Quantifying data information content to resolve seabed structure in geoacoustic inversion

Abstract

This paper considers the importance of quantitative model selection and general parameterizations in estimating and interpreting seabed profiles in geoacoustic inversion, with application to data collected on the New England Mud Patch. In particular, the seabed structure that can be resolved depends on the information content of the acoustic data set under consideration, which varies with a number of factors, including the physics of the seabed acoustic interaction, frequency content of the data, and measurement and theory errors. Quantitative model selection applied to general parameterizations ensures the inclusion of seabed structure that is reliably sensed by the data while avoiding spurious structure. Data sets considered here include ship noise, modal dispersion, and wide-angle reflection coefficients. In each case, seabed models consistent with the data information content are estimated as part of the inversion using trans-dimensional and/or Bernstein-polynomial parameterizations. Results for all data sets indicate a low sound-speed mud layer over higher-speed sand; however, the ability to resolve structure within the mud layer, such as a transition to higher speeds near the mud base and possibly a weak positive gradient in the upper mud, depends on the information content of the various data sets. [Supported by the Office of Naval Research.]