Advances in interbed multiples prediction and attenuation: Case study from onshore Kuwait Adel El-Emam* and Khaled Shams Al-Deen, Kuwait Oil Company; Alexander Zarkhidze and Andy Walz, WesternGeco Introduction Multiples contamination both surface and interbed related is a problem in almost all Middle East basins. The high acoustic impedance of carbonates and anhydrites layered with clastics is the major generator of these interbed multiples. These types of multiples are known to hinder the interpretation, fracture characterization, and inversion studies; they significantly complicate both the structural and stratigraphic interpretation within the zone of interest at the Cretaceous level as well as at the frontier Jurassic and Permian sections (El-Emam et al., 2001). Previous work has demonstrated marginal success in attenuating the main interbed multiples using the 1D multiple modeling technique post-migration thorough the analysis and identification of the major multiple generators using well data (El-Emam et al., 2005). This paper presents a case study describing the application of several data-driven multiple attenuation techniques using more advanced true-azimuth algorithms prior to prestack migration. The algorithms applied in this study are general surface multiple prediction (GSMP), extended interbed multiple prediction (XIMP) and deterministic interbed demultiple (DID). Multiple modeling and subtraction were performed on a high-resolution full-azimuth dataset acquired in northwest Raudhatain oil field, onshore Kuwait. Study Area A full-azimuth 3D survey with square-patch geometry was acquired over 115 km 2 in northwest Raudhatain oil field, (Figure 1), using a point-receiver acquisition system. The point-receiver interval was 6.25 m, staggered in four sublines to output 25-m group intervals after digital group forming (DGF). The full-azimuth, square-patch geometry with shot-line and receiver-line intervals of 200 m and station intervals of 25 m yielding a nominal fold over 900 was considered the most suitable design for this study, (Figure 2). Figure 2: Source-Receiver location map (to the right) and geometry template (to the left). Surface-related multiple analysis and prediction Prior to any multiple removal attempts, careful preconditioning is required to ensure the best possible signal-to-noise ratio. Several passes of ambient and coherent noise attenuation and intra-array perturbation corrections before digital group forming were performed to enable generating reliable multiple models. In order to effectively handle the multiples present in the data, the sequence started by addressing the surface-related multiples. 3D GSMP was used to predict and adaptively subtract the surface multiples. The advantage of this technique is that, almost no preconditioning is required in terms of interpolation, regularization, and extrapolation; these are carried out on-the-fly and all calculations were done from the smoothed surface, consequently this technique has minimal assumptions and successfully overcomes the challenges of sparse, missing, or irregularly spaced traces (Moore et al., 2008). This algorithm predicts the multiples at true azimuth, ensuring that the multiples model accurately matches the multiples in the input data. Figure 3 shows an example of a prestack time migration stack section before and after the process. Figure 1: NW Raudhatain oil field survey location map. SEG San Antonio 2011 Annual Meeting 3546
Although the results do not show a significant difference on the stack data, it is an important step in the multiple attenuation workflow because all the subsequent interbed multiple attenuation algorithms assume that the data are free of surface multiples. Figure 4 is an example of gathers and Figure 5 gives semblance plots that demonstrate that the surface-related multiples are mainly represented by a slow trend and are well attenuated by this technique. Interbed multiple analysis and prediction The XIMP algorithm is similar to the 3D GSMP and has most of its advantages. It is a true-azimuth 3D algorithm based on the method described by Jakubowicz (1998). The technique has the same requirements as conventional interbed multiple prediction (IMP) methods; it still requires identification of the multiple generators, but the iterative top-down methodology required in the past was replaced by simultaneous prediction for all identified generators. The adaptive subtraction workflow was modified to account for the simultaneous prediction. Several horizons were identified as multiple generators: namely Dammam, Rus, Hartha, Mutriba, Mishrif, Ahmadi, Mauddud, Zubair, Minagish, and Gotnia, these horizons were interpreted on the pre-migrated stack data for the entire volume. Each horizon was then used to predict its relevant multiples. Figure 6 shows two examples of the predicted multiple models generated by the Rus and Mutriba formations. Due to the relatively low signal-to-noise ratio of the shallow data, and to address specifically the Dammam-to- Rus multiple generating interval, the DID technique was employed. Originally designed to handle shallow watercolumn related multiples, it was later adopted to attenuate interbed multiples (Moore et al., 2006). This algorithm is designed to be applied in cases where the period of the multiple is already known. This technique employs a model-driven, non-linear multiple prediction approach and accurately derives both first- and higher-order multiple amplitudes. The subtraction of various combinations of these multiple models was tested using different QC tools such as semblance plots and well ties to determine the most appropriate results. Ultimately, the models from only seven generators (Dammam, Rus, Hartha, Mutriba, Mishrif, Ahmadi, and Mauddud) were identified as the major contributors to the multiple contaminations and used for final subtraction. Adaptive subtraction For various reasons such as imperfect geometry, inadequate sampling of wavefield, the assumptions of the algorithms, low signal-to-noise ratio, complexity of the geology, and others, the multiple models will always have some timing, amplitude, and phase errors that must be adaptively matched to the input reference data before subtraction. The various interbed multiple models predicted using XIMP and DID were simultaneously and adaptively matched to the input seismic data using least-squares filters and subtracted. This approach provided the ability to match each model in a window, and the results were determined by taking into account the quality of each multiple model. Conclusions Multiple contaminations in Kuwait seismic data impact the structural and stratigraphic mapping accuracy and reservoir characterization reliability within known reservoir formations. Previous studies concluded that those multiples cannot be easily attenuated using conventional methods; it was also concluded that these multiples respond well to the data-driven techniques such as IMP. This case study demonstrates the use of the latest industry multiple attenuation techniques that utilize 3D true-azimuth data-driven algorithms with no need for regularization or interpolation and produce superior results. In addition, all multiple attenuation algorithms are applied prestack and pre-migration; consequently, the subsequent velocity model building, subsurface imaging, and prestack inversion are deemed to be more robust in the absence of the multiples. The results have been verified through various QC tools including well ties (Figure 7) and seismic inversion; it is clearly shown that those multiples have not only been successfully attenuated in the reservoir level, but also in the overburden. The impact of effective multiple attenuation helps in improving the seismic image of the deep Jurassic targets below the salt and anhydrites of the Gotnia, which leads to better fault imaging and fracture interpretation by means of fracture cluster tracking and azimuthal analysis of different attributes, ultimately obtaining a better understanding of the hydrocarbon reservoir. Acknowledgements The authors thank Kuwait Oil Company and the Kuwait Ministry of Oil for their kind permission to publish this paper. Thanks also to Zhiming (James) Wu, Sonika, Bill Dragoset, Fred Hugand, Bruce Hootman, and Paul Ras for their input and support as well as to the WesternGeco Kuwait data processing center for preprocessing the data. Special thanks extend to Mr. Wael Zahran; Senior Geophysicist in KOC for reviewing and commenting on the abstract. SEG San Antonio 2011 Annual Meeting 3547
Figure 3: Prestack time migration stack before (left) and after (middle) GSMP, and after internal demultiple (right). High velocity interbed multiple Slow velocity surface multiple Figure 4: Gather (from left to right): before GSMP, after GSMP, after interbed demultiple, and all multiples model. Figure 5: Semblances; before GSMP (left), after GSMP (middle), and after interbed demultiple (right). SEG San Antonio 2011 Annual Meeting 3548
Figure 6: Raw interbed multiple models: Rus (left) and Mutriba (right). Figure 7: Well tie before (left) and after (right) interbed multiple attenuation. SEG San Antonio 2011 Annual Meeting 3549
EDITED REFERENCES Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2011 SEG Technical Program Expanded Abstracts have been copy edited so that references provided with the online metadata for each paper will achieve a high degree of linking to cited sources that appear on the Web. REFERENCES El-Emam, A., M. Abdullatif, and H. Al-Qallaf, 2001, Multiple attenuation techniques, case studies from Kuwait: 71st Annual International Meeting, SEG, Expanded Abstracts, 1317 1320. El-Emam, A., I. Moore, and A. Shabrawi, 2005, Interbed multiple prediction and attenuation: case history from Kuwait: 75th Annual International Meeting, SEG, Expanded Abstracts, 448 451. Jakubowicz, H., 1998, Wave equation prediction and removal of interbed multiples: 68th Annual International Meeting, SEG, Expanded Abstracts, 1527 1530. Moore, I., and R. Bisley, 2006, Multiple attenuation in shallow-water situations: Presented at the 68th Annual International Conference and Exhibition, EAGE. Moore, I., and W. Dragoset, 2008, General surface multiple prediction (GSMP) a flexible 3D SRME algorithm: Presented at the 70th Annual International Conference and Exhibition, EAGE. SEG San Antonio 2011 Annual Meeting 3550