An indirect approach to classify backscatter data for soft bottom habitat mapping

Abstract

The availability of wide area habitat maps is a key aspect for several marine assessment scenarios. The development of innovative, quick and low cost methods to produce such maps is therefore crucial to overcome the actual knowledge gap. Multibeam echosounder (MBES) backscatter data acquisition is largely used to collect geophysical data which are then post processed versus in situ data (sediment and/or biological samples) to classify the acoustic response (hereafter direct method). However, many parameters as roughness, porosity, presence of gas seeps, etc. strongly affect the results. In fact, a number of recent works highlight that the availability of backscatter and sediment data is not sufficient to evaluate the relationship between acoustic response and seabed substratum (De Falco et al., 2010; Ferrini and Flood, 2006; Goff et al., 2000; Sutherland et al., 2007). This is particularly true in big areas, which are often characterized by wide depth range and heterogeneous substrata and where in situ data can be insufficient. In order to exceed these critical aspects in backscatter data elaboration, a new procedure (indirect method) is presented in this work. The proposed classification method can be synthetized as follows: (1) MBES backscatter data cleaning and homogenization, (2) identification of "disturbed" areas (i.e. trawled areas, dredging areas, etc.), (3) MBES backscatter data clustering with identification of a high number of acoustic facies (a multiple of the sediment classes recorded by in situ samples), (4) association of different acoustic facies to a specific sampled sediment class. This procedure was tested in a study area located in the North Latium coast in the central Tyrrhenian Sea. This area extends 30 square kilometres with a depth range between 40 and 100 metres, mainly characterized by soft bottom. The application of the proposed method classifies six groups of acoustic facies allowing adequate control of the difference in backscatter response (i.e. due to attenuation with depth). Furthermore, this method provides encouraging results also in areas with a very low number of in situ samples. The comparison between direct and indirect methods confirms the usefulness of this approach for the identification of soft bottom habitats in wide areas.