Efficient Acoustic Reverse Time Migration With an Attenuated and Reversible Random Boundary

Pre-stack reverse time migration (RTM) based on the two-way wave equation has been proved to be the most accurate seismic migration method theoretically. However, it requires reverse-order access to the wavefield calculated in forward time. In recursion computing, such out-of-order access requires that most of the recursion history should be stored on the hard disk. For massive amounts of seismic data, loading the saved wavefield data from the disk during imaging has been the bottleneck of RTM, restricting its wide application. To solve this problem, the wavefield in forward time must be reconstructed in reverse order. Although the random boundary can avoid the disk requirement by creating random velocity around the computational domain when propagate the source function. However, the random wavefield reflected from the boundary can generate unwanted artifacts in the final images. In this paper, we develop an attenuated and reversible random boundary condition which is implemented by mixing the reversible attenuation and random boundary conditions. Similar to the random boundary scheme, the proposed method just needs to save the last one or two wavefield snapshots into the memory in forward process. It then reconstructs the source wavefield in reverse order, while greatly reduces the disk input and output (I/O) requirements. Taking the attenuated property into consideration, the artificial events reflected from the boundary can be eliminated. Thus, our method can improve the imaging quality largely compared with the random boundary scheme. Numerical results demonstrate that the RTM images with our proposed attenuated and reversible random boundary condition can not only eliminate the unwanted artifacts, but also improve the computational efficiency greatly.