Identification of twenty‐two bat species (Mammalia: Chiroptera) from Italy by analysis of time‐expanded recordings of echolocation calls

TLDR

Echolocation calls of 22 Italian bat species were characterized by their spectral and temporal features. The authors used time‑expanded recordings of 950 bats and applied quadratic discriminant function analysis, developing separate models for FM and FM/QCF call types. The discriminant models achieved up to 96% accuracy, with a genus‑level model reaching 94% and species‑level models ranging from 71% to 82%, enabling reliable identification for habitat surveys.

Abstract

Abstract Spectral and temporal features of echolocation calls emitted by 22 bat species from Italy (three rhinolophids, 18 vespertilionids and the molossid Tadarida teniotis ) are described. Time‐expanded recordings of calls from 950 bats of known identity were examined. Rhinolophus ferrumequinum , R. hipposideros , R. euryale and T. teniotis could be identified by measuring the call frequency of highest energy (FMAXE). Quadratic discriminant function analysis with cross‐validation was applied to calls from the remaining 18 species. A function based on start frequency (SF), end frequency (EF), FMAXE and duration (D) provided a correct overall classification of approximately 82%. A classification model at genus level that also comprised middle frequency (MF) and inter‐pulse interval (IPI) reached 94% correct classification. Two separate discriminant functions were devised for species emitting FM (frequency modulated) and FM/QCF calls (i.e. calls consisting of a frequency‐modulated component followed by a terminal part whose frequency is almost constant) respectively. The former function included SF, EF, FMAXE and D and provided an overall classification rate of 71%; the latter comprised EF, MF, D and IPI, and reached 96%. The functions may be applied to bat habitat surveys in southern Italy since they cover most of the species occurring in the area.