When Robert Dickinson began working at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado in 1968, global climate models were crude and wildly inaccurate.
“The code I worked with covered Los Angeles with snow,” he laughs. “That was a problem with modeling how frost forms on the ground.”
He eventually concluded that the most primitive part of the climate models, the bit that needed the most improvement, was the description of the land surface and how it interacts with the rest of the climate system. In the models, the surface tended to be boiled down to either water or land. Scientists didn’t know how, or have the computing power, to accurately account for things like soil moisture, albedo (how well the surface reflects sunlight), topography and vegetation. He refers to one early model that oversimplified soil moisture as the “bucket model.”
Dickinson has spent more than three decades working to improve how the land surface is represented in the models. To recognize that work, in 1996, the American Meteorological Society awarded him the Rossby Award, the highest honor bestowed by the society on an atmospheric scientist. That same year, he also received the Roger Revelle Medal, the American Geophysical Union’s highest award for contributions to the science of climate dynamics and to predictions of expected climate changes. He has also been inducted into numerous prestigious scientific organizations, including the National Academy of Sciences, the National Academy of Engineering and the Chinese Academy of Sciences.
Because of his effort and that of thousands of other researchers, climate models are far more robust today. But, he says, much remains to be done.
From Yellow Jacket to Longhorn
Dickinson set off for Harvard as an undergraduate in the late 1950s to study English, not science. But he enjoyed the physics and other science classes he took.
“I gradually thought that you need to go into something where you could see job opportunities,” he says. So he switched to a physics and chemistry major. He never lost his love of explaining complex subjects simply and clearly. “My success as a scientist is due to strong writing skills,” he adds.
He completed a doctorate in meteorology at the Massachusetts Institute of Technology, where his research focused on the theoretical dynamics of global atmospheric circulation. At that time, the idea of using computers to model global climate was just getting off the ground in cutting edge research labs and the work hadn’t quite filtered down to the students. Edward Lorenz, a professor of meteorology at MIT created one of the first simple, global atmospheric circulation models around 1960. That work led to the observation that complex systems are extremely sensitive to initial conditions, sparking an entirely new field of study: chaos theory. The example he is best known for is the “butterfly effect,” in which a small disturbance such as the beating of a butterfly’s wings in Brazil might yield a big change in the system, such as the birth of a tornado in Texas.
“I took one class from him on statistics and enjoyed it,” says Dickinson. “But like many students there, I wasn’t aware of what he was doing that he would later become famous for.”
Dickinson went to work at NCAR, one of the world’s premier climate research organizations. Then in the early 1970s, he and colleague Steve Schneider were asked to outline possible new research initiatives. During the process, he became fascinated by the idea of using computers to model climate. Now, 35 years later, he’s still working on those models.
As a professor at the University of Arizona in Tucson in the 1990s, he worked with a bright young post doc named Liang Yang on improving the land surface components of climate models. He and Yang, now a research associate professor at the University of Texas at Austin’s Jackson School of Geosciences, have continued to collaborate on research and co-author papers.
Together, they have made many improvements to the land and atmosphere components of the Community Climate Model, an open source model initially created with other collaborators at the University of Arizona and now hosted at NCAR. Dickinson and Yang will be able to work more closely now that they’ll be down the hall from each other. Dickinson is leaving his most recent university home at Georgia Tech in Atlanta, home of the Yellow Jackets, to be a professor at the Jackson School.
“Liang is good at identifying interesting problems to work on,” Dickinson says. “One of my skills is in refining definitions of what the science is and in describing it in writing.”
He also joins Kerry Cook, another leading climate scientist who has recently moved to the Jackson School from Cornell, and others in expanding the school’s climate research and teaching capabilities. Dickinson and Cook have both been trustees of the University Corporation for Atmospheric Research (UCAR), a nonprofit consortium of research universities that manages NCAR. Both are strongly committed to training a new generation of climate scientists.
“It’s a good opportunity to share students,” says Dickinson. “When students have several faculty, they have more people to turn to with questions, so they get more rounded guidance than if they have just one.”
When Dickinson began working on climate models in the 1970s, they were simple affairs.
Consider how they handled soil moisture. In the models, all soil had the capacity to hold 15 centimeters of water. If there was more water, the ground became saturated and the excess water would runoff. If there was less, it would be absorbed and evaporation would be reduced. If there was exactly 15 centimeters of water, then the land surface acted pretty much the same as the surface of the oceans in terms of evaporation. Dickinson calls this the bucket model. It turns out soils are far more complex when you fully account for their capacity for water, rates of evaporation and transpiration by plants.
The earliest models didn’t incorporate vegetation at all. When a colleague introduced a plant canopy into a climate model in the early 1980s, Dickinson was inspired to consider the myriad impacts that vegetation could have on the water cycle, wind patterns and the amount of solar radiation reaching the ground. Figuring out how to represent the land surface—especially surface water and vegetation—in models became a major focus of his research.
One difference between vegetated and non-vegetated land is that once water is in a plant’s leaves, its harder to get it back into the atmosphere than it is to get it out of non-vegetated soil. Leaves don’t act like wet surfaces. It takes a pull to get the water out of the leaves and into the air (transpiration).
“It’s a detail,” says Dickinson. “It’s all about trying to get details right. When you ignore any one detail, you can get hugely wrong answers.”
The physical structure of vegetation—from the height and density of leaf cover of individuals to spacing between plants—matters. So does the color of soil, which determines how much sunlight it reflects or absorbs.
More recently, Dickinson has turned to looking at how vegetation affects the absorption and reflection of incoming solar radiation. Current models treat a tree canopy as a flat surface with a uniform albedo. But in reality, a forest has gaps and variations in foliage density.
To take an even more complicated case, imagine vegetation is covered with snow. If a tree is entirely blanketed, its albedo will be quite high. Now imagine the snow is blown off the tree—more sunlight can filter through the canopy, yet the ground is now more reflective. If the tree is at a high latitude, the sun will be at a low angle during winter, casting longer shadows and thereby changing the albedo of the ground.
“All previous models had taken land as the same everywhere, there was no geographical distinction,” he says. “So I have tried to define land in terms of details instead of saying there was one universal property of land.”
He says there are still many aspects of the real world that need improvement in the models, such as thunderstorms or the effects of deforestation on rainfall. He relishes the challenge because it gives him the chance to make an impact on society.
“It’s in the top 10 of issues people care about like global disease and hunger,” he says. “So being a part of trying to resolve these issues is pretty exciting.”
Earlier this year, Dickinson served on a committee appointed by the U.S. Department of Defense to identify grand challenges in climate change research. One conclusion was that climate modelers need to develop the ability to make climate predictions on a regional rather than global scale and on the decadal rather than century scale. These are the temporal and spatial scales that are most useful to decision makers. Global circulation models, the standard tool of climate modelers today simply don’t have high enough resolution to offer practical help to the people who need it.
“The Nation needs detailed information regarding the magnitude and timing of the climate changes and of the consequent impacts to which human societies will have to adapt,” the scientists wrote in an unpublished draft report. “Such information would provide the basis for assessing the desirability of different adaptation options.”
Decisions on how best to manage water resources for example depend on predictions of future precipitation, flooding and droughts. In the old days, before satellites and computer models, those predictions were based on historical observations.
“But that only works statistically if you have an unchanging system,” says Dickinson. “If the system is rapidly changing, that doesn’t work. You can no longer say the future is the same as the past.”
As an example of the potential use of climate predictions, Dickinson says that in the Eastern U.S., water storage systems are designed to hold enough water to ride out one year of drought. A multi-year drought would empty their reserves. If water managers knew that a multi-year drought was likely in the near future, they would build bigger reservoirs.
There are at least two challenges in making model predictions with finer spatial resolution. One is computational power. Global climate models currently divide the world up into 100 square kilometer grids.
“Details involving local climate would start getting better if you could get down to a 10 square kilometer grid,” says Dickinson. “That doesn’t seem like much, just an order of magnitude, but the increase in computational power is really a factor of 1,000.”
The other challenge is to accurately model how the land interacts with the climate. On the small scale, the land surface—from pavement to forests to grasses to deserts to glaciers—can have a big impact.
The Department of Energy is now considering how it can redirect research efforts to implement the report’s recommendations.
Apart from the technical challenges of developing more useful climate models, Dickinson says there needs to be improvement on the human side of the equation. Because the climate system is so complicated, he says, scientists have developed many different ways to describe it in models. He sees this go-it-alone attitude as detrimental to the climate community.
“People have different descriptions and they each come up with their own answers,” he says. “They’re difficult to judge by anyone outside that particular research group.”
He says part of the problem is that scientists are trained to follow their own interests and not worry so much about how it might be useful for society. In his opinion, climate modelers should be trained more like engineers, to solve very specific problems. And they need to have more of what he calls social organization.
“To make climate models work, you have to have some agreement of ‘I do this and you do that and we’ll talk together and figure out how to do it together’,” he says.
For more information about the Jackson School contact J.B. Bird at email@example.com, 512-232-9623.