UCAR > Communications > UCAR Quarterly > Summer 2000 Search


Summer 2000


Community models and collaboration

Here I define community models as models that are widely available and used by the scientific community, include the community in their development and improvement, and have a significant level of documentation.

I recently had the pleasure of writing the introduction for the Community Climate System Model Plan (2000–2005). The evolution of community climate modeling at NCAR from the atmosphere-only Community Climate Model in 1981 to the Climate System Model (CSM) in 1994 and now to the Community Climate System Model (CCSM) in 2000 has been quite remarkable. The CCSM contains fully interacting component models of the atmosphere, oceans, land, and sea ice, and work is well under way to include interactive biogeochemical, ecological, and chemical processes in the entire system. Parallel and coordinated efforts are under way to couple the neutral atmosphere with the thermosphere. This effort will produce a scientific tool powerful enough to investigate quantitatively solar influences on tropospheric climate, a dream of UCAR and NCAR's father, Walter Orr Roberts.

Almost as impressive as the progress made in coupling the components of the earth system is the progress made in getting many smart and independent people from different disciplines and institutions to work together on a true community model. Like the weather, everybody talks about the need for cooperation, partnerships, strategic alliances, and interdisciplinary research, but few actually do anything about it, at least at the scale of the CCSM.

Another community of scientists has collaborated over the years on mesoscale models. The Penn State-NCAR Mesoscale Model was developed over a 30- year period with the help of hundreds of scientists and students. Other community mesoscale models have been developed at universities—for example, the Regional Atmospheric Modeling System under the leadership of Bill Cotton and Roger Pielke, Sr., at Colorado State University and the Advanced Regional Prediction System under the leadership of Kelvin Droegemeier at the University of Oklahoma. I estimate conservatively that together these three models are being used by more than 1,200 users at more than 600 institutions worldwide. Building on these efforts, the Weather Research and Forecast model is being developed jointly by the research and operational communities to serve as a cutting-edge model for research and operations.

From the vigor of these community modeling activities, and others, it is easy to conclude that the concept of community models is a good one. The advantages seem obvious. In the early days of numerical modeling, one or two people could develop a model for their own specific use if they had a working knowledge of basic numerical methods, computer programming, physical and chemical processes, and analysis and initialization techniques, and above all a good knowledge of the observed structure and behavior of the atmosphere. Now models include a wider range of physical, chemical, and biological processes; numerical and computational challenges have increased; and models have became general enough to apply to an enormous variety of phenomena and physical locations. It is impossible for a few people to carry out a state-of-the-art, balanced modeling program. A community of scientists, each contributing in creative, innovative ways, is needed to build a cutting-edge modeling system and apply this sophisticated scientific tool or facility to learn about the earth system.

In spite of these obvious advantages of community models, there may be drawbacks to the concept. Does the focus on only a few models stifle innovation and creativity? Do students and other users treat the models as "black boxes" in which they turn a few knobs, push a few buttons, and get marvelous output that can be visualized in four dimensions and presented at conferences, but without producing any new physical insights or without contributing to further model improvements? Are community models used by students to obtain "easy" theses and degrees and by scientists to produce "easy" publications? Are the promotion and reward systems in the universities and in laboratories like NCAR able to cope fairly with multiple collaborators and identify and reward the scientific contributions of each one? There are difficulties in maintaining community models, such as deciding when to make changes in the officially supported versions. Pielke and Cotton point out that most users make changes of some type in the supported versions, and that hundreds of users each making changes produces a "centrifugal force" on the model. They also point out that there is a danger of the official versions falling behind the capabilities of "maverick" versions developed in the field.

Upon reflection and discussion with colleagues, I believe that the advantages of community models greatly outweigh the disadvantages, if for no other reason than the multi- and interdisciplinary nature of the science requires that large numbers of experts work together. We cannot return to the days when one or two people could develop and apply models. The possible flaws in the community concept raised above can be addressed in a variety of ways: by having more than one, but not dozens, of community models; by requiring students and users of models to demonstrate creativity in how they use the models and interpret the results before the thesis is accepted or the paper published; by encouraging users to feed back their experiences to the team of model developers and so improve the models; by paying careful attention to the members of the teams that develop and apply the models to make sure that their contributions are recognized.

As we enter the next millennium, we are seeing a convergence of many of the community modeling approaches that have been carried out in parallel and pretty much separately over the past 35 years. Mesoscale models are being coupled with global climate models to study regional climates all over the world. Upper-atmosphere models are being extended downward toward the earth's surface to meet the neutral-atmosphere models as they grow upward. We are beginning to see many successful studies using fully interactive models of the earth system.

While exciting, these changes present a number of challenges, including a myriad of scientific ones of solving tough problems in individual components of the earth system (such as cloud-radiation interactions) and in coupling all of the components, many of which operate on vastly different temporal and spatial scales. The challenges are also computational—how to achieve the massive increases in sustained computer performance needed to run ensembles of simulations at high resolution for a large variety of scientific and assessment purposes. The social challenge presented by the need for a very large number of scientists and policy makers from quite different disciplines, backgrounds, and cultures to work together in a sustained and harmonious manner cannot be overemphasized. And finally, adequate financial resources must be provided in order to support the human resources and the infrastructure necessary to make the requisite progress on critically important global, national, regional, and local environmental problems. As we move toward a complete Earth System Model and higher- resolution, advanced weather-forecasting and regional climate models, we appreciate more than ever before the many dimensions of this "grand challenge" problem.


In this issue... Other issues of UCAR Quarterly
UCARNCARUOP

Edited by Carol Rasmussen, carolr@ucar.edu
Prepared for the Web by Jacque Marshall
Last revised: Fri Sep 1 16:44:56 MDT 2000