A study of the uncertainty in estimate seabed properties as a result of goeacoustic inversion algorithms

Abstract

This paper investigates the inherent uncertainty in estimated seabed properties derived from matched-field geoacoustic inversion algorithms. This algorithm-induced uncertainty must be considered when interpreting inversion results in terms of environmental changes as a function of time or space (eg., water column changes or range dependence). Fast simulated annealing (FSA), genetic

algorithms (GA), and adaptive simplex simulated annealing (ASSA) are compared by performing

multiple inversions of synthetic data (noise-free and noisy) and acoustic data measured over two

areas with different sediment types. ASSA is shown to produce the lowest uncertainty in inversion results for all cases, followed by GA and FSA. For one the sediment types the uncertainty is essentially negligible, while for the other case the uncertainty is significant compared to environmental

variations reported in the literature.