Required model input data are not always completely available and model structures are only a crude description of the underlying natural processes; therefore, model parameters need to be calibrated. Different model concepts (interflow, direct runoff), or rather the processes represented, such as infiltration, soil water movement, etc. are more or less dominating different sections of the runoff spectrum. Most models do not account for such transient characteristics inherent to the hydrograph. This falls together with uncertain input data (e.g. rainfall intensity on different scales and the rainfalls' spatial distribution, especially if rainfall is a predicted parameter). In this paper, we try to show a way towards a possible online evaluation of model uncertainty and the graphical means to communicate the uncertain forecast in a probabilistic but yet comprehensible way. This is based on the development of a flood forecasting system, which combines artificial neural networks with process models Monte-Carlo evaluation of the forecast uncertainty. The Monte-Carlo results are evaluated in an operationally applicable model and the uncertainty is colour coded in uncertainty plots of the forecast, which are updated at each time step. This provides operators in flood warning centres much more reliable and easy to interpret information about the probability of the forecast well in advance of the occurrence.