Building initial models for full-waveform inversion of shallow targets by surface waves dispersion curves clustering and data transform

Building accurate initial models for FWI is a very challenging task in near surface contexts. This is due to the complex behavior of the surface waves (SW) while interacting with the highly heterogeneous geophysical properties distribution. The geometrical dispersion of SW can be exploited to estimate the subsurface seismic velocity model. Commonly, SW dispersion curves (DCs) are used to retrieve the S-wave velocity (Vs) profiles, but recent studies (Socco et al., 2017; Socco and Comina, 2017) have proposed a method based on the concept of SW skin depth to estimate also P-wave velocity (Vp) profiles. An integrated DCs clustering and data-transform approach, based on SW data only, suggests the possibility of building accurate enough Vp and Vs initial models for FWI application. Here we perform a synthetic test where the estimated model is compared with the true synthetic model using a spectral element full-waveform modeling code. Results on a synthetic example with local high variations of heterogeneities motivate us to future more complex tests and real applications.