Atmospheric flow over Iceland has been simulated for the period January 1961 to July 2006, using the mesoscale MM5 model driven by initial and boundary data from the ECMWF. A systematic comparison of results to observed precipitation has been carried out. Undercatchment of solid precipitation is dealt with by looking only at days when precipitation is presumably liquid or by considering the occurrence and non-occurrence of precipitation. Away from non-resolved orography, the long term means (months, years) of observed and simulated precipitation are often in reasonable agreement. This is partly due to a compensation of the errors on a shorter timescale (days). The probability of false alarms (the model predicts precipitation, but none is observed) is highest in N Iceland, particularly during winter. The probability of missing precipitation events (precipitation observed but none is predicted by the model) is highest in the summer and on the lee side of Iceland in southerly flows.
Keywords: dynamical downscaling, Iceland, MM5, QPF, rain gauge data, validation