Automatic Identification of “Alytes obstetricans” Calls

Abstract

This article focuses on the procedure to automatically identify Alytes obstetricans vocalisations, an anuran species that emits calls when mating. In Luxembourg, 37 sites where the species was historically or recently recorded were monitored using automated sound recording systems (ARS) during spring and summer 2021. The huge amount of audio recordings collected were processed using scikit-maad, an open-source Python package dedicated to the quantitative analysis of environmental audio recordings. Our results show that the SVC method at high resolution presents the best results to predict A. obstetricans calls. With the help of the MAAD package, we were able to build several models that detect A. obstetricans calls with high efficiency, which seems to be a promissing alternative method to monitor the common midwife toad in Luxembourg.