Abstract EN:

In this thesis, we review the Receiver Function (RF) method, conceived many years ago in earthquake seismology, to see whether it can also be used in reflection seismology, and more specifically to see whether it can be applied to ocean-bottom-cable (OBC) data. The conventional RF method is used to determine the local PS wave response of a target zone below a multi-component 3C receiver and has been originally designed to process individual earthquake data of constant ray parameter p, acquired at the free surface. The target is illuminated from below. The converted PS wavefield generated at the receiver side is isolated from the global earth response by deconvolving the horizontal Ux component with the vertical Uz component, where Uz is assumed to contain only the impinging P waves (i. E. The unpredictable effective source function). Simultaneously to the source signature removal, the effect of the deconvolution can be subdivided into two steps: (1) P-PS wave separation and (2) multiples removal. The keyword is combination between components (by adaptive subtraction) or equivalently ratio between components. Our motivation is to reproduce these two steps with OBC data, in order to determine the (separated) primary PP and PS responses generated in the sub-seafloor area. However, there are several issues that require special attention when we implement the RF technique to OBC acquisition geometry. Firstly, the target (i. E. The sub-seafloor) is illuminated from above. Therefore there exist three types of incident waves at the receiver level: the upgoing P and PS wavefields (Pup and Sup as for land data) but also the additional downgoing P wavefield (Pdown). Secondly these wavefields are mixed between the components with time varying ray parameters, which precludes the possibility of applying the RF approach in the conventional time-offset domain. These problems can be addressed by taking into account the additional measurement of the pressure wavefield by the hydrophone Uh and by transforming the data in the ¿ -p domain (requiring fine receiver spacing usually afforded by OBCs). This transformation reorders the data by incidence angle at the receiver level, such that the pure upgoing PP and PS wavefields can be separated, based on polarization angle discrimination. This first step partially addresses the problem of multiples in the data by removing the downgoing (receiver side) water multiples, but it requires the knowledge of the seafloor properties as inputs. The other advantage of the ¿ ¡ p domain is that the water multiple reverberation becomes periodic. Remaining pure upgoing (source-side) water multiples are fully predictable (in contrast to overlapping source-side and receiver-side multiples) and can therefore be removed using predictive deconvolution (this is the required second step). Our adapted version of the RF technique uses the various ratios between components to estimate the elastic properties at the seafloor, as well as calibration operators, required for the decomposition. Our data-driven method can be automatically applied with a minimum of user-defined inputs, by taking advantages of the coherency between adjacent p traces and the redundancy of informations within the multiples. The strategy has been successfully applied to field data. Our results suggest several avenues for further processing