ELASTIC REVERSE TIME MIGRATION BASED ON WAVEFIELD SEPARATION

Abstract

Compared with other imaging algorithms (e.g., ray-based, one-way wave equation), reverse time migration (RTM) based on the two-way wave equation exhibits greater superiority, especially in handling steeply dipping structures. However, imaging with conventional single-component seismic data is unsuited for some complicated structures (e.g., gas clouds). Elastic RTM, which is based on the elastodynamic equation and uses multi-component seismic data to extract PP and PS reflectivity and subsurface information, can more consistently reproduce the characteristics of elastic wave propagation in real Earth media, resulting in seismic images that more accurately characterize the subsurface. To begin with, we exploit the first order stress-velocity equations to extrapolate the elastic vector wavefield, then the P- and S-wavefields are separated by computing the divergence and curl operator of the extrapolated particle-velocity wavefield. Then, imaging profiles with pure wave modes are computed by applying the source normalized cross-correlation imaging condition, thus avoiding crosstalk between unseparated wave modes. To address the polarity reversal problem of the converted image, we propose an alternative method in the common-shot domain. We also develop an efficient method that reconstructs the source wavefield in the reverse time direction to save storage in the GPU and to avoid large input/output in the elastic reverse time migration. During the forward modeling, the method only saves the particle-velocity wavefield of all time intervals within an efficient absorbing boundary and the total wavefields in the final time interval. When we extrapolate the receiver wavefield in the reverse time direction, we simultaneously reconstruct the total source wavefields via the saved wavefields. Numerical examples performed with the graben and Marmousi2 models have shown that the polarity reversal correction method works, and elastic reverse time migration can accurately characterize complicated structures.