naturally occurringfractures in reservoirs, it is a simple extension to char-acterize hydraulically induced fractures. This casestudy demonstrates how these azimuthally sectoredvolumes and their attributes may be used to map the

1Formerly University of Oklahoma, ConocoPhillips School of Geology and Geophysics, Norman, Oklahoma, USA; presently Chesapeake Energy,Oklahoma City, Oklahoma, USA. E-mail: Amanda.Trumbo@chk.com.

2Formerly Devon Energy, Oklahoma City, Oklahoma, USA; presently University of Oklahoma, ConocoPhillips School of Geology andGeophysics, Norman, Oklahoma, USA. E-mail: Jamie.Rich@ou.edu.

Manuscript received by the Editor 2 March 2013; revisedmanuscript received 13 June 2013; published online 24 October 2013. This paper appearsin Interpretation, Vol. 1, No. 2 (November 2013); p. SB51–SB59, 9 FIGS., 1 TABLE.

http://dx.doi.org/10.1190/INT-2013-0027.1. © 2013 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved.

t

Special section: Interpretation for unconventional resources

Interpretation / November 2013 SB51Interpretation / November 2013 SB51

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

hydraulically induced fractures in theBarnett Shale, Ft. Worth Basin.

Acquisition and processingIn April 2009, Devon Energy acquired

a wide azimuth, 51-km2, proprietary 3Dseismic survey over a highly developedarea of the Barnett Shale. Overall, theP-wave seismic data are of excellentquality with a high signal-to-noise ratiowith broad spectral content approaching100 Hz (Figure 1). Table 1 summarizesthe acquisition parameters. Offsets equalto or greater than the target depth wereacquired.

During processing, the azimuthal ve-locity variations were computed based

on the far-offset azimuthal variation in travel times.The data was prestack time-migrated, and four 45° wideazimuthally sectored volumes centered about 0°(North),45°, 90°, and 135° were generated for further analysis.Figure 2 shows the azimuth (spider diagram) of the mid-points inside representative CMP bins. Overall, there wasa good distribution over all azimuths justifying the azimu-thal processing. Although Table 1 shows the bin size tobe 110 × 110 ft (33 × 33 m), the bin size of the azimuthalsectored volumes was increased to generate “super gath-ers” at 220 ft (67 m) to increase fold. Figure 3a shows thefold of the 110 ft bins to be about 35, whereas the fold forthe super gathers in Figure 3b is about 55.

Geologic and production dataThe 3D survey is within an area that was heavily de-

veloped early in the history of the Barnett Shale (Mar-tineau, 2007). There are several hundred Barnett gaswells, approximately two-thirds of which are verticalor directional, and the remainder of which are horizon-tal. There is a substantial amount of geologic and pet-rophysical data related to these wells. Image logs wererun on sixteen wells prior to 3D seismic acquisition andtwo wells after the acquisition. Approximately a dozenmicroseismic experiments were run on another subsetof wells prior to acquisition. Four wells have conven-tional cores, eight wells have sidewall cores, and onewell has conventional and sidewall cores. In addition,two production logs were acquired. Completed zonesfor vertical wells consist of the Upper and LowerBarnett intervals. Horizontal wells are primarilylanded and completed in the Lower Barnett interval;however, there are seven horizontal Upper Barnettcompletions, including one with microseismic. Thesedata were used to corroborate the locations of thehydraulically induced fractures as derived from thesurface seismic.

Azimuthal attribute analysisTypically, seismic attributes are run on full-azimuth,

full-offset stacked data. In this study, a full suite of

Figure 1. Example of seismic data quality with interpreted horizons.

Figure 2. Spider diagram of the azimuthal coverage corre-sponding to the CMP bins. Note the azimuthal coverage iswide and close to uniform.

Table 1. Acquisition parameters used to allow subse-quent azimuthal processing.

Number of live lines 30

Number of stationsper line

120

Receiver line interval 660 ft (201 m)

Receiver group spacing 220 ft (67 m)

Shot line interval 880 ft (268 m)

Vibrator array interval 220 ft (67 m)

Patch size 26,180 × 25,520 ft(7980 × 7778 m)

Nominal bin size 110 × 110 ft (33 × 33 m)

Number of vibratorsweeps

8

Number of vibratorsper array

3

Sweep range 10–110 Hz, 10 s duration,3 db∕octave

Number of geophonesper group

6 in a 6 ft (2 m) circlearound station

SB52 Interpretation / November 2013

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

attributes was generated using each of the four azimu-thally sectored volumes. In total, 180 attributes that ex-ploit geometric, phase, frequency, and amplitudedifferences were calculated on each sectored volume.Horizons were picked independently on each azimu-thally sectored, stacked, and migrated volume. This pre-vents any residual time differences from manifesting asamplitude variations for the extracted attributes. Threehorizons above the zone of interest were chosen, plusone at the top of the Barnett Shale, one within theBarnett Shale, and one below. This resulted in 4320amplitude slices, 1080 for each azimuthal volume. Ana-lyzing anisotropy along multiple horizons provides ameans of differentiating azimuthal variability in thehydraulically fractured target, and “natural” anisotropyexpressed in the underlying and overlying fracturebarriers.

The wells are assumed to be producing from open,hydraulically induced gas-filled fractures, and if horizon-tally transverse or orthorhombic symmetry is present(one dominant set of vertically aligned open fractures),it should be possible to detect open fractures with sur-face seismic (Lynn, 2004). Variations are expected in theintensity and orientation of the induced fractures basedon the wide variety of microseismic responses observedwithin the survey (Figure 4). Knowledge of the fast and

slow azimuth directions, velocity anisotropy trends fromprevious seismic experiments (Simon, 2005), and behav-ior of the microseismic data all aid in confirming thisexpectation.

Ellipse fitting to azimuthal volumesGrechka and Tsvankin (1998) show how the NMO

velocity in a medium with horizontally transverse sym-metry is elliptical with respect to azimuth. Fitting an el-lipse to time delays in azimuthally sectored data hasbeen a standard approach to azimuthal velocity analysisin the past (Garotta, 1989; Grechka and Tsvankin, 1999).However, it has been demonstrated that the layer strip-ping necessary for defining interval azimuthal propertiesis very sensitive to the number of sectors and may re-quire an unrealistic number of sectors for reliable results(Lynn, 2007, 2011). Because the major objective in thisstudy was the characterization of the individual layerproperties, we instead chose to look at amplitudes whichdo not suffer the same shortcomings. We exploited thepotential azimuthal variations in amplitude by consider-ing azimuthal variations in attributes which are associ-ated with the waveform. To avoid the overprint of thetime-delay anisotropy, the same horizon was interpretedindependently on each of the azimuthally sectored vol-umes. These horizons were then used to extract a given

Figure 3. (a) Fold map of the 45° azimuthal sector at 110 ftbins. (b) Fold map of the 45° azimuthal sector at 220 ftsuperbins.

Figure 4. (a) Eccentricity e, rendered transparent when itsreliability R, given by equation 6, falls below an acceptablethreshold. (b) The same image showing the locations of eventsfrom six microseismic experiments. Notice the microseismicevents appear in areas of low reliability and low eccentricity.

Interpretation / November 2013 SB53

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

attribute for each corresponding azimuthally sectoredvolume. These attribute values AðθÞ were then fittedto an ellipse following the methodology of Thompson(2010)

AðθÞ ¼ a cos2 θ þ b cos θ sin θ þ c sin2 θ. (1)

By casting equation 1 for each azimuth θj in matrixform, we get

0BB@

cos2 θ1 cos θ1 sin θ1 sin2 θ1cos2 θ2 cos θ2 sin θ2 sin2 θ2: : : : : : : : :

cos2 θn cos θn sin θn sin2 θn

1CCA� abc

�¼

0BB@

A1

A2

: : :An

1CCA.

(2)

Equation 2 has the form A ¼ TC, where A is a vector ofthe measured attributes, and C is a vector of the un-known coefficients, which can be solved using least-squares

C ¼ ðTTTþ σIÞ−1TTA; (3)

where I is the identity matrix and σ is a small positivenumber used to stabilize the result. To find the majorand minor axes of the ellipse, the ellipse is rotatedby some angle β such that these line up with the localcoordinate system. For this problem, the equation of theellipse can be written as

AðθÞ ¼ λ1 cos2ðθ − βÞ þ λ2 sin2ðθ − βÞ; (4)

where λ1 and λ2 are the eigenvalues of the matrixformed by the general coefficients

�a bb c

�(5)

and where β is the azimuth of the eigenvector associ-ated with the eigenvalue λ1 (Figure 5). The eccentricity

e (or degree of anisotropy) of the best-fit ellipse is givenby

e ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1 −

�λ2λ1

�2

s: (6)

By construction λ2 ≤ λ1; therefore, 0 ≤ e ≤ 1. Thereliability of the azimuthal attribute is defined as:

R ¼ jλ1 − λ2jffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPnðAn−AðθnÞÞ2

4

r : (7)

The definition of reliability for azimuth was chosensuch that it is lower for an attribute that has little varia-tion in amplitude relative to the root-mean-square (rms)error. To determine a statistically significant thresholdfor reliability, data ellipses were fit to random data andreliabilities were calculated. For the random data set, amedian reliability of 5.1 was acquired over multiple iter-ations, with more than two-thirds of the random datafalling below a reliability of six.

After fitting ellipses to all 180 attributes for the sixhorizons, the eccentricity e, as seen in equation 6,and reliability, R, in equation 7, were mapped and com-pared. The median reliabilities for the horizons variedfrom four to seven. Those horizons with reliabilities be-low the random median value of 5.1 were binary attrib-utes that are unfit for ellipse fitting and they weresubsequently discarded from the analysis. The attrib-utes were ranked based on the median reliabilities, withthe top 10 attributes chosen for further analysis. One ofthese is the wavelet dominant frequency attribute, de-fined as a measure of central tendency of the bandwidth(Barnes, 1993). Figures 6 and 7 show the eccentricityand reliability of the wavelet dominant frequency attrib-ute, respectively for the Lower Barnett Shale and ViolaLimestone horizons.

Notice in Figure 6a that the degree of eccentricity islarger at the top of the Lower Barnett Shale than at thetop of the Viola (Figure 6b), indicating a different frac-turing regime at the interface at the top of the Barnettthan at the base. Figure 7a and 7b shows the reliabilitycalculated at the Lower Barnett Shale and Viola. Wherethe map is red, the reliability exceeds six. Clearly, mostof the data falls into this category indicating statisticalsignificance for the eccentricity values when com-pared to random data where two-thirds of the datahas a reliability value less than six. To further exploitdifferences between the horizons, we investigated thecorrelation of eccentricities by crossplotting the Mar-ble Falls top versus the Lower Barnett Shale top. It isassumed that the natural fracture orientation is similarin the Marble Falls and Barnett formations. Basedon microseismic data, induced fractures within theBarnett do not penetrate into the Marble Falls. The

Figure 5. Illustration of the parameters used for ellipse fit-ting and calculation of eccentricity and reliability.

SB54 Interpretation / November 2013

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

expectation is that a lack of azimuthal correlation be-tween the two attributes would be attributable to theinduced fracture orientations in the Barnett. Fromcrossplotting the ten attributes that exhibited the high-est eccentricity, three attributes related to spectral char-acteristics were found to have poor correlations betweenthe Marble Falls and the Lower Barnett Shale. Correla-tion coefficients ranged from 0.08 to 0.2. These attributeswere the Gabor-Morlet spectral components, the waveletenvelope derivative and the wavelet dominant frequency(Taner, 2000).

Next, we looked at the ellipse fit azimuths for theseattributes with poor correlations between the MarbleFalls and the Lower Barnett Shale. If orthorhombicsymmetry with one set of vertically aligned fracturesis present, one dominant azimuth direction would beexpected. Rose diagrams were created from the azi-muth of the best-fit ellipse for each of the four attributeson each of the four horizons, Marble Falls Limestone,Forestburg Limestone, Lower Barnett Shale, and the Vi-ola Limestone. The azimuths in Figure 8 show a domi-nant east–west orientation in the Marble Falls and Viola

whereas the orientations are significantly more randomwithin the Forestburg and the Lower Barnett.

Finally, the azimuths were compared with the eccen-tricities. Figure 9 displays the azimuth directions using acyclical colorbar blended with the eccentricities ingrayscale for each of the four horizons. Larger eccen-tricities appear black, while smaller eccentricities aretransparent. These maps show there are higher eccen-tricity values for the Marble Falls Limestone and the Vi-ola Limestone with lower values for the ForestburgLimestone and the Lower Barnett Shale, implying thecompetent, nonartificially fractured Marble Falls andViola Limestone are more anisotropic than the knownartificially fractured interval. This also contradicts theassumption of orthorhombic symmetry in the LowerBarnett Shale and suggests a more complex, inducedfracture network.

Incorporation of well dataWe have two means of directly measuring the frac-

ture azimuth and intensity of damaged rock: through

Figure 7. (a) Example of reliability calculated on the top ofthe Lower Barnett Shale. (b) Example of reliability calculatedon the top of the Viola Limestone.

Figure 6. (a) Example of eccentricity calculated on the topof the Lower Barnett Shale. (b) Example of eccentricity cal-culated on the top of the Viola Limestone.

Interpretation / November 2013 SB55

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

the use of image logs run before the stimulation proc-ess, and through microseismic events run during thestimulation process. Image logs and microseismic ex-periments were used to verify the finding of a complexinduced fracture orientation as seen from the surfaceseismic.

The use of image logs has increased substantiallyover the last few years and has become a common ad-dition to the standard suite of logging while drilling(LWD) tools used in the Barnett Shale. Information de-rived from these images include detection of naturaland drilling induced fractures, estimation of the densityand orientation of the fractures, classification of openand closed fractures, and the resulting interpretation ofstress states around the borehole (Quinn et al., 2008).This information is used to better place, orient, drill,and complete future wells.

Natural fracture distribution can vary within a basinand can make the difference between an uneconomicand economic well (Quinn et al., 2008). In the BarnettShale, the best-performing wells appear to be drilledin areas that do not have any preexisting fracture net-works. Similarly, induced fractures are also detrimental;tapping into these fracture networks may negatively im-pact production and lead to the loss of fracture energy tobreak new rock. To optimize the hydraulic stimulationprogram, it is necessary to obtain information aboutthe type, location, frequency, and direction of fractures(Janwadkar et al., 2007).

The picks relevant for this study were the open-con-ductive fractures and the closed fractures. To evaluatethe image logs, which represent one point within thesurvey, all of the picks were combined and interpretedtogether. The picks were separated into six groups:open-conductive fractures above and below the Lower

Figure 8. Rose diagrams showing the wavelet dominant frequency azimuth of the best-fit ellipse for the Marble Falls Limestone,Forestburg Limestone, Lower Barnett Shale, and Viola Limestone. Note the difference between the rich variation in azimuth of thehydraulically fractured Forestburg Limestone and Lower Barnett Shale and the strong east–west azimuth in the overlying MarbleFalls and underlying Viola Limestone fracture barriers.

SB56 Interpretation / November 2013

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

Barnett Shale, closed fractures above and below theLower Barnett Shale, and open-conductive and closedfractures above and below the Lower Barnett Shale.Multiple azimuths of the open-conductive fractures fromthe image logs were more abundant in the known hy-draulically induced interval (Lower Barnett Shale) than

above it (Marble Falls Limestone), supporting the obser-vations made from the surface seismic attributes.

The microseismic response in the Barnett Shale tendsto be less predictable than inmany other plays. The smalldifferential stress field combines with a complex naturalrock fabric to produce this highly variable response

Figure 9. Maps showing the azimuth of the ellipse fit blended with the eccentricity of the wavelet dominant frequency for theMarble Falls Limestone, Forestburg Limestone, Lower Barnett Shale, and Viola Limestone. Notice the greater frequency ofazimuths within the Forestburg Limestone and Lower Barnett Shale and greater eccentricities in the Marble Falls and ViolaLimestones.

Interpretation / November 2013 SB57

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

(Rich and Ammerman, 2010). The expectation is thatthe areas in which microseismic response occurs aremore likely to exhibit complex induced fractures. It isimportant to note that there are hundreds of wellswithout microseismic data, so it is impossible to com-pare in entirety the fractured areas to the eccentricity.However, in a comparison of the microseismic eventswith the ellipse fit eccentricities filtered by reliability;the microseismic appears primarily in areas of loweccentricity and low reliability (Figure 9). This sug-gests a complex fracture network resulting from thehydraulic stimulation of the well. This also leads toa reasonable expectation that the other areas withlow eccentricity are more likely to have inducedfractures.

ConclusionsAzimuthal variations in the seismic data can be ac-

centuated through the use of attributes. Mapping theeccentricity and the reliability of the best-fit ellipsefor the azimuthal attributes is interpreted as an indica-tor of hydraulically induced fractures. Comparison ofreliability of ellipse fit with random data indicates a sta-tistically significant azimuthal variation in the seismicattributes, with the spectral attributes exhibiting thestrongest reliability. The ellipse fitting of the attributesfurther verified that multiple azimuthal directions areseen within the fractured interval and are more com-plex than previously believed. Correlation of observedand inferred fractures through the use of image logs andmicroseismic support the observations of complex in-duced fractures. Although curvature, velocity aniso-tropy, and maximum horizontal stress all play a partin predicting how the rock will fracture, there is notone dominant azimuth the induced fractures prefer.This study has shown that, where the wells have beenhydraulically fractured, there is a very complex fracturefairway. It cannot be assumed there is orthorhombicsymmetry within the fractured interval and thereforethe areas of induced fractures cannot be completelycharacterized by azimuthal P-wave surface seismicalone. Although the eccentricity can be useful for highgrading infill locations, the use of converted wave datamay be necessary to provide additional insight in areasof hydraulically induced fractures with two or more setsof cracks.

AcknowledgmentsThe authors wish to thank Devon Energy and the

University of Oklahoma AASPI consortium for support-ing this work and granting permission to publish the re-sults. We also would like to recognize the many usefuldiscussions and suggestions from Heloise Lynn, MikeAmmerman, and Kurt Marfurt. Lastly, we thank the re-viewers whose valuable feedback and suggestions aregreatly appreciated.

ReferencesBarnes, A. E., 1993, Instantaneous spectral bandwidth

and dominant frequency with applications to seismicreflection data: Geophysics, 58, 419–428, doi: 10.1190/1.1443425.

Chopra, S., and K. Marfurt, 2007, Seismic attributed forprospect identification and reservoir characterization:Geophysical Developments, no. 11, SEG.

Garotta, R., 1989, Detection of azimuthal anisotropy: 59thAnnual International Meeting, SEG, Expanded Abstracts,861–863.

Grechka, V., and I. Tsvankin, 1998, 3-D description of nor-mal moveout in anisotropic inhomogeneous media:Geophysics, 63, 1079–1092, doi: 10.1190/1.1444386.

Grechka, V., and I. Tsvankin, 1999, 3-D moveout velocityanalysis and parameter estimation for orthorhom-bic media: Geophysics, 64, 820–837, doi: 10.1190/1.1444593.

Janwadkar, S., S. Morris, M. Potts, J. Kelley, D. Forten-berry, G. Roberts, M. Kramer, S. Privott, and T. Rogers,2007, Advanced LWD and directional drilling technolo-gies overcome drilling and completion challenges in lat-eral wells of Barnett Shale: SPE 110837.

Lynn, H., 2004, The winds of change anisotropic rocks-their preferred direction of fluid flow and their associ-ated seismic signatures — Part 1: The Leading Edge,1156–1162, doi: 10.1190/1.1825938.

Lynn, W., 2007, Uncertainty implications in azimuthalvelocity analysis: 77th Annual International Meeting,SEG, Expanded Abstracts, 84–87.

Lynn, W., 2011, Azimuthal interval velocity uncertainty:81st Annual International Meeting, SEG, Expanded Ab-stracts, 279–282.

Martineau, D. F., 2007, History of the Newark East fieldand the Barnett Shale as a gas reservoir: AAPG Bulletin,91, 399–403, doi: 10.1306/intro910407.

Quinn, T. H., J. Dwyer, C. Wolfe, S. Morris, and D. Coope,2008, Formation evaluation logging while drilling(LWD) in unconventional reservoirs for production op-timization: SPE 119227.

Rich, J. P., and M. Ammerman, 2010, Unconventional geo-physics for unconventional plays: SPE 131779.

Schoenberg, M., and C. M. Sayers, 1995, Seismic anisotropyof fractured rock: Geophysics, 60, 204–211, doi: 10.1190/1.1443748.

Simon, Y., 2005, Stress and fracture characterization in ashale reservoir, North Texas, using correlation betweennew seismic attributes and well data: M. S. thesis, Uni-versity of Houston.

Taner, M. T., 2000, Attributes revisited, http://www.rocksolidimages.com/pdf/attrib_revisited.htm, accessed29 December 2009.

Thompson, A., 2010, Induced fracture detection in the Bar-nett Shale, Ft. Worth Basin, Texas: M. S. thesis, Univer-sity of Oklahoma.

SB58 Interpretation / November 2013

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/

Amanda Trumbo is a geophysicist inthe reservoir characterization groupat Chesapeake Energy Corporationin Oklahoma City. After graduatingwith a degree in mathematics fromthe University of Oklahoma, sheworked as a geoscience technicianfor Devon Energy Corporation inOklahoma City. Upon completion of

her Masters in geophysics, also from the University ofOklahoma, she continued at Devon as a Geophysicistworking the Barnett Shale and the Permian Basin until2012.

Jamie P. Rich received a B.S. (2001)in applied physics with a geology con-centration and anthropology from In-diana University of Pennsylvania, andan M.S. (2003) and Ph.D. (2006) ingeophysics from the University ofOklahoma. He spent six years withDevon Energy before joining the fac-ulty at the University of Oklahoma

ConocoPhillips School of Geology and Geophysics in2012. His research interests include seismic attributes,characterization of seismic anisotropy, and microseismicmonitoring.

Interpretation / November 2013 SB59

Dow

nloa

ded

01/2

3/14

to 1

29.1

5.12

7.24

5. R

edis

trib

utio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms

of U

se a

t http

://lib

rary

.seg

.org

/