Uncertainty Analysis in the Estimation of River Discharges Based on Water Levels and Gradually Varied Flow Modeling

Effective water resource management and risk mitigation in urban environments require an accurate understanding of river flow dynamics. Traditionally, river discharge estimation relies on the rating curve method, which, despite being simple and low-cost, presents significant limitations in complex channels or under transient flow regimes. Aiming for greater accuracy in discharge estimates, this study employed a one-dimensional hydraulic model based on the Standard Step Method for gradually varied flows. High-quality in situ data, including detailed bathymetric surveys and water level measurements, were used for model calibration and validation. Discharge uncertainty was quantified through the estimation of hydraulic parameters (S₀ and n), comparing the Bayesian approaches GLUE and DREAM. The final uncertainty ranges in discharge estimates (Relative Error) using the DREAM method varied between 6.75% and 7.34%, while the GLUE method presented ranges between 7.31% and 8.64%. The DREAM method demonstrated superior performance, including faster processing speed. It was also observed that model accuracy tends to improve with decreasing water surface elevation, indicating the need for longer data collection intervals during extreme events.