| The goal of this study is to explore the short-term climate predictability of tropical Pacific sea surface temperature anomalies on the seasonal, interannual, and decadal time scales. The simple coupled atmosphere-ocean model developed by Battisti is used to evaluate the predictability of ENSO. Singular vector analysis is employed. To identify the roles of the annual cycle and the interannual ENSO cycle in determining the model predictive skill and error growth, we examine the optimal perturbation growth about several reference trajectories. We systematically investigate the relationship between the model forecast skill and optimal error growth for 6-month integrations about the actual model forecast trajectory for the period of record 1967-90.; In most of the cases investigated, there is only one structure that grows. The optimal perturbation growth is highly dependent on the phase of both the seasonal cycle and the ENSO cycle, and it is strongly dependent on the duration over which the perturbations are evolved. The maximum perturbation growth is achieved for integrations that pass through Northern summer, and the minimum growth takes place for integrations spanning Northern winter-early spring. The optimal perturbation growth for a 6-month integration starting in the peak phase or the cold phase of the ENSO cycle is generally small. The physical processes responsible for the phase dependence are identified. The initial and the final state patterns of the optimal perturbation, however, are both insensitive to the phase of the annual cycle and freely evolving ENSO cycle, and to the particular regions of phase space visited by the system. The structure of the optimal perturbation consists of an east-west dipole spanning the entire tropical Pacific basin, superimposed on a north-south dipole in the eastern tropical Pacific, and this structure has a simple physical interpretation.; There is a strong negative correlation between this model's forecast skill and the optimal error growth about the actual forecast trajectories, implying that the optimal error growth is in practice a useful indicator of the skill of the forecasts. Importantly, a large portion of the forecast error may be accounted for by the growth of the optimal errors in the sea surface temperature in the initial conditions alone. This suggests a need for a higher accuracy in observations of equatorial sea surface temperature. On the other hand, in about 4% of the forecasts, the optimal error growth in SST alone is insufficient to explain the poor forecast skill, suggesting that for these extreme cases, there are possible serious flaws in the model physics.; Finally, the seasonality of the model forecast skill and the optimal error growth exhibits a phase shift of 2-3 months from the pre-1976 to the post-1976 epochs. This decadal change is largely determined by the change in the actual forecast trajectories about which the errors are allowed to develop and not by a decadal change in the average annual cycle. (Abstract shortened by UMI.)... |
