Range-dependent matched-field geoacoustic inversion using a tonal source towed on circular tracks

We aim to estimate the acoustic properties and their spatial variability on the New England Mud Patch using data collected during the Seabed Characterization Experiment 2017. In particular, we use tonal signals between 303 and 953 Hz emitted by a source towed on circular tracks and recorded by a vertical line array. For Bayesian geoacoustic inversion, we develop a Metropolis-Hastings (MH) method that combines several ideas: (i) adaptation to learn "on the fly" the covariance matrix of the posterior probability density function (PDF) of the joint parameter vector, (ii) principal-component updates for identifying and jointly sampling correlated variables, (iii) parallel tempering to better explore multi-modal geoacoustic PDFs, (iv) an observation model based on adiabatic normal modes that can describe moderately range-dependent sound propagation, and (v) a model parametrization that allows geoacoustic parameters to change linearly with depth in each layer (with a specific focus on sound speed). We demonstrate that our MH method can accurately estimate mud-related geoacoustic parameters as well as the range dependence of the water depth and upper-sediment mud thickness. We also compare a range-independent variant of our approach with a range-independent trans-D reference method.