Regional distribution patterns can predict the local habitat specialization of arachnids in heterogeneous landscapes of the Atlantic Forest

Abstract

Aim: This study formally evaluates the ability of three models to use geographical data on species distribution to predict the habitat use patterns of species in heterogeneous landscapes. Location: Species and habitats in the Brazilian Atlantic Rain Forest were investigated. Methods: Based on empirical data on harvestmen and scorpions, we estimated the strength of species association with preferred habitat and classified them as habitat generalists or habitat specialists. We compared these empirical results with predictions made using data on species range size (model 1), species occurrence in biomes (model 2) and species occurrence in habitats within the biomes (model 3). Results: We used 1,278 records of eight harvestman and two scorpion species that had specific determination and enough sampling numbers to allow safe identification of habitat specialization. We observed the following: (1) the extension of species occurrence did not influence the strength of species–habitat association (estimated by IndVal), which led us to reject model 1; (2) species habitat specialization derived from occurrences in biomes was 60% coincident with the classification derived from empirical data. This value is not different enough from the value expected by chance for these data, which also led us to reject model 2; and (3) species classification derived from secondary data about the habitats used had a significant coincidence of 80% with the empirical classification, which led us to accept model 3. Main conclusions: For correct classification of species habitat specialization using secondary distributional data, we recommend that future studies consider using the most accurate information available on the habitats used by species. Especially for megadiverse and understudied groups, information about habitats used is not easy to obtain, so it is important for researchers and institutions to register and disseminate this information, which could support many other studies.