ABSTRACT Arid mountain regions face rising ecological risk and mismatched carrying capacity under climate change and human activities. However, an integrated pathway linking diagnosis, scenario analysis, attribution, and governance remains scarce. This study develops a closed‐loop framework that integrates dual‐axis diagnosis of Landscape Ecological Risk (LER) and Ecological Carrying Capacity (ECC), scenario zoning with the Patch‐generating Land Use Simulation (PLUS) model, and nonlinear attribution using eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanations (SHAP). From 2000 to 2020, risk expanded from isolated patches to continuous belts along transport corridors and urban frontiers. Meanwhile, mountain‐core carrying capacity stayed stable or improved, and the share of moderately high to high risk rose from 0.43% to 2.82%, underscoring a pronounced LER–ECC mismatch. Four‐quadrant zoning shows the conservation zone consolidated dominance, with a net increase of 2.37%. Frontier belts were dominated by restoration‐to‐cultivation transfers, and pressures continued to rise. Projections for 2030 and 2040 show an expanding conservation zone, contracting cultivation and control zones, and a 132.64% increase in restoration demand. These shifts redirect governance priorities toward oasis frontiers and transport corridors. XGBoost–SHAP attribution reveals that nonlinear interactions among land use, accessibility, and climate create high‐ECC, high‐LER frontier belts, marking them for intervention. Integrating the LER–ECC framework with PLUS‐based zoning and SHAP‐based driver analysis, this study connects diagnosis, projection, and management, advancing ecological understanding in arid mountain systems and offering a practical template for governance.
Li et al. (Tue,) studied this question.