Cases of large forecast errors over Iceland

Forty-eight hour numerical forecasts during a period of 5 years are studied with emphasis on cases of false alarms and missed windstorms at 850 hPa. The overall performance of the forecast system is very good. Windstorms from the southwest are very well predicted, there are a few false alarms in southeasterly winds and northeasterly windstorms tend to be underestimated by the forecast model. The false alarms are in many cases associated with fronts, where a slight shift of a position of the weather system in time may give a large difference in the forecasted and observed winds. Yet, the true value of the forecast may be high. We attribute an underestimation in the wind speed in northeasterly windstorms to non-resolved orography, leading to an underestimation of the corner effect SW-Iceland, and possibly to winds that are generated by a pressure gradient at the western side of the Iceland wake.