Combining genetic and linearized algorithms for a two-step joint inversion of Rayleigh wave dispersion andH/Vspectral ratio curves

Abstract

SUMMARY The joint inversion of Rayleigh wave dispersion and H/V curves from environmental noise measurements allows the retrieval of S-wave velocity profiles for the shallow subsoil. For this purpose, genetic and linearized algorithm have been combined in a two-step inversion procedure, that allows the principal drawbacks typical of the application of each algorithm separately to be overcome. In the first step, a genetic algorithm procedure is used to constrain the subvolume of the parameter space where the absolute minimum of the misfit function is located. In the second step, a linearized inversion algorithm, having as an initial guess the minimum misfit model deduced from the first step, is applied to force the inversion towards the optimal solution. To evaluate the feasibility and effectiveness of this approach, seismic noise recordings at a test site in the Po river valley (North Italy) have been analysed. Here, detailed geophysical and geological information is available along with earthquake recordings, which allow a well constrained definition of both the local shear wave profile and transfer function. Comparisons between theoretical and experimental S-wave velocity profiles and, above all, between the theoretical and experimental site response functions shows that this combination of inversion procedures can very efficiently to manage the extreme non-linearity of the problem.