A publication in Scientific Data

Improve the prediction of fluvial dike failures

The HECE research group (Urban & Environmental Engineering Research Unit / School of Engineering) publishes a database based on 54 laboratory experiments simulating fluvial dikes failures. The publication of this data in the journal Scientific Data (1) is the result of a vast research project that aims to make this data available to other teams in order to better understand a phenomenon whose consequences can be particularly dramatic.


iver dikes are structures commonly used to protect people and property from the risk of flooding. Facing an increase in extreme rainfall, some structures are subject to aging and are exposed to the risk of overflow, the moment when water is evacuated by passing over the dike. These phenomena weaken the dikes to the point where they fail.

The recent disaster at the Brumadinho tailing dam in Brazil is an example of a dam failure that has had catastrophic consequences. More than 300 people have been reported missing and the environmental impact is significant and difficult to quantify. "Monitoring these structures is necessary, but without any action it will be useless. It is therefore urgent to improve our ability to predict and avoid the failure of these structures”, explains Professor Benjamin Dewals, a researcher in the HECE group - Hydraulics in environmental and civil engineering - of the University of Liège. While there is extensive documentation on the design, construction, operation and maintenance of these structures, there are few databases shared with the scientific community that would allow researchers to better understand the mechanisms that support the formation of breaches in these dikes, in order to anticipate the risks of failure and plan the necessary interventions.


Improve the prediction of fluvial dike failures

Research carried out by the HECE laboratory (EU Research Unit / Faculty of Applied Sciences).

Carrying out detailed on-site measurements at the time of a dam failure is almost impossible. It would also be extremely dangerous. This is why scientists use physical and numerical modelling to measure and predict the behaviors that might be expected on a specific type of structure. "Here at the University of Liège, we are particularly interested in these questions by developing laboratory models on which flood scenarios are reproduced for the purpose of understanding, and for prevention above all. ", explains Sébastien Erpicum, head of the Construction Hydraulics Laboratory. The group of researchers has just published a set of data obtained from 54 laboratory experiments on fluvial dikes breaching due to overflow. In partnership with the Laboratoire d'Hydraulique Saint-Venant (Paris), these data were collected on two complementary experimental configurations, each composed of a main channel representing the river, an erodible lateral dike and a flood plain. "The data set covers seven series of tests," says Sébastien Erpicum, "involving hydraulic configurations with variable boundary conditions (inlet flow, downstream flow, etc.), main channel dimensions, as well as bottom and dike materials." Experimental data such as time series of levels and flows in the main channel and through the breach, as well as high-resolution 3D reconstructions of the geometry of the breach during its expansion have been produced.

These last measurements were made using a new non-intrusive laser profilometry technique developed for this research. "The reuse of the data collected will support efforts to improve our understanding of the physical processes underlying the failure of river dikes. It will also enable assessing the accuracy of numerical models, conceptual or detailed, for predicting dike failure, which is essential for flood risk management”, concludes Benjamin Dewals.

The data from these experiments will also advance the WOLF software, a unique tool for rainfall-runoffmodelling, flood risk analysis, hydraulic structure design, sediment transport analysis and morphodynamic modelling, as well as air entrainment modelling and contaminant transport by surface runoff (https://www.uee.uliege.be/wolf), which allows computer simulation of various flood scenarios and their potential consequences.

Scientific reference

Ismail Rifai, Kamal El Kadi Abderrezzak, Sébastien Erpicum, Pierre Archambeau, Damien Violeau, Michel Pirotton & Benjamin Dewals, Flow and detailed 3D morphodynamic data from laboratory experiments of fluvial dike breaching, Scientific data, 6, 2019.

Read also : Open data could have helped us learn from another mining dam disaster

Share this news