Scholarlink

<p>The rapidly increasing demand for groundwater in metropolitan areas has resulted in significant variations in groundwater quality in recent years. Identifying hydrochemical processes of groundwater and predicting groundwater quality are crucial for sustainable suitability assessments, particularly in regions with intense anthropogenic activities. In our study, 41 groundwater samples were collected from Chongqing, with 33 samples from the valley area and 8 from the ridge area, to investigate the factors controlling groundwater chemistry and influencing groundwater quality in this densely populated and economically developed urban region. The dominant hydrochemical facies in the study area were identified as Ca–HCO<sub>3</sub> in the ridge area and Na–HCO<sub>3</sub> in the valley area. Both qualitative (hydrochemical diagrams) and quantitative analysis (the Positive matrix factorization) signified that groundwater chemistry was influenced by both geological and anthropogenic factors, exhibiting spatial variation in controlling processes between valley and ridge areas. Municipal activities and the weathering of mudstone and sandstone were the dominant controls on groundwater chemistry in valley areas. Limestone dissolution, weathering of shale and feldspathic sandstone, and agricultural activities were predominant in ridge areas. Entropy-weight quality index (EWQI) and irrigation water quality index (IWQI) were applied to evaluate the groundwater suitability. The results indicated that 73.18% and 83.30% of groundwater samples in the study area are suitable for drinking and irrigation purposes, respectively. The sensitivity analysis confirmed the robustness of the groundwater suitability assessment. Three machine-learning algorithms were utilized for water quality prediction. Multivariable linear regression was demonstrated as a robust machine learning algorithm for drinking water quality prediction, with the strongest regression values (R<sup>2</sup> = 0.999), while the support vector machine was superior for irrigation water quality prediction with an R<sup>2</sup> of 0.853. The findings of this study may inform the treatment and management of groundwater resources in metropolitan regions globally.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Using positive matrix factorization, sensitivity analysis, and machine learning models to identify hydrochemical processes and predict groundwater quality in a typical metropolitan city

  • 期刊论文

  • 发表日期：

    2025年9月29日

查看全文
  • Zhan Xie,
  • Jinhang Huang,
  • Denghui Wei,
  • Yangshuang Wang,
  • Si Chen,
  • Shiming Yang,
  • Xingjun Zhang,
  • Chang Yang,
  • Junyi Li,
  • Yunhui Zhang

摘要

The rapidly increasing demand for groundwater in metropolitan areas has resulted in significant variations in groundwater quality in recent years. Identifying hydrochemical processes of groundwater and predicting groundwater quality are crucial for sustainable suitability assessments, particularly in regions with intense anthropogenic activities. In our study, 41 groundwater samples were collected from Chongqing, with 33 samples from the valley area and 8 from the ridge area, to investigate the factors controlling groundwater chemistry and influencing groundwater quality in this densely populated and economically developed urban region. The dominant hydrochemical facies in the study area were identified as Ca–HCO3 in the ridge area and Na–HCO3 in the valley area. Both qualitative (hydrochemical diagrams) and quantitative analysis (the Positive matrix factorization) signified that groundwater chemistry was influenced by both geological and anthropogenic factors, exhibiting spatial variation in controlling processes between valley and ridge areas. Municipal activities and the weathering of mudstone and sandstone were the dominant controls on groundwater chemistry in valley areas. Limestone dissolution, weathering of shale and feldspathic sandstone, and agricultural activities were predominant in ridge areas. Entropy-weight quality index (EWQI) and irrigation water quality index (IWQI) were applied to evaluate the groundwater suitability. The results indicated that 73.18% and 83.30% of groundwater samples in the study area are suitable for drinking and irrigation purposes, respectively. The sensitivity analysis confirmed the robustness of the groundwater suitability assessment. Three machine-learning algorithms were utilized for water quality prediction. Multivariable linear regression was demonstrated as a robust machine learning algorithm for drinking water quality prediction, with the strongest regression values (R2 = 0.999), while the support vector machine was superior for irrigation water quality prediction with an R2 of 0.853. The findings of this study may inform the treatment and management of groundwater resources in metropolitan regions globally.