U.S. Military Turns to New Supercomputers to Push the Limits of Weather Forecasting

Two new supercomputers named after a pair of iconic military meteorologists make up one powerful system now operational at Oak Ridge National Laboratory, where it supports worldwide U.S. Army and Air Force weather modeling and forecasting operations.

The roots of this sophisticated computational tool trace back to a $25 million contract unveiled in 2019. It was since built and recently delivered by Hewlett Packard Enterprise to the Tennessee-based national lab managing it, according to a press release.

“The system was formally accepted in January and became fully operational on Feb. 8, 2021,” Computing and Facilities Director for the National Center for Computational Science at ORNL Jim Rogers, told Nextgov via email Wednesday. The lab “continues to actively conduct research on this system.”

The supercomputers power projects steered by ORNL and Air Force Weather, or AFW, the branch’s meteorology division that leverages atmospheric and solar data to provide comprehensive terrestrial and space weather alerts, forecasts and analyses for military efforts spanning the globe. Air Force officials named the new machines “Fawbush” and “Miller,” in homage to meteorologists Major Ernest Fawbush and Captain Robert Miller, who in 1948 predicted the first tornado forecast at the Tinker Air Force Base, the release noted.

Supercomputers’ capabilities are expressed by quantifying the number of floating-point operations per second, or “flops,” they can perform. A one petaflop system can complete one quadrillion floating-point operations per second—and combined, the new systems meet a peak performance of 7.2 petaflops. That’s roughly 6.5 times faster than AFW’s existing system, the release said, adding that the boost will enable “larger computations at a higher resolution, increasing accuracy in global weather simulations from 17 kilometers between model grid points to 10 kilometers.”

AFW will retain and repurpose the prior production system, Rogers confirmed, noting “that is managed independently of the strategic partnership between ORNL and the Air Force.”

Fawbush and Miller also mark one of the first operational systems to be powered by the HPE Cray EX high performance computing architecture—previously known as Cray Shasta—designed explicitly for the next generation of supercomputing, or exascale. That same system underpinning the two is also set to power America’s three in-the-making exascale systems, including Frontier, which is slated for installation at Oak Ridge this year. The next step on the general scale of computing capabilities, exascale systems will be capable of at least one exaflop—a quintillion, or a billion billion—calculations per second.  

“The end-to-end [high performance computing] technologies made possible by the HPE Cray EX supercomputer will enable greater speed and dedicated performance to advance simulations in weather forecasting that were never made possible before,” Bill Mannel, HPE’s vice president and general manager for high performance computing said in the release.

Together with ORNL’s Computational Earth Sciences Division, the Air Force’s weather wing intends to use the system to help usher in “a new era of weather forecasting capabilities,” officials added. Potential next-level applications include remote sensing of a cloud-covered environment to help with mission navigation by forecasting the formation, growth and precipitation of atmospheric clouds.

“Researchers plan to achieve this by using comprehensive cloud physics that are not made possible with existing statistical regression models,” the release noted.

Among other uses, the advanced system will also aid military researchers in producing a global hydrology model involving hundreds of watershed and drainage basin simulations to help improve the accuracy of flooding- and streamflow-centered predictions for future events.  

Pointing to the unfolding research, Rogers noted that much of the current efforts “are focused on improved model physics, including better modeling of hydrology impacts as model resolution improves.” Officials are also exploring the applicability of graphics processing units to weather modeling. 

“ORNL and AFW have an active partnership that drives and directs this research,” Rogers said.