Abstract: ECMWF is currently running analysis and forecast ensembles. The Ensemble of Data Assimilation (EDA) provides background error statistics to the high-resolution 4D-VAR, and is used to generate the initial conditions of the perturbed members of the medium-range/monthly ensemble (ENS). ENS and the seasonal system-4 (S4) provide users with coupled ocean-atmosphere forecasts of the most likely scenario and its confidence, expressed for example in terms of probabilities of weather events. In this talk, the key characteristics of these three ensembles (EDA. ENS and S4) will be illustrated and their average performance briefly reviewed. The value of ensemble-based products to estimate forecast confidence in extreme weather events will also be discussed.