Walsh, J.E., V. M. Kattsov, W. L. Chapman,V. Govorkova, and T. Pavlova, 2002:
Comparison of Arctic climate simulations by uncoupled and coupled global models
Journal of Climate, 15, 1429-1446.

Simulations of present-day Arctic climate are assessed from suites of (1) thirteen global atmosphere-only models from the Atmospheric Model Intercomparison Project (AMIP-II) and (2) eight coupled atmosphere-ocean-ice models from the Data Distribution Center of the Intergovernmental Panel on Climate Change (IPCC). The assessment highlights the impact of coupling on the simulated Arctic climate, and also the improvement of the uncoupled models relative to a previous (early 1990s) phase of the AMIP project.  The across-model variance of the simulated air temperature is larger in the coupled models than in the uncoupled models, and the spatial pattern of the variance indicates that differences in the coupled models' simulated sea ice contribute to the larger variance of temperature.  The coupled models are also several degrees warmer than the uncoupled models during the winter half of the year.  As was the case with the earlier AMIP models, the simulated precipitation still exceeds the observational estimates, particularly over the terrestrial watersheds of the Arctic Ocean.  The bias is larger in the coupled models and is strongest during the cold season.  Both the coupled and the uncoupled models suffer from a bias of Arctic sea level pressure that will adversely impact the simulated sea ice motion and the ice thickness distribution.  The bias appears as a shift of mass from the Beaufort sector of the Arctic Ocean to the Asian coastal seas.  Improvements in simulated cloud coverage from AMIP-I to AMIP-II are apparent in a reduction of the across-model scatter of the AMIP-II cloud coverage and also in a more realistic annual cycle of the cloud fraction composited over the AMIP-II models.  The Arctic surface radiative fluxes vary widely among the AMIP-II models, especially under cloudy skies.