An application of Bayesian inference techniques to compare three competing sandy sediment models

Abstract

Many competing geoacoustic models exist to represent sandy sediments. In this work, Bayesian inference is employed to compare the sensitivity of the physical input parameters of three competing models: the viscous grain shearing model of Buckingham, the Biot-Stoll model, and the extended Biot model of Chotiros, using previously taken wave speed and attenuation laboratory measurements. Bayesian model selection techniques are then employed to assess the degree to which the predictions of these models can be used to explain the observed data. Since all three geoacoustic models require a large number of input parameters and many of these input parameters are difficult to measure, a main goal of this work is to use Bayesian techniques to determine which parameters can be resolved via inversion and what kinds of data are required for parameter resolution. [Work supported by ONR, Ocean Acoustics.]