75. Comparison Between Two Hydraulic Models (1D and 2D) of the Garonne River: Application to Uncertainty Propagations and Sensitivity Analyses of Levee Breach Parameters

In recent years, flooding hazard is usually assessed through numerical modelling. However, depending on the nature (e.g. 1D, 2D) and the breach characteristics (e.g. river geometry, bottom roughness, levees geometry) of the numerical model, the uncertainties on the corresponding parameters should be taken into account in a rigorous way, for improving the assessment of the simulated flooding hazard. In fact, levee behaviour during a flooding event is one of the major sources of uncertainties impacting the water level at a given location.

In this context, the objective of our work is to better understand the impact of uncertain parameters related to levee breaches, on the generated overflows, through Uncertainty Quantification (UQ) and Global Sensitivity Analysis (GSA) of these parameters.

With this purpose, two numerical models of the Garonne River were built and validated, between Tonneins and La Réole sections (for a river length of nearly 50 km): a 1D hydraulic model with storage areas, developed with HEC-RAS and a 2D model with TELEMAC-2D. These modelling approaches (1D and 2D) are classically used to carry inundation studies. Moreover, the simulated river reach is of interest as protected by a levee system to reduce the flood risk. These levees have been damaged during flood periods, by physical mechanisms as erosion due to overtopping for instance, such as during the 1981 historical flood event. The study evaluates the influence of levee breach parameters (breach triggering parameter, breach length and breach depth) on the maximum water level at four points located within the upper part of the study area, through UQ and GSA. These approaches are carried out with a meta-model built with 200 simulations runs using a Monte-Carlo approach for both models. In both cases, the breach parameters are uniformly distributed and randomly sampled in order to generate a large number of breach scenarios.

Globally, the Monte-Carlo and FAST (Fourier Analysis Sensitivity Test) analyses performed have shown some differences between the results coming from both meta-models and between the upstream and the downstream storage areas, more sensitive to levee breaches. These analyses also indicated the slight effect of the breach length parameter contrary to the triggering and depth breach parameters.