Geodesic Climate Model Uses Different Mapping Technique, Coordinates And Supercomputing To Improve Predictions

Working with oceanographers and mathematicians from other institutions, the Colorado State team has a cooperative agreement with the Department of Energy for $4.5 million to build the mathematical simulation over the next five years.

"This will be the first geodesic model that couples atmosphere and ocean -- in fact, the first geodesic model to deal with the ocean at all -- and it's also the first model to use 'hybrid' coordinates for both atmosphere and ocean," said David Randall, professor of atmospheric science and principal investigator. "It's a fairly radical step."

Because of the project's size and complexity, Colorado State scientists will coordinate with researchers at other universities and research facilities. Collaborators include Wayne Schubert of Colorado State; Akio Arakawa of UCLA, who developed one of the two "hybrid" coordinates used; Albert Semtner Jr. of the Naval Postgraduate School; Scott Fulton of Clarkson University; and John Baumgardner and Phillip Jones, both with the Los Alamos National Laboratory. David Bader is the project supervisor for the DOE.

"One of the problems with modeling the atmosphere and ocean is that the earth is round," Randall said. "A traditional way of locating a point on the earth is to use latitude and longitude to tell where you are. That causes problems at the poles, because meridians of longitude run together.

"The converging meridians make it necessary to use small intervals of time that make a model run very slowly. Tricks can be used to avoid this problem, but they cause problems of their own."

In the early 1990s Ross Heikes, now a doctoral candidate working with Randall, wrote his master's thesis on a simple "shallow water" model of a spinning globe covered with a thin layer of water. Heikes' innovation was to use a geodesic grid to map coordinates. Todd Ringler, a postdoctoral researcher also working with Randall, turned Heikes' effort into a full model incorporating the three-dimensional atmosphere as well as land masses.

An advantage of the new model, Randall said, "is that all of the grid cells are very nearly the same size. The poles are just like everywhere else in terms of the grid."

A major component of the DOE project is that the ocean and sea ice will now be represented on a geodesic grid like that used to simulate the atmosphere, although the ocean grid will be finer, Randall said. Coupling of the atmosphere and ocean models is simplified because they use similar grids.

In addition to the geodesic grids, the model will include other innovations. One perennial modeling problem is how to represent the vertical structures of the atmosphere or ocean. The new model will use "hybrid" vertical coordinates that drift up and down with the air (and water) as it moves. This approach eliminates problems caused as air moves up or down across the edges of layers in the model.

Another requirement of the DOE cooperative agreement is to design the model to take advantage of the latest supercomputers at DOE labs across the country.

Randall estimates that simulating a century's worth of climate with the full-coupled ocean-atmosphere model will require several weeks of round-the-clock supercomputer time -- "but I hope that someday we'll be able to do it over a weekend."

Ringler notes that the model would be "very well suited to weather forecasting," but the project members plan to use the model to study climate. Randall and colleagues at Colorado State have been working on global atmospheric modeling for more than a decade. Their work is also supported by the National Science Foundation and NASA. Randall expects that the DOE effort will result in a much-improved model with unprecedented capabilities.