Non-Linear Ecosystem Service Responses Under Coupled Climate-Land-Use Stress in Plateau Urban Agglomerations: Threshold Identification for Regional Sustainable Development

Global climate change and rapid urbanisation exert coupled disturbances on plateau urban agglomerations, where rugged terrain and land-use conversion exacerbate ecological fragmentation and function loss. Most existing research relies on linear analytical frameworks and cannot capture nonlinear ecological shifts, multi-factor coupling feedbacks or stability-transition thresholds for plateau regions; multi-scenario quantitative ecological assessments also remain limited. Taking the Central Yunnan Urban Agglomeration as a case, this study aimed to identify nonlinear ecosystem-service responses and statistically derived ecological breakpoints. Three Shared Socioeconomic Pathway (SSP) scenarios (SSP119, SSP245, SSP585) were integrated into a modelling chain: the Patch-generating Land-Use Simulation (PLUS) model projected 2030 land-use patterns (hindcast: 92.8% overall accuracy (OA), Kappa = 0.8887); the Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model quantified four key ecosystem services to compute the Multiple-Ecosystem-Services Landscape Index (MESLI); and eXtreme Gradient Boosting-SHapley Additive exPlanations (XGBoost-SHAP) decoded nonlinear drivers and abrupt thresholds. Built-up land increased from 1.15% (2000) to 2.39% (2020), with 2030 projected values of 2.97%, 3.15% and 3.35% under the three scenarios. MESLI reached maxima in mountain forests and minima in intensively developed intermontane basins. Land-use variables contributed >40% to mean-absolute SHAP importance, followed by precipitation and slope. Precipitation breakpoints differed across scenarios (1008 mm, 929 mm, 1065 mm), and an aridity-index threshold of 1.27 was detected for SSP245. Joint-response surfaces show favourable moisture–terrain conditions amplify forest-related multi-ecosystem-service benefits, which are substantially offset by urban encroachment. Statistical results further indicate hydrothermal thresholds are scenario-dependent instead of static, with divergent land-use-climate coupling strength between mountain and basin landscapes in plateau urban agglomerations.