Doi: https://doi.org/10.13157/arla.70.2.2023.sc1
Authors: Cristian PÉREZ-GRANADOS
E-mail: cristian.perez@ua.es
Published: Volume 70.2, July 2023. Pages 257-269.
Language: English
Keywords: automated detection, automated software, autonomous recording unit, bird monitoring, bird recognition, convolutional neural network and passive acoustic monitoring
Summary:
Bird vocalisations, like any other acoustic signals, attenuate over distance, and therefore their structure degrades progressively. Such degradation may have an impact on the ability of automated signal recognition software, to detect and correctly identify bird vocalisations. BirdNET is a recently launched automated bird song recogniser commonly employed by researchers and the public. However, few studies have assessed its performance and our current knowledge about how BirdNET performance may vary over distance or with species is very limited. I aimed to evaluate whether BirdNET's ability to correctly identify bird vocalisations of three bird species varied over distance or according to recorder type and target species using a playback broadcast from 10 to 150m away. BirdNET's ability to correctly identify bird songs varied among species and generally decreased over distance but did not vary among recorder types. Overall BirdNET recall rate, defined as the percentage of vocalisations detected, and correctly identified, by the software, was 59.9% (499 vocalisations correctly identified of 840 vocalisations broadcast). A significantly higher number of vocalisations were correctly identified when broadcast at 50m or closer (mean recall rate of 92.2%), when compared to vocalisations broadcast farther than that distance (mean recall rate of 34.9%). Recall rate was also significantly higher for the Grasshopper Sparrow and the Hooded Warbler, when compared to the Gray Vireo. The number of misclassifications varied over distances and did not follow a linear pattern. This study provides valuable information that may contribute to improved surveys and for expanding the use of BirdNET for surveying bird communities using passive acoustic monitoring.
Supplementary Material:
Downloadable 1
