No abstract

No abstract

No abstract

The predictability of the Nin˜o3.4 region, especially the skill loss for lead times longer than two seasons, is the target of this study. We use an equatorial version of a seasonal statistical model to identify a seasonal predictability barrier, the skill loss of the predictions which target the summer or autumn Nin˜o3.4 Index value, relative to those which target the winter or spring values. The variables of the basic model include an index for the subsurface anomalous state and another for the atmospheric variability. We develop different versions of the model, substituting some of its variables with others that contain tropical or extratropical information, produce a number of hindcasts with these models using two different prediction schemes, and crossvalidate them. The analysis shows that in winter and spring some skill improvements can be gained with the introduction of a particular variable or the other. However, these improvements are similar to the ones obtained using a forecast scheme that incorporates the complete solution of the stochastic model. Moreover, useful summer and autumn hindcast skill values are scored only with the model versions that include a representation of the extratropical feedbacks among its variables. Higher scores correspond to models that incorporate an index built from atmospheric temperature anomalies integrated from the surface up to the mid-troposphere, south of 208S.