Scholarlink

<p>Groundwater is a critical resource for ecosystem sustainability, human livelihood, and advancing socio-economic development. However, groundwater potential zones (GWPZs) in data-scarce environments, such as those in developing countries, remain a complex task due to limited hydrogeological data and infrastructure. To address this challenge, the present study integrates GIS-based machine learning models such as random forest (RF) and support vector machines (SVM) along with the analytic hierarchy process (AHP) to predict GWPZs in the western catchment of Abaya–Chamo Lakes Basin, Ethiopia Rift System. The model incorporates eight geo-environmental factors: geology, geomorphology, lineament density, drainage density, rainfall, slope, land use/land cover, and soil type. A geospatial inventory of 160 groundwater occurrence points, including 78 shallow wells, 38 boreholes, and 44 springs, was used: 70% for training and 30% for testing. GWPZs were delineated using index-based modeling, and model performance was evaluated using receiver operating characteristic (ROC) curves. The RF model identified 16.25% of the area as having very high groundwater potential, 22.48% as high, 44.24% as moderate, 12.04% as low, and 4.95% as very low. While all models exceeded the acceptable AUC threshold of 0.6, RF achieved the highest accuracy (AUC = 0.89), followed by AHP (0.87) and SVM (0.84), confirming RF’s superior predictive capability. These findings provide essential insights for policymakers and planners aiming to implement sustainable groundwater management strategies in data-limited areas.</p> Graphical abstract <p></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 zones using machine learning and MCDA methods in western catchment of Abaya–Chamo lakes basin, Ethiopia Rift System

  • 期刊论文

  • 发表日期：

    2025年12月16日

查看全文
  • Tilahun Wankie Wanjala,
  • Samuel Dagalo Hatiye,
  • Abunu Atlabachew Eshete,
  • Ayano Hirbo Gelebo

摘要

Groundwater is a critical resource for ecosystem sustainability, human livelihood, and advancing socio-economic development. However, groundwater potential zones (GWPZs) in data-scarce environments, such as those in developing countries, remain a complex task due to limited hydrogeological data and infrastructure. To address this challenge, the present study integrates GIS-based machine learning models such as random forest (RF) and support vector machines (SVM) along with the analytic hierarchy process (AHP) to predict GWPZs in the western catchment of Abaya–Chamo Lakes Basin, Ethiopia Rift System. The model incorporates eight geo-environmental factors: geology, geomorphology, lineament density, drainage density, rainfall, slope, land use/land cover, and soil type. A geospatial inventory of 160 groundwater occurrence points, including 78 shallow wells, 38 boreholes, and 44 springs, was used: 70% for training and 30% for testing. GWPZs were delineated using index-based modeling, and model performance was evaluated using receiver operating characteristic (ROC) curves. The RF model identified 16.25% of the area as having very high groundwater potential, 22.48% as high, 44.24% as moderate, 12.04% as low, and 4.95% as very low. While all models exceeded the acceptable AUC threshold of 0.6, RF achieved the highest accuracy (AUC = 0.89), followed by AHP (0.87) and SVM (0.84), confirming RF’s superior predictive capability. These findings provide essential insights for policymakers and planners aiming to implement sustainable groundwater management strategies in data-limited areas.

Graphical abstract