Bayesian inversion of reverberation and propagation data for geoacoustic and scattering parameters

Abstract

This paper applies nonlinear Bayesian inference theory to quantify the information content of reverberation and short-range propagation data, both individually and in joint inversion, to resolve seabed geoacoustic and scattering properties. The inversion of reverberation data alone is shown to poorly resolve seabed properties because of strong multi-dimensional correlations between parameters. Inversion of propagation data alone is limited by different correlations, but better constrains the geoacoustic parameters. However, propagation data are insensitive to scattering parameters such as Lambert's scattering coefficient. In each case the parameter correlations are inherent in the physics of the forward problem (reverberation and propagation) and cannot be overcome by processing or inversion techniques; rather, the inversion of more informative data is required. This is accomplished here by joint inversion of reverberation and propagation data, weighted according to their respective maximum-likelihood error estimates. Joint inversion of reverberation and propagation data collected on the Malta Plateau (Strait of Sicily) resolves both geoacoustic and scattering properties and achieves smaller uncertainties for all parameters than obtained by the inversion of either data set alone.