THE POTENTIAL PREDICTABILITY IN A 14-YEAR GCM SIMULATION

A 14-yr simulation of a GCM forced by observed SST and sea ice is comp ared with observations as well as a GCM simulation that used climatolo gical surface conditions. The low frequency (periods > 2 months) behav ior in both simulations and observations is examined, and it is found that the anomalous boundary conditions were the cause of much of the l ow-frequency variability in the simulations. Without the anomalous bou ndary conditions, the low-frequency spectra was often flat, suggesting that the internal variability was producing a white noise-like spectr a. The anomalous boundary conditions were found to be very important i n determining the low-frequency behavior of the model. If the future v alues of the SST and sea ice were known, then the predictability for c ertain variables could be quite high for low-frequency signals (period s > 8 months). Specific zones showed predictability for low-frequency signals in excess of 70% explained variance. These zones were often re lated to ENSO, as the Southern Oscillation is the strongest intradecad al phenomenon that is forced by the anomalous boundary conditions. Thi s study gives a lower bound on the variance explained by the anomalous surface forcings.