Paper Summary
Successful application of shallow water full-waveform inversion (FWI) requires a good starting model, the right data, and an optimal algorithm. Our implementation is based on a pseudo-analytic (PA) method that solves the two-way acoustic wave equation in VTI media. The non-linear inversion uses a regularization operator that combines smoothing and wavenumber filtering in order to minimize the shallow water acquisition footprint. We show, using a 3D dual sensors streamer survey from the North Sea, that the inversion of refracted and diving waves yields a high-resolution velocity model for the shallow sediments. The small-scale velocity variations obtained by FWI in this area correlate very well with the geological features in the migrated images. The FWI model yields better migrated images than the starting model from reflection tomography. Examining both the flatness of common-image gathers and the match between modelled and recorded data corroborates the accuracy of the FWI velocity model.