Scholarlink

<p>Groundwater performs a crucial role in meeting the water demand in the industrial Narayanganj district, Bangladesh, where excessive extraction for domestic, industrial, and agricultural needs has led to severe water scarcity. This paper aims to locate groundwater potential zones (GPZs) in the Narayanganj district through geospatial analysis utilizing the analytic hierarchy process&#xa0;(AHP) and five machine learning (ML) algorithms: random forest&#xa0;(RF), boosted regression trees&#xa0;(BRT), k-nearest neighbors&#xa0;(KNN), classification and regression trees (CART), and naïve bayes (NB). The dataset includes eight influencing factors for groundwater and inventory data. The study incorporates&#xa0;349 well and non-well locations as inventory data, which were randomly divided into 70% for training and 30% for testing. This research found that most models classified the northwest region as having very low groundwater potential, with NB and CART showing 40% and 30%, respectively. Using the ROC curve for validation, all models scored an AUC-ROC above 0.90. The results indicated very low to low GPZs in the northwest and west, while the eastern and southwestern parts of Narayanganj showed higher potential. The study's findings will assist policymakers in identifying new groundwater sources for long-term use in this industry-dominated area and provide crucial insights for sustainable groundwater management.</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

Prediction of Groundwater Potential Zone Using Machine Learning and Geospatial Approaches for an Industry-Dominated Area in Narayanganj, Bangladesh

  • 期刊论文

  • 发表日期：

    2025年3月31日

查看全文
  • Md. Mahmudul Hasan,
  • M. M. Shah Porun Rana,
  • Md Tasim Ferdous

摘要

Groundwater performs a crucial role in meeting the water demand in the industrial Narayanganj district, Bangladesh, where excessive extraction for domestic, industrial, and agricultural needs has led to severe water scarcity. This paper aims to locate groundwater potential zones (GPZs) in the Narayanganj district through geospatial analysis utilizing the analytic hierarchy process (AHP) and five machine learning (ML) algorithms: random forest (RF), boosted regression trees (BRT), k-nearest neighbors (KNN), classification and regression trees (CART), and naïve bayes (NB). The dataset includes eight influencing factors for groundwater and inventory data. The study incorporates 349 well and non-well locations as inventory data, which were randomly divided into 70% for training and 30% for testing. This research found that most models classified the northwest region as having very low groundwater potential, with NB and CART showing 40% and 30%, respectively. Using the ROC curve for validation, all models scored an AUC-ROC above 0.90. The results indicated very low to low GPZs in the northwest and west, while the eastern and southwestern parts of Narayanganj showed higher potential. The study's findings will assist policymakers in identifying new groundwater sources for long-term use in this industry-dominated area and provide crucial insights for sustainable groundwater management.