Sustainability of Oilseed Production Under Climate Change in the Canadian Prairies: A Machine-Learning-Enhanced Land Suitability Assessment for Canola and Flax

Oilseed production in the Canadian Prairies is highly sensitive to interacting climatic, soil, and landscape constraints, which has important implications for long-term agricultural sustainability under climate change. Canola and flax are economically significant Prairie oilseeds with distinct sensitivities to temperature and moisture variability; however, region-wide, crop-specific suitability assessments remain limited. In this study, we developed a machine-learning-enhanced Land Suitability Rating System (LSRS) framework to evaluate both historical and projected suitability for canola and flax across the Prairie agricultural region. Random Forest regression models were trained using spatial crop-density data in combination with observed climate normals and updated soil and terrain variables. Variable-importance scores were used to derive empirical climate–soil weighting, resulting in climate contributions of 70% for canola and 75% for flax. Model performance demonstrated strong internal agreement with observed spatial patterns (R2 ≈ 0.99). Canola suitability was primarily associated with mid- to late-season precipitation and maximum temperatures, whereas flax suitability was more strongly influenced by growing-season minimum temperatures. Future suitability was simulated using CanDCS-M6 downscaled CMIP6 projections under SSP2-4.5, SSP3-7.0, and SSP5-8.5. Projected warming and regionally variable precipitation changes produced spatially heterogeneous shifts in suitability, including localized gains in northern areas and increased climatic risk across parts of the southern Prairies under higher-emission scenarios. The proposed framework integrates empirical machine-learning insights with an interpretable suitability rating system, providing a scalable and policy-relevant tool to support climate-informed adaptation and sustainable oilseed production planning.