This work investigates the predictability of seasonal to inter-annual streamflow over several river basins in Norway through the use of multi-model ensembles. As GCMs do not explicitly simulate streamflow, a statistical link is made between GCM-forecast fields generated in December and average streamflow in the melting season May-June. Two models are tested for recalibration methods linking forecasted large scale circulation and moisture fields to streamflow. At the same time, various multi-model ensemble combination schemes are tested. As the inherent variability of the atmosphere requires, the seasonal forecasts are expressed probabilistically. The multi-model forecast is tested against climatology to determine the quality of the forecast.