UCAR > Communications > Highlights > 1996 Search

Computer models

How do you simulate 100 years of climate inside a computer?

It would be so convenient. If only the atmosphere were a self-contained system, scientists could predict its future without having to worry about the other parts of our planet.

However, in the physical as well as the social world, interrelationships run deep. The air around us is in the midst of constant exchange with the ocean, plants, land, ice, and sun. Each of these exchanges can be mimicked in time and space by a computer model. Until now, though, only a few large research centers have had the human and machine power to track all of these components of the earth system and follow the interplay among them.



The climate system model (CSM)


NCAR scientists and programmers, led by Byron Boville and Peter Gent; a steering committee; technical support and hardware from NCAR's Scientific Computing Division; more than 100 colleagues using the model at universities across North America and beyond


To improve computer simulations of global climate, both future and past, by enabling different parts of the earth system to be represented in an interactive fashion


By taking previous computer models of the atmosphere, ocean, sea ice, and land surface; updating them; linking them through flux-coupler software; and adding chemical, geological, and biological components over time

The climate system model will incorporate land, sea, vegetation, ice, clouds, and the atmosphere. Illustration by Elizabeth C. Johnston.


Created by NCAR's Climate and Global Dynamics Division (CGD) with contributions from universities and the National Oceanic and Atmospheric Administration; used by scientists worldwide on computers at NCAR's Climate System Laboratory or at their home institutions


Model configuration finalized at the end of 1995 and released to universities in May 1996; refinements ongoing

The NCAR Climate System Model (CSM) Project

This state of affairs changed in May 1996, when NCAR introduced a complex and powerful tool for its own scientists and for those at universities: a fully integrated climate system model (CSM). The gift wasn't unexpected, but it was duly appreciated. With the CSM and the Internet, a researcher at a rural school with modest resources can now operate one of the world's most complex, realistic climate models on a powerhouse supercomputer at NCAR--or obtain a scaled-down set of simulations on a workstation in her office.

The CSM's innovation lies in its couplings: ocean to atmosphere, atmosphere to land surface, ocean to sea ice, and several others. Such coupled models allow the monologue from a single computer program to become a conversation between two or more. For example, winds predicted by an atmospheric model drive ocean currents; in turn, the currents may allow colder water to well upward and cool the atmosphere above a certain patch of sea. Exchanging data back and forth in a high-speed dialogue, the programs nudge each other toward a more realistic view of the atmosphere.

"The National Science Foundation [NSF] saw in the early 1990s that a lot of institutions needed coupled models and weren't going to get them unless they built them themselves," says CSM development cochair Byron Boville. He and his NCAR colleagues knew that wasn't an option for many universities. Good climate models take years to develop, and they require multimillion-dollar supercomputers in order to produce high-resolution results quickly.

Such a model, they knew, could be built at NCAR. Indeed, it is part of NCAR's mission to provide for the university community what no single school can provide for itself. A classic example is the NCAR community climate model, launched in the 1980s. For 15 years, this package of software and support reigned as the tool of choice for university scientists studying long-term climate processes that unfold over many decades.

Useful as it was, the community climate model had its limits. Lacking coupling with the oceans and other parts of the climate system, it forced users to specify conditions outside the atmosphere itself, such as sea-surface temperature or the area covered by sea ice. Like a standard transmission in a car, each shift in these variables had to be done manually by the human modeler as the atmosphere evolved. In a coupled model, many of these variables can change automatically, in sync with the other unfolding parts of the earth system.

So, in line with its long history of community service, NCAR's Climate and Global Dynamics Division redirected its model development. Beginning in late 1993, the effort was infused with new NSF funds to create a full-fledged coupled climate model for the community.

There was ample material to start with. A number of NCAR scientists had developed their own two-way coupled models for specific tasks. Warren Washington and Gerald Meehl were shedding new light on El Niño with a coupled ocean/atmosphere model. Gordon Bonan was studying the interchange between land surface and the atmosphere. Both of these models were incorporated into the CSM, as was an ocean model from the National Oceanic and Atmospheric Administration.

It wasn't enough to bring the software together, though. Since the models were created separately by different groups of scientists, each had its own hallmarks: a different resolution, for instance, or a unique way of characterizing rainfall. Allowing a conversation among this ecumenical group of models called for a go-between. For this, the NCAR developers created what they call a flux coupler. It acts like a translator, passing the output of each model to the others and ensuring that the data are mutually understandable (set to the same resolution, for example). "As far as we can tell," says Boville, "the flux coupler is an entirely original concept."

What might the CSM deliver for society at large? Bonan's research provides a hint. In tracing the effect of agricultural practices such as deforestation, he has found substantial atmospheric signals, such as warmer and drier conditions in the Amazon, where clear-cutting is rampant. But it need not always be humans behind the change, says Bonan. "There are a lot of natural changes in vegetation over hundreds or thousands of years that also can have an impact on climate."

Peter Gent and Byron Boville flank an NCAR supercomputer. Photo by Carlye Calvin.

Climate changes in the past, rather than the future, have the attention of other researchers. Eric Barron (Pennsylvania State University) is a paleoclimate modeler who sees the CSM as a useful tool for university scientists like him who work to replicate ancient climates on microchips.

"The climate recorded in rocks and sediments reflects the integrated response of the earth system to a variety of different factors. Addressing paleoclimates by examining only the ocean or only the atmosphere has been very challenging. The CSM presents us the best opportunity to date to examine the rich record of global change recorded in earth history."

Through the late 1990s, NCAR will be issuing updates to the CSM. New models will join the coupled conversations, such as ones to describe the earth's hydrologic and ecologic cycles and the interchange of chemicals among soils, plants, and the atmosphere. University scientists will sit on a steering committee to oversee the model's development. The CSM will become increasingly portable, so that scientists can pull its code from the Internet and operate parts of it on their own workstations, or run the entire model at low resolution for testing linkages. (All the while, NCAR assistance will be only a phone call or e-mail away.)

For the biggest simulations, supercomputers will remain essential. NCAR's computing facility is prepared for the crunch with the recent installation of its Climate System Laboratory. It is one of only a handful of computer clusters worldwide devoted to massive simulations of global climate. The lab carries out billions of calculations per second in support of the CSM as well as models from other institutions.

The new face of climate modeling at NCAR is enlivened by an older dream, the one that founded the institution: people and tools assembled to serve the most ambitious goals of atmospheric scientists and the needs of society at large.

One hot model

Getting the lowdown on the high atmosphere

The summer of 1996 brought rampant forest fires to the southwestern United States and Alaska. It also brought new understanding of how fires and the atmosphere are interrelated. NCAR's Terry Clark and colleagues in the Mesoscale and Microscale Meteorology Division paired his mesoscale model--which had long been used to study thunderstorms and other small-scale weather features--with an Australian model of dry eucalyptus forest fires.

Fingers of flame about a kilometer apart are evident in this photo of th e Owens Valley, California, fire of July 1987. (Photo by Charles George, courtesy International Fire Science Laboratory.)

Using a variety of wind speeds, he and colleagues sharpened the model's resolution to as fine as 20 meters (65 feet) to study how a fire can alter the circulation patterns around itself. They found that a fire's growth depends not only on large-scale winds but on the balance between those winds and a fire's heat output. If the winds relative to an advancing fire line are weak, and the heating is particularly strong, a fire can force its own circulations, possibly resulting in unstable, "blow-up" conditions. Strong winds relative to a fire line, though literally fanning the flames, tend to produce a more stable regime in which the fire is less likely to create its own circulation.

The model also helped to explain the growth of kilometer-wide fingers of flame. These well-known features, long attributed to geography or fire-fuel variation, may instead be caused by inherent instability in a fire line at a given wind speed.

In about the same vertical distance as a drive from New York to Washington, the atmosphere changes from a well-mixed, dense molecular bath to a virtually empty realm. NCAR senior scientist Ray Roble has spent years creating a general circulation model (GCM) of the upper atmosphere, the transition zone between the bulk of the atmosphere--below 80 kilometers (55 miles)--and the void beyond 500 kilometers (300 miles). His models have gradually incorporated the dynamic, chemical, magnetic, and electric forces shaping the upper atmosphere.

Late in 1995, Roble and colleagues in NCAR's High Altitude Observatory completed the first direct link between upper- and lower-atmosphere GCMs. They connected his upper-atmosphere model with version 2 of the NCAR community climate model. The linkage was made possible largely due to the creation of the climate system model (see adjacent story), which gave the scientists the financial support and computer time they needed. Solar storms that fling electrons and ions toward earth have long been acknowledged as an influence on the upper atmosphere. However, the new modeling shows an unexpected influence from the other direction: atmospheric waves that extend upward from the lower atmosphere. "There's much more structure than we'd been able to model before," reports Roble.

Elsewhere in this publication...


For more information, contact Milli Butterworth, butterwo@ucar.edu.
Prepared for the web by Jacque Marshall
Last revised: Mon Apr 10 13:23:27 MDT 2000