71st EAGE Conference & Exhibition — Amsterdam, The Netherlands, 8 - 11 June 2009

Introduction

Seismic depth processing is the decisive step to reconstruct the structural geometry in the subsurface.Depending on the desired accuracy, both, the depth model building and the depth imaging can be verytime-consuming and costly steps. In order to increase the depth resolution and signal qualityespecially in data of varying fold or quality, the Common-Reflection-Surface (CRS) technique can beintegrated at all stages of the general depth imaging procedure.

In this case study, the depth processing is based on the initial CRS time processing of 3D seismic landdata from Mexico. At the surface, acquisition of these data had to deal with several inhabited areas,that caused strong variations of fold and data quality. In the subsurface, the Tertiary and Mesozoicsediments are disturbed by strong salt tectonics in parts of the survey. The low fold areas, and thecomplicated subsurface represented the main challenges for depth processing.

The CRS time processing provides high-resolution volumes of both, the CRS image, and the CRSstacking parameter or attributes, for subsequent depth processing. Due to an increased signal-to-noiseratio, the poststack time migration (PostSTM) of the CRS stack clearly outlines the salt areas. This isshown at time slices in Figure 1 in a comparison to prestack time migration.

The high information contents of the densely sampled CRS attributes may well be used for a fastinitial reconstruction of the velocity-depth model by CRS tomography. Especially with respect to thevarying data quality and the complicated subsurface, this approach has significant advantages incomparison to another fast method of initial model building, given by the Dix inversion of stackingvelocities. With the CRS tomography model, a first straightforward depth imaging can be performedby poststack depth migration (PostSDM) of the CRS stack. The result exhibits an excellent structuralresolution in comparison to conventional prestack time migration (PreSTM) . This first cycle of depthprocessing thus provides a fast and effective option to obtain preliminary depth sections that improveresolution beyond time processing.

Further depth processing using prestack depth migration (PreSDM), however, is required especially inthe region of salt intrusions. The salt bodies can be initially defined in the high-resolution CRSPostSDM, and inserted into the CRS tomography model. The subsequent iterative validation byPreSDM, and model updating benefit from the accuracy of the starting model which cuts down thenumber of iteration cycles. In addition, the quality of the PreSDM images can be strongly increasedby a CRS-based noise suppression and regularisation in the prestack data. This again facilitates themodel building and salt body definition.

CRS method

In the initial time processing, the CRS or Common-Reflection-Surface method was applied in order tocompensate for of the imaging problems in the low-fold areas. The CRS method was developedwithin the concept of macro-model independent imaging (e.g. Gelchinsky 1988, Jäger et al. 2001). Inthis concept, the imaging parameters are automatically obtained from local measurements in theseismic data.

The CRS method assigns the reflection events to local reflector elements with dip and curvature in thesubsurface. The CRS stacking parameters, the so-called CRS-attributes accordingly comprise thewavefield dip together with wavefront curvatures observed at the surface. Corresponding to theelaborate parameterisation of the subsurface reflection, the stacking contributions can be traced in theprestack data across several CMP locations. CRS stacking thus uses a much larger fold thanconventional NMO stacking, and obtains a much better signal-to-noise ratio, and event continuity.

71st EAGE Conference & Exhibition — Amsterdam, The Netherlands, 8 - 11 June 2009

Figure 1: Prestack time migration (top) versus Poststack time migration of the CRS stack (bottom).Note the impoved outline of the salt area in the CRS result.

Initial depth model building

For building the initial depth model, the CRS tomographic inversion of CRS wavefield dip andcurvature attributes was compared to conventional Dix inversion of stacking velocities. Dix inversionof stacking velocities is the most common and robust method that basically assumes a flat subsurface.The main charm of the Dix inversion is that it can be directly appended to any normal timeprocessing. However, with increasing dip and lateral variation of the subsurface structures, the Dixmodel strongly deviates from the exact velocity distribution in depth. In this case study, the deviationof the Dix model from the depth structure in an associated PostSDM volume was obvious even inregions of good data quality (Figure 2 left).

CRS tomography uses CRS attributes in order to calculate an interval velocity model in depth bymeans of grid tomographic inversion (Duveneck 2004). The required kinematic CRS attributescomprise the radius of the so-called NIP wave and the emergence angle α of the wave at the surface.The superior performance with respect to Dix inversion of stacking velocities mainly results from theincorporation of structural dip through the emergence angle α. For this case study, the correspondenceof the CRS tomography model to an associated PostSDM volume is much better than in the Dix case(Figure 2).

71st EAGE Conference & Exhibition — Amsterdam, The Netherlands, 8 - 11 June 2009

Figure 2: Dix model (left) versus CRS tomography model (right) with CRS-PostSDM sections

Depth migration and model update

The CRS tomography model is a smooth structural representation of the velocity distribution in depthwhich is especially well suited for depth migration. It was first used in PostSDM for transferring thehigh signal-to-noise ratio of the CRS stack to depth, allowing an initial definition of the salt body.Using this advanced initial model as a starting point for further depth processing, significantly cutdown the number of prestack depth migration (PreSDM) and model updating cycles.

Since the CRS attributes provide a very detailed local description of the seismic data, they may alsobe used to perform a local regularization and noise suppression in the prestack data by using thepartial CRS stacking operator. A CRS-based PreSDM strategy can include a regularisation of both,the offset and the azimuth distribution. A demonstration of the noise suppression by this strategy hasbeen given by Eisenberg-Klein et al. (2008). In this case study, a CRS-based noise suppression andregularisation was applied to the prestack input for PreSDM.

Figure 3 compares the result of a conventional model building and Kirchhoff PreSDM strategy to theCRS-based approach. The noise reduction by the partial CRS stacking is obvious, and additionalstructural features are revealed by a general increase of resolution. The definition of the salt region isespecially improved in the CRS-based imaging. As a consequence, this type of CRS-based PreSDMmay be helpful in situations where a strong noise level prohibits imaging by conventional PreSDM.

Conclusions

CRS time processing provides detailed local information on the seismic reflection events in the formof kinematic wavefield attributes. These so-called CRS attributes can be inverted into a depth modelby CRS tomography. Through the incorporation of the structural dip and the curvature of the NIP-wavefront this inversion is better constrained than vertical Dix inversion, providing a goodcorrespondence of velocity variation and structure. The CRS tomography model can be used as agood initial approximation of the subsurface velocity for both, poststack depth migration (PostSDM)and further model update in prestack depth migration (PreSDM). Significant CRS-based

71st EAGE Conference & Exhibition — Amsterdam, The Netherlands, 8 - 11 June 2009

improvements in depth imaging are achieved by using CRS attributes for data regularization and noisesuppression in PreSDM.

Acknowledgements

We thank PEMEX for the permission to present their data.

References

Duveneck, E., 2004, Tomographic determination of seismic velocity models with kinematicwavefield attributes: Phd thesis, University of Karlsruhe, Logos Verlag Berlin.

Eisenberg-Klein, G., J. Pruessmann, G. Gierse, and H. Trappe, 2008, Noise reduction in 2D and 3Dseismic imaging by the CRS method: The Leading Edge, 27, 258-265.

Gelchinsky, B., 1988, The common reflecting element (CRE) method (non-uniform asymmetricmultifold system): ASEG/SEG International Geophysical Conference, Exploration Geophysics,Extended Abstracts, 71-75.

Jaeger, R., Mann, J., Hoecht, G., Hubral, P., 2001: Common-reflection-surface stack: Image andattributes. Geophysics 66 (1), 97-109.

Figure 1:PreSDM from conventional and CRS-based model-building and imaging approach