A combined experimental and numerical strategy to assess the influence of model geometric distortion in laboratory scale modelling of urban flooding

Accurate modelling of urban flood hazard remains hampered by a lack of suitable validation data. In contrast with water marks and inundation extent, discharge partition in-between streets and flow fields are generally unknown, while they have a strong influence on flood risk (human destabilization, scour, contaminant transport). Laboratory experiments may provide a valuable complement to field data to achieve robust validation of flood hazard models. However, urban flooding is a multiscale phenomenon, with horizontal length scales (~ 103 m) are considerably larger than the vertical ones (~ 1 m). Therefore, recent experimental studies of urban flooding used geometrically distorted scale models, with a vertical scale factor smaller than the horizontal one. Though, little is known so far on the bias induced by model geometric distortion in the case of urban flooding. To address this issue, we have combined computational modelling with laboratory experiments. Based on 100+ numerical simulations, the bias induced by model geometric distortion on flow depth and discharge partition was found of the order of 10 %, which is not negligible compared to other uncertainties involved in urban flood hazard modelling. Moreover, when the geometric distortion is varied, the induced bias shows an intriguing non-monotonous evolution, which we could relate to a competition between frictional and secondary head losses. New tailored laboratory experiments are on-going on a “series” of laboratory scale models, i.e. several laboratory models representing the same urban layout at various scales. The outcomes of these experiments will make more robust the conclusions drawn from computational modelling.