Machine learning approaches to identify hydrochemical processes and predict drinking water quality for groundwater environment in a metropolis
Zhan Xie, Weiting Liu, Si Chen, Rongwen Yao, Yang Chang, Xingjun Zhang, Junyi Li, Yangshuang Wang, Yunhui Zhang
Abstract
Study region The study area is located in the urban area of Chongqing City, the largest metropolis in southwestern China. Study focus Various hydrochemical processes and water quality prediction are unknown, hampering the sustainable development of metropolis. In this study, geochemical model, entropy-weighted water quality index (EWQI), and machine learning (ML) methods were applied to explore the hydrochemical processes and predict the groundwater quality for drinking purposes. New hydrological insights for the region The self-organizing map classifies the groundwater samples into 2 clusters. Cluster 1, predominantly located along ridge areas, exhibited HCO 3 –Ca as the primary hydrochemical facie. Carbonate dissolution, cation exchange processes, and agricultural activities dominated the groundwater chemistry of Cluster 1. HCO 3 –Ca and HCO 3 –Na types were the dominant hydrochemical types of Cluster 2 in valley areas. Silicate weathering, cation exchange processes, and domestic sewage were the driving factors controlling the hydrochemistry of Cluster 2. EWQI results showed that 59.48 %, 31.90 % and 8.62 % of samples were excellent, good and medium for drinking, respectively. Four supervised machine learning methods were conducted to predict drinking water quality . Linear regression demonstrated the best correlation of 0.9999. The findings offer invaluable insights into groundwater suitability and evolution processes in a typical population density area and ensure a secure and sustainable domestic water supply worldwide.