Integrating ecological niche modeling and machine learning to optimize alfalfa variety deployment for climate-resilient forage systems

Rising demand for livestock products intensifies pressure on forage systems globally, yet continuous climate change threatens alfalfa ( Medicago sativa L.) productivity and stability due to mismatches between variety traits and local environments. Here we coupled MaxEnt ecological niche modeling with Random Forest regression to assess current and future (2021−2100) climatic suitability and yield potential for 27 major alfalfa varieties across China under three shared socioeconomic pathways (SSP1–2.6, SSP2–4.5, SSP5–8.5). Models achieved excellent predictive accuracy (mean AUC = 0.81–0.96). Under future scenarios, 11 varieties exhibited consistent yield increases (0.68–66.70% above current levels), while 12 showed persistent declines (1.75–39.67% losses), a bifurcation reflecting contrasting adaptive strategies. Optimized variety deployment in three representative regions achieved yield gains of 8.7–35.7%, translating to additional protein output of 0.44–1.46 t ha −1 and enhanced carbon sequestration of 4.30–14.28 t ha −1 . Our findings demonstrate that data-informed variety deployment can simultaneously enhance forage productivity and ecosystem services under climate change, providing a framework that is applicable to the same crop types in other regions and moderately transferable to neighboring countries with similar ecological conditions.