National Academies Press: OpenBook
« Previous: 3. COMPONENTS OF THE U.S. TOGA PROGRAM
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

4.
WHAT WE'VE LEARNED

TOGA was devoted to a study of ENSO in and over the tropical Pacific, its effects over the tropics and into midlatitudes, and its predictability. During the ten years of TOGA, the deployment of an observational system in the Pacific enabled us to observe the evolution of two warm phases of ENSO (1986–1987 and 1991–1992), one cold phase (1987–1988), and the prolonged warmth in the Pacific lasting from 1990 to 1994. The beginnings of a mathematical theory of the oscillatory aspects of ENSO have been developed. The predictability of sea surface temperature in the eastern Pacific, with lead times of about a year, has been demonstrated. Explorations of the effects of ENSO on the rest of the globe were pursued. The first short-range climate forecast made using coupled dynamical models predicted a year in advance the warming near the end of 1986. The TOGA years have seen major advances for the observational and theoretical understanding of ENSO. A series of regularly appearing data and prediction products capitalize on these advances. The meteorological and oceanographic communities worked together to bring about these accomplishments, reducing the barriers between their disciplines in the process.

We now have instruments to observe the evolution of the coupled atmosphere-ocean system in the tropical Pacific and theories to describe ENSO as a coupled atmosphere—ocean phenomenon. The observational progress made during TOGA has enabled us to begin evaluating theories of ENSO. The most apparently successful theory of ENSO is the delayed oscillator. This theory applies only to the completely regular case, although aspects of the theory apply to the irregular case as well. The underlying cause of the irregularity of ENSO can be narrowed to two possibilities: noise or nonlinearities, perhaps both. These observational and theoretical advances were major scientific lessons of TOGA. The program also provided lessons on how to conduct integrated research on the climate system.

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

OBSERVATIONS OF ENSO IN THE TROPICAL PACIFIC

Climatologies

ENSO is, among other things, an anomalous warming of the eastern Pacific. In order to define the anomalies, it is crucial to define the background, normal conditions against which the anomalies are measured. While simple in concept, this observational and definitional problem is complicated by the existence of both interannual (and longer) variability and intraseasonal (and shorter) variability. In the implicit definition used throughout this report, the climatology of a quantity is the sequence of monthly averages of that quantity for all the months of the year. Because both warm and cold phases of ENSO contribute to the monthly averages, a record long enough to include the effects of the slow interannual variability must be obtained. If the annual average is not stationary, the climatology will be unstable—i.e., different climatologies will arise from different averaging periods. Attempts to define a climatology by averaging only during “normal” periods—i.e., those without significant warm or cold phases of ENSO—will give an incorrect climatology if ENSO produces rectified effects.

Sea surface temperature is one of the key quantities that change during ENSO, and its climatology is therefore one of the most crucial. To date, climatologies of sea surface temperature have been obtained predominantly from historical records of in situ data (e.g., Reynolds 1982, Slutz et al. 1985), or by accumulating statistics from operational analyses (e.g., Reynolds and Smith 1994). Climatological winds have usually been obtained from historical records (e.g., Hellerman and Rosenstein 1983, Harrison 1989). The TOGA Observing System is providing large numbers of tropical data, which future climatologies will reflect.

The TOGA Observing System has produced some remarkable results on the climatologies of the oceanic subsurface thermal structure and subsurface circulation. Current-meter moorings have been in place at 110°W since 1980 and at 140°W since 1983. In conjunction with ATLAS moorings, they have yielded a remarkable picture of the behavior of the near-surface circulation (see, e.g., McPhaden and McCarty 1992). At 110°W (see Figure 7, middle), for example, the undercurrent is strongest in boreal spring when the winds are weakest, the sea surface temperature is warmest, and the surface currents have reversed to eastward. The warm sea surface temperature has no thermocline motion associated with it—indeed, the thermocline stays pretty much flat throughout the year. Clearly then, the main processes available to change sea surface temperature are the surface fluxes and surface advection, both affecting (and affected by) the mixing processes that determine the mixed-layer depth.

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

Models have been used to understand aspects of this evolution of the thermal structure (see, for example, Yin and Sarachik 1993 for a study using a two-dimensional nonlinear undercurrent model, and Philander et al. 1987 for a study using a three-dimensional ocean general-circulation model). The interpretation developed from modeling studies is that during boreal spring, when the winds are weak and the upward vertical advection of eastward momentum from the equatorial undercurrent is weakest, the pressure gradient is still strong and the surface currents are accelerated eastward, mostly by the pressure gradient. The surface currents then advect the temperature down gradient and warm the eastern part of the region. Vertical advection of momentum cannot be measured directly, but such advection plays an important role in the maintenance and variation of the undercurrent. As time goes on, and if the TOGA Observing System stays in place, we may expect to have a more complete, basin-wide view of the climatology of subsurface thermal variations.

Evolution of Warm and Cold Events

The canonical ENSO described by Rasmusson and Carpenter (1982)—on the basis of warm phases occurring in 1951, 1953, 1957, 1965, 1972, and 1976—showed a definite westward propagation of anomalies of sea surface temperature at the same time that isotherms of the total sea surface temperature moved eastward. This was possible because the isotherms of the annual cycle had a well defined westward propagation (see Figure 11a), and any anomaly added to the annual cycle would move isotherms farther east.

Descriptions of the propagation characteristics of the 1982–1983 warm phase are not consistent in the literature. However, it appears that the anomalies of sea surface temperature (see Figure 11b) warmed uniformly over large parts of the central Pacific, followed by warming near the coast of South America. It is clear that in some sense the anomalies of sea surface temperature did propagate eastward, but in another sense the warm sea surface temperature developed over a large region of the tropical Pacific without propagation.

Figure 9 shows the anomalies, relative to the climatology of Reynolds and Smith (1994), of sea surface temperature along the equator during the TOGA years. All points of view about the propagation of sea-surface-temperature anomalies can find evidence in this series. The anomalous warming in 1991 seems to have a component that propagated eastward, although in the boreal spring of 1991 warm anomalies appeared simultaneously in the east and around the dateline. The anomalous warmth of early 1993 seems to have set in simultaneously across the entire eastern Pacific, with no obvious phase propagation. The cooling of 1988 appears to have a definite westward propagation, although the zero line of the temperature anomaly shows no phase propagation.

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

Figure 11. Evolution of sea surface temperature in the tropical Pacific. The sections follow the equator from the western Pacific up to 95°W, then follow the climatological cold axis to the coast of South America, reaching 8°S. Panel (a) shows the mean climatology of the annual cycle, repeated for comparison with the other panels. Panel (b) shows anomalies for a composite El Niño, with the year of the warm peak designated year 0. Panel (c) shows the anomalies for 1981–1983. Note the difference in the contour intervals. (Reprinted with permission from Cane 1983, copyright American Association for the Advancement of Science.)

A more complete picture of the evolving warm and cold phases of ENSO (including subsurface thermal structure) is now available as a guide to modelers. The winds, subsurface thermal structure, sea surface temperature, and sea level are measured directly by the TOGA Observing System, and other quantities such as precipitation and surface fluxes are obtained by other methods. These

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

should become available in future years from the TAO array, which was completed at the end of TOGA.

The Warming During 1990–1994.

An increased incidence of warm phases of ENSO since about 1970 is clear in the instrumental record (Smith et al. 1994). The warming from 1990 to 1994 is unprecedented in the instrumental record (Trenberth and Hoar 1996). The Southern Oscillation Index (SOI) was negative during those five years, the western Pacific around the date line has been anomalously warm for most of that period, there were two warm phases of ENSO very close to each other (peaking in the boreal springs of 1992 and 1993) and another one well developed at the end of 1994. In addition, there was a persistent “horseshoe-shaped” anomaly of sea surface temperature developing early in 1990 (see Figure 12) extending from 20°S to 20°N in the east Pacific and crossing the equator at the dateline in the west. The predictions of sea surface temperature in the NINO3 region have been poorer during this time, with some notable misses by all dynamical and statistical models.

Figure 12. Total field (upper) and anomalies (lower) of sea surface temperature for January 1992. (From Climate Analysis Center 1992.)

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

Whether the decadal mean temperatures in the eastern Pacific have changed (caused by some possible natural fluctuation or by anthropogenic greenhouse warming) or whether the frequency of ENSO warm events has changed due to some other cause is, as yet, impossible to say. Also unusual in the period 1990–1994 is that the magnitude of the anomalies of sea surface temperature in the equatorial Pacific peaks in the boreal spring, coincident with the normal annual warming, rather than the normal case for ENSO with the magnitude of the warm anomalies peaking in late boreal winter.

Figure 13. Analyses of interannual and lower-frequency variations of sea surface temperature throughout the entire Pacific. Panel (a) shows the amplitude as a function of time for the leading (normalized) principal components with a high-pass (HP) filter, panel (b), and a low-pass filter (LP), panel (c), divided at 6 years. The tick marks in panel (a) indicate 1.0 standard deviations. The contour intervals in panels (b) and (c) correspond to 0.1 K per standard deviation. Negative contours are dashed; the zero contour is thickened. (From Zhang et al. 1996, reprinted by permission of the American Meteorological Society.)

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

Since the end of TOGA, the nature of the unusual warming of the early 1990s has gradually become clearer. Figure 13 shows analyses of both interannual and lower-frequency (periods longer than a few years) variations of sea surface temperature throughout the entire Pacific (Zhang et al. 1996; similar results were found earlier by Nitta and Yamada 1989). The pattern of the higher-frequency variability of sea surface temperature is clear—it is the ENSO phenomenon tightly confined to the equator, juxtaposed with a weak (opposite sign) North Pacific covariation. The lower-frequency variability is much less spatially confined to the tropics—it exhibits a horseshoe-shaped pattern similar to the pattern that persisted throughout the early 1990s, as shown in Figure 13. The amplitude function (the lower curve of Figure 13a) shows that the decadal mode was strong during the early 1990s and that it indeed provided a different background state for the more equatorially confined ENSO variations.

We now know that skill for predicting variations of sea surface temperature near the equator varies decadally (see, for example, Balmaseda et al. 1995 and Chen et al. 1995). It seems reasonable to speculate that the varying background state provided by the decadal mode is the cause of the decadal variation of ENSO prediction skill. Examination of the decadal time series in Figure 13 would seem to indicate that the decadal mode has been strong since the 1980s (when predictive skill was high), so that if this speculation is even partly true, other factors currently unknown must be involved. While the decadal mode has been identified in observations (Balmaseda et al. 1995, Chen et al. 1995) and simulated in a coupled general-circulation model (Latif et al. 1996) the mechanism for this mode is still not clear. Latif et al. (1996) argue that this mode is fundamentally a midlatitude mode with a tropical expression; they also provide a mechanism that depends essentially on midlatitude dynamics. Consistent with this view is the observation that the midlatitude expression of the decadal mode is stronger than the tropical expression, in contrast to the interannual mode, where the midlatitude expression is distinctly weaker. Whatever the cause, it becomes important to understand this decadal modulation of ENSO in order to improve prediction skill.

EFFECTS OF ENSO ON THE REST OF THE GLOBE

Tropics

The interactions on interannual time scales of the tropical oceans, communicating through the medium of the atmospheric circulation, were investigated by Latif and Barnett (1995), who conducted a series of model experiments and also analyzed observational data. Nagai et al. (1995) investigated the role of the Indian Ocean in the ENSO cycle. The results of both studies illustrate the key role of ENSO in generating interannual variability in all three tropical oceans.

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

Anomalies of sea surface temperature in the tropical Pacific force, via changes to the atmospheric circulation, anomalies of the same sign in the tropical Indian Ocean, as well as anomalies of opposite sign in the tropical Atlantic. The role of air-sea interactions in the tropical Indian and Atlantic Oceans is mainly that of an amplifier, by which ENSO-induced signals are enhanced in the ocean and atmosphere.

No evidence for zonal wave propagation around the globe, as proposed by Tourre and White (1996), was found in a modeling study by Latif and Barnett (1995). Consistent with observations, some eastward phase propagation was found in several model simulations. Anomalies of sea surface temperature in the Pacific induce a global response in winds. Because the different oceans exhibit different characteristics in response to low-frequency wind changes, the responses of the individual tropical oceans can by chance be timed to resemble a global wave, a coincidence which appears to be the case.

Middle Latitudes

While ENSO cycles have certain common and reproducible aspects in the tropics, their effects at higher latitudes are more variable. Large influences of ENSO, such as atmospheric wave trains propagating out of the tropics, can clearly be seen. However, the statistical reproducibility of these influences by the tropics on the middle latitudes is poor when the events are stratified solely according to whether or not a warm phase of ENSO is taking place. Modeling studies (admittedly with coarse atmospheric resolution, such as N.-C. Lau and Nath 1994) have indicated that variations of sea surface temperature in the tropics, characteristic of those found during the warm and cold phases of ENSO, affect the middle latitudes in recognizable patterns, and that midlatitude variations of sea surface temperature have only a small and indistinct effect. ENSO can have a variety of structures in the tropics, and each of these structures may induce a different reaction at higher latitudes. Furthermore, internally generated midlatitude variations may overwhelm the signal from tropical influences. These possibilities are briefly discussed here (see also, e.g., Trenberth 1995).

The primary source of ENSO is in the tropical Pacific. The observed global effects arise as the atmosphere transmits to higher latitudes the influence of anomalous heating in the tropics, driving changes to the large-scale overturning and thus anomalies in convergence and divergence. Regions of low-level convergence are mirrored in the upper troposphere by regions of divergence. The divergent outflow in the upper troposphere forces atmospheric Rossby waves (Sardeshmukh and Hoskins 1988), which propagate into the extratropics. The divergent outflow also provides an anticyclonic forcing when the Coriolis force acts on the outflow and, together with the advection of the Earth's vortic

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

ity by the anomalous divergent flow, helps create an anomalous Rossby-wave component (Sardeshmukh and Hoskins 1988, Rasmusson and Mo 1993). In the vicinity of the tropical heating, the response often takes the form of a pair of anomalous upper-tropospheric anticyclonic eddies that straddle the equator. The Rossby waves that emanate from these eddies can vary in wavelength and location according to the exact nature and scale of the forcing.

Several factors are important in determining the response of the extratropics to ENSO. First, rather small changes in both sea surface temperature and sea-surface-temperature gradients can greatly influence the locations of strongest convection and low-level convergence. The first-order response of the tropical atmosphere to anomalies of sea surface temperature is a shift in the location of organized convection. The second-order effect is a change to the character and intensity of the convection. Both effects lead to large anomalies in atmospheric heating, primarily through latent-heat release associated with precipitation. The spatial extent of the anomalous heating can vary considerably. In these respects, differences among the various ENSO cycles are large. These differences are apparent in the field of outgoing long-wave radiation in the tropics, where the response to sea-surface-temperature anomalies can be considered to be fairly direct. How much of this accounts for the huge differences among events in the extratropics is not yet known.

The second major factor in the response of the extratropics to ENSO comes from differences in the medium through which the forced Rossby waves propagate. These differences arise from the changing seasons and the changing location of the forcing. In addition, a random component arises from weather and weather-related variations, which dominate the extratropical circulation. Accordingly, the rather small influence on the extratropics from tropical forcing can be seen only if an average is taken over many synoptic events. The natural variability is therefore a form of unpredictable noise, while the signal caused by ENSO is regarded as potentially predictable and thus reproducible in a good model. For El Niño forcings, Kumar and Hoerling (1995) used an atmospheric general-circulation model to show that while the signal-to-noise ratio for the 200 mb geopotential height over the Pacific and North American extratropical region is roughly 0.2 for monthly means, it increases to about 0.6 for seasonal (3-month) means and to about 0.9 for 5-month means.

The above reasoning assumes a largely linear addition of the tropical influences on midlatitude patterns. This assumption is known from both observational studies (Horel and Wallace 1981) and modeling investigations (Hoskins and Karoly 1981) to be at least partially valid, although complications such as zonal inhomogeneity in the climatological background flow must be included (Branstator 1985). The Northern Hemisphere wintertime flow has a great deal of natural low-frequency variability arising from barotropic instability, and this

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

same barotropic energy supply can be tapped by disturbances initiated by forcing localized to the tropics (Simmons et al. 1983). In addition, changes in the extratropical circulation immediately begin to change the jet stream and the associated storm tracks, so that the heat and vorticity fluxes caused by the transient eddies in the extratropics are also altered (Branstator 1995). The impact of these secondary changes, however, can be as large as or larger than the influence of the direct tropical forcing (Held et al. 1989, N.-C. Lau and Nath 1990, Kushnir and Lau 1992). Fortunately, some of these influences appear to be fairly systematic, although differing considerably by region, depending on the climatological background flow (Trenberth and Hurrell 1994). It is therefore possible to parameterize the effects of the changes in storm tracks (Branstator 1995).

Another view of the role of the tropical forcing on the extratropics was put forward by Palmer (1993) and Molteni et al. (1993). They note the very large natural variability of the extratropical circulation, but with the existence of certain preferred regimes where more persistent flow patterns recur. The persistent patterns are presumably associated in some way with the distribution of land and sea, and with the climatological planetary waves. The patterns indicate that one effect of forcing by tropical sea surface temperature is to alter the frequency of occurrence and stability of certain pre-existing regimes, but with only minor changes in the regimes themselves. Thus preferred “teleconnection patterns”, such as the Pacific-North-American (PNA) pattern, may be excited. This helps explain why the response of the extratropical atmosphere to sea surface temperatures in the tropics can be highly nonlinear, as was found by Geisler et al. (1985).

The effects of ENSO are clearly global in extent. However, differences among ENSO cycles in the tropics are important and can become magnified in the extratropics. The anomaly patterns in the extratropics cannot be determined reliably by statistical means because the number of examples is not yet large enough to permit classification of the ENSO cycles into subtypes. Consequently, the best hope for predicting the remote influences of a particular phase of ENSO is good numerical modeling. Trenberth and Branstator (1992) discuss the criteria for performance of models if the simulations and/or predictions are to be useful. However, most models do not yet satisfy these demands. Improved models will allow the extratropical predictability that does exist to be better exploited.

ENSO and the Asian-Australian Monsoon

Although the heat source for the atmosphere associated with the western Pacific warm pool is of immense importance for, and has distinct and direct influences

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

on, the global climate (see, e.g., Palmer and Mansfield 1984), it is not the only region of intense atmospheric heating. In addition to the latent heating over the warm pool and the Indonesian archipelago, there is a distinct migration of the heating maximum over south Asia during the spring and early summer. The heating variability is asymmetric with respect to the seasons; the deepest convection is located north of the equator in the western Pacific and over south and east Asia. During the austral summer, most of the convection remains fairly close to the equator. The annual migration of this heating maximum is associated with the Asian-Australian monsoon system. Thus, the annual cycle of the tropical climate system is driven by heating gradients of the coupled ocean-atmosphere-land system, and the monsoon system is an integral part of the Pacific Ocean climate. Furthermore, there are year-to-year variations in the intensity and location of the heat sources during the boreal summer. These variations relate directly to strong and weak monsoon seasons over Asia (Webster and Yang 1992).

A portion of the variability of the monsoons can be directly related to the influence of ENSO. Table 1 shows that of the 20 drought summer seasons over India (i.e., summers with less rainfall than one standard deviation below the mean), 11 of them were during ENSO warm phases (E1 Niño years); none was during a cold phase. On the other hand, no years during an ENSO warm phase were “very wet” years (i.e., years with summer more rainfall than one standard deviation above the mean), and 8 of the wet years were during ENSO cold phases. Thus, ENSO appears to explain a considerable amount of the rainfall variance over India (Shukla and Paolina 1983). However, Torrence and Webster (1996) point out that there were periods in which there was little ENSO variability and a correspondingly small covariance with Indian rainfall. From 1915 to 1960, variability in NINO3 (the anomaly of sea surface temperature averaged over the region between 5°N and 5°S, 150°W and 90°W) was quite

Table 2. Rain Conditions for India and Associated Phase of ENSO

 

Total

During ENSO Warm Phase

During ENSO Cold Phase

Drought

20

11

0

Below-average rain

60

20

0

Above-average rain

61

2

16

Very wet

18

0

8

The relationship between the All-India Rainfall Index and the state of the Pacific Ocean for the period 1870 to 1991. See text for definitions of rain conditions. The table is an extension of one in Shukla and Paolina (1983).

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

small. In fact, the covariance between Indian rainfall and sea surface temperature in the western Indian Ocean was greater than the covariance between Indian rainfall and Pacific sea surface temperatures. The overall correlation between ENSO and Indian precipitation is dominated by two periods of very strong covariance, during 1890–1915 and during 1965–1980.

The waxing and waning of the relationship between the intensity of the monsoon and ENSO is tied to interdecadal variations of the planetary-scale climate system. Noting the failure of the relations between the SOI and Indian monsoon rainfall found earlier by Walker and his coworkers, Normand (1953) made the following assessment:

Unfortunately for India, the Southern Oscillation in June–August, at the height of the monsoon, has many significant correlations with later events and relatively few with earlier events… The Indian monsoon therefore stands out as an active, not a passive feature in world weather, more efficient as a broadcasting tool than an event to be forecast… On the whole, Walker's worldwide survey ended offering promise for the prediction of events in other regions rather than in India…

Clearly, Normand was commenting from the perspective of nearly forty years of poor relationships between the monsoon and ENSO. Troup (1965) made a similar point in noting that there had been significant changes in the character of ENSO during the period 1915–1965. Balmaseda et al. (1995) and Torrence and Webster (1995) showed much later that the character of ENSO itself had also changed through the decades.

The monsoon and ENSO are parts of a climate system that evolves from decade to decade and that involves tightly coupled components. It is difficult to discern where processes begin. The monsoon system may couple to the extratropics through mechanisms involving Eurasian snowfall (see, e.g., Barnett 1984, 1985; Yang 1996). While examining the relationship between the monsoons and ENSO, Barnett (1984, 1985) and Barnett et al. (1989) detected a large-scale propagating surface-pressure signal that moved through the Indian Ocean region into the Pacific Ocean, with time scales greater than two years. The signal appeared to originate in the Asian region, and Barnett (1985) suggested that it was associated with Eurasian winter snowfall.

In an attempt to develop preliminary investigations into monsoon predictability and the relationships between monsoons and ENSO, the TOGA Program created the Monsoon Experimentation Group (MONEG). A principal mandate of MONEG was studying, using integrations of atmospheric general-circulation models, the effects of prescribed anomalies of sea surface temperature on the monsoon. To focus its study, MONEG chose to investigate the monsoon

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

seasons of June–July–August during 1987 and 1988. As discussed by Krishnamurti et al. (1989a, b), these years were of particular interest because of their contrasting behavior. On one hand, 1987 was a severe drought year over both India and the African Sahel. On the other hand, 1988 was an above-average monsoon season for India, while rains over the Sahel were close to their longterm climatological mean. During these years, ENSO progressed abruptly from a warm phase in the summer of 1987 to a cold phase in 1988. This period was one in which the covariance between ENSO and the monsoon rainfall was strongest. The difference in sea surface temperature between the summers of 1987 and 1988 reached a maximum of 4 K in the eastern equatorial Pacific, flanked by differences of -1 K north and south of the equator. Other differences in sea surface temperature were evident around the globe, although the differences were much smaller in the Indian Ocean which tends to be in phase with the central Pacific.

In the first phase of coordinated experimentation, 17 atmospheric-modeling groups ran 90-day integrations with prescribed fields of sea surface temperature based on observations for 1987 and 1988. These are referred to as the MONEG integrations; full results from these integrations are discussed in WCRP 1992. The ability to simulate the basic monsoon climatology in a general-circulation model is not a trivial matter, and many of the models contributing to the MONEG-coordinated experiments suffered significant drift from the observed climate, often resulting in rather weak monsoon rainfall over some land areas (such as India). From the set of all MONEG integrations, it was apparent that a realistic simulation of the mean monsoon is a prerequisite to simulate interannual variability correctly, suggesting that the monsoon is an inherently nonlinear phenomenon.

Several models within the MONEG experimentation set (e.g., Palmer et al. 1992) were able to reproduce much of the observed coarse-grained interannual variability in the monsoon areas for 1987 and 1988, and the numerical experiments indicated that the skill was obtained from specifying the underlying (and evolving) anomalies of sea surface temperature, rather than from specifying the atmospheric initial conditions. This result is consistent with the Charney and Shukla (1981) hypothesis that most variability in the tropics is determined by variations in the state of the boundary.

The results (see, e.g., Palmer et al. 1992) showed overwhelmingly that variations in the monsoons were associated with observed variations in tropical Pacific sea surface temperatures. Observed sea-surface-temperature anomalies in the Indian and Atlantic Oceans did have some impact on the atmosphere over the Indian and Atlantic Oceans, respectively. Still, the role of interannual variability in the Indian Ocean on the Asian monsoon seems to be weaker than the remote effect of Pacific Ocean sea surface temperatures, at least during

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

ENSO years, and at least during the particular phase of the interdecadal variability that was studied. Since the completion of the MONEG integrations, Ju and Slingo (1995) have concluded that during cold phases of ENSO, warm anomalies of sea surface temperature in the tropical northwest Pacific may be of more direct importance in influencing the Asian summer monsoon that the cold equatorial anomalies of sea surface temperature in the central and eastern Pacific that determine the phase of ENSO.

In summary, it is clear that there is a predictable element in the summer monsoon of south Asia. At certain times during the last hundred years there are clear connections between ENSO and the strength of the Asian summer monsoon. Strong relationships exist between the Australian summer monsoon and ENSO with limited precipitation occurring during a warm phase and abundant precipitation during a cold phase. However, there are periods of low ENSO variance (e.g., 1920–1960) when there appears to be little connection between the monsoon and ENSO, and periods when the variability in the monsoon seems to lead ENSO variability. Questions remain as to the effects of the oceans on the monsoon during neutral phases of ENSO, or when ENSO has a different character from the 1987-88 episodes. Also, the effects of sea-surface-temperature anomalies on more regional quantities, such as country-wide seasonal mean rainfall, are largely unknown. From a practical forecasting point of view, these fine-grain issues are obviously of great importance.

Carbon Dioxide and ENSO

The eastern equatorial Pacific is a net source of carbon dioxide (CO2) for the atmosphere. Carbon is exchanged between the atmosphere and the ocean in proportion to the difference of the partial pressures of carbon dioxide (pCO2) in the atmosphere and in the ocean. In the eastern Pacific, the dominant control on pCO2 is upwelling to the surface of carbon-rich waters from beneath the thermocline (Barber and Chavez 1983). The controls on pCO2 in the western Pacific are by no means as clear (see Ishii and Inoue 1995), but seem to be dominated locally by temperature and salinity, both of which affect the solubility of CO2, with very little effect from upwelling because the carbon-rich waters are too deep.

When ENSO moves into a warm phase, the thermocline in the east deepens and the connection of the cold, deeper, carbon-rich ocean to the surface is broken. The result is that the outgassing of CO2 to the atmosphere ceases. The pCO2 in the west seems to change with changes in salinity, but these variations do not lead to major changes in the outgassing of CO2 to the atmosphere (Fushimi 1987). All other things being equal, the amount of CO2 in the atmosphere should decrease during warm phases of ENSO. However, variations in

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

the phase of ENSO drive changes in temperature and precipitation on land, and also affect the atmospheric concentration of CO2. Although the correlation of variations in the SOI and atmospheric CO2 (with variations in the SOI leading) was first reported more than two decades ago (Bacastow 1976), a quantitative understanding of the relative roles of the terrestrial versus the marine carbon pools in causing interannual fluctuations of CO2 remains controversial.

The atmospheric anomalies of CO2 observed during major ENSO warm events (1965, 1969, 1972, 1976, 1982, 1986, and 1991) are of order 1–2 ppmv; this should be compared to an annual-mean concentration now at 365 ppmv (increasing at a mean rate of approximately 1.5 ppmv/yr because of fossil-fuel combustion and deforestation). To put these values in perspective, note that a 1 ppmv increase in the global-mean atmospheric CO2 level requires an input of 2 Gton (2×1015 g) of carbon (C; 1 Gton of C is equivalent to 3.7 Gton of CO2). The balanced annual exchange of atmospheric CO2 with the global ocean is approximately 100 Gton C/yr; a nearly comparable exchange of approximately 70 Gton C/yr occurs between the atmosphere and the terrestrial biosphere. Therefore, the largest CO2 imbalances observed during climatically anomalous years of major ENSO phases represent only 1 percent fluctuations in the balanced exchange fluxes of atmospheric CO2 with the oceanic and terrestrial biospheric reservoirs of carbon.

The 1982–83 ENSO warm event provided the first opportunity to quantify with direct measurements the major reduction of the sea-to-air CO 2 flux from the equatorial Pacific Ocean. Feeley et al. (1987) and Keeling and Revelle (1985) estimated the reduction to be approximately 0.6 Gton C. The reduction in flux results from the greatly reduced equatorial upwelling of deeper waters, which are strongly supersaturated with CO2. The reduced ocean-to-atmosphere flux of CO2 in the eastern equatorial Pacific during the TOGA years has been consistently confirmed by surface pCO2 measurements throughout the development and decay of subsequent ENSO events in 1986–87 (Wong et al. 1993, Inoue and Sugimura 1992), and 1991–1994 (Feely et al. 1995, Inoue et al. 1996, Lefevre and Dandonneau 1992).

The pCO2 of the surface in the western equatorial Pacific is not controlled by upwelling, because the thermocline is far from the surface and the carbonrich waters cannot reach the surface by upwelling. There, the ocean-surface pCO2 is almost the same as the atmospheric value, and very little outgassing of CO2 occurs (Ishii and Inoue 1990). Furthermore, although there are measurable changes of pCO2 with changes in salinity, there is very little temperature change in the west Pacific to drive variations in the surface pCO2.

While a reduced ocean-to-atmosphere flux of CO2 during ENSO warm events is now well established, this variation alone is not sufficient to explain the observed fluctuations in regional or global atmospheric CO2 levels. The

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

pattern and timing of the CO2 variations during major ENSO warm events require a substantial, perhaps dominant, contribution from the terrestrial biosphere and the soil carbon pool. Ocean (only) general-circulation models seem unable to reproduce the magnitude or phasing of the observed atmospheric CO2 fluctuations (Winguth et al. 1994).

The argument for terrestrial dominance is most forcefully made by cointerpreting the atmospheric time series of total CO2 and the time series of isotopically substituted 13CO2. (The 13C/12C isotopic ratio of carbon dioxide is substantially altered during photosynthetic exchanges of CO2 with the terrestrial biosphere, but nearly unaltered by air-sea gas exchange.) Keeling has argued that the major pulse of CO2 to the atmosphere at the conclusion of major ENSO warm events comes from the terrestrial biosphere, largely caused by the drought and fire in southeast Asia that accompany the failure of the monsoon, and that, more generally, the flux anomalies of the ocean and terrestrial biosphere during an ENSO warm event are large and of opposing sign (Keeling et al. 1989, Keeling et al. 1995). An independent record of 13C by a different group shows a much smaller terrestrial contribution (Francey et al. 1995), and the topic remains controversial (Heimann 1995, Siegenthaler 1990).

The interpretation of the CO2 anomalies accompanying the latest, and unusually long, period of warmth in the tropical Pacific (1991–1994, see p. 94) is complicated by the Mt. Pinatubo eruption (June 1991) and the subsequent cooling of the atmosphere because of the veil of stratospheric volcanic aerosol, which likely had major effects on the temperature-dependent oceanic and terrestrial carbon reservoirs.

THEORIES OF ENSO

Advances in the theory of interannual variability in the tropical Pacific related to the coupling between the atmosphere and ocean can be divided into four categories: the mechanism of ENSO (including attempts to verify some existing hypotheses), the irregularity of ENSO, the potential for interactions between the annual cycle and ENSO variability, and theories of ENSO prediction. Each of these is discussed below.

The Mechanism of ENSO

The fundamental idea underlying the current explanation of ENSO is that interactions between the atmosphere and ocean can give rise to instabilities of otherwise stable atmospheric and oceanic systems. This can be understood most simply by imagining a warm anomaly of sea surface temperature that creates surface winds, which in turn enhance the warmth of the anomaly. The anoma

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

lies of sea surface temperature and winds would grow in concert, and therefore be unstable. The problem is to identify those interactions and modes in the combined atmosphere and ocean system that have this behavior, and to see which (if any) correspond to ENSO.

The theoretical study of coupled modes began with a paper by Philander et al. (1984) that used simple, linear, “shallow water” models for both the atmosphere and the ocean. The scheme parameterized anomalies of atmospheric convection as proportional to anomalies of sea surface temperature, and parameterized anomalies of sea surface temperature as proportional to thermocline depth. Philander et al. found an unstable coupled atmosphere-ocean mode. The mode was eastward moving, with characteristics in the ocean very much like those of a free oceanic Kelvin mode, and with the atmospheric convection following the positive anomalies of sea surface temperature. Because, as we have seen, sea surface temperature anomalies by and large propagate westward, and eastward-propagating isotherms of total sea surface temperature travel with about a tenth the speed of a free Kelvin mode, this mode could not be identified with the ENSO signal. The Philander et al. paper was notable, however, in showing that coupled unstable modes can exist.

Hirst (1986) performed a more complete study of coupled atmosphere-ocean modes using simple shallow-water models. He found that long-period (several years) and slowly growing (several months) modes always exist, but that the nature of the coupled modes depends crucially on the mix of processes that control sea surface temperature. When anomalies of sea surface temperature are assumed to be simply proportional to anomalies of thermocline depth, then the unstable coupled mode resembles a free, eastward-propagating Kelvin mode in the ocean, with atmospheric convection following. When anomalies of sea surface temperature are changed by surface advection, the unstable coupled mode resembles a free, westwardly propagating Rossby mode in the ocean, with atmospheric convection following. However, when the rate of change of sea surface temperature was set proportional to thermocline depth (as a simple parameterization of the effects of mean upwelling), the nature of the coupled modes changed dramatically. A slow, eastwardly propagating mode arose, with ocean characteristics that did not resemble any free mode. While this coupled mode still could not be identified with ENSO, it provided a cautionary note to observationalists that propagating sea-surface-temperature anomalies in the ocean could not necessarily be identified with known free modes. One serious puzzle raised by Hirst (1988, Hirst and Lau 1990) was that these modes seemed oblivious to the existence of boundaries—they were identical in bounded and unbounded basins.

The coupled atmosphere-ocean model of Zebiak and Cane (1987) was built specifically to model interannual anomalies of sea surface temperature in

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

the tropical Pacific. The key to the Zebiak and Cane model was the specification of monthly mean climatologies for both the atmosphere and the ocean, so that only anomalies about the specified climatology were calculated. The result of this coupled atmosphere-ocean model was a reasonable simulation of ENSO (see Figures 4–11 in Zebiak and Cane 1987). The model ocean consisted of a single baroclinic mode with a fixed-depth mixed layer. Anomalies of sea surface temperature were calculated with allowances for surface advection, mean and anomalous upwelling, and heat fluxes at the surface. The model atmosphere was a modification of the Gill (1980) model; it responded to anomalies of sea surface temperature and was also affected by the presence of mean convergences and divergences. In the coupled model, the anomalies of sea surface temperature grew in place and exhibited no particular direction of propagation. Clearly a mechanism for ENSO was contained in this model.

Battisti (1988) built a near replica of the Zebiak and Cane model to isolate the mechanism that produced ENSO-like oscillations. The model gave only regular ENSO-like oscillations but turned out to be appropriate for understanding a regular ENSO. (The differences between the Zebiak and Cane model and the Battisti model were explained only recently, by Mantua and Battisti (1995).) The oscillatory mechanism involved an intricate interplay among coupled instabilities, free equatorial modes, and changes in sea surface temperature. Consider a coupled instability in the eastern part of the Pacific such that, in the absence of boundaries, the anomaly of sea surface temperature would increase exponentially in time with a growth rate c and would be confined latitudinally to the region of the equator. The exponentially increasing surface winds, induced by the increasing sea surface temperature, would lie to the west of the sea-surface-temperature anomaly, as is usual in both models and observations. Because the winds cover a finite longitudinal extent, they tend to lower the thermocline to their east and raise the thermocline to their west, in accordance with the tendency of equatorial adjustment to bring the thermocline tilt in balance with the wind stresses (see Cane and Sarachik 1977). The downwelling of the thermocline to the east of the winds is consistent with the warming in the eastern Pacific. The upwelling (cooling) signal propagates freely westward with speeds and meridional structures characteristic of the lowest order symmetric Rossby mode, hits the western boundary, and returns behind wave fronts with speeds and meridional structure characteristic of Kelvin modes. After a time τ (the length of time it takes for the signal to return to the region of increasing sea surface temperature) the cooling signal, growing exponentially with an amplitude b as a function of its retarded time t-τ, reaches the east Pacific and begins to erode the warming signal, eventually cooling the region. This is possible only when the remote signal has an amplitude that exceeds the initial signal.

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

The work of Battisti and Hirst (1989) formalized the above description. They found that it could be encapsulated in a single equation, the so-called delayed (or retarded) oscillator equation for the sea surface temperature at a particular place, which can be expressed as

,

where the first two terms on the right-hand side form the linear part of the dynamics, b>c, and the final term allows the system to equilibrate at finite amplitude, with all coefficients real.

The coefficient c multiplies all terms of the thermal equation that locally change the temperature, including the damping by fluxes, while b multiplies only that term of the linearized thermal equation that arises from thermocline depth, i.e., mean upwelling of an anomalous vertical temperature gradient. Because the cooling signal travels a time τ before it shows up as sea-surface-temperature changes, it is effectively shielded from damping by surface fluxes and reaches the eastern Pacific undiminished. If the remote signal, multiplied by b, has a much greater ability to change sea surface temperature than the direct local effects, in the term multiplied by c, then growing oscillatory solutions can result. The cubic term is required for oscillatory finite-amplitude solutions. A somewhat similar argument is given by Cane et al. (1990).

The equation used by Schopf and Suarez (1988) was identical to the equation given above, with the crucial difference that they assumed c>b. For this condition, the linearized equation has only purely growing or decaying solutions, and any oscillatory solutions must arise from nonlinearity. Another way of looking at this is that while both of these models were formulated with the assumption that the coupled atmosphere-ocean mode was stationary and confined to the eastern basin, Schopf and Suarez postulated that the ocean dynamical adjustment time is large compared with the time scale associated with the air-sea interactions, and thus the coupled system is inherently nonlinear. The behavior of the retarded-oscillator equation as b changes magnitude from less than c to greater than c is discussed by Wakata and Sarachik (1994).

The relationship between the Hirst (1986, 1988) calculations of unstable modes, which did not at all resemble ENSO, and the ENSO-like mode in the Zebiak and Cane model remains unclear. The analysis by Hirst assumed a basic state that had no meridional structure for the upwelling. When the mean upwelling, with its rather narrow meridional distribution, was included, the mode found by Hirst was altered to become ENSO-like: it lost its eastward propagation and depended crucially on the boundaries (Wakata and Sarachik 1991b).

In a series of papers, Jin and Neelin (1993a, b) explored the instabilities that developed in a hierarchy of coupled models. Their study included an

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

oceanic general-circulation model coupled to a statistical atmospheric model, and also a simplified model with an equatorial strip of the coupled atmosphere-ocean. Only the essential ocean thermodynamics from the Zebiak and Cane coupled model were retained. In these studies, Jin and Neelin explored the range of coupled atmosphere-ocean modes that exist in the parameter space covered by varying the coupling strength, the ratio of the ocean dynamical adjustment time to the time scale associated with the sea-surface-temperature changes, and the relative strength of the upwelling versus horizontal-advection terms in the ocean thermodynamic equation. By varying the parameters within realistic values, introducing increasingly complete ocean (sea surface temperature) thermodynamics, and tracking the eigenmodes, Jin and Neelin demonstrated how the extreme (idealized) eigenmodes that arise from the problem with a homogeneous basic state (see, e.g., Hirst 1988) give way to an increasingly realistic stationary dominant eigenmode, the delayed oscillator. It is clear from these papers, and the review by Neelin et al. (1994), that (unless some important physics is being neglected, such as feedbacks from clouds) the delayed-oscillator mode is rather robust and is likely to be the dominant, unstable, coupled atmosphere-ocean mode, at least in simple models.

The question remains: Is the retarded oscillator the actual mechanism for ENSO? In the simple models, it appears to be. However, in many important ways, the simple models do not faithfully represent nature. More complete (and complex) coupled atmosphere-ocean general-circulation models have been built, and the question is being asked of these models.

N.-C. Lau et al. (1992) reported on the tropical interannual variability simulated with a coarse-resolution coupled atmosphere-ocean general-circulation model, in which the Kelvin waves were significantly distorted because of the resolution (4° latitude) and numerical algorithms. Moreover, the latitudinal extent of the upwelling, known from theoretical constraints (Wakata and Sarachik 1991b) to be fundamental to the nature of the interannual variability, was not resolved. As a result of, and consistent with, all the results above, the interannual variability in this low-resolution model seems to be well described by a slow, westwardly propagating, destabilized Rossby mode, and not by a delayed-oscillator mode. Resolution sufficiently fine to resolve both the Kelvin mode and the meridional extent of the upwelling is clearly important for testing the delayed-oscillator hypothesis.

The delayed-oscillator theory for ENSO is remarkably consistent with the ENSO simulated by the Hamburg coupled atmosphere-ocean general-circulation model (Latif et al. 1993), and it is qualitatively consistent with the results from the Geophysical Fluid Dynamics Laboratory (GFDL) high-resolution coupled model discussed by Philander et al. (1992). The ENSO simulated with these models shares many similarities with the observed

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

“canonical” ENSO (Rasmusson and Carpenter 1982). However, the interannual variability in a third coupled atmosphere-ocean general-circulation model, reported by Nagai et al. (1992), is rather weak compared to that observed, and does not seem consistent with the delayed-oscillator model of ENSO.

To what extent is the delayed-oscillator theory for ENSO supported by observations? Kessler (1990) has examined the observed wind-stress data, the anomalies of sea surface temperature, and the variability in the upper-ocean thermal structure obtained from the expendable bathythermographs launched by VOS. He concluded that the variability in the thermocline was consistent with that expected from the delayed-oscillator theory. Wakata and Sarachik (1991a) reached the same conclusion by examining the response of a “shallow-water” model to the long series of observed wind stress anomalies (Legler and O'Brien 1988).

While observations of the development and termination of the phases of ENSO are consistent with a delayed-oscillator theory of ENSO, there are clear inconsistencies between the observations and the delayed-oscillator theory during the onset of El Niño. Specifically, the strengthening of the trade winds that usually precedes the event is not a feature of the delayed-oscillator theory, nor of the simplified coupled atmosphere-ocean models from which the theory is derived. Li and Clarke (1994) used equatorial-wave theory and data from tide gauges to construct a record of the “observed” amplitude of equatorial Kelvin waves in the western Pacific. They correlated the reconstructed signal from the equatorial Kelvin waves with an index of the large-scale zonal-wind anomaly. Relying on those time series, they argued that the structure of the lagged-correlation curve was inconsistent with the delayed-oscillator theory of ENSO. The physics associated with the correlation structure reported by Li and Clarke was examined in a complementary analysis by Mantua and Battisti (1994), which used an ocean model forced with the observed anomalies of wind stress for the purpose of obtaining a signal of the equatorial Kelvin waves in the western Pacific (a signal that was highly correlated with the one derived by Li and Clarke using independent data and methods). Mantua and Battisti demonstrated that the delayed-oscillator theory did account for the termination of warm ENSO phases, but that the cold phases are not usually terminated by the ocean-adjustment process described by the delayed oscillator. The basic reason for this difference is that the ENSO phases are not nearly periodic. Nevertheless, the delayed-oscillator equation and theory can be viewed as the current paradigm for a regular ENSO cycle, remembering that such a regular cycle is not observed in nature. The delayed-oscillator mechanism for ENSO is internal to the coupled system of the atmosphere and the ocean. It is not possible to say that the cause of the oscillation resides in either the atmosphere or the ocean, nor

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

is it possible to identify triggers in either: The cause lies in the interaction of the atmosphere and the ocean, and the trigger in some small perturbation.

Finally, it should be noted that the basic idea of coupled atmosphere-ocean instability has been questioned. Penland and Magorian (1993) and Penland and Sardeshmukh (1995) have hypothesized that the tropical Pacific atmosphere and ocean system is stable in a global sense, and that the ENSO variability is best thought of as a response of the tropical Pacific system to stochastic forcing. Thus, without external forcing of the system there would be no ENSO. This hypothesis for ENSO appears to be at odds with the all the aforementioned studies on unstable modes of variability. If the forcing hypothesis is correct, the intermediate-level coupled atmosphere-ocean models and the hybrid coupled models with a numerical ocean and a statistical atmosphere should demonstrate stable, not oscillatory, behavior.

Irregularity of ENSO

The irregularity of ENSO has long been known. It is reflected in the spectrum of the SOI which is peaked, but not spiked, near a period of 40 months (Rasmusson and Carpenter 1982). Investigators have hypothesized many reasons for the irregularity, or aperiodicity, of ENSO. These hypotheses can be loosely grouped into four categories: (1) “noise” internal to either the atmosphere or the ocean and independent of the coupling between the two media; (2) inherent nonlinearity in the coupled atmosphere-ocean system (or nonlinearity in the coupling itself); (3) changes in the external forcing; and (4) interactions between ENSO and the annual cycle.

One suggested cause for irregularity of ENSO is noise internal to the atmosphere or ocean. Specifically, investigators have discussed the impact of the atmospheric weather and other short-term disturbances on time scales from weekly to interannual. Explicitly or implicitly, weather is invoked for ENSO irregularity in the studies of Schopf and Suarez (1988), Zebiak (1989), Goswami and Shukla (1991), and Penland and Magorian (1993).

Nonlinearity in the climate system has also been suggested as a source of irregularity for ENSO. Munnich et al. (1991) examined a streamlined, nonlinear coupled model and found that irregular interannual variability can result from the coupling of the atmosphere and ocean. Recently, Mantua (1994) analyzed the Zebiak and Cane numerical model, demonstrating that the irregularity in its modeled ENSO was caused by an interaction between two unstable coupled atmosphere-ocean modes. He concluded that, although a single ENSO cycle in the model was well described by the delayed-oscillator theory, the irregularity is caused by deterministic, nonlinear processes in the model associated with the interactions between the unstable coupled modes.

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

Forcing from outside the tropics or long-time-scale variations of climate might also cause irregularities in ENSO. Various investigators (e.g., Barnett et al. 1989) have proposed that disturbances propagating into the tropical Pacific from the extratropics or the Indian Ocean can trigger or modify ENSO. Preliminary investigations by Meehl et al. (1993) and Graham et al. (1994) implicate the effect of changing concentrations of atmospheric carbon dioxide. Finally, the definitive study of Mass and Portman (1989) eliminates volcanic eruptions as a forcing of ENSO.

Xie (1995), however, has suggested that the irregularity in ENSO may be explained by a simple superposition of the annual “mode” and the ENSO mode, rather than an interaction between these two dominant modes. This hypothesis for the irregularity in ENSO is based on the very different physical processes that seem to govern the interannual ENSO mode and the annual cycle (see, e.g., Köberle and Philander 1994).

It is not clear, however, to what extent the spectrum of interannual variability in the tropical Pacific is stationary. Direct observational data is not adequate to define even the multidecadal variability in the tropical Pacific. The low-frequency modulation of ENSO could come about from nonlinearity in the coupled ocean-atmosphere-land system, involving either the upper ocean or changes in the (shallow) thermohaline circulation in the subtropics. Unfortunately, the present intermediate models of the coupled atmosphere-ocean system are inappropriate for such studies. Furthermore, it is not yet possible to run coupled general-circulation models for the thousands of model years that are required to examine these issues.

It may be possible to examine low-frequency variability of ENSO using data from paleoclimate studies. The available proxy data have been increasing in variety and geographic extent. For example, proxy data from tropical corals have been used to infer equatorial upwelling (Lea et al. 1989) and extrema in anomalies of wind stress (Shen et al. 1992a, b; Cole et al. 1993). Proxy indicators for the SOI have been derived from coral (Cole and Fairbanks 1990) and tree-ring data (Stahle and Cleaveland 1993). Finally, a connection has been suggested between ENSO and increased marine emissions of dimethyl sulfide at high southern latitudes (Legrand and Feniet-Saigne 1991).

Annual Cycle and ENSO

The annual cycle does not seem to be necessary for ENSO to be realized. Many models without annually varying insolation have proven to be successful at simulating interannual variations that resemble ENSO. These models include the intermediate coupled atmosphere-ocean models (see, e.g., Schopf and Suarez 1988, Mantua 1994) and coupled atmosphere-ocean general-circulation

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

models (see, e.g., Philander et al. 1992, Latif et al. 1993). Furthermore, the annual cycle does not seem to be fundamental to the irregularity of ENSO in these models. One of the aforementioned models (Philander et al. 1992) was run without annually varying insolation, and the ENSO cycles that it produced occurred irregularly, nevertheless. However, ENSO is sufficiently linked to the annual cycle that it is possible to think of a canonical ENSO cycle, formed by compositing observations fixed to the calendar year (Rasmusson and Carpenter 1982).

Recently, simulations using coupled atmosphere-ocean general-circulation models that include annually varying solar forcing have been reported by Nagai et al. (1992) and Robertson et al. (1995a). In both of these studies, the annual cycles in the tropical Pacific of sea surface temperature, surface fluxes, and wind are qualitatively consistent with those observed. Both models develop coordinated interannual variations of the atmosphere and surface ocean in the tropical Pacific, albeit much weaker and only qualitatively similar in pattern to those observed. Overall, then, simulating interannual variability in the presence of a annually varying insolation continues to be a difficult problem. Some models have reproduced interannual variability of sea surface temperature and other models have reproduced the annual cycle, but simulation of the full spectrum of variability remains elusive. It should be recalled that the calculated annual cycle is an average over all the variability present in the system, hence it is not independent of interannual variability.

The processes most crucial for determining the annual cycle appear to be different from those most crucial for determining interannual variability. In particular, we now believe that interannual variability of sea surface temperature in the Pacific depends in an essential way on wind-driven variations of the thermocline, while heat fluxes at the surface act mainly to damp the interannual perturbations (Barnett et al. 1991). Annual variations of sea surface temperature depend critically on heat-flux variations at the surface, and thus in an essential way on radiative and cloud feedbacks (Köberle and Philander 1994). Low-level stratus clouds have presented particular problems. These clouds feed back positively on sea surface temperature, leading to more stratus accompanying low sea surface temperatures and less stratus accompanying high temperatures (when the stratus are replaced by trade cumulus). Existing models generally deal poorly with such clouds.

In some recent studies, investigators have postulated that annual-cycle forcing of the coupled atmosphere-ocean system is responsible for the irregularity of ENSO (Tziperman et al. 1994, Jin et al. 1994, Chang et al. 1994). These investigators studied interannual variability in numerical models by arbitrarily increasing either the annual cycle or the strength of the coupling between the atmosphere and ocean. Lacking annual-cycle forcing, the interan

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

nual variability in these models is periodic. As the coupling parameter is increased, the frequency of the interannual variability increases through a sequence of rational fractions of the annual cycle, but the ENSO-like cycle remains locked in phase with the annual cycle. The transition to higher frequencies is characterized by chaotic variability. In each of these studies, the physics associated with the ENSO mode appears to remain robust (e.g., in Jin et al. 1994 the ENSO mode is characterized by the delayed-oscillator physics) as the forcing is increased.

Several different coupled atmosphere-ocean models were used in these studies to argue for the importance of the annual cycle in producing irregularity of ENSO. However, even in chaotic regimes, under modest forcing, those models have spectra that are qualitatively dissimilar to the observed spectra. In addition, in none of these studies is the annual cycle actually forced by an accurate representation of the insolation cycle; rather, the amplitude of a parameter or process with an annual cycle oscillates. Nonetheless, these preliminary studies provide some insight as to how the annual cycle might interact with ENSO.

Optimal Growth of Disturbances and ENSO Predictability

The predictability of a deterministic system depends on the growth rates of the inevitably present initial errors. When these errors grow quickly, the system is less predictable than when initial errors grow slowly. The fastest-growing errors need not be normal modes of the system. In fact, for the general case of non-adjoint evolution operators, the fastest-growing errors change shape as they grow. Farrell (1990) has shown how the general problem of predictability can be identified with finding the optimals, i.e., the fastest-growing disturbances, of a given initial state. Blumenthal (1991) was the first to study theoretically the fastest-growing disturbances in the tropical Pacific. The finite-amplitude disturbances that grow most rapidly, in a system that is not linearly unstable, are said to experience “optimal growth”. He examined output from a freely evolving simulation obtained from the coupled atmosphere-ocean model of Zebiak and Cane (1987). The optimal perturbations were determined by deriving a linear autoregressive (Markov) model from a reduced set of output quantities. In his study, Blumenthal found that when a fixed-amplitude perturbation is applied to the model output, the maximum growth over nine months is realized when the perturbation is applied in February. He also argued that the dependence on season of the disturbance growth is consistent with the seasonal dependence of forecast skill reported by Cane et al. (1986).

Penland and Sardeshmukh (1995) built a similar autoregressive model using the observed sea surface temperature. In contrast to Blumenthal, they

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

reported that disturbance growth is not a function of season, and hypothesized instead that ENSO results from white-noise forcing of a globally linear system. Unfortunately, the results cannot be directly compared, for two reasons. First, the observed sea-surface-temperature data used to derive the Markov model are dominated by a record of unusually periodic ENSO events during the 1970s and 1980s. Hence, the regression model might have captured variability that is associated with ENSO, but does not in turn affect ENSO, and might lead to apparent (but false) predictability. In addition, because Penland and Sardeshmukh used sea-surface-temperature data but no thermocline data in their analysis, their hypothesis that the coupled system is globally stable required that the ocean dynamics equilibrate on time scales that are rapid compared with the changes in sea surface temperature. Analyses of observations by Clarke and Li (1994) and of ocean hindcasts by Mantua and Battisti (1994) indicated that this assumption is not valid for most of the ENSO cycle. Starting from a few months prior to the peak of a warm phase of ENSO, extending through that peak, and into the cold phase of ENSO, the observations are consistent with the physics of the delayed-oscillator model for ENSO. Thus, it is not surprising that Penland and Sardeshmukh found that the Markov model could not predict with skill the termination of ENSO phases.

The studies of Blumenthal (1991) and Penland and Sardeshmukh (1995) come to somewhat disparate conclusions concerning the predictability and stability of the coupled atmosphere-ocean system. Nonetheless, both studies suggest that the traditional approach to assessing the variability in the atmosphere-ocean system—identification of the fastest-growing normal modes—may not be the most instructive for understanding the predictability of the system. The linearized system need not be self-adjoint, either because of the coupling or because of the spatial variation of the background state, so the normal modes of the system need not be orthogonal. Thus, the fastest-growing normal mode cannot, in general, determine disturbance growth in the coupled models, and forecast skill is better explained by a study of the projection of the initial conditions on the optimal perturbations.

WORKING IN A LARGER COMMUNITY

Some lessons of the TOGA Program are difficult to measure. These include changes in the culture of science and public awareness of science. The TOGA Program strengthened the bond between oceanic and atmospheric scientists, permanently altering some of the institutional and disciplinary arrangements in the oceanic and atmospheric sciences. It contributed to an awareness, in both the scientific community and the public at large, that there are aspects of

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

seasonal-to-interannual climate variability that can be monitored, understood, and predicted—El Niño has become a household word.

The TOGA Program had an important influence on the culture of interdisciplinary research in the oceanic and atmospheric sciences. Knowledge of both the meteorology and oceanography of the tropical Pacific proved essential to unraveling the mysteries of ENSO. Although a small number of scientists have always been able to work across traditional disciplinary boundaries, TOGA taught a larger number how to do this. The TOGA Program increased the number of “amphibious” scientists, those who are equally at home with research in the atmosphere and in the ocean, or at least have a clear understanding of the nature of the coupled system. TOGA field programs, such as COARE, brought together oceanographers and meteorologists to formulate, plan, and implement new observational strategies for measuring and understanding processes that couple the ocean and atmosphere. Oceanic and atmospheric modelers collaborated on physically based coupled models for understanding and predicting ENSO. New strategies were developed for assimilating unprecedented amounts of ocean data to initialize coupled ocean-atmosphere prediction models. For the first time, ocean scientists were able to perform quantitative studies of the predictability of seasonal-to-interannual variations of large-scale ocean thermal and flow fields. The oceanographers and meteorologists who have collaborated within the TOGA Program for a decade intend to continue that collaboration in the fifteen-year post-TOGA Program, called GOALS (NRC 1994b).

Operational monitoring by the TOGA Observing System of interactions between the ocean and atmosphere encouraged cultural change. The TOGA Program promoted the free distribution of observational data and of numerical-model analyses of wind, surface and subsurface temperature, and other variables at sites remote from land. Many of the observations were made available within a day, and all of them within 30 days, of acquisition. The motivation for this requirement can be traced to scientists' frustration at their inability to observe the evolution of the 2- to 3-month onset phase of the 1982–83 El Niño. The introduction of satellite-communication technology greatly enhanced the ability to transmit and receive measurements from the VOS network and from moored and drifting buoys. Even though the 30-day oceanic operational time scale is much longer than the operational time scale for atmospheric observing systems, immediate distribution of TOGA data drove significant changes in the patterns of data collection and distribution in the ocean sciences. Basic oceanographic data had often been unavailable for periods as long as two years. Although meteorologists were accustomed to the rapid dissemination of operational atmospheric data, other information had often been viewed as proprietary, at least for a period of time. Research data, especially data for the oceans, and detailed results from numerical simulations usually were sequestered. The

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

TOGA Program, with its emphasis on analyses of the current state of climate systems, and on comparing the predictions of many modeling schemes with reality, encouraged and rewarded the immediate sharing of data. Many scientists learned that sharing their basic data, before it was fully analyzed, did not undermine their careers.

It is instructive to compare the TOGA Observing System in the Pacific Ocean combined with TOGA's experimental prediction models to the early attempts at establishing a numerical weather-prediction system for the United States. The implementation of the TOGA Observing System in the Pacific supported the development of numerous data-product bulletins and newsletters that began publication about a decade ago (see Appendix B for a listing). One of the first scientific uses of operational oceanographic information was demonstrated in August 1987 at the General Assembly of the International Union of Geodesy and Geophysics (IUGG), where an interpretative analysis of the June 1987 El Niño oceanographic conditions (sea surface temperature, near-surface currents, thermocline depth, sea-surface height, surface wind, and other variables) in the tropical Pacific was presented. Analyses and experimental short-range climate forecasts for the atmosphere and ocean are now routinely distributed to scientists and the general public through NOAA's Climate Diagnostics Bulletin.

In addition to changing the views and approaches of established scientists, the TOGA Program had an impact on the generation of climate researchers now in training. At present, at least ten major research universities have collaborative oceanic and atmospheric science programs. The TOGA Program recognized the need to entrain young researchers, and was instrumental in establishing the NOAA Postdoctoral Program on Climate and Global Change. In the past, postdoctoral fellowship programs had not generally been associated with research programs. (An exception is the Ocean Drilling Program, which also maintains a postdoctoral fellowship program.) The Climate and Global Change Postdoctoral Program awarded thirty fellowships by the end of 1994, covering a wide range of disciplines; ten of the fellowships were awarded in areas directly relevant to TOGA scientific objectives. A high percentage of the alumni of the program have received tenure-track positions at universities or senior research positions. It is envisioned that the NOAA Postdoctoral Program will continue to emphasize seasonal-to-interannual scientific objectives through its association with the GOALS Program.

Many of the scientific and institutional arrangements in existence prior to the TOGA Program were more of a hindrance than a help to such interdisciplinary activities. Oceanographers and atmospheric scientists found that their traditional working environments were inadequate for attacking large interdisciplinary problems such as those presented by ENSO. TOGA participants

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×

largely solved these problems. They developed a number of arrangements for coping with the realities of scientific and institutional activities that cross the boundaries of discipline and function. Achieving a balance among monitoring, modeling, empirical studies, and process studies required unprecedented cooperation among NOAA, NSF and other federal agencies when reviewing funding proposals and when filling crucial gaps during the implementation phase of the program. At the same time, meteorologists and oceanographers at universities and federal laboratories took a more active role in providing balanced scientific advice to the relevant federal agencies through the National Research Council's TOGA Panel and other structures. The coincidence of scientific interests and national priorities may allow the interdisciplinary collaborations developed during the TOGA Program to persist, and may help to define more clearly the relationship between climate research and societal needs. The development of the ability to predict seasonal-to-interannual climate variations has changed the ways in which many scientists think about their work and their obligations.

Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 81
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 82
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 83
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 84
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 85
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 86
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 87
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 88
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 89
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 90
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 91
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 92
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 93
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 94
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 95
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 96
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 97
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 98
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 99
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 100
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 101
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 102
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 103
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 104
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 105
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 106
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 107
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 108
Suggested Citation:"4. WHAT WE'VE LEARNED." National Research Council. 1996. Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program. Washington, DC: The National Academies Press. doi: 10.17226/5003.
×
Page 109
Next: 5. ORGANIZATION »
Learning to Predict Climate Variations Associated with El Nino and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program Get This Book
×
Buy Paperback | $54.00 Buy Ebook | $43.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

The TOGA (Tropical Ocean and Global Atmosphere) Program was designed to study short-term climate variations. A 10-year international program, TOGA made El Nino a household word. This book chronicles the cooperative efforts of oceanographers and meteorologists, several U.S. government agencies, many other nations, and international scientific organizations to study El Nino and the Southern Oscillation (ENSO).

It describes the progression from being unable to detect the development of large climate variations to being able to make and use rudimentary climate predictions, especially for some tropical countries. It examines the development of the TOGA Program, evaluates its accomplishments, describes U.S. participation in the program, and makes general recommendations for developing better understanding and predictions of climate variations on seasonal to interannual time scales.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    Switch between the Original Pages, where you can read the report as it appeared in print, and Text Pages for the web version, where you can highlight and search the text.

    « Back Next »
  6. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  7. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  8. ×

    View our suggested citation for this chapter.

    « Back Next »
  9. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!