The southern regions of China are rich in ion-adsorbed rare earth mineral resources, primarily distributed in ecologically fragile red soil hilly areas. Recent decades of mining activities have caused severe environmental damage, exacerbating ecological security (ES) risks due to the inherent fragility of the red soil hilly terrain. However, the mechanisms through which multiple interacting factors influence the ES of rare earth mining areas (REMA) remain unclear, and an effective methodological framework to evaluate these interactions dynamically is still lacking. To address these challenges, this study develops an innovative dynamic ES evaluation and earlywarning simulation framework, integrating Variable Weight (VW) theory and the Bayesian Network (BN) model. This framework enhances cross-stage comparability and adapts to evolving ecological conditions while leveraging the BN model's diagnostic inference capabilities for precise ES predictions. A case study was conducted in the Lingbei REMA. The main findings of the study are as follows: (1) From 2000 to 2020, the overall ES of the mining area exhibited a dynamic trend of deterioration, followed by improvement, and ultimately stabilization. (2) Scenario S27 (high vegetation health status and high per capita green space coverage) significantly reduces the probability of the ES reaching the extreme warning level. (3) The evaluation and simulation framework developed in this study provides a more accurate representation of the ES level distribution and its variations, with probabilistic predictions of ES demonstrating high accuracy. This study is of great significance for improving regional ES, supporting the optimization of ecological restoration strategies under multi-objective scenarios, and promoting the coordinated development of nature and resource utilization.

来源平台：JOURNAL OF CLEANER PRODUCTION