Surface Wave Analysis in Laterally Varying Media

This work studies the possibility of using surface wave analysis as a tool for a robust estimation of the S-wave velocity behaviour in laterally varying media. The surface wave method, in fact, can be effectively adopted for different purposes and at different scales, but I focused on the geo-engineering and geotechnical applications of surface wave analysis and also on the production of near-surface models for deep exploration: in both cases the aim is to retrieve the trend of the S-wave velocity in the first tens up to hundreds meters of depth of the subsoil. The surface wave method exploits the geometric dispersion proper of surface waves: in a non-homogeneous medium every frequency is characterized by a different phase velocity, as every frequency component travels through a portion of medium whose thickness is proportional to its wavelength. The curve associating every frequency component to its phase velocity is called dispersion curve, and it constitutes the experimental datum one uses for the solution of an inverse problem to estimate the behaviour of S-wave velocity in the subsurface. The inversion is performed by assuming a 1D forward modelling simulation and suffers from equivalence problems, leading to the non uniqueness of the solution. Despite its great ductility, the main limitation of surface wave method is constituted by its 1D approach, which has proved to be unsatisfactory or even misleading in case of presence of lateral variations in the subsoil. The aim of the present work is to provide data processing tools able to mitigate such limitation, so that the surface wave method can be effectively applied in laterally varying media. As far as the inadequacy of surface wave method in case of 2D structures in the subsoil, I developed two separate strategies to handle smooth and gradual lateral variations and abrupt subsurface heterogeneities. In case of smooth variations, the approach I adopted aims at “following” the gradual changes in subsoil materials properties. I therefore propose a procedure to extract a set of neighbouring dispersion curves from a single multichannel seismic record by applying a suitable spatial windowing of the traces. Each curve corresponds to a different subsoil portion, so that gradual changes in subsoil seismic parameters can be reconstructed through the inversion of dispersion curves. The method was tested on synthetic and real datasets, but proved its reliability in processing the data from a small scale seismic experiment as well. In the context of characterizing smooth 2D structures in the subsurface via the surface wave method, I also developed a procedure to quantitatively estimate the (gradual) lateral variability of model parameters by comparing the shape of local dispersion curves, without the need to solve a formal inverse problem. The method is based on a sensitivity analysis and on the applications of the scale properties of surface wave. The procedure can be devoted to different applications: I exploited it to extend a priori local information to subsoil portions for which an experimental dispersion curve is available and for an estimation of the lateral variability of model parameters for a set of neighboring dispersion curves. The method was successfully applied to synthetic and real datasets. To characterize sudden and abrupt lateral variations in the subsurface, I adopted another strategy: the aim is to estimate the location and embedment depth of sharp heterogeneities, to process separately the seismic traces belonging to quasi-1D subsoil portions. I adapted several methods, already available in literature but developed for different purposes and scales, to the detection of sudden changes in subsoil seismic properties via the analysis of anomalies in surface wave propagation. I got the most promising results when adapting these methods, originally developed for single shot configurations, to multifold seismic lines, exploiting their data redundancy to enhance the robustness of the analyses. The outcome of the thesis is therefore a series of processing tools that improve the reliability and the robustness of surface wave method when applied to the near surface characterization of laterally varying media.