In a laboratory at the University of North Carolina Charlotte, student researchers using scale-model NASCAR vehicles are exploring the next great frontier in auto racing safety. Using those same models in a water tank, they've already discovered a potential reason why roof flaps may have been less effective on a Sprint Cup car with a rear wing versus a spoiler. Other students are doing similar experiments in a virtual world, some of it at the request of a team that chooses to remain anonymous, all of it with the goal of preventing race cars from becoming airborne.

Every driver in NASCAR's national divisions wears a head-and-neck restraining device, every race track is coated in an impact-absorbent wall material, and the latest version of the race cars used on the Cup tour may very well be the safest on the planet. And yet, every year, cars still go hurtling up into the air and come crashing down again, evidence of an issue that continues to plague NASCAR and other motorsports series even in an era of safety advances on other fronts. It happened several times this past NASCAR season -- Brad Keselowski flipping onto his roof at Atlanta, Ron Hornaday and A.J. Allmendinger going end-over-end at Talladega, Kasey Kahne lifting into the air at Pocono -- just as it's happened for decades.

NASCAR races at restrictor-plate events have become infamous for their ability to toss vehicles into the air like children's toys. But this is far from just a NASCAR problem -- perhaps the most horrifying airborne crash this year came in the Indianapolis 500, when Mike Conway lifted off in an accident that left him with fractures in his left leg and lower back. Marc Webber flipped over and slid into a tire barrier in a harrowing Formula One crash in Spain. Airborne crashes have become such a concern across all racing disciplines that the subject has drawn the attention of researchers who operate outside of formal sanctioning bodies, some of whom are speaking as part of a Safety and Technical Conference at this week's International Motorsports Industry Show in Indianapolis.

"I don't think anybody's going to say, 'We're going to have this solved for next year.' Nobody's willing to stick their necks out like that," said Ray Leto, who heads U.S. efforts for the motorsports aerodynamics consulting firm TotalSim. "But I think quietly, in the background ... it's out there. People can see it. People, fans and drivers, are talking about it. It's another thing to incrementally move forward in safety in motorsports. Whether it's crash testing or safer walls or even fences around race tracks to keep cars on the tracks, all these things have evolved over the last 100 years of racing. You push the envelope, and somebody has to come up with the next solution."

Everybody, from NASCAR to the FIA to the sanctioning body of LeMans-style sports cars in Europe, wants to keep vehicles on the ground. And yet, the effort to do so is so complex that even testing carries with it a host of questions that are difficult to answer. How do you recreate a car spinning at such a rate of speed that it lifts into the air? How do you replicate a unique series of events on a race track, often involving multiple vehicles, that sends one car airborne? The HANS (Head and Neck Support) device and SAFER (Steel And Foam Energy-Reduction) barrier, the two most important life-saving devices in modern motorsports, came about through years of testing -- of slamming sled after sled into wall after wall, again and again, until developers got it right. But how do you prevent something you can't even test?

That's the quandary facing those battling airborne crashes, scientists like Leto and Mesbah Uddin, the latter an associate professor of motorsports at UNC Charlotte. Uddin's group uses tools like scale modeling and computational fluid dynamics (CFD) -- a virtual method of analyzing air flow -- to determine what causes a NASCAR race car to lift off. Some of their work is proprietary, being done for a team Uddin cannot identify. But some of it has also been published, including investigation into why roof flaps perhaps deployed less reliably on a winged car than one with a rear spoiler.

"What happens is, in the case of a spoiler, when we looked at extreme yaw, when the car is completely reversed, going in a 180-degree motion, the most extreme case found that the wake behind the car gets attached so there is no low pressure region right near the roof flaps, which sort of prevented the roof flaps [from deploying]," said Uddin, a former CFD analyst at Chrysler, who works with collaborator Peter Tkacik. "In the case of the spoiler, there was a huge low pressure region in vicinity of the roof flaps, so it helped in deployment of the roof flaps. ... I'm not saying this is the conclusive reason, I'm saying this is likely the cause. It needs further investigation."

This sort of experimentation isn't perfect. "The phenomena at the race track is way more complicated," Uddin admits. But CFD technology has progressed to the point where researchers can apply almost any force to a simulated car at any attitude, trying things that would be impossible to attempt in the real world. Computational simulation also allows scientists to create scenarios involving multiple vehicles, typically the kind of situation that produces spectacular airborne crashes at places like Talladega, with the goal of figuring out ways to prevent them.

"Obviously, it's difficult to conceive of every scenario there might be," said Leto, a former engineer in IndyCar racing. "You take like last year at Talladega, it's how the car was entering the banking and where it exactly spun and how many cars were in front of it. You can't simulate everything exactly. But I think the ideas people are thinking of is, coming up with a generic set of parameters where you could do some flat spins on tarmac or pavement, you can spin off into the grass and see what happens when the friction changes, you could do it with one or two cars in front on a progression of bank angles as well. You can do that through simulations over and over again."

NASCAR does its own research into the airborne issue, and has done so since the early 1990s, when high-flying crashes at Talladega and Michigan led the sanctioning body, with the help of car owner Jack Roush, to design the roof flaps that are now standard on all vehicles. "They were out front with this stuff," Leto said. The problem is, the roof flaps don't prevent all airborne crashes, something that's become quite evident with a spate of liftoff accidents at big, fast NASCAR tracks in recent years. Mike Fisher, managing director of NASCAR's Research and Development Center, says that's because most airborne crashes are the result of what he calls "assisted spin" -- another car pushing the vehicle through the acceleration process.

"That's especially challenging for us as well," Fisher said, "because the aerodynamic devices we have in place, they need a little bit of time to work. When the flaps deploy, they need a second or two to do what they need to do, which is to catch the air from the roof of the vehicle. But when you immediately go from straight down the race track to 180 degrees backward, or if you go through that rotation too quickly, a lot of the devices don't get the full opportunity to work, and that's why in some of those occasions we have liftoff. You've basically spun the car so quickly that you've put it at the liftoff point before it's been able to scrub speed or catch air. That's a problem, I wish I had the magic answer to that. But I don't think anybody would like to see us race cars by themselves."

Fisher, who is attending the IMIS event, said NASCAR works with manufacturer partners every year to try to raise liftoff speeds as part of its standard vehicle development process. High yaw testing -- rotation of the car past 90 degrees -- is part of that, and evident in moves like changing from the wing to the rear spoiler on the Cup car, or rolling out the new Nationwide models. The goal of preventing airborne accidents is "specifically baked into our vehicle development process," Fisher said.

But the testing challenges remain, and it's here where NASCAR differs somewhat from researchers working outside of sanctioning bodies. Modeling and CFD, Fisher said, provide one piece of the technical solution, one NASCAR doesn't want to rely solely upon in making decisions. NASCAR -- which, thanks to its partnerships, has some assets at its disposal that outside entities don't -- prefers to test the actual cars themselves, which can present its own hurdles given the size of the vehicles and the somewhat hazardous nature of the subject matter.

"We're somewhat limited in the number of wind tunnels available that are both large enough and where the operators are willing to let us test for cars lifting off, because obviously the thought of a car lifting off in a wind tunnel can be a little unsettling to the organization that owns that tunnel," Fisher said. "They don't want to damage it. But we fortunately have some partners, Dodge in particular has helped us in recent years, also General Motors, who allow us to turn our cars around at pretty high speeds in their tunnels to evaluate that. Modeling is one tool we can use to understand that, but you really have to back that up with physical testing."

Within NASCAR, there is what Fisher calls a "somewhat restrained level of confidence" in virtual testing, given that not everything on the computer screen correlates to the wind tunnel as perfectly as hoped. But as computer modeling becomes more savvy, Fisher didn't rule out NASCAR eventually utilizing some of the other efforts going on within the industry. For now, they're content to work along parallel fronts, each with the goal of learning a little more about what it takes for race cars to keep all four tires on the ground.

"I don't think any of these rules, whether it's IndyCar or NASCAR, have come about through people just guessing," Leto said. "Just like the roof flaps that NASCAR did 20 years ago, there's a purposeful process to going out there and testing these things. ... Everybody's working on it, I think."

The opinions expressed are solely those of the writer.