| description abstract | Numerical experiments are performed to isolate the cause of differences between the simulations of SST in the low-latitude Pacific of two coupled atmosphere?ocean general circulation models, the Center for Ocean?Land?Atmosphere (COLA) coupled model and the NCAR Climate System Model (CSM). The COLA model produces a more realistic simulation of the annual cycle of SST and interannual SST variability. The CSM has the more realistic annual mean wind stress and east?west SST gradient. The approach to finding the causes of these differences is to systematically eliminate differences in the physical parameterizations and numerics of the two models, and to examine the effects of these changes on the simulations. The results indicate that the atmospheric models rather than the ocean models are primarily responsible for differences in the simulations. There is no dominant process in the atmospheric models that explains the differences; both physical parameterizations (convection, surface flux formulation, shortwave radiation) and numerical schemes (vertical structure, moisture advection scheme) have significant effects. The effects of the parameterization changes on the annual mean SST are linear and additive, although tuning can cause apparent nonlinearity. In terms of the effects that directly impact the ocean, the different physics and numerics of the atmospheric models change the net heat flux into the ocean and/or the sensitivity of the wind stress to SST. These properties can be estimated by AGCM-only simulations with observed SST. Flux correction is then used to identify the process responsible for the difference between the coupled simulations. Heat flux is found to produce most of the difference, and with the sign that would be expected from the heat budget of the mixed layer. However, the larger sensitivity of the NCAR atmospheric model wind stress has a significant impact on extending the cold tongue into the western equatorial Pacific. | |
