Freeze Bath Inversion for Estimation of Geoacoustic Parameters

Abstract

This chapter describes a new approach for estimating geoacoustic model parameters by matched field inversion of broadband data. The objectives are to design an efficient method to invert accurate estimates of the geoacoustic parameters, and obtain statistical measures of the confidence limits. The essential requirements for the inversion are an efficient search mechanism for exploring the multidimensional model parameter space, and a cost function and propagation model that are appropriate for broadband data. The method should also be robust to sources of mismatch in the experiment, such as imprecise knowledge of the experimental geometry or of the geoacoustic model itself. In this approach, the cost function is based on a multifrequency processor that matches the measured waveform with modeled waveforms that are calculated by ray theory. The search process is a statistical freeze bath algorithm that provides a representation of the distribution of models that fit the data well. The efficiency of the search is improved by reparameterizing, using new parameters based on the covariance of the sampled models. The method is applied to synthetic data that simulate the environment of the Haro Strait tomography experiment.