In 2002, Saudi Aramco conducted its first 3D, 4-component (4C) ocean-bottom cable (OBC) seismic survey in the Arabian Gulf. The main objective was to delineate the middle Cretaceous Upper Khafji Sand Stringers Reservoir overlying the massive Main Khafji Sand Reservoir in the Zuluf field. The Upper Khafji Sand Stringers Reservoir in the Wasia Formation is typically characterized by weak acoustic impedance contrasts. A pre-survey modeling study, based on the logs of compressional (P) and shear-wave (S) velocities (Vp and Vs), indicated that converted compressional-to-shear waves (P-S) could better-image the structure and stratigraphy of the target reservoir. Commensurate with the objectives of the experiment, a pilot 100-square-kilometer survey was acquired with an inline swath-shooting geometry that employed two seabed receiver cables, with a symmetric split-spread deployment of the 4-C sensors. The acquisition geometry consisted of six sail lines per swath with a single-boat, dual-source, flip-flop configuration.
The data were processed through dual-sensor summation, horizontal-component rotation and P-P/P-S pre-stack time migration. Post-stack enhancement using non-stationary Gabor deconvolution proved beneficial in compensating for the missing high frequencies in the acquired converted-wave data. Well-to-seismic calibration for both P-P and P-S data at five wells aided in the interpretation of the data. Five horizons were interpreted and correlated between the P-P and P-S sections. The horizons were analyzed using both amplitude and interval times such that the lateral variations of the Vp/Vs ratio of the Upper Khafji Sand Stringers Reservoir could be mapped. A region of low Vp/Vs ratios in the northwest quadrant, obtained from the isochron interval-time analysis, was correlated with higher ‘net sand’ pay at a hidden well located in the middle of this region. These results were further corroborated by seismic facies analysis and provide a qualitative reservoir quality index in the Upper Khafji Sand Stringers Reservoir.
In 2002, Saudi Aramco conducted its first 3D, 4-component (4-C) ocean-bottom cable (OBC) seismic survey in the Arabian Gulf. The main objective was to evaluate seafloor multicomponent seismic technology as a means of delineating the Upper Khafji Sand Stringers Reservoir that overlies the massive Khafji Main Sand Reservoir in the offshore Zuluf field (Figures 1 to 3). The targeted Upper Khafji sandstone stringers occur at a depth of about 6,100 ft, vary in thickness from 10 to 70 ft, and are important for maximizing production from this field. This paper reviews the stages of the project from the pre-acquisition feasibility study to the final interpretation and data analysis results.
Figure 2 shows an E-W 800-km generalized stratigraphic cross-section of the lower and middle Cretaceous successions in Saudi Arabia; the section is datumed at the top of the Aptian Shu′aiba Formation (Figures 2 and 3). The Khafji sands were deposited in a fluvial-deltaic setting and were sourced from the west where the clastic section is thickest. Their depositional framework was controlled by sea-level fluctuations that reflected several transgressions and regressions. For example, the lower clastics of the Khafji Member prograded farthest eastwards after the late Aptian sea-level drop associated with top-Shu′aiba unconformity. In the Zuluf field, the Upper Khafji Sand Stringers Reservoir is well-developed (Figure 2).
Figure 3 shows the stratigraphic column in the Zuluf field, together with a representative gamma-ray log of the entire Khafji Member of the Wasia Formation. The gamma-ray log illustrates the difference in response between the Khafji Main Sand Reservoir (massive amalgamated deltaic sandstones between 6,100 and 6,300 ft) and the Upper Khafji Sand Stringers Reservoir. The latter consist of alternating sequences of fluvial-deltaic channel sandstones and occur between the Dair Limestone and the Khafji Main Sand Reservoir.
Due to the density of sea-surface facilities (platforms, rigs, etc.) it became evident that the data should be acquired with ocean-bottom cable (OBC), rather than a streamer. However, deciding upon the optimum type of OBC sensors (i.e. 2-component versus 4-component) was less straightforward. In order to assess the potential benefits from P-S converted-wave acquisition, in relation to the target reservoir, a detailed feasibility study was undertaken.
Full elastic modeling, based on density, sonic and full-waveform processed dipole-shear logs in five wells in the Zuluf field was an important part of the feasibility study. An example of the modeled P-P (vertical) and P-S (radial component) seismic response versus offset, at well A, is shown in Figure 4. The synthetic seismograms indicate that the P-P response from the Upper Khafji Sand Stringers Reservoir is weak and prone to polarity reversals at offsets of around 1,400 m (4,600 ft) that correspond to angles of incidence of 18°.
In contrast, the converted-wave P-S response is considerably stronger and monotonically increases within the expected usable offset range from zero to 2,000 m (6,560 ft). The top of the Shu′aiba Formation is a consistently strong reflection in both the P-P and P-S synthetic gathers. Notice also, in Figure 4, the correlation between low Vp/Vs (1.8 or less) and low gamma-ray (40 API units or less) in both the Safaniya and Khafji clastic reservoirs. This correlation is more evident in Figure 5b, which displays a cross-plot of Poisson’s ratio (or equivalently Vp/Vs) versus gamma-ray at well A. In contrast, Figure 5a shows that no obvious correlation occurs between Vp and gamma-ray. For example, for the clean sands with gamma-ray values of about 35 API, the Vp values range widely between 9,000 and 13,000 ft/sec. These observations imply that Vp alone cannot differentiate between sand and shale in the Khafji reservoir, while the Vp/Vs ratio can discriminate lithology and provides a reservoir-quality index for the pay sands.
In addition to the modeling study, a 3-component limited-offset vertical seismic profile (VSP) was acquired as part of the feasibility study. The analysis of the horizontal component data indicated that significant amounts of downgoing shear-wave energy was converted from P to S waves near the seabed. Moreover strong converted-wave energy was generated at the main horizons in the clastic section.
The above observations, coupled with the modeled strong P-S reflectivity at the top of the Khafji Reservoir (KFJR), pointed to significant potential benefits from multicomponent seismic. Accordingly, a pilot 3D-4C survey covering a surface area of 100 sq km was acquired over the southern part of the Zuluf field (Figure 1).
SURVEY DESIGN AND DATA ACQUISITION
Wide-versus narrow-azimuth acquisition geometries were numerically simulated as part of the feasibility and survey design study. Both orthogonal and inline shooting geometries were numerically tested. Patch geometries with shot lines that are orthogonal to receiver lines are generally preferred because they better-image complex structures. The data can also be used for azimuthal-anisotropy analysis for fracture detection (e.g. Angerer et al., 2006). Since neither fractures nor structural complexity are major features in the study area (maximum dips at the target are typically one to three degrees) the parallel inline shooting geometry was selected (Figure 6). This decision was further supported by the results of the numerical modeling, which indicated that an inline survey would result in higher nominal fold and lower footprint contamination than an equivalent crossline survey.
The average water depth in the survey area is 36 m. Source and receiver lines were oriented at 131/311 degrees azimuth in alignment with the predominant direction of underwater currents in the area. The source, an array of five 1,560 in3 air-guns, was operated at a depth of 5 m.
As a first step, shot and receiver static shifts were applied to bring the data to the mean sea level as the reference datum. Vertical component data were processed through surface-consistent dual summation of the geophone and hydrophone recordings in order to attenuate the water-layer reverberations and to regularize their frequency response. Subsequently, an adaptive subtraction process was applied to suppress mud-roll coherent noise that featured quite prominently. The pre-stack data were also characterized by the presence of strong guided-wave energy that restricted the useful offset range to a maximum of 1,800 m (5,904 ft) at the target reservoir level.
The horizontal component data were first oriented, with the aid of direct P-wave arrivals, and finally rotated into true radial and transverse components. Shear-wave receiver statics, analyzed with the aid of common receiver stacks on the radial component, were found to be moderate, with maximum values not exceeding ± 20 msec. Subsequently, both vertical and radial component data were processed through converted-wave pre-stack time migration (PSTM), using a Kirchhoff-based approach (developed by Wang and Pham, 2001, and similar to the equivalent offset migration of Bancroft et al., 1998). The converted-wave stacking velocity analysis was performed on a 450 x 500 m grid of P-S pre-stack migrated gathers generated from the full dataset after migration mapping (same locations as for the P-wave dataset). Percentage stacks, move-out corrected gathers, velocity-cube and semblance analysis served as tools for quality control. The final velocity field was used to correct for move-out the P-S-migrated gathers using a non-hyperbolic NMO correction.
The frequency content of the converted-wave data was exceedingly narrow, possibly as a result of poor receiver-cable coupling with the unconsolidated seabed sediments. Analysis of the transverse component data, in the early stages of data analysis, also provided some indication of vector fidelity problems, i.e. leakage between the inline and crossline horizontal component recordings. To address the issue of the missing high frequencies in the stacked P-S recordings, we attempted various poststack, high-frequency restoration techniques for the radial component data. The best results were obtained with Gabor deconvolution (Henley and Margrave, 2001). This is a Gabor transform-based deconvolution method, which, unlike conventional deconvolution methods, does not assume that the seismic data are stationary in time (Margrave and Lamoureux, 2001). Discarding the nonstationarity assumption provides an important advantage, especially when dealing with converted-wave P-S data, given the higher attenuation of shear waves.
A comparison between the migrated P-S stack sections, before and after Gabor deconvolution, shows significant improvements in vertical resolution (Figure 7). Moreover, the comparison of the spectra of the P-S data, before and after post-stack Gabor deconvolution, further demonstrates that signal amplitudes in the frequency range of 10–30 Hz have been boosted by at least 10 dB (Figure 8).
Five wells (A to E) with full suites of log curves, including dipole shear, were used for calibrating the well-to-seismic ties and for wavelet extraction. This type of calibration work is an important step prior to the interpretation and inversion of conventional P-P data; it becomes even more critical when dealing with the combined interpretation of P-P and P-S data. This is because mode-converted P-S reflectivity sequences can often be quite different than P-P reflectivities, resulting in reflection character differences that make event correlation between the P-P and P-S sections far from straightforward.
The first step of the calibration process was to compute synthetic P-P and P-S seismograms based on statistical wavelets that were extracted from the P-P and P-S volumes. These synthetics were next correlated with the field seismic traces in the vicinity of the available wells to obtain a preliminary calibration. The initial statistical wavelets were then upgraded into deterministically derived ones using the well logs in the wavelet-extraction process in order to obtain the wavelet phase. The well-to-seismic correlation was then repeated.
In general, adequate calibrations and laterally stable wavelets were obtained for both the P-P and the P-S data, with average correlation coefficients of about 0.56. The phases of the extracted P-P and P-S wavelets were 180° and 150°, respectively. Figure 9 shows the calibration results for the converted-wave (P-S) data in the vicinity of well B. The extracted wavelet is also shown. A 600-msec-correlation window, from 1,250 to 1,850 msec, was used for the calibration; the correlation coefficient was 0.58. The displayed synthetic P-S traces were computed by stacking synthetic gathers over the offset range from 100 to 1,500 m. Unlike P-wave synthetics that can be computed from the zero-offset P-P response, P-S synthetics can only be obtained from modeled offset gather stacks. This is because the amplitude of the converted-wave P-S response at zero offset is, theoretically, zero. Figure 10 displays the inline P-P stack section intersecting well B, with the calibrated P-wave synthetic trace inserted at the well location.
Using the results of the well-to-seismic calibration for guidance, we interpreted a number of key horizons. Some of them, for example the top Caprock Limestone (lower part of the Ahmadi Member, Figure 3) above the Khafji reservoir and especially the top Shu′aiba Formation, were sufficiently coherent to pick consistently and reliably in both the P-P and P-S volumes. Others, like the top Main Khafji Sand Reservoir (MKFJ), were difficult to pick, and were therefore left-out during the next step of data analysis, which consisted of the computation of Vp/Vs ratios.
Interpretation of the P-S volume, in particular, was difficult due to the low frequency and poor continuity of the seismic events. Attempts at automatic horizon picking of the converted-wave data proved unsuccessful and so we resorted to manual picking on a 20-by-20 inline-crossline grid. The use of the cosine phase seismic attribute superimposed with color on the seismic sections helped stabilize the interpretation by ensuring phase-consistency of the interpreted picks. As a quality-control measure, emphasis was placed on verifying that the interpreted horizon picks of the grid tied at the inline-crossline intersection points.
Figures 11a and 11b show interpreted inline and crossline P-S sections, respectively, centered at well B. The P-S synthetic is also shown at the location of the well. Notice the importance of having a good well-to-seismic tie for guiding the interpretation and the subsequent event-correlation process. The structural configuration of the interpreted key Shu′aiba horizon of the converted-wave data is shown as a PS-time horizon map (Figure 12).
Interval Vp/Vs Estimation
Following the interpretation and event correlation of the same reflections in both P-P and P-S
volumes, it becomes possible to compute interval Vp/Vs ratios for any selected pair of interpreted horizons. The ratio of compressional to shear-wave velocity is a powerful indicator of lithology, which is capable of differentiating between sand and shale, as discussed earlier in the feasibility study section. The so-called isochron method uses the P-P and P-S transit times between two horizons and has been successfully used in both carbonate and clastic settings (e.g. Miller, 1996; Macrides and Kelamis, 2000).
The Vp/Vs ratios were computed from the P-S and P-P two-way transit times, ΔTps and ΔTpp, using the formula Vp/Vs = 2 × ΔTps/ΔTpp – 1. In the present study, and in the absence of a consistent interpretation of the MKFJ (top Main Khafji Sand Reservoir), the analysis window was from the KFJR reflection (top Khafji Reservoir) to the top of the Shu′aiba reflection (Figure 13). The corresponding P-S stacked section in the vicinity of well B is also displayed in Figure 13. The ATps transit-time between the KFJR (top Khafji Reservoir) to top Shu′aiba reflections was 160 msec for well B. Although our immediate target was the 200-ft-thick Upper Khafji Sand Stringers (from KFJR to MKFJ in Figure 13), it was not necessary to use reflections from the top and base of that interval when performing the ΔTps/ΔTpp analysis. This is because the thicker section from KFJR to top Shu′aiba (approximately 500 ft) contains the 300-ft-thick Main Khafji Sand Reservoir, which is known to be uniform in thickness and lithology throughout the area. Therefore, any spatial variations that are observed in the ΔTps/ΔTpp ratio from this larger interval generally represent lateral lithology variations within the Upper Khafji Sand Stringers above the Main Khafji Sand Reservoir.
Furthermore, even if a consistent interpretation of the MKFJ was available, using it to estimate Vp/Vs would not be appropriate. This is because the KFJR to MKFJ interval (with ΔTps = 60 msec and ΔTpp = 40 msec, Figure 13) is too thin for the isochron method. For example Miller (1996) showed that the isochron method can become unstable when analyzing intervals thinner than 100 msec because small errors in the interpreted horizon time picks result in large errors in the estimated Vp/Vs ratios. This instability is similar to the case of the Dix method for estimating interval velocities from pre-stack gathers.
The final map of the Vp/Vs calculations from the isochron method is shown in Figure 14. The low Vp/Vs ratios (about 1.8) in the northwest quadrant appears to indicate a zone of cleaner sands and higher ‘net-sand’ pay within the Upper Khafji Sand Stringers Reservoir. This prediction was confirmed by introducing well F as a blind test (Figure 14). This well was not used in the well-to-seismic calibration as it lacked dipole shear logs. Calculations based on gamma-ray logs in that well, using a cut-off value of 35 API, indicated a ‘net-sand’ of 60% for the KFJR to MKFJ interval. This is in contrast with the lower ‘net-sand’ percentages, and higher Vp/Vs, estimated in other wells in Figure 14 (e.g. a ‘net-sand’ of 15% and Vp/Vs of 2.27 at well E).
Figure 15 shows a comparison of the logs and ‘net-sand’ calculations between wells E and F. These observations confirm that the obtained low values of Vp/Vs in the northwest quadrant (Figure 14) are diagnostic of higher ‘net-sand’ percentages in the Upper Khafji stringers. In that sense, the observed lateral variations in the computed Vp/Vs ratios in Figure 14 represent a qualitative sand-quality index map of the Khafji stringers reservoir in the Zuluf field.
In Figure 14, the range of Vp/Vs ratios is somewhat larger than expected for the known lithological variations in the area. This result occurs because the Vp/Vs ratios obtained from the isochron method represent the relative lateral variations and not the absolute values (Tatham, and McCormack, 1991). This is because noticeable differences in frequency content and phase characteristics between the P-P and P-S data (a situation that occurs in this study) readily introduces a bias in the Vp/Vs calculations.
Seismic Facies Analysis
Further to the estimation of Vp/Vs ratios using the isochron method that is based entirely on P-P and P-S reflection times, we also used the independent relative seismic amplitude information to analyze the lithology of the Upper Khafji stringers. A neural-network-based seismic-facies classification method was used for this study. The method is based on the analysis of the full-waveform seismic response of a targeted time window, here anchored at the top Shu′aiba reflection. This essentially pattern-recognition approach can detect lateral changes in the seismic character. Because the main objective was lithology-discrimination rather than fluid-detection, we performed the seismic facies analysis on the converted-wave P-S post-stack data.
Two approaches were used: (1) an interval map classification, and (2) a sample-by-sample volume classification. The results of the first method, using 15 facies, are mapped in Figure 16. Several other numbers of classes were tried, with 15 facies providing the most detailed and yet interpretable results. The time-window used for the classification of the seismic response was 120 msec and centered 100 msec above the top-Shu′aiba reflection in order to isolate (to the greatest extent possible) the Upper Khafji Sand Stringer Reservoir and avoid the basal shales of the Khafji Member.
The average model traces corresponding to each of the 15 facies are shown in Figure 17. In Figure 16, the highlighted zone of ‘reddish’ colored facies corresponding to Facies #2 to #5. The shape and areal extent of this zone bears considerable resemblance to the zone of low Vp/Vs ratios in Figure 14 and provides additional stratigraphic detail of its internal composition. The model traces of Facies #2 to #5 consist of a strong trough at the top of the window followed by a peak. According to the modeling results this is consistent with the occurrence of a clean sand stringer in the upper part of the Khafji Reservoir. Therefore it seems apparent that the two maps (Figures 14 and 16) provide complementary stratigraphic information by delineating Khafji sand-prone areas.
Figure 18 shows the results from the seismic facies volume classification of the P-S data for the entire Khafji interval, for a seismic section (inline) that crosses well B. This sample-by-sample volume-classification approach essentially provides additional temporal resolution. In contrast to the trace-by-trace interval classification method, it estimates sample-by-sample instantaneous facies based on seismic amplitudes in the analysis window. In contrast, the map facies analysis method (Figure 16) provides only an average seismic facies at each location, compressing the waveform information contained in the analysis window of the seismic trace into a single facies number.
In the volume classification results (Figure 18), the interval between the top of the Khafji Reservoir and the Main Khafji Sand Reservoir (KFJR to MKFJ interval) is characterized by a yellow-to-red colored facies above a blue colored facies. The upper yellow-to-red facies corresponds to a clean sand stringer consistent with low gamma-ray values in well B. The blue colored facies corresponds to a shaly interval of high gamma-ray values in the lower half of the interval. The mismatch between the computed instantaneous facies and the measured gamma-ray values in the lower part of the Main Khafji Sand (MKFJ) to top Shu′aiba interval is caused by a severe wash-out evidenced by the caliper log. The drilling-mud invasion in the wash-out zone, gives rise to the observed abnormal high gamma-ray count.
Conventional P-P and converted-wave P-S multicomponent seismic data were acquired in a pilot 3D OBC survey over the Zuluf field in the northern Arabian Gulf, in order to map the mid-Cretaceous Upper Khafji Sand Stringers Reservoir that overlies the Main Khafji Sand Reservoir. Five wells with full suites of logs, including dipole shear wave, were used to calibrate and interpret the seismic data for both the P-P and P-S volumes. Three methods were used to interpret the ‘net-sand’ pay in the Upper Khafji Sand Stringers Reservoir.
The isochron method used the P-P and P-S transit times to map Vp/Vs ratios over an interval that encompassed the Upper Khafji Sand Stringers Reservoir. A region with consistently low Vp/Vs ratios, which are diagnostic of higher ‘net-sand’ pay, was identified and confirmed by a blind-test well located in the middle of the low Vp/Vs region.
The second method used a neural-network-based seismic facies analysis of the converted-wave data to perform full-waveform classification of the seismic response over a seismic window encompassing the Upper Khafji Sand Stringers Reservoir. The result from this method classified 15 seismic facies that were mapped in the area. The distribution of sand-prone facies were in general agreement with the region of low Vp/Vs ratios identified with the isochron method.
The third method involved a sample-by-sample volume facies classification technique, which adds temporal resolution to the conventional trace-by-trace facies classification results. In some locations, it identified individual sand stringers in the Upper Khafji Sand Stringers Reservoir.
All three methods worked consistently towards detecting and delineating lithology variations in the Upper Khafji Sand Stringers Reservoir. The low-frequency content of the converted-wave data prevented us from fully resolving individual thin-sand stringers in the reservoir. Nevertheless, the results are expected to aid in the identification of sand-prone intervals in the overall Khafji Reservoir interval.
We thank Scott Eberle for his help in the data interpretation. Adil Al-Khelaiwi’s support and encouragement in the early stages of the survey evaluation design were instrumental for launching the project. We thank Turki Al-Ghamdi for his work and logistical support during the data acquisition phase of the study. The comments by GeoArabia’s editors and two anonymous referees are greatly appreciated. We thank the management of Saudi Aramco and the Saudi Arabian Ministry of Petroleum and Minerals for permission to publish this paper. The final design and drafting by GeoArabia Graphic Designer Arnold Egdane is appreciated.
ABOUT THE AUTHORS
Costas G. Macrides joined the Geophysical Research and Development Division of Saudi Aramco in 1993. He received a BSc in Physics from the University of Athens in 1980, and an MSc and a PhD in Geophysics from the University of Alberta in 1983 and 1987, respectively. He has been an Assistant Professor of Geophysics at the University of Manitoba and a Senior Research Geophysicist with Seis-Pro and Consultants in Calgary. Costas is currently a Geophysical Specialist with Saudi Aramco’s Geophysical Technical Services Division. His technical interests are currently focused on post-stack and pre-stack inversion, AVO, and multi-component seismology in oil and gas exploration. He is a member of the EAGE, SEG and DGS.
Fernando A. Neves is a Senior Reservoir Geophysicist with Shell International E&P seconded to Petroleum Development of Oman (PDO). He received a PhD in Geophysics (1996) from University of Cambridge, England and a Post-Doctorate from Colorado School of Mines, USA. Fernando worked for Petrobras (Brazil), Petroleum Geo-Services (USA and Norway) and Saudi Aramco (Saudi Arabia), in various capacities as Explorationist, Production Geoscientist, Seismic Interpreter and Reservoir Geophysicist. Since 2007 Fernando has been a staff member of the Exploration Directorate of PDO working in the quantitative interpretation team. He is an active member of the Geological Society of Oman, EAGE, SEG and AAPG. Fernando has published ten full articles and presented in several international conferences.