For the US military, accurate weather prediction is vital to both the planning and execution of worldwide missions. To extend its weather modeling capabilities, the US Air Force has joined the computing experts at the US Department of Energy’s (DOE’s) Oak Ridge National Laboratory (ORNL) in a strategic collaboration that includes procurement and operation of a new high-performance weather modeling computer system. Key members of the US Air Force and ORNL teams gathered on July 10 to kick off this project and tour the facilities supporting the new system.
The Air Force 557th Weather Wing provides the Air Force and Army with global- and regional-level numerical weather model forecasts. With the increasing scale of the requirements for the new system, the Air Force and ORNL identified an opportunity to leverage the capabilities of ORNL’s National Center for Computational Sciences (NCCS)—including expertise in high performance computing facilities and infrastructure, systems administration, computing procurement and acquisition, and system operations.
“We learned about the Air Force’s needs, and it was immediately clear that ORNL could help them solve their problems,” said Jim Rogers, NCCS director of computing and facilities. “We can integrate the Air Force weather team’s needs into our facilities in a cost-effective way, leveraging our capabilities to deliver exceedingly high availability to support their mission.”
As part of the collaboration, ORNL and Air Force Weather are exploring improvements to the numerical weather model, including the potential use of hardware accelerators, improved physical processes within the model, and existing capability gaps, including hydrological impacts. The partitioned system will run the forecast models as well as research and development codes that can dramatically advance the Air Force’s scientific capabilities for weather prediction.
“ORNL’s high-performance computing facilities and expertise across the computing spectrum will be a valuable partner as the Air Force looks to advance its weather modeling capabilities,” said ORNL’s Jeff Nichols, associate laboratory director of Computing and Computational Sciences. “Conversely, this new system will enable ORNL researchers to push the state of the art in next-generation computing technologies and evolve and refine the state of the art in weather research and prediction.”
The Air Force strives to maintain the most modern approach toward weather and Earth system modeling, a goal that requires harnessing the power of next-generation computing platforms such as those operated by the NCCS.
“The strategic partnership between the Air Force and ORNL provides the opportunity to leverage Oak Ridge’s world-class expertise to support our numerical weather forecast capabilities,” said US Air Force Director of Weather Ralph Stoffler.
In addition to the hardware deployment, ORNL researchers will apply their expertise in computational science and Earth system modeling to improve performance of the target Air Force model on the new computing resources and demonstrate novel science capabilities.
“Our group lives at the nexus of computing, climate, earth science, and weather,” said Kate Evans, group leader for the Computational Earth Sciences Group in the Computing and Computational Sciences Directorate. Through the research and development portion of the collaboration, Evans and her colleagues will use their expertise in computational performance research across multiple Earth system models to accelerate the GALWEM (Global Air-Land Weather Exploitation Model).
“We will use a combination of approaches to improve the model, some of which overlap with DOE strategies, creating a symbiotic collaboration with the Air Force and DOE,” said Evans.
The collaboration also allows the ORNL team to perform research to enable a predictive hydrology capability with a robust multimodel workflow. This workflow couples multiscale and multiphysics models for precipitation, runoff, streamflow, flooding, and impacts of flooding events. Along with weather forecasts, flood and drought prediction is of interest to a number of modeling and prediction centers, including the Air Force. Beyond model development, the multimodel workflow will utilize the evaluation, data processing, and model coupling expertise at ORNL.
The Air Force wants to later explore how advances in machine learning, combined with more traditional physics-based models, can advance its weather prediction capabilities.
ORNL is managed by UT-Battelle for DOE’s Office of Science. The Office of Science is the single largest supporter of basic research in the physical sciences in the United States and is working to address some of the most pressing challenges of our time. For more information, please visit https://science.energy.gov/.