A Bayesian framework for geoacoustic inversion of wind-driven ambient noise in shallow water

Abstract

Bayesian inversion is applied to estimate the joint posterior probability density (PPD) of geoacoustic parameters. The PPD is sampled by a reversible-jump Markov chain Monte Carlo (rjMCMC) algorithm, which uses an extended Metropolis-Hasting (MH) criterion that allows trans-D jumps between parameterizations, quantifying the uncertainly due to the lack of knowledge of the model parameterization. Sequential datsets are obtained by discretizing continuous-time recordings of ambient noise. Conventional beamforming was used to estimate the BL at 8 frequencies in the range 550 Hz to 1400 Hz. The BL data at 20 uniformly-spaced grazing angles from 14° to 90° is provided to the sequential Bayesian trans-D Monte Carlo algorithm for estimation of the PPD. The geoacoustic parameters and the depth of acoustic interfaces closely resemble the true profiles. Show more