Skip to main content
Dryad logo

Automated bird sound classifications of long-duration recordings produce occupancy model outputs similar to manually annotated data

Citation

Cole, Jerry; Michel, Nicole; Emerson, Shane; Siegel, Rodney (2022), Automated bird sound classifications of long-duration recordings produce occupancy model outputs similar to manually annotated data, Dryad, Dataset, https://doi.org/10.5061/dryad.x95x69pkr

Abstract

Occupancy modeling is used to evaluate avian distributions and habitat associations, yet it typically requires extensive survey effort because a minimum of three repeat samples are required for accurate parameter estimation. Autonomous recording units (ARUs) can reduce the need for surveyors on site, yet ARUs utility were limited by hardware costs and the time required to manually annotate recordings. Software that identifies bird vocalizations may reduce expert time needed, if classification is sufficiently accurate. We assessed the performance of BirdNET – an automated classifier capable of identifying vocalizations from >900 North American and European bird species – by comparing automated to manual annotations of recordings of 13 breeding bird species collected in northwestern California. We compared the parameter estimates of occupancy models evaluating habitat associations supplied with manually annotated data (9 min recording segments) to output from models supplied with BirdNET detections. We used three sets of BirdNET output to evaluate the duration of automatic annotation needed to approach manually annotated model parameter estimates: 9-min, 87-min, and 87-min of high-confidence detections. We incorporated 100 3-sec manually validated BirdNET detections per species to estimate true and false positive rates within an occupancy model. BirdNET correctly identified 90% and 65% of the bird species a human detected when data were restricted to detections exceeding a low or high confidence score threshold, respectively. Occupancy estimates, including habitat associations, were similar regardless of method. Precision (proportion of true positives to all detections) was >0.70 for 9 of 13 species, and a low of 0.29. However, processing of longer recordings was needed to rival manually annotated data. We conclude that BirdNET is suitable for annotating multispecies recordings for occupancy modeling when extended recording durations are used. Together, ARUs and BirdNET may benefit monitoring and, ultimately, conservation of bird populations by greatly increasing monitoring opportunities.   

Methods

We collected acoustic recordings from 44 locations within Carnegie State Vehicular Recreation Area in California, USA (37.6263°N, 121.5536°W). The recordings were collected using Wildlife Acoustics SongMeter 4 recording units - see text for more details about sampling duration. Of the 44 recording locations we only analyzed 34 in this paper because the remainder contained too much wind noise to be usable. In this dataset we include manual annotations (an expert observed listened to 9-min of a days long recording from each site), automated annotations (of >260 min of recordings from each site) generated by processing recordings with the BirdNET classifier software (https://github.com/kahst/BirdNET), and output of 4 occupancy models. We have also included R code for running analyses using the data the we provided to reproduce the results reported in our manuscript. Recording files were not uploaded due to the sheer space required to store such a dataset.

Usage Notes

The usage details for files are included in the comments of R code files included in this repository. Data files are supplied inside a zipped folder (Data.zip) that contains a subfolders with files needed for reproducing results in the text. All files included in the repository with the .R suffix are R scripts for running occupancy models and generating data summaries. See README file for more details.