Elastic full-waveform inversion of seismic ambient noise

Abstract Elastic full-waveform inversion (FWI) is appealing for an enhanced integration of wave physics propagation and the resulting improved characterization of the subsurface due to the reconstruction of P- and S-wave velocity models. While the high computational cost of elastic FWI can be controlled using adequate numerical methods, the increase in the nonlinearity of elastic FWI calls for dedicated strategies to design initial P- and S-wave velocity models and access low-frequency data. We believe that the recent surge in ocean-bottom sensor acquisition is an opportunity to develop such strategies. The seismic ambient noise recorded by an array of receivers can be stacked to reconstruct interstation Green's functions that can be further used as low-frequency input data for elastic FWI to generate low-resolution P- and S-wave velocity models. We illustrate two applications of this strategy at two different scales: the first for near-surface imaging in a landslide context, the second for the reconstruction of a high-resolution S-wave velocity model at the Alps scale. In both cases, the seismic ambient noise Green's functions are efficiently matched, and accurate velocity models are built.