shear-wave amplitude anomalies

Shear-wave amplitude anomaliesin south-central Wyoming

ROBERT R. KENDALL J-MICHAEL KENDALLAmoco Production University of Leeds

Amoco recently acquired four sur-face-seismic shear-wave lines, fourmulticomponent VSP’s and threecrossed-dipole sonic logs in south-cen-tral Wyoming. A feature of this region isthat areas of increased gas productioncorrelate with areas of increased frac-turing. After seismic processing of theshear-wave lines, it was observed thatreductions in the relative amplitude ofthe slow shear-wave (S2) seismic sec-tion with respect to the fast shear-wave(S1) section correlate with areas ofgreater gas production. These ampli-tude anomalies are attributed toanisotropy due to the preferential align-ment of vertical fractures where S1 ispolarized parallel to the fracture az-imuth and S2 is polarized perpendicularto the fractures. Furthermore, preferreddirections of fast shear-wave polariza-tion, and hence fracture strike, derivedfrom crossed-dipole sonic logs andmulticomponent VSPs are in agreementwith the directions determined from thesurface seismic. Modeling of wavepropagation effects due to the aniso-tropy illustrates how enhanced openfracturing may cause shear-wave ampli-tude anomalies. Seismic waveforms are

sensitive to the fracturedensityand ori-entation and the natureof the materialwithin the fractures. The modelling alsoshows that the range of offsets used instacking shear-wave data must be care-fully selected. Locations of polarityreversals and the onset of critical reflec-tions can vary dramatically with frac-turing.

Introduction. Evidence of anisotropyin fractured hydrocarbon reservoirs hasthe potential to be a powerful interpre-tation tool for characterizing the inten-sity and orientation of fractures. Thepresence of these fractures can dramat-ically improve production rates, mostnotably in tight formations. Amoco hasacquired, processed and interpreted thepure shear-wave components of four 9-component (9-C) seismic lines thatwere acquired in south-centralWyoming (Figure 1). Additionally,crossed-dipole sonic logs and multi-component VSPs have been acquired,processed and interpreted.

The zone of interest, at approxi-mately 3 100 m, is a mostly shale-dom-inated sequence with interbeddedsandstone layers that are roughly 10 m

Figure 1. General location of our study area.

thick. Deposition of the sandstone (off-shore barrier bar) represents a short-lived regressive pulse within a moredominant transgressive system. Thereservoir is primarily gas filled andanalysis of drill core has shown strongregional variations in the degree of frac-turing within these reservoirs.

A medium with preferentiallyaligned vertical fractures will exhibitazimuthal anisotropy (Crampin, 1984,Geophysical Journal of the Royal As-tronomical Society). The fast shear-wave will be polarized parallel to thecrack faces, while the slower shear-wave is polarized orthogonal to thecrack faces. Fracture orientation withinthe sandstone was determined from thefast shear-wave polarization directionusing the Alford technique. A layerstripping technique (Thomsen et al.,“Layer str ipping for azimuthalanisotropy of reflection seismic data,”in publication) was used to verify thatthe preferred direction was constant(N45°E 15°) with depth. These rota-tion angles were confirmed through theinterpretation of the crossed-dipole dataand multicomponent VSPs (N60°E

15°). Evidence of fracturing was in-ferred from variations in amplitudestrength between the fast and slowshear-wave components using a methodsuggested by Thomsen (“Reflectionseismology over azimuthally aniso-tropic media,” GEOPHYSICS, 1995) andapplied by Mueller (“Prediction of lat-eral variability in fracture intensityusing multicomponent shear-wave sur-face seismic as a precursor to horizontaldrilling in the Austin chalk,” Geophysi-cal Journal International). Here wefind that shear-wave amplitude anom-alies (SWAA) coincide with productionanomalies that are an order of magni-tude more productive than productionin the surrounding regions. Drilling hasconfirmed the interpreted fracture az-imuth (core measurements show thatthe fractures strike N60°E).

In order to better interpret the data,

AUGUST 1996 THE LEADING EDGE 913

seismic; crossed dipole sonic log; andmulticomponent VSP.

Figure 2. Final rotated shear-wave sections. The upper section is the fast or S1section while the lower is the slower or S2 section. Note the amplitude reduc-tion on the slower section at the identified level towards the left of the section(between arrows). Production coming from the zone located between the ar-rows is an order of magnitude greater than the surrounding regions.

Figure 3. A typical cross-dipolesonic log from the study area.From left to right the five tracksare: (1) the minimum and maxi-mum off-diagonal shear-wave en-ergy and depth. Areas of greatestanisotropy occur where there isthe greatest separation betweenthe maximum and minimumcurves (areas with a lot of greenfilling); (2) the gamma-ray log,well azimuth, well deviation andwell diameter (caliper); (3) theazimuth of the fast shear-wavepolarization; (4) the slownesses ofthe fast and slow shear-waves andslowness-based and interval-transit-time-based percentageanisotropy (see color scale attop); (5) the fast and slow shear-waveforms for one receiver pairand the analysis window (dashedlines).

we have modeled the seismic re-sponse of reflections from ananisotropic sandstone layer withpreferentially aligned fluid and gasfilled cracks. The modeling resultsconfirm the interpretation of theamplitude anomalies on the surfaceseismic and also reveal some of thesubtle variations that may occur dueto variations in fracture intensity,crack aspect ratio and crack fill ma-terial.

Data analysis. This involves threetypes of data: shear-wave surface

Shear-wave surface seismic. Thesedata were acquired using four horizon-tal motion vibrators (Amoco’s rotatingbaseplate Mertz 18’s). The signal tonoise ratio (S/N) was improved by run-ning the vibrators in a box pattern, asopposed to an in-line configuration. Itwas also observed that using more shortsweeps, as opposed to fewer long ones,improved the S/N. The optimal sweepturned out to be 6-36 Hz, linear over 16seconds, with a 7 s record length. Groupand shot intervals were 37 m, with theshots between the groups. An SGRrecording system was used to record the660 channels, producing a far offset of4000 m in a split-spread configuration.1 lo-fold coverage was achieved.

These data were processed using atrue amplitude approach, since we wereultimately interested in comparing am-plitudes on the fast (S1) and slow (S2)sections. Refraction statics were calcu-lated on the pure cross-line component,due to the minimal P-wave first arrivalenergy on these records, and that solu-tion was then applied to each of the fourcomponents. Residual statics were cal-culated and applied independently foreach component. Velocity analysis, aswith P-wave processing, was crucialand an average velocity function for thefour components was used for normal-moveout (NMO) corrections. The datawere stacked by limiting the angle ofincidence to less than 30” (angle stack),thus ensuring that only precriticalarrivals were used (see modeling sec-tion). A Radon filter proved to be veryeffective for removing most of the co-herent noise that often dominated thedata. Finally, we applied Alford rotationanalysis to determine the preferred di-rections and the data were then rotatedto that angle for the amplitude analysis.Figure 2 shows the data quality as wellas a typical SWAA. For this line, the op-timal rotation angle suggested a fracturestrike of N45°E. Three wells are markedin Figure 2. Two are above a SWAAanomaly (between the arrows on the leftside of Figure 2) and one is not (rightside of Figure 2). production in the twowells which lie above the SWAA wasan order of magnitude more than that inthe well that does not lie above theSWAA.

Crossed-dipole sonic log. Crossed-di-pole data were acquired and processed

914 THE LEADING EDGE AUGUST 1996

Gulf Publishing Company).

Figure 4. The four components of the VSP before (4a) and after (4b) Alford ro-tation into the direction of fast and slow shear-wave polarization.

Multicomponent VSP. Four wells inthis area were used to acquire multi-component VSPs. Azimuthal-aniso-tropy analyses using Alford rotationsand fast and slow shear-wave arrivaltime differences (time-delay) were usedto characterize the anisotropy. Figure 4(a, b) shows the four components of theVSP data from one well before andafter rotation. Clearly the S/N in thesedata is very high. The rotation analysisattempts to rotate the shear-wave datainto the fast and slow polarization direc-tions (main-diagonal components),thereby minimizing the energy on theoff-diagonal components. The postrota-tion data (Figure 4b) show a good mini-mization of the first arrival amplitudeson the off-diagonal components, withinthe identified window, confirming ourinterpreted azimuth. Figure 5 (a, b)shows the polarization angle (azimuth)and time delay respectively. The rotationanalyses and time-delay analyses ofthese data was done using the EAP soft-ware SWAP (Li and Crampin, GEO-PHYSICS, 1991). SWAP also has thefunctionality to do Winterstein (Winter-stein and Meadows, SEG Expanded Ab-stracts, 1990) layer stripping to isolatevariations in anisotropy with depth.These data showed a consistant fast-shear-wave azimuth with depth.

The polarization angle N60°E (Fig-ure 5a) is consistent with depth. There issome scatter in the azimuth, most no-tably in the shallower section (<1920m). The time-delay curve can be brokendown into 4 zones (Figure 5b). The ma-jority of the traveltime delay (henceanisotropy) is accrued in the first zone,from 0- 1920 m. This zone includes theweathering, subweathering and gener-ally less competent rock. Others(Kramer and Davis, 1992, TLE) havedocumented this shallow zone as beinghighly anisotropic. Zone 2 has been in-terpreted to be isotropic and extends toa depth of approximately 3 100 m. Zone3 is between 3100 and 3360 m and in-cludes the target interval. The per-centage anisotropy in this zone is about4% based on the time delay informationon Figure 5b. This is in agreement withthe 4-8% anisotropy that is observed forthe fracture zones on the cross-dipolesonic log.

Summary of SWAA method. Figure 6illustrates a fracture model that couldcreate a SWAA. We consider two near-normal-incident reflections from twopoints (marked 1 and 2). The sandstone

Figure 5. Fracture azimuth (5a) and time delay (5b) as determined using theprogram SWAP, Note the two predominant zones (1 and 3) of increased time-delay which indicates anisotropy and the more-or-less depth consistent azimuththat is clustered around N60°E.

has a higher impedance than the overly-ing shale and we assume the shale isisotropic. The reflection coefficient fororthogonally polarized shear waves willbe the same at location 1. That is, theshear wave polarized in the plane of thepage (in-line) will have the same reflec-tion coefficient as the shear wave polar-ized perpendicular to the plane of thepage (crossline). In contrast, the reflec-tion coefficients for the two orthogo-nally polarized shear waves will bedifferent at location 2. The shear wavethat is polarized in the in-line direction(perpendicular to the cracks) will detect

a less rigid, therefore lower velocity,sandstone at location 2 than that de-tected by the shear wave that is polar-ized in the cross-line direction (parallelto the cracks). Therefore, the shearwave that is polarized in the cross-linedirection will have a greater reflectioncoefficient (bigger amplitudes) than theshear wave that is polarized in the in-line direction. In the next section weshow that crack intensity, crack aspectratio and crack fill material all affectthis decrease in amplitude. Further-more, the polarization of the reflectedshear- waves at location 2 will be af-

Figure 6. Schematic diagram of a fracture model which could generate the ob-served shear-wave amplitude anomaly. Note that the ray propagates in a solelyisotropic media. For rays that propagate into the fractured region, we refer tothe in-line and cross-line polarizations as S2 and S1.


fected by the strike of the fractures, re-gardless of the source polarization. It isperhaps appropriate to refer to the ro-tated sections as S1 and S2, even thoughthe waves propagate entirely through anisotropic medium, since the shear-wavepolarizations are affected by the frac-ture strike.

Figure 2 shows an example of thisamplitude difference on a seismic sec-tion while Figure 7 illustrates the ampli-tude difference using an amplitudeextraction technique. The reflector is la-beled consistently on both figures as S 1and S2. Since the bed thickness is ex-tremely small compared to the totaldepth, traveltime differences betweenS1 and S2 accrued in the anisotropicthin-bed will be below the resolution ofthe seismic data. In contrast, the ampli-tude extraction technique is very sensi-tive to the anisotropy, even with thesenoisy data and the thin beds (Figures 2and 7). In other words, the amplitudetechnique (using SWAAs) has bettervertical resolution than do traveltimemethods.

Modeling. In order to interpret theSWAAs in terms of fracture intensityand orientation, one must first under-stand the nature of wave propagation insuch anisotropic structures. Modelingof these data has been done with a ray-based program (ATRAK) which han-dles general anisotropy, 3-D structureand nonplanar layers (Guest andKendall, 1993, Canadian Journal ofExploration Geophysics). Our resultsshow that the anisotropy can signifi-cantly affect the shear-wave ampli-tudes. Furthermore, the locations ofcritical points and nulls in reflection co-efficients are very sensitive to the natureof the anisotropy. Our model, which isbased on well results, has a thin (10 m)anisotropic layer which starts at a depthof 3095 m (Figure 8). The anisotropy isdue to the preferential alignment of thincracks within an otherwise isotropicsandstone. For simplicity, we have as-sumed that the surrounding rock isisotropic and has a linear velocity gra-dient with depth. Any anisotropy withinthe surrounding shales has been ne-glected. It is assumed that the fracturedsandstone exhibits azimuthal aniso-tropy (hexagonal symmetry) with the(horizontal) symmetry axis oriented inthe in-line direction (the fractures arestriking in the cross-line direction).

We adopt the approach of Hudson(“Overall properties of a cracked solid,”

Figure 7. Extracted amplitudes as identified on Figure 2. The arrows indicatethe region between which the anomaly, that is identified on Figure 2, is located.

Figure 8. Shear wave raypaths through the model of a thin fractured(anisotropic) sandstone layer embedded in an isotropic background medium.Note the splitting of rays at further offsets as they reflect from the top and bot-tom of the anisotropic layer.

Mathematical Proceedings of the Cam-bridge Philosophical Society, 1980) toestimate effective elastic constants ofthe fractured sandstone. Hudson’s ex-pressions are accurate to second orderin crack density, but this theory does notaccount for the effect of the cracksbeing hydraulically interconnected. Es-timates of the fracture attributes for ourarea of study have been obtained fromgeologic analyses of drill core. The de-gree of fracturing, which partially con-trols production in this area, is highlyvariable. As end-member models weconsider cases where the fracturingvaries from extreme to nonexistent(isotropic) and cases where the frac-tures are gas filled and fluid filled. Forthe fractured cases we consider a frac-ture density of 0.15 and a crack aspectratio of 0.00001, estimates of whichwere obtained from drill core.

It is instructive to first inspect theslowness and velocity surfaces for therange of anisotropy we are considering(Figure 9). Figure 9a corresponds to the

case for a wave propagating in theisotropic sandstone (not fractured) at lo-cation 1 in Figure 6. Figure 9 (b, c) cor-responds to plausible wavefronts in theanisotropic sandstone (fractured) at lo-cation 2. Note that these wavefronts arespherical for the isotropic case but notfor the anisotropic case and that twoquasi-orthogonally polarized shear-waves will propagate in the anisotropicregion. For wave propagation in the ver-tical direction ( x3), the fracturing haslittle effect on the qP-wave and trans-versely-polarized shear-waves (S1 orcross-line shear-wave). Recall that ac-cording to our model, the S1 waves arepolarized parallel to the fractures. Onthe other hand, the radially-polarizedS2 or in-line shear-waves (those wavespolarized perpendicular to the frac-tures) are much slower in the vertical di-rection. For wave propagation along thesymmetry direction ( x 1), both shearwaves propagate with the same velocityand are slow relative to the isotropiccase. The qP-waves in this direction are


Figure 9. Vertical sections of the slow-ness surfaces and polarizations(LHS) and wave (group velocity) sur-faces (RHS) for a medium withvertically aligned fractures withcrack-face normals oriented alongthe x 1 axis. The fracture density is0.15 and the crack aspect ration is0.00001. (a) No fracturing (isotropic);(b) gas filled cracks; (c) fluid filledcracks.

much slower for the gas-filled case thanthe fluid-filled case (compare Figures9b and Sc). Our numerical experimentsshow that the variation in S l-wave(cross-line) velocity from vertical tohorizontal propagation directions is notvery sensitive to the crack aspect-ratioor the inclusion material. In contrast theS2-wave (in-line) is very sensitive tothese parameters.

Wavefront folding, or triplications,will develop as the cracks become

“skinnier.” The fluid-filled cracks aremore sensitive to this than the gas-filledcracks.

Figure 10 (a, b, c) shows synthetic S-wave seismograms for reflections fromthe top of the thin sandstone layer. Inkeeping with the actual field recording,waveforms are modeled for offsets outto ~ 4 km. A 20 Hz Ricker wavelet isused. Despite traveling solely throughisotropic media the waveforms are verysensitive to anisotropy in the underlyingsandstone layer. This is due to polaritydependent variations in the velocitycontrast across the isotropic/anisotropicinterface. In the isotropic case the S-velocity increase across the interface isfrom 2000 m/s to 2530 m/s. Figure 10ashows that for near-normal-incidencethe cross-line and in-line amplitudes areequal in the isotropic case as expected.There is a null in the in-line reflectioncoefficient for offsets near 1600 mwhile there is one near 3 100 m, near thecritical angle, for the cross-line case.Figure 10 (b, c) shows the waveformsfor the anisotropic cases (gas and fluidfilled cracks respectively). The cross-line waveforms for near vertical arrivalsare essentially the same as those for theisotropic case. In contrast, the in-lineamplitudes are roughly half those of thecross-line case. This is due to the di-minished velocity contrast across theinterface due to the aligned fractures.

Waveform variations with offset forthe anisotropic cases are quite differentfrom those for the isotropic case. Incomparison, the point of critical reflec-tion for the cross-line-waveforms ismoved to further offsets because thecross-line velocities decrease in theanisotropic layer with increasing angleof incidence on the boundary. This isapparent on the wave surfaces if oneconsiders variations in wave propaga-tion in the region between the x 3 andx 1 directions (Figure 9). The in-linewaveforms are sensitive to the nature ofthe crack fill material. In comparison tothe isotropic case, the null in reflectionsfor the gas-filled case is at further off-sets while the null for the fluid filledcase is at nearer offsets. At far offsetsthe amplitudes for the gas-filled caseare much smaller than those for theother cases. The in-line amplitudes forthe fluid-filled case are almost as highas those for the isotropic case. This isdue to the high velocities near theemerging triplication on the wave sur-faces in the region 45° from the vertical(Figure 9c). In contrast, the velocities

Figure 10. Synthetic S-wave seismo-grams for reflections from the top ofthe thin fractured sandstone layermodelled using the program ATRAK.The synthetics are generated usingray theory and a 20 Hz Ricker wave-let. The cross-line and in-line recordsare shown for: (a) no fracturing in thesandstone layer (isotropic); (b) gasfilled cracks; (c) fluid filled cracks.

for the gas-f i l led case decreasesmoothly towards the horizontal (Fig-ure 9b). Our numerical experimentshave also shown that this in-line wave-form behavior at far offsets is also de-pendent on the crack aspect ratio.

Figure 11 (a, b, c) shows the com-


posite waveforms for reflections fromboth the top of the thin sandstone layerand the bottom of the layer. The layer isroughly l/10-wavelength in thicknessso therefore seismically detectable, butthere will be waveform interference ef-fects. All of the features observed forthe single reflection and discussedabove are still apparent. The only no-

table difference is that the in-line shear-wave (S2) amplitudes for the gas-filledcase are even weaker at the distant off-sets. Traveltime measurement of shear-wave splitting is not a diagnostic toolfor detecting fractures in such a thinlayer. Not only is the degree of splittingminuscule, the reflections for the topand bottom of the layer are indistin-guishable for the wavelengths we aredealing with. Instead reductions in re-flection amplitudes are strongly indica-tive of the fracturing. This is inagreement with the observed field dataand provides a plausible explanation(though nonunique) for the observedenhanced production within the regionsof SWAAs.

The reflection data from Figure 11(a, b, c) were NMO corrected andstacked (Figure 12a, b). Figure 12ashows the full-range stacks using all ofthe traces (o-3850 m) for the cross-line(S1) and in-line (S2) records. Figure12b shows the range-limited stacksusing offsets equivalent to those used inthe processing of the field data (o-3350m). The field data were angle-stackedusing only angles of incidence less than30”. This corresponds roughly to offsetsof 3350 m for a reflector at a depth of3 100 m. The range-limited stacks showvery clear SWAAs for both the gas- andfluid-filled fracture cases. The S2 am-plitudes are roughly half the S1 ampli-tudes for the anisotropic models. In

contrast, the S1 and S2 amplitudes arenearly the same for the isotropic case.The full-range stacks include the high-amplitude post-critical reflections at thefarthest offsets. The difference betweenthe fluid-filled case and the isotropiccase is no longer so obvious. The gas-filled case still shows the S2 amplitudesto be much smaller than the Sl ampli-tudes. Only precritical reflectionsshould be used in the stacks to avoid themore dominant high-amplitude andphase distorted postcritical reflections.

Conclusions. In our study area, lateralamplitude variations, as observed pre-dominantly on the slow-shear (S2) seis-mic sections, correlate with areas ofgreater gas production. Furthermore,preferred directions of fracturing de-rived from measurements of anisotropyon crossed-dipole sonic logs and multi-component VSPs are in agreement withface seismic. Anisotropic seismic mod-eling of these data illustrates that en-hanced open fracturing may be causingthe shear-wave amplitude anomalies.

Our waveform modeling has pro-vided a qualitative explanation for theobserved SWAAs in the data. Thewaveforms are very sensitive to thefracture density and orientation whichis valuable information for a drillingprogram. For reflections from an iso-tropic layer embedded in an isotropicmedium there is no difference between


the in-line and cross-line amplitudes inthe range-limited stacked shear-wavesections. In our modelling, this is nottrue for the full-range stack. In the caseof a fractured layer, we have shown thatthere can be large reductions in the re-flection amplitudes for the shear-wavesection that is polarized perpendicularto the fracture azimuth relative to theshear-wave section that is polarized par-allel to the fracture azimuth. It is im-portant to note that the degree oftraveltime separation caused by shear-wave splitting in the thin anisotropiclayer is less than a millisecond and istherefore not a useful tool for detectinganisotropy in such a thin layer. It is thewaveform effects which are diagnosticof the anisotropy.

There are many interesting aspectsof rock-fracture characterization thatwe have not addressed. The effects ofhydraulically interconnected cracks andequant porosity (Thomsen, “Elasticanisotropy due to aligned cracks inporous rock,” Geophysical Prospecting,1995) will be explored in the future. Wehave also not considered the effects ofmore complicated elastic symmetries.

For example the cracks and layeringmay not be orthogonal in orientation orthe matrix rock may have an intrinsicanisotropy.

Linked modeling and data interpre-tation provide valuable insights into thenature of seismic anisotropy and rock-fracture characterization. Obviouslydata interpretation will improve as webetter understand the nature of wavepropagation in anisotropic fracturedmedia. Data quality is a long way frombeing able to detect the somewhat sub-tle waveform features which are sensi-tive to crack aspect ratio and inclusionmaterial. Shear-wave AVO and AVAanalyses hold potential for obtainingsuch information. Finally, the applica-tion of these ideas to the 3-D domainwill add further insights.

Acknowledgments. We thank Steve Clawsonfor providing the details of the core analysisand production histories, Mike Mueller,Leon Thomsen, and Sean Guest for con-structive comments on the manuscript andMichele Simon for the editing of this issue.We also thank Amoco for allowing us to pub-lish these results.

Robert R. Kendall received his bachelor’s ingeophysics from theUniversity of Calgaryin 1988 and his mas-ter‘s in geophysicsfrom the ColoradoSchool of Mines in1992. He worked forSolid State Geophysi-cal from 1988 to 1990as a geophysicist andparty manager. In

1992 he joined Amoco and now works in theseismic processing group (STAT).

J-Michael Kendall received his bachelor’sand doctorate fromQueen‘s UniversityKingston, Ontario in1984 and 1991 re-spectively. He workedfor Chevron Canadafrom 1984 to 1986. Hewas a postdoctoral

, fellow at ScrippsInstitution of Ocean-ography from 1992 to

1993 and was an assistant professor at theUniversity of Toronto from 1993 to 1995.He is currently a university research fel-low/lecturer at the University of Leeds inEngland.


shear-wave amplitude anomalies

Documents