PUBLICATIONS

The progressive increase of marine macro-litter on the bottom of the Mediterranean Sea is an urgent problem that needs accurate information and guidance to identify those areas most at risk of accumulation. In the absence of dedicated monitoring programs, an important source of opportunistic data is fishery-independent monitoring campaigns of demersal resources. These data have long been used but not yet extensively. In this paper, MEDITS data was supplemented with 18 layers of information related to major environmental (e.g. depth, sea water and wind velocity, sea waves) and anthropogenic (e.g. river inputs, shipping lanes, urban areas and ports, fishing effort) forcings that influence seafloor macro-litter distribution. The Random Forest (RF), a machine learning approach, was applied to: i) model the distribution of several litter categories at a high spatial resolution (i.e. 1 km2); ii) identify major accumulation hot spots and their temporal trends. This approach, based on georeferenced information widely available in public databases, seems a natural candidate to be applied in other basins as a support and complement tool to field monitoring activities and strategies for protection and remediation of the most impacted areas.