Living on predictability: modelling the density distribution of efficient foraging seabirds

Abstract

In areas with regular fishing coastal fleets seabirds may benefit from the predictability of discards from fishing vessels, but it is not clear to what extent birds rely on this predictable resource and whether foraging is synchronized with the diel availability of discards. In this paper we investigate if a typical scavenger species, the yellow‐legged gull Larus michahellis, takes advantage of the temporal and spatial predictability of fish discards in the western Mediterranean Sea. The activity and distribution of the trawling fleet in this area is regulated and very predictable in time and space. We gathered aerial survey data across a relatively large area close to the coast to study the spatial distribution and density of L. michahellis, and modelled the density distribution of the species in relation to several oceanographic, ecological and temporal variables, using two different modelling approaches: MARS (multivariate adaptative regression splines) and GLM (generalized linear models). Our models suggest that the spatial density of trawlers at sea and the time of the day are the best explanatory variables of gull distribution, and that gulls concentrate in areas with vessels mainly during fish discarding time, supporting the hypothesis that gulls optimize time foraging to take advantage of fishery waste predictability. Additional surveys from the main gull roosting sites inshore support this hypothesis, as gulls start leaving to the sea just before fishing is completed and vessels begin discarding fish scraps when back to the harbour. This study represents one of the few examples of applying MARS to density distribution modelling, although its application to marine ecosystems should be conducted with caution because of large areas with real absence data. GLMs have shown to be more adaptable to such kind of data. Our data confirm the importance of fishery waste for L. michahellis, not only as a food resource but also as a major driver of their activity and distribution patterns. The ability of seabirds to predict accurately when a food resource will be available implies that modelling their distribution at sea needs to include such variables, both in spatial and temporal dimensions.