Unveiling nonlinear effects of drivers on cropland use function trade-offs and synergies through XGBoost-SHAP: Insights from the Loess Plateau
-
-
Abstract
Understanding the trade-offs and synergies among cropland use functions (CUFs) is essential for balancing food production, ecological restoration, and rural livelihoods. On the Loess Plateau, where terrace agriculture, severe soil erosion, and rural depopulation coexist, cropland serves as a critical interface linking human activities and ecological processes. However, the nonlinear effects of environmental, socioeconomic, and institutional drivers on CUF interactions remain poorly understood, particularly over long-term time series. This study integrated correlation analysis, XGBoost-SHAP, and Self-Organizing Maps to investigate the spatiotemporal evolution of three major CUFs—agricultural production (AP), social security (SS), and ecological maintenance (EM)—and to identify the nonlinear drivers of their trade-offs and synergies. The results revealed significant spatiotemporal heterogeneity in CUF relationships, with synergistic interactions dominating in the southeastern Loess Plateau and trade-offs concentrated in the northwestern regions. XGBoost-SHAP analysis indicated that slope, precipitation, and population density were the primary nonlinear drivers of CUF interactions, with threshold effects observed for each. The Self-Organizing Maps further identified distinct clusters of CUF combinations, corresponding to different stages of rural development and ecological restoration. These findings provide a nuanced understanding of how CUFs interact under complex mountain conditions and offer spatially explicit insights for land-use planning and policy formulation aimed at enhancing food security, ecological sustainability, and rural well-being on the Loess Plateau.
-
-