Benthosearcher: a machine learning based tool to ultra-fast, automatic characterization of vulnerable marine ecosystems

Abstract

The degradation and impoverishment of the seabed that has been detected during last decades is the result of numerous variables, among which are uncontrolled exploitation of the seabed concerning its vulnerability, based on trawling gear. Intending to bring the situation under control, European fishing authorities have concluded a series of proposals to promote, define and defend Marine Vulnerable Ecosystems (VMEs), among which one of the most controversial and with the greatest social and economic impact is the veto of trawling in numerous fishing areas/grounds that currently are exploited by a large c number of vessels from many countries of the EU, which see their livelihood in danger. In the process of proposing an alternative that is attractive to both parts, we propose an automatic, real-time tool (BentoSearcher) based on artificial intelligence so that trawlers have will have the autonomy to decide whether or not to cast the net on the seabed in which the vessel is operating based on the data of benthic species detected in previous fishing hauls or trips that characterize and allow to identify vulnerable seabeds.