Time domain sparsity promoting LSRTM with surface-related multiples in shallow-water case

In the SRME relation, multiples are expressed as the multi- dimensional convolution between the the vertical derivative of the surface-free Green’s function and the down-going receiver wavefield. This relation leads to the methods that separate the surface-related multiples and primaries. Instead of imaging separated multiples trivially, we introduced the SRME relation into wave equation by areal source injection, which costs nothing extra to involve multiples into the forward wavefields. The related image contains the phantoms components from cross-correlation between different-orders of events. Especially in shallow water, due to the strong magnitudes of multiples and their overlap with primaries, the artifacts contain strong phantoms not only of ocean bottom but also subsurface layer interfaces. We use sparsity-promoting inversions where the nicely curved structure of the subsurface models are detected to help cleaning up the phantoms. We reduce the costs by combin- ing randomized source subsampling with linearized Bregman method and get cleaned image with one data pass.