Integrating the AE-AM model and the SMI-P framework for water environmental security assessment: A case study of the Chushandian Reservoir basin

This is the authors' abstract. We don't add key points for this paper.

• Developed an AE-AM model within the SMI-P framework to assess water environmental security at a refined raster-scale. • Identified SPEI and RUSLE as the dominant drivers, with TWS, WC, WY, and SSM as secondary contributors. • Revealed the dynamic impacts of climate changes, soil erosion, and human activities on water environment security. • Applied SHAP and obstacle analysis to disentangle positive/negative contributions of indicators. • Verified model robustness and reliability through Monte Carlo simulation, expert scoring, and field validation. Water environmental security issues have become prominent with the rapid development of the economy and society. Scientifically conducting assessments is of great significance for identifying potential risks, protecting water resources, and promoting regional collaborative development. Integrating the AE-AM model and SMI-P framework, this study proposes a refined raster-scale water environmental security assessment for the case study of Chushandian Reservoir Basin from 2016 to 2023. The objectives are to identify key indicators, reveal spatiotemporal dynamics, and provide scientific support for watershed management. The results indicate that (1) SPEI and RUSLE are the dominant driving factors of water environmental security, while TWS, WC, WY, and SSM play secondary roles; (2) spatiotemporal analysis reveals a drought-wet-drought change pattern between 2016 and 2023, with relatively high security in water bodies and mountainous vegetation zones, but lower security in urbanized and intensively cultivated regions; (3) obstacle degree reveals the highest constrained factors are SPEI and RUSLE, while SHAP contribution further confirms the negative impacts (RUSLE and WC) and positive impacts (TWS, SPEI, WY, and SSM); (4) uncertainty and perturbation sensitivity analysis indicate data-layer and model-layer uncertainties are generally within acceptable ranges, and weight consistency with expert judgment demonstrates the robustness of the proposed methods. Overall, this study provides a dynamic and refined raster scale method for assessing water environmental security and offers valuable implications for adaptive management and basin governance.