Bayesian inference and model selection to investigate seabed shear-wave velocity profiles via interface-wave dispersion inversion

This paper applies a general geoacoustic profile parameterization based on Bernstein polynomial basis functions to consider the form of seabed shear-wave velocity profiles via Bayesian inversion of interface (Scholte) wave dispersion. Theory and observations indicate that the shear-wave velocity profile of seabed sediments of uniform composition and density often corresponds (approximately) to a power law function of depth. Hence, past inversions have often been based on a power-law parameterization without considering independently if this form is actually required by the data. The Bernstein parameterization, based on a linear combination of Bernstein-polynomial basis functions with the polynomial order determined by the Bayesian information criterion, is general and allows the form of the profile to be determined by the data, rather than by a subjective prior choice. In this paper, measured-data inversions are compared for power-law, Bernstein-polynomial, and layered trans-dimensional parameterizations to investigate the shear-wave velocity profile form.