Multiple-parameters inversion: Enhancing wave-matter interaction

Seismic waves honor wave propagation equation which is described by local model parameters. Full waveform inversion (FWI) attempts to distinguish automatically between reflection and transmission regimes through the implicit analysis of the different phases with variable signals expressed into seismic traces. Therefore, this approach, sensitive to phase and amplitude information, should require adequat description of different model parameters such as velocities, anisotropy coefficients and attenuation quality factors embedded into the spatial heterogeneous description of the model. Based on a single scattering approximation, FWI relies on the amplitude modulation in order to distinguish between model parameters at a point of the medium without considering important effects coming from spatial variations: only illumination (and curvature to lesser extent) is considered. Overcoming this limitation could be achieved by approximate scale separation specifying the wave-matter interaction often expressed through velocity/impedance parameterization or by preconditioning the model update through prior information. These additional strategies complement nicely the high-resolution performance of FWI without too drastic restriction in the model building. It does not overcome the intrinsic influence of large-amplitude phases compared to small-amplitude phases which is a characteristic feature of least-squares methods. Alternative strategies could be foreseen essentially based on stricter scale separation.