Uncertainty propagation in a sequential model for flood forecasting

The aim of this paper is the estimation of uncertainty in an online data assimilation model applied to a sequential, multiple-step-ahead flood forecasting system. The main aim of the forecasting system under consideration is the derivation of real-time forecasts of the water levels with the maximum possible lead-time. This is achieved through a two-level, sequential data assimilation procedure. In order to extend the maximum lead-time, we incorporate the forecasts obtained from the earlier stages of the forecasting system, both rainfall-water level and water level routing processes. The updating of the gain of each of the subsystems introduces nonlinearity into the system performance. The Generalized Likelihood Uncertainty Estimation (GLUE) technique is used to estimate the uncertainty of model predictions in the decomposed online forecasting system.