Comparison and combination of matched-field and modal-dispersion inversion for seabed geoacoustic profiles at the New England Mud Patch

This paper considers the information content for seabed geoacoustic inversion of recorded acoustic waveforms processed as modal-dispersion (MD) data (mode arrival times as a function of frequency) and matched-field (MF) data (multifrequency complex acoustic fields across a sensor array). These approaches are applied separately and combined in joint inversion, where the MD and MF data sets are derived from the same acoustic recordings collected during the 2017 Seabed Characterization Experiment on the New England Mud Patch. Unlike MD inversion, MF inversion requires knowledge of source and receiver depths, and the complex source spectrum must be estimated as part of the inversion. However, MF inversion is sensitive to seabed attenuation (MD is not) and more easily extended to higher frequencies, where mode filtering for MD data is challenging. Comparison of geoacoustic information content is facilitated here using trans-dimensional Bayesian inversion to sample probabilistically over the number of layers of the seabed model as well as the order of an autoregressive error model. Results indicate MF inversion resolves more detailed geoacoustic structure with smaller uncertainties, including good estimates of the attenuation profile for wideband (20-1504 Hz) inversions.