Supercomputers Create World's Most Detailed Simulations of Tornadoes

Thunderstorms are equally terrifying and captivating. At their most extreme, supercell thunderstorm — those with persistent rotating updrafts — can spawn tornadoes, causing wide-spread devastation. How do tornadoes form? What key ingredients do they require? And can they can be predicted to provide timely warnings to the public? These are all open questions.

"The National Weather Service has about a 70% false alarm rate, meaning that for about two out of three times when there's a tornado warning issued, there's no tornado," said Leigh Orf, an atmospheric scientist with the Space Science and Engineering Center at the University of Wisconsin-Madison. "We need to do better for people to heed these warnings."

Overall, the physics of tornadoes is well understood, but the small-scale aspects of their formation and how they are maintained within supercell thunderstorms require more research. Tornadoes cannot be fully studied in a laboratory, and field research of these rare, isolated events is difficult to plan (though scientists try).

"We don't have remote sensing technology that can capture the dynamics and physics of full storms with the needed resolution to get to the bottom of their mysteries," he said.

Creating Realistic Tornadoes

For the past decade, Orf has been on the leading edge of efforts to create – or in some cases re-create – virtual tornadoes using some of the most powerful supercomputers in the world. Most recently he has been using Frontera, at the Texas Advanced Computing Center — the National Science Foundation leadership-class supercomputer and the fastest system at any university in the world — to advance his work.

The software Orf developed is the first to create visualizations of fully resolved tornadoes and their parent supercells that look and behave like those seen in the real world.

"We need to throw a lot of computational power to get it right and resolve salient features," he said. "Ultimately, the goal is prediction, but the truth is, we still don't understand some basic things about how supercell thunderstorms really work. We do understand a lot; there's been a lot of numerical work prior to my own, and a lot of field work, but it's really a hard to answer questions like, will this supercell that just formed produce a tornado, and if so, will it be especially violent?"

Among the storms that Orf studies is the May 24, 2011 El Reno event — a tornado outbreak in Oklahoma that killed 11 people and injured 293. In the process of studying this storm, Orf has identified physical features that might provide critical clues to tornadogenesis.

"In these simulations, there's a lot of spinning going on that you wouldn't see with the naked eye," Orf said. "That spinning is sometimes in the form of vortex sheets rolling up, or misocyclones, what you might call mini tornadoes, that aren't quite tornado strength that spin along different boundaries in the storm."

Another feature detected in simulations that was "discovered" by Orf and his team is called the streamwise vorticity current, or SVC, a current of air that rotates on its side and get sucked up into the updraft, intensifying it.

Orf's simulations suggest these spinning currents of air are important, and fieldwork is starting to show the same. "Our simulations have shown that in some cases, weak subtornadic vortices can come together to help assemble a tornado," he explained.

The streamwise vorticity current, which in simulations is found in the rainy part of the storm, appears to be very important in helping to modulate how strong the updraft is near the ground, which is important for tornado formation and the maintenance of the tornado.

"Having a giant super-sucker vacuum cleaner updraft very near the ground is very conducive to forming and sustaining strong tornadoes," Orf said.

Orf's findings do not always agree with previous conceptual models, which can cause a bit of eyebrow-raising in the field. But the streamwise vorticity current is a feature that researchers are now looking for on radar and by flying drones and planes into storms.

Resolving Massive Storms

On Frontera, Orf is analyzing tornadoes with the highest spatial (10 meters grid spacing) and temporal resolution (with data saved every 1/5 of a second) ever attempted. The effort extends work that began on the Blue Waters supercomputer at the National Center for Supercomputing Applications, results of which were published in the September 2019 issue of Atmosphere.

Running such a simulation is no trivial matter. It took Orf three months to complete one full high-resolution tornado simulation on Blue Waters, and it involved on-the-fly tweaking of his code to make it work. Why do this heavy lifting? "It is only with this level of granularity that some features become evident."

His first efforts on Frontera involve re-simulating storms he has studied in the past to see how the results change as a function of different parameters. Frontera provides the opportunity to create dozens of simulations at very high resolution which can be compared in order to tease out the factors that result in the formation of powerful tornadoes – or result in no tornado.

"Very small changes early on in the simulation can lead to very big changes in the simulation down the road," Orf said. "This is an intrinsic predictability issue in our field. We're doing some of the frontier work to try to tease out these variables."

Supercell Mohawk

Orf doesn't only study the bottom of thunderstorms. He also investigates their tops. Recently he has focused on modelling above anvil cirrus plumes, a cloud "mohawk" that sometimes forms at the top of the anvil of supercell storms.

These features are important because they are visible from satellites and have been shown to be predictable indicators of damaging storms. They can be correlated 70% of the time to hail-producing events, which cause $22 billion in damages in the U.S. each year.

The plumes have been recognized as important features of supercell storms since the 1990s, but what causes them was somewhat of a mystery. Orf's high-resolution simulations provide clues to support a new hypothesis.

As air spreads out horizontally at the top of the thunderstorm updraft, some of it travels sideways in a thin jet with winds exceeding 90 meters per second. When that jet reaches supercritical velocity, it suddenly gets turbulent. This causes air from the troposphere (the lowest layer of Earth's atmosphere) to mix with stratospheric air (the layer above the troposphere), irreversibly drawing tropospheric cloud ice into the stratosphere where it is visible as a vertical cloud protrusion on satellite imagery.

"We've actually pinpointed the mechanism that appears necessary to form the above anvil cirrus plumes," Orf said. "Other people hadn't quite gotten to that point. Our work keys in on this specific jet feature that can only be found by running models at super high-resolution on machines such as Frontera."

Orf's team presented the findings at the American Geophysical Union Fall 2019 meeting and the American Meteorological Society 2020 Annual Meeting.

As a short-term forecasting tool, the plumes may help forecasters provide warnings earlier and make their predictions more accurate.

Visualizing Violent Storms

Understanding and communicating ideas about storms requires visual illustrations, and Orf is among the best at creating incredibly detailed and realistic scientific visualizations of these extreme events. His visualization work won awards including XSEDE's People's Choice award as well as Hyperion's Innovation Excellence Award, and visualizations of his tornadic supercells can be found at the American Museum of Natural History and Harvard Museum of Natural History.

This is partly due to some unique features of the software he developed for managing and visualizing model output. In addition to being high-resolution spatially, Orf has added code that lets him efficiently capture and preserve data from all the time slices of a simulation using a combination of data compression and novel I/O processes. "Visualizing data with this kind of spatial and temporal resolution really brings storms to life," he said.

With a new, large allocation of compute time on Frontera, Orf hopes to continue to make discoveries while developing the methods and codes he has worked on for decades. (He recently released open source versions of his codes to the community through GitHub.) His highest-ever resolution supercell simulations may uncover more small-scale features that help understand and ultimately predict tornadoes and other extreme weather events.

"For a tornado researcher, I'm a strange bird in that I really do enjoy high performance computing," Orf said. "I'm willing to put in the coding effort that's going to ultimately lead to a whole new framework of analysis that I think will push the field forward."