Ocean acoustic inversion for seabed geoacoustic properties

Abstract

Estimating geoacoustic properties of the seabed from ocean acoustic data provides a convenient in situ alternative to direct sampling (e.g., coring), with parameter sensitivities relevant to sonar applications. However, this requires solving a strongly nonlinear inverse problem, which is inherently nonunique. Hence, quantifying uncertainties for the recovered geoacoustic parameters is an important, but challenging, problem. This talk will describe a nonlinear Bayesian approach to geoacoustic inversion based on estimating properties of the posterior probability density (PPD), which combines information from observed data with prior information. An efficient Markov-chain method (Gibbs sampling) is applied to extract properties of the PPD, including optimal parameter estimates, marginal probability distributions, variances/covariances, and interparameter correlations. The inversion formulation is general, and will be illustrated with examples for a variety of approaches to geoacoustic inversion, including the inversion of acoustic field data, seabed reflectivity measurements, acoustic reverberation data, and ambient noise measurements.