Lightning Strikes

The shuttle sits on the launchpad, in stark black-and-white, a flash of lightning bowed out from cloud to ground beside it, wrapping around it like a closing parenthesis, in the photograph on Matt Pearce's office wall.

It's a constant reminder of the research team's goal: to forecast a lightning strike 20 to 30 minutes in advance—time to get the shuttle secure, get the crew out, for the driver of the rocket-fuel truck on the wide-open access road to reach cover

. . .

"The consequences could be disastrous if lightning hit the Shuttle. It'd blow up a square kilometer if it hit that truck," says Pearce, a graduate student in meteorology. "The driver'd never know what hit." He traces the truck's route on a map of the Kennedy Space Center, shakes his head. "Lightning is the costliest weather problem they have. There are all sorts of outdoor activities at the Space Center—lightning costs millions in delays. The pad and the shuttle stand up so high they're natural lightning rods. And Florida gets the most lightning per square kilometer of any place in the entire United States. There are in excess of 100 thunderstorm days per year in the state of Florida. So forecasting lightning is of obvious safety and economic interest to NASA."

In 1993 Penn State meteorology professor Greg Forbes took a research fellowship at Kennedy Space Center. Forecasters then sounded the alert whenever a thunderstorm crossed within a five- nautical-mile (9.25-kilometer) radius of several different sites within the Space Center—which gave little lead time to warn each and every location.

Although weather observations were improving, the effort to increase the lightning lead time, Forbes saw, was being strangely hampered by those same improvements. The "very high tech, very high resolution" data sets, as Pearce describes them, of LDAR (which detects electrical disturbances inside clouds), LLAP (a cloud-to-ground lightning detector), and NEXRAD ("it picks up all different types of weather phenomena, from precipitation to sea breeze fronts to severe storms to gravity waves") resulted, not in better forecasts, but in information overload.

"The problem was getting forecasters to use all this data in real time," says Pearce. "There are copious quantities of data. Enormous amounts. The forecaster simply doesn't have time to integrate all the sources of information. There are too many things to look at."

The solution, Forbes decided, was two-part. On one hand, there was meteorology: What determines when a thundercloud will loose that first bolt? or when the danger has passed? On the other hand was the computer software needed to winnow out the irrelevant data and leave the lightning picture clear.

"The fascinating thing," Pearce says, "is that these data sets are giving us the full information needed to understand the whole evolution of a thunderstorm process. Where is it going, is it strengthening, is electrical activity associated with it, is it going to be, which one of those five clouds will become a thunderstorm? No one understands the meteorology of lightning in the context of real-time weather forecasting. That's what we're trying to learn.

"We've got 52 cases from 1993 to '95, 52 individual days of data, to look at. Every situation is different. But we're trying to find a common pattern so that we can home in on certain features."

Rather than waiting until they have the lightning pattern down pat, however, Forbes, Pearce, and graduate student Steve Hoffert are simultaneously developing the artificial intelligence system that will allow the Kennedy Space Center's forecasters to sort lightning clouds from harmless lookalikes.

Their system uses Object Oriented Analysis and Design techniques, the kind of computer code used, for instance, in Microsoft Windows and most graphics programs, Pearce explains. "We're looking at lightning forecasting problems and breaking them down into components. OOAD allows us to organize the components of the problem to show the relationships between objects in the system. It's a way of thinking about the problem that lets us solve it in a more natural way."

The team hopes, in the next year, to have a prototype in place that Space Center forecasters can use to integrate the various data sets and evaluate the research results. "There's a strong possibility we'll get a 15- to 20-minute lead time on that first lightning flash by then.

"Forecasting for large areas isn't all that difficult," Pearce concludes, "but when you get down to the golf-course scale, the kilometer scale, it becomes a very different problem. You could never forecast when lightning is going to hit the pad—that's impossible. But you can forecast the probability of lightning over the launchpad area." And in time to get the shuttle—or at least the crew and the fuel—out of harm's way.

Matt Pearce and Steve Hoffert are master's degree students in meteorology, 415 Rider II, University Park, PA 16802; 814-865-3525. Their adviser, Greg Forbes, Ph.D., is associate professor of meteorology in the College of Earth and Mineral Sciences, 503 Walker Building; 863-2458. Their research is funded by NASA.

Last Updated December 01, 1995