Bayesian Geoacoustic Inversion With the Image Source Method

This paper develops a Bayesian approach to the image source method (ISM) for efficient inversion of seabed reflection data to estimate geoacoustic parameters and uncertainties. Based on the representation of layered seafloor-reflected signals by image sources, ISM is a very efficient method which provides the local sound-speed profile (SSP) of the sediment structure. It is a two-step method: first, the image sources are detected and localized from the recorded signals, and second, from these locations, the thickness and sound speed of each sediment layer are estimated from the Snell-Descartes law of refraction. This work focuses on the definition and construction of the image sources with a distinction between real and virtual image sources which has consequences on the uncertainties of ISM. The localization of the image sources is performed within a Bayesian formulation based on sampling the posterior probability density to estimate the median SSP and uncertainties. The algorithm is tested first on synthetic data, with results in excellent agreement with Bayesian travel-time inversion but a much lower computational cost. Finally, the Bayesian ISM is applied to at-sea data measured in the Scattering And ReverberAtion from the sea Bottom (SCARAB) experiment, which took place near Elba Island off the west coast of Italy in 1998, and the resultant sediment SSP agrees well with previous results of other geoacoustic inversion methods.