Dual-dimensional evaluation and zoning of cotton planting suitability based on the fusion of climate, soil, and topography data

To explore the prediction performance of multivariate data on cotton planting suitability and to support planting zone planning in Xinjiang, China, a dual-dimensional evaluation framework integrating climatic, soil, and topographical data is developed. This framework combines an XGBoost-based yield potential model with a MaxEnt-based ecological potential model, to overcome the limitations of single-dimensional evaluations. Results show that both models exhibit high prediction performance. The Coefficient of Determination ( R 2 ) and Root Mean Square Error (RMSE) of the yield potential model are 0.79 and 131.34 kg/ha, respectively. The Area Under the Curve (AUC) value of the ecological potential model is 0.91. Furthermore, through SHapley Additive exPlanations (SHAP) interpretability analysis and jackknife test, Accumulated Evapotranspiration (AET), Gravel Content (GC), and Terrain Slope (TS) were identified as the dominant influencing variables among climate, soil, and topography variables, respectively. Subsequently, the region was classified into four categories using the Multi-Criteria Decision Analysis (MCDA) framework: core planting zones, priority planting zones, restricted planting zones, and harsh planting zones. The most suitable areas for cotton planting are primarily distributed along both slopes of the Tianshan Mountains. These findings provide actionable guidance for cotton production as well as practical insights and reference for promoting scientific planning and sustainable management of cotton planting in arid regions.