TY - JOUR T1 - The determinants of alpine butterfly richness and composition vary according to the ecological traits of species JF - bioRxiv DO - 10.1101/002147 SP - 002147 AU - Vincent Sonnay AU - Loïc Pellissier AU - Jean-Nicolas Pradervand AU - Luigi Maiorano AU - Anne Dubuis AU - Mary S. Wisz AU - Antoine Guisan Y1 - 2014/01/01 UR - http://biorxiv.org/content/early/2014/01/27/002147.abstract N2 - Predicting spatial patterns of species diversity and composition using suitable environmental predictors is an essential element in conservation planning. Although species have distinct relationships to environmental conditions, some similarities may exist among species that share functional characteristics or traits. We investigated the relationship between species richness, composition and abiotic and biotic environment in different groups of butterflies that share ecological characteristics. We inventoried butterfly species richness in 192 sites and classified all inventoried species in three traits categories: the caterpillars diet breadth, the habitat requirements and the dispersal ability of the adults. We studied how environment, including influence butterfly species richness and composition within each trait category. Across four modelling approaches, the relative influence of environmental variables on butterfly species richness differed for specialists and generalists. Climatic variables were the main determinants of butterfly species richness and composition for generalists, whereas habitat diversity, and plant richness were also important for specialists. Prediction accuracy was lower for specialists than for generalists. Although climate variables represent the strongest drivers affecting butterfly species richness and composition for generalists, plant richness and habitat diversity are at least as important for specialist butterfly species. As specialist butterflies are among those species particularly threatened by global changes, devising accurate predictors to model specialist species richness is extremely important. However, our results indicate that this task will be challenging because more complex predictors are required. ER -