Global optimisation by simulated annealing for common reflection surface stacking and its application to low-fold marine data in southwest Japan

Abstract

The common reflection surface (CRS) stack is an alternative method of producing a zero-offset stacked section with a higher signal-to-noise ratio (SNR) than the conventional normal moveout (NMO)/dip moveout (DMO) stack method. Since, however, it is difficult to determine global optimal parameters for the CRS stack method by the conventional three-step search method, especially for complex structures and low-fold data, we investigate the ability of simulated annealing (SA) to optimise our estimation of these parameters. We show a detailed but practical procedure for the application of SA to the CRS stack method. We applied the CRS stack method with SA to numerically modelled seismic reflection data, and to multichannel marine seismic data over complicated geological structures around the Median Tectonic Line (MTL) in Japan. We used the results of the conventional three-step search algorithm as the initial model for the SA search and showed that with this approach SA can estimate CRS parameters accurately within a reasonable number of calculations. The CRS stack method with this approach provided a clearer seismic profile with a higher SNR than either a conventional NMO stack method or a conventional CRS stack method.