The Storm Maker
Robert Wilhelmson’s computer simulations of severe storms, like tornadoes, have garnered scientific acclaim and an Academy Award nomination.
On April 19, 1996, the air thickened and the sky darkened to a pea green over a 300-mile swath of central Illinois. It was classic tornado weather that didn't disappoint. Between 1 p.m. and 9 p.m., more than 30 tornadoes ripped across the prairie. They threw trucks into trees and flattened parts of Urbana, east of the University of Illinois. In nearby Ogden, they left 350 of the town's 800 residents homeless. Although tornado watches had been issued hours earlier, in the end, some of the communities directly in the path of the tornadoes had only 10 minutes warning.
As an atmospheric scientist, LAS's Robert Wilhelmson knows why, a half-century after the National Weather Service issued its first official tornado warning, these storms still often catch us by surprise. A complex set of atmospheric conditions must snap together perfectly to unleash a tornado from a thunderstorm. Scientists understand most of the telltale forces. But what eludes them is a foolproof means of identifying which storms will produce tornadoes. Like detectives trying to solve a murder from a sketch artist's rendering, they are missing the incriminating fingerprints.
Wilhelmson has a strategy for making a positive ID. Using massive computing power and graphics that rival those of a Hollywood blockbuster, Wilhelmson is re-creating severe storms so that scientists can analyze them and discover how these powerful natural systems work. His approach is novel in that it bridges art and science. Not only have his simulations garnered scientific acclaim, but one was also nominated for an Academy Award. Others have been featured in Scientific American and National Geographic and on PBS. Millions of people saw his tornado simulation that was part of the 1995 Omnimax film, Stormchasers. Not many scientists can claim their work may save lives and entertain.
Big, Bad Storms
Looking at Wilhelmson's eye-catching simulations, it is easy to forget the serious nature of his research. Tornadoes are among the most destructive forces on Earth. The worst among them form in storms called supercells—the towering thunderstorms that unleash the energy equivalent to several atomic bombs within their one- to two-hour lifespans.
Conditions are favorable for these menacing storms whenever cold, dry air overlies warm, moist air and there are large changes in wind speed with height. A slight instability shoves the warm air upwards—heat rising off a plowed field will do—inaugurating a powerful cycle of updrafts and downdrafts. Warm air in supercells spins slowly as it rises through the thunderhead; air cooled by evaporation rushes downward, spreading outwards as it hits the ground and producing gusts of up to 75 MPH. These downdrafts are rotating and sometimes slip beneath the updrafts. Friction and the tilting of these winds on end often trigger tornadoes.
Computer simulations are ideal for helping identify the elusive tornado triggers because, unlike what is depicted in movies like Twister, real tornadoes are hard to find and observe.
The storm systems that give rise to damaging tornadoes are huge, often stretching across several states, even though individual storms within the systems may be only 10 to 20 miles wide. The National Weather Service monitors their progress with a nationwide grid of weather observing systems consisting of weather balloons, Doppler radar stations, surface stations, and geostationary satellites. Stormchasers, who follow the storms from the ground, may also call in data.
Because observing stations on the Earth's surface are spaced anywhere from 50 to 200 miles apart, the resulting weather maps are too crude to capture the small storm changes that trigger tornadoes. For that, says Wilhelmson, "We would ideally need observing systems on every street corner to track storms with the necessary degree of detail."
Conceivably, simulated storms can provide that kind of block-by-block resolution. They require good data—from real storms—and programs that can make these data behave like real storms. The latter is extremely tricky. The equations governing storm behavior are so complex and detailed that unless researchers devised shortcuts, or approximations, solving the equations would take literally hundreds of years, even on the fastest computers available. The artistry of simulation lies in streamlining the computing without compromising accuracy. Wilhelmson says researchers will know if they have approximated the physics accurately when their simulation behaves like the real thing. "We know what surface weather conditions are associated with evolving severe storms. We also have video and still pictures. If the overall behavior of our simulation matches the general observed behavior, we know we've likely got it right."
The key word is behavior. The goal isn't to create storms with all their "visually observed detail." Scientists are more interested in the storm's dynamics—the updrafts, downdrafts, and rotation—that are usually invisible or masked by clouds. This is where the graphics become important.
Better Than Real
A single two-and-a-half-hour storm simulation can easily produce a trillion bits of data. Without a sophisticated means of interpreting that data, the information generated would be so overwhelming as to be nearly useless. The graphical techniques used by Wilhelmson's team not only depict storm evolution and structure but also elucidate the relationships among data fields such as wind, temperature, moisture, and pressure.
Wilhelmson's techniques for elucidating data also generate impressive imagery. In 1989, his animation of a thunderstorm won 14 awards, including an Academy Award nomination. The techniques he and his team employed for visualizing air currents and water movement are similar to those used in Twister.
In the mid-1990s, Wilhelmson began experimenting with virtual reality. Using a room-sized virtual reality device called the CAVE, he was actually able to step inside a simulated tornado and view it from the inside out.
More recently, Wilhelmson's team has successfully overcome the technological limitations for simulating the evolution of a tornado from a thunderstorm, which was necessary for them to begin identifying the forces that trigger tornadoes.
Which brings us back to the tornado of 1996. The data from this storm are being used by Wilhelmson in the search for tornadic fingerprints. If his team can make their model of this storm system behave realistically, then dissect it for the hallmarks of tornado triggers, scientists and forecasters will be able to correlate that information with what Doppler can detect. "By coupling our understanding of tornadoes with live Doppler radar, we'll be better able to determine if a storm will produce a tornado in this county or the next."
Though this capability may be years away, we may take comfort in knowing that some day forecasters will recognize whether the storm looming on the horizon will level our community or rumble past.
By Holly Korab
Collaborators in severe storm research at Illinois have included the following students or post-docs: Michael Bradley, Harold Brooks, Ching-Sen Chen, Steve Chin, Kelvin Droegemeier, Brian Jewett, Bruce Lee, Ed Mlodzik, Luis Muñoz, Steve Peckham, Lou Wicker, and Guangming Zhou. Collaborators in visualization and virtual reality have included: Matthew Arrott, Mark Bajuk, Colleen Bushell, Stephen Fangmeier, V. J. Jaswel, M. McNeill, Jeff Terstriep, Jeffory Thingvold, David Wojtowicz, and Jeff Yost. External collaborator: Joseph Klemp. Group support has been provided by Crystal Shaw.