As part of the AMIP polar diagnostic subproject, the surface air temperatures and energy budgets of 19 models have been examined. Over Greenland and the Arctic Ocean, the models' annual mean temperatures vary by up to 10 degrees C, 1995: the Greenland surface air temperatures are biased by the models' smoothing of the topography. Over the northern land areas (poleward of 60°N) and Antarctica, the range among models is as large as 15°C. The models capture the latitudinal and seasonal variability of Arctic temperatures, although a cold bias of 1-3°C is apparent over the northern continents, especially during spring over Eurasia. A warm springtime bias over the Arctic Ocean is smallest in those AMIP models in which the prescribed albedo of sea ice is highest. Summer temperatures over the Arctic Ocean are negatively correlated (r=-0.48) with the mean total cloud fractions of the models. However, there is little correlation between total cloudiness and mean surface air temperature during winter. In most models, the interannual variations of the simulated and observed polar temperature show little correlation, although one model (the DERF model) developed for long-range forecasting shows skill over northern Eurasia.
The components of the surface energy budget vary substantially from model to model. Over the Arctic Ocean, the annual mean shortwave radiation ranges among models from 30 to 60 Wm-2, 1995: the net surface longwave radiation ranges from -20 to -65 Wm-2 , 1995: the annual mean sensible heat flux ranges from -15 to +6 Wm-2 , 1995: and the annual mean evaporation ranges from 0.15 to 0.75 mm day-1. The model-to-model range of several of these quantities is even greater over Greenland and Antarctica. The longwave radiative and sensible heat fluxes tend to compensate the model-to-model differences of incoming shortwave radiation at the surface. The surface shortwave radiation, in turn, correlates negatively with the models' spring and summer cloud fractions. The radiative flux differences among the models are large enough to produce sea ice thickness changes of more than 10 cm yr -1 when used to force a thermodynamic sea ice model.