Objective

Provide sufficient computational resources to the NASA Seasonal to Interannual Prediction Project (NSIPP), commensurate with its computing requirements, and thereby enable the project to achieve its objective:

To demonstrate the utility of satellite observations in predictions of short-term climate variations, and to establish the cost-effective blend of remote surface observations and subsurface data necessary for a seasonal-to-interannual climate prediction capability.

ESS Accomplishments

• Procured and installed a 512 processing element (PE) and disk upgrade to the HPCC/ESS SGI CRAY T3E system at GSFC in March 1998 through funding provided by NASA's Office of Earth Science.
• Partitioned the system (via system software) to restrict the additional 512 PE's to the NSIPP Project.
• Worked with NSIPP Project members to configure all aspects of the NSIPP partition to enable optimal utilization and throughput.

NSIPP Accomplishments

• Ran the global ocean model at 2/3 by 1 degree resolution, achieving 2 gigaFLOPS performance on 128 PE's before optimization.
• Ran a 3000-year simulation of a coupled atmosphere-land surface model at 2 by 2.5 degree resolution.
• Ran the global atmosphere model at .5 by .5 degree resolution on 512 PE's and achieved 20 gigaFLOPS performance.
• Ran an ensemble Kalman filter assimilation on 128 PE's.

ESS Approach

• Demonstrate viability of the CRAY T3E platform for the NSIPP Project via a Guest Investigator allocation on the HPCC T3E.
• Procure a cost-effective (better than 16 peak megaFLOPS per dollar) upgrade of 512 CRAY T3E-600 PE's, which had been traded in by other SGI customers who had upgraded to faster processors.
• Work in partnership with NSIPP to configure and maintain the full CRAY T3E system to meet the needs of both ESS and NSIPP users.

NSIPP Approach

• Parallelize, optimize, and validate output of the global ocean model and coupled atmosphere-land surface Model independently on the CRAY T3E.
• Couple the models to form the Coupled General Circulation Model (CGCM) and integrate an ocean data assimilation system to initialize the coupled system.
• Run simulations in production mode, leading to prediction runs in production mode.

Significance

From a scientific standpoint, the CRAY T3E upgrade will enable NASA (through NSIPP) to make a significant, competitive contribution to national and international leadership in short-term climate prediction.

From a technology standpoint, this upgrade marks the first time that an HPCC resource was deemed critical to the success of a core NASA science mission such that additional NASA funding was identified for a procurement. This is a key example of the transition of HPCC technology into the general NASA science community, a major goal of the HPCC program.

Status/Plans

The NASA Office of Earth Science issued a NASA Research Announcement, NRA-98-OES-07, on July 1, 1998 entitled, "Investigations to Support the NASA Seasonal-to-Interannual Prediction Project," which will open the NSIPP portion of the CRAY T3E to the broader research community. The closing date of the NRA is September 24, 1998.

NSIPP will run a full coupled ocean/atmosphere model in production mode for several decades by October 1998. NSIPP will run coupled model forecasts in production mode using ensemble techniques for ocean data assimilation by October 1999.

For more information on the NSIPP Project, see URL:
http://nsipp.gsfc.nasa.gov

Points of Contact

Lisa Hamet Bernard
SGI/CRAY T3E Manager
NASA Goddard Space Flight Center
Lisa.Bernard@gsfc.nasa.gov
301-286-9417

Michele Rienecker
Principal Investigator, NSIPP
NASA Goddard Space Flight Center
Michele.Rienecker@gsfc.nasa.gov
301-614-5642

Max Suarez
Principal Investigator, NSIPP
NASA Goddard Space Flight Center
suarez@janus.gsfc.nasa.gov
301-286-7373