Quantifying the discharge forecast uncertainty by different approaches to probabilistic quantitative precipitation forecast

Abstract. A probabilistic approach to flood prediction over the Reno river basin, a medium-sized catchment in Northern Italy, has been tested using two different meteorological ensemble systems. The future precipitation scenarios are provided either by an analogue-based technique (statistical approach) or by a limited-area ensemble prediction system (dynamical approach), then used as different inputs to a distributed rainfall-runoff model. The ensemble of possible future flows so generated allows to convey a quantification of uncertainty about the discharge forecast. The probabilistic discharge forecasts, based on the precipitation forecast provided by the two ensembles, are then compared to the deterministic one obtained by the rainfall-runoff model fed on precipitation input provided by a non-hydrostatic meteorological model, run at 7km of horizontal resolution. For this case study, the dynamical approach appears to be more feasible in providing useful discharge ensemble forecast than the statistical one, because the observed large spread among members obtained with the analogue method makes difficult to issue real-time flood warnings.