Characterization of soundscapes with acoustic indices and clustering reveals phenology patterns in a subtropical rainforest
Long-term biological and phenological monitoring has become essential for conservation in the face of rapid climate change. In this study, we utilized long-term passive acoustic recording data. We employed a combination of 14 features extracted from acoustic indices and unsupervised clustering metho...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-02-01
|
| Series: | Ecological Indicators |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1470160X2500055X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Long-term biological and phenological monitoring has become essential for conservation in the face of rapid climate change. In this study, we utilized long-term passive acoustic recording data. We employed a combination of 14 features extracted from acoustic indices and unsupervised clustering methods to classify the soundscapes of Taiwan’s subtropical rainforests. Our results demonstrated that in environments with complex soundscapes, this approach effectively distinguished predominant acoustic elements, including cicadas, orthopterans, rain, and frogs, constituting more than 10–20% of the total audio recordings, and identified smaller yet significant groups, such as avian dawn choruses, accounting for approximately 2% of the recordings. The clustering results enabled the description of dynamic changes in the soundscape throughout the year. In the subtropical rainforest, rain and wind affected the soundscape from October to March, whereas bird songs were prominent only in the early mornings from February to May, which were subsequently replaced by cicada calls that continued until late August. The nocturnal soundscape was dominated by frog calls and orthopteran stridulations in the aquatic and forest habitats. Correlations among the vocal activities of several representative groups, temperature, and rainfall were found. Our study confirms that acoustic indices can extract meaningful ecological features, and unsupervised algorithms offer valuable insights into biodiversity exploration data-scarce regions. The combination of these methods has led to the development of non-species-specific soundscape classification, which not only facilitates the monitoring of phenological dynamics across multiple biological groups in the face of climate change but also lays the foundation for further exploration of key taxa. |
|---|---|
| ISSN: | 1470-160X |