Geostatistical modelling of multibeam backscatter for full-coverage seabed sediment maps

Extensive seabed sediment mapping is highly relevant to describe marine ecosystems and to quantify the distribution and extent of benthic habitats. Compared to traditional mapping methods, primarily based on bed sampling, multibeam echo sounding (MBES) is a time-efficient tool to acquire high-resolution bathymetric and backscatter data over large areas. We use a Bayesian method for unsupervised acoustic sediment classification (ASC) of MBES backscatter data of the Cleaver Bank, Netherlands Continental Shelf. On these sparsely distributed backscatter datasets, we tested and evaluated different Kriging algorithms, showing that Ordinary Kriging results in a reliable map. We introduce a new approach to classify interpolated MBES backscatter based on the Bayesian method for producing full-coverage sediment maps. Comparison to a traditional sediment map and in situ measurements shows that this approach resolves lateral heterogeneities (kilometers). When evaluating the high-resolution sediment map obtained from the Bayesian method, based on the actual backscatter, mapping laterally heterogeneous sediments significantly improved (meters). In order to create the optimal sediment map, we aimed to integrate ASC into existing maps, which, however, requires quantified spatial uncertainties of both considered maps. Finally, the low discrimination power of MBES backscatter for coarse sediments is highlighted as a shortcoming of current ASC mapping.