•  
  •  
 

ESTIMATING THE ABUNDANCE OF COMMON BIRD SPECIES USING THE DENSITY OF SONGS PRESENT ON SOUND RECORDINGS**

Abstract

Avian monitoring is in a state of transition, with traditional in-person point-counts being slowly replaced with automated monitoring methods using Autonomous Recording Devices (ARDs). This presents many new challenges, including how to estimate abundance and density of birds using only single, un-arrayed, sound recorders. With traditional point-counts, the number and location of each bird is recorded, along with observation times, allowing estimates of abundance or density to be calculated. Bird number and sampling area are two key values needed to estimate density, but they are not directly provided by sound recordings. Some methods, including cue-counting and sound-pressure distance-estimation, have been proposed to estimate these values indirectly, but validation of these methods has been limited so far. We used cue-counting, the frequency of songs, to estimate the abundance of singing males. In the summer of 2025, we conducted paired point-counts at 31 random locations in Chattahoochee River National Recreation Area. During 12-minute intervals, we recorded the species, number, timing, and estimated distances for singing birds. Concurrently, we collected a sound recording for each survey. We developed a Bayesian model utilizing capture-recapture and distance-sampling to estimate abundance and density for several common bird species at each location using the point-count data. We manually annotated a random sample of 50 recording segments in Raven Pro, labeling all bird songs present. We developed a custom BirdNET model and validated it by calculating recall and precision on the manually annotated files. We then used this trained model to count the density of songs present on the full set of recordings. The song density from the recordings was then compared to number of birds observed during the point-counts for each site. Multiple methods of cue-counting analysis were tested to determine which best predicted the observed abundance of birds singing males.

Acknowledgements

Thank you to the National Park Service for their generous provision of funding, equipment, software, and logistical support, and to the Georgia Ornithological Society for funding to support this fieldwork.

This document is currently not available here.

Share

COinS