Source Analysis of Groundwater Chemical Components in the Middle Reaches of the Dawen River Based on Unsupervised Machine Learning and PMF Source Analysis

作者：Xinqi Wang, Xinqi Wang, Zhenhua Zhao, Hongyan An, Lin Han, Mingming Li, Zihao Wang, Xinfeng Wang, Xinfeng Wang, Zheming Shi · 发表于：Water · 年份：2025 · DOI：10.3390/w17202924 · 被引用次数：3 · 研究领域：Environmental and Agricultural Sciences、Hydrological Forecasting Using AI、Water Quality and Pollution Assessment

Groundwater chemical composition often exhibits complex characteristics under the combined influence of anthropogenic activities and natural geological conditions. Accurately distinguishing between human-derived and naturally occurring constituents is crucial for formulating effective pollution control strategies and ensuring sustainable groundwater resource management. However, conventional hydrogeochemical analytical methods often face challenges in quantitatively differentiating these overlapping influences. In this study, 66 groundwater samples were collected from the midstream section of the Dawen River Basin, an area subject to significant anthropogenic pressure. An integrated approach combining hydrogeochemical analysis, Self-Organizing Map (SOM) clustering, and Positive Matrix Factorization (PMF) receptor modeling was employed to identify sources of chemical constituents and quantify the proportional contributions of various factors. The results indicate that: (1) The predominant groundwater types in the study area were Cl·SO4·Ca. (2) SOM clustering classified the groundwater samples into five distinct groups, each reflecting a dominant influence: (i) natural geological processes—samples distributed within the central geological mining area; (ii) agricultural activities—samples located in intensively cultivated zones along both banks of the Dawen River; (iii) hydrogeochemical evolution—samples concentrated in areas with impermeable surfaces on the eastern and western ...