Seafloor change detection using multibeam echosounder backscatter: case study on the Belgian part of the North Sea

To characterize seafloor substrate type, seabed mapping and particularly multibeam echosounding are increasingly used. Yet, the utilisation of repetitive MBES-borne backscatter surveys to monitor the environmental status of the seafloor remains limited. Often methodological frameworks are missing, and should comprise of a suite of change detection procedures, similarly to those developed in the terrestrial sciences. In this study, pre-, ensemble and post-classification approaches were tested on an eight km2 study site within a Habitat Directive Area in the Belgian part of the North Sea. In this area, gravel beds with epifaunal assemblages were observed. Flourishing of the fauna is constrained by overtopping with sand or increased turbidity levels, which could result from anthropogenic activities. Monitoring of the gravel to sand ratio was hence put forward as an indicator of good environmental status. Seven acoustic surveys were undertaken from 2004 to 2015. The methods allowed quantifying temporal trends and patterns of change of the main substrate classes identified in the study area; namely fine to medium homogenous sand, medium sand with bioclastic detritus and medium to coarse sand with gravel. Results indicated that by considering the entire study area and the entire time series, the gravel to sand ratio fluctuated, but was overall stable. Nonetheless, when only the biodiversity hotspots were considered, net losses and a gradual trend, indicative of potential smothering, was captured by ensemble and post-classification approaches respectively. Additionally, a two-dimensional morphological analysis, based on the bathymetric data, suggested a loss of profile complexity from 2004 to 2015. Causal relationships with natural and anthropogenic stressors are yet to be established. The methodologies presented and discussed are repeatable and can be applied to broad-scale geographical extents given that broad-scale time series datasets become available.