Habitat suitability modeling with Random Forest as a tool for fish conservation in Mediterranean rivers

Prediction of fish presence is needed in many branches of regulated river management, including the definition of environmental flows and habitat restoration measures for wildlife conservation. Research on river ecology has indicated that fish species distribution can be related to habitat attributes, and models with high predictive performance can be obtained by combining biotic and abiotic habitat descriptors. In four selected reference sites of the Cabriel River (province of Cuenca, Spain), the presence of Eastern Iberian barbel (Luciobarbus guiraonis) was related to environmental variables linked to different mesohabitat characteristics. By means of Random Forest (RF), the data collected in the field were used to predict fish presence for two key bioperiods: migration and spawning (April-June) and rearing and growth (July-September). The aims of this study are (i) to select the most important habitat attributes for the fish presence (ii) to evaluate how biotic interactions among fish species affect habitat use and (iii) to examine the feasibility of using RF in building habitat suitability models for fish. Random Forest provided an indicator of variables’ importance and the most parsimonious model was selected to define the lowest number of variables to be surveyed for future model applications, e.g. habitat restoration measures and prediction of areas with high habitat suitability which should be conserved. The preliminary results of this research were discussed, as well as possible future developments, showing potentials and limitations of Random Forest in building habitat models for fish.