Elastic vertically transversely isotropic full-waveform inversion: From synthetic to field data

Abstract

Representing the earth as an elastic medium leads to a more accurate approximation of the seismic response than can be obtained by assuming an acoustic medium. However, this representation requires a larger number of unknown parameters to be introduced and a more complex seismogram to be considered. Consequently, it is important to carefully deal with two challenging issues, the impact of strong surface waves and the crosstalk among parameters, for land seismic multiparameter full-waveform inversion (FWI). Under the elastic vertically transversely isotropic assumption, a robust FWI approach is developed by combining the optimal transport (OT) misfit function, the Gaussian time window, and the suitable parameterization ( v n, η, δ, v s, ρ). Without specific preprocessing, surface waves usually dominate the inversion process, leaving the body waves not fully exploited. Correspondingly, the subsurface model is only updated at a shallow depth. The OT misfit function can balance the contributions of the surface and body waves to model updates in an automatic manner. In addition, the Gaussian time window further improves the weights of the body waves during the inversion. We efficiently exploit the entire seismogram with these two approaches. In this configuration, the performances of several parameterizations chosen from a radiation pattern analysis are evaluated. The proposed inversion strategy is applied to the Middle East model and a field data application. In both cases, it outperforms other inversion strategies. The entire land seismic data are efficiently exploited and the multiparameters are simultaneously reconstructed.