Down-to-earth models:
Bringing ground cover
into climate simulations

Some of the hardest pieces to place in our planet's jigsaw of climate include forests, crops, and pavement. A new tool for studying land-atmosphere exchange is bolstering NCAR's flagship climate model and providing fresh views of the global atmosphere we'll experience in the century to come.

When Gordon Bonan steps into the pine-studded foothills behind NCAR’s Mesa Lab, he sees a view unlike the one cultivated by his graduate training. "As an ecologist, you’re trained to think that each part of the landscape is unique and different," says Bonan, who earned his doctorate in environmental sciences at the University of Virginia. "You’re taught to characterize the differences and not the unifying elements."

Having leapt from ecology into climate science years ago, Bonan is now reaching back across that disciplinary gulf. He’s one of a handful of researchers working on a new view of how plants and the atmosphere mesh and how this intricate network can be translated into future-mapping software. In a new Community Land Model (CLM), forged under Bonan’s leadership, the interplay between landforms and the atmosphere is being depicted more realistically than ever before.

Along with any backyard gardener, "ecologists have known for a long time that climate affects where plants grow and how well they do," notes Bonan. By the same token, "climate modelers knew that vegetation is an important determinant of climate." Yet the two elements stayed separate for years. In the mid-1990s, NCAR launched an ambitious effort to build a new, multifaceted global climate model for use by university and NCAR researchers.

Most parts of the new Community Climate System Model, or CCSM—atmosphere, ocean, and the like—were adapted from existing components. However, if this nuanced approach was to succeed, it had to include a far richer set of land-air relationships than modelers had used before. So CLM was built essentially from scratch.

Daunting as this was, it gave Bonan (above) and his colleagues a rare opportunity: to create a model in which plants blossomed, shifted territory, and exchanged gases with the atmosphere. The first version of CLM debuted in January 2002 as part of the second version of CCSM. This new land model wasn’t perfect—its soils were a bit too dry, for instance, and the dynamic vegetation hadn’t yet been added—but the first iteration looked and behaved much as its creators had hoped.

As with other parts of CCSM, hundreds of scientists worldwide provided feedback on CLM through a working group. About a dozen of these experts worked especially closely with NCAR’s terrestrial sciences group, "getting their hands dirty in active development," says Bonan. Much of this work took place on an upgraded set of IBM supercomputers based at NCAR (see page 4).

Satellite data from NASA and NOAA provided a crosscheck for present-day simulations that helped build confidence in the model’s depictions of past and future. Paul Houser (NASA Goddard Space Flight Center) worked closely with Bonan to steer the model’s creation with satellite-based tools in mind. For example, the model simulates the two uppermost layers of topsoil at 2 and 5 centimeters thick (about 0.8 and 2.0 inches). This corresponds to the depths that satellite-based microwave sensors can best profile. "Many models have just one soil moisture [index] for the entire root zone," says Houser. "It’s very hard to extract temperature and moisture for the top few centimeters from that."

A living planet in cyberspace

Not even the world’s most potent supercomputers can recreate every tree on Earth. In order to make CLM practical, the land-modeling team broke the planet’s surface into rectangles that correspond to CCSM’s atmospheric grid—2.8 degrees on each side, or about 240 by 320 kilometers (150x200 miles) for the midlatitudes. Inside each square, the resident plants and landforms are identified by percentages rather than by actual locations. This freedom- within-constraint technique saves valuable computer time, while retaining the detail needed for a clear-eyed picture of land surface that can evolve with climate.

"If the climate warms up, you’ll see tundra convert to forest in the model," says Bonan. "If it’s too dry, you’ll see forest converting to grassland."

The dynamic elements in this new model will go far beyond vegetation. Already, CLM follows rain and snow flowing through rivers into the ocean, and the model is incorporating plant-based emissions, such as volatile organic compounds, that affect air chemistry. Future versions will track great clouds of dust, kicked up into the atmosphere across Asia and Africa, that can shape cloud formation and temperature markedly over the short term (and shift climate in the long term, should their frequency change).

Perhaps the most eagerly sought feature of the land model is its treatment of carbon. The building block of life itself, carbon—in the form of carbon dioxide that helps warm our climate as it accumulates in the air—is casting question marks at the future of our environment. An upcoming version of CLM will allow plants to add and remove carbon in amounts that vary over space and time, much as they do in the real world. This should help provide a clearer reading of how much our atmosphere might warm in the years to come.

The indirect effect of snow cover on carbon dioxide is one of the many processes that show up in new detail in CLM. "The temperature of the land surface underneath snow was coming out way too cold in previous models," says Robert Dickinson (Georgia Institute of Technology), who developed his own land model while at NCAR in the 1980s. His was one of the first such models to trace the effects of vegetation on global climate. What it didn’t do was portray the effect of snow on dead vegetation below it. The unrealistically cold surface prevented plants from decaying. In reality, the land surface under a blanket of snow can hover near freezing, allowing plants to decay and emit carbon dioxide. CLM includes this insulation effect; it also allows for true-to-life patches of snow cover within grid boxes, rather than a too-uniform layer.

On the carbon trail

The coevolution of climate and the carbon cycle is being studied using CCSM as part of a grand sequence of experiments dubbed the Flying Leap. Chaired by Inez Fung (University of California, Berkeley) and Scott Doney (NCAR and Woods Hole Oceanographic Institution), the project’s goal is to jump into new territory for climate simulations. In a series of benchmark studies, future levels of carbon dioxide will be calculated by including projected fossil-fuel emissions as well as the cycling and storing of carbon from land and ocean sources. The idea is to investigate whether greenhouse warming may be accelerated by a warming-induced release of terrestrial and oceanic carbon. The Leap will contribute to an intercomparison project—cosponsored by the World Climate Research and International Geosphere-Biosphere Programmes—designed to help scientists understand and predict climate change.

Roughly 30% of the carbon leaving the atmosphere each year remains unaccounted for. Midlatitude forests from Asia to North America to Europe probably absorb much of this, but observations don’t have enough regional clarity to prove the case. A volley of conflicting studies has led to confusion over how much carbon dioxide might be absorbed by various nations’ forests, and in turn, how much credit those countries might get for their forests within global CO2 management schemes such as the Kyoto Protocol. Large sections of present-day forest didn’t even exist 200 years ago. Much as the tropics are now being deforested, the expansionist cultures of Europe and the embryonic United States chopped down trees with gusto in the 1700s. Forests have since made a comeback across large stretches of the eastern United States. Even so, "a significant part of the world has been converted from natural vegetation to cropland," notes Bonan. "Changes in land cover over the next century could be as important as greenhouse gases in determining climate."

In 2001, as part of its third major assessment since 1988, the Intergovernmental Panel on Climate Change quantified the impact of land cover change on global warming for the first time. However, Bonan notes that "there are huge error bars," a sign of how much remains to be learned.

Fresh ideas from young scientists

An infusion of recently hired scientists at the start of their careers has boosted NCAR’s ability to study carbon cycling and other land-air processes. Peter Thornton is working to quantify the interactions of carbon levels in the land, atmosphere, and ocean and how enhanced levels of CO2 might affect vegetation. Natalie Mahowold is examining large-scale dust sources and their impact on biogeochemistry and climate.

Another new arrival, Britt Stephens, and fellow NCAR scientist David Schimel cochaired an innovative workshop in 2002 on global CO2 data collection and use in models. Using virtual instruments and atmospheres and an imaginary budget, participants carried out a CO2 competition on NCAR‘s powerful supercomputers. The goal was to see which group’s instrument network could produce the most benefit in tracking regional sources and sinks of CO2.

The energy from NCAR’s new arrivals, and the resulting ability to interact with university colleagues more extensively, has made a difference, says Bonan. "Our model development has really accelerated over the past three years." CLM, he adds, is "taking on a vitality of its own. It’s no longer a series of subroutines within a climate model—it’s a model in itself."



A framework for better climate models

Models like the Community Climate System Model that blend atmosphere, land, ocean, and sea ice will soon be able to draw on a shared framework. NCAR is a key partner in the Earth System Modeling Framework (, a $10 million, three-year project sponsored by NASA. It will allow some of the most widely used climate and weather models and data assimilation systems to work together, taking advantage of vast amounts of observational data and stimulating cross-disciplinary interactions.

Software engineers at NCAR are building a core infrastructure with tools for cross-component communication, time management, performance profiling, and other common functions. "An application running on the framework will resemble a sandwich," says NCAR’s Cecelia DeLuca, one of the partnership’s three technical managers. The bottom slice of bread is the infrastructure, with utilities and data structures that allow developers to build applications more easily. The top slice is the superstructure for coupling that allows model components to work together. In between is the filling: software written for specific applications. The framework will make it much simpler to compare approaches from different sources.

"This is the first step in a progression that will enable the modeling community to use its resources more effectively," says DeLuca.

Building cities in reverse

From a home base surrounded by the corn and wheat fields of the U.S. heartland, Johannes Feddema is trying to pin down how cities affect our climate. Feddema (University of Kansas), an NCAR visitor in 2002, is working with Gordon Bonan on a new portion of CCSM that simulates city landscapes.


Climate assessments take pains to remove the heat-island effect that warms cities beyond levels observed in the surrounding rural area. But even if cities are a drop in the bucket of global climate, they’re where millions of people live, so the urban impact on perceived climate—the way people experience the air—is disproportionate. That influence is getting bigger, as suburbs continue to sprawl and heat-absorbing pavement covers the land.

Feddema is mapping where cityscapes have grown since the Industrial Revolution. He’s using a "backward model" that takes away people and buildings as one moves back in time. Lessons learned in linking the built environment to climate will then be extended into the future, using estimates of future population growth. The work is part of a larger study examining soil degradation, urbanization, and other impacts on land cover and regional climate.

Will cotton be king in the 21st century?

Like lenses switched by an optometrist during an eye exam, the sharpening of resolution in a climate model might bring newfound clarity—or it might simply make things fuzzy in a different way.

In a multiagency study, NCAR’s Linda Mearns and colleagues set out to test the value of such improvements by seeing how increased spatial resolution affected the output of models that linked climate to agriculture and economic output. The team focused on the agriculture-rich U.S. Southeast, tracking the evolution of crop yields while bringing carbon dioxide levels in the atmosphere to double their present-day values.

At higher resolution, the crop projections tended to become gloomier. Some plants thrived in a warmer climate, though. "Cotton was the winning crop," says Mearns. "It was the only one that did better in all of the model configurations." Soybeans suffered the most: unable to shift to an earlier sprouting cycle, they wilted in the longer, hotter summers. A separate set of model runs tested whether farmers could abate the impacts through crop management. Adaptation does make a difference, says Mearns— "assuming that farmers completely understand the climate changes and respond immediately to them."

Photo: National Cotton Council

On the Web

NCAR Community Climate System Model
NASA Land Data Assimilation Scheme


UCAR > Communications > Highlights > 2002 Search
Highlights 2002