RELEASE NO: 98-46

NASA TO SUPPORT EL NINO PREDICTION STUDIES WITH ONE OF WORLD'S MOST POWERFUL SUPERCOMPUTERS

NASA's Seasonal to Interannual Prediction Project (NSIPP) will use an upgraded CRAY T3E-600 supercomputer at Goddard Space Flight Center, Greenbelt, Md., to support scientific and computational efforts to predict seasonal to interannual climate variations. The augmented CRAY T3E will enable NSIPP to run models capable of predicting phenomena such as El Nino and its associated atmospheric effects felt in many regions around the globe.

The NSIPP and a science team of investigators from universities and other institutions will have access to 512 new processors in the upgraded CRAY supercomputer at Goddard. The total system of 1,024 processors, 131 billion bytes of memory and 1.2 trillion bytes of online disk space will perform nearly 400 billion floating-point operations per second (400 gigaflops) on a standard benchmark, ranking it among the world's five most powerful supercomputers.

Goddard's CRAY T3E can do in one second what would take every person in the United States using hand-held calculators over 40 years to perform.

"We plan to enhance these computational capabilities in support of our Earth Science objectives and establish Goddard as the lead center for Earth science supercomputing internationally," said Dr. Ghassem Asrar, NASA Associate Administrator for Earth Science. "The challenge is to implement large-scale Earth system models, run them in a timely fashion and then transfer the technology to the operational agencies such as the National Oceanic and Atmospheric Administration."

NSIPP scientists will combine comprehensive satellite observations with global climate models. Since the ocean, with its large heat capacity, contains the memory of short-term climate variability, the project will build a new ocean data assimilation system to ingest the satellite data and provide the initial conditions for predictive model runs. Experimental predictions using past El Nino events for verification will assess the ability to forecast future events.

"The new technology will enable us to develop the best system—coupled climate models and data assimilation system—for taking full advantage of NASA's satellite observations for this problem," said Dr. Michele Rienecker, NSIPP's principal investigator. "We will be able to conduct ensembles of runs to give a realistic statistical characterization of uncertainty in the forecasts." NSIPP has developed a global general circulation model that couples models of the oceans, atmosphere, land surface and sea ice. The parallel model, capable of running on many computer processors, is a product of research funded by the Earth and Space Sciences Project of NASA's High Performance Computing and Communications Program.

"This system upgrade can be seen as an Agency commitment to scaleable parallel computing for operational supercomputing," said Lee Holcomb, NASA Chief Information Officer. "It culminates more than 20 years of NASA investment in parallel computing technology development."

Climate models divide the globe into a grid of layered columns, solving the relevant equations in each column layer and then assembling the full results. With 512 processors, NSIPP will be able to use a finer grid resolution than possible so far, with a column 1/2-degree wide (or 30 miles over the continental United States) in the atmosphere model, for example. "We know that model resolution impacts the ability to simulate the ocean as well as the atmosphere and land surface in a realistic manner," Rienecker said.

The CRAY T3E upgrade is occurring in two stages, with 384 processors installed in March and 128 processors scheduled for availability in May. A next-generation parallel supercomputer is planned for the year 2000.

Additional NSIPP information may be obtained on the World Wide Web at the URL: 
http://nsipp.gsfc.nasa.gov/

Earth and Space Sciences Project details are at the following URL:
http://esdcd.gsfc.nasa.gov/ESS/

-30-