A method has been developed which can be used to retrospectively simulate and understand the evolution of sea surface temperature (SST) using reanalysis surface fluxes and a coupled GCM. The method has two parts: 1) weather noise surface fluxes are extracted from the reanalysis surface fluxes by removing the ensemble mean response of an AGCM ensemble to the observed SST evolution; 2) an interactive ensemble (IE) CGCM, in which a single OGCM is coupled to the mean of an ensemble of the AGCMs so the atmosphere internally generated weather noise is reduced, is then forced by this noise. The IE CGCM will reproduce the observed SST evolution except for errors in the model, the reanalysis surface fluxes, and the ocean initial state, as long as there are no internal ocean “weather noise” or coupled instabilities. External forcing must also be taken into account properly.

The method is applied both to understanding low frequency variability in CGCM simulations and to diagnosing the mechanisms of observed low frequency variability.

The results indicate that using the NCEP reanalysis surface fluxes and state of the art coupled GCM, the observed North Atlantic low frequency SST evolution for the latter half of the 20th century can be simulated reasonably well. Local heat flux weather noise forcing plays a major role. Ocean dynamics also has important effects on the SST evolution.