The Influence of Rating Curve Uncertainty on Flow Conditions in the River Vistula in Warsaw

The flow forecasting system for the middle reach of the River Vistula consists of a distributed flow routing model and rainfall-runoff modules. The MIKE 11 flow routing model, applied in the study, requires specification of a number of input variables, including roughness coefficients, initial conditions, river geometry and boundary conditions. All listed variables introduce errors into the system that have an impact on the uncertainty of flow predictions. This work focuses on the estimation of uncertainty of flow predictions resulting from inaccuracy of boundary conditions and parametric uncertainty. In particular, we are interested in the influence of a stage-discharge relationship (rating curve) uncertainty applied as the downstream boundary condition. The influence of parameter and rating curve (RC) uncertainty on model predictions is studied using the Generalised Likelihood Uncertainty Estimation GLUE framework. We parameterised downstream RC using a power law. The RC parameters were optimised together with the model roughness coefficients. The optimal parameter values were subsequently used as mean values of a priori distribution within the GLUE approach to derive the flow routing model prediction uncertainty. The results were superior to those without RC uncertainty taken into account only in Warsaw Port Praski gauging station. The proposed methodology was applied to the Middle Vistula reach between Zawichost and Warsaw Port Praski gauging stations. Model performance was estimated using both water level and discharge observations at gauging stations situated along the river reach.