In February and March 1996, the French NWP model Arpège was run to give operational forecasts five days ahead. These forecasts were a part of preparations for FASTEX (Fronts and Atlantic Storm-Track Experiment), whose field phase took place in 1997. The extensive collection and availability of model output during this period gave an excellent opportunity to study in real time the behaviour of different NWP models during the period of the year when strong winds and rapid changes in weather conditions are frequent. In this paper, we study a case where the ECMWF model predicted an incorrect pressure gradient at Northeast Iceland, while Arpège correctly simulated the gradient. The model errors are connected to anomalies in the mid-tropospheric vorticity field, which is in general quite different in the two models. The differences in the vorticity fields are discussed in relation to orographic effects.