How an atmospheric general circulation model (AGCM) can reproduce the atmospheric aspects of climate, when given the boundary condition, is investigated in a series of AGCM experiments. In these experiments, the sea surface temperature (SST) computed by an integration of a coupled general circulation model (CGCM) is used to force the AGCM. This AGCM is the same as the atmospheric part of the CGCM.
A systematic difference in the mean climate is found between the AGCM and the CGCM in the tropical Pacific. This difference is apparent when and where the local SST is warmest. In the CGCM, the precipitation occurs to the west of the warmest SST region and does not last long probably due to a negative feedback between SST and cloud-radiation process. On the other hand, the precipitation in the AGCM tends to persist over the warmest SST region. The above difference results in a slightly but significantly different mean climate in the tropics in the AGCM compared to that in the CGCM. This also results in a larger estimate of interannual variability of AGCM than CGCM in the tropics.
The result of this experiment implies that a caution is needed for a use of an AGCM with boundary conditions derived from a CGCM for climate change studies.