Elastic Anisotropic Full Waveform Inversion of Ocean Bottom Node Data: The Gorgon Case Study

This project aims at applying 3D seismic imaging by Full Waveform Inversion (FWI) on a marine industrial dataset collected by a dense array of multi-component Ocean Bottom Nodes (OBNs) for imaging the Gorgon petroleum field (Australia). While 3D FWI was mostly applied with simplified (acoustic) physics in exploration geophysics during last decade, the methodological and HPC challenges taken by this project are to account for elastic anisotropic physics to reconstruct multiple classes of parameters (anisotropic P and S wavespeeds, density) from the four components of the stations (hydrophone, three geophones). FWI is a nonlinear data fitting problem for parameter estimation tackled with iterative local optimization methods. The main computational burden is the simulation of wave propagation for multiple sources at each FWI iteration. The modeling engine relies on the spectral element method on unstructured hexahedral meshes to benefit from h adapativity and mesh complex bathymetry accurately. The FWI code combines two levels of MPI parallelism by source distribution and domain decomposition.  The researchers aim at designing a FWI workflow that allows us to push the inversion at the highest possible frequency to reconstruct a representative target of the Gorgon field. The demand is 570 000 node hours on MareNostrum5 GPP.