|3D OBC wave-equation multiple modeling: a case study on the Schiehallion converted-wave dataset|
This paper presents the application of 3D wave-equation multiple modeling (WEMM), which is used to remove source-side P-wave multiples from converted-wave data.
The process was successfully tested on the Schiehallion UKCS ocean-bottom cable (OBC) dataset that suffered from a severe multiple problem due to a dipping seabed and high-reflectivity, variable-thickness Balder coal.
WEMM using synthetic interpreted horizons to represent the major lower multiple generators produces improved results compared to WEMM using a full reflectivity from an image stack. It also demonstrates that in areas of complex strong multiple generators WEMM is more effective than conventional 3D tau-p deconvolution.
Among the reasons to acquire multicomponent data, the combination of the hydrophone and vertical geophone is probably one of the most attractive because it allows for effective demultiple. Many examples have been published regarding P-wave multiple attenuation (Amundsen, 2001; Osen and Amundsen, 2001) from P-wave data; however, there is a lack of published case histories for the application on converted-wave (PS) processing.
A recent increase in multicomponent acquisition activity has broadened the diversity of geological scenarios. Hence, there is a pressing need for algorithms capable of handling 3D multiples. Case studies on P-wave datasets have been published (Pica, et al.,2005; Xia, et al., 2006). However, PS multiple removal techniques are mainly limited to conventional methods having strong limitations: they assume 1D, or near 1D geology and they are only effective in a simple geological context combined with a relatively flat and smooth seabed.
The technique presented in this paper is an adaptation of WEMM developed by Stork et al. (2006) to OBC data. Application is shown to a converted-wave dataset acquired over the Schiehallion oilfield.
In this paper, converted-wave data refers to a P (compressional) wavefront propagating from the source downward to the reflecting interface, where it is converted to an S (shear) wavefront propagating upward to recording geophones located on the seafloor. Each receiver is a group of three orthogonal geophones and a hydrophone (P) sampling the 3D wavefield. The geophone records processed by 3D rotation produce the vertical (Z) and horizontal datasets: radial (particle motion in source-receiver direction) and transverse (particle motion perpendicular to source-receiver direction). As a convention, we will refer to radial data when referring to PS or converted-wave data and we will refer to the combination of P and Z records as PZ data.
Shear waves do not pass through the water layer without converting back to P-waves. This conversion is, in most cases, weak and will be primarily recorded on the vertical, not the horizontal, geophones. Therefore, the principal multiples in PS data are source-side surface multiples. These are P-wave multiples and, therefore, can be modeled using a P-wave reflectivity model. The location of the sources at the sea surface allows generation of the multiples from a full sea level image file.
The WEMM process can model multiples using two methods (Stork, et al., 2006). The first is to propagate an impulsive source wavefield through the earth model and then upon reaching the surface, take it back through the subsurface so as to simulate surface multiples to a given order. The second method is to begin with a shot or receiver record and propagate the data through the earth model only once. Both of these methods have been used for conventional surface P-wave data. In this paper, we apply an adaptation of the second method to converted-wave data.
The inputs to WEMM are:
1. 3D PS receiver gathers
2. Reflectivity model – image file (depth migration) or synthetic model
3. Depth-interval velocity model.
The reflectivity model can be obtained from a streamer depth image or synthetic horizons such as a seabed profile (for water layer multiples only). Because they are OBC data, it is likely that the PZ data will be processed at the same time. Depth-migrated PZ data can be used for the image file if there is adequate water depth/spatial sampling to fully image the water bottom and near-surface multiple generators. Note that no pre-processing of the PS data is required to obtain the image file, as the image file is P-wave only.
The WEMM was tested with data from the BP Schiehallion 3D OBC survey acquired in 2006 in a crossline shooting configuration (Figure 1). The aim of the converted-wave processing was to determine if a reliable 3D image could be obtained for the reservoir located at 3.4 s converted-wave two-way time, and if additional reservoir characterization could be delivered from the shear-wave data.
The seismic wavefield was recorded by two patches of six cables laid on the seafloor with 350 m spacing, each cable containing 240 receiver groups. The nominal detector group spacing was 25m, as was the shot spacing. Shot line spacing was 250 m. As stated previously, our focus was the converted-wave data only.
The main problem affecting this dataset is the presence of strong multiples that interfere severely with the primary data. These multiples have proved difficult to remove with previous techniques, due to the dipping seafloor and a rugose high-reflectivity Balder coal layer above the main target zone (Figure 2).
As described previously, the WEMM process was run on 3D receiver gather. Two types of reflectivity models were tested: the seabed synthetic and the full towed-streamer depth-migrated image. Both approaches modeled multiples up to 30 Hz, which was sufficient to cover the multiple frequency content. The modeled multiples were then adaptively subtracted from the input data and the output data were depth-migrated to assess the benefits.
Figures 3, 4, 5, 6, and 7 show the same converted-wave stacked section through the survey, parallel to a receiver cable, with different demultiple applied. The benefits of the WEMM process can be further assessed by examining autocorrelations at the bottom of each stacked section.
Figure 3 shows the reference stacked section where no demultiple processes were applied. Seabed-generated multiples dominate the stacked section down to the target area as indicated by the arrows.
Figure 4 and 5 show the results of applying the WEMM processing using the image stack and the synthetic seabed respectively. In Figure 5, only seafloor source-side-related multiple are modeled and subtracted. The comparison between Figures 3, 4 and 5 highlights the benefits of the WEMM: top Balder water-bottom first- and second-order multiples are attenuated.
Further testing of the reflectivity input to WEMM led us to compare the use of a synthetic seabed with the full depth-migrated stack. Considering the seafloor to be the main multiple generator, it was clear that the image stack seabed quality is a key factor for WEMM success. In our data example, the seabed was not well imaged. Despite the fact that the image stack contained the whole reflectivity series, compared with the synthetic seabed approach, the WEMM demultiple results using the image stack were inferior, in general, to the results obtained using the synthetic seabed (Figures 4 and 5).
Following this latest approach, the second main multiple generator after the seabed was added to the process. WEMM modeled the P-wave source-side multiple generated by the top Balder reflector. The new multiple model was adaptively subtracted once P-wave seabed multiples had been removed from the data. The new results showed clear improvements compared to the single synthetic seabed horizon approach (e.g. left side of figure 5 and 6).
Finally, the best WEMM results were compared with a more conventional demultiple approach.
3D tau-p predictive deconvolution processing was used as a benchmark to assess the WEMM results. 3D tau-p demultiple has shown good results on many surveys, mostly in the North Sea where the geology and seabed vary smoothly. However, the 3D tau-p transform has limitations related to dipping seabed and complex geology.
Both of these limitations are observed on the Schiehallion dataset where we see dipping seabed and strong variable multiple generators (Figures 1 and 2). For cross-spread geometry, 3D tau-p is commonly applied in the cross-spread domain due to spatial sampling issues in the receiver domain. In this configuration, with converted-wave data, receiver-to-receiver variations due to statics, coupling and noise can cause further problems with the use of 3D tau-p deconvolution.
The 3D tau-p depth-migrated stack (Figure 8) shows a high level of residual multiple compared with the WEMM stack section (Figure 6). Top Balder water-bottom multiples are still very strong.
Despite encouraging results, the WEMM technique has some limitations.
1. The acquisition geometry can be an issue if the source locations are too sparse.
2. The availability of a depth-migrated stack can also be problematic. However OBC data are frequently acquired in locations where a conventional marine survey was performed.
3. The quality of the depth-migrated stack is important for the success of the technique; however, using the main multiple generators as synthetic horizons is a very powerful alternative.
This case study presents the methodology and application of wave-equation 3D modeling of source-side P-wave multiple from a converted-wave dataset. The WEMM process has shown that, in more geologically complex areas, it is capable of attenuating more multiple energy than the conventional 3D tau-p deconvolution approach. It was also illustrated that the reflectivity model is a key factor for the success of WEMM: In our example, main multiple generators supplied as synthetic horizons produced better results than the full reflectivity stack.
The authors thank the Schiehallion partners for permission to present this dataset and the management from BP and WesternGeco for allowing publication of this work. Also thanks to Steve Campbell, Ian Moore, and Neal Zimmermann for their contributions.
Amundsen, L., 2001, Elimination of free-surface related multiples without need of the source wavelet: Geophysics, 66 , No. 1, 327-341
Osen, A., and L. Amundsen, 2001, Multidimensional multiple attenuation of OBS data: 71st Annual International Meeting, SEG, Expanded Abstracts, 817-820.
Pica, A., G. Poulain, B. David, M. Magesan, S. Baldock, T. Weisser, P. Hugonnet, and Ph. Herrmann, 2005, 3D surface?related multiple modeling, principles and results: 75th Annual International Meeting, SEG, Expanded Abstracts, 2080-2083.
Stork, C., J. Kapoor, W. Zhao, W. Dragoset, and K. Dingwall, 2006, Predicting and removing complex 3D surface multiples with WEM modeling – an alternative to 3D SRME for wide azimuth surveys?: 76th Annual International Meeting, SEG, Expanded Abstracts, 2679-2683.
Xia G., J. Etgen, and K. Matson, 2006, Multiple attenuation on OBS/OBC data by extended wavefield extrapolation: 68th Annual International Meeting, EAGE, Extended Abstracts.