EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Assigning Pyrenean Vegetation Plots to Phytosociological Alliances Using Supervised Machine Learning

Raúl Pérez Prats, Gwendolyn Peyre, Xavier Font

Journal of Vegetation Science · 2026-11 · 37 (6)

DOI: 10.1111/jvs.70188 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Vegetation Science, online erschienen am 13.11.2026. Verfasst von Raúl Pérez Prats und 2 weiteren Autor:innen. Die Arbeit nennt 55 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Aims

Artificial intelligence (AI) is becoming an increasingly important tool for describing and understanding ecological patterns. In vegetation science, supervised machine‐learning methods can learn from previously labeled plot databases and rapidly assign vegetation plots to preexisting syntaxonomic units. Here, we evaluated the ability of machine‐learning and deep‐learning techniques to assign Pyrenean vegetation plots to preexisting phytosociological alliances recorded in the Information System for the Iberian Peninsula and Macaronesian Islands (SIVIM) database. Location The Pyrenees.

Methods

We relied on 27,354 plots from the SIVIM database, organized into 167 alliances and 49 classes. We used two data representations, sparse and reduced, and ran supervised classification models with two gradient‐boosting tree algorithms, lightGBM and XGBoost, and one neural network approach, multilayer perceptron (MLP). Each assignment was done at the alliance level, upscaled to the class level and then compared with the original phytosociological assignments made by the authors of the relevés. Ward's hierarchical clustering was performed using the 1 − Rand dissimilarity computed from the final partitions to assess their similarities.

Results

Overall, the algorithms showed good performance in alliance assignment, although accuracy varied with data representations. MLP achieved both the highest (0.850) and lowest (0.795) accuracy values on sparse and reduced data, respectively. XGBoost showed the most consistent performance across data representations.

Conclusions

All three algorithms reproduced the alliance assignments recorded in SIVIM with good overall accuracy, although performance depended on the data representation. The best‐ and poorest‐assigned alliances and classes offered insight into differential algorithm performance and highlighted units whose prediction could be improved with additional data or expert examination. Our study supports the combined use of machine‐learning and expert‐based approaches for vegetation plot assignment, with promising applications from local research management to large‐scale biogeography.

Zitieren

APA

Prats, R. P., Peyre, G., & Font, X. (2026). Assigning Pyrenean Vegetation Plots to Phytosociological Alliances Using Supervised Machine Learning. Journal of Vegetation Science, 37(6). https://doi.org/10.1111/jvs.70188

Vancouver

Prats RP, Peyre G, Font X. Assigning Pyrenean Vegetation Plots to Phytosociological Alliances Using Supervised Machine Learning. Journal of Vegetation Science. 2026;37(6). doi:10.1111/jvs.70188

BibTeX
@article{Prats2026,
  author = {Prats, Raúl Pérez and Peyre, Gwendolyn and Font, Xavier},
  title = {Assigning Pyrenean Vegetation Plots to Phytosociological Alliances Using Supervised Machine Learning},
  journal = {Journal of Vegetation Science},
  year = {2026},
  volume = {37},
  number = {6},
  doi = {10.1111/jvs.70188}
}
Kostenlos starten →