The inversion of spectral ratio H/V in a layered system using the Diffuse Field Assumption (DFA)

Abstract

In order to evaluate the site effects on seismic ground motion and establish preventive measures to mitigate these effects, the dynamic characterization of sites is mandatory. Among the various geophysical tools aimed to this end, the horizontal to vertical spectral ratio (H/V) is a simple way to assess the dominant frequency of a site from seismic ambient noise. The aim of this communication is contributing to enhance the potential of this measurement with a novel method that allows extracting from the H/V the elastic properties of the subsoil, assumed here as a multilayer medium. For that purpose, we adopt the Diffuse Field Assumption from both the experimental and the modeling perspectives. At the experimental end, the idea is to define general criteria that make the data processing closely supported by theory. On the modeling front, the challenge is to compute efficiently the imaginary part of Green’s function. The Cauchy’s residue theory in the horizontal wavenumber complex plane is the selected approach. This method allows both identifying the contributions of body and surface waves and computing them separately. This permits exploring the theoretical properties of the H/V under different compositions of the seismic ambient noise. This answers some questions that historically aroused and gives new insights into the H/V method. The efficient forward calculation is the prime ingredient of an inversion scheme based on both gradient and heuristic searches. The availability of efficient forward calculation of H/V allows exploring some relevant relationships between the H/V curves and the parameters. This allows generating useful criteria to speed up inversion. As in many inverse problems, the non-uniqueness issues also emerge here. A joint inversion method that considers also the dispersion curves of surface waves extracted from seismic ambient noise is presented and applied to experimental data. This joint scheme mitigates effectively the non-uniqueness.