Résumé
We present the study of sensitivity and uncertainty analyses of the flash flood propagation test case,modelled by the 1D shallow water equations, usually viewed as being non linear. The potential of first-order Local Sensitivity Analysis for uncertainty analysis is assessed through a comparison with globalapproach, based on Monte Carlo simulation, via statistical estimators (mean and variance). Consideringthe maximum intensity of the flood as the only uncertain parameter, the difference between outputvariances obtained from local and global remains lower than 20% of global results for an inputvariation of up to plus or minus 80% (corresponding to an input variance of 42% of the nominal value).The sensitivity analysis with respect to four input parameters is then estimated using global indices ofSobol’ and compared to local results. The fact that the two methods provide very close results confirmsthe great advantage of the local approach that requires only three model runs instead of severalthousands for the global approach.