The preceding sections have described the evolution of the TOGA observing system and how it has contributed to scientific progress in studies of short-term climate variability during the TOGA decade. Development of this observing system was a major technological achievement, which revolutionized climate monitoring programs by stimulating increased demand for real-time ocean data delivery. The data from this observing system were essential to fostering advances in many aspects of TOGA research, including the following: (1) documentation of the ENSO cycle and related phenomena, such as the mean seasonal cycle and intraseasonal variability, with unparalleled resolution and accuracy; (2) testing of ENSO theories, such as the delayed oscillator; (3) development of new theoretical concepts relating to ocean-atmosphere interactions on seasonal-to-interannual timescales; (4) development of oceanic, atmospheric, and coupled ocean-atmosphere models; and (5) development of ocean data assimilation systems for improved climate analyses and for initializing climate prediction models. In short, measured against the goals of TOGA stated in section 1, the TOGA observing system was a tremendous success.
It is fortuitous that TOGA spanned a decade in which there was both a large swing from El Niño to La Niña conditions (19861989) and a period of prolonged anomalous warming (19911995). The dramatic change from El Niño to La Niña during the first half of TOGA heightened awareness about the importance of the cold phase of the ENSO cycle [e.g., Trenberth and Branstator, 1992; Halpert and Ropelewski, 1992] and afforded the opportunity to examine sharp contrasts between extreme climatic conditions in the Pacific and their impacts worldwide [e.g., Palmer et al., 1992]. On the other hand, the period 19911995 was unprecedented when viewed in the context of modern instrumental records dating back to the last century. The warm conditions evident during 19911995 have been interpreted as a single warm phase ENSO event, in which case it would be the longest in the past 100 years [Trenberth and Hoar, 1996]. An alternative interpretation is that 19911995 was characterized by three distinct warm events [Goddard and Graham, 1997], implying a recurrence rate significantly higher than the average 3-4 years expected from historical records. Either interpretation identifies 19911995 as unique in the modern record.
It is interesting to compare the evolution of warm events in Plate 1 with the Rasmusson and Carpenter  composite, which was based on El Niño events from the 1950s to the 1970s. Rasmusson and Carpenter  suggested that anomalous surface warming occurs first off the South American coast, peaking in MarchMay, then progresses westward along the equator into the interior basin, reaching a "mature phase" in DecemberFebruary. Subsequently, warm SST anomalies and associated westerly wind anomalies weaken and eventually disappear by the following May. There were features common among the El Niño events observed during TOGA, such as anomalous warming in the equatorial cold tongue and large-scale weakening of the trade winds in the central and western Pacific. However, like the 19821983 El Niño prior to TOGA, none of these warm events evolved strictly according to the canonical Rasmusson and Carpenter  composite.
Significant differences in duration, phasing, and spatial warming patterns observed during events of the 1980s and early 1990s defy easy categorization. Most pronounced warmings in the eastern and central Pacific in the 1990s, for example, occurred in boreal winter 19911992, boreal spring 1993, and boreal fall 1994. This disparate timing of maximum warm anomalies raises questions about the dynamical links between the seasonal cycle and the evolution of El Niño. Moreover, South American coastal warming did not generally precede maximum SST anomalies in the equatorial cold tongue, as in the Rasmusson and Carpenter  composite. Deser and Wallace  had earlier found that coastal warmings appear to be only loosely coupled to the broader basin-scale manifestations of El Niño, a result that appears also to apply to warm events observed during the TOGA decade. Also, considering the 1993 and 19941995 warmings as separate events, their duration was significantly shorter than the norm of 1218 months for El Niños of the past.
Consistent with the complexity of the observed interannual variability, tests of ENSO theories using data prior to and during the TOGA decade suggest that more than one set of mechanisms can give rise to ENSO timescale warm and cold events in the tropical Pacific. The delayed oscillator theory, for example, can often, but not always, be invoked to explain the termination of ENSO warm events. On the other hand, delayed oscillator physics cannot generally account for the onset of warm ENSO events. New physical hypotheses are being formulated regarding the ENSO cycle, based on the failure of existing theories to explain the full range of observed variability.
The unusual warm conditions prevailing near the date line in the equatorial Pacific during 19911995 raise questions about the relationship between the ENSO cycle and decadal timescale variability. The persistent warm anomalies are the reflection of a decadal timescale variation that has higher latitude manifestations in North and South Pacific SSTs [e.g., Latif et al., 1997; Wallace et al., this issue; Zhang et al., 1997]. This decadal mode may result from decadal modulations in the intensity and/or frequency of ENSO events, or it may be a mode of coupled ocean-atmosphere variability with dynamics distinctly different from those of ENSO. In either case the decadal timescale of this variation and its manifestations at higher latitudes suggest a link to decadal timescale processes that maintain the equatorial thermocline [Fine et al., 1987; McPhaden and Fine, 1988]. These processes involve the ocean thermohaline circulation which couples the tropical ocean to the subtropical and higher-latitude North and South Pacific Ocean [e.g., McCreary and Lu, 1994; Lu and McCreary, 1995]. Decadal timescale variations in the overlying atmospheric circulation at midlatitudes [Trenberth and Hurrell, 1994; Latif and Barnett, 1995; Zhang et al., 1997] alter patterns of air-sea heat exchange, providing a mechanism by which the formation of thermocline water masses can be affected in density surface outcrop regions [Miller et al., 1994]. A theory for self-sustaining decadal time scale oscillations involving ocean-atmosphere interactions and heat transports between the tropical and extra-tropical oceans has been proposed recently by Gu and Philander .
Observed variability during TOGA also suggests a possible connection between El Niño and global warming. Average SSTs in the tropical Pacific were unusually high during the 1980s and 1990s, at the same time that there was a trend for warmer global surface air temperature. The tropical Pacific SSTs were warmer because of a greater intensity, frequency, and/or duration of warm ENSO events. Two recent studies [Kumar et al., 1994; Graham, 1995] based on atmospheric model simulations forced with observed SSTs for the 1980s and 1990s suggested that the warming of global surface air temperature for this period may have been induced by the warming of SST in the tropical Pacific. Tropical Pacific SSTs in these simulations were prescribed from observations, however. It is possible that the character of ENSO changed and that SSTs were warmer because of anthropogenic greenhouse gas warming [Trenberth and Hoar, 1996]. There is no consensus on this issue, and recently, Cane et al.  argued that global warming should lead to a cooling of the tropical Pacific. Clearly, resolution of the questions concerning ENSO, decadal variability, and anthropogenic greenhouse gas warming will require considerably more research.
TOGA demonstrated the synergy that can emerge from the combined use of data and dynamical models. As a measure of progress, prior to TOGA, there was no system of routine data assimilation for tropical ocean climate analyses and no routine short-term climate prediction efforts. However, during TOGA, models were used to help design the observing system, and data from the observing system were then used to foster model development and to initialize models for short-term climate prediction. Now many ENSO prediction modeling groups have been established [National Weather Service, 1997], and prediction models, initialized with TOGA data sets, show significant skill for lead times of up to 1 year. The skill of these predictions is likely to improve as we learn more about the underlying dynamical processes involved in ENSO and as models and assimilation systems improve.
TOGA also demonstrated the synergy that can emerge from the combined analysis of satellite and in situ measurements. In situ measurement systems provide high-accuracy information on both surface and subsurface ocean variability, the latter of which is not directly accessible to satellites. In situ measurement systems also provide necessary data for ongoing calibration and validation of satellite retrievals. The strength of the satellite data, on the other hand, is their near-global coverage and uniform time-space sampling characteristics. Unfortunately, the full potential for satellite missions for climate research during TOGA was not realized in part because most of the satellite missions were sponsored for reasons other than climate research and some (like TOPEX/POSEIDON) were originally intended as one-time experimental missions. Similarly, the launch of NSCAT was so often delayed that eventually it fell outside the TOGA time frame. Coordination between agencies and countries sponsoring satellite missions did not always succeed because of uncertainties in funding, payload development, and launch dates. This lack of coordination led to a 2-year gap in altimeter measurements between the U.S. Navy Geosat mission and the ERS-1 mission. Nonetheless, the tremendous value of those satellite data that were acquired during TOGA bodes well for the future application of satellite measurements to ocean climate studies.
As a result of TOGA, we are now entering a new era of climate research and forecasting. The World Climate Research Program (WCRP) has embarked on a 15-year (19952010) study of Climate Variability and Predictability (CLIVAR), one element of which, the Global Ocean-Atmosphere-Land Studies (GOALS) program focuses on seasonal-to-interannual variability [National Research Council, 1994b; World Climate Research Program, 1995b]. Also, a newly instituted International Research Institute for Climate Prediction (IRICP) will begin to issue routine short-term ENSO forecasts, conduct research on ways to improve those forecasts, and help to coordinate the use of the forecast products for various socioeconomic applications [International Research Institute for Climate Prediction Task Group, 1992]. Likewise, some national meteorological centers are already routinely issuing climate forecasts [e.g., National Centers for Environmental Prediction, 1996], and others intend to do so in the near future.
The success of these research and forecasting activities requires that essential elements of the TOGA observing system be continued for the foreseeable future. Explicit guidance on the development of post-TOGA climate observing systems is contained in the reports of various planning committees that have considered the observational needs of future climate programs [e.g., National Research Council, 1994b; Ocean Observing System Development Panel, 1995]. These reports are unanimous in their recommendations to continue the observing system developed under TOGA in support of short-term climate prediction. For some components of the observing system this may require transfer of the responsibility for long-term, systematic measurements from the research community to the operational oceanographic and/or meteorological communities. Effecting this transition will be challenging because there is no precedent for institutionalizing an observing system built entirely within the framework of a climate research program.
The need for long-term support of critical climate measurements has motivated planning for the Global Climate Observing System (GCOS) as well as the climate module of the Global Ocean Observing System (GOOS). These emerging international programs, modeled loosely on the World Weather Watch for weather forecasting, are intended to foster and coordinate measurements for a wide range of climate applications. As national commitments were essential in developing the TOGA observing system, so will they be essential in maintaining the observing system after TOGA. GOOS and GCOS are at different stages of evolution in different countries involved in supporting climate observations, complicating coordination at the international level. However, CLIVAR and GCOS/GOOS have recognized the merits of collaboration to ensure that an effective post-TOGA observing system is maintained. Therefore, in the near term, it is almost inevitable that the post-TOGA observing system will be maintained under a mix of research and operational support.
In the meantime it is of paramount importance that the existing data stream not be interrupted. Tremendous effort was expended in developing an adequate infrastructure to support the collection of critical data sets during TOGA. This infrastructure, involving cooperative relationships between research institutions and government agencies in several countries, was established through painstaking evaluation and oversight by the international scientific community over the course of 10 years. This infrastructure is fragile; premature curtailment or disruption of observational efforts could have disastrous and long-lived effects on the development of future climate observing systems. Thus a conservative approach must be adopted in recommending changes to either observational strategies or to the organizational framework in which the observations are supported. Conservatism does not imply that the observing systems for post-TOGA climate studies should be static in their design, though. On the contrary, the observing system should be flexible enough to take advantage of new advances in technology. Likewise, it is essential that there be ongoing assessments of the observing system design and that these assessments be guided by scientific priorities.
Much of this paper has dealt with the TOGA observing system in the tropical Pacific, where TOGA focused its effort as a first priority. Clearly, adequately observing the tropical Pacific was a sine qua non for making progress on understanding and predicting ENSO. In contrast, scientific questions relating to the climatic impacts of ocean-atmosphere interactions were not as thoroughly explored in the other two ocean basins, and resources were too limited to allow for uniform development of observing system components throughout the global tropics during the TOGA decade. Nonetheless, as a consequence of TOGA, our understanding of ocean-atmosphere interactions in the Indian and Atlantic Oceans has significantly improved. New hypotheses have emerged, such as the role of the Indian and east Asian monsoons in ENSO [e.g., Webster and Yang, 1992] and the role of both Pacific and Atlantic SST variations in affecting climate in the Atlantic basin [e.g., Servain, 1991; Zebiak, 1993; Delecluse et al., 1994]. Also, while there is ongoing debate about the origin of ENSO-related SST anomalies in the North Pacific and their effects on climate variability over North America [e.g., Lau and Nath, 1994], even stronger decadal timescale variations in North Pacific SSTs have recently been documented [e.g., Zhang et al., 1997]. The relationship of these decadal variations to ENSO and to global climate variability, in general, needs to be better understood. Thus geographic expansion of in situ observational efforts should be carefully considered as part of the post-TOGA climate research agenda.
Return to previous section or go to Appendix A
PMEL Outstanding Papers
PMEL Publications Search