基流
水流
水文气象
水文学(农业)
环境科学
水资源
季风
流域
基流
水文模型
水资源管理
统计
缺水
地下水
估计
样品(材料)
气候变化
作者
Raj Pal Singh,Hiren Solanki,Vimal Kant Mishra
摘要
Baseflow is critical for maintaining river flows, supporting ecosystems, and meeting water demands during dry periods. Yet, large‐scale estimation of baseflow remains challenging due to the lack of direct observations. Here, we present HyGBR‐Baseflow, a machine learning framework that estimates baseflow by anchoring the learning to reference points (RPs) and read directly from streamflow recession. The framework uses four graphical and five digital filter methods as predictors, together with hydrometeorological variables. To address the scarcity of reliable training data during monsoon season, our framework uses pseudo‐reference points (PRPs), derived from the mean baseflow of the three best‐performing baseflow separation methods and assigns reduced sample weights (0.2–0.5) compared to true RPs. We applied HyGBR‐Baseflow to 227 stations across 18 major Indian River basins for 1961–2021 and evaluated its performance using Nash–Sutcliffe Efficiency, Kling–Gupta Efficiency, refined index of agreement, and R 2 . Overall, the HyGBR‐Baseflow framework consistently outperformed individual baseflow separation methods, with the largest improvements during the monsoon season across diverse hydrological regimes. The baseflow trend analysis showed pronounced declines in several catchments. Even the Ganga and Brahmaputra basins, which comprise catchments where baseflow contributes more than 70% of the total streamflow at most stations, showed declining baseflow trends. Additionally, the long‐term baseflow index indicated higher baseflow contributions in humid catchments compared to semi‐arid and dry sub‐humid regions. The HyGBR‐Baseflow provides a scalable, accurate, and observation‐anchored nonlinear reconciliation framework for baseflow estimation that can support groundwater–surface water interaction studies, water resources planning, drought monitoring, and climate change impact assessments across India.
科研通智能强力驱动
Strongly Powered by AbleSci AI