December 1996 — The El Niño/Southern Oscillation (ENSO) phenomenon, an eastward shift of warm water in the tropical Pacific and associated effects on the atmosphere, is at the heart of global interannual climate variability. The just completed, decade-long Tropical Ocean/Global Atmosphere (TOGA) program was dedicated to understanding and working toward predicting ENSO by bringing together oceanographers and atmospheric scientists in a coordinated observational and numerical modeling research program. TOGA has not answered all the questions: We have not uncovered the physical mechanisms of the elusive ENSO "trigger" nor have our best coupled air/sea numerical models been as successful in predicting the rather irregular ENSO signal of the 1990s as they were in predicting the regular events of the 1980s and hindcasting the events of the late 1960s through the 1970s.
Prediction is the ultimate goal of ENSO research. It is also the ultimate test for an ENSO model and the theory underlying the model. During the last decade, a number of forecast models have shown predictive skills in both retrospective and real time forecasting, and they are now being used for routine ENSO prediction. Nevertheless, the skill of even the best available models is far from perfect, and there is still considerable room for improvement in modeling, observation, and forecasting techniques.
Factors that limit the current skill of ENSO forecasts include:
It seems likely that the inherent predictability limit for ENSO is years rather than weeks or months, though more theoretical study is needed in this area. The observing system is improving, but still far from satisfactory. Thus a challenge facing the modelers is to improve model forecasts by making the most reasonable and efficient use of available data.
In the past few years much effort has been devoted to assimilating various observational data into the initial state of forecast models. The most common approach is to improve the initial ocean conditions by assimilating observations of sea surface temperature, thermocline (region of rapid temperature decline) depth, or sea level into an ocean model prior to coupling it with an atmosphere model. One problem with this approach is that no attention is paid to the ocean-atmosphere interaction during initialization, so the coupled system may not be well balanced initially and may experience a shock when the forecast starts. A new initialization/assimilation procedure significantly improves the predictive skill of one of our most promising coupled models, which was constructed by Mark Cane and Steve Zebiak (Lamont-Doherty Earth Observatory).
In the new methodology the model is initialized in a coupled manner, using a simple data assimilation scheme in which the coupled model wind stress anomalies are "nudged" toward observations. The new procedure improves the model's predictive ability as measured by a variety of statistical scores. It also eliminates the so-called "spring prediction barrier," a marked drop of skill in forecasts that try to predict across the boreal spring, found in many previous ENSO forecast systems. The success of the new initialization procedure is attributed to its explicit consideration of ocean–atmosphere coupling, and the associated reduction of initialization shock and random noise.
As an example, the forecasts made by the improved model are compared to observations in the figure above right in terms of the sea surface temperature anomaly averaged over an area in the eastern/central equatorial Pacific (5°S to 5°N and 90°W to 150°W). The model is capable of forecasting ENSO more than one year in advance. The large warming and cooling events in the 1980s are particularly well predicted. However, the model does a poorer job for the 1970s and 1990s: The 1976-77 event is largely overpredicted, and the short warm episodes in 1993 and late 1994 are missed.
Although the predictive skill of this model is most likely limited by its highly reduced physics, the skill of more sophisticated coupled ocean-atmosphere general circulation models presently does not exceed that of the model described above, at least in terms of the tropical Pacific sea surface temperature. In order to predict the global impact of ENSO, a two-tiered approach appears to be reasonable: A physics-simplified, coupled model is first used to predict tropical sea surface temperature fields, and these fields are then used as boundary conditions for a more-complete-physics, global atmospheric general circulation model to predict the global distribution of atmospheric disturbances. Scientists are rigorously pursuing this kind of research.
A second area of progress concerns improved understanding of the coupling between different depths and different regions of the ocean. A popular ENSO paradigm that emerged in the late 1980s was based on the observed, rather regular rythms of ENSO conditions during a span of 25 years before 1990. The "delayed oscillator" mechanism emphasizes eastward-propagating equatorial wave processes,* westward-propagating off-equatorial signals, and their asymmetric reflections at eastern and western boundaries respectively. Despite the irregularities in the 1990s ENSO, this wave propagation/reflection paradigm is still compelling; it can accomodate irregularities in the ENSO signal by combining the tropical signal with longer-term variability in the subtropics.
A number of studies have sought to understand how tropical variability is linked to the mid latitudes. Ocean circulation may provide the links via several different pathways that are summarized schematically at right. These are not simple, direct north-south flows; the existence of vigorous zonal current systems complicate the picture. In the upper layers of the ocean, upwelled waters along the equator flow into the subtropics, mainly through the mid-latitude western boundary current (the Kuroshio). There is an additional interior ocean pathway, through the eastward Subtropical Countercurrent, that more directly feeds subtropical sites where surface water moves deeper into the ocean. These interior pathways are associated with a recirculating tropical gyre in and just below the mixed layer in the northeastern tropics. Below the mixed layer, thermocline water from the subtropics to the tropics zigzags almost zonally across the basin, succeeding in flowing toward the equator only along zonally narrow, southward flowing conduits. The low-latitude western boundary currents serve as the main southward circuit for the subducted (water moving from the surface to depth), subtropical thermocline water.
A model constructed by the authors also indicates important direct flow of thermocline water through the ocean interior, confined to the far western Pacific (away from the low-latitude western boundary currents) along 10°N. These southward flowing waters are then swept eastward by the North Equatorial Countercurrent, finally penetrating to the equator in the central and eastern Pacific. The water pathways in the subtropical thermocline essentially reflect the surface gyre circulation.
Along with our colleagues Ronghua Zhang (University of Rhode Island) and Antonio J. Busalacchi (NASA Goddard Space Flight Center), we have examined the interannual variability of these subtropical/tropical pathways and found important propagating subsurface ENSO signatures in the subtropical Pacific. There appears to be continual movement of subsurface, basin-scale anomalies that can then affect sea surface temperature (SST) anomalies, especially in sensitive regions where the thermocline is shallow. These SST anomalies can then trigger coupled air/sea interactions. A clear pattern of moving anomalies is less obvious at the sea surface. The systematic subsurface propagation is reminiscent of the delayed oscillator: eastward along the equator, westward off the equator with apparent further propagation along the eastern and western boundaries. Off the equator, subsurface propagation of anomaly patterns initiates an SST anomaly in the North Equatorial Countercurrent regions of the western Pacific, which then intensifies and moves into the equatorial waveguide, consistent with the mean water pathways found above. We speculate that this could be a mechanism for initiating coupled, air-sea interactions that can begin to evolve as an ENSO event. The cycling time of the subsurface anomaly patterns may determine the ENSO's frequency. We look forward to continuing our investigations to solidify these assertions.
One challenge for the newly established Climate Variability (CLIVAR) program will be to uncover the ENSO triggering mechanism and enable intelligent design of a long-term ocean and atmosphere monitoring system. CLIVAR is the oceanographic and atmospheric scientific community's new program of climate prediction. Its focus is on understanding the coupled air/sea system's variability on seasonal-to-interannual-to-interdecadal time scales for the purpose of determining predictability, and then prediction. Those observations would then feed into coupled air/sea numerical models for the purpose of long lead time forecasting, much like the present-day weather forcasting systems. However, the interannual ENSO signal does not exhibit a simple ryhthm; there are clearly influences on longer (decadal) time scales that need to be considered. There are clues as to what those signals might be (for example, the North Atlantic Oscillation—see Clara Deser's article), but we are still in the early stages of identifying these signals.
The natural system is not easily divided according to time scales; it is a fully nonlinear system. If we are to understand and eventually predict global interannual variability, we must not limit ourselves to monitoring the air/sea system over a few interannual cycles. Permanent monitoring systems are needed. It is the primary charge of CLIVAR to help design such a monitoring system while, at the same time, supporting the evolution of the numerical prediction systems that will issue the forecasts.