Luncheons

Processing and Analysis of 3-D PS-wave Data for Fracture Detection: Green River Basin, Wyoming Case

James E. Gaiser

James E. Gaiser
Westerngeco, Denver, Co

Monday, January 7th, 2002 – 10:30 AM
Telus Convention Centre - 8th Ave SE, Calgary

Abstract

Summary

The investigation of S-wave birefringence (splitting) using 3-D converted P to S-waves (PS-waves) is an important tool for characterizing reservoir fractures. In azimuthally anisotropic media, fracture intensities are directly related to traveltime differences between the fast and slow S-waves. Fracture orientations can also be determined from the polarization direction of the fast S-wave. These effects are accurately analyzed in a 3-D/3-C survey from the Green River basin in Wyoming to preserve meaningful azimuthal variations in amplitudes and traveltimes. Estimates of the principal PS-wave fast and slow directions (PS1 and PS2) are made early in the processing to guide propagation azimuth limitations on the data for key processing steps including surface-consistent statics and moveout velocities. In preparation for advanced fracture analysis techniques, the data are processed in common-azimuth volumes and then all azimuths are combined using 2Cx2C Alford rotation into a single group after azimuthal residual statics. Ratios of fast (PS1) and slow (PS2) average velocity are particularly important to identify the vertical extent of overburden anisotropy, as well as reservoir horizons. Overburden effects can be removed by 2Cx2C rotation and layer-stripping analyses before characterizing deeper horizons. In addition, less quantitative attributes, which are very sensitive indicators of S-wave splitting, can be employed such as residual off-diagonal 2Cx2C amplitudes after overburden removal and isochron differences between PS1 and PS2.

Introduction

Recent interest in the use of PS-waves to help characterize fractured reservoirs has prompted the acquisition of several multicomponent surveys around the industry. Ata and Michelena (1995) used three 2-D lines centered over a well to quantify fracture information. Although the spatial coverage was sparse, azimuthal anisotropy appeared to be caused by two fracture systems. A small 3-D/3-C survey collected in the Wind River basin in Wyoming to calibrate a larger P-wave effort had some measure of success in characterizing fracture anisotropy (Gaiser, 1999; and Grimm et al., 1999).

The objective of this study was to use a PS-wave seismic survey in the Green River basin in Wyoming to quantitatively identify fractured areas in a naturally fractured Cretaceous sandstone reservoir at depths between 3,000 and 4,500 m. A 3-D/3-C survey was designed and acquired to provide wide azimuth and offset coverage at the target. The receiver lines were oriented E-W and a diagonal brick shot pattern was acquired to yield a CMP fold of approximately 24 over 50 km2. A detailed processing methodology was developed to preserve the effects of S-wave birefringence and prepare the data volume for further fracture analysis. Three analysis techniques are presented which take advantage of the wide azimuth data to provide important attributes for characterizing fractures. Also, 2Cx2C Alford (1986) rotation adapted for PS-waves (Gaiser, 1999) provided a means to combine the multi-azimuth data into a single 2Cx2C volume (the 4-components: PS11, PS12, PS21, and PS22).

Data Processing

Processing of the vertical component, P-wave data proceeded using conventional time processing techniques. Initial processing of the horizontal data included rotation to the radial and transverse components in a source-centered coordinate system (Gaiser, 1999), geometric spreading corrections, surface-consistent deconvolution, and time-variant spectral whitening. Source statics computed from the P-wave processing were also applied to the PS-wave data, as well as elevation corrections at the receivers. Preliminary stacking velocities were estimated and an initial common-conversion point (CCP) binning correction was applied to the data. Five passes of residual receiver statics were computed while iterating with additional passes of velocity analysis and anisotropic, depth-dependent CCP binning.

Two key well locations within the survey were identified for further analysis of preferred PS-wave directions. At each of these locations, a large azimuth supergather measuring 536 by 670 m (17 x 21 CCP gathers) was extracted and sorted into limited azimuth gathers. All statics were applied and the data were NMO corrected, muted, and stacked. This was done for both the radial and transverse components and resulted in azimuthal stack traces every 10 degrees (Fig. 1). The transverse component showed clear polarity reversals every 90 degrees and the radial component demonstrated a variation in traveltime with azimuth. Based on these results, the fast PS1 direction was determined to be approximately N135°E and the slow PS2 direction, N225°E at both well locations.

The results of the supergather analysis also suggested that, by limiting the radial component data to the principal directions, better quality PS-wave reflections could be obtained. The data volume was limited to the PS1 and PS2 propagation directions (+/-22.5°), stacked and migrated. Additional residual receiver-static corrections were then computed using these limited azimuth volumes and improved stacks. While the previously described PS1 and PS2 volumes proved to be better quality than the all-azimuth product, additional propagation directions, and the transverse data needed to be incorporated into the final results.

In preparation for subsequent fracture detection analysis, the entire data volume, both radial and transverse components, was divided into eight common-azimuth sectors: 0° to 360° incrementing by 45° with a tolerance of +/-22.5°. The transverse component data was processed using the same deconvolution operators, statics, and velocities estimated from the radial component data. All volumes were then migrated using the same migration velocity field. This resulted in 16 separate common-azimuth volumes of radial and transverse data. These components exhibited azimuthally varying traveltimes and, to combine them into a single dataset for improved fold and enhanced signal, 2Cx2C rotations were applied. Each 2Cx2C set was rotated into the preferred fast (PS11) and slow (PS22) directions (N135°E and N225°E) and stacked to create one set of 2Cx2C data for further analysis. Again, this increased the fold and resulted in improved signal quality when compared to the initial azimuth-limited stack.

However, small residual time shifts between PS11 and PS22 were observed in the data for each of the eight common-azimuth directions and components. To correct for this and improve the combined stack, azimuth-consistent static corrections were computed to align the radial component data in both the PS11 and PS22 propagation directions. After applying these corrections to the radial and respective transverse components for each azimuth direction the resulting stacks were significantly improved (Fig. 2).

Birefringence Analysis

One of the most important steps in using PS-waves for fracture detection is to quantify the overburden azimuthal anisotropic properties and remove the detrimental effects S-waves suffer. These properties include the orientation of the principal S-wave directions, used to identify fast and slow waves for processing, and the differential velocity between the fast and slow waves. Using either azimuth supergathers from preprocessing analyses or CDP-migrated data, these properties can be determined from 2Cx2C rotation and Winterstein and Meadows (1991) layer-stripping methods.

Another important property of the overburden is the vertical extent. One approach to estimate the extent of S-wave azimuthal anisotropy in the overburden is to analyze PS11 and PS22 velocity ratios (Gaiser, 1996) as a function of two-way time. Figure 3 shows an analysis between the fast and slow data located at the well in the northern part of the survey. The vertical axis is PS11 two-way time and the horizontal axis is the ratio of Vps11/Vps22 average velocity. Variable density represents positive cross-correlation coefficients between the two waves. Depending on the velocity ratio, PS22 is stretched (< 1.0) or compressed (> 1.0), and then correlated with the PS11 at predefined window times. Contours superimposed on the plot indicate constant time delays of PS22 in milliseconds.

A maximum correlation trend can be clearly interpreted and is indicated by the dashed line. This corresponds to the time-variant velocity ratio of Vps11/Vps22. Above 1.0 s the trend is unknown. However, below 1.0 s the trend increases to a maximum at about 1.3 s and then roughly follows a 30 ms PS22 time delay. In the absence of the upper 1.0 s of data, the base of the overburden can be interpreted at about 1.5s. It represents an interval of relatively homogeneous, azimuthally anisotropic material with constant orientation confirmed by azimuth-supergather analyses. Notice that the trend increases to the 40 ms PS22 time delay contour at about 2.5 s, indicating an increase in S-wave birefringence.

Further processing of the overburden involves layer stripping, or removing the azimuthal anisotropic effects imparted on the PS-wave data by this shallow interval. This can be accomplished over a time window from 0.0 to 1.6 s. The resulting fast S-wave direction and slow S-wave time delays of the overburden are smoothed spatially and then applied to the poststack data.

After layer stripping the overburden, the data are in position for further analysis to determine principal S-wave directions and percent anisotropy for deeper intervals. Figure 4 shows the 2Cx2C-inline section that intersects the well at the vertical white line. PS11 and PS22 are the principal components and are aligned down to the base of overburden at about 1.6 s. Also, the off-diagonal components (PS12 and PS21) have been minimized down to this same event.

One approach to analyze target horizons below the overburden is to interpret residual amplitudes on the off-diagonal components spatially and temporally. These amplitudes result when there are changes in the S-wave birefringence properties, such as the increase at 2.5 s in Figure 4, and a change in the principal directions. Such changes may be too subtle to be quantified by traditional layer-stripping methods or the velocity ratios, but amplitudes are sensitive to these variations in S-wave properties and can give qualitative insights into regions where fracturing may be more intense and change orientation. Figure 5 shows example maps of residual amplitude at about 1.9 and 2.5 s (arrows in Fig. 4). The horizontal white line indicates the location of the inline data from Figure 4.

A more quantitative approach is to measure traveltimes between reflections bracketing targets for both the PS11 and PS22 waves in the volume in Figure 4. By comparing isochron differences between PS11 and PS22, a spatial representation of percent anisotropy can be interpreted for lateral variations in fracture intensity. The assumption here is that there is little or no change in the orientation of the principal axes, since rotations are not involved. Constant principal-axis directions are observed in azimuth-supergather analyses located at the two wells indicated in Figure 5.

The third approach, which is most quantitative, involves further layer stripping below the overburden. It is important to remember that the property we are analyzing for fracture detection is the transmission effect on S-waves: the birefringence orientation and the time separation. Off-diagonal components are minimized further by 2Cx2C rotations to estimate any change in the direction of the principal axes. After minimization, the separated PS11 and PS22 waves are correlated and PS22 is aligned with PS11. These properties can then be interpreted as lateral variations in fracture orientation and intensity.

Conclusions

Early estimation of the principal S-wave orientation is critical to optimize processing for fracture characterization. This can be accomplished using azimuth supergather analyses at selected locations. Once these directions are determined, propagation azimuths may be limited to improve signal quality for various data processing steps. Rotating to the fast and slow PS-wave directions can improve surface consistent static corrections and velocities. Rotating back to radial and transverse and processing common-azimuth volumes allowed all the data to be combined using 2Cx2C rotation and used in subsequent analyses. This increased fold helped improve the signal quality, but only after azimuth-consistent static corrections were computed and applied.

Vps11/Vps22 velocity ratio analysis is a valuable tool to quantify the vertical extent of S-wave birefringence as a function of time. As this ratio varies in time, it indicates different layers in the subsurface where birefringent properties may have changed. One of the most important of these layers is the overburden, which can be removed effectively by 2Cx2C rotation and layer-stripping analyses, and leads to quantifying the fracture properties at target horizons. Three analysis techniques (residual off-diagonal amplitudes, PS11 and PS22 traveltime isochrons, and layer stripping) provide a broad range of interpretation tools and attributes that can help identify lateral variations in fracture properties.

Acknowledgments

We thank Jonathan Fried at WesternGeco for his help in processing the converted-wave data for this study. Thanks also to EOG Resources, Tom Brown Inc., Basin Exploration and ExplorTech for their participation and for providing valuable insights into the nature of the reservoir.

References

Alford, R.M., 1986. Shear data in the presence of azimuthal anisotropy: Dilley, Texas. 56th Ann. Internat. Mtg., Soc. Expl. Geophys., Houston, Expanded Abstracts, 476-479.

Ata, E., and Michelena, R.J., 1995. Mapping distribution of fractures in a reservoir with P-S converted waves. The Leading Edge, 14, 664-676.

Gaiser, J.E., 1996. Multicomponent Vp/Vs correlation analysis. Geophysics, 61, 1137-1149.

Gaiser, J.E., 1999. Applications for vector coordinate systems of 3-D converted-wave data. The Leading Edge, 18, 1290-1300.

Grimm, R.E., Lynn, H.B., Bates, C.R., Phillips, D.R., Simon, K.M., and Beckham, W.E., 1999. Detection and analysis of naturally fractured gas reservoirs: multiazimuth seismic surveys in the Wind River basin. Wyoming. Geophysics, 64, 1277-1292.

Winterstein D.F., and Meadows M.A. 1991. Shear-wave polarizations and subsurface stress directions at Lost Hills Field. Geophysics, 56, 1331-1348.

Biography

James Gaiser, Principal Research Geophysicist: James Gaiser received his M.S. degree in geophysics from the University of Utah in 1977 and his Ph.D. in geophysics from the University of Texas at Dallas in 1989. In 1977, he joined ARCO in their geophysical analysis and processing group, and moved to research and development in 1981 where he worked on vertical seismic profiling, elastic wave anisotropy and shear-wave seismology. In 1992, he joined Western Geophysical research in Englewood, Colorado as a senior research geophysicist and is currently coordinating WesternGeco's converted-wave research in Denver. His research interests are in 3-D multicomponent, converted-wave seismology, imaging in anisotropic media and noise attenuation. He is a member of SEG, EAGE, EEGS and RAS.