National Center for Computational Sciences

Climate scientists look to the past to prepare for the future

Weather is not always kind. Tropical storms, blizzards, and other weather “mood swings” are some of the more temperamental traits of a climate’s personality.

Researchers at the Oak Ridge National Laboratory Leadership Computing Facility are using the world’s fastest supercomputer for open scientific research to recreate the last century’s global climate and examine some of its more destructive weather tantrums. By understanding the past, scientists may be able to more accurately predict future extreme weather under climate change.

“We can represent the weather for every extreme event in the historical record for which we have observations,” said Gilbert Compo of the University of Colorado, principal investigator of the Surface Input Reanalysis for Climate Applications (SIRCA) project.

To confidently project weather extremes into the twenty-first century, it is necessary to make sure daily data from the nineteenth and twentieth centuries are accurately reflected, Compo explained.

The SIRCA project is conducted under the auspices of the Cooperative Institution for Research in the Environmental Sciences Climate Diagnostics Center of the University of Colorado and the National Oceanic and Atmospheric Administration Earth System Research Laboratory. As part of the project, researchers are reconstructing global weather conditions in six-hour intervals from 1850 to the present. An international initiative called Atmospheric Circulation Reconstructions over the Earth facilitates collection of more observations for the project and their applications.

Compo and his team expect the results of the SIRCA project will help scientists better understand extreme weather throughout recent history. “The datasets should be good for looking at extreme events like colder outbreaks and heat waves,” Compo said.

Some major climate events of the nineteenth and twentieth centuries had tragic consequences, and scientists hope to better anticipate future trends with the help of historical simulations.

Three million Americans fled from the Great Plains during the 1930s Dust Bowl, when dust storms rolling over the plains left half a million people homeless. Dust clouds picked up the dry soil of fields left barren by drought. The dust was thick enough to clog lungs and cause a condition known as dust pneumonia that killed hundreds.

Another weather tantrum was the Children’s Blizzard in January 1888, which killed 200 people on the Great Plains who were exposed to the subzero temperatures and blasts of snow. Half of the victims were children trying to get home from school during the blizzard.

Compo also expects to follow the warming trend seen in the Arctic from the 1920s to the 1940s. By effectively tracking droughts, storms, and other highs and lows of the past, he thinks improvements can be made to current climate models, which are expected to predict similar extremes in the future.

The climate model for the SIRCA project will be run on the Department of Energy’s Jaguar supercomputer at ORNL. Initial datasets from SIRCA’s predecessor, the 20th Century Reanalysis Project, are reconstructing global climate and weather from the 1870s to the present. Scientists expect that project to be completed this fall. SIRCA is projected to extend the team’s reconstructions from the 1870s back to the 1850s at a spatial resolution two times higher than the Century Reanalysis by fall 2012.

The SIRCA and Century Reanalysis projects are the first of a three-phase reanalysis undertaking enabled by the Innovative and Novel Computational Impact on Theory and Experiment program. The task will map weather conditions in the troposphere and stratosphere, the lowest portions of the Earth’s atmosphere.

It takes a lot of computer power to reconstruct a century and a half of weather. “To generate the 1850 to 2011 dataset, we expect to need 60 million [processing] hours,” Compo said. “Without the supercomputer we couldn’t do [the SIRCA study] at all.”

Not only will researchers make forecasts for every six-hour period, requiring the use of hundreds of processors, but they will also make 56 forecasts for each period.

“Think of a newspaper weather map with the highs and lows for the day,” Compo said. “It’s like having 56 maps.”

The SIRCA “weather maps” discern details of the Earth’s surface as close as 100 kilometers apart. Compo and his team plan to zoom in on hurricanes, severe storms, and floods. For these extreme events, the computer will generate maps showing details 60 kilometers apart. Zooming on these details will require 100 times the computing power of the other weather maps.

An algorithm called the Ensemble Kalman Filter pulls this collection of datasets—or ensemble—into a single reconstruction that is closest to the actual weather conditions. The computer algorithm recasts the observational data into three-dimensional datasets of past weather conditions.

“In traditional numerical weather predictions, you combine the observations you have with a single numerical model integration,” Compo said.

What comes first, observational data or the algorithm used to generate new forecasts?

“It’s a chicken and the egg issue,” Compo said. “We combine the forecast with observations to change the forecast to be more consistent.”

Consistency is key in climate modeling. Predictions of weather extremes require reliable computer modeling simulations, which help ensure future climate mood swings won’t take us by surprise.

— by Katie Freeman
Katie Freeman is a science writing intern for the National Center for Computational Sciences.