Uncertainty quantification on a real case with telemac-2d

In this study the software TELEMAC-2D (www.opentelemac.org) is used with the OpenTURNS library (www.openturns.org) to quantify the uncertainty on a real hydraulic case. The used approach is based on the chaining of OpenTURNS and TELEMAC-2D using the SALOME platform (www.salome-platform.org) in order to implement a Monte Carlo-like algorithms. Each uncertain parameter (inlet discharge, friction coefficient) is associated to a statistical distribution (defined using OpenTURNS). A sufficient number of TELEMAC-2D runs are achieved with respect to the predefined random entries in order to guarantee the convergence of the studied Monte Carlo-like algorithms. EDF’s cluster has been used to run the simulations. Indeed, to handle the uncertainty with the Monte Carlo method, it is important to run a lot of simulations in order to have reliable results. The obtained results are analysed twofold: On one hand, the effect of variability of random inputs is assessed at some specific points (assumed to be around a fictive point of interest). On the other hand, a global statistical analysis all over the domain is done. A spatial distribution of the mean water depth and its variance is obtained. These results are of utmost importance for dimensioning of protecting dykes. Furthermore, they are very useful when establishing scenarios for flood managing. However, Monte Carlo technique that while generic and robust is also computationally expensive. Ways to lower the cost typically require to replace the pure random sampling that form the backbone of the Monte Carlo method by alternative sampling methods such as the Latin Hypercube Sampling approach and the quasi-Monte Carlo method based on low discrepancy sequence. The present work aims to compare the behavior of these Monte Carlo-like algorithms. This work shows that, thanks to the availability of important computer resources and to an optimized software, we are able to consider Monte Carlo-like algorithms for uncertainty quantification of real hydraulic models. This critical conclusion was, even an unfeasible dream, couple of years ago.