Different bat guilds perceive their habitat in different ways: a multiscale landscape approach for variable selection in species distribution modelling

Context Unveiling the scale at which organisms respond to habitat features is crucial to understand how they are influenced by anthropogenic environmental changes. We implemented species distribution models (SDMs) based on multiple-scale landscape pattern analysis for four bat species representative of different foraging guilds: Nyctalus leisleri, Rhinolophus hipposideros, Myotis emarginatus and Pipistrellus pipistrellus. Objectives (a) to assess the environmental factors and the influence of scale on the habitat suitability of bats; (b) to develop an objective methodology to select the best performing variables from a large variable dataset. Methods We performed the study in central Italy (Tuscany): 381 variables were derived from topographical and habitat maps using a moving windows analysis set at three spatial scales (1, 5 and 10 km) that are ecologically meaningful for bats. For each species, we ran 381 univariate models to select the variables for multivariate SDMs. Results All the variables retained in the SDMs described spatial pattern indices underlining the importance of landscape structure for species distribution. Species reacted differently in terms of both scale and landscape pattern. P. pipistrellus only responded to variables at 10 km; N. leisleri and M. emarginatus did so at two scales (5 and 10 km); whereas R. hipposideros also responded to variables at 1 km. Conclusions Our findings make it possible to tailor SDMs according to species-specific landscape pattern requirements at appropriate scales. Our approach, which can be easily extended to other taxa and different spatial scales, represents a significant step towards more effective land management planning.