Evidence‐based management of invasive trees: Using ensemble species distribution models to prioritise control in Rwandan restored montane forests

Biological invasions increasingly threaten tropical montane forests, yet the environmental drivers shaping invasion risk in African Afromontane landscapes remain poorly understood. Gishwati‐Mukura Biosphere Reserve in Rwanda exemplifies this challenge, where invasive species threaten forest recovery and native species conservation. We evaluated patterns of two invasive species ( Acacia melanoxylon and Acacia mearnsii ), and an exotic non‐invasive Alnus acuminata to identify species‐specific invasion risk and the environmental drivers shaping their establishment.

Vegetation surveys were combined with ensemble species distribution models integrating climatic, soil, hydrological, vegetation indices and disturbance‐related predictors. Model performance, spatial evaluation metrics and variable importance analyses were used to quantify habitat suitability and identify dominant ecological drivers.

Field surveys revealed strong differences in species dominance between two forests. A. melanoxylon strongly dominated plots in Gishwati Forest, whereas only A. mearnsii was consistently detected in Mukura Forest. Habitat suitability modelling revealed a high invasion risk from A. melanoxylon in Gishwati, whereas A. mearnsii exhibited a lower and more spatially restricted suitability pattern, particularly within Mukura Forest. A. acuminata showed consistently low suitability across both forests, suggesting strong environmental limitations on its establishment. Ensemble modelling improved spatial predictions and reduced residual spatial autocorrelation relative to individual models, highlighting its reliability for predicting forest invasion risk. Invasion drivers varied between the two forests. In Gishwati Forest, invasion risk was primarily associated with variables related to disturbance and soil structure, including bulk density and coarse fragment content, consistent with ruderal invasion strategies. In Mukura Forest, however, invasion patterns were influenced by climatic and hydrological variables, particularly warm‐quarter precipitation and soil moisture, suggesting the importance of environmental filtering in shaping species establishment.

Practical implication. Our findings highlight the need for forest‐ and species‐specific management strategies, guided by local environmental drivers of invasion risk. In broader application, ensemble SDMs can inform invasion management and restoration planning across Rwanda and comparable montane forest systems.