During the past decades, full waveform inversion (FWI) has federated a lot of innovative researches in the field of quantitative seismic imaging. Today, 3D FWI has been shown to be feasible to build high-resolution velocity models of the subsurface. These high-resolution velocity models can be used as improved background models for reverse-time migration and other migration techniques but also as quantitative information for structural interpretation of the subsurface. FWI as well as reverse-time migration have also been key motivating factors for innovations in the field of acquisition with the emergence of wide azimuth and long offset acquisitions and the recording of low frequencies.

Since 2006, the SEISCOPE consortium has been in the forefront of the research on FWI, during which we have developed new tools for seismic modeling for FWI both in the time and in the frequency domains, we have promoted wide-aperture acquisitions for FWI to fill the gap between the long-wavelength and short-wavelength reconstructions, we have shown the feasibility of 3D acoustic FWI, and we have pushed FWI beyond the acoustic approximation for the reconstruction of several classes of parameters such as shear wavespeed, attenuation and anisotropy parameters.

On the basis of the breakthroughs that were achieved during the 2006-2011 time slot and the numerous open problems which remain, we have relaunched a second round of the SEISCOPE consortium (referred to as SEISCOPE II) in 2013. For achieving this goal, we have built a research group where complementary skills in the fields of applied mathematics, theoretical and applied geophysics and computational geosciences are represented.

Our main scientific objectives is focused on multiparameter FWI in an attempt to evolve beyond the acoustic approximation by including anisotropy, attenuation and elastic features. Multi-parameter reconstructions are useful for high-quality migrations but may also be used for petrophysical inferences. We have gained some experiences during SEISCOPE in our ability to reconstruct several classes of parameters such as the shear wave velocity in land and marine environments, anisotropic parameters and attenuation. However, we need more elaborated optimization algorithms such as Gauss-Newton and full Newton methods as well as more robust regularizations and prior informations to be able to manage multiple classes of parameter with different footprints in the data. Sensitivity to the initial model is also a key point for which we work both on asymptotic method and misfit function design in FWI.

A research proposal for the ongoing round the SEISCOPE consortium (SEISCOPE II - 2019-2021) can be provided upon request to Romain Brossier (email: romain.brossier_at_univ-grenoble-alpes.fr).

Mis à jour le 5 March 2019