Modeling climate-change impacts on suitable habitats of Myristica yunnanensis

Myristica yunnanensis is an endangered, China‐endemic tree of medicinal and spice value whose habitat is threatened by climate change and human disturbance. We provide the first quantitative habitat‐suitability assessment to guide conservation. We compiled 21 vetted occurrences and environmental predictors (19 bioclimatic variables, ASTER GDEM topography, HWSD v2.0 soils, and GLC-SHARE land use) across China, focusing interpretation on Yunnan. We modelled current suitability and future projections (2050s, 2070s) under SSP126, SSP245, and SSP585 using a MaxEnt model and a biomod2 ensemble (GLM, GAM, RF, GBM), with cross-validation. Model performance was evaluated with AUC; variable importance used jackknife/permutation; and robust areas were identified by overlap between binarized MaxEnt and ensemble outputs. All models performed well (AUC > 0.90); MaxEnt performed best (AUC = 0.997) but produced more conservative suitable areas, whereas the ensemble predicted a broader extent. Temperature-related variables dominated importance: Temperature Seasonality (bio3), Max Temperature of Warmest Month (bio4), and Mean Temperature of Driest Quarter (bio9). Under current climate, unsuitable area predominates. Under SSP126/245, suitable area slightly contracts; under SSP585, moderate and high suitability expand, most strongly in the 2070s. Overlap mapping identifies stable core refugia concentrated in southeastern Yunnan and southern Hainan. Our results clarify thermal controls on M. yunnanensis distributions, quantify model-dependent uncertainty, and translate consensus into spatial priorities. The identified refugia provide actionable targets for habitat protection, restoration, and assisted management under warming. The workflow—integrating single-model and ensemble SDMs with consensus overlap—offers a transferable template for conservation planning of other threatened tropical–subtropical plants.