73rd EAGE Conference & Exhibition incorporating SPE EUROPEC 2011 Vienna, Austria, 23-26 May 2011

A048How to Obtain a Seamless Dataset for a Pre-stack Multi-survey Merge Using 5D DataReconstruction?J.L. Rivault* (CGGVeritas) & A. Motagally (CGGVeritas)

SUMMARYPre-stack merging multi-survey datasets is often challenging in terms of processing effort as well as timeand human resources. It is problematic not only due to issues arising from survey matching (due todifferences in sources and receivers characteristics) but also from variations in sampling, which mayrelate to acquisition azimuth, bin size, shot/receiver line spacing. For such datasets, it is important toregularise the sampling of the data in order to harmonise fold and minimise migration artefacts. In thispaper we demonstrate the use of 5D data reconstruction for the benefit of multi-survey merging bymapping all the different surveys to a single acquisition design. The result is a dataset with a singleacquisition configuration which has a constant fold and a significantly reduced level of migration noise. Italso allows the preservation of azimuthal information by migrating in the COV domain. Its shows that 5Ddata reconstruction fully preserves the original seismic aspect allowing to easily migrate pre-stack multi-survey vintages in a single pass, which reduces the work of the geophysicist.

73rd EAGE Conference & Exhibition incorporating SPE EUROPEC 2011 Vienna, Austria, 23-26 May 2011

Introduction

Pre-stack merging multi-survey datasets is often challenging in terms of processing effort as well as time and human resources. From a processing point of view, it is problematic not only due to issues arising from survey matching (due to differences in sources and receivers characteristics) but also from variations in sampling. Sampling difference issues may relate to acquisition azimuth, bin size, shot/receiver line spacing, etc. These sampling differences require specific parameterisations at various stages of the processing (such ground-roll attenuation, demultiples, migration, etc.). From a production point of view, handling all these complex issues in a constrained time schedule can lead to a compromise on quality, or delay on delivery. For such datasets, it is important to regularise the sampling of the data in order to harmonise fold and minimise migration artefacts. With marine datasets, which are almost mono-azimuth, 3D regularisation techniques have been successful. For wide azimuth datasets, a key point is to preserve the azimuthal information up to the end of the processing. 3D regularisation can be applied to multi-azimuth data by working on azimuth sectors, but this has problems due to holes in coverage particularly at near and far offsets. The optimum solution is to process in Common Offset Vector (COV) domain, but with different acquisition designs it is not possible to simply process in COV in a seamless sense. In this paper we demonstrate the use of 5D data reconstruction for the benefit of multi-survey merging by defining the coordinates of a template dataset to which all surveys are mapped. The result is a dataset with a single acquisition configuration which has a constant fold and a significantly reduced level of migration noise. It also allows the preservation of azimuthal information by migrating in the COV domain. Additionally, managing a single survey instead of a multi one, reduces the work of the geophysicist.

Method

Pre-stack merging different datasets is often a challenge. In the past techniques such as DMO/Reverse DMO (AMO), trace rejection and 3D interpolation have been successful for standard marine acquisition, for which there is minimal azimuth dependency and trace duplication. By working in azimuth sectors, such processes have also been used for land data processing. However, this generally has only limited success due to highly irregular sampling in offset-azimuth volumes, e.g. large holes at near and far offsets. 5D interpolation can succeed where lower dimensional algorithms fail as it makes use of the continuity of data between offset and azimuth volumes to fully respect the wide azimuth nature of land data. This is achieved by decomposing the data in all recording dimensions simultaneously (Trad, 2009). By using a multi-dimensional Fourier transform that handles irregular data, it is possible to recreate theoretical acquisition datasets whilst preserving amplitude variations with offset and azimuth (AVO and AVAz). This project utilised an interpolation engine which decomposes the input data in the inline, crossline, offset, azimuth and frequency dimensions simultaneously. The algorithm performs a forward irregular Fourier transform to build a representation of the data in the frequency domain. The reverse transform is then used to generate data on targeted source and receiver coordinates. The forward Fourier transform was based on the anti-leakage Fourier transform (Xu et al, 2005) which minimizes spectral leakage and interpolates energy beyond aliasing (Poole, 2010).

Data example

The dataset was composed of 9 different acquisitions covering 7000 sqkm. 5000 sqkm were from modern acquisition (cross-spread, bin size 12.5 × 12.5m). The remaining 2000 sqkm were also cross-spread, but had different orientations, bin size, and receiver/shot line spacing. As the goal of the project was to generate a pre-stack time migration (PreSTM) on the new acquisition grid, entering the raw irregular data would lead to massive migration artefacts, generated by sub-optimal destructive interference of the migration operator. The proposed solution was to design a new survey (matching the new acquisition design but densified by a factor of 2 in terms of receiver and shot line spacing) for the whole area (Figure

73rd EAGE Conference & Exhibition incorporating SPE EUROPEC 2011 Vienna, Austria, 23-26 May 2011

1). The dense design was chosen based on the best parameters from each original survey as well as minimising the COV tile size. This provided a seamless acquisition design, with homogenous fold, allowing then to easily process the complete survey in the COV domain, without any acquisition holes or irregularities. As the different surveys all had cross-spread design, mapping them to a single one was fully justified in terms of interpolation reliability.

Figure 1: Survey layout where three surveys are overlapping

a) Surveys 2010, 2006, 2004, b) Shot and receiver coordinates for the original surveys, c) Shot and receiver coordinates after 5D data mapping

The first step was to create the ideal geometry which resulted in complete records in a surface consistent manner as could have been recorded on the field. The process consisted of creating blank traces with idealized coordinates, one shot record having constant shotpoint coordinates for all its traces, and specific receiver coordinates for each trace. This was achieved by using proprietary software to create an acquisition layout (creation of seismic processing support files (SPS) for recorder). Once the geometry was built, 5D interpolation was used to fill the target traces using 5D data reconstruction. The algorithm used spatial windowing in the inline, crossline, offset, and azimuth directions to build each output trace. The result was a new seismic dataset having offset and azimuth distribution equivalent to the main acquisition (Figure 2).

0 500 1000 1500 2000 2500 -2000 -1000 0 1000 2000 Offset (m)

Figure 2: Acquisition design a) Offset distribution (%) of output dataset, b) Azimuth distribution of output dataset

a) b)

a) b) c)

%

73rd EAGE Conference & Exhibition incorporating SPE EUROPEC 2011 Vienna, Austria, 23-26 May 2011

The resulting CMP gathers fully preserve the original seismic events, but with much finer sampling in offset and azimuth domain (Figure 3).

Figure 3: CMP gather trace distribution (offset and azimuth)

a) Input CMP gather with its native geometry b) Output CMP gather with the targeted geometry Figure 4 compares a stack section before and after 5D data reconstruction. The stack before interpolation (Figure 4, left) exhibits a very irregular fold resulting in poor signal continuity and variable noise level. After 5D data reconstruction (Figure 4, right), we see the stack has high constant fold which harmonises the amplitudes and enforces continuity along the line.

Figure 4: Stack comparison before and after 5D interpolation densified mapping

a) Original data binned on the final grid (irregular fold from 0 to 120 b) 5D interpolated densified stack (regular fold at 360)

Having the complete survey with the same cross-spread geometry allowed migration of the data in COV domain, preserving potential azimuthal dependency of the velocity, contrary to standard migration in offset domain, where azimuth information is lost. The COV migrated gather (Figure 5) illustrates azimuthal amplitude preservation when sorted in azimuth increasing within offset classes (known as snail gather). It gives the opportunity to estimate azimuthal residual move out to properly flatten the CIG.

a) b)

a) b)

Fold Fold

73rd EAGE Conference & Exhibition incorporating SPE EUROPEC 2011 Vienna, Austria, 23-26 May 2011

Figure 5: COV PreSTM CMP gather (top graph: offset, bottom graph: azimuth)

a) PreSTM CMP sorted in snail, b) CMP sorted in offset after azimuthal RMO

Conclusions

Data mapping using 5D interpolation is a powerful tool for pre-stack data merging and migration. By working on all spatial coordinates simultaneously it is possible to infill and merge the data with good amplitude preservation of the original seismic information as well as to obtain an improved stack image. Once the data has been merged onto a common grid it is possible to perform a single migration rather than migrating each survey separately. Preservation of the azimuthal velocity dependency during the interpolation process is critical for recovering correct geological information from the image. This methodology doesn’t avoid the necessity to properly match the different surveys (e.g. amplitude and phase matching), and also proper global statics and velocity solutions needs to be solved prior to run 5D interpolation. Also, it is significantly simplifying the project management, and so guarantee a proper result in a proper time frame.

Acknowledgements

CGGVeritas thank Repsol Exploracion Murzuq S.A. (REMSA) for their support and permission to show the examples. Thanks to Ahmed Motagally and his team, who worked very hard to provide on time suitable data and comprehensive support.

References

Poole, G., 2010, 5D data reconstruction using the anti-leakage Fourier transform: 2010 EAGE Barcelona Trad, D., 2009, Five-dimensional interpolation: Recovering from acquisition constraints. Geophysics, 74, V123. Trad, D., 2008, Merging surveys with multidimensional interpolation: Back to Exploration – 2008 CSPG CSEG CWLS Convention Xu, S., Zhang, Y., Pham, D., and Lambare, G. [2005] Anti-leakage Fourier transform for seismic data regularization. Geophysics, 70, 87-95.

CMP COV distribution a) b)