A Vision for the Future
The panel believes that successful implementation of the recommendations given in the previous chapter would go a long way toward making both modeling for societal benefit and modeling purely for understanding much more effective. The panel also notes that the integration of operations and research for both modeling and observations is a prerequisite to successfully producing and delivering useful climate information. The panel is compelled to augment its view that addressing modeling alone is not enough, and that climate modeling can be made more effective by the creation of operational entities that maintain observing systems, produce and disseminate climate information, and are responsible for coordinating certain functions with the climate research community and the weather forecasting operational community.
High-end climate modeling depends on observations and on research. Likewise, climate research depends on high-end modeling and observations. Observational data assimilated into a comprehensive coupled climate model enables the verification and enhancement of model code to produce accurate and continual climate analysis. From panel deliberations and acknowledgements of the robust linkages between research, observations and climate modeling, the panel endorses previous NRC reports (NRC, 1998b, 1999a, 1999b, 1999c, 2000a, 2000b) that called for the development of a sustained climate observing system. In this section, the panel presents a vision of an operational entity that facilitates the synergy among sustained climate observations, high-end modeling, and research while identifying, creating, and producing climate information useful to society.
7.1 CLIMATE RESEARCH AND CLIMATE OPERATIONS
We have distinguished modeling for societal benefit and modeling for understanding. Modeling for societal benefit is product oriented, requires regular and systematic runs of climate models, and is user driven and user evaluated. Modeling for understanding is freer, competitive, driven and evaluated by scientific priorities, evaluated by peers, and generally less capable of being organized or constrained. We recognize that these distinctions are not universal and are therefore imperfect, but we have found it useful to make these distinctions in order to consider the needed organizational aspects for each type of modeling.
It is the nature of all climate research that a proper balance between process studies, background observations, and modeling best advances the understanding of the entire system. It is the difficulty of climate research that, while it is easy to put together field programs of limited duration to measure poorly understood processes, it is almost impossible to sustain measurements on climatic time scales. The climate research community neither has the infrastructure for doing so nor are sustained observations amenable to the usual peer review process, because sustaining observations is not itself research.
Climate operations has an analogous structure. It needs an observing system and it produces model products using high-end modeling, some for analyzing and improving the observing system and some for diagnostic and predictive information. A prime function of climate operations is the design and delivery of climate information products that benefit society and put demands on the observational system and on modeling. These societal functions are qualitatively different from research functions. They tend to have time constraints determined by the nature of the decision to be made, they require specific products to be delivered in forms most useful to decision makers, and they are judged by a different standard from curiosity-driven research. In practice the difference between this type of product-driven research and curiosity-driven research shows up as differences in resources and organizations required, which, in turn, implies different modes of management and funding. Product-driven research tends to be large scale (beyond the scale of a single principal investigator), more expensive, and more highly centralized. The additional resources required are justified in terms of benefits to users with the ultimate evaluation done not by modelers but by the users themselves. In particular, operations can (and must) sustain infrastructure and can (and must) sustain an observing system.
7.2 MUTUAL INTERACTIONS AND MUTUAL BENEFITS BETWEEN CLIMATE RESEARCH AND CLIMATE OPERATIONS
Operations provide enormous benefits to research and are most likely to be successful when interacting strongly with research. For example, the
current weather observations network is maintained by the Weather Service for construction of weather forecast products. The long records of upper air observations, taken globally since the 1950s, are an invaluable resource for atmospheric researchers. Indeed, it is hard to imagine the upper-air network being maintained in the research domain in a principal investigator mode of operation. It is also hard to conceive of weather operations existing without the development and improvement of numerical weather forecasting, which arose in the research domain. It is this synergy between the weather forecasting operations and atmospheric research that is most valuable both to researchers and to society.
But we note the asymmetry between the two types of activities. Centralized, expensive, and ongoing operations can contribute greatly to curiosity-driven research, but research that is decentralized and organized predominately in a principal investigator mode cannot produce the extensive regular and systematic products demanded by society. It can, however, design and help develop these products.
It is this asymmetry between research and operations that must be recognized for us to present our vision. We envision a modeling activity that responds and contributes to both research and societal requirements. Because of the noted asymmetry, each function can be fulfilled only with the involvement of the other. Because the analysis of the state of the climate system also involves model assimilated data and because model development requires a constant confrontation between observations and models, we can diagram the needed interactions as in Figure 7-1.
As in a similar model for research (Figure 2-2), the interactions and balance of components of Figure 7-1 are crucial. Research, in the usual distributed principal investigator mode, is needed to define and understand the climate problem and constantly improve the components of models through deepened understanding derived from process studies, small-scale models, and the diagnosis and analysis of observations. Sustained observations are the basis for our knowledge about the climate system. It involves long-term, accurate, and calibrated measurements of all components of the climate system. This element may use existing observations of the atmosphere, ocean, land, hydrosphere, etc. taken for different purposes but must assure they are maintained as climate observations (NRC, 1999b). The high-end modeling component synthesizes the research and assimilates the sustained observations to provide climate information products for use by researchers and society.
It is the concept behind Figure 7-1 that forms our larger vision of effective climate modeling in the United States. Modeling is thus put in its proper context: as synthesizer of research, as assimilator of observations, and as producer of products for society. It is in the implementation of this concept that the details of our vision must be fleshed out in the relationships implied by the arrows and in the organization and resources needed for the components. It is clear that by considering large-scale modeling and its proper context we have once again come upon the needed functional components of a Climate Service.
Within the conceptual picture of Figure 7-1, we begin with the high-end modeling component. The magnitude of the problem is large because it involves modeling the climate system and assimilating large amounts of data, including satellite data. As long as it involves any of the several activities detailed in Chapter 4, the climate models involved will be large, as close to comprehensive as possible, requiring simulations of thousands of model years to develop and many hundreds to apply.
An effective modeling activity for both operations and research should have models and model components developed and improved in the research domain, run in the operations domain, and analyzed in both domains. The process should be ongoing and cyclic and should involve focused parts of the research community in every phase of the cycle. It should have centralized parts in order to accomplish its operational mandate and may have a degree of decentralization for the research functions (consistent with need to run on high-end computers). This model should avoid the duplication and disorganization endemic (and probably necessary) to a successful small-scale research enterprise.
We see the optimal way of fulfilling this requirement as a small number of operational Centers (either new or existing, colocated or otherwise in touch), each devoted to a different societal need (i.e., producing a different product), each adequate in resources to its task, interacting with
each other and the research community through a mechanism of exchange involving a common modeling infrastructure.
In addition to the development and operation of a high-end climate model, each center should include a facility (people, storage devices, and machines) for the storage and distribution of model codes and output to the research community and for the collection of parameterizations and diagnostics from the research community, all in common formats. Common diagnostic tools would be developed and used by all large modeling centers and by the researchers interacting with them. Research funding and Center computer time should be made available to the research community for model improvement. The Centers themselves would have a certain amount of funding available for the research they deem necessary for advancing their tasks.
To develop and evaluate the needed models, the Centers and the distributed research community must cooperate. The Centers build and run the models. They make available to the research community the output of the models. While universities do not have the resources to run the models, they can certainly diagnose their output. The university research community thus has access to the latest model output and the Centers gain an ongoing diagnosis of the state of the models.
Similarly, in the building of the models the parameterization of unresolved processes (e.g., clouds, mixing in the ocean, soil moisture) in the Center models should be competed for by interested researchers. The research community would gain the funding and resources (monetary and computer) for better defining climate processes, and the Centers would gain the expertise and perhaps manpower to run the extensive tests needed to find out if new parameterizations improve the model. One of the absolutely necessary functions of the Centers' interaction with the research community is enabling and facilitating the arduous process that takes researchers from the analysis of data from the synoptic network or from field programs to the development of improved parameterization schemes for use in climate models. Again, crucial to this interchange is the standardization of protocols of data and codes common both to the Centers and to the research community. This would also guard against duplication because the codes would be available to and used by all modeling Centers and all members of the research community.
Observations are critical for defining the state of the models and for providing the analyses against which the models can be tested. The climate models of course need to be tested against an analysis provided by a model different from the one being tested. Because much of the model atmospheric data is ingested for purposes of weather prediction at NCEP, it is important that the operational climate centers be tightly connected to NCEP.
The output of the Centers' analysis models should be made available
to the research community. This should include ocean analyses, land analyses, coupled-climate analyses, and atmospheric reanalysis. The research community will benefit from having the various analyses to diagnose, and the operational climate observational community (when they exist) will benefit from the diagnoses. The Centers will have improvements in observing design for their own diverse purposes as one of their tasks, so that both observations and modeling will benefit from an optimized observing system.
We do not mean to imply that the entire climate research community should be engaged with climate operations—this would be neither practical nor desirable. But the benefits to be gained from having climate research interacting with climate operations would stimulate research and enrich operations to an extent that the benefits of interaction would be hard to overlook.
7.3 FROM VISION TO REALITY
Climate Research and Climate Operations are not interchangeable and both are needed to construct and disseminate climate information products for the benefit of society. Climate Operations will be expensive, with the major cost being the climate observing system. Because of the integrated nature of the functions needed for Climate Operations, high-end modeling must be considered an essential part of operations.
The nation needs the best possible climate information on which to base decisions about the future. The panel has no doubt that the nation will, at some point in the future, choose to institute Climate Operations. An effective high-end climate modeling activity is an essential step on the way.