We begin by making some non-standard distinctions, solely for convenience in the following discussion. We will define “ENSO prediction” as the process of predicting the SST in the tropical Pacific a month to a year or so in advance. We will use “short-term climate prediction” to refer to the procedure of predicting the climatic conditions in the global atmosphere or over land away from the tropical Pacific a month to a year in advance. The utility of this distinction is that ENSO prediction can only be accomplished by coupled models, whereas short-term climate prediction, which depends in part on the thermal forcing due to the distribution of regions of persistent precipitation and is partly determined by the SST distribution in the tropical Pacific, can be accomplished by a global atmospheric model (with predicted tropical SST specified) but can also be accomplished by a fully coupled climate model. The distinction will become clearer in our discussion of one-tiered and two-tiered prediction below.

The possibility that coupled climate models, whether simple or complex, can predict aspects of the future evolution of ENSO is not at all obvious. The atmosphere is known to be of limited predictability, basically because it is chaotic: inevitable small errors in the initial conditions grow and, depending on the growth rate, eventually limit the skill of prediction after a given time.