NASA Uses Recovery Act Funding to Upgrade Climate Modeling Computer


SAN FRANCISCO — NASA Goddard Space Flight Center’s climate modeling team is adding powerful computer processors and developing new tools to strengthen its climate simulation capabilities and improve data dissemination with financial assistance from U.S. President Barack Obama’s economic stimulus package.

That package, known as the American Recovery and Reinvestment Act, included $1 billion for NASA. A portion of that funding earmarked for improving the U.S. space agency’s supercomputing capabilities is helping to pay for sophisticated processors built by Intel Corp. to improve Discover, the IBM computer that serves as the workhorse in Goddard’s extensive climate simulation campaign. Discover also hosts NASA’s modeling work for various projects including the space agency’s contribution to the Intergovernmental Panel on Climate Change (IPCC), a scientific organization established by the United Nations.

Scientists and engineers rely on Discover to help them process the massive amounts of data obtained by NASA’s space-based sensors. The new Intel Nehalem processors are twice as efficient in handling climate and weather codes as the processors Discover used previously to crunch that data. “So we can either get twice as much work done, or do the same work twice as fast,” said Phil Webster, chief of the Computational and Information Sciences and Technology Office (CISTO) at Goddard in Greenbelt, Md.

That improved speed and performance is particularly important in climate simulation because the computer models require so much computation. Climate simulation is based on a set of physical variables, such as temperature, moisture and wind velocity, measured at data points on a grid around the Earth. Adding points to that grid improves the accuracy of climate models but it also increases the demand for processing power.

With their speed, the new processors allow scientists to enrich the grid and conduct high-resolution simulations that include climate features such as clouds that they were unable to include in previous models. “Things like cloud formation happen on the order of 2 kilometers and right now, the best simulations we can do have about 8-kilometer resolution,” Webster said. “Each time you increase the resolution by a factor of two, it increases the computing by about a factor of 10. … It is an incredibly complex system that we are trying to model and use to make accurate predictions.”

Not only will the improved simulations help scientists understand the long-term impact of global climate change, they will also improve the daily forecasts Goddard’s Global Modeling and Assimilation office runs in support of NASA satellite instrument teams. Those daily forecasts are based on the Goddard Earth Observing System model, which currently offers 27-kilometer resolution. Additional processing power is offering resolutions in ongoing testing as high as 3.5-kilometer resolution.

CISTO officials began working with the Nehalem processors in August when the group added the first batch of 4,128. The second batch, another 4,128 processors being purchased with Recovery Act funds, is scheduled for delivery sometime in November or early December, Webster said.

Recovery Act money is also allowing CISTO to accelerate the distribution of climate simulation data by integrating the NASA Center for Computational Sciences, Goddard’s enormous supercomputing facility, into the Earth System Grid, an effort led by the Department of Energy’s Scientific Discovery Through Advanced Computing to give climate scientists greater access to simulation and observation data.

“The goal of the Earth System Grid is to let scientists focus on the science instead of focusing on finding the data they need and worrying about data formats,” said Dean Williams, Earth System Grid principal investigator and software project leader at the Program for Climate Model Diagnosis and Intercomparison at the Lawrence Livermore National Laboratory in Livermore, Calif. “If scientists want the monthly mean data showing the difference of sea surface temperature in January, they can find it.”

To make it easy to search for that data, however, the information must be offered in a form that all scientists can understand. Another challenge is managing the vast data archive, Williams said. The Earth System Grid offers researchers access through two web portals, one for general climate data and another for IPCC. The Earth System Grid will host and disseminate a multimodel archive to be used for the IPCC’s fifth assessment report, a project started in July and scheduled to conclude in 2013 to chronicle worldwide scientific research on climate.

Currently, the Earth System Grid’s general portal does not include data derived from space-based sensors. In an ongoing project scheduled to be completed in 2010, Williams’ team plans to incorporate data from various instruments — including the Atmospheric Infrared Sounder on NASA’s Aqua satellite and the Microwave Limb Sounder carried by Aura — into the archive to make it available to researchers. “We want to make NASA data readily available to climate scientists in a form that they can understand,” Williams said.