Sampling methods for uncertainty quantification in source localization and geoacoustic inversion in the ocean

Abstract

Iterative and sequential Bayesian filtering approaches have been successfully employed for the estimation of select features of received acoustic signals - namely, arrival times and amplitudes of paths that have interacted with the propagation medium. These are subsequently utilized in source localization and environmental property estimation. Sequential filtering has the advantage of relating multipath arrival times across spatially separated hydrophones of a receiving array, providing tighter estimates of arrival times and amplitudes and, thus, probability densities with a reduced spread in inversion. We here present a new implementation of linearization of the relationship that links source location, water depth, and sound speed to the received sound field using arrival time estimates extracted with sequential filtering from time-series received at an array of phones in the ocean.