The parsimonious time series models used within the Data-Based Mechanistic (DBM) modelling framework have been shown to provide reliable accurate forecasts in many hydrological situations. In this work the DBM methodology is applied to forecast discharges during a flash flood in a small Alpine catchment. In comparison to previous work this catchment responds rapidly to rainfall. It is demonstrated, by example, that the use of a radar-derived ensemble quantitative precipitation forecast coupled to a DBM model allows the forecast horizon to be increased to a level useful for emergency response. A treatment of the predictive uncertainty in the resulting hydrological forecasts is discussed and illustrated.