Passive acoustic monitoring of cave-dwelling bats with a sonotype classifier

Abstract

The aim of this study in the Republic of Congo was to test a new method to conduct a quantitative study of cave-dwelling bat communities using acoustics. This area is characterised by limited knowledge of the acoustic repertoire of bats. Over a period of 19 months for a total of 398 nights, we carried out 11 individual capture sessions to build a species inventory, record reference sounds and set up a passive acoustic monitoring protocol (PAM) at the exit of two caves. We used the Tadarida automatic sonotype classifier to classify acoustic vocalisations of bats in caves. For this, we enhanced the Tadarida classifier library with reference recordings of bats captured in both caves. Due to the acoustic overlap of several species, we grouped them into five distinct acoustic units using a posteriori classification based on four distinct parameters: call shape, acoustic frequency, a harmonics index and the identification probability. A random manual control stratified by sonotype showed an accuracy ranging from 82% to 98% depending on the group. This study is the first local application of a bat sonotype classifier designed and developed to function globally. It confirms the possibility of undertaking quantitative assessments of bat communities with relatively minimal effort, even in areas with limited knowledge of their acoustic repertoires.