Automated acoustic recorders identify higher occupancy for at-risk birds compared to point counts

Background/Question/Methods

: In the eastern US, widespread timber harvests and herbivore overpopulation have led to a loss of ideal forested habitat for eastern forest birds. This habitat loss has affected species such as the Wood Thrush (Hylocichla mustelina), Cerulean Warbler (Setophaga cerulea), and Golden-winged Warbler (Vermivora chrysoptera), whose populations have all declined. Although large-scale forest management is being implemented across eastern forests to restore preferred habitats for these birds, traditional point count surveys have proven costly and time consuming for determining restoration success at large scales. Here, we compare Wood Thrush, Cerulean Warbler, and Golden-winged Warbler site occupancy within eastern forest habitats using both traditional point count surveys and a novel method utilizing automated acoustic recordings. We deployed AudioMoth acoustic recorders and performed point counts within two sites containing 190 locations from mid-May to mid-July 2021 within the central Appalachian region of Pennsylvania. To determine naïve occupancy using acoustic recordings, we developed an automated birdsong classifier using OpenSoundscape, a free and open-source Python package for analyzing bioacoustic recordings. We manually listened to the top scoring clips generated from the classifier to verify focal species song presence and compared the resulting occupancy results to naïve occupancy reported from traditional point counts.

Results/Conclusions

: For Wood Thrushes, point counts at site 1 found 131/166 (78.9%) locations naïvely occupied compared to 144/153 (94.1%) for automated acoustic recorders and point counts at site 2 found 8/24 (33.3%) locations naïvely occupied compared to 11/22 (50.0%) for automated acoustic recorders. For Cerulean Warblers, point counts at site 1 found 18/166 (10.8%) locations naïvely occupied compared to 56/153 (36.6%) for automated acoustic recorders and point counts at site 2 found 2/24 (8.3%) locations naïvely occupied compared to 5/22 (22.7%) for automated acoustic recorders. No Golden-winged Warblers were observed at site 1 using either method, but at site 2, point counts found 4/24 (16.7%) locations naïvely occupied compared to 2/22 (9.1%) for automated acoustic recorders. Increases in naïve occupancy by using automated acoustic recorders for Wood Thrushes and Cerulean Warblers compared to point counts may be due to differences in sampling effort. Automated acoustic recorders represent 100+ hours of continuous sampling compared to two 10-minute point counts performed during the same time period. We suggest that acoustic recorders are complementary to, and may outperform, point counts for monitoring forest birds.