Random forest is the best species predictor for a community of insectivorous bats inhabiting a mountain ecosystem of central Mexico.

Bats are nocturnal animals that can be identified by recording and analysing quantitatively their echolocation calls. For this task, many studies have used both parametric and non-parametric approximations with a variety of results. This urges the necessity of developing more call libraries, that should be analysed using the different statistical approaches to test their performance. This could be relevant in countries holding high biodiversity where the knowledge of the variation in the call structure among species is still scarce. We constructed and validated a call library from bats inhabiting a mountain ecosystem of central Mexico using the Linear Discriminant Function, Artificial Neural Network and Random Forest approaches. We recorded and analysed 2,325 pulses from 114 individuals and 16 bat species of the families Vespertilionidae, Mormoopidae, Molossidae, and Natalidae. The Random forest model (81.3%) was the better species predictor over the artificial neural network and the discriminant function analysis (69% and 62.1%, respectively). Our work is one of the few attempts to do this exercise that has been conducted in Mexico. The library can be useful as a starting point of research in other regions of the highlands in central Mexico where the information is still scarce. [ABSTRACT FROM AUTHOR]