END- TO- END SEASONAL TO INTERANNUAL PREDICTION
Working Group Participation
Edward S. Sarachik, Chairman
Otis B. Brown
Moustafa T. Chahine
William E. Easterling
Roger B. Lukas
Piers J. Sellers
Peter J. Webster
Designated Federal Liaison: J. Michael Hall
Rapporteur: Frank Eden
WORKING GROUP SUMMARY
Edward S. Sarachik,
Edward S. Sarachik, Chairman
In terms of climate prediction, the last ten years have witnessed a revolution in our ability to observe, understand, and predict a year in advance the fundamental dynamics of the El Niñ o/ Southern Oscillation system. Success to date suggests that further research and development could lead to climate predictions that can provide advanced information to reduce the impacts of such destructive natural climate fluctuations as droughts, which lead to forest fires and crop failures; floods, which lead to loss of life and stoppage of river commerce; and heat and cold waves, which lead to human misery and deprivation.
We make a prediction every time we expect this year's summer to be basically the same as last year's. It is this expectation of the regular return of the seasons that is confounded when unusual spells of weather cost us time and money because our expectations turn out to be false. The need to predict, when possible, the actual state of the climate, months to a year or so in advance, motivates programs on seasonal -to interannual prediction. What we now have the ability to accomplish motivates a great deal of scientific observation, research and modeling. The science is fundamental, yet the payoffs are short term and tangible.
Progress by a determined community of government and university meteorologists, oceanographers, and hydrologists with multiagency support (led by National Oceanic and Atmospheric Administration (NOAA) Office of Global Programs) has been rapid and remarkably successful: We already have begun to predict aspects of El Niñ o in the tropical Pacific, and these forecasts have benifited countries affected by El Niñ o (Peru, Brazil, Australia, Chile, and Columbia, the Philippines, and the U.S. Pacific Islands). Progress over the next few years will determine whether this predictive capability can be developed fully for use within the United States.
In the early days of climate research, science was the province of a few agencies, often with diverse objectives. The U.S. Global Change Research Program (USGCRP), aided by the Office of Management and Budget (OMB), allowed the agencies to focus their resources and to function in a coordinated way with advice provided by the National Research Council (NRC). As a result of these programs on seasonal to interannual variability, we have moved from a time in which the El Niñ o phenomenon could barely be observed, to a time in which data on the actual state of the surface and subsurface tropical Pacific to a depth of 500 meters, along with predictions based on these observations, are accessible to any researcher via desktop computers.
Science Questions The creation and evolution of USGCRP programs on seasonal to interannual variability are based on four fundamental scientific questions:
USGCRP RECORD IN UNDERSTANDING SEASONAL TO INTERANNUAL CLIMATE VARIATIONS Through programs developed under the USGCRP (with the cooperation of OMB and support from Congress)--primarily TOGA (Tropical Oceans Global Atmosphere), its successor program CLIVAR/GOALS (Global Ocean Atmosphere Land System), and GEWEX (Global Energetics and Water Experiment)--we have begun to understand seasonal to interannual climate variations in limited regions of the Earth, especially the phenomenon referred to as El Niñ o. We can now see, understand, and predict (to a degree usable for some regions of the world) the climate variations that characterize El Niñ o. We have also begun to appreciate the role of land processes and hydrologic systems in seasonal to interannual climate variability or predictability.
Some remarkable achievements over the last ten years have pioneered short-range climate prediction and indicated a path to the eventual prediction of seasonal to interannual climate variations over the U.S. These include the following:
These accomplishments have arisen from focused U.S. contributions to international programs, including TOGA, GOALS, and GEWEX. However, a great deal of activity in USGCRP agencies on seasonal to interannual climate has not been part of these focused efforts and therefore has not been nearly as effective in advancing the highest priorities.
OPPORTUNITIES FOR USGCRP IN SEASONAL TO INTERANNUAL CLIMATE VARIABILITY AND PREDICTABILITY Based on the results of the TOGA program, the research community believes that future opportunities for the USGCRP will best be achieved in the context of
Such a demonstration project is reflected in national and international global change program documents that describe the need for research programs, such as CLIVAR/ GOALS and GEWEX, and call for the establishment of an international research institute (IRI) for seasonal to interannual climate prediction. Planning documents for elements of the World Climate Research Program (WCRP) and the U.S. Seasonal to Interannual Climate Prediction Program (SCPP) point to the establishment of an IRI as an important mechanism to
The broad outlines of such a demonstration project can be diagrammed as shown in Figure 3.
Since all useful forecasts are local, a large-scale forecast is, by itself, not sufficient for practical application. Local data (models, statistical data, etc.) must be added to the large-scale forecast to produce a regional forecast. This regional forecast is then used for application to a sector. Different applications may require different types of local forecasts: for example, applications to fisheries may require, among other things, ocean temperature, whereas applications to agriculture and water resources will require, among other things, rainfall amounts.
In this context, an end- to- end prediction system can be defined as consisting of the following steps:
Implementation Implementation of the concept of end- to- end prediction requires a number of things that can be diagrammed as shown in Figure 4.
The strong interaction and balance among all the elements in the figure are crucial. End- to- end seasonal to interannual prediction requires the development of coupled atmosphere- ocean- land models. It requires that observations be available and a procedure developed for initializing the forecasts. It means that remote and in situ observations must be combined for this initialization and that an efficient data system must be established for this combination. It requires a procedure for validating predictions. It requires that poorly understood or modeled processes be investigated and sets priorities for these processes. Since climate information, to be useful, must be brought down to the local level, it requires adding local information and making region- specific forecasts. Then, the sector of applica-tion and its normal mode of operation in the absence of additional information must be identified and understood. Finally, the information must be combined with the forecast and presented to the user in a way that guarantees maximum utility.
The basic implication of this concept is that it guides, in a focused way, what needs to be done; provides a measure of the value of an activity in terms of its role in the end- to- end system; indicates gaps or imbalances in the activities (what is not being done); provides useful results on both a short-term and an ongoing basis; and has a built-in means of evaluation: the skill of prediction and the success of the applications. Conversely, this end- to- end activity is integral: no part of it can be compromised without affecting the ultimate skill of the prediction and the usefulness of the applications.
The working group participants identified some priorities within individual components of this integrated program on seasonal to interannual climate prediction.
Research is needed to enhance the understanding of a crucial, but poorly understood, aspect of climate models: (1) land- atmosphere interactions, with initial emphasis on land-atmosphere interactions over the Mississippi and the Amazon basins, and (2) the characteristics and predictability of precipitation in this region and other land regions that affect seasonal to interannual predictability (GEWEX).
A general observing system for end- to- end predictions must be some combinationof in situ and remote observations and must lead to model-assimilated data.
The reasons for this principle are numerous: Remote systems generally require surface information continuously. This information is used for continuous calibration and to ameliorate gaps that always arise from remote observations. Conversely, in situ observations can never be global; they require remote measurements to achieve global coverage. Both types of observations must contribute to the initialization and validation of predictions and, therefore, to a model- assimilated data product.
We can identify the priorities for seasonal to interannual prediction:
The quantities are not prioritized among atmosphere, land, and ocean, and only for the ocean are relative priorities identified (italicized quantities represent the highest priorities). Note that precipitation occurs in all three lists. Maintenance of the CLIVAR/GOALS observing system in the tropical Pacific and its appropriate expansion combining in situ and remote observations (including Mission to Planet Earth) over other oceans and over land are essential.
Process studies can be observational, theoretical, or computational and can range from pencil-and-paper calculations to large observational field programs. In order to apply to end- to- end prediction, they must focus on those inadequacies in the models, observations, or applications that affect the skill of prediction or the success of applications.
The skill in seasonal to interannual prediction within the U.S. is still insufficient to be used effectively but it is being developed in a planned, phased process. This process begins by further improving the skill of predicting of El Niñ o in the tropical Pacific; then expanding the regions of application around the tropics (including the monsoon regions of North America, especially Arizona, Texas, and New Mexico; South America; and Southeast Asia); next investigating predictability in midlatitude areas (including the U.S. West Coast and Southeast) that derive their predictability from the remote effects of El Niñ o; and finally, investigating whatever predictability may be further exploited from atmosphere- ocean- land interactions totally outside the tropics (CLIVAR/GOALS and GEWEX).
These process studies are best pursued via U.S. contributions to the high- priority international programs CLIVAR/GOALS and GEWEX, and via successful implementation of the U.S. SCPP, including establishment of an IRI.
EVALUATION OF USGCRP PROGRAM MANAGEMENT Accomplishments thus far have resulted in a new paradigm in which the concept of end- to- end prediction motivates and guides all program components and determines the priorities and balance among program elements.
The concept of end- to- end prediction can also be used to focus and evaluate relevant research by imposing a discipline on the process and defining the priorities for a carefully balanced program. This balance is crucial: since all elements depend on each other, no element can be compromised without damaging the entire enterprise. It presents a method of R&D in which success can be demonstrated by the development of forecast skill and by the money and lives saved by applications of predictive information. The program requires careful coordination, good advice and oversight, and a stable and balanced funding profile, with focused contributions by the agencies involved in seasonal to interannual prediction. This country has an enthusiastic and able body of scientists eager to tackle the scientific problems involved in developing end- to- end prediction on these time scales. The return for investment now will pay off in the short run and eventually lead to a permanent prediction capability that will benefit the entire country.
In this context, the working group identified some program management principles that must apply in supporting and managing a demonstration research program on end- to- end seasonal to interannual prediction.
Success requires a management structure in USGCRP (with OMB, the Office of Science and Technology Policy (OSTP), and the Congress) that will
The working group emphasized that these requirements are not currently being fully met.
OPPORTUNITIES FOR INTERACTION WITH OTHER ELEMENTS OF USGCRP Seasonal to interannual climate variability interacts strongly with other elements of the USGCRP. Only a few examples are given here.
Decadal to Centennial Variability and Change The attachment to this appendix provides some details on the connections between research on seasonal to interannual climate variability and investigations of decadal to centennial climate change. Examples include the following:
Atmospheric Chemistry Since cumulus convection in the tropical Pacific has the time dependence of El Niñ o, and since it both directly transports water vapor (and other trace gases) into the stratosphere and affects the height of the tropopause, there will be a modulation of stratospheric-tropospheric exchange.
MISSION TO PLANET EARTH/EARTH OBSERVING SYSTEM (MTPE/EOS) AND SEASONAL TO INTERANNUAL PREDICTION
CONCLUSION The U.S. public responds to what it reads and experiences and has come to expect predictions of heat waves, destructive hurricanes, excess rainfall leading to floods, and spells of drought. The skill for seasonal to interannual prediction within the United States at the moment is too low to be used effectively. However, it is being developed by a planned, carefully phased process that begins by concentrating on regions where predictability has been proven, particularly El Niñ o in the tropical Pacific. This process then concentrates on international programs such as CLIVAR/GOALS and GEWEX, and on implementation of the U.S. SCPP, including the IRI.
Intersection of Seasonal to Interannual and
Decadal to Centennial Climate Variability and Prediction
Roger B. Lukas
Roger B. Lukas
The past few years have seen ENSO variations in the tropical Pacific unlike anything in the past 100 years. The probability of observing this type of variability by chance is 1 in 2,000 if the recent climate record is stationary with respect to S- I variability. Thus, the inescapable conclusion is that S- I variability is nonstationary, and it remains to determine whether this is a characteristic of natural variability on longer time scales or whether it is related to enhanced greenhouse warming.
Recent analysis showed that the amplitude and phase of the annual cycle in the SOI have varied substantially during the 1900s. It is well established that the existence and character of model ENSOs depend on the annual cycle that is either produced by the model or specified a priori. One might view ENSO as a perturbation of an unstable annual cycle.
A recently discovered global mode of the ocean- atmosphere- land system involving winter warming over northern land masses and winter cooling over northern oceans showed that surface temperature anomalies varied out of phase on short time scales, but they have been locked into a warm phase over land masses for at least the past two decades.
Together, these results suggest that decadal time- scale processes are interacting with ENSO. Further, it appears that these modulations are impacting the recent prediction skill for ENSO. Thus, it is very important for the seasonal to interannual climate component of CLIVAR and USGCRP to work in collaboration with the decadal to centennial climate component to understand the mechanism(s) responsible for these modulations of ENSO.
Some hypotheses can be advanced to explain these and related observations. Two involve tropical- extratropical linkages within the ocean, operating on much longer time scales than such linkages in the atmosphere. One hypothesis involves long oceanic Rossby waves generated along the eastern boundary of the Pacific during ENSO, and their subsequent propagation westward across the basin and interaction with the atmosphere through sea-surface temperature (SST) variations. Another hypothesis involves the interplay of the shallow thermohaline overturning cell in the North Pacific coupling the tropical and subtropical wind- driven gyres, with anomalous heat and freshwater flux forcings in the subtropical gyres (forced in part by ENSO) manifest later as equatorial thermocline anomalies.
A combination of monitoring, modeling, and process research is appropriate to pursue one or more of these hypotheses. Such an integrated approach to understanding the decadal modulations of ENSO provides motivation for continuing observations in a research context. Existing elements of the GOALS (former TOGA) observing system and the ongoing World Ocean Circulation Experiment (WOCE) program already provide a large-scale monitoring context for the upper Pacific Ocean. A sequence of process studies is proposed to address the processes that are critical to these (and other possible) hypotheses in order to ensure that they are properly captured in coupled models that can be used to rigorously test the motivating hypotheses. Such an approach has been used quite successfully during TOGA.