Statistical analysis of high-frequency multibeam backscatter data in shallow water

The seabed type of most shallow water areas is very often not homogenous but patchy in space and changeable with time. Recent studies have shown that statistics of acoustic backscatter from an inhomogeneous seabed deviate significantly from a conventional Rayleigh model and may show distributions with multiple modes and heavier tails. In this work, seabed backscatter statistics are analysed using the data collected with a RESON SeaBat 8125 multibeam sonar system (operating at 455kHz) in the regions of Recherche Archipelago and Cockburn Sound in Western Australia. The 455-kHz acoustic backscatter data were collected over areas consisting of different seabed types, including sand, rhodolith, seagrass and bedrock. Statistical models of backscatter intensity from these different seabed types were investigated for different values of the incident angle. The goodness of fit for the model with the experimental distributions was assessed with the nonparametric Kolmogorov-Smirnoff (KS) test statistic. Before compensation for the angular dependence of backscatter, the Rayleigh mixture probability distribution function provides the best fit to all experimental distributions at near-nadir angles of incidence. Within moderate and oblique angle intervals, the lognormal distribution fits experimental distributions best around the centre of the distributions and the Rayleigh 3component mixture distribution in most cases provides the best fits to the tail of the distributions. After the angular correction, the log-normal distribution fits empirical distributions best in all cases, but the Rayleigh mixture distribution offers the best fits to the tail of distributions for seagrass and bedrock seafloors.