On the adaptive component of a real-time flood forecasting model

An adaptive semi-distributed model for real-time flood forecasting in small and medium basins is investigated. The model uses a simulation component, which represents both the rainfall-runoff process and the routing of the observed discharge in an upstream section, and an adaptive component which relies upon the adjustment of initial soil water content at successive time origins of forecast. The model was tested over three basins in Central Italy ranging from 136 Km(2) to 2035 KM2. The model effectiveness is fairly good, as shown by analysing the errors on the prediction of time to peak characteristics and the values of the persistence coefficients. Furthermore, the model structure seems to be appropriate, as indicated by the behaviour of the evolution in time of the initial soil water content which exhibits a very limited range of variability when adapted in real-time.