Data for publication "Comparing machine-learned and engineered acoustic features for estimating animal species richness in the Cape Region of South Africa"
Domain:
environment and energygeospatialRecord type:
This archive includes sound clips (.wav files) of bird, frog and insect vocalizations for 70 species and soundscape components (Anthropophony, Biophony, Geophony, Interference - ABGI) in the Greater Cape Floristic Region (GCFR) of South Africa. These data were used for training and validating convolutional neural network (CNN) models for species detection (CapeNet model) and ABGI classes (ABGI AI model). Details on these data are explained in the paper by Clark et al. (2026) titled "Comparing machine-learned and engineered acoustic features for estimating animal species richness in the Cape Region of South Africa". These data are available for use without restrictions, with no warranty on data quality or utility for a given application. We request that any work that does use these data cite the Clark et al. (2026) paper.
Clark, M. L., Salas, L., Lee, A. T. K., Seymour, C., Turner, A., Couldridge, V., Schackwitz, W., Buainain, N., Bezerra, C. P. de A., Schneider, F. D., & Ferraz, A. (2026). Comparing machine-learned and engineered acoustic features for estimating animal species richness in the Cape Region of South Africa [Manuscript submitted for publication].
Associated code for training CNN models, performing inference, and applying post-classification corrections can be found in the GitHub archive https://github.com/mateocla…
Raw acoustic data from the BioSoundSCapes project are available at: https://doi.org/10.3334/ORN…
If you use the recordings, please cite this data paper: Turner, A.A., Clark, M.L., Salas, L. et al. BioSoundSCape: A bioacoustic dataset for the Fynbos Biome. Sci Data 12, 1432 (2025). https://doi.org/10.1038/s41…
These data were collected as part of the BioSoundSCapes project, funded by NASA’s under awards 80NSSC22K0830 and
80NSSC23K1459. This work was conducted under Permit numbers CRC/2024-2025/005--2023/V1 from South African National Parks and CN32-87-27034 from CapeNature.
----------------------------
This repository includes the following archives:
- abgi_wav_data.zip - clipped sound data used in training the ABGI AI. See the GitHub for the trained model.
- capenet_wav_data.zip - clipped sound data used in training the CapeNet species detection CNN. See the GitHub for the trained model.
- data.zip - large csv files used by analytical R code in the GitHub.
- models.zip - fitted Random Forest models
- umap_tpd_modeling.zip - BirdNET embeddings PCA and UMAP statistics; TPD files