Chapter 14

Case Studies of Multicomponent Seismic Data forFracture Characterization: Austin ChalkExamples

Xiang- Yang LiEdinburgh Anisotropy ProjectBritish Geological SurveyEdinburgh, Scotland, U, K.

Michael C. MuellerAmoco EPTGHouston, Texas, U.S.A.

AbstractShear wave studies of multicomponent seismic data were done along the Austin Chalk trend in

Texas. Six surface seismic lines of four-component shear wave data from Pearsall and Giddingsfields and three zero-offset vertical seismic profiles (VSPs) from three sites with different produc-tion rates were studied to demonstrate the applications of shear wave splitting for fracture reser-voir delineation. The three seismic lines (l-3) from Pearsall field formed a classic experiment forstudying shear wave splitting. They have different (line) azimuth with respect to the regional frac-ture strike (parallel, perpendicular, and at -4O”, respectively), are in areas with different hydrocar-bon production, and have different split shear wave behavior. The anisotropy along line 1 is small,which correlates with the absence of nearby commercial production. There is an increasing trendin anisotropy from line 2 to 3, which correlates with line 2 being close to and line 3 being withinthe producing Austin Chalk acreage. The trend of anisotropy variation along line 3 also correlateswith the distribution of producing oil wells along that line. In particular, production from the threehorizontal wells (Wl-W3) drilled nearby correlates with the variation in shear wave polarization,time delay, and amplitude. Line 4 from Giddings field had a horizontal well drilled on it, and mudlogs were obtained for identifying fracture zones. The fracture swarms identified by the mud logsare coincident with the dim spots identified from the section of the slow split shear wave. Lines 5and 6, also from Giddings field, demonstrate classic S, dim spot and S, versus S2 mistie behavior,respectively.

The three VSPs, from three wells with different production rates, show different shear waveresponses. VSP 1, from a nonproductive well, shows minimal shear wave time delay and no ampli-tude anomalies. VSP 2, from a water-producing well, shows some amount of splitting but noanomalies in shear wave attributes. VSP 3, from an oil-producing well, shows clear shear wavesplitting and anomalies in shear wave amplitudes. Amplitude corrections are necessary for pre-serving and recovering shear wave anisotropy information associated with the target, and the lin-ear transform technique (LTT) simplifies the processing sequence for examining anisotropy inshear wave reflection data. Stacked polarization logs can be used for identifying local variations inpolarization (often associated with local variations in crack geometry) and for better imaging ofsubsurface structure, as in line 3.

337

338 Li and Mueller

IntroductionThe study of seismic waves propagating through the

earth has greatly improved our understanding of theearth’s interior and has developed into one of the mostimportant tools for petroleum exploration. However, un-til the 1970s most theories were developed mainlyaround the propagation of P-waves in isotropic solids(Aki and Richards, 1980; Crampin, 1981). Most appli-cations were limited to P-wave energy recorded onsingle-component vertical geophones or pressure-sensi-tive hydrophones (Tatham and McCormack, 1991).Compared to shear wave recording, P-wave recording ischeaper and more robust and often yields better structur-al imaging of the subsurface. Moreover, the effects ofanisotropy on P-wave are often small and difficult to de-tect in one-component P-wave data (Winterstein, 1990).Thus, the relevance of shear wave and seismic anisotro-py to hydrocarbon exploration and development was notwidely recognized.

Gardner and Harris (1968) and Tatham and Stoffa(1976) predicted significant anomalies associated withgas saturation in the velocity ratio of P-wave to shearwave (VJV,) and identified V,lV, ratios as a possible in-dicator of lithology and even porosity variations. Keithand Crampin (1977) and Crampin (198 1) showed thepotential importance of shear wave anisotropy and iden-tified shear wave splitting (birefringence) as a diagnos-tic feature of shear wave anisotropy, in which the polar-ization, time delay, and amplitude of shear waves are re-lated to the in-situ stress orientation and the alignmentand intensity of fractures and cracks. These new insightsinto shear wave behavior, together with the emergenceof digital processing, opened new aspects in shear waveexploration. In the 1970s this led to the development ofSH-wave survey (one-component horizontal SH sourceand receiver) for studying Vdv, ratio and, in the 1980sto the acquisition of multicomponent seismic data usingboth vertically and horizontally polarized sources andmulticomponent receivers for studying shear waveanisotropy.

Over the past decade, the use of joint P- and S-wavesurveys and the analysis of VJV,, ratios have been wide-ly reported (e.g., Tatham, 1982, 1985; Ensley, 1984;Winterstein and Hanten, 1985). Tatham and McCor-mack (1991) have documented these applications. Ob-servations of shear wave anisotropy have also becomewell-documented in multicomponent seismic surveys(e.g., Crampin, 1985; Crampin et al., 1986; Alford,1986; Willis et al., 1986) and in earthquake recordings(Crampin et al., 1985). Crampin and Love11 (1991) havereviewed these developments. In the present study, we

focus on the use of multicomponent seismic data and theanalysis of shear wave anisotropy for applications to pe-troleum exploration.

It has long been speculated that the study of shearwave anisotropy can be used to characterize fracturedreservoirs (e.g., Crampin, 1987; Lewis et al., 1991).Higher oil production in fractured reservoirs is often as-sociated with higher fracture intensity and hence greatershear wave anisotropy. Thus, shear wave anisotropymay be directly correlated with oil production (Lewis etal., 1991; Mueller, 1991). Cliet et al. (1991) reported thecorrelation of shear wave anisotropy with oil productionrates in a multi-offset three-component vertical seismicprofiling (VSP) survey in the Romashkino field, Russia.Davis and Lewis (1990) and Lewis et al. (1991) alsocorrelated the degree of shear wave anisotropy with oilproduction in a three-dimensional, three-component re-flection survey in Silo field, Wyoming.

The Austin Chalk trend in Texas is a natural labora-tory for testing practical applications of shear waves inanisotropic rocks. This is because the chalk has little po-rosity, and vertical aligned fractures provide the majorpathway for oil. To analyze shear wave anisotropy in theAustin Chalk trend, Amoco Production Company ac-quired several lines of shear wave data, and severalstudies of shear wave anisotropy along the trend havebeen published. Alford (1986) established the presenceof shear wave splitting and an algorithm for accommo-dating it using data from the chalk trend. Thomsen(1988) suggested that reflection amplitude anomalies(between the split shear waves) constituted a sensitivemeasure of local difference in anisotropy, useful for de-tecting fractured reservoirs. Confirming this suggestion,Mueller (199 1) identified lateral fracture concentrationsfrom the differential amplitudes of split shear wavesfrom the Austin Chalk in the Giddings field of centralTexas. Li et al. (1993a) investigated variations in shearwave attributes with varying line azimuth and compareddifferent processing methods for preserving and enhanc-ing shear wave splitting. Yardley and Crampin (1993)studied three VSPs from different sites along the trendand examined the different response of shear wave split-ting for varying production rates.

Here we review these shear wave studies along theAustin Chalk trend. We outline the processing and inter-pretation methods for fracture reservoir delineation anddiscuss the underlying principles of shear wave splittingassociated with the chalk, including its physical proper-ties and seismic characteristics. We then present inter-pretations and illustrate how the fracture parameters canbe deduced from variations in shear wave attributes suchas polarization, time delay, and amplitude. We show that

Multicomponent Seismic Data for Fracture Characterization 339

the variations in anisotropy can be correlated with oilproduction in the study area and that the variation inshear wave amplitude can be correlated with fractureswarms encountered from horizontal drilling, as in-ferred from production records and mud logs.

Overview of the Shear WaveTechnique

When a shear wave propagates into an effectivelyanisotropic medium (e.g., resulting from aligned frac-tures), it necessarily splits into two modes that have dif-ferent polarizations and travel with different speeds. Fornear-vertical propagation, the faster split shear wave ispolarized along the fracture strike and the slower one ispolarized nearly orthogonal to it (Crampin, 1985). Fur-thermore, the time delay and differential amplitude be-tween the two shear waves are both proportional to theintensity of fracturing. This forms the basis for fracturedelineation using multicomponent shear wave data. Thissection first reviews the shear wave information contentin terms of shear wave polarization, time delay, and nor-mal reflectivity. We then discuss the development ofmulticomponent acquisition and processing techniquesnecessary for studying shear wave splitting.

Shear Wave Attributes

The characteristics of split shear waves diagnosticof fracture distributions are referred to as shear wave at-tributes. These include the polarization of the leadingshear wave, the time delay between the two shearwaves, the differential reflectivity at interfaces, and thedifferential attenuation and scattering along the raypath.In general, polarization and time delay are quantitativeattributes, but are difficult to measure in thin-layered orweakly anisotropic reservoirs. Differential reflectivity atnormal incidence is a qualitative attribute, but measur-able for thin-layered or weakly anisotropic reservoirs.An integration of all three attributes with other geologicinformation is needed for a reliable interpretation offractured zones. Here we briefly review the three mostwidely used attributes: the polarization, the time delay,and the reflectivity at normal incidence. We discuss theirrelative merits and compare their potential for detectingfractures.

Yardley et al. (1991) have provided a detailed dis-cussion on the viability of using shear wave amplitudeversus offset (AVO) for fracture detection. Although thetwo shear waves have different AVO signatures (the re-

flectivity of the SV-type wave may have a zero crossingat relatively small offset due to conversion), it is difficultto attribute this difference to fracturing because suchdifference in AVO signatures also exists for SV- and SH-waves propagating in isotropic media. The pattern of theAVO curves remains the same with increasing fracturingexcept the reflectivity at normal incidence. Yardley et al.(1991) concluded that for fracture detection, the differ-ential rejectivity at normal incidence is a first-order ef-fect and can be measured and interpreted reliably, whilethe ofSset-dependent rejiectivity is a second-order effectand more subtle and difficult to interpret.

Polarization

In theory, for near-vertical propagation, the leadingshear wave is polarized parallel to the fracture strike ofan effectively anisotropic medium when that mediumpossesses a single (or dominant) vertical, aligned frac-ture set. Thus, the polarization of the fast split shearwave can be used to infer the fracture strike of the medi-um through which it has propagated from seismic datarecorded on the surface or in boreholes. However, themain factor affecting the use of polarization informationat the target is the anisotropic effects in the overburden.A shear wave generally splits whenever it passesthrough an anisotropic medium. Thus, if the overburdenis anisotropic, the polarization recorded at the surface isdetermined only by the anisotropic symmetry in theoverburden; the anisotropic symmetry information inthe target is then concealed by that of the overburden.Considering that anisotropy is ubiquitous (Willis et al.,1986) and often seen in the overburden (Crampin andLovell, 1991), inferring the fracture orientation of thereservoir from surface recordings requires layer-strip-ping techniques (MacBeth et al., 1992; Li, 1995; Thom-sen et al., 1995a). Because of these complications, Yard-ley and Crampin (1991) suggested that VSPs may bebetter used to study shear wave polarizations at target.

Note that the layer-stripping technique was first de-veloped by Winterstein and Meadows (1991) for accom-modating shallow anisotropy in shear wave VSP data;Winterstein (199 1) and Meadows et al. (1994) suggestedthat the same techniques for VSPs could be equally ap-plied to surface seismic data. However, to best accom-modate the anisotropy effects in the overburden, a re-vised technique for surface data may be necessary withdifferent time shifts for the diagonal and off-diagonal el-ements, as demonstrated by MacBeth et al. (1992), Li(1995), and Thomsen et al. (1995a).

340 Li and Mueller

Time Delay

The time delay between the fast and slow split shearwaves depends on the propagation path length and di-rection, the symmetry of the medium, and the degree ofanisotropy. Because the degree of anisotropy is mainlydetermined by the fracture density in the media, timedelay is a valuable attribute for the inference of fracturedensity from a seismic section, and the fracture densitycan be quantitatively determined if the number of frac-ture sets and their orientation are known. Time delaysare usually measured from the stacked sections of thefast and slow shear waves. The advantage of using thetime-delay attribute lies in its simplicity. It is relativelyeasy to measure, and overburden anisotropy can be cor-rected for simply by taking the interval time delay be-tween the top and bottom reflections of the target if thefracture orientation does not change with depth. Layerthickness can be simply handled by normalization.However, for a thin or weakly anisotropic reservoir, in-terval time delays may be too small to resolve reliably inthe presence of noise, particularly if a significant part ofthe anisotropy is in the overburden and near the surface,as demonstrated by Li et al. (1993a). Also, time delayscontain no information about fracture orientations,which must be deduced from other attributes. Squires etal. (1989) and Lewis et al. (1991) presented examples ofinterpreting interval time delays from field data.

Normal Reflectivity

Thomsen (1988) suggested that the differential am-plitudes between fast and slow split shear waves can beused to identify fracture zones in stacked seismic sec-tions if these fracture zones are terminated at reflectingboundaries such as lithologic boundaries, as commonlyseen in outcrops. In such cases, the anisotropy caused bythe fractures tends to affect the slow wave (with polar-ization perpendicular to the fracture strike) more thanthe fast wave (with polarization parallel to the fracturestrike). The presence of the fractures lowers the velocityof the slow wave and causes a change in slow wave im-pedance contrast with respect to that of the fast wave,hence a difference in reflection strength in the stackedfast and slow sections occurs. This phenomenon is oftenreferred to as differential rejlectivity at normal incidencebecause stacked sections are often considered as seismicreflections at normal incidence. Note that anisotropycaused by macrofractures, whose scale length is compa-rable to wavelength, can affect both the fast and slowsplit shear waves (Liu et al., 1993).

Mueller (1991) first presented an example using

normal reflectivity to identify fracture swarms in theAustin Chalk in south Texas (the results of which are in-cluded in this paper), where the fractures are confined inthe brittle chalk (Corbett et al., 1987). The reflectivityattribute is of greater value in identifying the intenselyfractured zones in the seismic section rather than inquantifying the exact values of the fracture intensity, al-though methods have been published to quantify thefracture intensity from reflectivity (Spencer and Chi,199 1; Li and Crampin, 1993~). Because this qualitativeinformation is usually sufficient in most cases, the dif-ferential reflectivity at normal incidence is thus a usefulindicator of fracture swarms. The interpretation isstraightforward and similar to traditional amplitudeanalysis, and thin-layered or weakly anisotropic reser-voirs can be resolved. The main difficulty lies in main-taining and recovering the true amplitude informationduring processing.

Four-Component Geometry

The use and study of shear waves requires polarizedsources and receivers. In the middle 1970s when hori-zontal receivers and large horizontal motion vibroseissources became available, shear wave exploration andmulticomponent surveying began to develop. Up tonine-component data can be recorded using differentconfigurations of three-component sources and receiv-ers. These components consist of various combinationsof three mutually orthogonal source motions and receiv-er directions (combinations of vertical, Hl and H2, or z,X, and y sources and receivers) (Figure la). In practice,several configurations of sources and receivers havebeen used depending on the purpose of the survey.These include (1) three-component acquisition (verticalsource recorded by vertical, Hl, and H2 receivers, or zz,zx, and zy components) in VSPs for correlating seismichorizons (Hardage, 1991); (2) vertical and H2 (or SH-wave) surveying (zz and yy components) (ConocoGroup Shoot, Ensley, 1984; Winterstein and Hanten,1985; Tatham and McCormack, 1991); and (3) verticaland Hl (or SV-wave) surveying (zz and zx components)for mapping geologic structure and lithology (Dohr,1985).

In the 1980s following the completion of the Cono-co Group Shoot in 1978, Amoco Production Companyacquired four-component (xx, xy, yx, and yy) and five-component (XX, xy, yx, yy, and zz) surveys (Alford, 1986;Willis et al., 1986). The purpose of these surveys was toinvestigate whether shear wave splitting could be usedfor reservoir characterization. This configuration hasseveral distinct features. First, under the assumption of

Multicomponent Seismic Data for Fracture Characterization 341

SourcesX Y

Z

X

Y

(b)

Fig. 1. Multicomponent geometry and coordinatesystem. (a) Full nine-component geometry: threeorthogonal sources (columns) and three orthogonalreceivers (rows). The shaded area is the five-componentgeometry (Alford, 1986) examined here. (b) Coordinatesystem: x represents the in-line source and geophonedirection; y, the cross-line source and geophone direction;S,, the fast split shear wave; S,, the slow split shear wave;and a, the S, polarization azimuth as measured from thex direction.

weak anisotropy, the vertical or P component (ZZ com-ponent) is decoupled from the four horizontal or shearwave components (XX, xy, yx, and yy). Thus, processingof five-component data can be separated into two parts:processing of the zz component and processing of thehorizontal or shear components. Second, the four hori-zontal components allow unique determination of thefast and slow split shear waves (Alford, 1986; Thomsen,1988; Li and Crampin, 1991 a, 1993a), which is dis-cussed later in this section.

In the examples shown in this paper, all surface datawere acquired using four-component geometry. Thefour-component shear wave survey with two horizontalsources and two horizontal receivers, as shown inFigure 1 a (shaded), form a data matrix of traces d(t):

d( t ) = xx(t) xy (t)’yx (f) YY (f>I (1)

where the left column, xx(t) and yx(t), represents the in-line (x-axis) traces from the in-line and cross-line sourc-es, respectively, and the right column, xy(r) and yy(t),represents the cross-line (y-axis) traces from the in-lineand cross-line sources, respectively. This data matrixconcept is important for processing and interpretingshear wave splitting. From here onward, the phraseshear wave data will refer to the data matrix unless oth-erwise specified.

Processing Techniques

The purpose of processing multicomponent shearwave data is to preserve and recover the three major at-tributes (polarization, time delay, and principal reflectiv-ity). The shear wave data matrix represents a tensorwavefield, and conventional P-wave data processingmethods are not directly applicable. The basic approach,as reported in the literature, is to separate the tensorwavefield into the principal vector wave fields. By re-stricting oneself to the shear wave plane of polarization(assuming no change along raypaths), the vector fieldscan then be processed separately by the conventionalscalar method, including statics correction, velocityanalysis, normal moveout correction, stacking, and othersignal enhancing routines such as f-k filter. These con-ventional methods are well established for P-wave dataprocessing (Yilmaz, 1987). The approach here is basedon the assumption that polarizations do not change withdepth or offset, and that once the polarizations are sepa-rated, anisotropy effects such as varying velocity withoffset on the principal split shear waves are negligibleover small spreads (Li, 1992; Li and Crampin, 1993b).With this approach, the separation of the split shearwaves in the recorded data matrix is the crucial step inthe processing sequence that allows many of the con-ventional scalar processing steps to be applied to eachcomponent separately. Ultimately, only a few steps, suchas those associated with the accommodation of anisotro-py (rotations) and near-surface distortions, are handledwith a tensor or vector wavefield (multicomponent) ap-proach.

For azimuthally anisotropic media with a uniformsymmetry axis, Alford (1986) and Thomsen (1988) pro-posed a rotation algorithm that rotates the source and re-ceiver axes numerically to minimize the energy in theoff-diagonal elements (xy and yx components) of thedata matrix. Similar results can be obtained by rotating

342 Li and Mueller

analytically (Murtha, 1988; Lewis et al., 1991). Li andCrampin (I 991 a, 1993a) suggest a linear transformtechnique (LTT) as an alternative to the rotation tech-nique. The LTT contains four linear transforms, whicheffectively recast the complicated shear wave motionsin the recording coordinate system into linear motions inthe transformed coordinate system. Then, the polariza-tion estimation and split shear wave separation can beeasily made from these linearized components. This isequivalent to a singular value decomposition of the datamatrix. In principle, all these methods assume that theamplitudes of the two split shear waves are eigenvaluesof the data matrix.

Another major challenge in processing the splitshear waves involves near-surface and overburden am-plitude corrections to recover the shear wave informa-tion associated with the target zone. The shear wave isinevitably affected by acquisition errors, near-surface ir-regularities, and overburden anomalies. Winterstein andMeadows (199 1) developed a layer-stripping algorithmfor the VSP context and applied it to reflection data(Winterstein, 1991). MacBeth et al. (1992), Li (1995),and Thomsen et al. (1995a) modified this stripping pro-cedure for reflection data. Li (1994) presented anotherprocedure for amplitude correction in multicomponentshear wave data which includes a prestack surface-con-sistent correction and poststack overburden scaling.

The data examples presented here have been pro-cessed using different separation algorithms. For com-pleteness, we cover some basic equations for processingsplit shear waves.

Multicomponent Amplitude Corrections

According to Li (1994), factors affecting the targetamplitudes in the multicomponent data can be dividedinto two groups: one related to the surface (or near sur-face) and the other to the subsurface. The surface-relat-ed group includes source and receiver distortions due tointeractions with the near-surface. These effects can becorrected for by using a modified surface-consistentprocedure for multicomponent seismic data before sepa-ration of split shear waves and stacking. The subsurface-related group includes attenuation, scattering, geometricspreading, anisotropy, and other undesirable wavefieldproperties in the overburden. These complications canbe corrected by an overburden-correction scheme afterseparation of the split shear waves and stacking. Notethat the effects of an anisotropic overburden may intro-duce different moveout velocities and AVO responsesfor the fast and slow split shear wave. Thus, differentstacking velocities may be needed for stacking the fast

and slow split shear waves; different amplitude scalingfactors can be applied to correct the anisotropic AVO re-sponses.

Assuming the earth is a linear system for seismicwaves satisfying the convolution model (MacBeth et al.,1994), we have, in the time domain,

d ( t ) = 0(t)*g(t)*m(t)*s(t), (2)

where o(t) is a two-by-two matrix representing the offsetresponses for the four components. For a data matrixcontaining mixed modes of split shear waves, o(t) canbe assumed as a scalar o(t) that is the same for all fourcomponents, and for data matrix containing separatedsplit shear waves, o(t) is a diagonal matrix representingdifferent offset responses for the fast and slow splitshear waves. Also, g(t) is a diagonal matrix representingthe receiver geophone responses, m(t) is a two-by-twomatrix representing the subsurface response for shearwave splitting, and s(t) is a diagonal matrix representingthe source responses. The symbol * denotes convolu-tion.

Equation (2) is a direct extension of Taner and Koe-hler’s equations for one-component P-wave seismicdata (Taner and Koehler, 1981). In a similar way andwith similar assumptions, one does not need to computethe complete frequency-dependent surface (or near-sur-face) response. Moreover, amplitude corrections can beimplemented by multiplying the seismic trace by a sca-lar, which is equivalent to adding a constant value to thelog of amplitude spectrum. This procedure is similar toconventional P-wave processing (Taner and Koehler,1981) although there are two source scalars, two geo-phone scalars, and two (for separated split shear waves)or one (for mixed modes) offset scalars. Statistical solu-tions of o(t), g(t), and s(t) can then be obtained using theleast-squares method (see Li, 1994, for details). Thus,the subsurface medium response m(t) can be obtained.We then proceed to separate the split shear waves fromthe medium response m(t), or the data matrix after sur-face-consistent deconvolution. In the following sections,we use D(t) (note capital letter) to represent the data ma-trix, after surface-consistent correction, that is, D(t) isthe same as m(t) after surface-consistent correction.

Separation of Split Shear Waves

Assuming orthogonally polarized split shear waves,we can simulate the medium response of shear wavesplitting by a two-component eigensystem, with theeigenvectors representing the polarizations of the two

Multicomponent Seismic Data,for Fracture Characterization 343

split shear waves and the eigenvalues representing theiramplitudes. For the four-component geometry (XX, xy,yx, and yy) shown in Figures la and lb, in the time do-main, the medium response for shear wave splitting (orthe data matrix after surface-consistent correction) canbe written as

D(t) = C(a)A(t)C~(a), (3)

where superscript T is the transpose operator, a is thepolarization azimuth of the fast split shear wave(Figure lb), D(t) is the medium response for shear wavesplitting or the data matrix after surface-consistent de-convolution using equation (2), C is the orthogonal rota-tion matrix, and A(t) is the diagonal transfer function forthe two split shear waves S, and S, (also see Alford,1986). We then have

C(a) = cosa sina .[ I3-sina cosa

A ( t ) = s, (t> 0i 10 s,(t) *(4)

Note that the orthogonality of split shear waves is strict-ly true only for phase propagation and is a good approx-imation only for near-vertical group propagation(Crampin, 1981; for nonorthogonal split shear waves,see Li et al., 1993b). We stress again that equation (3)holds only if the source and geophone responses havebeen carefully compensated by surface-consistent de-convolution and only for near-vertical propagation oversmall spreads. If we take the inverse of the rotation ma-trix, equation (3) becomes

A ( t ) = C~(cc)D(t)C(a). (5)

Equation (5) forms the basis of Alford (1986) rotation.The data matrix can be rotated numerically using a se-ries of angles. The angle that minimizes the energy ofthe off-diagonal elements of the rotated data matrix isthe polarization azimuth a.

Another way to determine the rotation angle and theamplitudes of the split shear waves is to use the LTT (Liand Crampin, 1993a). The four linear transforms are de-fined as

Substituting equation (3) into equation (6) and makingsome manipulations, we have

L 1SW x(t) =rl (t> S(t) 1. Icos2a 0[ I[ [Sl (t) -S2(t)] 0 1 .sin2a 1 0 [Sl (t) +S2(t)]

Thus, the time series S,(t) and S,(t) are separated fromthe static geometry factors a after transformation; thetransformed vectors [g(t) q(t)]’ and [x(t) L(t)lT areeigenvectors representing linearly polarized motions inthe time domain in the displacement plane; the rotationangle a can be determined from vector [c(t) q(t)lT as theJacobi rotation angle:

1 r 2~5(t+wl(t+Q 1a = i tan-l

4Lc ,(&+t)) -T-p(t+T)] . (8)

r 1

The summation is over a chosen time window, orthe entire recorded time, and z represents the startingpoint of the chosen window. If we apply equation (8) foreach time sample instantaneously (e.g., the length of thetime window is one sample), then we obtain one anglefor each sample. This is referred to as a polarization log,or instantaneous polarization trace, which is discussed inmore detail later. The fast and slow waves can also beeasily determined from the transformed vectors fromequation (7):

Sl (t) -S2(t) = c(t) cos2a+q (t) sin2a,Sl (t) +S2(t) = C(t).

(9)

The simple arithmetic of equation (6) and the resultingseparation of time series from geometry factors in equa-tion (7) are the main advantages of the LTT. After sepa-ration of the split shear waves, the conventional P-wavestacking procedures can be performed separately for theprincipal wave components. Overburden correction iscarried out on poststack sections.

Overburden Correction

Overburden amplitude corrections compensate fornear-surface and overburden anisotropy effects, attenua-tion, scattering, and transmission loss. This is similar tothe surface-consistent approach for prestack amplitude

344 Li and Mueller

correction. Here a statistical approach for the overbur-den, assuming subsurface consistency, is taken to imple-ment the overburden correction for surface seismic data(Li, 1994). Note that the subsurface consistency impliesthat the reflectivity of the overburden horizon is consis-tent for adjacent common depth points (CDPs). The pro-cedure can be summarized as follows:

1. Select a quality horizon in the overburden as a ref-erence level. An average pattern of the amplitudevariation along this horizon in the overburden iscalculated for each CDP location on the S, and S2stack sections.

2. Similarly, define for the target horizon an averagepattern of the amplitude variation calculated foreach CDP.

3. Derive a common overburden scaling factor fromthese two curves, assuming the scaling factor isconsistent for adjacent CDPs.

This method is based on the model that the target issandwiched between the overburden and a half spaceand that the overburden response can be simplified witha frequency-independent scaling factor (Li, 1994).

Shear Wave Splitting in Austin ChalkThe multicomponent seismic data we selected in

this study consist of six lines of four-component surveysfrom two different fields and three VSPs from three dif-ferent sites over the Austin Chalk. Following a reviewof the geologic setting in the study area, we discuss theacquisition parameters for these data sets and theirunique features. We then review the characteristics ofshear wave splitting along the Austin Chalk and com-pare different ways to display shear wave splitting. Weintend to demonstrate that clear shear wave splitting inthe Austin Chalk can be observed both in multicompo-nent surface data and in VSPs, which makes it possibleto characterize fracture swarms in the chalk.

Geologic Setting

The study area lies in an intraplate basin, and thegeneral tectonic setting and stress regime have been de-scribed by Zoback and Zoback (199 I). The outcrop pat-tern of the Austin Chalk is shown in Figure 2 ;sou it dipsabout 2” toward the Gulf of Mexico. Extension is occur-ring toward the Gulf of Mexico, and the maximum hori-zontal stress parallels the coast. The nature of the frac-turing in the chalk has been described by Corbett et al.(1987), who analyzed data from cores and surface out-

M E X I C O

Fig. 2. Map showing the locations of the study areas. Theshaded area marks the Austin Chalk outcrop, and thearrows show the regional fracture strike obtained fromshear wave data. Adapted from Yardley and Crampin(1993).

crops. Fracture orientations trend parallel to the regionalstress directions. Studies of shear wave anisotropy in theregion find that the polarization direction of the leadingsplit shear wave parallels the stress pattern (Raikes,1991; Mueller, 1991; Li et al., 1993a).

The area covers parts of Dimmit, Zavala, Frio, andLa Salle counties in the southwest (Figure 3a) and De-vine, Giddings, and Burleson counties in the northeast(Figure 2). The subsurface structures are almost hori-zontal layers because of the small dip. Primary oil andgas plays in the area are in Cretaceous rocks, includingthe Olmos Sandstone, San Miguel Sandstone, and Aus-tin Chalk (Ames, 1990). In the southwestern part of thearea, oil and gas are produced from the Olmos Sand-stone; in the northeast, oil and gas are produced from theSan Miguel Sandstone and the Austin Chalk.

Of these major formations, the Austin Chalk has at-tracted continuous interest since the initial oil discoveryin the 1920s (Stapp, 1977). Interest in the Austin Chalkhas been renewed with the advent of horizontal drillingtechnology (Kuich, 1989; Ames, 1990). In Dimmit, Zav-ala, Frio, and La Salle counties, the chalk is at 2000 mdepth and has a thickness of -300 m (Scott, 1977; Stapp,1977). In Devine, Texas, which is closer to the outcrop,the chalk is at 1000 m depth and is 100 m thick. In Gid-dings field (in Fayette and Burleson counties), the chalkis at -2500 m depth and is thin, with a thickness of only30-100 m. The strike of the Austin Chalk is northeast-

Multicomponent Seismic Data for Fracture Characterization 345

(4

. .ZAVALA

0. .

.- .,*

*+* . ..0* ;

.

. . -. ..

** ,..

:-

- : _

(! LINE 1 I nnYl”rVP 1*,9 STRIKE

. .

FRIO

.

*... .

. .

. .* 0.

l .

.-. .*. *. ..; -- .:

-.

0-

.*. .*

-:

�Oo*

- .*.

.l .

.%oc

0

20%.*�☺o: 0 I * -

.

. .. I

I

9d 0 MILE 10

DIMMIT FEET 07I

.

. L A S A L L E

----•-- - - -

I

1 -a, . . a _--,Line 1: 0.5% Line 2: 1 .u% une 3: 1.5%

Fig. 3. (a) Locations of the threesurface lines in Pearsall field inDimmit and Frio counties. Thedots mark oil wells (courtesy ofAmoco Production Company).Note the location of VSP 1. (b)Map showing part of line 3 andthe surrounding area. The rowsof circles are surface projectionsof the horizontal wells; the largeopen circle at one end indicatesthe surface location of a well andthe small solid circle at the otherend indicates the horizontallydisplaced total depth of the well.WI, W2, and W3 are the threewells selected for study; thedashed lines project thehorizontal wells to the survey linealong fracture strike.

346 Li and Mueller

southwest, and the regional structural setting is parallelto that of the Gulf Coast Basin. Weeks (1945) andDravis (1979) have reviewed the regional geology andsedimentology of the Austin Chalk.

The Austin Chalk was formed from a fine-grainedcarbonate mud containing skeletal remains of algae(Corbett et al., 1987). The porosity of the chalk in thesubsurface is as low as 6% (Corbett et al., 1987). Withsuch low porosity, reservoir development is almostwholly dependent on accessing fractures, and successfuldrilling depends on identifying and penetrating highlyfractured chalk (Kuich, 1989; Mueller, 1991).

The two major fracture systems in the Austin Chalkare regional and fault-related. Both fracture systems arebelieved to be uniformly oriented and parallel to thestrike of the structures (Stapp, 1977). There are threemajor reasons for this. First, according to Scott (1977),the Austin Chalk was deposited on a relatively flat sur-face in deep-water environments and is expected to haveuniform facies. Second, the regional fractures were cre-ated by the downwarping of a flat depositional surfacein response to subsequent deposition of great thickness-es of Tertiary sands and shales. This downwarpingstretched the Austin Chalk uniformly along the dip di-rection of the structure. Third, all large subsurface faultsin the Cretaceous strata are parallel to the strike of struc-ture and have similar displacements. There is no evi-dence of age difference in the various fault zones (Cor-bett et al., 1987), and all movements are assumed to becontemporaneous and the regional stress field uniform.

Although the strike of fractures in the chalk is ex-pected to be uniform, the fracture intensity in dipmeterlogs and field mapping is not. This is probably due tominor differences in local stress and variations in thecontent of skeletal algae remains; thus, fracturing ismore likely to occur in some areas (Scott, 1977; Stapp,1977; Corbett et al., 1987). Thus, fractures form clustersand swarms which are the major exploration targets forhorizontal drilling (Kuich, 1989; Mueller, 199 1, 1994).The same uniform stress fields as in the Austin Chalkalso reoriented the pore space in the Tertiary overburdenalong the strike of the fractures. Thus, an overall uni-form anisotropy symmetry direction is expected in thebasin, with possible minor local variations. This situa-tion, together with the “layer-cake” structure, provides afavorable environment for studying shear wave splittingin seismic reflection surveys.

Data Volume and Field Acquisition

Figure 2 shows the Austin Chalk trend and the dis-tribution of shear wave data sets. Geographically, the

data are from three fields: three surface lines (lines l-3)and one VSP (VSP 1) are from the Pearsall field in Dim-mit and Frio counties in the southwestern part of thetrend; one VSP (VSP 2) is from a site in Devine, Texas,in the central part of the trend; and three surface lines(lines 4-6) and one VSP (VSP 3) are from the Giddingsfield in Fayette and Burleson counties in the northeast-ern part of the trend.

The six surface seismic lines were recorded in afour-component shear wave survey design with 121 in-line and 12 1 cross-line receiver channels both for in-lineand cross-line (horizontal) source orientations and witha station interval of 34 or 50 m (-110 or 160 ft). (Detailsof acquisition parameters for lines l-3 from Pearsallfield are documented by Li et al., 1993a; those of lines4-6 from Giddings field are documented by Mueller,199 1.) The three VSPs were acquired with three orthog-onal sources and three-component geophones, forming anine-component geometry. However, only the four-com-ponent horizontal subset is analyzed here. Details oftheir parameters are documented by Yardley and Cramp-in (1993).

.

The three seismic lines from Pearsall field form aclassic shear wave experiment for studying shear wavesplitting. Three lines with different azimuths to the re-gional fracture strike, and located in areas with differenthydrocarbon production rates, demonstrate different be-havior of the split shear waves. Figure 3a shows the lo-cations of the three survey lines and well distributions inthe Pearsall field. Line 1, striking about N 40” E, is par-allel to the strike of the subsurface faults and subsurfacefractures. There were few exploration wells drilled inthis area of southwestern Dimmit County. Line 2, trend-ing N 50” W, is nearly perpendicular to the fracturestrike, and line 3, trending about north-south, is at -40”to the fault and fracture strike. These two lines are locat-ed near the juncture of Dimmit, Zavala, Frio, and LaSalle counties, where the area has been heavily drilled(Figure 3a). Significant oil fields operated in this areaare the Pearsall field in Frio County (Champion, 1936)and the Big Wells field in Dimmit and Zavala counties(Layden, 1976). Recent horizontal drilling in the AustinChalk has also occurred here (Kuich, 1989; Ames, 1990;Mueller, 1991). Three horizontal wells (Wl, W2, andW3) have been drilled along line 3.

Line 4 (Figure 4a), from the southern part of Gid-dings field, had a horizontal well drilled on it with sup-porting mud logs for identifying fracture zones. Thus,any fracture zone, as identified by shear wave attributessuch as differential time delays and amplitudes betweenthe two splits shear waves, could be carefully calibratedwith the horizontal well. Figure 4a shows the local line

Multicomponent Seismic Data for Fracture Characterization 347

(a)-..b..

-. ‘. *.

,.

lfRegional

Fracture Strike.

09

E.-c

4.0

4.1

4.2

4.3

4.4

4.5

4.6

4.7

4.8

4.9

5.0

. . -. -.%

Giddings

*. *.*.

tN

*.‘.

‘.. .

‘.*.

LEASE BOUNDARY LEASE BOUNDARY

<-00

AUSTINCHALK

Fig. 4. (a) Location of line 4 and horizontal well W4 within the lease boundary (outer square). The light dashed line is line4, with the small squares representing dim spots interpreted from the S2 section (below). The solid line trending northwestrepresents the horizontal borehole, with the open boxes representing fracture swarms determined from mud logs. Thesolid dots are the surface locations of producing wells drilled to the Austin Chalk. (b) Part of the S2 seismic sectionshowing amplitude dim spots, with heavy vertical bars marking the lease boundary; the reversed L-shaped line marks theborehole. The seismic section has the same lateral scale as the map in part (a). Adapted from Mueller (1991).

Li and Mueller

layout for line 4. The regional fractures and fault systemin this area strike southwest-northeast (Figure 2), andline 4 strikes about N 40” W. which is approximatelyperpendicular to the regional fracture strike. W4 marksthe horizontal well with mud logs taken to verify thefractures, which are interpreted from mud logs by thepresence of calcite crystals that form on the intersectedfracture faces. Lines 5 and 6 are from the northernreaches of Giddings field in Burleson County (Figure 2),and VSP 3 was obtained from a well near line 5. Thesetwo lines in Burleson County strike N 40” E, and thefracture strike at this location has rotated to a east-westorientation (Figure 2).

The three VSPs also form a classic experiment. VSP1 is from a dry hole on line 1 in Dimmit County in anarea that has had little oil production (Figure 3a); VSP 2is from a water-producing well and VSP 3 is from anoil-producing well. Of these wells, only the chalk layerat VSP 3 is producing oil. This allows comparison of theanisotropy attributes for different production rates. Thethree VSPs are all near-offset nine-component VSPs,thus allowing accurate determination of the anisotropyparameters. We not only use the VSPs to demonstratedifferent features of shear wave anisotropy as the pro-duction varies but also VSP 1 and 3 to correlate hori-zons with the shear wave surface data.

The six surface lines illustrate some basic featuresof shear wave splitting, such as coupling and decouplingof wave modes as line azimuth changes, misties of fastand slow events as the two shear waves propagatethrough the subsurface with different speeds, and differ-ential amplitudes between the fast and slow shear wavesas the slow wave attenuates more than the fast wave.

Shear Wave SplittingRecording shear waves with horizontal sources and

receivers is generally less robust than recording P-waves with vertical sources and receivers. The couplingof a horizontal source with the surface may be weak,and the horizontal phones may be more open to variousnoise that often propagates horizontally, such as windand surface waves. Furthermore, because of the difficul-ties associated with the near surface, such as strong at-tenuation and scattering, low-velocity weathered layers,and irregular topographic variations, there have been se-rious concerns in the industry as to whether qualityshear wave data can be acquired and whether shearwave splitting can be observed. Here, we use held datato examine these concerns.

Pearsall Field

Figures 5a, 5b, and 5c show one shot-gather datamatrix each from lines 1, 2, and 3 in the Pearsall field.These figures show how field data change for differentline azimuths. Coherent primary reflection events can beclearly identified in all three data matrices. Comparisonof the diagonal elements (XX and yy) with the off-diago-nal elements (xy and JX) of the data matrix shows thatboth line 1 and line 2 (Figures 5a and 5b) have strong,coherent shear wave events in the diagonal elements butalmost no coherent signals in the off-diagonal elements.In contrast, line 3 (Figure 5c) has strong, coherent shearwave events with approximately equal energy in bothdiagonal and off-diagonal elements. These figures illus-trate the effects of shear wave splitting in a basin havingazimuthal anisotropy with uniform symmetry directions.Lines I and 2 are parallel to the vertical planes of sym-metry, so that the in-line source excites the in-line re-ceiver, and the cross-line source excites the cross-linereceiver, and there is little cross-coupling. Line 3, in anintermediate direction, results in the in-line and cross-line source orientations each exciting both split shearwaves, and there is strong cross-coupling. Comparisonof the in-line source components (XX and xy) with thecross-line source components (YX and yy) shows that allthree lines have larger coherent noise (the S-to-P con-verted arrivals) in the in-line source components thanthe cross-line source components above 3.0 s. This co-herent noise is mainly S-to-P converted waves becausethe cross-line source does not favor S-to-P conversions.The XX components show the largest coherent noise,while the yy components show almost no coherentnoise.

Thus, although there is noise in the recorded datamatrix, quality primary shear waves can be recorded,and shear wave splitting is observed, as shown inFigure 5c. Note that here we display data in a matrix for-mat and compare the distributions of energy among thedifferent components for interpreting shear wave split-ting. Color displays of instantaneous polarization (Liand Crampin, 1991b), which can be calculated usingequation (8), are another way of displaying thisinformation.

Figures 6a and 6b show polarization logs after LTThas been applied to the data matrices of line 1(Figure 5a) and line 3 (Figure 5c), respectively. The po-larization log of line 2 shows similar features to that ofline 1 and is not shown here. Note that polarization fil-tering flags the noise and incoherent polarizations as un-

0.0

1.0

2.0

3.0

4.0

0.0

1.0

2.0

3.0

4.0

Multicomponent Seismic Data for Fracture Characterization 349

(a) Line 1X X

1 31 61 91 121

XY1 31 61 91 121

YX1 31 61 91 121

YY1 31 61 91 121

Fig. 5. Comparison of shot data matrix selected from the three lines in Pearsall field: (a) line 1, (h) line 2, and (c) line 3.Rows are common geophone orientations, and columns are common source orientations for in-line (x) and cross-line 0)horizontal components. The station interval is 50 m (165 ft), and the far offset is 3000 m (9900 ft). Relative amplitudesamong the components are preserved by applying the same gain function to all four components. Note that the shear wavefirst breaks are weak and incoherent, likely caused by the attenuation due to source and receiver groups (Li et al., 1993a).(see parts b and c on next two pages)

3.50 Li and Mueller

(b) Line2

2ii.-l -

i!.-I-

1.0

2.0

3.0

4.0

XY1 31 61 91 121

0.0

1.0

2.0

3.0

4.0

YX1 31 61 91 121

YY1 31 61 91 121

Multicomponent Seismic Data for Fracture Characterization 351

(c) Line 3

xx1 31 61 91 121

0.0

1.0

3.0

4.0

XY YY1 31 61 91 121 1 31 61 91 121

YXYX11 3131 6161 9191 121121

0.0

1.0

Eii 2-o.-I-

3.0

4.0

352 Li and Mueller

specified polarizations represented by the white back-ground. In Figure 6a, the incoherent polarizations be-tween the in-line (XX and xy) and cross-line source com-ponents (yx and yy) are muted (Li and Crampin, 199 1 b),while in Figure 6b, polarizations of O”-15” are muted.The polarization log of line 1 (Figure 6a) shows twoevents: green (in-line polarization of 0’) and red (trans-verse horizontal cross-line polarization of 90”), indicat-ing that the in-line and cross-line polarizations from thesource are preserved. This can imply either a plane-lay-ered isotropic structure or, as here, an anisotropic struc-ture in which the reflection lines are along directions ofvertical symmetry planes. In contrast, the polarizationlog of line 3 (Figure 6b) shows blue and orange eventswith polarizations of -50” and 40” from the in-line di-rection, counterclockwise for negative angles and clock-wise for positive angles. This is a typical occurrencewhen source polarization lies between the symmetryplanes of an azimuthally anisotropic medium. Compari-son of Figure 5 with Figure 6 shows that color displaysof instantaneous polarization are highly efficient for an-alyzing shear wave polarizations.

Giddings Field

Thus far, we have discussed one feature of record-ing shear waves in anisotropic rocks using the threelines in the Pearsall field: the variation in event coher-ence on the four components as the line azimuth variesin an azimuthally anisotropic medium. Using lines 5 and6 from the Burleson County part of Giddings field, wenow show two other features of recording shear wavesin anisotropic rocks: mistie of events and differentialamplitudes between fast and slow shear waves as theytravel at different speeds.

Figure 7 shows two stack sections from line 5 repre-senting the two split shear waves. The fast shear wave ispolarized east-west, and the slow shear wave north-south. From top to bottom, there is an increasing mistie,indicating that anisotropy is present not only in the near-surface layer but also in deeper layers. At the level ofthe Austin Chalk (just below 4 s two-way traveltime) acumulative mistie of 70 ms is observed, representing2% average shear wave anisotropy from the surface tothe top of the chalk (just below 4 s). Note that there is noincreasing mistie from 3 to 4 s, which confirms that thethick Tertiary sandstone above the Austin Chalk is iso-tropic.

Figure 8 shows two stack sections from line 6 forthe two split shear waves. Similar to line 5, the fastshear wave is polarized east-west and the slow shearwave north-south. At the Austin Chalk level, a mistie of

70 ms and a differential amplitude between the fast andslow shear waves can be observed. The reflections fromthe chalk formation in the S, section are consistent instrength and continuity, while those in the S, sectionvary strongly. There is a prominent weak amplitudeanomaly in the center of the event that corresponds to alarge fracture swarm.

Data Processing

Because the data sets showed different responses toanisotropic rocks, slightly different processing sequenc-es were applied to different survey lines. We discuss theprocessing sequences separately.

Lines 1 and 2

Lines 1 and 2 are parallel and perpendicular to thepresumed crack strike, respectively. Figures 5a and 5bshow that the two split shear waves along lines 1 and 2are separated so that one-component P-wave processingtechniques can be applied to each component separately(Lynn and Thomsen, 1990; Li et al., 1993a). Because theoff-diagonal elements contain little signal energy(Figures 5a and 5b), they can be omitted in an earlystage of processing. Table 1 summarizes the processingsequence used for lines 1 and 2. The main points to noteare as follows:

1. Even when the polarizations are separated, the ve-locities of the two shear waves are different andvary with offset. Li and Crampin (1993b) suggest-ed velocity and moveout equations for CDP gath-ers in anisotropic media. For small spreads, the ef-fects of varying velocities with offset are thoughtto be negligible, and the wide offset traces must becarefully muted, as done here. However, the dif-ference between the stacking velocities of the twoshear waves is not necessarily negligible eventhough it may be small, and different stacking ve-locities may be needed to stack the two shearwaves.

2. Careful amplitude balancing is usually necessaryin processing vibrator data. Traces at small offsetstypically have large amplitudes that decreasesharply with increasing offset; sources at differentorientations and locations may vary in strengthand coupling. Thus, prestack multicomponent sur-face-consistent corrections are usually necessaryto condition the data matrix, as indicated byequation (2).

-57-63-69-75-81

-111-117-123 h-129 I

I-135

-141 E'.-+

-103-171-177

Fig. 6. Polarization logs after applying the linear transform technique (LTT) to (a) the shot data matrix in Figure 5a,line 1, and (b) the shot data matrix in Figure 5c, line 3. Positive angles in the color code indicate polarizations measuredclockwise from the in-line direction, and negative angles indicate polarizations measured counterclockwise.

354 Li and Mueller

Table 1. Conventional Processing Sequences

Procedure

Data editing

Sequence of Steps

Geometry inputTrace editingData statics using velocity 3000 ft/sMulticomponent surface consistent amplitude correction

Velocity and residual statics CDP sortVelocity analysisNM0 correctionFour-component scaling: balancing trace by off-diagonal recordsStacking: form pilot traces for statics estimationResidual statics pickingSolve for surface consistent residual staticsIterate above sequence

Final Stacking

LINE 5 HORIZON MISTIENORTHlSOUTH

CDP sortNM0 correctionResidual statics applicationStackOverburden amplitude correctionDisplay

l-

2-

E0 3-Ei=

4-AUSTINCHALK

5-

- 2

- 3

- 4AUSTINCHALK

Fig. 7. A natural polarization direction comparison forline 5 in Giddings field. On the left side of the plot, thenorth-south particle motion is displayed, and on theright side, the east-west particle motion. The sections arejoined at the same surface point. A dynamic mistie ofevents can be seen where the sections join. After Mueller(1991).

(a) Sl 1000 ft-3.5

3al AUSTINE

4.5CHALK.-

F

W3.5

2g 4 .5 AUSTIN.- CHALK+

5.5

Fig. 8. Final stacked (a) S, and (b) S, sections of line 6 inGiddings field using the processing sequence in Tables 1and 2. The S, section shows an amplitude dim spot alongthe Austin Chalk.

Multicomponent Seismic Data.for Fracture Characterization 355

3. In practice, it is often assumed that the fast andslow split shear waves have the same field staticsand the same residual statics. However, becauseshear wave splitting is due to the polarizations re-acting differently to the anisotropic structure, thisassumption can only be a first-order approxima-tion. Assuming equal statics, the problem is to en-sure that the same statics are applied to all compo-nents. The fast and slow split shear waves mayhave different stacking velocities (Thomsen,1988; Li, 1992) and it may be necessary to carryout two separate passes of velocity analysis for S,and &.

Line 3

Line 3 shows strong energy on cross components.Separating split shear waves provides an extra challengein addition to the problems encountered in processinglines 1 and 2. Before one can proceed to separate thesplit shear waves, one has to compensate for source, re-ceiver, offset, and other near-surface factors. The datamatrix after compensation may then be considered as aproper representation of the medium response for shearwave splitting. However, amplitude corrections are sen-sitive and can be tricky; thus, careful procedures have tobe taken to ensure minimum distortion and artifacts tothe data. Always check the quality control (QC) plots ofthe scaling factors before applying the corrections. Thesource and receiver factors should only show short-wavelength and minimum corrections as implied by thesurface consistent criteria; the offset factor should showa smooth and almost exponential variation with offset.One can also check the observation logs to confirm anysignificant abnormality in the source and receiver scal-ing factors. In practice, we find that these scalingfactors, particularly the offset scaling factor, provide anoptimum amplitude balance between the near- and far-offset traces but preserve the relative amplitude betweenthe components and CDP locations. Without such cor-rections, the trace amplitudes are not balanced, and theS/N ratio of the stacking section may be very low (seeLi, 1994). A conventional amplitude scaling may bal-ance the trace amplitude, but it may also distort the rela-tive amplitudes between the components.

Here the multicomponent surface consistent correc-tion, as shown in equation (2) and the LTT, as describedin equations (6)-(9), are used for processing line 3. Themulticomponent surface-consistent correction involvesGause-Seidal iteration to decompose the logarithm am-plitude spectrum based on equation (2) (Li, 1994). TheLTT processing is also straightforward. First, LTT is ap-

plied to the shot data matrix to transform the four com-ponents (XX, xy, yx, and yy) into the three LTT compo-nents: S,, S,, and the polarization log. Second, the con-ventional processing sequence in Table 1 is applied sep-arately to the S,, &, and polarization components to givethe final stacked S,, S2, and polarization sections. Notethat the processing of polarization logs is slightly differ-ent and usually involves a polarization filter to removeunwanted polarizations (Li and Crampin, 199 1 b).

To show the effect of LTT, we have includedFigure 9a which is the output after applying LTT to theshot data matrix in Figure 5c. Comparing Figure 9a withFigure 5c shows that the energy in the off-diagonal pan-els is significantly reduced. The polarization log afterapplying LTT to Figure 5c was shown and discussedpreviously in Figure 6b. Note that Figure 9a is not asgood as Figures 5a and 5b in terms of residual energy inthe cross components, which highlights the limits of thedata processing technique for separating split shearwaves, as compared with physical separation of the twoshear waves.

To compare the LTT results with the rotation tech-nique, Figure 9b shows the output after rotating the shotdata matrix in Figure 5c by 40” (angle derived fromFigure 6b; see earlier discussion). Comparing Figure 9awith 9b shows that the energy in the cross componentsobtained by LTT is less than that obtained by conven-tional rotation. This indicates that the split shear wavesare optimized better by LTT than by rotation. This is be-cause in the implementation of the rotation used here,the rotation angle is constant, whereas LTT allows asample-by-sample and trace-by-trace variation in thepolarization angle (Li and Crampin, 1993a). Li et al.(1993a) explain in more depth why LTT performs betterthan simple rotation.

Lines 4,5, and 6

Line 4 is perpendicular to the fracture strike and isprocessed as outlined for line 2.

Lines 5 and 6 also show significant shear wave split-ting, as demonstrated in Figure 7 (for line 5) andFigure 8 (for line 6). The separation of shear waves inlines 5 and 6 is performed using the rotation method, asa comparison with LTT in line 3. The process of rotationanalysis can be summarized as follows. First, the dataediting sequence in Table 1 is applied to all four compo-nents. Again note that the editing sequence in Table 1contains the procedure of amplitude corrections whichare essential for balancing the trace amplitudes whilepreserving the characteristics of shear wave splitting,such as the differential amplitudes of the two shear

356 Li and Mueller

waves. This requires careful quality control to ensurethat only the short-wavelength and minimum source andreceiver corrections are applied (see section on line 3 formore discussion). Second, the velocity and statics se-quence in Table 1 is applied to the yx or xy component todetermine the stacking velocities and residual statics.Note that the yx and xy components are expected to beabout equal in noise-free azimuthal anisotropic mediacontaining uniform symmetry (same crack orientation)throughout the depth range. In such cases, a single passof the velocity and statics sequence is sufficient tochoose off-diagonal records for preparing stacked datamatrices for rotation. Third, the preprocessing sequencefor rotation in Table 2 is applied to all four componentsseparately and a stacked data matrix is obtained. This isimmediately followed by the rotation sequence inTable 2. At this stage, the stacked data matrix can beused to determine the rotation angle, and this angle usedto rotate the shot data matrix to separate the four-com-ponent data into two-component (S, and S,) data sets.The last stage is to apply the velocity and statics se-quence and the final stacking sequence in Table 1 to theS, and S> components separately to obtain the finalstacked S, and Sz sections.

Vertical Seismic Profiles

The LTT (from equations 6 to 9) is used to deter-mine the fast direction in the VSPs at the three sites, andthe data are then rotated to give the fast and slow sec-tions. The time delays are calculated using cross-correlation of the fast and slow section. To compare thedifferential amplitude of reflected fast and slow shear

waves, quadrilateral f-k filters are used to separate theupgoing and downgoing wavefields. Before the f-k lil-ters, a depth-dependent amplitude scaling factor is ap-plied to the data to balance the amplitudes betweendifferent depth levels; after the f-k filters, this scalingfactor is removed to recover the amplitude variations.Before the amplitude ratio of the upgoing arrivals on theslow and fast sections can be calculated, the amplitudeof the downgoing arrival for the fast and slow waves ateach geophone level is used as a scaling factor to nor-malize the amplitude of the upgoing arrival for the cor-responding fast and slow waves at the same geophonelevel. This means that changes in source strength be-tween the in-line and cross-line sources and sphericaldivergence effects can be neglected. The amplitude ratiois then calculated at each level for the reflections fromthe Austin Chalk for all three VSPs.

InterpretationThe interpretation of fracture zones is based on the

analysis of shear wave attributes, including polariza-tions, time delays, and amplitudes (dim spots) at normalincidence. Wherever possible, other well informationsuch as production figures and mud logs are used to cal-ibrate the results. We first discuss the results of the threelines in Pearsall field, then the results of line 4 in Gid-dings field, and compare the three VSPs from the threesites with different production rates. Finally, we give aregional overview of all the results. Note that we em-phasize the use of shear wave attributes at normal inci-dence as interpreted from stacked sections. This is be-cause the variations in shear wave attributes at normal

Table 3. Processing Sequences for Separating Split Shear Waves: Comparison of RotationTechnique with the Linear Transform Technique (LTT)

Procedure

Preprocessing for rotation

Sequence of Steps

CDP sort;NM0 correctionRobust average scalingResidual statics applicationStack

Rotation sequence Rotation angle estimation: rotation analysis, or other techniquesApply rotation angle to shot dataSeparate data into S, and S2 data sets

LTT Apply LTT to shot dataSeparate data into S,, SZ, and polarization components

Multicomponent Seismic Data,for Fracture Characterization 357

sii.-l-

(a) Line 3: LTT

1310.0

S61 91 121

6,s: ,)')jl" , : ,':;j'/ :;, ,!I

1.0

2.0

3.0

4.0

1.0

1 31 61 91 121/ 11 I' 8a i,!;lI/ ,,t,',' '1 ,

11 3131 6161 9191 121121

Fig. 9. (a) The shot data matrix of Figure SC after applying LTT and (b) after applying a 40” rotation. The coherent off-/diagonal energy in both is less than in Figure SC, but the off-diagonal energy in (a) is substantially less than in (b). (seepart b on next page)

358 Li and Mueller

(b) Line3: RotatedS

0.0 '31 61 91 121

, ,' 11, "'/' ,, i

0.0

1.0

2.0

3.0

4.0

I ,__,, 11,,1: :/,,,,,‘, ,, /I, :1,, /I, :1

,,I’,. ,‘I1 :‘, I I.’

3.0

4.0

1 31 61 91 121

I

1 31 61 91 121

Multicomponent Seismic Data for Fracture Characterization

incidence are first-order effects, as compared to theiroffset-dependent variations which are more subtle anddifficult to interpret for fracture characterization.

Pearsall Field: Lines 1,2, and 3

This section presents the final stacks of the threelines for S, and S2 waves and the stack polarization logof line 3. We demonstrate a positive correlation of theoverall anisotropy with production and a correlation ofamplitude dim spots in the stacked section with fractureswarms encountered by horizontal drilling as inferredfrom production records.

Variations in Anisotropy

Figures 10, 11, and 12 show parts of the final stacksections of lines 1, 2, and 3 for the fast and slow waves.Figure 13 shows the stack polarizations of line 3. Onecan see that the stacks are “ringy” (particularly line 3 ofFigures 12a, 12b), which is caused by the narrow band-widths of the field data. As shown in Figure 5, the domi-nant frequency is 10 Hz with a bandwidth of less than15 Hz. For such narrow band data, predictive deconvo-lution does not function properly for reducing reverbera-tions. Note that to improve the reliability of event corre-lation in ringy (narrow bandwidth) data, such as in line3, it is essential to integrate all available information in-cluding local VSP corridor stack and sonic synthetics,local P-wave survey lines, and other regional surveylines.

Figure 14a shows the cumulative time delays of theGeorgetown Formation below the Austin Chalk, whichis calculated from the cross-correlation of S, and S, sec-tions for a 400 ms window from one cycle above the topof the chalk. Figure 14b shows the interval time delaysbetween the Olmos Sandstone and the Austin Chalk forall three lines. These results show clear variations inanisotropy among the three lines. The average anisotro-py in line 1 is small, less than 0.5%. This is calculatedfrom the interval time delay in Figure 14b and normal-ized by the traveltime. Note that the interval time delaysfor line 1 are O-l 0 ms (Figure 14b), and the GeorgetownFormation appears at 2.5 s (Figure 10). This is consis-tent with variations in other attributes such as small dif-ferential amplitude variations and the small mistie ofevents in Figure 10. Similarly, the average anisotropyfor lines 2 and 3 can be calculated from Figure 14. Line2 possesses about 1% interval anisotropy and line 3about 1.5%. This is also consistent with the greater vari-ations in amplitude and mistie in Figures 11 and 12.

Also, anisotropy increases from south to north on lines 2and 3.

Correlation of Anisotropy with Production

Variation in anisotropy on the three lines can beroughly correlated with oil production. Figure 3a showswell locations in the study area. Few wells have beendrilled near line 1, and it is more than 10 mi (16 km)from neighboring production. Line 2 is at the edge of thePearsall and Big Wells fields in Frio, northern Dimmit,and Zavala counties, and there are about 10 wells within1 mi (0.6 km) of the line. Line 3 is through the center ofthe Pearsall field, and many producing wells have beendrilled along or near it. The percentages at the bottom ofFigure 3a summarize the overall anisotropy in all threelines, as calculated from Figure 14: line 1, 0.5%; line 2,1%; and line 3, 1.5%. This shows that the overall oilproduction along or near the three reflection lines ap-proximately correlates with the anisotropy (time delays)observed along the three lines, where line 1 with thesmallest time delays has little production nearby, line 2with intermediate time delays shows intermediate pro-duction, and line 3 with the largest time delays has thegreatest production nearby.

Variation of anisotropy along line 3 can also be cor-related to oil production along or near the line.Figure 14c shows the distribution of horizontal wellswithin 1 mi (0.6 km) on either side of the line. Mostwells are distributed in the northern part of the line (sta-tion numbers ~700). There are only two wells at thenorthern end (station numbers <loo), and only one wellat the southern end of the line (station numbers ~800).The time delays along line 3 in Figure 14a and 14b showwide variation. In Figure 14a at the northern end, the de-lay rapidly builds up to 55 ms, remains at about thesame level until station 400, and then uniformly de-creases to about 35 ms at station 1000 toward the south.Thus, the trend of variation of time delays broadly cor-relates with the overall trend of well distribution alongthe northern part of the line. The interval time delay inFigure 14b shows a similar trend.

The overall trend of the variation in the stacked po-larization logs of line 3 in Figure 13 can also be correlat-ed with the oil production along the line. Note that wedisplayed 400 traces here to show the coherency of thepolarizations, while in Figure 12, we displayed only 200traces to highlight the variations in amplitudes. Corre-sponding to a high production area from stations 399 to449, the polarizations are more scattered (more whitebackground in the lateral variations), and the polariza-

(4 Sl702 652 602 552 502 452 402 352 302

W S2702 652 602 552 502 452 402 352 302

Fig. 10. Final stacked S, and S2 sections of line 1 using the processing Fig. 11. Final stacked S, and Sz sections of line 2 using the processing

sequence in Table 1: (a) S, section (xx component) and (b) S2 section (vy sequence in Table 1: (a) S, section (yy component) and (b) S2 section (xx

component). component).

(40.0

2.0

c

g 3 .0‘F

4.0

6.0

Sl376 326 276 226 176 126 76 26

:r: OLMOS:5gfdJSTIN CHALK

E;

FHOSSTON

W S*376 326 276 226 176 126 76 26

Multicomponent Seismic Data for Fracture Characterization 361

64

(b)

q3 L52Sl

WI CRACK,I /STRIKE

AUSTIN CHALKGEORGETOWN

iALK‘OWN

tc) 0.6 ,I

z‘CI= 0.4.ZE

fv) 0.2

E

-*- Overburden........ Austin Chalk

0.0 A A A401 3&l 301

CDP #

Fig. 12. Stacked (a) S, and (b) S, sections of line 3 using the processing sequence in Tables 1 and 2. Wl, W2, and W3 arehorizontal wells; the bars at the top represent the well azimuths, and the bar at the far right shows the crack (fracture)strike. The white boxes mark the amplitude dim spots, and the two pairs of arrows on the left mark two time windows foramplitude analysis (top = overburden, bottom = Austin Chalk). (c) Comparison of windowed rms amplitudes in the over-burden and target; windows are shown by two pairs of arrows in parts (a) and (b). The dot-dashed line is the rms ampli-tude in the overburden window, and the dotted line is the amplitude in the Austin Chalk window. The dashed straight linerepresents mean amplitude; the arrows mark significant dim spots.

362 Li and Mueller

0ui

Multicomponent Seismic Data for Fracture Characterization 363

(a)100

z 60- cl -- - Line Line Line 2 3 1

., I1 I 1 I I I I I I I

1035 935 835 735 635 535 435 335 235 135 35

STATION POINTS@I-”

‘;iE -‘-’ Line 1

- 30cl._..... Lhe 2

cn- Line 3

1 0 3 5 9 3 5 835 735 635 535 435 335 235 135 35

STATION POINTS

l o 3 5 9 3 5;” , 9,5 7f5 6:s 5?5 475 3:5 2?5 la5 ‘I

9iii32.

!r

1029 929 829 729 629 529 429 329 229 129 29

STATION POINTS

Fig. 14. (a) Comparison of cumulative time delays abovethe Georgetown strata between the fast and slow shearwaves measured from the stacked sections of lines 1,2,and 3 (Figures 10, 11, and 12) from the Pearsall field. (b)Comparison of interval delays between the OlmosSandstone and the Austin Chalk. (c) Distribution ofhorizontal producing wells within 1.6 km (1 mi) of line 3.

tion events from the Austin Chalk are less continuousand often broken into vertical variations. Note that thewhite background is associated with noise and nonco-herent polarization events of 0” to 15”. Dim and brokenS, events can cause broken and noncoherent polarizationevents which, after stacking, are attenuated toward 0”.Clearly, identification of broken and scattered polariza-tion events in color-coded sections is easier than identi-fication of dim and broken S, events in the amplitudesections. Note that such variations are subtle and com-plicated, and only when the nature of such variations ofpolarization are better understood, from more case stud-ies, can such interpretation be confidently made.

In summary, shear waves in the three reflection lines

show typical features of shear wave splitting, althoughthe anisotropy along line 1 is weak and marginal. Thevariation in overall anisotropy between the three linescan be correlated roughly with commercial oil produc-tion in the study area. Variations in anisotropy along line3 can also be correlated with oil production along thatline. High production areas are found to be associatedwith scattered polarizations and broken polarizationevents in the stacked polarization logs.

Correlation of Amplitude Dim Spots withFracture Swarms

We selected line 3 and the three horizontal wellsdrilled nearby for a detailed analysis of amplitude varia-tions and their correlation with fracture swarms, as en-countered by horizontal drilling near line 3. Figure 3b isan enlarged map from Figure 3a showing the location ofthe survey line and the distribution of the horizontalwells. Figure 3b shows a number of horizontal wells inthe area (indicated by rows of circles), all drilled in theAustin Chalk. The three wells close to the line and withsimilar horizontal distances-W 1, W2, and W3-wereselected for study. Well W 1 trends northwest-southeast,at about 60” from the regional fracture strike; W2 trendssouthwest-northeast, parallel to the fracture strike; andW3 is southeast-northwest and perpendicular to strike.The production rates of the three wells are shown inTable 3, with Wl the most productive, W2 the least pro-ductive, and W3 moderately productive.

For similar completion programs in similarly drilledhorizontal wells, the production rate (flow rate) of a hor-izontal well is mainly determined by the amount of frac-tures encountered. This is in turn determined by thelength and azimuth of the horizontal well and the frac-ture intensity of the zone penetrated by the well. Withthis in mind, and by comparing Figure 3b and Table 3,one can infer that ( 1) local fracture swarms are presentat all three sites, (2) wells parallel to fracture strike inter-cept fewer fractures than wells 60” or more from thefracture strike, and (3) the zone penetrated by Wl maybe more intensively fractured than the zone penetratedby W3, or W 1 penetrates more fracture swarms than W3(Li, 1996). Now we examine how the variation in ampli-tudes can be correlated with these fracture swarms.

Figure 12 shows two groups of dim spots (whiteboxes) that are caused by shear wave differential reflec-tivity at normal incidence (and not by AVO effects).Figure 12c compares the windowed amplitude varia-tions in the Austin Chalk with those in the overburden.(Windows are marked by two pairs of arrows on the leftside of Figures 12a and 12b.) The right-hand boxes con-

364 Li and Mueller

Table 3. Production Records of Three Horizontal Wells WI, W2, and W3l

Well Azimuth Horizontal Maximum CumulativeName Distance Production Production Period

(ft> (bbl/day) (bbl) (months)

Wl N 25” W 3000 600 108,000 14

w 2 N 40” E 2000 300 37,000 19

w 3 N50”W 3000 370 74,600 21

‘Regional fracture strike is N 40” E. Data supplied by Amoco Production Company

tain two dim spots extending from CDP 235 to 245 andfrom 260 to 275, and those on the left contain one dimspot from 345 to 365. To see how the horizontal wellsare intercepting these dim spots, we projected the hori-zontal wells to the survey line along the fracture strike(Figure 3b); the projected wells were then superimposedon the seismic sections (Figures 12a and 12b). Compari-son of Figures 3b and 12 shows that Wl extends from225 to 265 and intercepts two dim spots, W2 is at theedge of the third dim spot (CDP 370), and W3 extendsfrom CDP 355 to 395 and intercepts part of the thirddim spot (355-365). Thus, WI is likely to interceptmany more fractures per unit of drilling distance thanthe other two wells since it intercepts more dim ampli-tudes than the others. This may be one of the reasonsthat Wl yields more than W2 and W3, demonstratingthe correlation between the dim spots in the seismic sec-tions and the local fracture swarms encountered by thehorizontal wells. Dim spots appear in both the S, and S,sections here, implying the presence of macrofracturesthat attenuate both the fast and slow split shear waves,as demonstrated by Liu et al. (1993) at the ConocoBorehole Test Facility.

We believe that the dim spots in the S, section inFigure 12a are more likely to be genuine rather than arti-facts of data processing, such as the overburden correc-tion. There are two reasons for this. First, after examin-ing the original data, we find that dim spots are alsopresent in the stacked S, section before the overburdencorrection, but less obvious when compared to the sec-tion after the correction. Thus, the correction appliedhas preserved and enhanced the original features in thedata. Second, both micro- and macrofractures are likelypresent in the Austin Chalk, and in certain areas, one setof fractures may be more developed than the others(Scott, 1977; Stapp, 1977; Corbett et al., 1987). Notethat a fracture is classified as a microfracture if its

length is less than 0.1 of the shear wavelength; other-wise it is considered to be a macrofracture. A mediumcontaining aligned microfractures is equivalent to a ho-mogeneous anisotropic medium, according to the equiv-alent medium theory (e.g., Hudson, 198 1). However, formacrofractures, the equivalent medium theory breakdowns (Ebrom et al., 1990). For the seismic data studiedhere, the shear wavelength is 200-250 m.

Giddings Field: Lines 4,5, and 6

In the previous example, amplitude dim spots werecorrelated with fracture swarms inferred from produc-tion records in horizontal wells. In the present example,a more precise correlation of amplitude dim spots withfracture swarms is made by examining the mud logsfrom a test horizontal well drilled on line 4.

The structure in the Giddings field can be roughlyseparated into four anisotropic layers. As shown inline 5 of Figure 7, the first layer (from the surface to 2.4s two-way traveltime) has a mistie of 50 ms representing2% shear wave anisotropy. The second layer (2.4-3.0 s)has an increasing mistie of 20 ms (interval delay) repre-senting 3% anisotropy. The third layer (3.0-3.8 s) has noincreasing mistie and is isotropic, while the fourth layer(Austin Chalk interval, 3.8-4.3 s) has an increasingmistie of 20 ms (interval delay) representing 4% anisot-ropy. A strong anisotropic target with an isotropic caplayer creates an ideal situation for studying the fracturesin the Austin Chalk.

From line 6 of Figure 8, a cumulative mistie of 70ms at the Austin Chalk level can also be observed. Fur-thermore, there is a prominent dim spot in the S2 section.The S, section at the chalk level shows strong and con-sistent amplitudes, while the S, section shows a weakamplitude anomaly in the center of the event.

Similar dim spots can be seen in the S, section of

Multicomponent Seismic Data for Fracture Characterization 365

2500 25&

2520 2520 g

2540 2540 8

2560 2560 jj

2580 2580 ‘g

2500 2600 z

2620 2620 $c

600 550 500 450 400 350 300 250 200 150 100 50 0

Horizontal Dispacement (m)Fig. 15. A display adapted from the horizontal mud log for the well drilled to test the S, amplitude anomalies inFigure 4b. The short vertical lines intercepting the borehole are fractures partially filled with calcite. After Mueller(1991).

line 4 (Figure 4b) which correspond to fracture swarmsconfirmed by horizontal drilling. A test horizontal wellwas drilled toward the anomaly, and mud logs were tak-en to verify the fractures by the presence of calcite crys-tals in the mud system which often grow on the fracturefaces of the Austin Chalk trend. The trajectory of thehorizontal well is shown in Figure 15; the vertical linesintersecting the borehole represent fractures identifiedfrom the mud log. Figure 4b shows a part of the S, sec-tion, which highlights the amplitude anomaly and its in-terception with the horizontal well. The fractures andthe amplitude anomalies are all projected onto the sur-face and shown in Figure 4a, where the solid line is thesurface projection of the horizontal well and the parallellight dashed line is line 4. The black boxes along thesurvey line are amplitude dim spots interpreted from theS, section in Figure 4b. The boxes along the well projec-tion are fractures identified from the mud logs(Figure 15). A good correlation clearly exists betweenthe dim spots in the S, section and the fracture swarmsidentified by the mud logs. Note that dim spots appearonly in the S, section here, which implies the dominanceof microfractures (with length less than 20 m) in thisarea.

Comparison of the Three Vertical SeismicProfiles

We now examine the shear wave response in thethree multicomponent VSPs, which are from three dif-ferent sites: Dimmit (VSP l), Devine (VSP 2) and Bur-leson (VSP 3) (see Figure 2). The three wells are dry,water producing, and oil producing, respectively.

Figures 16a, 16b, and 16c show the four-componentdata matrices of the three VSPs. Note that they are allnear-offset VSPs and that the orientation of the down-

hole geophone can change direction during acquisition.Reorientation is often needed in subsequent processing,although in Figure 16, the tools have not been reorient-ed. VSP 1 was acquired without supporting offset VSPsfor tool orientation purposes. Because of tool rotation,the energy of the direct arrival switches from one com-ponent to another as depth varies. Some bad traces werecaused by loose coupling of the geophone in VSP 3(Figure 16~).

For the three VSPs, anisotropy parameters were cal-culated using the linear transform technique (LTT),which allows for unknown geophone orientation.Figure 17 shows polarization and time delay estimatesfrom these sites. The top row shows the polarizationmeasurements, and the bottom row shows the corre-sponding time delay estimates. Polarization angle esti-mates of the leading split shear wave were about N 30”E for the Dimmit VSP (which is close to the surfacemeasurement of N 40” E), N 55” E for the Devine VSP,and N 100” E for the Burleson VSP (which is also closeto the surface measurement of an east-west orientationfor line 5). Note that there is a 10” difference in mea-surements of the polarization between the surface andVSP results and that the measurements were constantwith depth. However, no observable increases in timedelay were associated with the Austin Chalk layer, andany increase in time delay must be less than one sampleinterval (2 ms) (bottom row of Figure 17). The chalk isabout 300 m thick at the Dimmit VSP and 100 m at De-vine. No buildup of time delay occurs in the chalk, sug-gesting that there is little birefringence in the AustinChalk in the Dimmit and Devine VSPs and that theanisotropy observed must be attributed to the overbur-den. This is in agreement with the fact that the Dimmitwell is a dry hole and that the Devine well has only pro-duced minor rates of water.

(a) DIMMIT: VSP 1

1.0 1.5One-wa$nme (ij5

1.0 1.5 2.0

BURLESON: VSP 3

04 DEVINE: VSP 2 s

0.3 0

E-Y 0.45f

; 0.6 0

0.75

0.3 0

E-Y 0.45E

; 0.60

0.75

1.5 0.0 0.5One-way time (s)

Fig. 16. Data matrices as recorded from the three zero-offsetVSPs from (a) Dimmit, (b) Devine, and (c) Burleson counties.Relative amplitude is maintained for different components ofthe data matrix, but not for different depth levels. Only thewell at Burleson is producing oil from the Austin Chalk. Notethe unequal amplitude between theyx and xy components inpart (a) caused by the tool rotation.

One-way time (s)

Multicomponent Seismic Data,for Fracture Characterization 367

(a) DIMMIT: VSP 1o.o- .

..os- ..

.LO- .

P ;dfi- t’jj .‘t:

f20- z AmllN CNALK

I .f

-*.2% >

.

I..*

3.0 I ,* I0 50 lm 150

Polarlzatbn An* (E of N)

0.0.

.

0.6 -.

i.

10 .

g :fi’6 .

8 ii

20 I AwlN CHALK

‘*:*

26 .!

1

i'.

-1 I0 lb ;o 3il ;o

The Delay Ons)

(b) DEVINE: VSP 2o.o- ..O.l- .

..

02-.

0.3-

0.4-

1

.!

O.I- z

B0.e * z

:>\

0.7-* * .

- .AUSTIN CHALK

c .0.8-

;>(?

as-

L O I I *I0 50 loo la0

Polarkatlon Anglo (E of N)

(c) BURLESON: VSP 3

p’2.0- )

r, WSnNCwl

2530

, 10 loo lso

Polarlzatlon Angle (E of N)

0.0 .. . I ,.,I0.1 :1 .

0.2 :1.

0s I.II

0.4- :i’

0.S 3’I**

O.S- ;a.

;:0.7- i ’ AUSTIN Cwu

. .!o.s-

‘I’a.ao-

L”i-7-mrdTInlo MY (m)

..2.0 I

.!I AmllNCHALK

2swTime Delay (ms)

*ICl

3

Fig. 17. Comparison of polarizations (top row) and time delays (bottom row) estimated from the three VSPs: (a) VSP 1from Dimmit, (b) VSP 2 from Devine, and (c) VSP 3 from Burleson. The spurious results around 115” in thepolarizations in the Dimmit VSP (top left) correspond to a 90” shift due to multiple solutions of the data matrix. Thehigh time delay values for the shallow geophone levels in the Dimmit VSP are due to a 13 m difference in the in-line andcross-line source positions leading to an apparent time delay between arrivals from each source. Compared to the two-way-traveltime delay in line 1 (where the Dimmit VSP is located; Figure 14a), the time delay in the Dimmit VSP issmaller and a picking error of one or two samples is likely in the surface seismic data. Adapted from Yardley (1992).

It is not possible, however, to determine if the chalkat Burleson is anisotropic using transmitted shear wavesbecause the chalk here is thin (only about 30 m), eventhough VSP 3 was acquired in a producing well. Fieldstudies show that thin layers are more likely to be heavi-ly fractured (Barthelemy et al., 1992). The study of re-fleeted arrivals may provide the best way of studying

anisotropy in thin layers (Thomsen, 1988; Yardley et al.,1991). Reflected amplitudes in the zero-offset VSP datawere analyzed to see if they could give a more accuratedetermination of the anisotropy in the Austin Chalk. Fig-ure 18 shows a comparison of the reflected amplitude ra-tios from the three VSPs, calculated using the process-ing sequence previously described.

368 Li and Mueller

to-

$

5 ts- x

%xx

x xx

P x xx

xx20- wslw CWLY

2.5-

3 .0 I I0 .0 0 .5 10 1.5 2s

ta) DIMMIT: VSP 1 (W DEVINE: VSP 2

-

E& .,5 a5%0

0.4 L

0.0 .*:

0.7

0.0

10.84

dl--) 0.0 0 .5+i0

~D=0.091D=O.ZO

I: *

.*

. * ..

.

Wsmi CM

I15

Amplitude ratio Amplitude ratlo Amplltude ratlo

os-

- to-Es

Fi,,_

(‘) BURLESON: VSP 30.0 Ix x: AR=0..34sD=0.18. Y: AR=O.84 sD=o.20- IsolRoPlc

D

Fig. 18. Amplitude ratios (ARs) calculated for the Austin Chalk reflector at each VSP site: (a) VSP 1 from Dimmit,(b) VSP 2 from Devine, and (c) VSP 3 from Burleson. The ARs were calculated at each geophone level, and all values wereaveraged to give a single number for each reflector (given as AR values in the figure). Although the results are scatteredbecause of weak reflectors, they cluster around the 1.0 line, as expected for isotropic reflectors. The exception is atBurleson, where both the in-line (x) and cross-line (y) data give amplitude ratios below unity, implying an anisotropicreflector. After Yardley and Crampin (1993).

Figure 18 shows that the average amplitude ratiosfor the chalk reflectors for the Dimmit and Devine VSPsare close to unity, indicating that the chalk layer is iso-tropic as expected from the analysis of transmittedwaves. The amplitude ratios at Burleson are less thanunity, implying an anisotropic chalk layer. The BurlesonVSP is the only one of the three data sets to be collectedin a producing well. Because the hydrocarbon reservoirsin the Austin Chalk are fractured, they are expected tobe anisotropic to shear wave propagation, which is con-firmed by Figure 1%. Analysis of the reflected ampli-tudes at the three VSP sites indicates that only the Bur-leson well penetrates an anisotropic chalk layer, whichagain correlates with the fact that the Burleson well isoil producing. Without detailed knowledge of the veloc-ity structure and pulse shape it is not possible to quanti-fy the amount of anisotropy in the reflecting layer. How-ever, for these sites, a positive correlation is seen be-tween anisotropy and productivity in the chalk layer.

DiscussionThrough shear wave studies along the Austin Chalk

trend, we have demonstrated that the overall degree ofanisotropy, as quantified by differential shear wave trav-eltime reflection surveys, can be broadly correlated withregional oil production. Dim spots identified from varia-

tions in differential shear wave amplitude can be exactlycorrelated with the fracture swarms encountered by hor-izontal drilling, as inferred from production records andmud logs.

The thickness of Austin Chalk varies along thetrend. If the interval of a target is too small, it is impossi-ble to measure interval polarization and time delays reli-ably from surface seismic data. Thus, polarization andtime delay measurements often give only an estimate ofthe anisotropy with low spatial resolution. In contrast,variations in amplitude can give a better resolution, asshown for lines 3 and 4.

Both the surface data and VSPs show that anisotro-py occurs both above and within the Austin Chalk(Figures 7 and 17). We have assumed here that thesetwo kinds of anisotropy have the same symmetry axisand the same strike, as suggested by the geologic evi-dence and as indicated by the polarization measure-ments from the three VSPs (Figure 17). With the uni-form, nearly layer-cake structure seen here, we can reli-ably identify the orientations of the fractures in thechalk by the polarizations of split shear waves observedat the surface.

In the presence of near-surface complications, shearwave splitting in reflection surveys may be subjected todistortions where raypaths pass twice through the sur-face layers (Yardley and Crampin, 199 1). These distor-

Multicomponent Seismic Data for Fracture Characterization

tions are much less in VSPs, where the raypaths passonly once through surface layers, and the recorded po-larizations of split shear waves are in the undisturbedrock mass within a few wavelengths of the geophone.Despite this, our examples have shown that surface re-cordings of split shear waves possess sufficient subsur-face information for the purpose of delineating fracturereservoirs. However, as the use of shear wave technolo-gy expands, it is becoming important to extend thismethodology to cases where fracture orientation maychange with depth. For VSP surveys, Winterstein andMeadows ( 1991) demonstrated vertical variations infracture azimuth. Meadows et al. (1994), Thomsen et al.(1995a), and Chaimov et al. (1995) demonstrated amethod and case history of vertical fracture azimuthvariation for the surface seismic reflection case. Studiesof surface geometry sensitivity and processing methodsinclude MacBeth et al. (1992) and Li (1994, 1995); webelieve that more studies should be done in this direc-tion. Another approach may be based on data matrixasymmetry studies (Li and MacBeth, 1995).

Line 3 shows variations in both the fast and slowsections, while the data of line 6 (Figure 8) shows varia-tions only in the slow section. It is possible that this dif-ference is caused by the length of the fractures involved.For microfractures, the resulting anisotropy modifiesonly the reflectivity of the slow wave, and only the slowwave can then show anomalies in amplitude variations.In the case of large (formation scale) fracturing and suf-ficiently developed orthogonal fracture sets, the result-ing anisotropy may affect both the fast and slow waves.In both cases, the ability to observe the differential am-plitude contrasts or the amplitude dim spots requires anappropriate impedance contrast at the interface (Yardleyet al., 1991; Li and Crampin, 1993~). With large imped-ance contrasts, the difference in the amplitudes of shearwave reflections from fractured and unfractured rockcould be much less, and other methods may prove morediagnostic of anisotropy (e.g., Thomsen et al., 1995b).Processing procedures can also introduce side effectsthat degrade the phenomenon. By the use of multicom-ponent amplitude correction and the LTT, the informa-tion contained in stacked amplitude sections is pre-served and enhanced. In some cases, broken and scat-tered polarization events on the stacked polarizationsection are diagnostic (line 3, Figure 13).

ConclusionsThe shear waves in the three reflection lines from

Pearsall field show typical characteristics of shear wavesplitting in both shot data matrix and color displays of

polarization logs. The average fracture orientation of N40” E in this area agrees with other geologic and geo-physical results. In areas where commercial oil produc-tion is absent, the average cumulative time delay of thesplit shear waves is less than 10 ms, corresponding toabout 0.5% differential shear wave anisotropy or less. Inareas close to major production fields, the average delayis about 30-40 ms, corresponding to about 1.5% shearwave anisotropy. In major production areas, the delay is50-60 ms, corresponding to about 2.5% anisotropy.Three horizontal wells drilled near line 3 were selectedfor a more detailed correlation of production withanisotropy. Well W 1, drilled about 60” to the fracturestrike in an area with relatively high anisotropy, wasmost productive. Well W2, parallel to the fracture strikeand drilled in an area with less anisotropy, was least pro-ductive. Well W3, drilled in the same area as WI butperpendicular to the fracture strike, was moderately pro-ductive.

The shear waves in the surface lines from Giddingsfield show classic amplitude dim spots between the fastand slow waves. The fast wave, polarized parallel tofracture strike, show strong continuous events, while theslow waves, polarized perpendicular to fracture strike,possess weak and dim spots associated with fractureswarms. For line 4, the existence of fracture swarms isindicated by mud logs taken from a line-parallel hori-zontal well. The regional fracture strike in this areachanges from southwest-northeast at the southern edgeof the held to east-west in the north.

The VSPs show that Austin Chalk anisotropy maynot be detected by polarization and time delay variationin the transmitted downgoing wavefield from the VSPs,but may be detected by variations in amplitude on thereflected upcoming wavefield. The polarization mea-surements from the VSPs agree with those from the sur-face data within -10” error.

The use of multicomponent surface consistent cor-rection and overburden correction helps in preservingand recovering the shear wave anisotropy informationassociated with the Austin Chalk. The use of the LTT al-lows the generation of both stacked amplitude sectionsand stacked polarization logs of the split shear waves.Lateral variations in polarization (often associated withlateral variations in crack geometry) can be identifiedfrom the polarization logs, and subsurface structures canalso be better imaged than with conventional stack sec-tions. Areas of high production are found to be associat-ed with scattered polarizations and broken polarizationevents in the stacked polarization logs, which corre-spond to dim amplitude spots in the stacked fast andslow shear wave sections.

370 Li and Mueller

AcknowledgmentsWe thank Amoco Exploration and Production and

British Petroleum for contributing to the EdinburghAnisotropy Project, and for permission to publish reflec-tion and VSP data from Pearsall and Giddings fields(Amoco) and VSP data from Devine, Texas (BP). Wealso thank Gareth Yardley and Stuart Crampin for per-mission to include their VSP study in this review. Wethank David Booth and Enru Liu for their comments onthe manuscript, and Don Winterstein for his detailed re-view of the paper. The Edinburgh-based work on thisproject was supported by the sponsors of the EdinburghAnisotropy Project and the Natural EnvironmentResearch Council. It is published with the approval ofthe EAP Sponsors and the Director of the British Geo-logical Survey (NERC).

ReferencesAki, K., and Richards, P. G., 1980, Quantitative seismology:

theory and method, I: W. H. Freeman and Co.Alford, R. M., 1986, Shear data in the presence of azimuthal

anisotropy: 56th Ann. Internat. Mtg., Sot. Expl. Geo-phys., Expanded Abstracts, 476-479.

Ames, G. L., 1990, Oil and gas developments in south Texasin 1989: AAPG Bull., 74, 144-148.

Barthelemy, P., Tremolieres, P. and Andrieux, J., 1992, Frac-ture spacing related to bed thickness: geological or me-chanical bed limits?: Presented at 4th EAPG Conf.,156-157.

Chaimov, T. A., Beaudoin, G. J., Haggard, W. W., Mueller,M. C., and Thomsen, L. A., 1995, Shear wave anisotropyand coalbed methane productivity: 65th Ann. Internat.Mtg., Sot. Expl. Geophys., Expanded Abstracts, 305-308.

Champion, 0. R., 1936, The stratigraphy and structure of thePearsall field, Frio County, Texas: AAPG Bull., 20, 1.5 14.

Cliet, C., Brodov, L., Tikhonov, A., Marin, D., and Michon,D., 1991, Anisotropy survey for reservoir definition:Geophys. J. Int., 107,417-427.

Corbett, K., Friedman, M., and Spang, J., 1987, Fracture de-velopment and mechanical stratigraphy of Austin Chalk,Texas: AAPG Bull., 71, 17-28.

Crampin, S., 1981, A review of wave motion in anisotropicand cracked elastic media: Wave Motion, 3, 343-39 1.

Crampin, S., 1985, Evaluation of anisotropy by shear wavesplitting: Geophysics, 50, 142- 152.

Crampin, S., 1987. Geological and industrial implications ofextensive dilatancy anisotropy: Nature, 328, 491-496.

Crampin, S., and Lovell. J. H., 199 I, A decade of shear wavesplitting in the Earth’s crust: what does it mean, what usecan we make of it, and what should we do next?: Geo-phys. J. Int., 107, 387-407.

Crampin, S., Bush, I., Naville, C., and Taylor, D. B., 1986, Es-timating the internal structure of reservoirs with shearwave VSPs: The Leading Edge, 5, No. 11, 35-39.

Crampin, S., Evans, R., and ocer, S. B., 1985, Analysis ofrecords of local earthquakes: the Turkish DilatancyProjects (TDPl and TDP2): Geophys. J. R. Astr. Sot., 83,l-16.

Davis, T., and Lewis, C., 1990, Reservoir characterization by3-D, 3-C seismic imaging, Silo field, Wyoming: TheLeading Edge, 9, No. 11, 22-25.

Dohr, G. P., 1985, Seismic shear waves, part A: theory: Geo-physical Press.

Dravis, J. J., 1979, Sedimentology and diagenesis of the Up-per Cretaceous Austin Chalk formation, south Texas andnorthern Mexico: Ph.D. thesis, Rice Univ.

Gardner, G. H. F., and Harris, M. H., 1968, Velocity and atten-uation of elastic waves in sands: Sot. Prof. Well Log An-alysts, 9th Ann. Logging Symp., Ml -M 19.

Ebrom, D., Tatham, R., Sekharan, K. K., McDonald, J. A., andGardner, G. H. F., 1990, Dispersion and anisotropy inlaminated versus fractured media: an experimental com-parison: 60th Ann. Internat. Mtg., Sot. Expl. Geophys.,Expanded Abstracts, II, 1416-1419.

Ensley, R. A., 1984, Comparison of P- and S-wave seismicdata: a new method for detecting gas reservoirs: Geo-physics, 49, 1420- 143 1.

Hardage, B. A., 1991, Vertical seismic profiling, part A: prin-ciples: Pergamon Press.

Hudson, J. A., 1981, Wave speed and attenuation of elasticwaves in a material containing cracks: Geophy. J. R. Astr.SOL, 64, 133-150.

Keith, C. M., and Crampin, S., 1977, Seismic body wave inanisotropic media: Geophys. J. R. Astr. Sot., 49,181-243.

Kuich, N., 1989, Seismic fracture identification and horizontaldrilling: keys to optimizing productivity in a fracturedreservoir, Giddings field, Texas: Trans. Gulf Coast Ass.Geol. Sot., 39, 153-158.

Layden, R. L., 1976, Big Wells field, Dimmit and Zavalacounties, Texas, in Braunstein J., Ed,. North American oiland gas fields: Am. Assn. Petr. Geol. Mem. 27, 145- 156.

Lewis, C., Davis, T. L., and Vuillermoz, C., 1991, Three-di-mensional multicomponent imaging of reservoir hetero-geneity, Silo field, Wyoming: Geophysics, 56,2048-2056.

Li, X.-Y., 1992, Shear wave splitting in reflection surveys:theory, methods, and case studies: Ph.D. thesis, Univ. ofEdinburgh.

Li, X.-Y., 1994, Amplitude corrections for multicomponentsurface seismic data: 64th Ann. Internat. Mtg., Sot. Expl.Geophys., Expanded Abstracts, 1505-1508.

Li, X.-Y., 1995, Detecting polarization changes with depth inmulticomponent reflection surveys: 57th Ann. Internat.Mtg., Euro. Asso. Expl. Geophys., Expanded Abstracts,co34.

Li, X.-Y., in press, Fracture reservoir delineation using multi-component data: Geophys. Prosp., 44.

Multicomponent Seismic Data,for Fracture Characterization 371

Li, X.-Y., and Crampin, S., 199 1 a, Linear transform techniquefor analyzing shear wave splitting in four-componentseismic data: 6 I st Ann. Internat. Mtg., Sot. Expl. Geo-phys., Expanded Abstracts, 1, 5 l-54.

Li, X.-Y., and Crampin, S., 199lb, Complex component anal-ysis of shear wave splitting: case studies: Geophys. J.Int., 107,605-6 13.

Li, X.-Y., and Crampin, S., 1993a, Linear transform tech-niques for processing shear wave anisotropy in four-component seismic data: Geophysics, 58, 240-256.

Li, X.-Y., and Crampin, S., 1993b, Approximations to shearwave velocity and moveout equations in anisotropic me-dia: Geophys. Prosp., 41, 833-857.

Li, X.-Y., and Crampin, S., 1993c, Variations of reflection andtransmission coefficients with crack strike and crack den-sity in anisotropic media: Geophys. Prosp., 41, 859-882.

Li, X.-Y., Mueller, M. C., and Crampin, S., 1993a, Case stud-ies of shear wave splitting in reflection surveys in southTexas: Can. J. Expl. Geophys., 29, 189-215.

Li, X.-Y., Crampin, S., and MacBeth, C., 1993b, Interpretingnon-orthogonal split shear waves in multicomponentVSPs: 55th Ann. Internat. Mtg., Euro. Asso. Expl. Geo-phys., Expanded Abstracts, CO09.

Li, X.-Y., and MacBeth, C., 1995, Interpreting data matrixasymmetry and polarization changes with depth in multi-component reflection surveys: 65th Ann. Internat. Mtg..Sot. Expl. Geophys., Expanded Abstracts, 7 19-722.

Liu, E., Crampin, S., Queen, J. H., and Rizer, D. W., 1993,Velocity and attenuation anisotropy caused by microc-racks and macrofractures in a multiazimuth reverse VSP:Can. J. Expl. Geophys., 29, 177- 188.

Lynn, H. B., and Thomsen, L. A., 1990, Reflection shear wavedata collected near the principal axes of azimuthal anisot-ropy: Geophysics, 55, 147- 156.

MacBeth, C., Li, X.-Y., Crampin, S., and Mueller, M. C.,1992, Detecting lateral variability in crack parametersfrom surface data: 62nd Ann. Internat. Mtg., Sot. Expl.Geophys., Expanded Abstracts, 8 16-8 19.

MacBeth, C., Zeng, X., Yardley, G., and Crampin, S., 1994,Interpreting data matrix asymmetry in near-offset, shearwave VSP data: Geophysics, 59, I 76- 19 I.

Meadows, M. A., Winterstein, D. F., and Szymanski, D. L.,1994, New observations of shear wave polarizationsfrom surface and VSP data at Lost Hills field, California:Presented at the 6th Internat. Workshop on SeismicAnisotropy.

Mueller, M. C., 1991, Prediction of lateral variability in frac-ture intensity using multicomponent shear wave surfaceseismic as a precursor to horizontal drilling in the AustinChalk: Geophys. J. Int., 107, 409-415.

Mueller, M. C., 1994, Characterization of fracture swarms inchalked reservoirs: Abstracts of the AAPG-EAPG JointConf. on Chalk Reservoirs, Copenhagen.

Murtha, P. E., 1988, Estimation of the rotation transformationangle for shear wave data acquired in azimuthally aniso-tropic regions: Presented at the 3rd Internat. Workshopon Seismic Anisotropy.

Raikes, S. A., 199 1, Shear wave characterization of BP De-vine test site, Texas: 61st Ann. Internat. Mtg., Sot. Expl.Geophys., Expanded Abstracts, 65-68.

Scott, R. J., 1977, The Austin Chalk-Buda trend of south Tex-as: Trans., Gulf Coast Assoc. Geol. Sot., 27,164-168.

Spencer, T. W., and Chi, H. C., 199 1, Thin-layer fracture den-sity: Geophysics, 56, 833-843.

Squires, S. G., Kim, C. D., and Kim, D. Y., 1989, Interpreta-tion of total wave field data over Lost Hills field, KernCounty, California: Geophysics, 54, 1420- 1429.

Stapp, W. L., 1977, The geology of the fractured Austin andBuda formations in the subsurface of south Texas: Trans.,Gulf Coast Assoc. Geol. Sot., 27, 208-229.

Taner, M. T., and Koehler, F., 198 1, Surface consistent correc-tions: Geophysics, 46, 17-22.

Tatham, R. H., 1982, Vp/Vs and lithology: Geophysics, 47,336-344.

Tatham, R. H., 1985, Shear waves and lithology, in Dohr, G.,Ed., Seismic shear waves, part B: applications: Geophys.Press, 87- 133.

Tatham, R. H., and Stoffa, P. L., 1976, V,lV,-a potential hy-drocarbon indicator: Geophysics, 41, 837-849.

Tatham, R. H., and McCormack, M. D., 1991, Multicompo-nent seismology in petroleum exploration: Sot. Expl.Geophys.

Thomsen, L. A., 1988, Reflection seismology over azimuthal-ly anisotropic media: Geophysics, 53, 304-3 13.

Thomsen, L. A., Tsvankin, I., and Mueller, M. C., l995a, Lay-er stripping of azimuthal anisotropy from reflection shearwave data: 65th Ann. Internat. Mtg., Sot. Expl. Geo-phys., Expanded Abstracts, 289-292.

Thomsen, L. A., Tsvankin, I., Mueller, M. C., l995b, Adapta-tion of split shear wave techniques to coalbed methaneexploration: 65th Ann. Internat. Mtg., Sot. Expl. Geo-phys., Expanded Abstracts, 30 l-304.

Weeks, A. W., 1945, Balcones, Luling, and Mexia fault zonesin Texas: AAPG Bull., 29, 1733-1737.

Willis, H. A., Rethford, G. L., and Beilanski, E., 1986, Azi-muthal anisotropy: occurrence and effect on shear wavedata quality: 56th Ann. Internat. Mtg., Sot. Expl. Geo-phys., Expanded Abstracts, 479-48 1.

Winterstein, D. F., 1990, Velocity anisotropy terminology forgeophysicists: Geophysics, 55, 1070- 1088.

Winterstein, D. F., 1991, Method of layer stripping to predictsubsurface stress regimes: U.S. Patent 5 060 203.

Winterstein, D. F., and Hanten, J. B., 1985, Supercritical re-flections observed in P- and S-wave data: Geophysics,50, 185-195.

Winterstein, D. F., and Meadows, M. A., 199 1, Changes inshear wave polarization azimuth with depth in Cymricand Railroad Gap oil fields: Geophysics, 56,1349-1364.

Yardley, G. S., 1992, Identification and characterization offractured reservoirs by analysis of shear wave anisotropy:Ph.D. thesis, Univ. of Edinburgh.

372 Li and Mueller

Yardley, G., and Crampin, S., 1991, Extensive dilatancyanisotropy: relative information in VSPs and reflectionsurveys: Geophys. Prosp., 39,337-355.

Yardley, G., and Crampin, S., 1993, Shear wave anisotropy inthe Austin Chalk, Texas, from multioffset VSP data: Casestudies: Can. J. Expl. Geophys., 29, 163-176.

Yardley, G., Graham, G., and Crampin, S., 1991, Viability ofshear wave amplitude versus offset studies in anisotropicmedia: Geophys. J. Int., 107,493-503.

Yilmaz, O., 1987, Seismic data processing: Sot. Expl. Geo-phys., 2.

Zoback, M. D., and Zoback, M. L., 1991, Tectonic stress fieldof North America and relative plate motions, in Slem-mans, D. B., Engdahl, E. R., Zoback, M. D., and Black-well, D. D., Eds., Neotectonics of North America: Geol.Sot. Am., 339-366.