Full-waveform inversion in a VTI elastic Earth: a parameterization and crosstalk study

Five parameters are needed to describe VTI anisotropy in 2-D elastic media. With conventional streamer data and considering that we have a kinematically accurate NMO velocity (vnmo) and η (or horizontal velocity, vh) obtained from tomographic methods, we first show that a parameterization given by vh, η and allows us to focus the inversion to update vh (for traveltime and amplitude fitting) and (for amplitude fitting near zero offset). Having an inaccurate δ affects the inversion parameterized by the vertical velocity vv , δ, and more than the inversion parameterized by vh, η and . Then, we wonder what are the effects of not updating density and shear-wave velocity on the inversion results with the vh, η, and parameterization. For this analysis, using radiation patterns derived in the elastic case and a modified version of the Marmousi II model, we see that PP-waves are mostly helping to recover vh and , while PS-waves are mostly helping to recover η. For all scattering angles and wave modes, and ρ are strongly coupled. Keeping ρ and vs unchanged during the inversion is a valid strategy if their background models are good enough. For small vs and ρ errors (< 20%), the vs perturbation will map into η while ρ will map into . For large errors (≈ 100%), vs adversely affects the recovery of η and might leak into vh while ρ will preferentially leak into , then η and, in smaller proportions, into vh.