Author(s): Tanmoy Das; Subhasish Das
Linked Author(s):
Keywords: Bias correction; GloFAS; XGBoost; Rainfall runoff; HEC-HMS; Flood
Abstract: Global-scale reanalysis products have created new opportunities for streamflow estimation in ungauged basins, yet hydrological model outputs still suffer from significant uncertainties due to systematic biases. In this study, we applied an extreme gradient boosting (XGBoost) machine learning model to correct bias in GloFAS discharge data and evaluated the potential of the improved GloFAS discharge as a calibration and validation benchmark for the rainfall-runoff model. Three input scenarios were evaluated to achieve the optimal configuration for XGBoost. The study was conducted in a poorly gauged and flood-prone river basin, Shilabati, in eastern India. Results indicate that neither GloFAS discharge (Scenario 1) nor engineered rainfall features (Scenario 2) as model input can capture the complex non-linearity in observed discharge. Incorporating time series of lag and rolling window statistics of GloFAS discharge as model input (Scenario 3) substantially improves the predictive performance of XGBoost by capturing a richer hydrological memory and flow dynamics. The well-improved GloFAS discharge also performs well as a calibration and validation benchmark in the HEC-HMS model for event-based flood estimation, indicating its reliability for rainfall-runoff modelling.
Year: 2026