Estimation of poroelastic parameters from measured reflectivity

Abstract

The reflectivity of sound at a water-sediment interface contains information about the properties of the sediment. Estimates of these properties can be obtained by optimizing the match between the magnitude of the measured reflectivity, as a function of grazing angle and frequency, and values computed using a poroelastic sediment model. A traditional simulated annealing approach yields such parameter estimates but little information about the uncertainty associated with those estimates. An alternate approach computes the posterior probability distributions (PPDs) for the parameters using a fast Gibbs’ sampler approach. The PPDs not only indicate the most probable values for each parameter, but the widths of the distributions also provide an understanding of the uncertainty and the dependence of the data-model mismatch on each parameter. Both of these methods are applied to reflectivity measured in a laboratory tank over a frequency band of 77–147 kHz and grazing angles of 7–49 deg. [Work suppported by ONR, Ocean Acoustics.]