How predictable is El Niño?

Nobody anticipated that El Niño would be weak and prolonged in 1992, but brief and intense in 1997/98. Why are various El Niño episodes so different, and so difficult to predict? The answer involves the important role played by random atmospheric disturbances (such as westerly wind bursts) in sustaining the weakly damped Southern Oscillation, whose complementary warm and cold phases are, respectively, El Niño and La Niña. As in the case of a damped pendulum sustained by modest blows at random times, so the predictability of El Niño is limited, not by the amplification of errors in initial conditions as in the case of weather, but mainly by atmospheric disturbances interacting with the Southern Oscillation. Given the statistics of the wind fluctuations, the probability distribution function of future sea surface temperature fluctuations in the eastern equatorial Pacific can be determined by means of an ensemble of calculations with a coupled ocean-atmosphere model. Each member of the ensemble starts from the same initial conditions and has, superimposed, a different realization of the noise. Such a prediction, made at the end of 1996, would have assigned a higher likelihood to a moderate event than to the extremely strong event that actually occurred in 1997. (The rapid succession of several westerly wind bursts in early 1997 was a relatively rare phenomenon.) In late 2001, conditions were similar to those in 1996, which suggested a relatively high probability of El Niño appearing in 2002. Whether the event will be weak or intense depends on the random disturbances that materialize during the year.