GOALS (Global Ocean-Atmosphere-Land System) for Predicting Seasonal-to-Interannual Climate: A Program of Observation, Modeling, and Analysis (1994)

A new age of climate prediction is beginning as the groundbreaking effort of the 10-year (1985–1994), international, Tropical Ocean and Global Atmosphere (TOGA) program comes to an end. Before the TOGA program began, it was not possible to carry out even the minimal observational and modeling efforts needed to predict short-term (seasonal-to-interannual) climate variations. Now, 10 years after the inception of TOGA, the necessary observing systems and models are in place for a concerted effort to accomplish this objective of short-term prediction. The program proposed in this report—the Global Ocean–Atmosphere–Land System, or GOALS, program—is designed to capitalize on the observational and theoretical progress made by TOGA and to meet the new challenges of climate prediction.

The most familiar variation of the climate is the annual cycle. The change of seasons associated with the annual cycle is also perhaps the most predictable variation. In the middle latitudes, where the United States is located, this cycle is often described in terms of warm (or hot) summers and cool (or cold) winters. In the tropics, the annual cycle is more evident in changes in precipitation and winds. Most people assume that the annual cycle is well characterized and understood. When the TOGA program began, however, the regular progression of the seasons was not well described over much of the oceans. Significant gaps still remain in our understanding of the global annual cycle.

There is great societal value in the seemingly trivial forecast that

the conditions for each season or month will be close to that period's climatological mean. However, the progression of the annual cycle is not identical from year to year. The deviations from the normal, or averaged, annual cycle are referred to as climate anomalies or variations. Skill in climate prediction is measured by how well forecasters predict the variations from climatological values. Prediction of climate variations with lead times of a season to a few years is the ultimate objective of GOALS.

This introductory chapter sets the scene for the proposed GOALS program. It outlines the developments that led to the establishment of TOGA, including the discovery of the phenomenon called ENSO—for El Niño/Southern Oscillation—as a major theme in climate research. Following are descriptions of TOGA, its major objectives, and its achievements, which have prepared the way for future developments.

Chapter 2 introduces the GOALS program proposed by the Climate Research Committee of the National Research Council (NRC). It explains that GOALS would expand upon the original TOGA focus by laying a broader foundation for dynamical prediction of global climate variability at seasonal-to-interannual time scales. The chapter presents the scientific objectives of GOALS and lists nine high-priority science questions that the program should address.

Chapter 3 discusses spatial variability and temporal variability as they relate to predicting climate on seasonal-to-interannual scales. With respect to spatial variability, the chapter discusses the complexity of interactions among the upper ocean, the atmosphere, and land, as well as the interactions between the tropics and extratropics, and the effects of all these interactions on climate. It then describes seasonal, annual, interannual, and decadal variability as they relate to GOALS objectives.

Chapter 4 highlights the four major elements—modeling, observations, empirical studies, and process studies—envisioned for GOALS. It outlines the general sequence in which the work of the program would proceed. A section on each of the four elements describes the challenges within that area and their relation to other components of the program.

The data acquisition and data management needs of GOALS are discussed in Chapter 5. The objectives for a detailed data management plan are set out. The scientific and programmatic linkages of GOALS to other ongoing and proposed research efforts are addressed in Chapter 6. Chapter 7 describes the organizational structure proposed for GOALS and the relationship of the U.S. program to CLIVAR, an international program on climate variability and predictability de-

veloped under the auspices of the World Climate Research Program (WCRP).

Appendices to the report provide: background material on the present status of short-term climate prediction (Appendix A), the agenda and list of participants at the March 1993 study conference organized by the GOALS Steering Committee of the Climate Research Committee (Appendix B), and a list of acronyms and other initials (Appendix C).

It is well known that day-to-day atmospheric fluctuations are not predictable beyond about 2 weeks because of the chaotic nature of atmospheric dynamics. However, it has also been recognized that there is predictability in the midst of chaos. The interactions among the various physical components of the climate system—namely, the atmosphere, the upper oceans, the land, and the cryosphere (snow and ice)—produce long-period variations in the climate system that enhance its predictability. In particular, the relatively slow-changing surface conditions lead to a predictable slow variation on the statistics of the atmosphere (Namias, 1969, 1975; Charney and Shukla, 1981; Shukla, 1984). These surface conditions include sea-surface temperature (SST), sea ice, snow cover, soil moisture, vegetation type, and surface-land temperature. An outstanding example of long-period climate variation is produced by interactions between the tropical oceans and atmosphere. The recognition that the coupled tropical ocean and global atmosphere is potentially predictable led to the establishment of the TOGA program, described in the next section.

Before looking at the beginnings of the TOGA program, it will be useful to discuss briefly four terms—Southern Oscillation, El Niño, La Niña, and ENSO (El Niño/Southern Oscillation).

Southern Oscillation is the term Sir Gilbert Walker (Walker and Bliss, 1932) coined in the early part of this century during his attempt to predict the year-to-year fluctuations of India's monsoon rainfall. He described the Southern Oscillation as:

We now know that Walker was observing the two phases that manifest the Southern Oscillation in the tropical Pacific—El Niño and La Niña , or the warm phase and the cold phase, respectively.

The most widely known of the two phases is El Niño, which has come into the popular vocabulary in recent years, partly because of the severity of some recent ''warm episodes.'' For example, in 1972 an El Niño event along the coast of South America was accompanied by alarming declines in the anchovy population, which had major repercussions for the local fishing industry and the world commodities market (Barber, 1988). The winter of 1976–1977, which coincided with a warm episode, brought drought in California and record cold and fuel shortages in much of the central and eastern United States. And the most devastating El Niño of the past century was the 1982–1983 ENSO "event" (as the large-scale El Niño warm episodes have come to be called). The most recent ENSO events were in 1986–1987 and 1991–1992. Thus, El Niño has attracted worldwide attention in recent years with the vastness of its reach and the severity of the weather conditions and economic dislocations that accompany it.

Historically, however, it was quite the opposite perception that accounted for its name. El Niño (the Spanish name for the child Jesus) was used to refer to the southward-flowing current off the west coast of South America that appeared shortly after Christmas some years. Locally it would bring heavy rains, which in turn brought an abundance of vegetation—including crops, grass for grazing, and other products—that seemed like gifts.

It is now known that the anomalous surface-water temperatures off the South American coast during the El Niño and La Niña phenomena often extend thousands of kilometers offshore and are but one aspect of anomalous oceanic and atmospheric conditions throughout the tropical Pacific. This irregular alternation of anomalously warm and cold temperatures and the concomitant rainfall variations across the tropical Pacific Ocean are referred to as ENSO. The ENSO phenomenon occurs every 4 years or so. Although it is often irregular and develops in different ways, many of its slowly evolving features can be simulated and predicted by models that include only the tropical Pacific Ocean and the atmosphere above it.

It is now well understood that ENSO is a coupled phenomenon arising from interactions between the atmosphere and the ocean, so that dynamical simulation and prediction are based on coupled atmosphere–ocean models. The ENSO signal offers the best initial hope for climate predictions, and models are now beginning to be able to predict ENSO events in advance: the warm phases of the 1986–1987 and 1991–1992 events were predicted more than a year in advance by some models (though mispredicted by others).

Other important interannual variations of climate are due to processes and mechanisms other than those associated with ENSO. To

achieve the objectives of predicting climate variations (for example, flooding and drought in the global tropics and in the midlatitudes), observations and scientific study must be expanded beyond the tropical Pacific. Nevertheless, the work of the TOGA program laid the foundation for future developments.

TOGA evolved from loosely coordinated research efforts that gained momentum in the early 1970s. Pioneering empirical studies by Jacob Bjerknes provided evidence that the long-term persistence of the global climate anomalies (with respect to the averaged annual cycle) associated with Walker's Southern Oscillation (Walker and Bliss, 1932) is closely associated with slowly evolving SST anomalies in the eastern and central equatorial Pacific. Bjerknes argued that the periodic strengthening and weakening of the southeasterly trade winds affect equatorial SSTs, which, in turn, influence the large-scale patterns of precipitation. Bjerknes viewed the shifts in rainfall patterns as forcing for the global anomaly pattern embodied in the Southern Oscillation (Bjerknes, 1966, 1969).

During the 1970s¹, empirical studies of the tropical upper ocean led to the hypothesis that the evolution of equatorial SST anomalies during the onset of a warm episode in the eastern Pacific could be explained by a coupling between the decreasing southeasterly trade winds and equatorial wave activity in the tropical upper ocean through the mechanical forcing by surface wind stress (Wyrtki, 1975). Simplified numerical models of the upper ocean provided strong support for this hypothesis and provided an interpretation of ENSO that could be tested on the basis of more detailed field observations and dynamical numerical models (for example, Busalacchi and O'Brien, 1981).

At the same time, atmospheric scientists began to exploit the historical records of surface meteorological data collected from volunteer observing ships, as well as upper-air data. From these they obtained a more comprehensive and detailed picture of the spatial patterns of SST and surface-wind anomalies associated with ENSO (Rasmusson and Carpenter, 1982). Empirical evidence of remote connections between the tropical anomalies and the middle- and high-

latitude atmospheric circulation in the Northern Hemisphere (Horel and Wallace, 1981) was supported by analytical and numerical investigations of the response of the atmosphere to imposed SST or heating anomalies characteristic of ENSO episodes (Hoskins and Karoly, 1981; N.-C. Lau, 1985).

Although plans to begin a concerted study of ENSO in these pre-TOGA years already existed, the intense warm episode of 1982–1983 galvanized the tropical climate research community into action. Several factors contributed to the new determination to understand and predict the ENSO phenomenon: this extreme episode of 1982–1983 started without being noticed; a number of measuring programs² were in place to capture the details of the evolving episode as never before possible; the episode was of particular interest because it differed substantially from the "composite event" previously described by Rasmusson and Carpenter (1982); and, finally, the climatic consequences of this episode were very large in scale.

The early to middle 1980s were marked by increased activities in measuring the atmosphere-ocean system³ and understanding the coupling between the atmosphere and the ocean in terms of surface winds and SSTs, especially in the tropics. In the ocean, the observed equatorial thermocline variations (the thermocline being the level or layer of the ocean [usually above 300 meters in the tropics] where the temperature gets rapidly colder with increasing depth) were shown to be simulated reasonably by simple ocean models, when forced with the climatologically varying winds, determined from data collected over long periods from volunteer observing ships in the tropics and then subjectively analyzed into wind fields (Busalacchi and O'Brien, 1981; Busalacchi et al., 1983). This success indicated that it is only the largest-scale aspects of the wind fields that are responsible for the observed large-scale thermocline variations. The SST was shown to be reasonably simulated by ocean general circulation models (GCMs) in terms of the wind forcing and a parameterized heat-flux forcing at the surface (Philander and Seigel, 1985).

For the atmosphere, the simple Gill model (Matsuno, 1966; Webster, 1972; and Gill, 1980), its extensions (for example, Zebiak, 1986), and

atmospheric GCMs were proving able to simulate the large-scale aspects of surface winds in terms of the thermal forcing of the atmosphere. By the middle 1980s, it was becoming more widely accepted that SST anomalies in any of the three tropical oceans (that is, Atlantic, Indian, and Pacific) can, at least in principle, induce global patterns of atmospheric circulation anomalies. Predictability studies with atmospheric GCMs showed that the potential predictability of the tropical atmosphere is far greater than that of the extratropics (Charney and Shukla, 1981).

With elements of ENSO simulation in place and the will to understand being strong, the TOGA program was born, both nationally (NRC, 1983) and internationally (WMO, 1985). Its three overall objectives were (WMO, 1985):

1. To gain a description of the tropical oceans and the global atmosphere as a time-dependent system, to determine the extent to which this system is predictable on time scales of months to years, and to understand the mechanisms and processes underlying that predictability;

2. To study the feasibility of modeling the coupled ocean–atmosphere system for the purpose of predicting its variations on time scales of months to years; and

3. To provide the scientific background for designing an observing and data transmission system for operational prediction if this capability is demonstrated by coupled ocean–atmosphere models.

From the inception of TOGA, the interaction of the ocean and atmosphere in the tropics was recognized as a central issue in predictability and prediction.

The TOGA program began on 1 January 1985 and will end on 31 December 1994. Its progress was reviewed by the NRC (1990) and at an international conference held in 1990 (WCRP, 1990). The program has been remarkably successful in addressing the three TOGA goals stated above. However, it should be noted that essentially all the work and progress have been in and over the tropical Pacific Ocean. By focusing on understanding and predicting ENSO, the largest identified signal of interseasonal and interannual variability in the circulation of the atmosphere and the tropical Pacific upper ocean, the TOGA program was able to define clearly its observational requirements and to motivate process studies. Eventually, the requirements of the developing coupled atmosphere–ocean models established the priori-

ties for the communication, organization, and synthesis of the various TOGA data sets.

As a result of the TOGA program, the coupled atmosphere–ocean interactions responsible for the ENSO phenomenon are thought to be basically understood. It is now known that the atmosphere-ocean system over the tropical Pacific has some aspects that are predictable a year or more in advance. An observational system has been put in place to monitor the tropical Pacific and to provide the initial data for coupled predictions of aspects of the ENSO phenomenon in and around the tropical Pacific. Predictions are being made with simplified statistical and coupled models, and prediction systems are being developed with more complicated and inclusive general circulation models. (See Appendix A for a more complete description of the current status of seasonal-to-interannual prediction developed under the TOGA program.)

Two ongoing subprograms have been developed to address the goals of TOGA:

1. The TOGA Observing System and, in particular, the TOGA TAO (Tropical Atmosphere Ocean) array—a network of about 65 moored thermistor chains in the tropical Pacific measuring surface meteorological data and subsurface thermal data (Hayes et al., 1991); and

2. The TOGA Program on Prediction (T-POP), a national research program concentrating on seasonal-to-interannual predictions using coupled atmosphere–ocean models (Cane and Sarachik, 1991; Sarachik, 1991).

The TOGA program has also fostered a large international process study in the western tropical Pacific, the TOGA Coupled Ocean–Atmosphere Response Experiment (COARE) (Webster and Lukas, 1992). This process study is examining the mutual interactions between the ocean and atmospheric convective activity over the ocean and is addressing the coupled modeling of these interactions. An operational Pacific Ocean modeling and data assimilation effort designed to synthesize irregularly taken ocean data into dynamically consistent fields of information has been started and continues to be run at the Climate Analysis Center of the National Weather Service (Derber and Rosati, 1989; Leetmaa and Ji, 1989; Miyakoda et al., 1990). This effort has focused the attention of researchers on the problems of ocean-data assimilation and has been useful as a testbed for investigating the impact of operational ocean data.

The TOGA program has also initiated the planning for an Inter-

national Research Institute for Climate Prediction (IRICP; Moura, 1992). The proposed research institute will concentrate on the prediction of aspects of ENSO and on the successful utilization of these forecasts by countries directly affected by interannual variations of precipitation caused by ENSO.

Not the least of the accomplishments of TOGA has been the cooperation operation of two diverse research communities, the meteorological and the oceanographic communities, in pursuing common TOGA goals.