Résumé
The first-order second moment (FOSM) method is widely used inuncertainty analysis. This method uses a linearization of the function that relates theinput variables and parameters to the output variables. This simplification occasionallyleads to problems when the mean value of the input variable is close to a local orglobal maximum or minimum value of the function. In this case, the FOSM computesartificially a zero uncertainty because the first derivative of the function is equal tozero. An improvement to the FOSM is proposed, whereby a parabolic reconstructionis used instead of a linear one. The improved FOSM method is applied to a floodforecasting model on the Loire River (France). Verification of the method using theMonte Carlo technique shows that the improved FOSM allows the accuracy of theuncertainty assessment to be increased substantially, without adding a significantburden in computation. The sensitivity of the results to the size of the perturbation isalso analysed.