Abstract
Understanding the circadian rhythm of the calling behaviour of wild animals can guide efforts to monitor and conserve rare and endangered species using sound. Here, we use passive acoustic monitoring to investigate the vocal behaviour of the crested argus (Rheinardia ocellata) in Kon Chu Rang Nature Reserve, Gia Lai Province, Vietnam. We had three main objectives: (i) to investigate the performance of BirdNET transfer learning for automated detection of crested argus calls; (ii) to investigate the environmental predictors of crested argus calling; and (iii) to qualitatively investigate seasonal patterns of calling. We recorded continuously for 4-5 days at 40 recording points in 2021, and at 30 points in 2023. We also recorded the calls of crested argus at four fixed points from 2022 to 2023 to explore patterns of seasonal variation. For automated detection, we found acceptable performance with only 30 high-quality training samples (F1 score = 0.70). Our top model for calling during the 24 h period only included the time category, and we found that there was peak calling activity at dawn and dusk. We found peak calling activity during March and April. Our findings can contribute to planning effective monitoring of the critically endangered crested argus.This article is part of the theme issue 'Acoustic monitoring for tropical ecology and conservation'.
