Simulations of an atmospheric general circulation model run with sea surface temperature and sea ice amounts prescribed to observed values for the period 1979-1994 are compared with "control run" simulations by this atmospheric model coupled to two different ocean models. The atmospheric GCM is the NCAR Community Climate Model Version 3 (CCM3) and the coupled models are the NCAR Climate System Model (CSM) and the US Department of Energy - sponsored Parallel Climate Model (PCM). The CCM3's simulation with prescribed SST and sea ice is a contribution to the Atmospheric Model Intercomparison Project (http://www-pcmdi.llnl.gov/amip), and the CSM and PCM's simulations are contributions to the Coupled Model Intercomparison Project (http://www-pcmdi.llnl.gov/cmip).

Examining r.m.s. errors in spatial and time variations, we find that all three simulations have common weaknesses that presumably orginate in the atmospheric model. Prominent among these (in AMIP and CMIP runs in general, in addition to the simulations examined for this study) are cloudiness, meridional and vertical wind velocities, and tropopause temperatures colder than observed, especially in the Southern Hemisphere. As expected, replacement of the CCM3's "perfect ocean" with either the CSM's or PCM's ocean model degrades the level of agreement with observations for most fields. Nevertheless, r.m.s. errors in atmospheric variables from the coupled simulations are surprisingly similar to those from the simulation in which SST and sea ice are prescribed to observations. This result holds even for some aspects of interannual variations despite the fact that the coupled simulations represent long-term climate equilibria without reference to particular calendar years. For the seasonal and hemispheric components of the 24 atmospheric fields examined, the increase in r.m.s. error that results from switching from a "perfect ocean" to a coupled model is typically less than one-third of the r.m.s. variations in the corresponding observed fields. However, the normalized r.m.s. error increase is several times greater in the tropics.

Neither the CSM nor the PCM employs "flux adjustments" at the ocean-atmosphere interface, and the variables examined in this study are taken from the end of 300-year simulations by these models. We conclude that modern coupled ocean-atmosphere models without flux adjustments can provide simulations of comparable quality to those provided by atmospheric models driven by observed ocean and sea ice boundary conditions; furthermore, the climatology simulated by the coupled models can be stable for several centuries. Our future work will examine other pairs of AMIP and CMIP simulations as they become available.

This work was performed under auspices of the U.S. Department of Energy by University of California Lawrence Livermore National Laboratory under contract No. W-7405- Eng-48.