Defining a procedure for integrating multiple oceanographic variables in ensemble models of marine species distribution

Marine species distribution models are largely used in ecology and fishery science to support the management of marine exploited resources. This study was focused on defining a procedure to develop an ensemble model, combining 5 differents modeling approaches. For each approach, we implemented spatial training and test data set to evaluate the best performance of 9 models through a set of indicators. These models include a different combination of covariates, starting with the simplest model (depth, year, and spatiotemporal variables) to the most complex one (including also oceanographic variables and effort). We applied this procedure to develop a robust ensemble of models to describe accurately the distribution in space and time of European hake (Merluccius merluccius) in the Adriatic and Ionian sea.