Since 1987 - Covering the Fastest Computers in the World and the People Who Run Them

October 7, 2005

Field Tests Unite Weather and Climate Models

Nicole Hemsoth

Researchers from NASA's Goddard Space Flight Center and seven other government and academic institutions have created four new supercomputer simulations that for the first time combine their mathematical computer models of the atmosphere, ocean, land surface and sea ice. These simulations are the first field tests of the new  Earth System Modeling Framework (ESMF), a shared software infrastructure that promises to accelerate research aimed at improving U.S. predictive capability ranging from short-term weather  forecasts to century-long climate change projections and more rapidly move the results of that research into operational prediction systems.

Under a partnership, groups from NASA, the National Science Foundation, the National Oceanic and Atmospheric Administration, the Department of Energy, the Department of Defense and research universities are using ESMF as the standard for coupling their weather and climate models to achieve a realistic representation of the Earth as a system of interacting parts. Having a standard will unify much of the modeling community. ESMF makes it easier to share and compare alternative scientific approaches from multiple sources, uses remote sensing data more efficiently, and eliminates the need for individual agencies to develop their own coupling software.

“The development of large Earth system applications often spans  initiatives, institutions, and agencies and involves the geoscience, physics, mathematics, and computer science communities. With ESMF,  these diverse groups can leverage common software to simplify model  development,” said NASA ESMF principal investigator Arlindo da Silva, a scientist in GSFC's Global Modeling and Assimilation Office.

NASA's Earth-Sun System Technology Office/Computational Technologies (ESTO/CT) Project funds the field tests and overall ESMF development. The partners on the field tests are DOE's Los Alamos National Laboratory, the Massachusetts Institute of Technology, NASA's Jet Propulsion Laboratory, NOAA's Geophysical Fluid Dynamics Laboratory and National Centers for Environmental Prediction, NSF's National Center for Atmospheric Research and the University of California, Los Angeles.

The newly completed field tests, known as interoperability  experiments, show that this new approach in coupling models works as envisioned. For instance, ESMF enables an NSF-NASA atmosphere model  to ingest conventional and satellite observations from NCEP's data analysis system. The coupling then produces global temperature and  wind outputs similar to those from NCEP's operational coupled models.  Although most of the experiments would require exhaustive tuning and validation to be scientifically sound, they already show that ESMF can be used to assemble coupled applications quickly, easily, and  with technical accuracy.

“These interoperability experiments illustrate the role ESMF can play  in integrating the national Earth science resources,” da Silva said. ”Using existing data assimilation technology from NCEP, the finite-volume Community Atmosphere Model, or fvCAM, was able to ingest  conventional and satellite observations, a capability that could open the door to using the fvCAM for weather as well as climate prediction.” The fvCAM, which includes land surface capabilities, was developed by NCAR, with key components from GSFC.

The second experiment again uses NCEP's data assimilation technology, but this time couples it with the Aries atmosphere model originally developed by the NASA Seasonal-to-Interannual Prediction Project. Aries is typically also coupled with an ocean model to run  experimental forecasts of phenomena such as El Nino and its effects  on precipitation. Because they both use ESMF, these two  interoperability experiments enable the intercomparison of systems  for satellite data assimilation.

The third experiment, combining a GFDL atmosphere-land-ice model with an MIT ocean-sea ice model (known as MITgcm), may ultimately bring new insights into ocean uptake of carbon dioxide and other important atmospheric gases and how this process affects the climate.

In an early independent adoption of ESMF technology, UCLA researchers have successfully coupled their Atmospheric General Circulation Model  to the MITgcm for the first time and inserted ESMF into an existing  coupling of their model to the LANL Parallel Ocean Program model. They made experimental predictions of the El Nino/Southern Oscillations with the coupled models using initial states provided by JPL's Estimating the Circulation and Climate of the Ocean (ECCO) project. These preliminary results validate ESMF performance in terms of scientific fidelity, thereby demonstrating the software's ability to serve in a production modeling system. Scientists can then use the ESMF-enabled couplings to compare the forecasting skills between different model combinations. The results also support the importance  of ECCO products for improving short-term climate forecasts.

Demonstrations of the software and the field tests took place at the Fourth ESMF Community Meeting, held at MIT in July. It attracted more than 100 attendees from the U.S. and other  countries. This year's meeting kicked off with the first ESMF on the Grid Workshop on July 20. Shujia Zhou of the CT Project and Northrop Grumman IT, Chris Hill of MIT, and Cecelia DeLuca of NCAR organized  the workshop. The event included presentations on grid software and grid implementations of ESMF and other Earth and space science  software frameworks.

The Community Meeting's main session on July 21 demonstrated the maturation ESMF has undergone over the past year. CT project manager Jim Fischer noted that ESMF “is graduating from NASA development funding to multi-agency coordinated funding.” On the technical front, ESMF now allows concurrent execution of models, as explained by NCAR's Nancy Collins. This capability enables ensemble forecasting: running multiple models with slightly different starting conditions. In addition to the field tests, presenters from GFDL, MIT, the Naval  Research Laboratory, NCAR and NCEP detailed efforts to replace their organizations' coupling software with ESMF.

Max Suarez of GSFC described development of the GEOS-5 atmosphere model, the first model  completely implemented with ESMF. As related by Don Anderson of NASA Headquarters, GEOS-5 and two other NASA models are being tested for hurricane forecasting in the Florida State University Superensemble. The Community Meeting concluded on July 22 with a hands-on tutorial and an industry partners forum.

Reaching beyond the NASA-funded partnership, the ESMF development team distributes the software to the scientific community via the Internet. NCAR, home of the core implementation team, released ESMF  Version 2.2.0 on July 13.

This article originally appeared in the CISTO News, a publication of  the Computational and Information Sciences and Technology Office at  NASA's Goddard Space Flight Center. The CISTO News is available at:


Share This