EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

The evolutionary structure of acoustic learnability: a deep learning approach to neotropical birdsong

Camilo Andrés Cortés-Parra iD, Héctor J Hortúa iD, Juan Camilo Ríos-Orjuela iD

Royal Society Open Science · 04.11.2026 · 13 (11)

DOI: 10.1098/rsos.260394 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Royal Society Open Science, online erschienen am 11.11.2026. Verfasst von Camilo Andrés Cortés-Parra und 2 weiteren Autor:innen. Beteiligte Einrichtung: Maestria en Ciencia de Datos, Escuela Colombiana de Ingenieria Julio Garavito, Bogotá, Colombia. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Passive acoustic monitoring offers a scalable solution for biodiversity assessment in the Neotropics, although classifying hundreds of sympatric species in complex soundscapes remains challenging. We developed a deep learning framework for large-scale avian classification by training convolutional neural networks on recordings from 667 Neotropical bird species across northern South America. EfficientNetV2L performed best, achieving 94.48% accuracy, 94.30% macro F1-score and 0.998 macro ROC-AUC. After training, we conducted a post hoc analysis of species-level F1-scores using phylogenetically informed models (phylogenetically generalized least squares (PGLS) and phylogenetic generalized linear mixed models (PGLMM)) to test morphological, ecological and geographic predictors under phylogenetic control. Broad traits explained little interspecific variation: the best-supported PGLS accounted for approximately 2.2% of the variance, whereas the null model received the strongest support among PGLMM candidates. Geographic range size showed the most consistent negative association with F1-score, while morphological and ecological predictors added little explanatory power. Performance nevertheless showed weak but significant phylogenetic signal, and frequent confusions tended to involve more closely related species. Gradient-weighted class activation mapping and Monte Carlo Dropout provided complementary descriptions of saliency and predictive uncertainty. Overall, deep learning performed effectively for regional biodiversity monitoring and provided a comparative framework for assessing how much of the variation in acoustic classification performance could be explained by broad biological predictors.

Schlagwörter

Stichwörter der Autor:innen

acoustic classificationbioacousticsbiogeographyconvolutional neural networksexplainable AImodel uncertaintysong evolution

Autor:innen und Institutionen

  • Maestria en Ciencia de Datos, Escuela Colombiana de Ingenieria Julio Garavito, Bogotá, Colombia.
  • Grupo de Morfología y Ecología Evolutiva, Instituto de Ciencias Naturales, Universidad Nacional de Colombia, Bogotá, Colombia.
  • Escuela Colombiana de Ingenieria Julio Garavito, Bogotá, Colombia.
  • Grupo INUB-MEDITECH, Instituto de Neurociencias, Universidad El Bosque, Bogotá, Colombia.
  • Departamento de Ciencias Biológicas, Universidad de los Andes, Bogotá, Colombia.
  • Departamento de Biología, Universidad Pedagógica Nacional, Bogotá, Colombia.

ORCID-Profile: Camilo Andrés Cortés-Parra, Héctor J Hortúa, Juan Camilo Ríos-Orjuela

Zitieren

APA

Cortés-Parra, C. A., Hortúa, H. J., & Ríos-Orjuela, J. C. (2026). The evolutionary structure of acoustic learnability: a deep learning approach to neotropical birdsong. Royal Society Open Science, 13(11). https://doi.org/10.1098/rsos.260394

Vancouver

Cortés-Parra CA, Hortúa HJ, Ríos-Orjuela JC. The evolutionary structure of acoustic learnability: a deep learning approach to neotropical birdsong. Royal Society Open Science. 2026;13(11). doi:10.1098/rsos.260394

BibTeX
@article{CortsParra2026,
  author = {Cortés-Parra, Camilo Andrés and Hortúa, Héctor J and Ríos-Orjuela, Juan Camilo},
  title = {The evolutionary structure of acoustic learnability: a deep learning approach to neotropical birdsong},
  journal = {Royal Society Open Science},
  year = {2026},
  volume = {13},
  number = {11},
  doi = {10.1098/rsos.260394}
}
Kostenlos starten →