A method is presented to compare the statistical properties of surface air temperature, sea surface temperature, precipitation, global radiation, 500 mbar height, sea level pressure, and wind by a GCM with those of observations. As an illustration the control run of a UKMO 11-layer GCM for five successive winters (DJF) is compared with observations for NW Europe. In this comparison large differences are found in (monthly) mean values, standard deviations and autocorrelation coefficients for various elements.
In general, this GCM winter run creates too low temperatures over land, too high temperatures over sea, large temperature variability but small pressure variability and too small autocorrelation coefficients for time-lags of more than three days. Too many wet days are created with too little precipitation. The 500 mbar circulation is veered with respect to reality and shows a peculiar through over the North Sea. The wind at 100 m is backed and the geostrophic wind is somewhat underestimated. No realistic transition of the surface temperature from land to sea is observed.
These results suggest that care has to be taken in interpreting direct model output on a regional scale such as NW Europe. Given the limitations of this study, the relatively short GCM run and difficulties related to comparing GCM grid points with station observations, further work along these lines is desirable, preferably on the basis of the output of more recent GCM versions.
JJ Beersma. GCM control run of UK Meterological Office compared with the real climate in the NW European winter
KNMI number: WR-92-02, Year: 1992, Pages: 32