The performance of a global coupled atmosphere-ocean general circulation model, HadCM2, in simulating present-day climate in Sweden and Northern Europe is studied. The study is motivated by the use of HadCM2 within the SWEdish regional CLimate Modelling programme SWECLIM. In particular, HadCM2 provides the boundary data for the first regional high-resolution climate simulations conducted at the Rossby Centre.
HadCM2 is probably one of the best present global climate models, and several aspects of the simulated control climate are in reasonable agreement with observations. However, some quantitative shortcomings are also evident. In Northern Europe, the simulated summer temperatures are a few degrees colder than those observed, and their interannual variability appears too large. In winter, a local warm bias occurs around the Baltic Sea, which is too warm to get an ice cover in the model. The effect of the Scandinavian Mountains on the distribution of precipitation is distorted by the model's modest (2.5° lat x 3.75° lon) resolution. The simulated total cloudiness generally exceeds that observed.
In comparing the model results with the records of individual climate stations, the first problem is to derive the model-simulated value at the station location from the original discrete grid box values. Several choices of doing this are examined and, at least in comparison with observed temperature and precipitation at Swedish stations, the conclusions frequently depend on how the model-simulated station value is defined.
It is also studied how the climate in the 10-year HadCM2 control run used for dynamical downscaling at the Rossby Centre differs from a longer (240-year) control run made with the same model. In most cases, the differences between the 10-year and 240-year runs are smaller than those between the former and observations. Nevertheless, some of these differences are large enough to be kept in mind when comparing the 10-year control run with, for example, a similar run with increased CO2.