Data-driven vulnerability assessment of the Nansi Lake Basin water resource system

Study region Nansi Lake Basin, a water-scarce region in northern China. Study focus This study conducted a data-driven vulnerability assessment of the Nansi Lake Basin water resource system. The Driver–Pressure–State–Impact–Response conceptual model was applied to define a three-tier vulnerability evaluation system comprising a target layer, 5 first-level indicators, and 26 second‐level indicators. Particle swarm optimization (PSO) was employed to determine the optimal hyperparameters of a long short-term memory (LSTM) network trained on data comprising randomly generated evaluation indicator values and their corresponding vulnerability grading standards. This hybrid PSO–LSTM model was used to assess the spatiotemporal vulnerability of the Nansi Lake Basin water resource system from 2010 to 2021, and its rationality was verified by comparison with the variable fuzzy assessment method. New hydrological insights for the region The hybrid PSO–LSTM model exhibited high accuracy ( R 2 > 0.999 for both training and validation samples) and stability, indicating its suitability for vulnerability assessments. Temporally, the vulnerabilities in the Nansi Lake Basin and its six subregions fluctuated downward over the study period, evolving from highly vulnerable in 2010 to slightly vulnerable by 2021. Spatially, vulnerability decreased faster in the western than eastern subregions. Finally, the obstacle degree model identified the causes of vulnerability to provide a scientific foundation for the utilization and management of water resources in the Nansi Lake Basin.