Calibration of river hydrodynamic models: Analysis from the dynamic component in roughness coefficients

Abstract

Hydrodynamic modelling is an important tool for several studies, such as flood forecasting and analysis of energy and mass transport. In rivers, the most common calibration parameter is the roughness coefficient, that accounts for the friction term. However, while uncertainties are associated to this parameter definition, it is often defined as constant over time or space. Even though sophisticated techniques are available for model calibration, unsteady flow routing applications usually involve simplified methods, such as trial-and-error and Monte Carlo strategies. These practices can be subjective, time consuming and dependent on user experience; on the other hand, automatic approaches may generate unrealistic values and require large computational efforts. This study introduces a new strategy to calibrate one-dimensional hydrodynamic models, accounting for temporal and spatial variation while maintaining computational thrift. Among multiple random roughness coefficients, a time series is generated according to a linear relationship with observed water depths or flow for five control sections; the resulting daily series of parameters are the set for simulation inputs. The algorithm was combined with the hydrodynamic module in the SihQual model (Hydrodynamic and Water Quality Simulation); when compared to observed data, it showed to be suitable to predict discharges and water levels for different years over 85 km of the Iguaçu river, located in Paraná (Brazil). Compared to Monte Carlo and trial-and-error techniques for calibration, the HTC method (Hydrodynamic Temporal Calibration) was able to match their results or even improve simulations in some cases, requiring less user interaction than the other approaches. The study highlights the relevance of including a temporal component in model calibration, aiming to reduce the overall inherent uncertainty of this process.