Analyzing Converted-Wave Seismic Data: Statics, Interpolation, Imaging, and P-P Correlation
Converted-wave (P-S) seismic processing requires a specialized approach due to its mixed wavetype. To facilitate this approach, a set of algorithms has been developed starting from data preparation and ending with the correlation techniques for P-P and P-S time sections.
A new statics estimation algorithm is developed to estimate S-wave receiver statics. This algorithm operates in the common-offset domain and relies on f-x prediction to generate pilot traces. Good results using the f-x statics estimation on the synthetic Marmousi example with random statics suggests that this technique is very useful especially in complex structural regimes. This is especially true where hyperbolic velocity assumption is violated or velocity estimation before statics is not possible. These features of the new technique are important for the converted-wave data and are demonstrated with Blackfoot data.
After statics corrections are applied, I use f-x prediction filters to interpolate missing traces. At this stage, the prestack data are assumed to be evenly sampled and are ready for further multi-channel processing.
Some fundamental equations for converted-wave traveltimes are derived and implemented in a velocity analysis method based on P-S prestack migration. This analysis uses the previously estimated P-wave background velocity to create the pseudo S-wave velocity field.
Based on both velocity fields I use a new Asymmetric MoveOut Correction (AMOC) technique to transform the P-S data into pseudo P-P reflection data. With this approach, Swave receiver statics can be estimated and a non-iterative velocity processing flow can be developed.
The processing is finalized with a rigorous analysis of the correlation between P-P and the corresponding P-S time sections. Two analysis techniques are developed for this purpose. The first technique is based on non-linear optimization of the correlation function, while the other is an approach to match data in the logarithmic time domain. Both correlation techniques produce a good visual match on both synthetic data and the Blackfoot data.