Information about climate1 is used to make decisions every day. From farmers deciding which crops to plant next season to mayors in large cities deciding how to prepare for future heat waves, and from an insurance company assessing future flood risks to a national security planner assessing future conflict risks from the impacts of drought, users of climate information span a vast array of sectors in both the public and private spheres. Each of these communities has different needs for climate data, with different time horizons (see Box S.1) and different tolerances for uncertainty.
Over the next several decades, climate change and its myriad consequences will be further unfolding and possibly accelerating, increasing the demand for climate information. Society will need to respond and adapt to impacts, such as sea-level rise, a seasonally ice-free Arctic, and large-scale ecosystem changes. Historical records are no longer likely to be reliable predictors of future events; climate change will affect the likelihood and severity of extreme weather and climate events, which are a leading cause of economic and human losses with total losses in the hundreds of billions of dollars over the past few decades.2
Computer models that simulate the climate are an integral part of providing climate information, in particular for future changes in the climate. Overall, climate modeling has made enormous progress in the past several decades, but meeting the information needs of users will require further advances in the coming decades.
In an effort to improve the United States’ capabilities to simulate present and future climate on local to global scales and at decadal to centennial time scales, the National Oceanic and Atmospheric Administration (NOAA), the National Aeronautics and Space Administration, the Department of Energy, the National Science Foundation, and the intelligence community requested that the National Research Council (NRC) produce a strategic framework to guide progress in the nation’s climate modeling enterprise over the next 10-20 years. In response, the NRC appointed the Committee on a National Strategy for Advancing Climate Modeling with the task to engage key stakeholders in
1 Climate is conventionally defined as the long-term statistics of the weather (e.g., temperature, precipitation, and other meteorological conditions) that characteristically prevail in a particular region.
2 Total losses from weather- and climate-related disasters are estimated to exceed $700 billion for the time period of 1980-2009 and to exceed $50 billion in 2011 alone from the more than 14 weather- and climate-related disasters in that year. See http://www.noaa.gov/extreme2011 (accessed October 11, 2012).
BOX S.1 INFORMATION FROM CLIMATE MODELS
Climate models skillfully reproduce important, global- to continental-scale features of the present climate, including the simulated seasonal-mean surface air temperature (within 3°C of observed [IPCC, 2007c], compared to an annual cycle that can exceed 50°C in places), the simulated seasonal-mean precipitation (typical errors are 50 percent or less on regional scales of 1,000 km or larger that are well resolved by these models [Pincus et al., 2008]), and representations of major climate features such as major ocean current systems like the Gulf Stream (IPCC, 2007c) or the swings in Pacific sea-surface temperature, winds, and rainfall associated with El Niño (AchutaRao and Sperber, 2006; Neale et al., 2008). Climate modeling also delivers useful forecasts for some phenomena from a month to several seasons ahead, such as seasonal flood risks (Figure 1).
FIGURE 1 Climate models can deliver useful forecasts for some phenomena a month to several seasons ahead, such as this spring flood risk outlook from NOAA’s National Weather Service for 2011. See Chapter 1 for more details. SOURCE: http://www.noaa.gov/extreme2011/mississippi_flood.html (accessed October 11, 2012).
Beyond these advances, however, the climate modeling community aspires to make substantial further progress in the quality of climate projections, especially on regional space scales and decadal time scales, to deliver the types of climate projections with sufficient resolution and accuracy needed by users. For example, Figure 2 shows projected changes to water runoff for later this century.
FIGURE 2 Longer-time-scale climate projections can assist in long-term planning. The figure shows projected changes in annual average runoff by the middle of the 21st century. See Chapter 1 for more details.
SOURCE: USGCRP, 2009.
a discussion of the status and future of climate modeling in the United States over the next decade and beyond; to describe the existing landscape of domestic and international climate modeling efforts; to discuss, in broad terms, the observational, basic, and applied research, infrastructure, and other requirements of current and possible future climate modeling efforts; and to provide conclusions and/or recommendations for developing a comprehensive and integrated national strategy for climate modeling over the next decade and beyond (see Appendix A for the statement of task and Box S.2 for a description of the committee’s activities).
A NATIONAL STRATEGY FOR ADVANCING CLIMATE MODELING
The U.S. climate modeling community is diverse and contains several large global climate modeling efforts and many smaller groups running regional climate models. As a critical step toward making more rapid, efficient, and coordinated progress, the committee envisions an evolutionary change in U.S. climate modeling institutions away from developing multiple completely independent models toward a collaborative approach. A collaborative approach does not mean only one center of modeling; rather it means that different groups pursue different niches or methodologies where scientifically justified, but within a single common modeling framework in which software, data standards and tools, and even model components are shared by all major modeling groups nationwide. An overarching thread of the committee’s vision is to promote unification of the decentralized U.S. climate modeling enterprise—across modeling efforts, across a hierarchy of model types, across modeling communities focused on different space and time scales, and across model developers and model output users.
BOX S.2 THE COMMITTEE’S REPORT PROCESS
The committee held five information-gathering meetings over the course of a year, including a large community workshop, to interact with a range of stakeholders from government labs, federal agencies, academic institutions, international organizations, and the broad user community. The committee examined previous reports on how to improve climate modeling in the United States and interviewed key officials and scientists (see Appendix B for a complete list) to help draw lessons from these reports. The charge to the committee emphasized decadal to centennial time scales, but because of the overlap of issues between decadal and intraseasonal to interannual (ISI) time scales, as well as the potential benefits of testing climate models at shorter time scales, the committee believed it was important to extend the focus of the report to shorter time scales, including ISI time scales.
FIGURE S.1 Driven by the growing need for climate information and the coming transition to radically new computing hardware, a new generation of climate models will be needed to address a wide spectrum of climate information needs. A national strategy consisting of four key unifying elements and several other recommendations can help to achieve this vision.
The committee recommends a national strategy for advancing the climate modeling enterprise in the next two decades, consisting of four main new components and five supporting elements that, while less novel, are equally important (Figure S.1). The nation should
1. Evolve to a common national software infrastructure that supports a diverse hierarchy of different models for different purposes, and which supports a vigorous research program aimed at improving the performance of climate models on extreme-scale computing architectures;
2. Convene an annual climate modeling forum that promotes tighter coordination and more consistent evaluation of U.S. regional and global models, and helps knit together model development and user communities;
3. Nurture a unified weather-climate modeling effort that better exploits the synergies between weather forecasting, data assimilation, and climate modeling; and
4. Develop training, accreditation, and continuing education for “climate interpreters” who will act as a two-way interface between modeling advances and diverse user needs.
At the same time, the nation should nurture and enhance ongoing efforts to
5. Sustain the availability of state-of-the-art computing systems for climate modeling;
6. Continue to contribute to a strong international climate observing system capable of comprehensively characterizing long-term climate trends and climate variability;
7. Develop a training and reward system that entices the most talented computer and climate scientists into climate model development;
8. Enhance the national and international information technology (IT) infrastructure that supports climate modeling data sharing and distribution; and
9. Pursue advances in climate science and uncertainty research.
The elements of this strategy are described in more detail below. If adopted, this strategy provides a path for the United States to move forward into the next generation of climate models to provide the best possible climate information for the nation.
ELEMENTS OF A NATIONAL STRATEGY FOR ADVANCING CLIMATE MODELING
Evolve to Shared Software Infrastructure
The entire climate modeling enterprise is computationally intensive. Over the past 15 years, major climate modeling groups have been forced to devote increasing attention to software engineering. One catalyst was a disruptive hardware transition in the late 1990s from vector to parallel supercomputing. It was viewed with trepidation but the climate modeling community adapted well, in part by moving toward common software infrastructure for basic operations like data regridding and coupling between model components.
All indications are that increases in computing performance through the next decade will arrive not in the form of faster chips, but by connecting far more of them, requiring new approaches optimized for massively parallel computing and customized to particular computer designs. A renewed and aggressive commitment to innovatively
designed common infrastructure across the U.S. climate and weather modeling communities is needed to successfully navigate this transition without massive duplication of effort that greatly slows overall progress.
This idea of a common software infrastructure is not new or controversial. More than a decade ago, approaches such as the Earth System Modeling Framework were pioneered for this purpose and have become influential and fairly widely used, but no one approach has become a nationally adopted standard. Individual U.S. modeling centers have developed different forms of such infrastructure, upon which they now depend, and have learned from those experiences.
Now is the time to aggressively develop a new common software infrastructure to be adopted across all major U.S. climate modeling efforts. Such an infrastructure could be an important tool in facilitating a more integrated plan for U.S. climate modeling. The committee’s vision is that, in a decade, all major U.S. climate models—global and regional—will share a single common software infrastructure that allows interoperability of model components (e.g., atmosphere, land, ocean, or sea ice), even when developed by different centers, and that supports a common data interface. The proposed infrastructure would
• facilitate the migration of models to new, possibly radically different computing platforms (Figure S.2);
• support a research effort to develop high-end global models that execute efficiently on such platforms, enabling cloud-resolving atmospheric resolutions (~2-4 km) and eddy-resolving ocean resolutions (~5 km) within as little as a decade;
• allow centers to easily share model components and design hierarchical model frameworks with individual components simplified or specialized as needed for applications such as paleoclimate or weather forecasting and data assimilation (Figure S.2 and Box S.3);
• allow the academic community, other external modeling groups, and core modeling centers to work together more easily, because different model configurations could be run using very similar scripts; and
• harmonize outputs and file structures from all models, benefiting the model analysis and applications communities.
Decades of experience have shown that a full palette of modeling tools—a “model hierarchy”—is required across various scales and with different degrees of complexity with respect to their representation of the Earth system. The common software infrastructure is envisioned as a tool for linking together a model hierarchy, making it portable to a variety of computer architectures and user friendly for education, aca-
FIGURE S.2 The development of a common software infrastructure that interfaces between the climate modeling computer code and the computing hardware has two important advantages: (1) it will facilitate the migration of models to the next generation of computing platforms by isolating the climate modeling computer code from the changes in hardware and (2) it will allow the interoperability of climate model components, for example to enable the testing of two different atmospheric component models, without having to adapt the component models to different hardware platforms.
demic research, and exploratory science. Within this hierarchy, potential new modeling and evaluation approaches can be tested and compared, and improvements from one type of model can be easily transitioned to other models. It is a manageable investment (at least on a national scale) to carefully design, document, and refine one software infrastructure, and once users have learned it, their experience is transferable to using other model configurations and their output data structures. The committee recommends a community-based design and implementation process for achieving a national common software infrastructure. Although this goal has risks, costs, and institutional hurdles, the committee believes they are far outweighed by its benefits.
BOX S.3 SOFTWARE INFRASTRUCTURE ANALOGY TO OPERATING SYSTEM ON A SMARTPHONE
The software infrastructure described in this report can be thought of as similar to the operating system on a smartphone. The software infrastructure is designed to run on a specific hardware platform (analogous to a specific phone), and climate modelers develop model components (analogous to apps) to run in the software infrastructure to simulate parts of the climate system such as the atmosphere or ocean.
Currently, different modeling centers in the United States have different software infrastructures (operating systems) that run on different pieces of hardware; this is similar to comparing the iPhone to the Android. This means that climate model components (apps) written for one software infrastructure will not work with another (similar to how iPhone apps will not work directly on an Android).
Ultimately, the vision is that the U.S. modeling community could evolve to use the same common software infrastructure (operating system), so that model components (apps) could be interchanged and tested versus one another directly. This would also mean that when the hardware (phone) advances, the software infrastructure (operating system) can be updated to continue to work with the new hardware without having to completely rewrite the climate model components (apps).
The common software infrastructure alone will not allow climate models to take full advantage of the advances in computation of the next 10-20 years. A vigorous research program is needed to improve the performance of climate models on the highly concurrent computer architectures that will be the way forward in the coming decade. The common infrastructure will facilitate the sharing of such an advance across models and modeling centers and thus support this national effort to push the computational frontiers of climate science.
Convene a National Climate Modeling Forum
To help bring together the nation’s diverse and decentralized modeling communities and implement the new common software infrastructure, the committee recommends the establishment of an annual U.S. climate modeling forum in which scientists engaged in both global and regional climate model development and analysis from across the United States, as well as interested users, would gather to focus on timely and important cross-cutting issues related to U.S. climate modeling. While modelers can learn about each other’s progress at conferences and through scholarly journals, this can be slow, haphazard, and inefficient. The goal of the proposed forum is to promote better coordination among scientists involved in major global and regional
modeling efforts across the United States and the user, applications, and analysis communities. These forums could
• serve as a mechanism for informing the community of the current and planned activities at the core modeling centers;
• provide a venue for fostering important interactions among scientists in the core modeling efforts and those at other institutions, including universities;
• facilitate a more coordinated approach to global and regional model development and use in the United States, including the design of common experiments using multiple models and the formation of joint development teams;
• provide an important vehicle to enhance and accelerate communication among climate modeling groups at research and operational modeling centers;
• offer an opportunity to facilitate the development and implementation of a shared national software infrastructure through sustained, regular interactions between the infrastructure software developers and model developers and users;
• offer a vital opportunity for end users of climate model information to both learn about the strengths and limitations of models, and provide input to modelers on the critical needs of end users that could feed back into the model development and application process; and
• provide an opportunity for regular broad-based discussion of strategic priorities for the national climate modeling enterprise.
The development of this approach would benefit greatly from additional resources specifically targeted to such integrative activities, and from support from a strong coordinating institution to integrate activities across multiple agencies. Organizations such as the American Meteorological Society (AMS), the American Geophysical Union (AGU), or the World Climate Research Program could in theory serve this role, but the U.S. Global Change Research Program might be a natural choice for organizing the forum given its mission to coordinate climate research activities in the United States.
Nurture a Unified Weather-Climate Modeling Effort
Unified weather-climate prediction models are increasingly an important part of the spectrum of climate models. Testing a climate model in a “weather forecast” mode, with initial conditions taken from a global analysis from a particular time, allows evaluation of rapidly evolving processes such as cloud properties that are routinely observed. Such simulations are short enough to test model performance over a range of
grid resolutions relevant not only to current but also to prospective climate simulation capabilities. Transitioning to a unified weather-climate prediction approach is a major effort that requires substantial infrastructure. This approach is being successfully used by the UK Met Office, a leading international modeling center. In the United States, no weather or climate modeling center has yet fully embraced this philosophy, though several centers have some capability for weather forecasting, climate simulation, and data assimilation.
The committee recommends an accelerated national modeling effort that spans weather to climate time scales. One method to achieve this would be nurturing at least one U.S. unified weather-climate prediction system capable of state-of-theart forecasts from days to decades, climate-quality data assimilation, and reanalysis. This prediction system would be but one effort within the U.S. climate modeling endeavor. It would be most effective if it involved a collaboration among operational weather forecast centers, data assimilation centers, climate modeling centers, and the external research community, which would need to work together to define a unified modeling strategy and initial implementation steps. To facilitate cross-fertilization with other climate modeling efforts, this effort should take advantage of the common software infrastructure and community-wide code and data accessibility described in the rest of this committee’s strategy. Its success would be judged by simultaneous improvement of forecast skill metrics on all time scales.
Develop a Program for Climate Model Interpreters
By improving climate models, the scientific community has made considerable progress in the past decades in its capability to project future climate and its impacts. Nonetheless, important details about future climate remain uncertain. Simultaneously, addressing the wide spectrum of user climate information needs is outpacing the limited capacity of people within the climate modeling community. Effective communication about climate change and its uncertainty to science managers and decision makers is a crucial part of advancing our national climate modeling capability. There is no simple formulaic way to communicate uncertainty; as climate models and their available outputs become more sophisticated, those looking to use this information struggle to keep up.
Climate information is already being provided by a number of public and private entities in various capacities, and there have been numerous other calls for the provision of more extensive government-run climate information services. The committee chose to not weigh in on the debate about the appropriate role for the federal government
in providing climate services. Rather, the committee notes the need for qualified individuals who can provide credible information to end users based on current climate models, wherever they work.
To address this need, the committee recommends developing a national education and accreditation program for “climate model interpreters” who can take technical findings and output from climate models, including quantified uncertainties, and use them in a diverse range of private- and public-sector applications. The education component could be a degree or certificate program offered by universities with adequate expertise in climate science and modeling, and the accreditation could be through a national organization that has a broad reach and is independent of any agency or modeling center, such as the AMS or the AGU. The training of climate interpreters is not envisioned as the solution to address all user needs for climate information, but rather as a crucial step that benefits any system for any of the various mechanisms that bridge the climate modeling and user communities.
Sustain State-of-the-Art Computing Systems for Climate Modeling
Climate simulation is difficult because it involves many physical processes interacting over a large range of space and time scales. Past experience shows that increasing the range of scales resolved by the model grid ultimately leads to more accurate models and informs the development of lower-resolution models. Therefore, to advance climate modeling, U.S. climate science will need the best possible computing platform and models.
The committee recommends a two-pronged approach that involves the continued use and upgrading of dedicated computing resources at the existing modeling centers, complemented by research into more efficient exploitation of the highly concurrent computer architectures that are expected in the next 10-20 years.
The community has been able to exploit other extreme-scale computing facilities that are not solely dedicated to climate as resources of opportunity. Continuing to do so will likely prove useful, but access to these external systems can be unreliable, and they often have operating protocols that are not suited to the very long simulations often needed for climate models. The committee debated whether the current combination of institution-specific computing and use of external computer resources of opportunity was the best national strategy for climate computing. The pros and cons of a national climate computing facility were weighed, and it was concluded that such
a facility would be beneficial only if it were created in addition to the current computing capabilities at the modeling centers. An expensive new national climate computing facility would be most attractive and least risky in an environment of sustained budget growth for climate science and modeling, which would allow it to be pursued in parallel with other critical investments in climate modeling.
Continue to Contribute to a Strong International Climate Observing System
Observations are critical for monitoring and advancing understanding of the processes driving the variability and trajectory of the climate system. The evaluation and improvement of climate and Earth system models is thus fundamentally tied to the quality of the observing system for climate. A national strategy for climate modeling would be incomplete without a well-maintained climate observing system capable of comprehensively characterizing long-term climate trends and climate variability. Maintaining a climate observing system is an international enterprise but requires strong U.S. support that has come under serious threat. Over the next several decades, it is imperative to maintain existing long-term data sets of essential climate variables, in tandem with innovative new measurements that illuminate Earth system processes that are still poorly characterized.
Develop a Training and Reward System for Climate Model Developers
Model development is among the most challenging tasks in climate science, because it demands synthetic knowledge of climate physics, biogeochemistry, numerical analysis, and computing environments as well as the ability to work effectively in a large group. The committee recommends enticing high-caliber computer and climate scientists to become climate model developers using graduate fellowships in modeling centers, extended postdoctoral traineeships of 3-5 years, and rewards for model advancement through clear well-paid career tracks, institutional recognition, quick advancement, and adequate funding opportunities.
Enhance the National IT Infrastructure That Supports Climate
Modeling Data Sharing and Distribution
The growth rate of climate model data archives is exponential, and maintaining access to these data is a growing challenge. Observational data about the Earth system are also becoming much more voluminous and diverse. The climate research community,
decision makers, and other user communities desire to analyze and use both types of data in increasingly sophisticated ways. These trends imply growth in resource demands that cannot be managed in an ad hoc way. Instead, the data-sharing infrastructure for supporting international and national model intercomparisons and other simulations of broad interest—including archiving and distributing model outputs to the research and user communities—should be systematically supported as an operational backbone for climate research and serving the user community. Beyond stabilizing support for current efforts, the United States should develop a national IT infrastructure for Earth system climate observations and model data that builds from existing efforts, so as to facilitate and accelerate data display, visualization, and analysis both for experts and for the broader user community. Without substantial research effort into new methods of storage, data dissemination, data semantics, and visualization, all aimed at bringing analysis and computation to the data, rather than trying to download the data and perform analysis locally, it is likely that the data might become frustratingly inaccessible to users.
Pursue Advances in Climate Science and Uncertainty Research
To meet the national need for improved information and guidance over the coming decades, U.S. climate models will have to address an expanding breadth of scientific problems while improving the fidelity of predictions and projections from intraseasonal to centennial time scales. The committee finds that climate modeling in the United States can make significant progress through a combination of increasing model resolution, advances in observations and process understanding, improved representations in models of unresolved but climate-relevant processes, and more complete representations of the Earth system in climate models. As a general guideline for most effectively meeting future climate information needs, climate modeling activities should focus on problems whose solution will help climate models better inform societal needs, and for which progress is likely given adequate resources. With such focus, advances in Earth system modeling may yield significant progress in the next decade or two for a number of scientific questions, including sea-ice loss, icesheet stability, land-ocean ecosystem and carbon-cycle change, regional precipitation changes and extremes, cloud-climate interaction, and climate sensitivity.
As these challenges are faced and models grow in complexity, they are likely to exhibit an increasingly rich range of behavior, full of surprises and unexpected results. Therefore, the committee emphasizes that it is unwise to promise that successive generations of models will invariably result in firmer predictive capability. Progress on these challenges is important, however, to develop a fuller understanding of the climate
system, reducing the likelihood of unanticipated changes and improving climate models in the long term.
Uncertainty is a significant aspect of climate modeling and needs to be properly addressed by the climate modeling community. To facilitate this, the United States should more vigorously support research on uncertainty, including understanding and quantifying climate projection uncertainty, automating approaches to optimization of uncertain parameters within models, communicating uncertainty to both users of climate model output and decision makers, and developing deeper understanding on the relationship between uncertainty and decision making.
Climate models are among the most sophisticated simulation tools developed by mankind and the “what-if” questions we are asking of them involve a mind-boggling number of connected systems. As the scope of climate models has expanded, so has the need to validate and improve them. Enormous progress has been made in the past several decades in improving the utility and robustness of climate models, but more is needed to meet the desires of decision makers who are increasingly relying on the information from climate models.
The committee believes that the best path forward is a strategy centered around the integration of the decentralized U.S. climate modeling enterprise—across modeling efforts, across a hierarchy of model types, across modeling communities focused on different space and time scales, and between model developers and model output users. A diversity of approaches is necessary for progress in many areas of climate modeling and is vital for addressing the breadth of users’ needs. If adopted, this strategy of increased unification amidst diversity will allow the United States to more effectively meet the climate information needs of the nation in the coming decades and beyond.