Advancing wetland groundwater pollution zoning: A novel integration of Monte Carlo health risk modeling and machine learning

Wetlands serve as crucial water reservoirs, providing essential water resources for the surrounding regions. However, elevated ion concentrations in wetland groundwater may pose health risks to local populations. This study focused on Judian Lake and its adjacent areas, proposing an innovative multimodel coupled uncertainty propagation framework to establish an integrated "process characterization-risk quantification-source management" methodology. The Entropy-Weighted Water Quality Index (EWQI), deterministic and Monte Carlo-based probabilistic health risk assessments, Principal Component Analysis-Absolute Principal Component Score-Multiple Linear Regression (PCA-APCS-MLR), and Self-Organizing Map-K-means (SOM-K-means) clustering were used. Results indicated that over 50 % of the water resources in the study area were suitable for drinking and irrigation purposes. F^-, Mn, NO₂^-, and NO₃^- posed non-carcinogenic risks to both adults and children, with NO₃^- being the most severe. Monte Carlo indicated that for high-concentration pollutants (Mn, NO₂^-, and NO₃^-), source control measures should prioritize concentration reduction, whereas for low-concentration pollutants (F^-), minimizing exposure pathways was necessary. The PCA-APCS-MLR model suggested that NO₃^- primarily originated from agricultural activities, while F^- mainly came from the weathering and dissolution of fluorite. SOM-K-means divided the study into four clusters, of which cluster III was the most polluted.