Predictive performance of the BRT model relating the beach species distributions to environment, averaged across the species (presence/absence) over the regional scale.

*Mean and se estimated within model building. Standard error (se).Estimates of predictive performance were calculated using 10-fold cross validation and a tree complexity of 3. Learning rate (lr), the number of trees (n.trees), and the number of positive observations (pa) are shown. AUC: area under the receiver operating characteristic curve. The final environmental variables included in the models are shown in Table 2. Beach species included in the final model: Bathyporeia pelagica, Donax trunculus, Eurydice affinis, Eurydic pulchra, Pontocrates arenarius, Scolelepis squamata, Sphaeroma rugicauda, Talorchestia brito and Talitrus saltator.