Geoacoustic inversion using global search

Abstract

The purpose of an experiment in ocean acoustics is often to estimate a representative set of parameters for a given environmental model and forward propagation model. Global search is a systematic approach to obtain these parameters relative to manual comparisons of the observed and modeled fields. The SAGA-code provides an engine for doing this comparison. The array data type can be transmission loss, complex amplitude, or reverberation. Several choices of state-of-the-art forward models are available as well as several choices of objective functions (least squares, Bartlett, broadband cross correlation, etc.). Finally, the optimization can be carried out either by exhaustive search, genetic algorithms, or simulated annealing. In a Bayesian approach, the result of an inversion is the a posteriori probability density for the estimated parameters from which all information such as mean, higher moments, and marginal distributions can be extracted. These distributions are useful in assessing data, environmental model, and obtained estimates. In a recent time reversal mirror experiment [Kuperman et al., J. Acoust. Soc. Am. 103, 25–40 (1998)], a vertical source array was transmitting to a vertical receive array at range 6.3 km in a 125-m waveguide. It will be investigated how a multisource inversion performs relative to a single source inversion for one way transmission.