Groundwater suitability assessment for irrigation and drinking purposes by integrating spatial analysis, machine learning, water quality index, health risk model

Abstract

An in-depth understanding of nitrate-contaminated groundwater quality and risks is important for groundwater management. Hydrochemical characteristics and driving forces of groundwater quality and non-carcinogenic risks of nitrate were revealed by the integrated approaches of self-organizing map analysis, spatial visualization by geography information system, entropy and irrigation water quality indices and human health risk model. Groundwater samples were subdivided into two clusters by SOM analysis. Cluster Ⅰ including three samples were Ca-SO₄ type and cluster Ⅱ of remaining 136 samples were Ca-HCO₃ type. Hydrochemical compositions of two cluster samples were dominated by water-rock interaction: (1) calcite and gypsum dissolution for Cluster I samples, and (2) calcite dissolution, silicate weathering and positive cation exchange for Cluster Ⅱ samples. Nitrate contamination occurred in both Cluster I and Ⅱ samples, primarily induced by agricultural nitrogen fertilizer. The EWQI results showed that 90.97% in total groundwater samples were suitable for drinking purpose, while the IWQI results demonstrated that 65.03% in total groundwater samples were appropriate for irrigation purpose. The HHR model and Monte Carlo simulation indicated that the non-carcinogenic nitrated risk was highest in children. Exposure frequency was the most sensitive factor (86.33% in total) influencing the total non-carcinogenic risk, indicated by sensitivity analysis.