ACADEMIA

With Blue Waters, atmospheric scientists believe they'll be able to gain important insights into the formation of dangerous twisters, leading to earlier and more accurate warnings.

In many ways, tornadoes remain a mystery. While their powerful winds can leave broad swaths of devastation—damaged and destroyed homes and businesses, injuries and fatalities—twisters are a challenge for atmospheric scientists to study. It's difficult (not to mention hazardous) to be in the exact right place at the precisely the right time to see one in action, and models that capture all of the critical small-scale details within a dynamic storm system are computationally demanding.

That's where the sustained-petaflop Blue Waters supercomputer comes into the picture. The University of Illinois' Bob Wilhelmson is the principal investigator for a Petascale Computing Resource Allocation project. The National Science Foundation's PRAC program enables scientists (in this case Wilhelmson and collaborators at Illinois and other institutions across the country) to work closely with the Blue Waters team, preparing their codes to take full advantage of its several hundred thousand cores.

One target for Wilhelmson's work is tornadogenesis—the process by which a tornado forms. Many tornadoes, including all of the extremely powerful storms that measure EF3, EF4, and EF5 on the Enhanced Fujita Scale, are spawned by supercell thunderstorms. Forecasters can identify conditions in which tornadoes are likely to form, but it's not yet possible to predict exactly when and where tornadoes will form, how strong they will be, or what path they will follow.

"A lot of the key aspects of tornadoes are just not understood," says Brian Jewett, a University of Illinois atmospheric scientist and a member of the PRAC team. "After all these years, how the tornado forms is still a source of considerable controversy."

To address these questions, the PRAC team is working with CM1, a sophisticated model of atmospheric dynamics and thermodynamics, in order to simulate the interaction between the birth of a tornado from a supercell in the ultra-high resolution needed to adequately represent important small-scale features that influence the evolution of the tornado. For example, without Blue Waters, the team does not believe they could resolve thin curtains of precipitation, tens of meters thick, which potentially transfer large amounts of angular momentum from cloud-base to ground.

Another intriguing small-scale feature is the near-surface inflow layer. "Sometimes you'll see a racing jet of wind a little above the ground that is howling toward the tornado," Jewett says. "This is all going on in the lowest 100 meters above the ground. That's a pretty small zone, but it's a detail that has to be captured. If you don't get these small-scale features, you probably aren't getting the structure of the tornado correct."

Jewett says more detailed simulations using Blue Waters may provide valuable insights into the environmental and particular storm conditions that support tornado development and longevity.

"Large, long-lived tornadoes are the ones that cause a lot of the damage and fatalities," he says. "We don't really understand the conditions that allow them to form and to persist. Why do some tornadoes last for an hour or more and other last for seconds?"

Better understanding of what conditions lead to the most dangerous tornadoes can be used to improve the lower-resolution models used in routine forecasting, leading to more precise warnings about the most dangerous twisters.

Getting breakthrough results on Blue Waters won't just be a matter of using existing storm-tornado modeling codes on a more powerful computer. The PRAC team also is working to improve the CM1 model, originally developed by team member George Bryan at the National Center for Atmospheric Research, and the microphysics using the expertise of Matt Gilmore at the University of North Dakota.

"Major modifications to the code are being considered in order to obtain the computational performance necessary to run it at the resolution and with the physics we need to accurately simulate the tornado," Jewett says. "Without them, we will have to either use less sophisticated physics or lower resolution, both of which we do not want to do."

Wilhelmson says it is also necessary for the team to decide what simulations will be run on Blue Waters at higher resolution than ever before and how the vast amount of data generated during these simulations will be analyzed and visualized in a way that improves our understanding and ability to predict severe storms and when tornadoes can be expected to form within them.

"Preparing for Blue Waters is a challenging task that requires the unique expertise of this entire talented team," he says.

Team members
George Bryan, National Center for Atmospheric Research
Matthew Gilmore, University of North Dakota
Brian Jewett, University of Illinois at Urbana-Champaign
Joe Klemp, National Center for Atmospheric Research
John Michalakes, National Center for Atmospheric Research
Leigh Orf, Central Michigan University
Glen Romine, National Center for Atmospheric Research
Bill Skamarock, National Center for Atmospheric Research
Lou Wicker, National Oceanic and Atmospheric Administration
Robert Wilhelmson, University of Illinois at Urbana-Champaign (principal investigator)
Paul Woodward, University of Minnesota