ABSTRACT
ABSTRACT: Numerical Rainfall-runoff (RR) models are used for the prediction of river floods and form an important component of Flood Early Warning Systems (FEWS). RR-models often contain large uncertainties due to model errors and errors in input data. In RR-models the uncertainties in the input are often dominant compared to those of the model. Data assimilation is an efficient and often applied method to reduce those uncertainties and improve the model’s performance. This paper describes a data assimilation technique used to reduce the uncertainty in the input (i.e. rainfall data) in the RR-model HYMOD. The data assimilation technique used is the Ensemble Kalman filter (EnKF). One of the main advantages of the EnKF is that it is a generic method and is also suitable for non-linear models. Here, the feasibility of the EnKF as a tool for input correction is investigated. The present results show that the EnKF is capable of reducing the uncertainties in the rainfall by 10% and those in the runoff by about 50%. The runoff forecasted by the model after the assimilation showed improvements for the first 8 to 10 days.
