Elastic full waveform inversion in anisotropic media: Methodology and Application to ocean bottom node data from the North Sea

Join Zoom Meeting: https://us02web.zoom.us/j/83804091653?pwd=WTdNS1RqY0hqNmpyRnRaMmFuZW1odz09 Meeting ID: 838 0409 1653 Passcode: 496025 Jury : Thomas BOHLEN (Geophysikalisches Institut, Karlsruhe, Germany): Referee Romain BROSSIER (Isterre, Grenoble, France): Referee Jean-Paul MONTAGNER (IPGP, Paris, France): Examiner Phuong-Thu TRINH (TOTAL, Pau, France): Examiner Paul WILLIAMSON (TOTAL, Houston, USA): Guest Satish SINGH (IPGP, Paris, France): Thesis director Mark NOBLE (MINESParisTech, Fontainebleau, France): Thesis co-advisor Full Waveform Inversion (FWI) has emerged as a powerful class of algorithms to retrieve quantitative and high resolution models of mechanical properties of the subsurface, such as density or (visco-)elastic parameters. However, FWI suffers from high computing costs and from the intrinsic limitations of the embedded inversion algorithm. Focusing on Earth local imaging in marine environments, these limitations led to the initial restriction of FWI to a fluid characterization of the Earth, where only the P-wave velocity was updated. Benefiting from increased computing resources, recent efforts produced a successful and much needed 2D inversion strategy for both P- and S-wave velocities in soft seabed environments. In order to meet modern applications needs, this thesis focuses on more contemporary recording devices, and partially relaxes the isotropic limitation. Over the years, ocean-bottom cables are indeed replaced by ocean-bottom seismometers/nodes (OBS/OBN). Without the use of the reciprocity principle, where the roles of the airgun sources and the OBS/OBN receivers are interchanged, the computing cost would be prohibitive. Consequently, I provide a guide of such a use of reciprocity in elastic anisotropic 2D/3D FWI. Furthermore, the available soft seabed strategy was developed in isotropic environments, while the real Earth is anisotropic. After the necessary adaptations, I assess the performance of the strategy for vertical velocity retrieval when slightly inaccurate remaining anisotropy parameters are used, those parameters not being updated. I focus on a specific kind of anisotropy called “vertical transverse isotropy”. I also propose a variant strategy for real applications that would not take attenuation into account for instance. Finally, I support my synthetic and theoretical findings with a real dataset from the North Sea.