Author: Nasridinov, Rustamjon Baxtiyorjon o‘g‘li

Annotation: In this study, the interaction between surface water and groundwater, as well as the processes of contamination, were digitally modeled using physics-based artificial intelligence and mathematical modeling methods. An integrated approach combining the MODFLOW–MT3DMS physical model and the Physics-Informed Neural Network (PINN) artificial intelligence model was employed. The Zarafshan Valley lowland was selected as the study area, focusing on the analysis of water table levels, flow velocity, nitrate pollution, and biological degradation processes.The results demonstrated that the AI-based surrogate model operates 12–20 times faster than the traditional physical model while maintaining an accuracy level of approximately 95%. Variations in irrigation intensity and temperature significantly affected the dynamics of pollutant dispersion, accelerating infiltration and diffusion processes. Through the hybrid AI+Physics approach, it became possible to assess ecological safety, identify contamination zones, and manage water resources in real-time mode.The findings of this research hold both scientific and practical significance for water resource management, environmental safety monitoring, and the implementation of Digital Twin technologies in hydrological systems.

Keywords: artificial intelligence, physical modeling, groundwater, surface water, contamination, advection–diffusion, surrogate model, PINN, digital twin, environmental safety.

Pages in journal: 106 - 121