Comparison of five modelling techniques to predict the spatial distribution and abundance of seabirds.

Abstract

The number and complexity of modelling techniques used to predict species distributions has increased substantially over the past decades (Hegel et al., 2010), and several comparisons of model performance have been carried out for terrestrial species (e.g., Elith and Graham, 2009; Elith et al., 2006; Segurado and Araújo, 2004). In contrast, the marine environment is less studied and more challenging given its dynamic nature (Ready et al., 2010; Robinson et al., 2011; Wakefield et al., 2009). Furthermore, seabirds are highly mobile species, and their presence at certain locations varies temporally depending on whether an area is used during the breeding season, as a migration stopover, or as moult refuge (Tremblay et al., 2009). A comparison of the performance of different models that predict distributions and abundances of seabirds based on shipboard survey data has to our knowledge only been explored for one coastal species (Yen et al., 2004), yet the bourgeoning interest in the identification of pelagic marine protected areas warrants a comparison of newer distribution modelling techniques. Here we compare the performance of five modelling techniques to predict the occurrence and abundance of a migratory seabird species outside of the breeding season. The Balearic Shearwater (Puffinus mauretanicus) is a critically endangered species that breeds only at the Balearic archipelago in the western Mediterranean, and migrates to the North-East Atlantic after the breeding season (Brooke, 2004). The species suffers from high adult mortality at sea (Oro et al., 2004), and most research efforts have focused on understanding foraging ecology and distribution during the breeding season in the Mediterranean (Bartumeus et al., 2010; Louzao et al., 2006a, 2006b). Marine protected areas are needed for Balearic Shearwaters throughout its range, and although both Spain and Portugal have delineated marine important bird areas (IBAs, Arcos et al., 2009; Ramírez et al., 2008), most of these areas were still not legally protected as of October 2011 (BirdLife Inter-national, 2010a). Our model comparison aims to inform seabird conservation managers about the performance of modelling techniques that can be used to predict the spatial distribution and abundance of seabirds for the identification of marine IBAs or protected areas. We tested model predictions against independent data and com-pared predicted distributions with the locations of existing marine IBAs to evaluate whether our model results agree with IBAs that were identified with a variety of different methods (Ramírez et al., 2008). Thus, we provide information on which modelling techniques are useful for seabirds, and identify areas that may war-rant protection to benefit the Balearic Shearwater.