Document Type


Publication Date



Modelling floodplains adequately is crucial for numerous water management applications. Many of these applications require a large number of simulations or long term analyses, such as optimization problems at catchment scale, uncertainty analyses or real time control of hydraulic structures to prevent flooding. Therefore, models with a limited calculation time are necessary. Various computationally efficient models exist that describe the flow in rivers, but the modelling of floodplains is often overlooked. This study focuses on the computationally efficient lumped modelling of floodplains. Two different data-driven modelling approaches are proposed that predict the inundation level in the floodplain and the flow between the river and the floodplain for a given set of river stages. On the one hand, the flow is estimated using a set of optimized weir equations that are deemed to describe the flow over hydraulic structures. On the other hand, hybrid models are studied that combine physical principles with artificial neural networks. Such networks are highly flexible and can learn from the data provided during the model set-up procedure. Two set-up procedures are investigated to ensure that physically sound networks with good generalization capabilities are obtained. The derived methodologies are tested on a case study by emulating the results of a detailed full hydrodynamic model of the Dender River in Belgium. The results show that the hybrid models can predict the water levels and flows in/to floodplains with great accuracy. The novel modelling approach that incorporates artificial neural networks outperforms the conceptual models with rigid equations. Both types of models are more than 10.000 times faster when simulating events compared to a 1D hydrodynamic model that solves the full de Saint-Venant equations. Because the model build-up requires expert knowledge and can be time-consuming, a tool was developed that assists the modeller during the calibration procedure. When simulation results from detailed models and information on the detailed model lay-out are presented, the tool suggests an optimal model structure and accompanying set of parameters.


Session S5-01, Special Session: Computational Intelligence in Data Driven and Hybrid Models and Data Analysis I



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.