Bayesian inversion of frequency-averaged reflection data

Abstract

This paper presents a nonlinear Bayesian inversion of high-resolution seabed reflection data to estimate visco-elastic parameters of the upper sediments. The inversion is applied to data from two sites in the Strait of Sicily. One site is characterized by low-velocity, silty-clay sediments, resulting in data with a well-defined angle of intromission. The second site is characterized by high-velocity clayey sand, resulting in a critical angle. The data were frequency averaged from 500–2000 Hz and inverted for visco-elastic parameters of a half-space seabed model. The likelihood function employed in the inversion is based on the assumption of independent, Gaussian-distributed data errors, with the standard deviation included as a nuisance parameter in the inversion. Statistical tests are applied to the data residuals a posteriori to validate these assumptions. Good results are obtained for sediment compressional-wave velocity, compressional attenuation, and density; shear parameters are less-well determined although low shear-wave velocities are indicated. The Bayesian analysis provides a quantitative comparison of inversion results for the two sites in terms of the resolution of specific geoacoustic parameters, and indicates that the geoacoustic information content is significantly higher for angle-of-intromission data.