Environmental characterization of microhabitats used by amphibians in the Tensift region of Morocco: An explanatory assessment using Artificial Neural Networks

Abstract

An adequate understanding of the relationship between amphibians and their habitat has been among the main challenges in herpetology in recent decades, particularly given the role of global change in the rapid declines of this group worldwide. Using the Artificial Neural Networks approach (ANN), we examined the environmental factors determining the occurrence of amphibians in the aquatic ecosystems in Tensift region of Morocco. We applied this modeling technique to 14 environmental factors and the presence of amphibian species collected from 40 sites. The results showed that the ANN is a useful approach to evaluate the effects of habitat factors on species occurrence. The model correctly classified all species with high performance. The best result was obtained for Bufo spinosus data, with a recognition percentage of 93.6% and a prediction performance of 99.4%. Of all factors studied, altitude was key in explaining the species distribution and richness, followed by hydroperiod and conductivity, for almost all species. The importance of other factors varied according to species. Principal Component Analysis differentiated a community composed by three species of Bufonidae (Bufotes boulengeri, Sclerophrys mauritanica and Barbarophryne brongersmai) that are close to Hyla meridionalis, while Bufo spinosus, Discoglossus scovazzi and Pelophylax saharicus were influenced by other environmental factors. The results provide important new information that will support conservation decision making for the protection of amphibian populations and their habitats in the studied region