Propagation of discharge uncertainty in a flood damage model for the Meuse river

Uncertainty analysis plays an important role in the decision- making process. It can give decision makers a complete idea of how different measures will affect the whole river system. Thus it helps decision makers to make a better choice among measures in a more systematic manner. In case of flood damage reduction projects, uncertainty analysis helps to evaluate the main decision criterion – expected annual damage. The aim of this paper is to investigate the propagation of discharge uncertainty, which is one of the main uncertainty sources in a damage model, into expected annual damage. The discharge uncertainty considered here includes model uncertainty (choice of different probability distributions) and sampling errors due to finite gauge record lengths. The calculated uncertainty in the discharge varies between 17 percent for a return period of 5 year and 30 percent for a return period of 1250 year. A first order method is used here to explore the role of discharge uncertainty in the expected annual damage model. The results from the damage model indicate that both model uncertainty and sampling errors are important, with the latter being somewhat more important. The Log-Pearson Type 3 gives a much smaller uncertainty range of the expected annual damage than the other three distribution models used. The uncertainty is aggravated when propagated into the damage results. The uncertainty in the damage reduces a great amount when the sample size increases to n=80. The results derived from the first order method in fact give two bounds of uncertainty, which is an overestimate in this case.