- Copyright © 2003 Society of Exploration Geophysicists
Multicomponent seismic data, combining P-wave and converted P-to-SV wave (C-wave) wavefields, provide independent measurements of rock and fluid properties. Unlike P waves, C waves are minimally affected by changes in pore fluids, and in cases of azimuthal anisotropy, will be split into two modes (fast and slow) with differing polarization. The 4C, 3D ocean-bottom cable (OBC) multicomponent seismic data discussed here were acquired in shallow water (<300 ft) offshore Louisiana over approximately 455 miles2 (Figure 1). Because these data are still being marketed to interested oil and gas operators, only data above targeted oil and gas reservoirs (<5000 ft) were used. The P-wave migrated data extend to only 1 s and the C-wave migrated data volume to only 2 s.
The initial objectives of the survey were to improve P-wave reflection quality by combining hydrophone and vertical-geophone data and to improve structural interpretation in the presence of “gas clouds” with C-wave data. Our additional research objectives were to evaluate seismic attributes, such as VP/VS velocity ratios and Poisson's ratio derived from P-wave and C-wave data, for characterizing shallow seafloor sediments and to develop practical poststack 3D data interpretation methodologies to capitalize on these unique characteristics. These research findings will be used later to analyze seafloor properties across gas hydrate prospects. Gas (methane) hydrates in the northern Gulf of Mexico (GOM) typically are not manifested by the classic bottom simulating reflector (BSR) and seem always associated with P-wave data “wipeouts” in gas clouds (verbal communication, Harry Roberts, Louisiana State University). The inability of conventional P-wave surveys to image the internal stratigraphy of gas hydrate systems in the GOM warrants further study. Diagnostic multicomponent seismic attributes that locate shallow hazard zones just above marine gas-hydrate deposits can optimize well placement strategy and identify areas for commercial exploitation of gas hydrates.
Preliminary analysis of the OBC data reveals that signal frequency content is about three octaves, 10–70 Hz, for the P-wave data and about two octaves, 10–30 Hz, for the C-wave data. Gas clouds pervade the shallow portion of the P-wave data volume (Figure 2), obscuring structural and stratigraphic features within the gas-charged intervals. C-waves (Figure 3) penetrate these P-wave wipeout zones and allow interpretation of previously obscured features. Shallow data are dominated by processing artifacts (muting), making interpretation at these levels challenging. Many shallow imaging problems are linked to the design of the acquisition survey and the subsequent data processing. This is not surprising because the targets were deeper hydrocarbon reservoirs. A survey designed to optimize the imaging of shallower targets would, of course, improve shallow data quality.
Correlation of P-wave and C-wave data
The greatest problem posed to the interpreter of C-wave data is correlating events with P-wave data. In the absence of shallow sonic (dipole)logs and multicomponent VSP data, evaluating both data volumes simultaneously is the only option an interpreter has for establishing P-wave to C-wave event correlation. A common first intuition is to try correlating events between the two data volumes in section view, as required in 2D data sets. A search of the P-wave and C-wave sections was made to find characteristic features that should be expected on both data sets and thus provide a correlation basis or “nail” point. The shallow section contained few geometric features such as structural or stratigraphic terminations that could be used to constrain the interpretation. An attempt to use fault surfaces for P-to-C depth correlation was made; however, the shallow faults were sufficiently steep to make the time correlation ambiguous (on the order of 50+ ms). Four “depth equivalent” horizons were picked on P-wave and C-wave sections on the basis of fault correlation. Estimates obtained from this initial depth equalization attempt revealed an unrealistic increase of VP/VS ratio with depth, so a further search for a viable correlation was undertaken.
The importance of robust correlation between P-wave and C-wave data cannot be overemphasized. Miscorrelation of key stratal surfaces will lead to erroneous seismic-based attributes (VP/VS ratio, Poisson's ratio, shear modulus, and bulk modulus) and make accurate interpretation impossible. A key issue is the comparison of P-wave and C-wave reflection characteristics or attributes at the same stratal level. Without an accurate P-to-C correlation, such attribute comparisons make no sense. With this in mind, it was determined that a correlation method based on attribute maps was needed for accurate depth registration of P-wave and C-wave images. Fortunately, this data set is a 3D data volume, thus providing a unique opportunity to use horizontal data slices as a basis for depth correlation. The complex sinuous geometry of stream channels and incised valleys produces geometric signatures on horizon slices that should uniquely correlate P-wave and C-wave times. Ambiguity should be limited to a fraction of channel thickness, and should differ only if there is a significant change in elastic moduli with depth of the channel fill. Displaying attributes in map view facilitates the search for correlation nails that are starting points to map equivalent stratal surfaces throughout each data set.
Amplitude and data similarity time slices were extracted from both data volumes at 4-ms intervals for comparison. Amplitude time slices, although valuable for evaluating general subsurface structural trends, failed to identify unique seismic events common in both P and C data sets and were not conducive to correlation. To generate data similarity time slices, a minimum crosscorrelation statistical measure was applied over an analysis window (40 ms for P-wave and 80 ms for C-wave) to optimize both structural and stratigraphic features imbedded in the data volumes. This attribute searches the data volume and systematically flags areas that have significant lateral changes in seismic reflection character, which is of great value for defining structural features (faults) or stratigraphic variations. P-wave data similarity time slices with unique characteristics were the baseline images to which the entire suite of C-wave data similarity time slices was visually compared until a compatible match, or nail, was found. The most definitive correlation was seen on the deepest data (best data quality). This example (Figures 4a and 4b) correlates P-wave time of 796 ms to C-wave time of 1964 ms. We believe this depth registration of P and C image times is accurate to better than 2%, or about 20 ms in P-wave time. P-wave characterization (Figure 5a) of these subtle stratigraphic features can be matched, in map view, to its C-wave equivalent (Figure 5b); however, the same features in section view (Figure 6) are very difficult to correlate. Channel features A through F marked on the time slices (Figures 5a and 5b) are labeled on the sections in Figure 6 to illustrate the different stratal surfaces imaged by P and C data.
Larger, more prominent stratigraphic features like incised valley systems (Figure 7) typically have a vertical footprint of more than 40 ms (P-wave), allowing for more P-to-C correlation ambiguity. For the feature in Figure 7, C-wave energy images the entire incised valley system, but the P-wave energy fails to delineate the incised valley edges in the northeast portion of the image space. Initial suspicion was that the image differences were caused by structural influence (the incised valley was out of the time-slice plane) of the data. Consequently, stratal surfaces were initiated at the nail positions and mapped throughout the two data sets. Data similarity attributes were extracted along mapped stratal surfaces, effectively removing any structural influence (i.e., flattening the horizon). Resulting images (Figure 8) closely resembled the time slices, indicating that C-wave data image different stratal surfaces from P-wave data in the northeast portion of the study area. C-wave resolution, and lack of P-wave sensitivity, confirms the advantages gained by using multicomponent data in lateral facies discriminations. As with the previous case, trying to correlate seismic events using vertical section views (Figure 9) would be very difficult, implying that 2D seismic surveys may not allow accurate VP/VS analyses, especially in shallower intervals. Areas with steep faulting, void of unique antithetic or intersecting faults, increase the possibility of mis-ties between P-wave and C-wave data sets and can lead to correlation ambiguities on the order of 50 ms or more. Depth registration of P-wave and C-wave data should not focus on such areas.
Varying sensitivities between P-wave and C-wave velocities to porosity, pore fluid content, and lithology allow the ratio of P-wave and C-wave velocities (VP/VS) to be utilized by interpreters. P-wave and C-wave velocities associated with a sequence can be used to calculate key, grossly averaged, elastic constants of the material within that sequence. Once seismic events identified in both P-wave and C-wave data are correlated and stratal surfaces subsequently mapped, time intervals can be calculated. When the source efforts are P-wave and S-wave (9 component), the ratio of correlated time intervals (Δts/Δtp) across any analysis window are equal to the velocity ratio VP/VS. In the case of OBC data, where the source effort is P-wave (downgoing) to generate C-wave (upgoing), the VP/VS relationship must compensate for mode conversion. In such a case the following relationship applies: 1
Several horizons were correlated between the P-wave (Figure 10a) and C-wave (Figure 10b) data and time intervals calculated between horizons of interest. In addition, water-bottom information was extracted from trace headers and converted to two-way traveltimes to allow characterization of the shallowest seafloor sediments. The Δts time intervals were then divided by the Δtp time intervals to calculate the ratios of correlated time intervals in equation 1. The VP/VS ratio attribute can then be used to identify specific lithologies (Table 1). The values in Table 1 apply to consolidated sediments, not to unconsolidated sands and shales in shallow seafloor strata. However, the general principle documented in Table 1 applies to shallow siliciclastics, where shale has a higher VP/VS value than sandstone. VP/VS in shallow seafloor strata can range from 5 to almost 13, with shale-dominated facies having larger VP/VS values than the sand-dominated facies.
Some problems with this approach are obvious. In zones influenced by gas clouds, velocity sags distort the true structure. The P-wave data volume is particularly sensitive to this effect. An interpreter can choose a mapping horizon influenced by the gas cloud, or force a pseudohorizon through the gas cloud to estimate the subsurface structure. In the first case, mapping horizons that sag with the gas cloud generate structural maps that are not accurate. However, if both upper and lower bounding surfaces of an analysis window are affected equally by the gas cloud, time-dependent attributes (VP/VS) calculated within the analysis window should not be greatly affected. A series of VP/VS attribute maps were generated for each interval using this approach. As expected, the VP/VS ratio decreases with depth, indicating that the correlations are reasonable. These maps (Figure 11) serve as quality control mechanisms to ensure that depth registration between P-wave and C-wave stratal surfaces is correct. Intervals that have been miscorrelated will have unreasonable (high or low) VP/VS anomalies that reveal themselves as mirror (opposite low or high) anomalies in the next interval. As intervals of interest thin, the chance of miscorrelation increases. All attempts at thin-bed analysis should rely on these quality control mechanisms to ensure robust interpretations because chances of mis-ties increase proportionally with diminishing thickness.
Figure 11 shows VP/VS ratios for the deepest interval (Figure 11a) through the shallowest interval (Figure 11d). A number of localized high VP/VS values are on either side of some faults. These anomalies may be meaningful or may be the result of picking strategies within the gas chimneys or the result of local migration problems. Missing portions of Figure 11a result where the C-wave event is below 2 s, the base of our image volume. Note the exceptionally low VP/VS ratios occurring locally along the downthrown side of the growth faults. If valid, these anomalies could represent local gas accumulations or increased sand content. Figure 11b shows the VP/VS ratio for the H2-H3 interval. The value generally ranges from 3 to 4. The lower values near 3 suggest a sand-prone interval south of the main east-west fault and on the central edge of the data. The gradual increase to the south gives confidence that this is a valid lithologic indicator in this interval. The southwest corner of Figure 11c has localized high VP/VS values on the downthrown side of a normal fault, raising suspicion that there may be a correlation error between the two data volumes. One explanation is that the high VP/VS values are imaging a paleoshelf relic deposited at an earlier time, which is geologically consistent with the basinward increase of VP/VS ratio (shale prone). Figure 11d shows the VP/VS ratio for the shallowest interval. The value generally ranges from 5 to 10. The lower values near 5 suggest a sand-prone interval north of the main east-west fault and in the southwest edge of the data. Such anomalies are worthy of further investigation.
The VP/VS ratio maps can be converted into Poisson's ratio using the relationship: 2
As seen from the formula, no new information is introduced by using this attribute. However, some disciplines are more familiar with this elastic constant, and it may produce an image having different textures by altering the dynamic range of the color display. These differences, when comparing the VP/VS ratio (Figure 11d) with the Poisson's ratio (Figure 12), may reveal subtleties in reflection character related to lithofacies variations that are not readily apparent by viewing just a single attribute. Comparing a variety of multicomponent seismic attributes (VP/VS ratios or Poisson's ratios, complex trace attributes, etc.) may be useful in diagnosing important lithologic variations and pore fluid conditions in intervals of interest.
Seismic attribute maps are useful to identify lateral facies variations not readily apparent when viewing vertical seismic sections. Combining VP/VS ratio and statistical seismic attributes (amplitude, complex trace, spectral, and sequence) from the same stratigraphic sequence can optimize shallow marine sediment characterization. Root-mean-square (rms) amplitudes were extracted from the interval between water bottom and the shallowest nailed horizon in both data sets. P-wave (Figure 13a) rms amplitudes are sensitive to sandstone-bearing sequences and are characterized by strong amplitudes (dark-blue) occurring locally along the downthrown side of the fault (southwest edge of data). Corresponding C-wave rms amplitudes (Figure 13b) in the same area are characterized by little amplitude variation, indicating the amplitude response in the P-wave data set may be due to in-situ gas. Note that same anomaly was identified by VP/VS ratio analysis and subsequently identified as a sand-prone interval. The second area of interest, a possible sand-prone interval north of the main east-west fault, was previously identified on Figure 11d. The high-amplitude responses in both P-wave and C-wave suggest that this area is void of in-situ gas, and the VP/VS ratio response may be due to a shallow sand-rich incision. By visually comparing attributes from both data sets, interpreters can capitalize on unique P and S wave propagation principles, determining subsurface characteristics with greater confidence.
Developing robust 3D interpretation-based methodologies will allow geoscientists to exploit the unique characteristics of multicomponent data. A survey designed to optimize the imaging of shallower targets would improve shallow data quality. Time-dependent attributes like VP/VS and Poisson's ratios strongly depend on accurate depth registration between P-wave and C-wave seismic events. Depth registration of P-wave and C-wave images should rely on correlating map-based images (either time slices or horizon slices), and correlation between section (2D) views should be done sparingly. Quality control mechanisms are vital to ensure robust correlation and to produce valid multicomponent seismic-based attributes for shallow marine characterization. These methodologies can serve as a road map for interpreters attempting to optimize oil and gas well placement strategies that avoid shallow hazard zones in the northern Gulf of Mexico and may identify potential areas for commercial exploitation of gas hydrates.
“3D seismic discontinuity for faults and stratigraphic features: The coherence cube” by Bahorich and Farmer (TLE, 1995). “Rock lithology and porosity determination from shear and compressional wave velocity” by Domenico (Geophysics, 1984).
This article was prepared with the support of the U.S. Department of Energy under Cooperative Agreement DE-FC26-00NT41024. However, any opinions, findings, conclusions, or recommendations expressed herein are those of the authors and do not necessarily reflect the views of the DOE. As an industry partner, Seitel Data with WesternGeco (the 4SIGHT alliance) has provided a license for showrights to the 3D data. Landmark Graphics Corporation provided software for the basic 3D seismic interpretation via the Landmark University Grant Program. Published with the permission of the director, Bureau of Economic Geology, The University of Texas at Austin.
Coordinated by Rebecca Latimer