Predicting summer fin whale distribution in the Pelagos Sanctuary (north‐western Mediterranean Sea) to identify dynamic whale–vessel collision risk areas

Abstract

Abstract Mediterranean fin whales aggregating in the Pelagos Sanctuary in summer to feed are exposed to vessel collision risk, particularly from high‐speed ferries. This study developed models to predict summer fin whale distribution using a generalized additive model (GAM) and MaxEnt, with the aim of providing a tool to identify potential high whale–ferry collision risk areas along ferry routes within the Pelagos Sanctuary during summertime. Models were trained using sightings data collected in the summer months of 2009–2018 on board ferries crossing the central area of the Pelagos Sanctuary. Environmental predictors were bathymetry and mean sea surface chlorophyll concentration of the annual spring bloom period. The predictive ability of GAM and MaxEnt was assessed using existing knowledge of summer fin whale distribution in the region. GAM (deviance explained = 20.2%) predictions matched documented distributions more closely than that of MaxEnt, with highest predicted fin whale occurrence in deep offshore waters (>2000 m) encompassing the central north‐western and western regions, and in the south‐eastern region, consistent with known fin whale habitats within the Pelagos Sanctuary. Inter‐annual variability was evident, influencing collision risk areas. Collision risk was estimated as a function of the overlap between the predicted probability of fin whale occurrence and ferry density estimated from Automated Identification System data. Ferry routes that cross the northern and eastern regions of the Pelagos Sanctuary presented relatively higher collision risk. Areas with changes in risk intensity between the years were temporally and spatially dynamic: some appeared intermittently throughout the study period while others persisted over consecutive years or recurred in different years. Due to the vastness of the Pelagos Sanctuary, vessel speed reduction maybe a more practical measure to manage collision risk than re‐routing shipping lanes. A combination of Seasonal Management Areas and Dynamic Management Areas approaches could be adopted for high‐risk areas.