Multi-Method Identification of Characteristic Groundwater Pollutants Under Aquifer Structural Constraints and Their Indication of Environmental Processes.

Accurate identification of groundwater contamination factors is critical yet challenging at industrial sites with complex pollution sources and heterogeneous aquifers. This study developed a multi-method integrated framework combining pollution index evaluation, statistical analysis, and machine learning to achieve stratified identification of characteristic pollutants across different aquifers. Applied to a chemical plant site, shallow groundwater pollutants were identified through pollution index, correlation analysis, and random forest/XGBoost models. Deep groundwater pollutants were discriminated based on pollution grade distributions and exceedance characteristics, followed by spatial analysis. The results showed clear aquifer-dependent differences in groundwater contamination. In shallow groundwater, total hardness showed the highest correlation with pollution levels (ρ = 0.62), followed by SO4 2- (ρ = 0.56), COD (ρ = 0.55), and NH4 +-N (ρ = 0.45). Machine learning results further indicated that nitrogen- and sulfur-related indicators, especially NH4 +-N and SO4 2-, were the dominant factors controlling shallow groundwater pollution. Spatially, NH4 +-N mainly exhibited point-source pollution characteristics, whereas SO4 2- showed a multi-source distribution pattern. In contrast, deep groundwater showed lower contamination levels and a more limited spatial extent, suggesting restricted downward migration of pollutants. The proposed framework effectively reveals aquifer-dependent characteristic pollutants and provides practical support for groundwater pollution identification and management in chemical plant areas.