Using multibeam echosounder datafor a GIS-ready seafloor characterization in the Adriatic Sea

Multibeam Echosounder (MBES) is increasingly becoming suitable for marine habitat mapping and various classification methods, diversifying in terms of classification algorithms, automation and MBES involved data features, are requiring scientific attention. On the other hand, the improved quality and volume of MBES data, as well as the increased demand for mapping products, is driving the desire for repeatable seafloor mapping approaches, analogous to the automated classification of satellite imagery in terrestrial remote sensing. In this context, the proposed study deals with Kongsberg EM3002D data and investigates how to tailor a proprietary hybrid classification approach, successfully developed for terrestrial Land Cover mapping, to seafloor characterization. Dataset includes full-coverage bathymetry, backscatter intensity and their derived textural/statistical features. Concurrent clam fisheries surveys (striped venus Chamelea gallina) give the chance to carry out sample-specific quantitative correlations and investigate how shellfish beds can affect MBES backscatter and consequently be used to detail the seafloor characterization. Experimental results show how to improve the overall acoustic seabed classification first integrating MBES data along with their derivatives and later combining information at both pixel and object level to detail the seafloor map legend in terms of probable presence of clams. The final habitat data is GIS-ready, hence suitable to be directly used for marine spatial/temporal planning and resource management.