3D reverse time migration using a wavefield domain dynamic approach

Abstract

3D Reverse-time migration (RTM) is a powerful technique for imaging complex geologic structures. This approach requires a significant computational effort, demanding a high amount of memory for storing the source's wavefields, consequently leading to a high cost to perform the imaging condition. Thus, this work aims to reduce these problems by introducing a dynamic approach (DA) that considers the sparsity of the wavefield in the first periods of propagation. The RTM combined with the DA (RTM-DA) approximates the computational domain to the propagation domain, which is the region delimited by the wavefront. In practical terms, the computational domain expands together with the wavefront, reflecting in a very significant economy of memory and reduction of processing time when compared to the conventional RTM, which we denominate as a static approach (RTM-SA). To reduce the sparsity of the wavefields in the first periods of propagation, we have built an empirical 3D filter that maps each timestep of the wavefield and gives the coordinates to approximate the computational domain to the propagation domain. We compare both approaches using the 3D SEG/EAGE Salt model and demonstrate that the RTM-DA is more efficient than the RTM-SA in terms of memory consumption and computational time, preserving the quality of the seismic image.