Atmosphere in a box: NCAR's first GCM

Photo of man looking at film
Warren Washington was among the first climate scientists to animate output from general circulation models.

Computer models of global climate became indispensable tools in the last decades of the 20th century as society began to grapple with the impact of human-produced greenhouse gases. In the early 1960s, though, the challenge was simply to reproduce the general features of the atmosphere circulating around the globe—thus, the quest for a general circulation model, or GCM.

Two young NCAR scientists were among the pioneers who set out to achieve this formidable goal. Warren Washington came to the center in 1963, not long after completing his doctorate at Pennsylvania State University. A few months earlier, Akira Kasahara had joined NCAR after taking part in pioneering modeling efforts at universities in Japan and the United States. The two had met at the Courant Institute of Mathematical Sciences in New York, a hotbed of research into atmospheric modeling. They shared a keen interest in pushing the limits of the relatively modest computing resources then available. In a 1998 interview, Kasahara recalled that when Washington approached him about pooling their skills to build a model, “I was quite excited, because I had been thinking about the same thing.”

As it does now, global modeling hinged on a set of equations put forth by Norwegian meteorologist Vilhelm Bjerknes in the early 20th century to describe atmospheric motion. Scientists had longed to use those equations to calculate the present behavior and future state of the atmosphere: in the 1920s, British researcher L.F. Richardson had envisioned a “forecast factory,” where hundreds of clerks would carry out manual calculations. However, the goal of numerically simulating the state of the atmosphere remained out of reach until computers came on the scene. By the early 1960s, two teams—one at the U.S. Weather Bureau, the other at the University of California, Los Angeles—had set out to build full-fledged GCMs.

Photo of man standing next to computer magnetic tapes
By the end of the 1960s, Akira Kasahara and colleagues were storing the increasing amounts of data from NCAR’s climate simulations onto nine-track magnetic tapes.

Could NCAR match these efforts? Washington was encouraged by the work of Cecil Leith, then at Lawrence Livermore National Laboratory and later at NCAR. “His heroic effort of building single-handedly a comprehensive atmospheric model impressed many people in the field, including me,” says Washington.
In order to run a GCM, Kasahara and Washington knew they would need a substantial hunk of NCAR’s computational muscle. The center’s original plans had put the emphasis on processing aircraft and satellite data rather than modeling. But the two scientists found an ally in NCAR’s associate director, Philip Thompson, who helped arrange for time on the center’s first computer, a Control Data Corporation 3600.

Using schematic continents and observed ocean temperatures, Kasahara and Washington kicked off their model runs with a cold start—uniformly cool temperatures across the globe—to see if the model’s inclusion of solar and oceanic heat would produce realistic weather. “This was an exciting time,” recalls Washington. “The model allowed us to play ‘god’ a bit.” Indeed, within a week of model time, high and low pressure centers materialized, and after 40 days and nights, the simulated climate had fallen into a rough equilibrium. Cold and warm fronts scooted across midlatitudes; doldrums appeared in the tropics. In short, it was a recognizable atmosphere, which made the project a resounding success.

Through several iterations over the late 1960s and 1970s, Kasahara and Washington refined their model. At the same time, a handful of groups across the world launched global modeling efforts of their own. Most universities, though, were still shut out of the action. In response, NCAR began shifting to a new Community Climate Model aimed at providing university scientists with software as well as technical support. The first CCM, built in the early 1980s, drew on a potpourri of innovative formulations developed at Australian and European labs and at NOAA’s Geophysical Fluid Dynamics Laboratory. As the CCM grew in scope and influence, it evolved into the Community Climate System Model, one of the world’s leading comprehensive global models, and its recent successor, the Community Earth System Model.

Two men sitting at table with book shelves behind them
Akira Kasahara and Warren Washington both remain active at NCAR.

Along with their landmark modeling, Kasahara and Washington were pioneers in another sense. Their GCM team, which included two women programmers, was uncommonly diverse for the 1960s. Washington went on to become a leader in science policy as well as modeling: he served on the National Science Board and advised five U.S. presidents in various capacities, and he remains active in collaborative modeling research involving NCAR and the U.S. Department of Energy. Though Kasahara is now retired, he and Washington continue to work with colleagues at NCAR and elsewhere on projects of interest.


Today — CCSM's working groups:  a model community

Photo of Minghua Zhang

"Working on the CCSM has been a continuous learning experience."

—Minghua Zhang, Stony Brook University, co-chair of the CCSM atmosphere model working group

Each of the dozen or so leading models of the Earth system created by researchers worldwide—depicting atmosphere, ocean, land, ice, and more—has its strengths and weaknesses. When it comes to human capital, though, few can match NCAR’s Community Climate System Model. The CCSM is freely available to scientists worldwide, and NCAR provides extensive support to help guide users. The model itself is largely a product of collaboration between researchers and developers at NCAR and their colleagues at universities and laboratories across the nation and beyond.

One of the main vehicles for making CCSM better is its lineup of 12 working groups. Each is devoted to a component or process, from the atmosphere, ocean, and land models to paleoclimate and biogeochemistry. As the CCSM grows in complexity, new groups arise. Most are co-chaired by an NCAR scientist and a researcher from a university or federal laboratory.

Photo of two women sitting at a lecture
Mingjie Shi (University of Texas at Austin) and Kendra Castillo (Purdue University) were among the land surface modelers who met with CCSM developers at a working group meeting in February 2010.

“I believe that the working group process has been one of UCAR’s most successful activities,” says Robert Dickinson (Georgia Institute of Technology). While at NCAR, Dickinson led the creation of NCAR’s first model to account for forests, grasslands, and other surface features. As he notes, no single institution can provide the expertise needed to pull together increasingly complex global models. However, he says, “That expertise can be obtained by a system that allows participation by all interested scientists capable of providing the needed help. This has been the approach of the CCSM working groups, and overall they have been a remarkable success.”

Each working group typically meets several times a year, together examining the latest model improvements and scrutinizing the minutiae that can make or break a model’s fidelity and tractability. In June, all of the groups convene at the CCSM Annual Meeting, which draws hundreds of users and developers.

NCAR’s Marika Holland has led studies that employ the CCSM to highlight the risk of dramatic ice retreat in future Arctic summers. Holland co-chairs the working group on polar climate with Elizabeth Hunke (Los Alamos National Laboratory). As Hunke notes, “The group at NCAR that works directly on sea ice model improvements is quite small, and yet we have a very sophisticated sea ice model component within the CCSM. This is largely due to collaborations fostered and maintained through the working-group structure.” For example, outgoing (short-wave) radiation is a critical element in Arctic climate, where clouds and ice can change the balance enormously. An improved technique for depicting such radiation was recently developed for CCSM4 by Bruce Briegleb (NCAR) and Bonnie Light (University of Washington).

For several reasons—its wide usage, its distributed governance, and its complexity—“the CCSM may not always be the world’s first model to implement a brand-new technique,” says NCAR’s James Hurrell, head of the CCSM Scientific Steering Committee. However, he adds, the benefits of openness and community access are well worth it.

Map of the world
For the 2007 assessment report of the Intergovernmental Panel on Climate Change, modelers bumped up the resolution of the atmospheric and land-surface components of the CCSM. The simulation produced a level of detail on fronts and storm systems rarely seen in global models.