Predictability of ENSO on interannual and decadal timescales

The predictability and performance of the Zebiak-Cane (ZC) model for forecasts of the El Nino Southern Oscillation (ENSO) is assessed for both interannual and decadal variability. The ZC model is an intermediate coupled model for simulating the large-scale physics of ENSO. Historically, it has shown forecasting skill in the Nino-3 region that is on par with the skill of more sophisticated coupled general circulation models.; A low-order, coupled, 4D-Var data assimilation scheme is developed for initialization of seasonal to interannual model forecasts. Using this initialization approach, retrospective forecasts are performed from January 1975 to December 2003. It is shown that initializing the model at the end of the assimilation window (as in the traditional adjoint approach) yields 1--9 month forecasts that outperform those initialized with a standard nudging assimilation technique. However, the new scheme does not outperform the operational skill of the ZC model.; In a related study, the predictability of decadal ENSO variability is assessed. The ZC model is shown to be capable of producing sequences of variability that exhibit shifts in the time-mean state of the eastern equatorial Pacific that resemble observations of tropical Pacific decadal variability. The model's performance in predicting these shifts (when initial conditions are perturbed with uncorrelated noise) is compared with two native forecasting strategies. It is found that the ZC model consistently outperforms the naive forecasts, suggesting that its performance cannot be attributed to chance. Forecasts initialized during anomalously warm and anomalously cold decades are shown to have the highest predictability. These modeling results suggest that, to a moderate extent, the state of the tropical Pacific in one decade can predetermine its time mean state in the following decade. However, even in this idealized context decadal forecasting skill is too low to be useful. Results are discussed in the context of their implications for the ongoing debate over the origin of decadal variations in the Pacific.; Finally, the role of tropical wind stress in forcing the 1976--77 climate shift is studied. The 1976--77 shift towards a warmer time-mean state of the tropical Pacific upper ocean is the paradigmatic example of Pacific decadal variability. It is shown that the observed surface warming in the tropics after 1977 is consistent with thermocline variations produced in a linear shallow water model forced with the observed tropical wind stress. Restricting the wind forcing to within 5 degrees of the equator does not substantially alter this result. While the question of how tropical wind anomalies are generated is not addressed, this study does suggest that observed variability does not require an oceanic link to the midlatitudes. (Abstract shortened by UMI.)...