Overview of the ENSO System

Predicting ENSO

Types of models for predicting ENSO

The prediction of ENSO (that is, the likelihood of an El Niño or La Niña) is based on models that predict sea-surface temperatures in the equatorial Pacific Ocean. There are two general types of these models.

The first type is called a "dynamical model" which consists of a series of mathematical expressions that represent the physical laws that govern how the ocean and atmosphere behave. To make a forecast, dynamical models are given the current conditions in the ocean and atmosphere and then a computer "does the math" to determine what the future conditions (out to six months or more in advance) will be.

The second general type of model is a "statistical model". These models use observations of the past to make predictions of the future. To make a forecast with a statistical model requires a long history of observations, generally of the same kind that would be used as input for dynamical models, but extending far back in time, by as much as 30 to 50 years. This long record of observations is used to identify key features of the ocean and atmosphere that often occurred prior to subsequent changes in sea surface temperatures in the tropical Pacific. Examples of precursors for El Niño are:

• An increase in the total heat content in the western Pacific ocean
• An increase in ocean temperatures in the western Pacific ocean at certain depths
• A weakening of the easterly trade winds in the central equatorial Pacific ocean.

Statistical models are "trained" on the long history of these precursor events along with the subsequent ENSO condition, so that when they are given the current observations they are able to predict the likelihood of various possible ENSO conditions for the upcoming several seasons. This type of model does what a forecaster might informally do from his or her own experience regarding what happens before an El Niño or La Niña. The difference is that the statistical model does this objectively and quantitatively. In contrast to dynamical models, the mechanisms that cause ENSO changes remain unknown in statistical models, as the model simply predicts on the basis of what has happened before. Statistical models can range from the very simple, such as analogs or simple regression, to more complicated schemes such as nonlinear canonical correlation analysis or neurological networks.

Advantages and disadvantages of prediction models

Dynamical models have some advantages over statistical models:

• Dynamical models are usually thought to be more scientific, because they explicitly use the physical equations and thereby attempt to accurately capture events in terms of their physical causes and effects.
• Dynamical models are able to handle unprecedented climate events, since the basic physics would apply equally well to novel situations as to familiar situations. Statistical models can only see new situations as extrapolations of historically observed ones, and run the risk of missing any new "rules of the game" that may come about only in the new situation.
• The accuracy of statistical models depends on the quality of the historical data used to train them. If the data quality is poor throughout the long period, then the forecasts will also be poor.

• Dynamical models require much greater computer power than statistical models, because the physical equations are more complex than statistical equations. (We are talking about the difference between a PC and a supercomputer!)
• Dynamical models approximate some of the physics of the ocean and atmosphere because they operate on spatial scales that are too small to be represented in the model. For example, the growth and precipitation of individual cumulus clouds in the tropics cannot be treated on an individual cloud basis, and must be approximated by a formula so that the results come out about right in an overall sense. This indirect treatment can compromise the accuracy of the resulting forecast.  

In real ENSO forecasting situations, the skills of dynamical and statistical models have been found to be approximately equal. While many oceanographers and atmospheric scientists expect dynamical models to prove superior as computer power increases and more is learned about ENSO physics, this has not yet been clearly demonstrated. Therefore, both kinds of models continue to be used. As statistical models are much quicker and less expensive to use, their skill is often thought to represent a baseline against which the skills of the more expensive dynamical models can be judged.

Physical barriers to prediction

Regardless of what type of ENSO forecast model one uses, forecasting ENSO is considerably more difficult during certain seasons of the year than others. Individual El Niño or La Niña episodes tend to develop between the months of April and June, and, once developed, last until the following February through May. Thus, once an episode has developed in early northern summer, forecasting its evolution through the remainder of its life cycle is not difficult. A much harder task is to forecast what will happen between March and June, when a forecast is being made in the preceding January through April. The difficulty in forecasting at this time of year is often called the "spring barrier" (in the Northern Hemisphere), or the "autumn barrier" (in the Southern Hemisphere).

After April has finished, while there still is uncertainty, it starts becoming easier to see in the latest observations how the stage is being set for the remainder of the calendar year and the first few months of the following year. By June, the uncertainty becomes still less: if there is nothing new developing, the chances of new development are small. While ENSO forecasting is most difficult through the late northern spring, the spring barrier is not impenetrable. Signs of changes in the ENSO state, such as increased heat content in the western equatorial Pacific Ocean, are available, so that at least a probability forecast can be made through the spring barrier. As April, May and June come along, such probabilities normally become more robust.