Scholarlink

<p>Reliable estimation of water availability in ungauged catchments remains a major challenge in water resources planning, particularly in regions with limited hydrometric observations. Accurate assessment of water yield is essential for reservoir planning, irrigation development, hydropower projects, and climate-resilient water management. This study presented a comparative framework for regionalization-based water yield estimation using three complementary approaches: (i) the GR2M conceptual rainfall-runoff model combined with inverse distance weighting (GR2M-IDW), (ii) rainfall-runoff regression relationships developed at the monthly scale, and (iii) machine learning algorithms for predictive modelling. Hydrological and climatic datasets from 33 gauging and discharge sites distributed across the Betwa, Chambal, Ken, Sindh, Son, and Tons sub-basins of the Ganga River System were analysed. The GR2M model was calibrated and validated at each site, followed by parameter transfer through spatial interpolation. Regression models were developed for monsoon months, while several machine learning algorithms were evaluated for runoff prediction. The GR2M-IDW framework demonstrated satisfactory transferability, with Nash-Sutcliffe efficiency values reaching 93.4% in hydrologically homogeneous catchments, although performance decreased in physiographically complex regions. Monthly rainfall-runoff regression models produced coefficients of determination ranging from 0.59 to 0.82, indicating their suitability for practical engineering applications. Among the machine learning techniques, Gradient Boosting and Huber Regressor exhibited the highest predictive performance, with accuracies exceeding 88% across several basins. The results indicate that conceptual modelling coupled with regionalization provides a physically meaningful and operationally practical solution for ungauged basins. Regression approaches offer transparency and ease of implementation, whereas machine learning techniques enhance predictive capability and scalability. The proposed framework provides a useful basis for water yield assessment and project formulation in data-scarce river basins.</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

Bridging Data Gaps in Water Resources Planning: Comparative Regionalization for Ungauged Basins

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • R. K. Jaiswal,
  • R. V. Galkate,
  • S. Mandloi,
  • S. Indwar,
  • Pushpanjali Kumari

摘要

Reliable estimation of water availability in ungauged catchments remains a major challenge in water resources planning, particularly in regions with limited hydrometric observations. Accurate assessment of water yield is essential for reservoir planning, irrigation development, hydropower projects, and climate-resilient water management. This study presented a comparative framework for regionalization-based water yield estimation using three complementary approaches: (i) the GR2M conceptual rainfall-runoff model combined with inverse distance weighting (GR2M-IDW), (ii) rainfall-runoff regression relationships developed at the monthly scale, and (iii) machine learning algorithms for predictive modelling. Hydrological and climatic datasets from 33 gauging and discharge sites distributed across the Betwa, Chambal, Ken, Sindh, Son, and Tons sub-basins of the Ganga River System were analysed. The GR2M model was calibrated and validated at each site, followed by parameter transfer through spatial interpolation. Regression models were developed for monsoon months, while several machine learning algorithms were evaluated for runoff prediction. The GR2M-IDW framework demonstrated satisfactory transferability, with Nash-Sutcliffe efficiency values reaching 93.4% in hydrologically homogeneous catchments, although performance decreased in physiographically complex regions. Monthly rainfall-runoff regression models produced coefficients of determination ranging from 0.59 to 0.82, indicating their suitability for practical engineering applications. Among the machine learning techniques, Gradient Boosting and Huber Regressor exhibited the highest predictive performance, with accuracies exceeding 88% across several basins. The results indicate that conceptual modelling coupled with regionalization provides a physically meaningful and operationally practical solution for ungauged basins. Regression approaches offer transparency and ease of implementation, whereas machine learning techniques enhance predictive capability and scalability. The proposed framework provides a useful basis for water yield assessment and project formulation in data-scarce river basins.