EUVIMEDEUVIMED entdecken
Crossref

Disentangling the invasion landscape: Spatially-explicit explainable machine learning reveals the heterogeneous drivers of Solenopsis invicta distribution in China

Lin Chen, Yu Liu

Journal of Environmental Management · 2026-01 · 397 · 128366

DOI: 10.1016/j.jenvman.2025.128366 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Environmental Management, 2026-01. Verfasst von Lin Chen und 1 weiteren Autor:innen. Beteiligte Einrichtung: College of Horticulture and Forestry Sciences, Huazhong Agricultural University, Wuhan, 430070, China; Hubei Engineering Technology Research Center f…. Hauptthemen laut MEDLINE: Machine Learning, Introduced Species, Ants. Die Arbeit wurde bisher 4-mal zitiert (Crossref) und nennt 74 Literaturangaben. Gefördert von: Huazhong Agricultural University.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

BACKGROUND: Solenopsis invicta, the red imported fire ant, is a globally significant invasive species causing severe ecological and economic damage. Its recent rapid expansion in China necessitates accurate risk assessment tools to guide management. While machine learning models are powerful predictors, their "black box" nature often limits their utility for developing targeted control strategies. METHODS: We compiled 1,791 verified occurrence records of S. invicta in China and a suite of 12 environmental variables. We evaluated six machine learning models using a rigorous 5-fold spatial cross-validation to account for spatial autocorrelation. The optimal model (CatBoost) was interpreted using post-hoc explainability techniques, including SHapley Additive exPlanations (SHAP) and the novel GeoShapley method, to identify key drivers and analyze their complex, spatially-varying effects. RESULTS: The optimal model accurately predicted high habitat suitability in the subtropical and tropical regions of southern and southeastern China (AUC = 0.90 ± 0.06). The three most influential drivers were precipitation of the driest month (PDM), gross primary productivity (GPP), and population density (PD). SHAP-based analyses revealed critical non-linear relationships and ecological tipping points, such as a sharp decline in suitability when PDM exceeds 19.5 mm. Critically, the GeoShapley analysis demonstrated significant spatial heterogeneity in driver importance; for example, the negative impact of high PDM was most pronounced in the humid southeastern coast, while the importance of anthropogenic factors was highest along major transportation corridors extending into central China. CONCLUSION: Our study provides a high-resolution, robustly validated habitat suitability map for S. invicta in China. The novel application of spatially-explicit explainable machine learning moves beyond traditional risk mapping by identifying where specific environmental and anthropogenic drivers are most influential. This provides a powerful, nuanced framework for developing targeted, evidence-based, and spatially-differentiated management strategies to combat the ongoing invasion.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Machine LearningIntroduced SpeciesAntsChinaAnimalsEcosystemFire Ants

Stichwörter der Autor:innen

Anthropogenic factorsExplainable machine learningInvasive speciesSolenopsis invictaSpatial heterogeneity

Autor:innen und Institutionen

  • College of Horticulture and Forestry Sciences, Huazhong Agricultural University, Wuhan, 430070, China; Hubei Engineering Technology Research Center for Forestry Information, Huazhong Agricultural University, Wuhan, 4300…

Förderung

  • Huazhong Agricultural University

Zitieren

APA

Chen, L., & Liu, Y. (2026). Disentangling the invasion landscape: Spatially-explicit explainable machine learning reveals the heterogeneous drivers of Solenopsis invicta distribution in China. Journal of Environmental Management, 397, 128366. https://doi.org/10.1016/j.jenvman.2025.128366

Vancouver

Chen L, Liu Y. Disentangling the invasion landscape: Spatially-explicit explainable machine learning reveals the heterogeneous drivers of Solenopsis invicta distribution in China. Journal of Environmental Management. 2026;397:128366. doi:10.1016/j.jenvman.2025.128366

BibTeX
@article{Chen2026,
  author = {Chen, Lin and Liu, Yu},
  title = {Disentangling the invasion landscape: Spatially-explicit explainable machine learning reveals the heterogeneous drivers of Solenopsis invicta distribution in China},
  journal = {Journal of Environmental Management},
  year = {2026},
  volume = {397},
  pages = {128366},
  doi = {10.1016/j.jenvman.2025.128366}
}
Kostenlos starten →