Bayesian Inversion of Multimode Interface-Wave Dispersion From Ambient Noise

Abstract

This paper applies nonlinear Bayesian inversion to estimate seabed shear-wave speed profiles and their uncertainties using interface-wave dispersion curves extracted from ocean ambient noise, and compares the resolution of seabed structure for fundamental mode and multimode data. Ambient noise recordings were collected for 2.15 h at hydrophones of an entrenched ocean bottom cable in the North Sea. Scholte-wave dispersion curves for the fundamental mode and several higher order modes within the frequency range 0.25–3.9 Hz are extracted from cross correlations of noise recordings at sensor pairs via the slowness-frequency transform. The Bayesian information criterion is used to determine the preferred model parameterizations in terms of the number of sediment layers supported by the data for inversions based on the fundamental mode alone and on the first three modes. Adaptive hybrid optimization and Metropolis–Hastings sampling are applied to estimate the optimum a posteriori shear-wave speed models and to compute marginal posterior probability distributions and profiles. The results show quantitatively that multimode inversion provides higher resolution of shallow shear-wave speed structure with smaller uncertainties at all depths than inversion of the fundamental mode alone.