Landscape ecological risk dynamics and scenario-based regulation in the Gannan Plateau, China: A coupled PLUS-LER-BRT-ABM framework

Landscape ecological risk (LER) assessment is essential for sustainable land management in ecologically fragile alpine regions, yet existing approaches typically adopt one-way workflows that disconnect landscape simulation, risk assessment, driver analysis, and scenario optimization. This study developed a coupled PLUS–LER–BRT–ABM framework with closed-loop coupling between the landscape-simulation and agent-based modules, and applied it to the Gannan Plateau over 1985–2045. Results showed a persistent northeast–southwest risk gradient, with the high-risk (HighER) share rising from 35.2% in 1985 to 37.8% in 2024. Boosted regression tree (BRT) analysis attributed 38.8% of relative influence to socioeconomic factors and 8.7% to grazing intensity. By 2045, the ecology-priority (EP) and comprehensive-coordination (CC) scenarios reduced mean regional LER by 0.82% and 0.88% relative to the baseline (BL), whereas human-activity control (HC) achieved only 0.05%. In a matched open-loop control, agent actions could not modify the precomputed PLUS trajectory; open and closed paths remained identical before the first 2029 refresh, after which feedback supplied 15.0% and 21.3% of the total EP and CC reductions. Counterfactual separation further showed that FVC recovery dominated the normalized single-channel decomposition, accounting for 86.8% of the summed isolated-channel effects. These results show that five-year feedback changes the simulated risk pathway and that ecology-active regulation outperforms development control alone.