Temporal patterns in Malaysian rainforest soundscapes demonstrated using acoustic indices and deep embeddings trained on time-of-day estimation

Rapid urban development impacts the integrity of tropical ecosystems on broad spatiotemporal scales. However, sustained long-term monitoring poses significant challenges, particularly in tropical regions. In this context, ecoacoustics emerges as a promising approach to address this gap. Yet, harnessing insights from extensive acoustic datasets presents its own set of challenges, such as the time and expertise needed to label species information in recordings. Here, this study presents an approach to investigating soundscapes: the use of a deep neural network trained on time-of-day estimation. This research endeavors to (1) provide a qualitative analysis of the temporal variation (daily and monthly) of the soundscape using conventional ecoacoustic indices and deep ecoacoustic embeddings, (2) compare the predictive power of both methods for time-of-day estimation, and (3) compare the performance of both methods for supervised classification and unsupervised clustering to the specific recording site, habitat type, and season. The study's findings reveal that conventional acoustic indices and the proposed deep ecoacoustic embeddings approach exhibit overall comparable performance. This article concludes by discussing potential avenues for further refinement of the proposed method, which will further contribute to understanding of soundscape variation across time and space.