National Center for Computational Sciences

Scientists simulate climate response to volcanic gas emissions to test model’s accuracy

The 1991 eruption of Mount Pinatubo in the Philippines spewed 20 million tons of sulfur dioxide gas into the atmosphere, the bulk of it in only 9 hours. The sulfur dioxide circulated around the globe in about 2 weeks, interacting with oxygen molecules along the way to become a sulfate aerosol that would remain in the atmosphere for several years.

Scientists at Oak Ridge National Laboratory and the National Center for Atmospheric Research (NCAR) are using the Cray XT5 supercomputer, Jaguar, at the National Center for Computational Sciences (NCCS) to simulate how the climate system reacts to the atmospheric increase in aerosols from volcanic eruptions.

Eruption aerosols circulate higher in the atmosphere than gases released by power production and other human activities. Solar radiation bounces off the aerosol particles, directing some of the radiation away from the earth and causing short-term cooling. In contrast, anthropogenic, or human-produced, gases in the lower atmosphere trap solar radiation, warming the planet’s surface over longer periods of time. Volcanic emissions can mask the long-term warming caused by human activity.

Using Jaguar, the fastest supercomputer for open science research, the team will be able to follow the impact of aerosols on smaller scales than was possible in the past. It will monitor timescales in decades, not centuries, and space scales in geographical regions, not the entire globe.

“For the first time we’ll actually be able to include more detailed information about changes in atmospheric aerosols and run the atmospheric model at a 40-kilometer resolution to assess the impact,” said Kate Evans, principal investigator on this project to predict climate change on decadal timescales. The project began running on Jaguar in January as one of 21 petascale early science projects, or large-scale endeavors given early access to use most of the supercomputer’s processing cores.

The sudden influx of aerosols into the atmosphere following a volcanic eruption makes a bold statement relative to other factors that influence climate.

“It’s analogous to plucking a guitar string against a background of silence,” Evans said. “The volcano is perturbing the system on a new frequency.”

Aerosols from volcanic eruptions typically clear away within 2 to 3 years as gravity pulls the aerosol particles back down. Conversely, anthropogenic emissions are ongoing and may alter climate long into the future.


Although several types of gases, including carbon dioxide, are emitted during a volcanic eruption, the release of sulfur dioxide and the resulting sulfate aerosol are especially important because the sulfate particles scatter incoming solar radiation. With less energy reaching the ground, the average global temperature can decrease up to a degree Fahrenheit. This short-term cooling contrasts with long-term global warming, predicted to be 3 to 10 degrees Fahrenheit over the next century.

Global models cannot yet predict climate change based on escalating greenhouse gases for regional scales. Scientists confidently use current climate models to predict climate change as a global average, but they are still exploring whether the models can accurately predict the unique impacts of climate change from region to region. Where one region may experience drought, another may experience rising water levels, depending on many interconnected variables of climate change.

Looking back to leap forward

Evans and her team are using the NCCS supercomputer to simulate more than 30 years of global climate, from 1978 to the present. The team includes five other researchers at ORNL—James Hack, leader of ORNL’s Climate Change Initiative; John Drake, leader of ORNL’s contribution to the Climate Science Computational End Station, which will play a role in the upcoming Intergovernmental Panel on Climate Change’s fifth assessment report; James Rosinski, a computational scientist who designed the model runs; Pat Worley, a computer scientist who will help the team scale the climate modeling codes to make the best use of the supercomputer; and Lianhong Gu, a climate scientist in the Environmental Sciences Division who will study atmosphere interactions with land and ecosystems. The team also includes three researchers at NCAR—climate scientist Julie Caron, computational scientist John Truesdale, and atmospheric chemist Jean-Francois Lamarque.

Simulating global climate requires enormous processing power to calculate temperature, pressure, and other variables for millions of points on the globe every 150 seconds for more than a quarter century. The Jaguar XT5’s 1.4-petaflop processing speed allows more than a quadrillion calculations per second, enabling the project’s researchers to achieve the planned level of detail.

The Community Climate System Model considers four key components of the earth’s climate system: atmosphere, ocean, land surface, and ice. Using a version of the Community Atmosphere Model computational code developed at NCAR and run on Jaguar, Evan’s team is configuring the atmosphere and land components to run at a resolution ten to 40 times the spatial degrees of freedom typical of most global climate models.

The team must test the model for accuracy at the finer resolution by running it through a period of recent history. Observational data from weather balloons and satellites are compared to data generated by the model. When observations can confirm the results of computer modeling, scientists can better determine the model’s ability to predict climate response.

The project is slated to use more than 8,000 of the supercomputer’s processors for a single run, Rosinski said. Four times that number may be required as multiple runs are carried out. The goal of running the model several times is to build a collection of datasets of slightly different results that represent climate variability. Once averaged, the collection of data will yield a more accurate prediction.

“If the model has the ability to predict the system response to changes in the aerosols,” Hack said, ”then we will have increased our confidence in being able to detect the response to longer-term anthropogenic changes.”

— by Katie Freeman

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