Automated ecoacoustic monitoring: Evaluating classifier performance for common cuckoo (cuculus canorus) detection
Whitehead, Helen C; Davies, William J; Gashchak, Sergey; Wood, Michael D
Authors
Helen C Whitehead
William J Davies
Sergey Gashchak
Michael D Wood
Abstract
This study evaluates the effectiveness of automated bioacoustic classifiers in detecting the vocal activity of the common cuckoo (Cuculus canorus) within the Chornobyl Exclusion Zone (CEZ). Data were collected from 12 recording locations using Wildlife Acoustics Songmeter 3 (SM3) units during May 2015. Acoustic data were analysed using Kaleidoscope Pro, focusing on the left channel to avoid duplicate detections. The classifier scanned recordings for target sounds based on specified signal parameters and employed cluster analysis to categorise events. The classifier demonstrated high precision (95.1%) and a low false negative rate (0.5%), with most misclassifications due to other bird species with similar frequencies. The recall rate averaged 61%, varying from 40% to 100% across different recorders. Variations in recall rate were influenced by habitat structure, environmental noise, and distance between the target species and the recorder. This study highlights the potential of automated classifiers in ecoacoustic monitoring.
Make cities and human settlements inclusive, safe, resilient and sustainable
Protect, restore and promote sustainable use of terrestrial ecosystems, sustainably manage forests, combat desertification, and halt and reverse land degradation and halt biodiversity loss
