Imagerie sismique des milieux visco-acoustiques et visco-élastiques à deux dimensions par stéréotomographie et inversion des formes d'ondes: applications au champ pétrolier de Valhall

摘要

Geophysics is a very insightful field to gain inferences on the internal structure of the Earth at different scales, and on the past events related to its slow evolution. It also constitutes a major issue in our society, as the resources are becoming increasingly rare, and because it is of great consideration for town and country planning. Full waveform inversion is an optimization process differs from other seismic imaging approaches by its ability to extract the full information content of the seismic data to construct a high-resolution quantitative imaging of one or more physical parameters. I discuss in my thesis several issues related to this method and I illustrate them with applications to multi-component Ocean Bottom Cable (OBC) data recorded at the Valhall oilfield in North Sea. I first discuss the footprint of anisotropy in vertical transverse isotropic media on the seismic Valhall dataset by comparing the subsurface models built by full waveform inversion when anisotropy is taken into account or not in the seismic modeling. I show some biases in the velocity reconstruction when the imaging is performed in the isotropic approximation, these biases leading to mispositioning of the reflectors at depth. %The meaning of the velocities is different in subsurface and in depth, where they are representative of horizontal and NMO velocities respectively. I then address the problem of constructing an initial velocity model for full waveform inversion from wide-aperture data. The selected approach is an extension of stereotomography, a slope tomographic method, where the traveltimes and the slopes of the reflected and refracted arrivals are sequentially introduced in the inversion process through a multiscale approach. The potential of the method is discussed based on the synthetic Valhall case study before the application to the real Valhall data-set. The final chapters are devoted to the reconstruction of several classes of parameters within the visco-acoustic and visco-elastic approximations. In order to reduce the nonlinearity of the inverse problem, I propose a methodology based on the hierarchical reconstruction of the parameter classes, and on the progressive introduction of the different data components in the inversion process. I first test different inversion strategies on the synthetic Valhall case to build the compressive velocity, the density and the attenuation parameters, before their application to the real data within the visco-acoustic approximation. I reconstruct in a second step the shear velocity from the three components of the sensor (one hydrophone and a vertical and horizontal geophones). Quality of the results is assessed with various tools, and a geological interpretation of the results is proposed.