Climate change projections rely heavily on computer models to provide a physical foundation for the climate’s future. At present, the vast majority of U.S. and international resources for climate modeling are dedicated to the development of so-called Earth system models (ESMs). Such models are constructed to numerically simulate Earth’s climate with the greatest possible fidelity and are thus built to include the most comprehensive range of physical, chemical, and biological processes that can be handled on today’s most powerful computing systems.
The immense complexity of these ESMs arises from the need to generate the most accurate projections of climate change in the coming decades. These projections play a crucial role in the assessments of the Intergovernmental Panel on Climate Change; they also inform climate-related decision-making in a wide variety of sectors.
Although this modeling approach is important for making accurate projections, it imposes substantial limitations when it comes to obtaining a fundamental understanding of the Earth system. The large number of simulated processes and the high resolution at which the simulations are typically performed require that these complex simulations be run on very expensive supercomputers. This requirement greatly limits our ability to explore the models’ sensitivities to different system components and climate forcings. As a consequence, our ability to deeply understand the behavior of these models is limited.
Earth system models may be good for simulating the climate system but may not be as valuable for understanding it.
In a nutshell, ESMs may be good for simulating the climate system but may not be as valuable for understanding it. So we have now added a new set of tools within the Community Earth System Model (CESM) project: a hierarchy of simpler models to foster this understanding. Specifically, we are happy to announce that the next version of CESM will include two simple atmospheric models: a “dynamical core” and an “aquaplanet.”
A Hierarchy of Models
Isaac Held alerted the climate science community to a widening divide between simulating and understanding the climate system more than a decade ago [Held, 2005]. He noted that a gap had developed between idealized models, which can sometimes be thoroughly understood, and the complex, high-end models (such as ESMs) that comprehensively simulate the climate system but cannot be completely understood because of their complexity.
This gap is particularly problematic for many researchers and students, who often have to work with limited computational resources. Furthermore, simple climate models are key to educational activities in university classrooms and as part of graduate research: They provide an entry point that enables students to master modeling techniques and concepts before moving on to more complex methods.
Held [2005, p. 1610], building on earlier suggestions [Schneider and Dickinson, 1974; Hoskins, 1983], emphasized that the gap between understanding and simulation could be bridged by developing a hierarchy of climate models:
Consider, by analogy, another field that must deal with exceedingly complex systems—molecular biology. How is it that biologists have made such dramatic and steady progress in sorting out the human genome and the interactions of the thousands of proteins of which we are constructed? Without doubt, one key has been that nature has provided us with a hierarchy of biological systems of increasing complexity that are amenable to experimental manipulation, ranging from bacteria to fruit fly to mouse to man. Furthermore, the nature of evolution assures us that much of what we learn from simpler organisms is directly relevant to deciphering the workings of their more complex relatives. What good fortune for biologists to be presented with precisely the kind of hierarchy needed to understand a complex system! Imagine how much progress would have been made if they were limited to studying man alone.
What is needed now is a set of simpler models, ideally embedded within the same modeling infrastructure as the ESMs.
So how might one proceed in developing such a hierarchy of models for the Earth system? One end of the hierarchy already exists: the most complex ESMs. What is needed now is a set of simpler models, ideally embedded within the same modeling infrastructure as the ESMs. This embedding would allow the simple models to selectively make use of the same components (e.g., the cloud scheme) as the ESM and would enable users to easily set up, run, and analyze any model along the hierarchy, from the simplest all the way to the most complex.
The Community Earth System Model
In the United States, the CESM project is the natural choice for building this hierarchy of climate models. The ESMs made available under the CESM project are developed, maintained, and documented specifically to serve the entire community of climate scientists. The CESM project has a wide user base; all released model versions are thoroughly tested and routinely ported to the latest machines available, an arduous task not easily accomplished by individual investigators in a typical university setting. Crucially, the CESM can be freely downloaded and used by anyone.
With these considerations in mind, we (the authors) informally approached colleagues to explore the general sentiment about developing simpler models for climate science research. In October 2013, we held a special session on “Simpler Models” at the Annual Member’s Meeting of the University Corporation for Atmospheric Research (UCAR). In November 2013, two of us sent a letter to approximately 50 university faculty members engaged in climate modeling research, asking about their interest in developing a model hierarchy within CESM: The response was overwhelmingly positive. We also gathered feedback from the larger CESM user community, and it became clear there was widespread support for the initiative. In late 2014, we received a formal green light from the CESM leadership and the National Science Foundation to start developing an officially supported set of simpler model configurations.
First Steps Toward a Hierarchy of Models
We are now happy to report that after 2 years of work, the fruit of our labors will soon appear. Specifically, with the impending release of CESM 2.0, two idealized atmospheric model configurations—a dynamical core and an aquaplanet—will be made available to the climate science community.
In many ways, these two components constitute the bookends of the atmospheric model hierarchy. The dynamical core model solves the fluid dynamical equations alone, all other physical processes having been immensely simplified. At the other extreme, the aquaplanet model is nearly identical in complexity to the atmospheric component of the ESM itself, the only simplification being that in the model, all landmasses have been eliminated and oceans cover Earth’s entire surface.
Earlier versions of such models have existed, in one form or another, at many modeling centers at various times, but the novelty is that these models will now be officially incorporated into the most current version of CESM. From CESM 2.0 onward, simpler climate model configurations will be made available, maintained, documented, and updated for use by the entire climate science community. In many ways, however, this is only the beginning.
Filling in the Hierarchy
What happens next? First, we hope that colleagues working in climate science will take advantage of these two model configurations and use them in their research. Second, now that we have opened the door and allowed simpler models to enter the CESM project, we invite colleagues to take the initiative in suggesting and contributing other simplified model configurations to “fill up” the hierarchy. For instance, work is underway for so-called single-column configurations of both the atmosphere and ocean models. We also expect idealized configurations of other components (e.g., the land model) to be released in the near future.
How should the community now go about filling up the levels of the hierarchy? We recommend the following guiding principles for moving forward.
First, there should be a clearly demonstrated need for a simplified model. Second, the community members who are spearheading each effort need to pair up with one or more partners at the National Center for Atmospheric Research to collaborate in the model development. Third, project leaders must assess the resources needed to develop the new simpler model and identify some avenue for funding the development. Finally, developers of simpler models should submit scientific papers to the peer-reviewed literature, explaining the rationale for the usefulness of the new models and describing what novel understanding of the climate system is to be gained from using these simpler models. Ideally, these activities would be coordinated across the community to avoid duplication of efforts.
Toward Models of Lasting Value
Models in the hierarchy must be of lasting value.
We conclude by emphasizing one crucial point in Held’s proposal: Models in the hierarchy must be of lasting value. ESMs are constantly under development to promptly incorporate the latest findings or methods. However, we believe that at least some of the models in the hierarchy need to be forcefully shielded from the relentless cycle of model improving and updating. If those models are well chosen, their value will come precisely from the fact that they are not being updated. Because they remain unchanged, we will be able to understand them in great depth and thus close the gap between simulation and understanding—the ultimate motivation of this entire exercise.
Acknowledgments
The authors wish to express their gratitude to Eric DeWeaver, director of Climate and Large-Scale Dynamics in the division of Atmospheric and Geospace Sciences at the National Science Foundation, for his strong and continued support. They are also very grateful to Bill Large, director of the Climate and Global Dynamics laboratory at the National Center for Atmospheric Research, for his invaluable help.
References
Held, I. (2005), The gap between simulation and understanding in climate modeling, Bull. Am. Meteorol. Soc., 86, 1609–1614, https://doi.org/10.1175/BAMS-86-11-1609.
Hoskins, B. J. (1983), Dynamical processes in the atmosphere and the use of models, Q. J. R. Meteorol. Soc., 109, 1–21, https://doi.org/10.1002/qj.49710945902.
Schneider, S. H., and R. E. Dickinson (1974), Climate modeling, Rev. Geophys., 12, 447–493, https://doi.org/10.1029/RG012i003p00447.
—L. M. Polvani (email: lmp@columbia.edu), Columbia University, New York, N.Y.; A. C. Clement, University of Miami, Fla.; and B. Medeiros, J. J. Benedict, and I. R. Simpson, National Center for Atmospheric Research, Boulder, Colo.
Citation:
Polvani, L. M.,Clement, A. C.,Medeiros, B.,Benedict, J. J., and Simpson, I. R. (2017), When less is more: Opening the door to simpler climate models, Eos, 98, https://doi.org/10.1029/2017EO079417. Published on 25 September 2017.
Text © 2017. The authors. CC BY 3.0
Except where otherwise noted, images are subject to copyright. Any reuse without express permission from the copyright owner is prohibited.