Analysis of the hydrogeochemical characteristics of groundwater and identification of pollution sources in facility agriculture areas using self-organizing neural networks

Abstract

Facility agriculture is a modern intensive cultivation method that is widely seen as the future of global agriculture. However, large-scale emissions of concentrated pollutants during production pose serious threats to groundwater quality. Identifying the sources of pollutants and assessing source-specific risks are critical for developing effective risk mitigation strategies. In this study, a combination of methodologies including Self-Organizing Maps (SOM), K-means clustering, factor analysis, and ion ratio analysis were utilized to investigate pollution risks in a typical facility agriculture area in Shouguang City, Shandong Province, China. The groundwater quality in the study area is poor and slightly alkaline, with NO₃⁻ being the main pollutant. The chemical composition of groundwater in the aquifer is influenced by both human activities (41.89%, such as agricultural activities) and natural processes (58.11%, such as water–rock interactions). Furthermore, pollution sources in the study area were spatially categorized into two clusters: Cluster 1, mainly located on the right bank of the Mi River, is primarily related to urban domestic sewage discharge, and Cluster 2, primarily on the left bank of the Mi River, is mainly related to agricultural activities. The average concentrations of Cl⁻ and Na⁺, both of which have high mobility, are significantly higher in Cluster 2 than in Cluster 1, suggesting that the groundwater system in Cluster 2 is relatively closed, resulting in higher ion concentrations and pollution levels. These findings provide valuable insights for the prevention, control, and remediation of groundwater pollution in the study area, and in facility agriculture regions generally.