AI-enhanced system dynamics simulation for sustainable urban development under water constraints: A case study of Langfang, China.

Water-scarce cities undergoing rapid urbanization must sustain economic growth while complying with stringent water-resource and environmental constraints. However, long-term policy simulations are often affected by uncertainty in exogenous drivers, particularly population dynamics. This study develops an AI-enhanced system dynamics (SD) framework that couples a BP neural-network ensemble forecaster with a multi-subsystem SD model integrating water resources, environment processes, and urban development. The forecasting module nonlinearly combines GM (1,1) and Logistic models to generate annual population trajectories, which are used as exogenous boundary inputs for SD simulations. The SD model captures feedback among water supply and allocation, reclaimed-water reuse, pollution abatement, economic growth, industrial upgrading, and urban-rural coordination. The framework is applied to Langfang, Hebei Province, a severely water-scarce city in the Beijing-Tianjin-Hebei region, using data from 2000 to 2021 and simulating five policy scenarios for 2020-2035. The BP ensemble achieves a mean relative backcasting error of 0.08%, outperforming GM (1,1) (0.40%) and the Logistic model (0.42%). Under the coordinated development scenario (S5), characterized by a 3.2% annual reduction in industrial water use per CNY 10,000 of industrial value added, a 60% reclaimed-water utilization rate, and an urbanization rate of 78%, GDP reaches CNY 1.005 trillion by 2035. Reclaimed water accounts for 24.4% of total supply, wastewater discharge declines to 55.6% of the S1 level, and the urban-rural income ratio narrows to 1.74. Compared with single-objective pathways, S5 provides a more balanced trajectory across economic growth, water security, pollution control, and social equity. The proposed AI-SD framework provides a reproducible approach for reducing exogenous-input uncertainty and supporting multi-objective policy design in water-constrained urban systems.