Bayesian matched-field inversion for shear and compressional geoacoustic profiles at the New England Mud Patch

This paper estimates depth-dependent profiles of shear- and compressional-wave geoacoustic properties for seabed sediments at the New England Mud Patch through matched-field inversion of broadband (20-2000 Hz) acoustic data recorded at a 14-element vertical line array due to a combustive sound source. Trans-dimensional Bayesian inversion is applied to sample probabilistically over the number of seabed layers and corresponding layer depths and geoacoustic properties, as well as over the order and parameters of an autoregressive error model. This approach, based on a parallel-tempering implementation of birth/death reversible-jump Markov-chain Monte Carlo sampling, combines objective, data-driven model selection and quantitative parameter/uncertainty estimation. Results indicate low shear-wave speeds (∼30 m/s) with small uncertainties over most of the upper mud layer, increasing in underlying transition and sand layers, with values in good agreement with in situ probe measurements (for the mud) and nominal values (for sand). The compressional-wave attenuation profile is well estimated but shear-wave attenuation is poorly constrained. Comparison of results for inversions both with and without shear-wave parameters and consideration of inter-parameter correlations indicate that estimates of compressional-wave parameters, including attenuation, are not substantially influenced by shear-wave effects, with the possible exception of the sand layer.