Mid-July 2021 unprecedented torrential rain fall across parts of Western Europe, claiming more than 200 lives and causing tens of billions of euros’ worth of damage. Exactly five years to the day after the disaster, researchers at the HECE (Hydraulics in Environmental and Civil Engineering) laboratory at the University of Liège are taking stock of a long-term scientific project. Reconstruction of the flood, validation of numerical models, the role of log jams beneath bridges, and factors contributing to damage to homes: four major scientific studies, often conducted in collaboration with international partners, have utilised the exceptional data collected in the wake of the disaster.
Triggered by the Bernd low-pressure system, whose advance was blocked by a high-pressure system to the east of the continent, the rainfall in July 2021 remained virtually stationary for several days. The Vesdre and Amblève catchment areas, in eastern Belgium, were among the hardest hit. Verviers, Pepinster and many other municipalities paid a heavy price. For the engineers at ULiège, this tragic event also provided an unprecedented field of study.
As the first detailed hydrological analysis of the Vesdre and Amblève catchment areas, this research reveals the staggering intensity of the event. The estimated peak flow at Eupen exceeded four times the hundred-year return period flow, and some rain gauges recorded, over two days, nearly double the volume associated with a 200-year return period. As most of the measuring stations on the Vesdre had been swept away, researchers Christophe Dessers and Pierre Archambeau had to reconstruct the flood using a runoff model calibrated against dam levels. Their main conclusion is that no single model is sufficient to predict such an extreme event and that it is necessary to combine several approaches and several sources of rainfall data, the quality of which has a significant impact on the results.
It is widely believed that the accumulation of field data guarantees the reliability of numerical models. Following the flood, the Wallonia Public Service mobilised nearly 200 staff to record more than 16,000 high-water marks, one of the most comprehensive post-disaster datasets in the world. However, whilst the WOLF model developed in Liège performs excellently, this study shows that the way in which observed and simulated data are compared can cause validation scores to vary by more than 20 per cent. Pratik Chakraborty and Benjamin Dewals draw a clear lesson from this: the quantity of data is not a miracle solution, as its collection, documentation and processing are just as important.
Log jams - those accumulations of debris that block bridges and exacerbate flooding upstream - were the subject of a systematic survey of 71 bridges spread across six Belgian and German rivers. The findings are striking. Nearly half of the accumulated material was man-made - such as construction waste, vehicles and even caravans - rather than the usual driftwood. On 85 per cent of the structures, the water reached or exceeded the deck level. Led by Sébastien Erpicum in collaboration with TU Delft and RWTH Aachen, this research shows that these denser log jams - which have never been factored into conventional risk models – call for a rethink of bridge design and serve as a reminder of everyone’s responsibility regarding objects that could be swept away.
The factors determining the damage sustained by homes during flash floods - a type of flooding that remains poorly documented - were scrutinised across more than a thousand buildings surveyed in Belgium and Germany. Machine learning analysis confirms the central role of water level, whilst revealing other variables that are often overlooked, such as living space, building type, wall materials, the presence of sediment and the vulnerability of older people. Led by Daniela Rodríguez Castro in collaboration with the GFZ Helmholtz Centre and the University of Potsdam, the study calls for damage models specifically tailored to flash floods.
Taken together, these four studies point to the same conclusion: in the face of floods on such a scale, the tools of the past are no longer sufficient. Reconstructing events, rigorously validating models, taking log jams into account, and detailed damage modelling - each advance refines our understanding of the phenomenon and, above all, is already translating into practical applications: more reliable risk maps, better-designed structures, and more robust forecasts. This research is currently ongoing, and the findings are gradually being incorporated into planning decisions and crisis management measures. At a time when climate change is set to make such extreme events more frequent, it is only on this basis that Wallonia and its neighbours will be able to face the next flood better prepared.
L'édition 2026 de la Nuit des Chercheur·es se tiendra le vendredi 25 septembre 2026. Impliquée dans la Nuit des Chercheur·es depuis sa création, l'ULiège vous invite à participer à cette manifestation de culture scientifique.
nGHOST2Market to track stray light from space telescopes and GrAVITe, an optical gyroscope to stabilise high-precision instruments.
Among the grants awarded by the FRS-FNRS this year, 119 will finance research carried out by scientists at the University of Liège.
