Framing uncertainties in flood forecasting with ensembles

Abstract

Under the assumption of rational decision making flood forecasts produce economic benefits only if their application reduces the uncertainties of future developments. In general a forecasting system cannot provide the exact future value of the predictand. There are many different sources of uncertainties. In this chapter tools are presented which can be used to characterize them. It is focused on ensemble methods. Ensembles do not only provide descriptions of uncertainties. They can be combined with data assimilation to produce best guess forecasts based on the Bayes' theorem. The forecasting system can be adapted indirectly to the actual state of information by selecting an ensemble member which is in good agreement with observed data. A widely used approach, based on this methodology, is the Ensemble Kalman Filter (EnKF). The EnKF can be used to update state variables of hydrological models. In addition meteorological ensembles forecasts and parameter ensembles are discussed. A case study, demonstrating the applicability of ensemble forecasts, closes this chapter. © 2011 Springer Science+Business Media B.V.