Identification and assessment of urban groundwater pollution based on Monte Carlo health risk modeling and self-organizing map (SOM).

With the acceleration of urbanization, groundwater quality is severely threatened, posing health risks to public health and the environment. In reality, the highly heterogeneous geological background and hydrochemical relationships not only increase the difficulty in identifying pollution, but also pose significant challenges in groundwater risk prevention and control. Taking Guangzhou, a megacity in the bustling Pearl River Delta, as the research object, this study leverages the ability of machine learning to mine complex data and the advantage of Monte Carlo simulation in quantifying random uncertainties to deeply analyze the driving factors, pollution sources, and potential health risks of urban groundwater chemistry. Through the SOM model, a cluster analysis was conducted on 107 groundwater samples, resulting in the classification of 5 distinct water chemical groups with significant characteristics (Clusters I - V). Among them, human activities (domestic pollution) have a significant impact on groundwater chemistry mainly in Cluster I (a total contribution of 52.9 %). Manganese (Mn) and nitrate (NO3-) pose non-carcinogenic risks, with the highest risks observed for Mn in Cluster II and NO3- in Cluster III. Monte Carlo simulation indicates that the overall health risk in the region is low, with only 1.3 % of children potentially exposed to high HI conditions. This study provides valuable insights for the precision prevention, control, and risk management of complex groundwater pollution in cities such as those in the Pearl River Delta.