Model-based signal processing and elastic wave propagation for deterministic deduction and statistical inference in exploration seismology

VANDERBAAN devant le jury composé de : Michel CAMPILLO .............. Examinateur Chris CHAPMAN ................ Rapporteur J-Michael KENDALL ............ Rapporteur RJean Paul MONTAGNER ......... Examinateur Nikolai SHAPIRO .............. Rapporteur Satish SINGH ................. Directeur de thèse Abstract: An important aspect of seismology is the study of elastic wave propagation in the Earth. Exploration seismology employs these elastic waves to detect the Earth's inner resources in a mostly non-destructive way. Successful exploration demands for the application of a large variety of methods based predominantly on deterministic deduction and statistical inference. Deterministic deduction uses the classical laws of physics to derive particular conclusions. Statistical inference applies the laws of probability to reach the most likely solution given quantitative data. Deterministic deduction and statistical inference are thus complimentary cornerstones of exploration seismology and vital in any interpretation. Successful interpretation requires however first of all that the recorded seismic data are of sufficiently high quality and resolution. Signal processing is therefore another cornerstone of exploration seismology. It ensures that signal can be separated from noise. This is most commonly achieved by assuming a specific mathematical model for the signal, and by transforming the data to a suitable domain. Applications used range from wavelet estimation and deconvolution via PP/PS-wave separation to extraction of reflected signals. Statistical inference and deterministic deduction become feasible once the data are of sufficiently high quality. Yet the power of deduction varies greatly with the capacity and ingenuity of the person attempting the interpretation. The physical laws of elastic wave propagation can reveal the Earth's resources if properly applied, yet often lead to overly complex mathematics if implemented in their most general forms. The key to successful interpretation often lies in simplifying the picture such that the basic elements appear without sacrificing vital detail. The Earth's properties can be studied by analysing the phenomena of multiple-wave scattering and anisotropic wave motion. Described applications range from inferences on intrinsic versus apparent attenuation in the Earth, optimal source frequencies in subbasalt imaging, anisotropy-parameter estimation, common-conversion-point sorting, amplitude analysis and traveltime computations.