Part I — CRS stack: Global optimization of the 2D CRS-attributes

Abstract

Abstract The common-reflection-surface (CRS) stacking method simulates zero-offset (ZO) seismic traces by means of summing the amplitudes of seismic reflection events of multi-coverage data by using a stacking operator defined for the midpoint and half-offset coordinates. For two-dimensional media, this operator depends on three kinematic attributes, also referred to as CRS attributes. The main problem in using the CRS stacking method is the determination of the three CRS attributes from multi-coverage seismic data for each sample point in the simulated ZO section. Usually, the CRS attributes are determined in several steps, through searching individual attributes on the prestack seismic gathers and on post-stack sections. This multi-step solution may accumulate errors or inaccuracies in the CRS attributes, which can compromise the quality of the ZO stacked section and affects their successful application to other processes. In this study, the optimization problem of the CRS method is solved by means of multidimensional global optimization, which uses an objective function based on a coherence measure (semblance) of the traces in the multi-coverage data. The simultaneous determination of the three CRS attributes is performed by applying the Simulated Annealing (SA), Modified Simulated Annealing (MSA) and Very Fast Simulated Annealing (VFSA) global optimization algorithms. These algorithms were compared in terms of their effectiveness, efficiency and robustness. The results show that the VFSA and MSA algorithms are more suitable for determining the CRS attributes and generate high signal-to-noise ratio ZO stacked sections. Both algorithms are highly effective, but VFSA is more efficient than MSA. The classical SA algorithm has a low efficiency and effectiveness for determining the CRS attributes. In our proposed single-step global optimization strategy to search CRS attributes, we use the stacking velocity model derived from conventional velocity analysis to determine the lower and upper limits of the optimization search space for RNIP‐ We use real data with low fold and high noise level to illustrate the proposed optimization strategy and show that its result is of higher quality than that obtained with CRS stack based on the well-known extended-pragmatic search strategy.