The evolution towards integral and integrated catchment modelling is impeded by a large number of factors, whereof the two most important are the incompatibility of the existing models and the computational effort needed to run simulations. Incompatibilities arise because of differences between software suppliers, time and space scales, the level of detail, the modes of applicability of the different models and many more. Calculation times of full hydrodynamic models limit the use of integrated catchment models for flood prediction systems, real-time control, uncertainty analysis and other applications.
This report presents a conceptual grey-box modelling approach for river systems that can replace or assist existing hydrodynamic models. These conceptual models try to emulate the results of the detailed hydrodynamic models, based on a concatenation of reservoir-type elements. Their explicit calculation scheme makes them very useful for integrated catchment modelling and for applications that require long term or a large number of simulations, like forecasting, real-time control, sensitivity analyses etc.
The report first gives an overview of all selected conceptual model components that are used to calculate water quantity variables like water levels, discharges, depths and velocities. Almost all model components are based on interpolation tables to acquire sufficiently low calculation times. Subsequently, the calibration of the conceptual model components is discussed. The available data from measurement points and gauging stations is usually insufficient to calibrate conceptual models. Therefore, use is made of the simulation results of detailed fully hydrodynamic models (in this case: MIKE11 models). The procedure to transform a detailed hydrodynamic into a lumped conceptual model is described.
The presented conceptual modelling approach is tested and demonstrated on two case studies: the rivers Dender and Zenne, located in the central part of Belgium. Both models include the main river from the Walloon border until the confluence with the tidal river Scheldt, and the main tributaries. The simulation results of both models were compared and it is shown that the surrogate models have only minor loss of accuracy, compared to the detailed models, and an important gain in calculation time.
Finally, the models were subjected to a number of robustness tests to investigate how they react outside their calibration range. The results show that the model cannot handle important changes to the parameters of the hydraulic structures, but is capable of producing reliable results when the gate regulations are changed or when more extreme events are simulated.