Deep learning-based analysis of groundwater pollution sources and health risk evaluation around typical irrigation areas

The ecological environment in western Jilin is fragile. To investigate the changes in groundwater quality and their impact on human health following the implementation of land consolidation projects, as well as to achieve the sustainable goal of clean water, we aimed to address the limited application of triangular stochastic simulations. Our research focused on evaluating water quality using deep learning methods, and assessing the health risks associated with pollutants in groundwater. In this study, a Monte Carlo random simulation model was developed, which is particularly suitable for the complex groundwater environments surrounding irrigation districts. The results showed that the chemical type of groundwater in three different aquifers were HCO3-Na+-Ca2+. The three-nitrogen index of the phreatic water monitoring site exceeded 82.79%, indicating that fertilizers applied for agricultural have polluted groundwater. Besides, Class IV monitoring points around the irrigation area accounted for 66.07%. Both phreatic and confined water were affected by dissolution-secondary enrichment. Non-carcinogenic and carcinogenic risks in different aquifers were ranked as follows: Quaternary phreatic water > Quaternary confined water > Neogene confined water; the highest non-carcinogenic risk was on the order of 10(-8), the carcinogenic risk of As3+ was on the order of 10(-2), Monte Carlo model was more sensitive to changes in data.