Away from the tropical Pacific ocean, El Nino and La Nina are associated with relatively minor changes in the probabilities of atmospheric states. Even so, it is important to estimate such changes accurately for each event, as they can imply substantial changes in the probability of extreme anomalies. They can also affect the predictability of seasonal averages. Changes of both mean and variance are relevant in this context. The mean signals need not be symmetric with respect to El Nino and La Nina. Even for El Nino and La Nina events considered separately, they may depend upon the unique aspects of the sea surface temperature (SST) fields for each event. As for changes of variance and higher moments, little is known at present. This is a concern especially for precipitation, whose distribution is strongly skewed in areas of mean tropospheric descent.

These issues will be examined in this talk in observations and general circulation model (GCM) simulations of the northern winter season. For the observational analysis, the 40-year (1958-98) global 'reanalysis' data generated at the National Centers for Environmental Prediction (NCEP) were stratified into El Nino, La Nina, and neutral winters. The GCM analysis was based on GCM runs made with prescribed seasonally evolving SSTs for El Nino, La Nina, and neutral winter conditions. A very large number of seasonal integrations, differing only in initial atmospheric states, were made for each of these tropical SST conditions. The ensemble was large enough that the changes of probability even in regions not usually associated with strong El Nino signals could be ascertained.

The results indicate a large asymmetry in the response to El Nino and La Nina conditions, not only in the mean but also the variability. In general the response for El Nino conditions was stronger, but also MORE VARIABLE, than for La Nina conditions. The source of the increased variability was traced to increased variability of precipitation in the central equatorial Pacific. The mean response was also highly nonlinear in the sense that in several regions the El Nino and La Nina probability distributions of lower tropospheric temperature and precipitation were distinguishable from their 'neutral' counterparts, but were indistinguishable from one another. These and other results confirm that the remote signals for individual El Nino and La Nina events can differ substantially from the historical signal obtained by pooling and averaging disparate events in the historical record. They also highlight the importance of understanding and determining the changes to the full probability distribution, and not just shifts of the mean, to extract the full benefit from seasonal predictions of SST in the tropical Pacific basin.