Estimating geoacoustic parameters from broadband data from the New Jersey shelf

Abstract

This paper presents the geoacoustic inversion of broadband data to estimate the sea bed parameters of SWARM95 site. The data set used in this study contains air gun signals that were transmitted at the depths of 25 and 45 m, respectively, below the sea surface, while the source ship was maneuvering at small speed (towing); the signals then were collected at the ranges around 5 km by a vertical line array. A canonical geoacoustic model was generated first according to previous experimental research results, for the purpose of a parameter sensitivity study; then both hybrid optimization and Bayesian inversion techniques were applied to the real data. The data error covariance matrix is estimated to resolve the spatial correlated data error in Bayesian inversion approach. Since the ratio of range and water depth is rather large in this study, only the geoacoustic parameters at the top of the sediment are sensitive in the inversion. The optimization results show the consistency of sensitive geoacoustic/geometric parameter estimations over the ranges and different source depths. Bayesian inversion gives the estimation uncertainties of the geoacoustic properties by marginal probability distributions. [Work supported by ONR.]