Inversion of 3D multi-component OBC data from the North Sea in the visco-elastic approximation

Postdoc offer

Frame

Full waveform inversion (FWI) is a high resolution seismic imaging technique based on an iterative data fitting procedure (Virieux & al. 2017). The match between observed data and synthetic data computed through the numerical solution of partial differential equations is iteratively improved following local optimization methods. Since 2010 and a spectacular application on 3D field data from the North sea (Sirgue & al. 2010), FWI has become a standard velocity building tool in the exploration industry. Most of the applications are performed however on marine streamer data in the acoustic approximation, with few exceptions of land data inversion in the elastic approximation.

Nonetheless, recent development in acquisition pushes forward the deployment of multi-component Ocean Bottom Stations (OBS) or Ocean Bottom Cables (OBC). Aside the possibility to record the particle velocities in the three direction of space (geophones) on top of the pressure variation (hydrophones), such acquisition devices make it possible to decouple the receiver and source systems, giving access to wider/larger azimuthal coverage for a given target. Such illumination is particularly helpful to obtain broad wavenumber reconstruction of the subsurface mechanical properties with FWI, as the intrinsic resolution attainable by FWI is directly linked to the illumination angle.

From such multi-component data, it is theoretically possible to reconstruct simultaneously P-wave and S-wave velocity models in the elastic approximation. Such characterization of the subsurface is much richer in information than a single P-wave velocity model, especially regarding the estimation of the presence of fluids. We want to assess in this project the feasibility of such an approach.

Project

For several years, we have developed in the SEISCOPE project a FORTRAN90 package named SEM46, dedicated to full waveform modeling and inversion in the 3D visco-elastic approximation. This package is based on a spectral element discretization of the elastodynamics equations. It has been developed and ported on many HPC architectures, showing excellent scalability properties for problems involving up to several tens of billions of degrees of freedom. It has been applied to the inversion of different land datasets, at different scales (near surface, exploration scale, deep crustal scale, Trinh & al. (2018) ; Irnaka & al. (2020) ; Lu & al. (2020)).

Recently, we have integrated a fluid/solid coupling modeling engine in SEM46 to make it possible to model the propagation of waves in a visco-elastic medium below a fluid layer (Cao & al., 2022). This update has been performed to make it possible to invert for 4C OBS/OBC data in the visco-elastic approximation with SEM46.

This project will thus be devoted to the application of SEM46 to a high quality North Sea 4C OBC data in the visco-elastic approximation we have been provided by one of our sponsor. We have already acquired a significant experience in the inversion of this data set in the 3D acoustic approximation (Kamath et al., 2021 ; Pladys et al., 2022 ; Provenzano et al., 2022). These acoustic inversion results will serve as a reference point for the visco-elastic inversion. The following questions will have to be tackled during the project

• pre-processing of the multi-component data (verification of the orientation, filtering, muting)
• initial design of the S-wave velocity model
• computational cost and stability of an inversion in the full model

In particular, a sequential strategy where the shallow part of the S-wave velocity model could be inverted first based on the inversion of the Sch\"olte waves (fluid/solid interface waves) might be considered.