3.
COMPONENTS OF THE U.S. TOGA PROGRAM

The U.S. TOGA Program examined seasonal-to-interannual climate variability using observational, theoretical, and numerical techniques. In this chapter, we report on the major observing, process-study, and modeling and prediction efforts of TOGA. TOGA organized a large observational program, many of the instruments of which remain in place as the TOGA Observing System. Process studies, especially TOGA COARE, were performed to fill gaps in our knowledge of crucial dynamics on smaller spatial and temporal scales. Models, especially coupled models, and prediction systems were developed to address many aspects of ENSO. The interrelations are summarized by Figure 3, which indicates that the components all interact to motivate and justify each other. These components are now beginning to interact with applications of climate information and studies of the impacts of climate variations.

Over the ten-year lifetime of TOGA, U.S. federal agencies spent over $230 million to fund U.S. efforts directly contributing to the international TOGA Program. This funding represented roughly half of the direct contributions from the eighteen nations that were members of the Intergovernmental TOGA Board

Figure 3. The major conceptual components of TOGA.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program 3. COMPONENTS OF THE U.S. TOGA PROGRAM The U.S. TOGA Program examined seasonal-to-interannual climate variability using observational, theoretical, and numerical techniques. In this chapter, we report on the major observing, process-study, and modeling and prediction efforts of TOGA. TOGA organized a large observational program, many of the instruments of which remain in place as the TOGA Observing System. Process studies, especially TOGA COARE, were performed to fill gaps in our knowledge of crucial dynamics on smaller spatial and temporal scales. Models, especially coupled models, and prediction systems were developed to address many aspects of ENSO. The interrelations are summarized by Figure 3, which indicates that the components all interact to motivate and justify each other. These components are now beginning to interact with applications of climate information and studies of the impacts of climate variations. Over the ten-year lifetime of TOGA, U.S. federal agencies spent over $230 million to fund U.S. efforts directly contributing to the international TOGA Program. This funding represented roughly half of the direct contributions from the eighteen nations that were members of the Intergovernmental TOGA Board Figure 3. The major conceptual components of TOGA.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program (see p. 121). A partial enumeration of international funding for TOGA can be found in IGFA 1993. Figure 4 provides a bar chart of U.S. spending by year, which ranged from $15 million to $39 million. For comparison, the 1994 budget for the USGCRP was $1443 million (Subcommittee on Global Change Research 1995), of which $37.5 million was directly attributed to TOGA. The near-doubling of the budget from 1989 to 1992 reflects the buildup for the TOGA COARE field program (see p. 51) of 1992–93. These totals do not include the costs of aircraft time and ship time, which were budgeted separately. The TOGA Program benefited from activities related to predicting seasonal-to-interannual climate variations that were not considered part of Figure 4. Funding for U.S. TOGA. The bar chart shows total funding, by year, for U.S. efforts contributing to the TOGA Program. The pie chart shows U.S. funding, by program, for the last year of TOGA. See the text for descriptions of activities that were included or excluded from the totals, and for the program abbreviations. (Data courtesy of M. Patterson, NOAA.)

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program TOGA. For example, no costs of satellite-based instruments are included in the totals given above and in the figure. However, TOGA made use of operational meteorological satellites operated by NOAA (National Oceanic and Atmospheric Administration) and research satellites operated by NASA (National Aeronautics and Space Administration) for multiple scientific purposes. The totals for TOGA do not include most of the international meteorological observing system on which TOGA relied, but do include the specialized observing system that was deployed as part of TOGA in the tropical Pacific. The TOGA Program also benefited from work by the National Weather Service to provide operational long-term outlooks and to develop numerical models for short-term climate variations, but those costs are not included. Several U.S. federal agencies participated in TOGA. A breakdown of their contributions for the final year of the program is shown in the Figure 4 pie chart. A similar chart for 1989 can be found in NRC 1990, which also includes a discussion of the funding mechanisms for TOGA. The statistics on which these figures were based came from an annual survey conducted by the U.S. TOGA Project Office. Funding levels include only resources requested and directed specifically for TOGA research. NOAA funding came from three offices: the NOAA Climate and Global Change Program (labeled C&GC, which includes its predecessor, the NOAA TOGA Program), the NOAA Equatorial Pacific Ocean Climate Studies (EPOCS) Program, and the National Ocean Service (NOS). National Science Foundation (NSF) funding came from the Ocean Sciences Division (OCE) and the Atmospheric Sciences Division (ATM). NASA funding came from the Mission to Planet Earth (MTPE) Program. The Office of Naval Research participated only during 1992–1993, for TOGA COARE, and is therefore not shown in the pie chart for the last year of the program. Participation by various agencies in specific field programs is noted below in this chapter. OBSERVATIONS OF ENSO The TOGA Observing System was initially designed to observe the evolving warm and cold phases of ENSO, and then provide the resulting data to scientists immediately. As predictions of the phases of ENSO began to show skill, priorities shifted to measuring those quantities most useful for initializing and evaluating coupled models. We now have the ability to measure the quantities of highest priority—sea surface temperature, surface winds, and subsurface thermal structure—plus some other quantities of interest.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program If we were to stand on the equator with our backs to the coast of Ecuador and begin moving westward along the equator at walking speed—coincidentally, about the speed of the fastest ocean Rossby mode—it would take about nine months to reach the east coast of Halmahera Island in the far western Pacific. Over this vast expanse of ocean, only a few Volunteer Observing Ship (VOS) lines were in place throughout the 1960s and 1970s (see Figure 1 of Rasmusson and Carpenter 1982). VOS lines are series of observations of temperature, humidity, winds, cloudiness, and upper-ocean conditions made on a voluntary basis by the crews of non-research ships while the ships are traversing their normal routes. These limited observing resources left vast regions of the equatorial Pacific, almost as wide as the entire equatorial Atlantic, unmeasured. The problem for observing ENSO was to measure enough of the tropical Pacific, both at the surface and at depth, and to do it frequently enough to resolve the evolving atmospheric winds and ocean thermal structure. At the beginning of TOGA (see Figure 5, top), there was no obvious solution to this problem. It was envisioned in the original TOGA Plan (sections 2.1 and 2.2 of NRC 1983) that the VOS observations would be augmented by a small number of drifting thermistor chains and fixed moorings, by additions to the upper-air network of the World Weather Watch, and by satellite measurements. As explained below, the original concept of the TOGA Observing System proved inadequate, and more creative and extensive means had to be developed. During the TOGA years, additional systems were put into place to reach the goals of TOGA. By the middle of the program's life, the TOGA Observing System looked as shown in Figure 5, middle, and by the last half of 1994, the TOGA Observing System looked as shown in Figure 5, bottom. In addition to the need to observe regions previously under-observed, there was also a need to arrange for the prompt transmission of observational data to scientists (transmission “in real time”). This requirement necessitated changes to oceanographic instruments, such as moorings that recorded data on tape but did not transmit them, as well as changes in the practices of oceanographers. The scientists involved in TOGA made enormous progress in the development of observing systems that provided data in real time. TOGA Observing System The TOGA Observing System is a primary legacy of the TOGA Program. The following discussion touches on the broad range of observations used to address the scientific objectives of TOGA. Some of these observations were not established by TOGA, but were used, championed, or sustained by scientists participating in TOGA. As noted above, the TOGA Observing System was initially designed to observe ENSO. As such, it concentrated on the tropical oceans,

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program Figure 5 (at left). The TOGA In Situ Pacific Basin Ocean Observing System. The top panel is as of the start of TOGA in January 1985, the middle panel as of 1990, and the bottom panel as of the end of TOGA in December 1994. Moored buoys with thermistor chains and wind-measuring equipment are shown with diamonds, and those with current meters added are shown with squares. Island and coastal tide gauges that reported to the TOGA Sea Level Center are shown with circles. Drifting buoys are indicated schematically with short arrows; each arrow represents ten actual drifters. Tracks of volunteer observing ships releasing expendable bathythermographs are indicated by long curved lines, thick lines representing 11 or more transects per year and thin lines representing 6–10 transects per year. By the end of TOGA, most measurements were being reported in real time via satellite data relays. (Courtesy of M. McPhaden, NOAA/PMEL.)

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program especially the Pacific. A review of the TOGA Observing System was also presented in NRC 1994a. Building on the sparse operational network in place at the beginning of the program, TOGA built a diverse system incorporating observations that were: meteorological and oceanographic, land-based and ocean-based, remotely sensed (mostly from space) and in situ, funded by operational agencies and by research programs, driven by predictive model requirements and by curiosity-driven desires for elucidation of physical mechanisms in the climate system, obtained by standard means and by novel technologies, and solutions to TOGA data needs and to the data needs of other scientific and operational constituencies. The scope of the system was broader than the idealized (and not as yet established) observing system envisioned by TOGA Objective #3: “… an observing and data-transmission system for operational prediction.…” Such a system, tuned to provide just the data required for “operational prediction”—not more, not less—remains an ideal. A method for optimizing the observing system, using clear results from accepted prediction models, is still in development. It is likely that this ideal system will include a number of substantial enhancements and augmentations of the present-day system, and perhaps some selective winnowing of observations that do not carry much information into the prediction process in proportion to their cost. A limit on any such winnowing is that while some particular measurement series may prove to be of marginal utility for TOGA purposes, the data time series may be of great value for some other set of scientific studies or operations, and those other constituencies should be included in discussions about the future modification of the observing system.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program Table 1. TOGA data requirements   Quantity Spatial Resolution Temporal Resolution Accuracy 1 Upper-air winds 500 km (900 mb, 200 mb) 1 day 3 m/sec 2 Tropical wind profiles 2500 km (every 100 mb) 1 day 3 m/sec 3 Surface pressure 1200 km 1 day 1 mb 4 Total-column precipitable water 500 km 1 day 0.5 g/cm2 5 Area-averaged total precipitation 2° latitude by 10° longitude 5 days 1 cm 6 Global sea surface temperature 2° latitude by 2° longitude 30 days 0.5 K 7 Tropical sea surface temperature 1° latitude by 1° longitude 15 days 0.3–0.5 K 8 Tropical surface wind (1) 2° latitude by 10° longitude 30 days 0.5 m/sec 9 Tropical surface-wind stress (1) 2° latitude by 10° longitude 30 days 0.01 Pa 10 Surface net radiation 2° latitude by 10° longitude 30 days 10 W/m2 11 Surface humidity 2° latitude by 10° longitude 30 days 0.5 g/kg 12 Surface air temperature 2° latitude by 10° longitude 30 days 0.5 K 13 Tropical sea level (2) 1 day 2 cm 14 Tropical ocean subsurface temperature and salinity (3) (3) (3) 15 Tropical ocean surface salinity 2° latitude by 10° longitude 30 days 0.03 PSU 16 Tropical ocean surface circulation 2° latitude by 10° longitude 30 days 0.1 m/sec 17 Sub-surface equatorial currents 30° longitude (at 5 levels) as available 0.1 m/sec (1) While the accuracy values are given for 30-day averages, daily values are required for the resolution of 30–60 day oscillations. (2) Resolution as permitted by available sites and satellite altimetry. (3) See discussion in chapter 3 of NRC 1994a.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program A good overview of TOGA requirements for observations and the methods that were used to meet some of those requirements can be found in sections 2–7 of the International TOGA Implementation Plan (fourth edition, International TOGA Project Office 1992). Table 1, extracted from that plan, shows the atmospheric and oceanographic quantities desired, with associated sampling intervals and accuracies in both space and time. The Implementation Plan also provides brief rationales for the numbers presented. Mostly, the accuracies and sampling requirements are based on observed amplitudes of ENSO signals in the ocean and atmosphere. Prediction models and observing-system simulation experiments (OSSEs) are not yet able to reliably guide the design of complete observing systems. Sections 3–7 of the Implementation Plan summarize well the current status and near-term outlook for upper-air instruments, moorings, drifters, VOS, and island tide gauges, respectively. These were the principal tools of the TOGA Observing System for in situ measurements. Their condition at the end of 1994 is the TOGA legacy for observations insofar as TOGA played a role in creating or sustaining them. A particular measurement platform usually contributed to measurements of more than one quantity of interest. For example, the TOGA TAO moored array provides data on surface winds, surface-air temperature, sea surface temperature, surface atmospheric pressure, and upper-ocean thermal structure, with additional data on upper ocean currents at some of the array elements. Space-based observations were generally not designed specifically by or for the TOGA community, yet they have proven crucial for measuring certain basic quantities of the ocean and atmosphere, and will therefore be discussed as part of the system. To explain how and why the TOGA Observing System developed as it did, we provide a history of the observing system with an emphasis on the overall motivation and on the scientific and technical problems that arose at crucial stages along the way. It begins with a discussion of how traditional types of oceanographic measurement platforms were developed to address the problems of TOGA. By the time TOGA ended, a variety of options were available to measure what were, by then, agreed-upon quantities of highest priority. A final section indicates how, and how well, these high-priority quantities were measured. Volunteer Observing Ships Merchant ships, recruited under the auspices of the World Weather Watch to take standard surface observations for meteorological purposes, have been the major source of surface data over the open oceans for many years, and undoubtedly will continue to be for some time. TOGA heightened our awareness of both the utility and drawbacks of volunteer observing ships (VOS) as a data

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program source. It also advocated the continuation of VOS measurements and pointed out ways in which they could be improved and enhanced. Participants in TOGA championed, with some success, the identification of additional VOS candidates in data-poor tropical regions, improvement in the promptness and reliability of data communication using conventional high-frequency radio or newer links (e.g., communication satellites), the aggressive recovery of additional data for archiving from platforms such as fishing vessels that do not provide immediate data transmission, the upgrading of the accuracy of shipboard instruments and methods, and the development of hull-contact sensors for ocean measurements (thereby easing installation and calibration problems). Modernizing and bringing such a far-flung system to maximum effectiveness is not a one-shot process; continual pressure has been and will be required. A particularly effective “add-on” to basic VOS surface observations has been the inclusion of regular expendable bathythermograph (XBT) observations of upper-ocean thermal structure to the routine of selected VOS that traverse scientifically critical routes. Much of this work has been planned and carried out in concert with the World Ocean Circulation Experiment (WOCE) because a given ship often crosses regions of significance to both programs. A joint TOGA and WOCE set of route charts and sampling-interval suggestions was produced. Not all the recommended routes were sampled, and even fewer were sampled as frequently as TOGA objectives required, but there has been progress in recent years. While the accumulated VOS reports are of great value in our knowledge of the background state and climatology of the ocean, the tracks are too wide apart in the tropics and visited too infrequently to permit the study of the evolving warm and cold states characteristic of ENSO (see the figure on p. 30 of NRC 1994a). Surface Drifters Lagrangian drifters closely follow the ocean currents in which they sit. They are an effective means of obtaining broad, basin-wide coverage of sea surface temperature, sea-level pressure, and near-surface currents. In the late 1970s, Doppler-ranging from the French navigational system ARGOS became operational on NOAA TIROS (Television Infrared Observation Satellite). It was a cost-effective technique for listening to and locating radio transmitters on the ocean surface. The development of TIROS spawned the design and construction of a large number of ocean-surface drifters for measuring ocean circulation and as platforms for a variety of meteorological sensors. The drifters designed in the 1970s typically used large surface floats and “window-shade” drogues, a combination that degraded their ability to follow surface currents accurately and limited their useful lifetimes. During the planning of the World Climate Research Programme (WCRP 1983) several re

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program searchers realized that a redesigned drifter, one that was lightweight, long-lived, and readily deployable from ships, could make widespread measurements of sea surface temperature, surface pressure, and surface velocity to accuracies of ±0.3 K, ±1 mb, and ±1 cm/sec, respectively. By 1985, Draper Laboratory of the Massachusetts Institute of Technology, NOAA's Atlantic Oceanographic and Meteorological Laboratory (AOML) and the Scripps Institution of Oceanography (SIO) had produced competing drifter designs. During the period 1985–1989, field measurements of the water-following capability of the drifters were made using vector-measuring current meters (VMCMs) attached to the top and bottom of the drogue attached to the drifter (Niiler et al. 1987, Niiler et al. 1994). Several modeling studies of drifter behavior in steady upper-layer shear and linear wave fields were also carried out (Chereskin et al. 1989). These studies provided a basis for interpreting possible differences between the movement of the drifter and the movement of the water. In 1978, under the auspices of the NOAA/EPOCS program, the deployment of small groups of drifters began in the tropical Pacific (Hansen and Paul 1984). Under the auspices of TOGA in 1988, a basin-wide process experiment, the Pan-Pacific Surface Current Study, began. Its technical objectives were, over a three-year period, to learn to use VOS to maintain a group of 160 drifters and to select the most robust elements from the competing drifter designs. Its scientific objectives were to obtain a basin-wide field of surface currents and sea surface temperature of the tropical Pacific for the purpose of studying processes that determine the evolution of sea surface temperature. By 1991, these objectives had been realized and a WOCE/TOGA Lagrangian Drifter design was published (Sybrandy and Niiler 1991). These new drifters had a half-life of over 400 days at sea. They were deployed routinely, with over 95 percent survival, from VOS by a single able-bodied seaman. Costs for tracking and communication were reduced to one-third of the costs incurred using the older design, with no reduction of data quality. New sensors were also developed. Barometers were added to the drifters in 1991. In 1992, new salinity sensors were deployed during TOGA/COARE. At the end of 1994, there were nearly 500 drifters, deployed from VOS or aircraft, in the global ocean. The greater portion of these drifters reported data through the Global Telecommunication System (GTS) to operational meteorological and oceanographic centers. The data from these drifters were processed at the Drifter Data Center (housed at NOAA/AOML), and are available over the Internet using anonymous ftp (file-transfer protocol) with about a one-month delay. Fourteen countries contribute resources to this drifter program, and seven U.S.-government agencies use the WOCE/TOGA Lagrangian Drifter as an operational instrument.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program The Global Drifter Program in support of climate research and monitoring is now in place. It is an element of both the Global Ocean Observing System (GOOS) and the Global Climate Observing System (GCOS). TOGA played the principal role in the inception and implementation of this drifter program. Of primary interest to TOGA were sea surface temperature data from drifters. These data are critical because they provide well-distributed benchmarks for adjusting and correcting the inaccuracies and biases in maps of sea surface temperature derived from other sensors (such as those on satellites or VOS). Of great interest in oceanography, but of somewhat less importance to ENSO predictions, is the ability of drifters to determine accurately, with minimal contamination from wind and other noise, large-scale patterns of the surface currents. Drifters are being constructed in and deployed by several nations, using both research and operational funding. A census of 250 units in the tropical Pacific was stated as a goal for TOGA, and at the end of 1994 was approximately met. An additional set of 50 meteorological drifters (compared with a stated requirement for TOGA of 100) in the southern oceans was sustained primarily to obtain surface meteorological data, especially sea level pressure, in this data-sparse region. These drifters, the so-called FGGE (pronounced “figgy”, after the First GARP Global Experiment) buoys, have a different hull form, so their displacements are less useful than the new Lagrangian drifters for mapping surface currents. Tide-Gauge Network. Tide gauges measuring sea-level changes have been maintained for many years in harbors and lagoons of Pacific islands. Some of these gauges date back to shortly after World War II. In 1972, NOAA shut down the upper-air and sea-level stations at Canton Island and Christmas Island, right at the beginning of the NORPAX program. This action was taken as the 1972–73 El Niño was beginning, creating an unfortunate break in critical time series that were more than 20 years long. Jacob Bjerknes, who had used the long time series from Canton Island in his research on ENSO, wrote a protest letter to NOAA in hopes that this key station would be maintained. His letter fell on deaf ears. Such was the lack of appreciation for climate monitoring at that time. Klaus Wyrtki of the University of Hawaii, whose research using sea level made the case for the importance of ocean dynamics in climate (Wyrtki 1973), immediately went to work to rebuild the sea-level network of Pacific islands to ensure long time series for future climate research. He proposed a sea-level monitoring network for climate-research purposes. NSF funded the proposed network under the NORPAX program, and Wyrtki began installing tide gauges on various islands. In addition, Wyrtki expanded this network by forging relationships with numerous other Pacific nations that maintained tide gauges

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program circulation models. The complexity and difficulty of the task proved greater than foreseen (Neelin et al. 1992). It was well understood that the ocean component had to produce the correct sea surface temperature in response to the fluxes of heat and momentum from the atmosphere. It was also understood that the modeled atmosphere had to produce the correct surface fluxes in response to the sea surface temperature. What was not appreciated is how sensitive each model is to errors in the other. The original attempts to couple the atmosphere to the ocean ran into problems; the final state turned out to be far from a realistic one (see, e.g., Manabe and Stouffer 1988). In order to restore a more realistic climate state, the artificial expedient of “flux corrections” was adopted, viz., the arbitrary imposition of surface fluxes designed to force the modeled state to a realistic one (Saucen et al. 1988). These flux corrections have proven detrimental to simulations of climate change because they always tend to bring the climate state back to the original one, regardless of the state to which the changes in the forcing should impel the model. Improved coupled models have produced realistic interannual variability, representative of ENSO, without the need for flux corrections (Philander et al. 1992, Nagai et al. 1992), but only in the absence of the annual forcing of insolation. The annual cycle is subtle and difficult to simulate. It involves processes in the atmosphere and ocean different from those governing interannual variability (Köberle and Philander 1994). In particular, heat fluxes play a more important role in the annual cycle than the interannual cycle. For example, stratus clouds seem to be crucial for modulating the heat fluxes needed to produce the annual cycle, especially in the coastal regions of the equatorial and South Pacific (Klein and Hartmann 1993). In the cold-tongue regions, stratus clouds influence the far-eastern Pacific (say within 1000 km of the coast), where the variations of the heat fluxes through the ocean surface are in phase with the variations of sea surface temperature. However, stratus clouds appear to be unimportant for modulating interannual variations of the cold tongue in regions away from the coast, where the variations of heat fluxes and sea surface temperature are out of phase. The unrealistic warmth of all existing coupled general-circulation models at the far eastern part of the cold tongue has been attributed to the inability of the models to successfully simulate the stratus clouds and their effects (Mechoso et al. 1995). At this writing, coupled models are just beginning to produce combinedannual and interannual sea-surface-temperature variability without theartificial expedient of flux corrections (Latif et al. 1993, Schneider and Kinter 1994, Robertson et al. 1995a), although the modeled variability is not yet totally realistic (Mechoso etal. 1995). The challenge remains for coupled models to simulatesuccessfully both annual and interannual variability in the tropics.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program PREDICTION All objectives of the TOGA program were related to improving prediction, or quantifying predictability, of short-term climate variations. Progress towards achieving these objectives ranks among the greatest successes of TOGA. Not only was the basis of climate predictability established through ENSO theories and models, but by the end of TOGA, regular and systematic experimental ENSO predictions had been underway for several years. It is generally thought that the intrinsic variability of the atmosphere creates a barrier to predictions beyond approximately two weeks (Lorenz 1965, 1982; Charney et al. 1966). However, the slowly evolving nature of the ocean, together with the strong coupling between the ocean and atmosphere, allows that barrier to be avoided. Charney and Shukla (1981) pointed out that the muted high-frequency atmospheric variability of the tropics made the atmosphere there highly dependent on its boundary conditions, and therefore much more predictable on time scales associated with the slow variation of those boundary conditions. Prediction of sea surface temperature, a boundary condition for the atmosphere, allows the forecast of concomitant atmospheric statistics. Predictions of eastern tropical Pacific sea surface temperature can be made a season to a year or more in advance (e.g., Latif et al. 1994). While weather cannot be predicted further ahead than a week or two, seasonal climatic variables for particular regions of the world directly affected by ENSO can be predicted. The TOGA period has seen this concept developed and demonstrated in practice. The TOGA Program fostered development of both statistical and dynamical prediction models, established a coordinated prediction program that is the prototype of a multinational climate-prediction enterprise, and created operational climate predictions. Prediction products are routinely disseminated and used worldwide. Many diverse definitions of forecast skill have been used to evaluate weather predictions. At this early stage in the development of climate forecasts, only simple measures of skill are in regular use. The most common measures are (1) the correlation of a forecast with the NINO3 index and (2) the average of the root-mean-square error over the length of a forecast. These two measures give a simple indication of how well phase and amplitude have been predicted over the length of a record. More sophisticated measures compare the success of a forecast to the success of persistence and/or climatology. It should be noted that there is predictability in the sea-surface-temperature fields simply because of their persistence. The autocorrelation of the index

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program called NINO3 (the anomaly of sea surface temperature averaged over the region between 5°N and 5°S, 150°W and 90°W), for example, shows a 0.5 autocorrelation with a lag of 6 months. Clearly, the correlation of a prediction of NINO3 with observations must beat this value in order to demonstrate useful skill. With a lag of 3 months, the autocorrelation of NINO3 is about 0.7. Skill in predicting sea surface temperature, as measured by the correlation of observed and predicted time series, is expected to have its greatest value for lead times exceeding 6 months, when the value of persistence forecasts is low. Inoue and O'Brien (1984) provided the first ENSO predictions using a numerical model. In their prediction scheme, a dynamic ocean model forecast the evolution of the thermocline depth in the eastern Pacific, assuming that surface wind-stress anomalies remained fixed at their initial values. The model clearly demonstrated a predictive potential at lead times of several months, even though sea-surface-temperature and atmospheric fields were not explicitly forecast. Prediction schemes for ENSO based on statistical models were introduced by Graham et al. (1987a, b). In these schemes, sea-level pressure or tropical Pacific winds are used to predict sea surface temperature. A significant issue in the construction of such models is the appearance of artificial skill, i.e., the tendency of a statistical scheme with many degrees of freedom to produce results which are arbitrarily good in hindcast mode (reproducing the data used to construct the model), but poor in true forecast mode. Graham et al. minimized artificial skill by reducing the degrees of freedom in the model through the use of empirical orthogonal functions (EOFs). Others have used principal-oscillation patterns (Xu and von Storch 1990; Penland and Magorian 1993). The results of these schemes are clearly superior to persistence forecasts, providing correlations of about 0.5 with lead times of 9 months for predictions of sea surface temperature in the eastern Pacific (Latif et al. 1994, Barnston and Ropelewski 1992). Prediction correlations at this level using statistical models demonstrate clearly the large-scale, low-frequency nature of ENSO, as well as a significant degree of linearity in its evolution over time scales of several seasons. By the close of the TOGA Program, monthly statistical forecasts were being made by the Climate Analysis Center for 3-month-mean sea surface temperature in several regions of the tropical Pacific and Indian Oceans (Graham et al. 1987a, b; Barnston and Ropelewski 1992). Forecasts were made with lead times up to 12 months. Results indicated that the 170°W-120°W region of the equatorial Pacific is the most predictable. For forecasts with a lead of 6 months, the anomaly correlation score for the last decade over this region is 0.7. Skill is greatest for boreal winter forecasts made at the end of the summer. Forecasts made during winter for the following summer exhibit considerably less skill.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program Statistical forecasts outperform persistence in the equatorial Pacific under most conditions. Similar statistical techniques have also been implemented to forecast North American air temperature and precipitation (Barnston et al. 1994), but with forecast correlations for seven-month lead times not exceeding 0.3. The first coupled atmosphere-ocean model applied to ENSO prediction was that of Zebiak and Cane (1987). Only departures from monthly-mean atmospheric and oceanic climatologies were explicitly calculated. As the climatology was specified from observations, the mean state was guaranteed to be realistic. Atmospheric dynamics were approximated by a steady, linear, “shallow water”* system, following the work of Gill (1980) and later modified by Zebiak (1982, 1986) for use in modeling ENSO. Oceanic dynamics were also approximated by linear, shallow-water equations, following the work by Busalacchi and O'Brien (1980, 1981). The Zebiak and Cane model simulated a realistic ENSO. Analyses by Zebiak and Cane, Battisti (1988), Suarez and Schopf (1988), and Schopf and Suarez (1988) highlighted the crucial role of ocean dynamics in supplying a delayed negative feedback that sustained the oscillations in the models. The “delayed action oscillator” theory of ENSO (Battisti and Hirst 1989; Wakata and Sarachik 1991a; Jin and Neelin 1993a, b; Neelin and Jin 1993) is now the most widely accepted explanation of ENSO (see Battisti and Sarachik 1995 for more details). Regular predictions of ENSO on an experimental basis, using the Zebiak and Cane model, began in 1985 (Cane et al. 1986). Figure 10 shows the evolution of the correlation skill for a set of hindcasts of the NINO3 index for El Niño events of 1976 and 1982–1983, and the nonevent years of 1977–1979. The forecasts of these different periods were all largely successful at lead times of a year or more. The overall correlation skill (i.e., the correlation between the predicted and observed index) for NINO3 forecasts during this period was above 0.6 for lead times up to 11 months. Such success is particularly remarkable in view of the highly simplified initialization procedure that was used. Initial conditions of the ocean for each forecast were obtained by forcing the ocean model with analyzed surface winds based on data from VOS (Goldenberg and O'Brien 1981), often marginal in quality because of poor coverage. Initial conditions of the atmosphere were derived from the ocean-model simulation of sea surface temperature. Thus, no observations of sea surface temperature or subsurface temperatures were utilized. That such a simple scheme could allow skillful predictions strongly reinforces several prevailing ideas about ENSO: it is a large-scale, low-frequency signal; the important “memory” of the coupled system resides in the pattern of upper-ocean heat content; and, finally, the heat-content variability is largely controlled through surface forcing by wind stress, *   The so-called “shallow water” equations are approximations used for many fluid systems, including the atmosphere, that have much greater horizontal expanse than depth.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program Figure 10. Correlations for predictions of anomalies of equatorial sea surface temperature using fully coupled atmosphere-ocean models. The curve labeled “Lamont” shows results obtained with the coupled model of Zebiak and Cane (1987). The curve labeled MPI shows results obtained with the coupled general-circulation model of Latif et al. (1993). The curve labeled “Lamont (20 cases)” shows the anomaly correlations for the Lamont model when applied to the same 20 cases to which the MPI model was applied. Also shown, for reference, is the result of persistence forecasts. (Reprinted with permission from Latif et al. 1994, copyright Springer-Verlag.) in accordance with linear, shallow-water dynamics. The results from coupled models justify an optimistic assessment of the additional predictive potential of more detailed models and assimilation systems that could take advantage of all available observations. We can clearly see from the figure that while the overall skill of the Zebiak and Cane model is useful, there are specific times when the forecast skill is considerably higher than other times. Conversely, predictions made during boreal spring are considerably worse than the mean. The skill of the Zebiak and Cane model has been recently increased by more careful initialization, which not only reduced the nowcast error, but also increased the correlation skill out to more than a year (Chen et al. 1995). The first ENSO forecast using coupled numerical models, made in early 1986 (Cane et al. 1986), predicted a warming event for later that year. The forecast for January 1987, made with a lead time of one year, was essentially

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program accurate, although the rapid warming in 1986 occurred about 3 months later than had been predicted. Barnett et al. (1988) compared the predictions for 1986–1988 made using the Zebiak and Cane model with predictions from the statistical forecast model developed by Graham et al. (1987a, b), and found similar performance at lead times of 9 months or less. Evaluated over a longer record, the performances of the dynamical and statistical models are comparable for lead times of 4 months or less, but dynamical predictions are increasingly superior at longer lead times (Latif et al. 1994). In 1991, after informal discussions with members of the TOGA Panel, the TOGA Project Office began a program to further develop the new field of short-term climate prediction. The program, the TOGA Program on Prediction (T-POP, described in Cane and Sarachik 1991) was designed to be inclusive, supporting all the ongoing research in the United States on short-term climate prediction. Its focus was the development of prediction systems—models, data-assimilation techniques, and data quality control—based primarily on coupled atmosphere-ocean general-circulation models. In addition to prediction research, T-POP concentrated on making experimental predictions and on inter-comparisons of predictions for designated time periods. Forecasts using coupled models require initialization of only the ocean components of the model because, in most instances, information about the atmospheric initial state is lost within the first two weeks of the model run. The simplified prediction systems (e.g., the one developed by Cane and Zebiak) use the history of wind stress to initialize the state of the ocean. During TOGA, a real-time ocean-analysis system (Ji et al. 1994b) was implemented at NMC. The NMC system provided forecasts of sea surface temperature for the tropical Pacific, as well as forecasts of rainfall and surface temperature around the globe. The skill level of this model for about the first two seasons was higher than that of other systems because an observed ocean thermal state is used for initialization rather than a proxy based on the history of the winds (similar results were obtained by Rosati et al. 1996). Forecast experiments where subsurface data were not used in the assimilation show a loss of skill, measured by correlation coefficients, from about 0.8 to 0.6 at two seasons. Because much of the global atmospheric variability depends on variations of tropical sea surface temperature, any model that forecasts sea surface temperature can be used with any atmospheric general-circulation model to forecast atmospheric variations around the globe. This process has several advantages in opening up the field of climate forecasting to groups that do not have the resources to run fully coupled models. Groups that possess only atmospheric models can make seasonal-to-interannual climate forecasts by using one or all of the several, now routinely available, forecasts of sea surface temperature.

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program The TOGA goal of establishing the existence of seasonal-to-interannual climate predictability has been realized. Actual predictions of those aspects of climate associated with ENSO have become available in increasing numbers from a suite of dynamical and statistical models. Stemming from these achievements has been a growing sense of the opportunity to capitalize more fully on the established predictive potential for societal and economic benefit. While the examples provided in chapter 6 demonstrate actual benefits of climate forecast applications, they point to the even greater potential that would accompany a more organized and focused climate prediction and application activity. It was with this motivation that the concept of an International Research Institute for Climate Prediction (IRICP, see chapter 7) was developed. As of the end of TOGA, routine predictions of sea surface temperature based on ENSO dynamics were being made using the Zebiak and Cane model, various statistical models, and the NMC coupled model. These predictions were published in the NOAA Climate Diagnostics Bulletin and the NOAA Experimental Long Lead Forecast Bulletin. Additional forecast products were available from several other groups, both national and international. Products consisted of predicted indices (e.g., NINO3), as well as predicted fields, at lead times ranging from 1 to 18 months. The routine dissemination of such products attested to a high level of maturity of ENSO prediction, and signaled the movement into an operational mode in which practical application of climate predictions can begin. However, experience has shown that there can be strong intradecadal variations in the skill of forecasts. During the 1980s, when ENSO was regular and of large amplitude, forecast skill was high. Events during the 1990s have proven more difficult to predict (see section on The Warming During 1990–1994, p. 85). Appendix B lists known prediction products. TOGA PRODUCTS As a result of the building of the TOGA Observing System and the development of a predictive capability for aspects of ENSO, a large number of products (both observational and predictive) have become available (see Appendix B for a list). These products encapsulate the variety of efforts and progress of TOGA. They are generally available, they are used and consulted widely, they have many sources (purely observational, model-assimilated, and predictive), and they are available in different media (paper and electronic) through a variety of distribution networks. These products were motivated by TOGA, a research program, but have transcended it. Normally, when a research program ends, what is left are the data and the resulting knowledge. TOGA has accomplished more than that. The key to understanding this is TOGA's role in the development of short-term

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program climate prediction. Products exist, and will continue to be issued on a regular basis, because predictions will continue to be made as long as they are found to be useful. TOGA has provided products because the community that produces and uses short-term climate prediction demands that the ocean measurements started by TOGA be continued and be made widely available. An analogy can be made between maintenance of the upper-air observing system in support of weather prediction and maintenance of the ocean observing system in support of climate prediction. The upper-air network has been maintained for about 50 years not so much because it is producing good science (which it is), but rather because it has proven useful to society. We have gained a global view of the atmosphere and its variations because weather prediction is recognized to be valuable. Similarly, the ocean-data products developed in conjunction with ENSO predictions are already used by society and have a recognized value beyond their scientific worth. We expect that our knowledge of the ocean will be expanded because of the maintenance of an ocean observing system in support of climate prediction. Reanalysis Reanalysis is a tool that allows consistent data sets to be produced for climate studies. It is needed because existing weather analyses are inadequate for use in climate studies. Climate data sets were needed during TOGA both for diagnostic studies and for evaluation of predictive systems using historical events. For studies of interannual climate variations, a consistent record spanning several decades would be very useful. The long record of weather forecasts and analyses produced by national weather centers would seem to be an ideal data set for climate studies. However, many changes—mostly undocumented—have been made in the various components of the weather forecasting system. Changes in the quality-control protocols, in the data-assimilation models and methods, in the initialization procedures, and in the weather prediction models all contribute to spatial and temporal inhomogeneities in the long-term record of weather analyses. The only way to “undo” these changes and to produce a spatially and temporally homogenous set of analyses suitable for climate studies is to repeat the entire procedure with the best available weather-data assimilation systems. NMC agreed to reanalyze the entire climate record beginning with 1960, work that it continued as NCEP. In response to a request from T-POP, the first five years to be reanalyzed will be those from 1985–1990, to facilitate the intercomparison of predictions that T-POP has undertaken for a common period. NASA Goddard has also performed a five-year reanalysis, for March 1985 to February 1990, as part of the Earth Observing System (EOS) project. NASA is also

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program preparing a specialized reanalysis of data from the western Pacific in support of COARE. NCEP and the ECMWF have also performed reanalyses of data from the COARE period. Reanalysis at a single point in time does not solve the problem posed by further additions to the climate record. To have a useful, growing climate record, reanalysis must be done regularly, say every ten years or so. Reanalysis not only allows an orderly growth of the climate record, but also allows data that did not arrive in time for the original analyses (and are therefore normally wasted) to be put to use. Regular reanalysis provides continuing incentives for uncovering and rehabilitating old data. TOGA CD-ROM Project. The ITPO worked past the end of TOGA to make the TOGA data available on CD-ROM. A wide range of data sets for the ten-year TOGA period, provided by the TOGA data centers and by other data centers in various nations, are freely available to the international research community on CD-ROM. As a first step, data sets for 1985 and 1986 were published (Halpern et al. 1990) on one CD-ROM, which has been widely distributed. Subsequently, CD-ROMs with data up to 1990 were prepared at the Jet Propulsion Laboratory (JPL) for distribution in 1994, with further CD-ROMs to follow. PROBLEMS AND SHORTCOMINGS Although the U.S. TOGA Program accomplished much, it did encounter problems. Discussion of some of these can also be found in NRC 1992. Whether some aspects of the program should have, or could have, been handled differently is a matter of judgment. On some of the issues mentioned below, no consensus has been reached on whether these were truly shortcomings of the program. Problems involved: the overall scope of the program, the development of observational systems, the timing and relevance of the field programs, and the cooperation among all the participating individuals and organizations. However, resolution of some of these problems within the scope and lifetime of the TOGA Program was one of the strengths of the program. The original plan from the United States (NRC 1983) for the program that became TOGA concentrated solely on ENSO in the tropical Pacific. The international TOGA plan (WCRP 1985), however, took a more expansive view of the problem and proposed a more general study of seasonal-to-interannual variability throughout tropical oceans and the global atmosphere, with the objectives previously listed (p. 19). These broader objectives were accepted by the U.S. community (NRC 1986). However, these objectives proved overly

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program ambitious. The financial resources from the United States, and the rest of the international research community, for such an ambitious program were not available. Nor was any great effort made to distribute the study and financial responsibilities among the participating nations to cover the broader objectives of the global program. Even if greater financial resources had been available, it is not clear that a sufficient number of trained scientists interested in these problems would have been immediately available. In practice, TOGA resources were focused on studying interannual variability in the tropical Pacific region, falling back to a strategy more closely resembling the 1983 plan. This limited focus may have contributed to the overall success of TOGA. However, even the great attention paid to ENSO left unanswered many questions about the annual cycle in the most studied region of the tropical Pacific. The original plans for TOGA relied heavily on proposed and planned observations from satellites. Insufficient attention was paid to the political and operational realities of satellite programs, and the timing of when satellite-based observations of the physical quantities most needed for TOGA would really be available and reliable. When the satellite programs on which the original plans relied failed to materialize, the TOGA Program had to develop alternate strategies. While the resulting TOGA Observing System, and its array of moored buoys in the tropical Pacific, was one of the great achievements of the program, much of the system was not available until nearly the end of the program. The usefulness of the system for making forecasts of seasonal-to-interannual climate variations has still not been fully evaluated. The best mix of satellite-based and in situ observations for research or operational forecasting is still unknown. Problems also remain as to how to arrange for research observing systems to become part of the operational observing systems. Furthermore, continuity is not assured for observations of the many geophysical quantities needed for studying and predicting seasonal-to-interannual climate variations. The relevance of some of the field programs of TOGA to the larger program has been the subject of debate. The several field programs discussed earlier in this chapter, especially the large COARE field program, were some of the most expensive parts of TOGA. Though there is little dispute that these programs, especially COARE, were, and will continue to be, valuable for understanding air-sea interaction, some scientists would have preferred that the resources had been allocated to examining processes on the larger spatial scales more clearly relevant to directly attaining the objectives of TOGA (although new funds were found for the COARE field program). COARE was not completed until near the end of TOGA and the data from the field program are still undergoing intense analysis; it therefore did not have a large direct influence on TOGA. Though COARE has already produced much excellent science, the

OCR for page 22
Learning to Predict Climate Variations Associated with El Niño and the Southern Oscillation: Accomplishments and Legacies of the TOGA Program extent to which it will produce knowledge applicable to the basic objectives of TOGA is not yet known. The TOGA Program required cooperation among scientists from several disciplines, several government agencies, and scientists and organizations from several nations. The required cooperation did not always develop smoothly. Although by the end of TOGA, participating atmospheric and oceanic scientists had learned to work together, and the larger community was viewing the atmosphere and oceans in a more unified way, at the beginning of the program lines of communication were not as open (see p. 106). Data management, though kept relatively inexpensive, relied heavily on the work of the scientists collecting the data, so that not all data were immediately available to the larger community or the operational centers that could use them for real-time model initialization. However, data accessibility improved as the program evolved, to the extent that all data from the TAO array are now freely available in real time. Furthermore, for the results from TOGA to be valuable to a large user community, applications must be developed (see p. 124), but communication and understanding between physical and social scientists is still difficult, and no coherent strategy for identifying and effecting applications has yet arisen. Interagency cooperation presented difficulties because of the differing objectives and operating styles of the agencies involved. NOAA had a specific operational mission guiding its research. NASA had a research strategy more linked to development of space-based technology then to specific research problems. NSF preferred to respond to the best scientific proposals it received without promising funds for a program. These agencies struggled to coordinate their activities. The problems are structural and will face any future large programs for climate research. The facts that this report concentrates so heavily on the U.S. efforts, that not all financial data from participating U.S. federal agencies are uniform and available for all years of the program, that financial data from other countries are limited, and that some issues remained unresolved through the preparation of this report on the relationship of the advisory and review mechanism (e.g., the TOGA Panel) to the program and the several sponsoring agencies all point to the difficulties of coordinating the large international TOGA Program. However, such problems confront all large cooperative ventures.