Scholarlink

<p>Arsenic (As) is a geogenic groundwater contaminant, and its mobilization involves complex, spatially heterogeneous mechanisms. This study first seeks to identify the mechanisms and then determine the conditions under which they operate, to understand the heterogeneity of influences on As mobilization. A prominent As-affected region was selected to model groundwater quality and the physicochemical properties of the overlying soil, using a large dataset of 66,183 observations fed into machine learning algorithms. The models were interpreted to evaluate the relative influences of the parameters and their threshold concentrations above which mobilization is likely. The results show that 94.6% of the study area has an As concentration exceeding 10&#xa0;µg/L (the WHO limit), while the modelling achieved satisfactory accuracy, with a kappa statistic greater than 0.5. The SHapley Additive exPlanations interpretation framework results indicate that As desorption from reductive dissolution of ferric oxyhydroxides is dominant, it highlights the role of particle-size fractions in the overlying soil layer. Iron concentrations above 2&#xa0;mg/L and sand content in the range of 45–85&#xa0;g/kg, acting as conduits, are indicative of higher arsenic severity, while finer particles serve as the source. This study formulates a comprehensive strategy for identifying regions susceptible to elevated arsenic mobilization, employing a methodology that integrates groundwater chemistry with the physical properties of the overlying soil and is adaptable across diverse geographical settings. Overall, it advances research on As management, especially in developing countries, by offering a robust, economically feasible, and accessible delineation approach for As-severe zones.</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

Idealizing the role of soil particle size fractions in arsenic mobilization mechanisms and severity in groundwater using an explanatory framework of machine learning and SHAP analysis

  • 期刊论文

  • 发表日期：

    2026年7月13日

查看全文
  • Tuhin Subhra Konar,
  • Rijurekha Dasgupta,
  • Gourab Banerjee,
  • Saroj Kumar,
  • Subhasish Das,
  • Asis Mazumdar,
  • Arunabha Majumder

摘要

Arsenic (As) is a geogenic groundwater contaminant, and its mobilization involves complex, spatially heterogeneous mechanisms. This study first seeks to identify the mechanisms and then determine the conditions under which they operate, to understand the heterogeneity of influences on As mobilization. A prominent As-affected region was selected to model groundwater quality and the physicochemical properties of the overlying soil, using a large dataset of 66,183 observations fed into machine learning algorithms. The models were interpreted to evaluate the relative influences of the parameters and their threshold concentrations above which mobilization is likely. The results show that 94.6% of the study area has an As concentration exceeding 10 µg/L (the WHO limit), while the modelling achieved satisfactory accuracy, with a kappa statistic greater than 0.5. The SHapley Additive exPlanations interpretation framework results indicate that As desorption from reductive dissolution of ferric oxyhydroxides is dominant, it highlights the role of particle-size fractions in the overlying soil layer. Iron concentrations above 2 mg/L and sand content in the range of 45–85 g/kg, acting as conduits, are indicative of higher arsenic severity, while finer particles serve as the source. This study formulates a comprehensive strategy for identifying regions susceptible to elevated arsenic mobilization, employing a methodology that integrates groundwater chemistry with the physical properties of the overlying soil and is adaptable across diverse geographical settings. Overall, it advances research on As management, especially in developing countries, by offering a robust, economically feasible, and accessible delineation approach for As-severe zones.