Ship-noise based geoacoustic inversion via particle filtering of vertical specific acoustic impedance

Abstract

This paper presents a sequential approach to infer sediment geoacoustic properties from the observation of vertical specific acoustic impedance due to ship noise. This acoustic quantity does not require knowledge of the source and is sensitive to ocean bottom properties including density. The approach is demonstrated for the characterization of sediment off the small Senegalese coast during EHL-IRD joint experiments [ECOAO 13]. The noise field due R/V Antea sailing parallel to the coast was recorded on a vertical, multi-wavelength pressure-gradient array (EHL) from which impedance data was derived. A particle filter (PF) simultaneously tracks the range variations of impedance at a number of discrete frequencies in order to output a sequence of environmental parameter estimates with their associated uncertainties in the form of posterior probability densities (PPDs). The range-averaged inversion results are in good agreement with those produced by a classical batch inversion method based on a genetic algorithm (GA). Apparent inhomogeneity of the ocean bottom is observed, which is consistent with the sieving analysis of sediment grab samples collected at two different locations. When compared to batch processing, the computational efficiency and robustness of particle filtering are due to the capacity of iteratively updating the estimated PPDs, as is demonstrated by implementing the inversion with different particle sizes, of 200, 300 and 400.