Comparison of Streamflow Estimated by Image Analysis (LSPIV) and by Hydrologic and Hydraulic Modelling on the French Riviera During November 2019 Flood

Short and heavy rainstorm events often lead to flash floods on the French Riviera coastal catchments: during the historical 2nd October 2015 flood, a peak streamflow value between 185 and 295 m3/s was estimated on the Brague River at Biot at 10:30 P.M., while the streamflow was around 1 m3/s at 6:30 P.M. at the same section. If the measurements of such streamflow values are highly important (for flood statistical analysis, flood modeling, hydraulic structure design), such measurements are dangerous when they require an operator to manipulate an instrument in or near the river. Alternative methods can be used, such as video analysis, by analyzing a sequence of images and locating the displacement of patterns on the water surface. Thus, the velocity field at the flow surface can be determined and then used for estimating flow discharge on specific cross sections. In this work, we applied two different Large-Scale Particle Image Velocimetry (LSPIV) algorithms (Fudaa-LSPIV and OpyFlow) to several videos of the November 2019 floods within the Brague catchment, in order to estimate streamflow values. The obtained streamflow values have then been compared to values estimated through available observations, and also to the results of (i) rainfall-runoff modeling and (ii) hydraulic modeling on the same sections. Both LSPIV estimations, rainfall-runoff simulations and observations are coherent on the studied sections, showing the interest of combining such different and independent techniques in order to estimate flood streamflow values.