A national-scale probabilistic ecological risk indicator for land use change: An improved Sharpe ratio framework with nonlinear driving threshold identification across China

Global climate change and intensified anthropogenic land use have profoundly altered ecosystem structure and function, resulting in increasingly complex ecological risk (ER) patterns. However, ecological risk induced by land use change (ERLUC) remains constrained by insufficient consideration of land use uncertainty, ecosystem service (ES) trade-offs, and nonlinear mechanisms. This study develops a probability-based ERLUC framework for China by integrating land use simulation (LCM), ecosystem service modeling (InVEST), an improved Sharpe ratio-based risk indicator, and LightGBM-SHAP. The framework quantifies future land use transition probabilities and ecosystem service responses from 2020 to 2035, while identifying dominant drivers and ecological thresholds.ERLUC exhibited significant spatial heterogeneity across China. High-risk areas were mainly concentrated in rapidly urbanizing regions such as the North China Plain, whereas low-risk areas were primarily distributed in the interior Qinghai–Tibet Plateau and arid regions of Northwest China. Under future land use scenarios, ecosystem services showed contrasting responses: carbon sequestration remained stable (−0.02%), water yield increased slightly (+1.16%), while habitat quality and soil conservation declined by 1.46% and 9.58%, respectively. LightGBM-SHAP analysis identified NDVI as the dominant factor controlling ER differentiation, contributing 17.7%, followed by bio15 (13.3%), bio23 (10.9%), and precipitation (10.4%). Nonlinear analysis revealed a clear NDVI threshold effect, with ecological risk decreasing rapidly as vegetation coverage increased and showing predominantly negative responses when NDVI exceeded approximately 0.15–0.20.This study provides a probability- and mechanism-oriented framework for national-scale ERLUC assessment, enhancing understanding of how uncertain land transitions, ecosystem service responses, and nonlinear environmental drivers jointly shape ecological risks. The identified spatial patterns and ecological thresholds support differentiated land management and ecological restoration strategies.