Simulation and prediction of sulfamethazine migration, transformation and risk diffusion during cross-media infiltration from surface water to groundwater driven by dynamic water level: Machine learning coupled HYDRUS-GMS model.
Journal:
Journal of environmental management
Published Date:
Nov 30, 2024
Abstract
Seasonal water level fluctuations in rivers significantly influenced the cross-media migration, transformation, and risk diffusion of antibiotics from the vadose zone into groundwater. This study developed a coupled model integrating machine learning (ML) with HYDRUS-3D and GMS to accurately predict sulfamethazine migration under dynamic water levels. The predictive accuracy (E≥0.98) of this ML-HYDRUS-GMS model was enhanced by accounting for seasonal water level fluctuations and biogeochemical variability. Significant seasonal differences presented with sulfamethazine diffusion in the vadose zone with the migration rate decreased from 0.06 m/d to 0.02 m/d with the transition from wet to dry seasons. After 6 years of infiltration, it reached groundwater, where lateral migration rates, influenced by seasonal flow variations, were 0.12 m/d in the wet season and decreased to 0.07 m/d in the dry season, with a diffusion range extending to 217 m over 100 years. This discrepant continuous filtration of sulfamethazine and the succession of metabolic pathways induced toxicity range to expand by 65.6 m and the risk to increase to warning level. Sulfamethazine underwent oxidative breakdown in aerobic vadose zone conditions, while anaerobic groundwater conditions led to hydrogenation and reduction, increasing its migration distance.