Scholarlink

<p>Water scarcity significantly constrains agricultural productivity and rural livelihoods across the Middle East. In northern Jordan, where annual precipitation ranges between 400–600 mm and water availability falls below 700 m³/capita, innovative approaches are required to enhance local water supply. This study develops a hybrid framework integrating the Analytic Hierarchy Process (AHP) with a multilayer perceptron (MLP) neural network to identify optimal rainwater harvesting sites in Irbid Governorate. Eight spatial conditioning factors were analyzed: slope, land use/land cover, hydrologic soil group, curve number, stream order, drainage density, lineament distribution, and runoff potential. The AHP consistency ratio (0.07) confirms reliable expert judgment. The MLP model demonstrated strong predictive performance, achieving 88% accuracy, 86% precision, 87% recall, and an F1-score of 0.86 under cross-validation. Spatial validation using 127 field-verified structures showed 83.5% agreement, confirming the model’s ability to capture real-world siting patterns. Results indicate that approximately 83% of the basin is suitable for rainwater harvesting, with 52% classified as highly suitable (suitability ≥ 0.70). Integration with drainage networks identified 35 optimal locations, including 15 check dams along major drainage corridors and 20 percolation ponds in agricultural lowlands. Feature importance analysis highlights slope (25%) and hydrologic soil group (22%) as dominant controls, consistent with hydrological theory. The resulting spatial configuration supports a multi-scale water harvesting strategy aligned with integrated water resources management principles. This study provides field-ready, prioritized locations and a robust decision-support framework to enhance water security and support sustainable resource management in water-scarce regions.</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

Integrating GIS and Neural Networks for Sustainable Water Harvesting Site Selection

  • 期刊论文

  • 发表日期：

    2026年5月16日

查看全文
  • Marwa Elshebli,
  • Yousef Sharrab

摘要

Water scarcity significantly constrains agricultural productivity and rural livelihoods across the Middle East. In northern Jordan, where annual precipitation ranges between 400–600 mm and water availability falls below 700 m³/capita, innovative approaches are required to enhance local water supply. This study develops a hybrid framework integrating the Analytic Hierarchy Process (AHP) with a multilayer perceptron (MLP) neural network to identify optimal rainwater harvesting sites in Irbid Governorate. Eight spatial conditioning factors were analyzed: slope, land use/land cover, hydrologic soil group, curve number, stream order, drainage density, lineament distribution, and runoff potential. The AHP consistency ratio (0.07) confirms reliable expert judgment. The MLP model demonstrated strong predictive performance, achieving 88% accuracy, 86% precision, 87% recall, and an F1-score of 0.86 under cross-validation. Spatial validation using 127 field-verified structures showed 83.5% agreement, confirming the model’s ability to capture real-world siting patterns. Results indicate that approximately 83% of the basin is suitable for rainwater harvesting, with 52% classified as highly suitable (suitability ≥ 0.70). Integration with drainage networks identified 35 optimal locations, including 15 check dams along major drainage corridors and 20 percolation ponds in agricultural lowlands. Feature importance analysis highlights slope (25%) and hydrologic soil group (22%) as dominant controls, consistent with hydrological theory. The resulting spatial configuration supports a multi-scale water harvesting strategy aligned with integrated water resources management principles. This study provides field-ready, prioritized locations and a robust decision-support framework to enhance water security and support sustainable resource management in water-scarce regions.