The project, awarded jointly to ORNL and Los Alamos National Laboratory, will allow the labs to develop a climate model that will provide a scientific basis for evaluating policy alternatives. "It's difficult to assess the effects of policies if we do not have accurate models," said John Drake of ORNL's Computer Science and Mathematics Division. "This funding will allow us to develop tools to provide accurate information to policymakers looking at the potential effects of increased greenhouse gases."
A climate model solves the complex physical equations that balance atmospheric and ocean flows with incoming solar radiation and weather events. The project focuses on model development and the software design and engineering required for the climate research community to develop and maintain high-quality climate prediction capability that effectively uses high-performance computers.
High-performance computers that have thousands of processors acting simultaneously require scalable software to take advantage of unprecedented computer power. The focus of the DOE effort is on coupling the atmosphere, ocean and sea-ice sub-models in a way that efficiently uses the supercomputers at ORNL and other national laboratories.
"We have two primary goals," Drake said. "First, to restructure and redesign the climate model components with a view to enhancing performance on a range of computing platforms, and, second, to extend the model for more realistic climate simulations by including interactive carbon cycles."
ORNL's Center for Computational Sciences is one of two DOE high-performance computing research centers established in 1992 as part of the federal high-performance computing and communications initiative. Goals of the initiative include utilizing and moving to maturity state-of-the-art computing systems, using high-performance computing to help solve "grand challenges" and increasing the use of high-performance computing in the industrial sector.
ORNL's IBM and Compaq supercomputers boast more than 1.5 teraflops (1.5 trillion calculations per second) computing power.
Others involved in the project are Argonne National Laboratory, Lawrence Livermore National Laboratory, Pacific Northwest National Laboratory, the National Center for Atmospheric Research and NASA.
Funding for this work was made available through DOE's Scientific Discovery through Advanced Computing initiative. ORNL is a DOE multiprogram facility operated by UT-Battelle.