Espe Yocham and her father-in-law, Sonny, stood in the late-afternoon sun amid the shattered heap of glass, plywood and aluminum siding that was her trailer house in Blanchard, Okla. Several yards away, her silver Mustang sat upside down in the red-brown mud.
Across the field were the crushed remains of another trailer, that of Sonny`s youngest daughter and her family. Espe`s legs and shoulder were covered with ugly purple bruises; she had just returned from the hospital where her husband, Brandon, lay in intensive care with broken ribs and a punctured lung.
”The trailer started shimmying and shaking,” she recalled. ”I grabbed my baby, Elvis-then all of a sudden everything just exploded.”
It would require no stretch of the imagination to believe that a bomb had in fact gone off, so intense was the devastation. But this scene was a pure work of nature-and a surprisingly common one. Thunderstorms, and the lightning, winds, hail and tornadoes they spin off, take a huge if little appreciated toll every year. In the United States alone, the annual losses exceed $1 billion.
Lightning kills about 100 people and starts more than 10,000 fires. Hail smashes windows, gouges unprotected cars and aircraft, and robs farmers of 1 percent of their crops. About 165 people are swept away in flash floods following storms, while tornadoes claim 100 lives. At any given time, 2,000 thunderstorms are in progress around the world, releasing energy equivalent to several atomic bombs.
For all of the advances in weather prediction brought about in the last decade by satellites, radar and supercomputers, when it comes to forecasting these most damaging of storms, meteorologists acknowledge that they are still coming up short. Predicting tomorrow`s temperature is easier than foretelling tomorrow`s rainfall; predicting broad, regional weather patterns even five days in advance is easier than determining exactly where tomorrow`s thunderstorms will develop, how long they will last or how severe they will be.
One shortcoming is lack of up-to-the-minute data. The most accurate weather information still comes from temperature, air pressure, humidity and wind readings from balloons that are launched every 12 hours from only about 100 points across the country. Storms can form, devastate an area, then dissipate in much less time. And storms tend to be tens of miles across, while hundreds of miles stretch between balloon launch sites.
The mysteries that still surround the behavior of thunderstorms present even more fundamental problems for forecasters. Precisely which atmospheric conditions precipitate the towering, accelerating growth of clouds into thunderheads; how the vast, twisting cyclone of air that develops in the heart of some clouds gives birth to tornadoes; how electrical charges build up to trigger lightning: All are questions crucial for improved prediction, and all are questions begging for answers.
Much of what we do know about thunderstorms has come from scientists at the National Severe Storms Laboratory (NSSL, pronounced ”nissel”) in Norman, Okla. Norman is in the heart of what meteorologists call ”Tornado Alley,” a region that stretches from central Kansas to Texas. All around Norman is evidence of the frequency and ferocity of local thunderstorms. Most trees lean north, bent by the prevailing winds. A local carwash hands out rain checks that are good for 24 hours: If a storm spatters mud on your freshly washed car, come back and they will rewash it, free of charge.
Researchers stuck inside
Most of science`s standard techniques are of little help. Scientists can`t take atmospheric readings in the middle of a tornado, nor can they examine a dead storm or experiment with a live one in the lab. Instead, meteorologists are learning to study storms indirectly, by reconstructing their wind patterns from the damage they leave in their wake, or by using a variety of remote sensors, such as radar that can measure speed and direction of winds aloft and special lasers that measure precipitation size.
On the recent Saturday afternoon when the storm ripped through the Yochams` trailer, NSSL researchers watched from the roof of the lab as the towering purplish-gray clouds advanced across the countryside toward Norman. At the storm`s base, an inky ”shelf” cloud jutted out like a large flat foot; to the meteorologists, it signified that the storm, still several miles off, was producing high winds. Fat lightning flashes arced from cloud to cloud, accompanied by increasingly sharp cracks of thunder. Then a chilly wind rose, and big drops of rain began to fall. The researchers hurried inside, and within minutes, the rain was so thick that no one could see across the NSSL parking lot.
Thunderstorms this spectacular are rare, even for Tornado Alley. But NSSL scientists were stuck in the lab instead of out studying the storm because earlier in the day they predicted that the clouds gathering over Oklahoma would not coalesce into a significant storm. Their botched forecast points to the central irony of storm research: Scientists cannot predict storms well enough to reliably catch them.
Why do storms form at night?
For the past month, David Jorgensen, head of NSSL`s mesoscale-research division, and his team have been probing storms with a National Oceanic and Atmospheric Administration (NOAA) research plane that rotates among different labs around the country. With the plane, researchers are able to range far and wide in search of storms.
This season, the Jorgensen team has focused, in part, on huge circular storm clusters, known as ”mesoscale convective complexes” (MCC`s), which were first identified by NSSL Director Robert Maddox when he was a graduate student in the 1960s.
In the Midwest, especially, MCC`s are a potent meteorological force. They produce twisters and damaging hail, yet they supply the region with most of its rain during the growing season. Currently, many of scientists` questions- why storms often form at night, when during their life cycle they spawn tornadoes and how they get so big-go unanswered. But the research flights are helping to change that. Under the right atmospheric conditions, about 1 storm in 10 develops into a long-lived ”severe” thunderstorm capable of spawning tornadoes, large hail and fierce winds. In these storms, the current of rising air can attain speeds of close to 100 m.p.h., and ice particles inside the storm may tumble about for hours, growing to the size of walnuts, tennis balls or even larger. The intense updrafts can also be transformed into a whirling column of air that gives rise to tornadoes.
Flying around a storm
The pattern of winds within a storm in progress is thus of crucial interest to the meteorologists. To measure these winds without risking their lives, researchers avoid the violent up-and-down drafts at the storm`s center. They instead fly in box patterns around a storm while scanning it with radar and other sensors that provide a detailed three-dimensional view of the storm. Probably the most crucial piece of electronics-a Doppler radar-is mounted in the rear of the craft. Like conventional radar, Doppler measures the overall intensity and distance of a storm by bouncing microwaves off wind-driven rain and hail. But the Doppler also measures the velocity of particles moving toward or away from the radar, giving meteorologists a glimpse of circulation inside the storm.
Normally, after a storm has roared through, it takes a while for the atmosphere to destabilize enough for another storm to develop. But only a few hours after the storm passed through Norman that Saturday, the sky was again ominous.
At 9 p.m., Jorgensen`s plane took off into a thunderstorm complex that by midnight would cover most of Oklahoma and part of North Texas. The first task on the night`s agenda was gathering data that will be used to develop a reliable hail-detection system. The pilot flew to Yukon, Okla., where researchers on the ground were probing the clouds with Doppler radar. As the plane circled, scientists on board measured hailstones with the plane`s sensors.
The research team was also investigating one of the mysterious aspects of thunderstorms: how a nonelectrified cloud becomes an enormous electrical generator, with negatively charged ice particles segregated at the bottom and positive ones at the top.
To tackle this question, lightning experts John Hallett of the Desert Research Institute in Reno, Nev., and Steven Rutledge of Colorado State University in Ft. Collins needed a vertical profile of the storm`s electrical field. The pilot swung the plane into a slow spiral descent from 24,000 to 10,000 feet while Hallett recorded the electrical field with onboard sensors. At 2 a.m., Jorgensen called it quits, and the plane turned toward home. Many of the flight crew were dragging with fatigue, but the research team was exuberant. The storm produced a wealth of data, which the scientists expect to spend a good two years analyzing.
The researchers on this flight observed three out of four of a severe storm`s fearsome features: strong winds, hail and lightning. Capturing data about the fourth, tornadoes, is perhaps the most pressing need but also the most difficult to accomplish. Tornadoes form quickly, with little warning.
`Wicked witch is dead`
On the Sunday after the storms hit Norman, another team from the laboratory was picking through the Metzger family`s flattened vegetable garden near Blanchard, Okla. Four-year-old Bubba Metzger danced around the garden chanting, ”Ding, dong, the wicked witch is dead.”
It was a sunny afternoon, and NSSL staffers Ken Howard and Doug Forsyth were examining the storm`s damage to determine whether it produced a small tornado as it raced through Grady County. Twisters leave a characteristic trail with debris flung in every direction.
The Metzgers escaped relatively unscathed. Over all, the damage to their property gave an inconclusive picture of the storm. Lawn chairs, parts of the well house and tree branches generally were flung southeast, the direction the storm was moving. But a canoe was tossed north into a tree-a hint that twisting, cyclonic winds may have skipped through the area.
A couple of miles down the road, Forsyth pulled over next to a man clearing branches and shingles from his yard. ”You are a day too late,” he said curtly to Forsyth. The Weather Service issued a warning the previous day for high winds, though not for tornadoes. In response to the researcher`s queries, the man said that he does not know whether a funnel passed through because he took shelter in his basement.
”To them, one weatherman is the same as another,” says Forsyth, who has never worked as a forecaster. ”It`s tough going out the next day, hearing that we didn`t give them enough warning. But if we don`t go out, we can`t do a better job next time.”
Forsyth drove off, and within minutes the devastation at the Yochams is in view. To a nonmeteorologist, the scene is pure chaos: a shattered television, an upended stove, and clothing, toys and kitchenware scattered in the mud. But Howard and Forsyth see order in the disorder. Most of the debris lay northeast of the toppled concrete blocks that once supported the trailers, suggesting that they were tossed by a weak tornado. Twenty-five feet from the flipped car was a relatively undamaged pickup truck; erratic and superlocalized damage is characteristic of twisters. Over all, says Forsyth, the damage is consistent with a small tornado, perhaps about 75 feet in diameter, embedded in intense, straight winds of 60 to 80 m.p.h. As he will later write in his report to the Weather Service, the tornado moved in a northeasterly direction and probably touched down only for about a mile.
The picnic prediction
Hand in hand with these efforts to better understand the genesis and development of storms are initiatives to improve real-time reporting of weather information to forecasters.
”We miss a lot of the important weather events because we are taking pinprick samples,” says NSSL`s Howard. Data collected at Weather Service offices around the country are relayed to the National Meteorological Center in Camp Springs, Md., where they are fed into a supercomputer that churns out national forecasts. The forecasts are whisked back to the field offices, along with charts of upper-atmospheric readings across the country. To make a detailed local prediction, forecasters digest the computer output, charts and satellite-and conventional radar images of the region.
The system produces reasonably accurate short-range statewide forecasts.
”Right now, when we say rain tomorrow in Colorado, we are right 7 times out of 10,” says Philip Merilees, a meteorologist at the National Center for Atmospheric Research (NCAR) in Boulder, Colo. But it falls down when it comes to predictions on a local scale, such as whether tomorrow`s picnic will be rained out.
A $1 billion overhaul of the National Weather Service promises to remedy many of the shortcomings of atmospheric sampling today. By the mid-1990s, a $700 million nationwide network of Doppler radars will replace today`s World War II-vintage conventional radars. A prototype for the new system, dubbed NEXRAD for NEXt-generation RADar, currently is being tested in Norman. With Doppler radar, forecasters should be able to predict a strong tornado as much as 20 minutes before it occurs.
A new weather workstation also promises to help forecasters with 3-to-12- hour predictions. The Program for Regional Observing and Forecasting Services, called Profs, has developed a prototype that is being tested in Denver. The workstation will give forecasters access to the new data, and allow them to overlay satellite or radar information with other images on one screen.
Some meteorologists wonder whether in the Profs era the flood of data will overwhelm forecasters. Experts are especially concerned by studies showing that when forecasters are given a lot of data, they are more confident of their predictions even when the predictions are no more accurate.
Criticism has also been raised about automating the weather-reporting network, which currently uses humans to relay the data; they often add important observational details, such as the presence of clouds on the horizon or the time thunder began. Such data would be missed by automatic sensors.
How accurate will forecasters get at predicting damaging thunderstorms?
Even the next-generation sensors will not catch everything.
According to Ron Alberty, director of the NEXRAD support facility in Norman, most tornadoes will escape detection in the Doppler era. While the really big twisters are the easiest to spot, 90 percent are too short-lived and diminutive to be picked up by a Doppler unless the funnel descends less than 25 miles from the radar.
The experts concur that they may never be able to forecast an individual storm much more than an hour in advance. The problem is the limits set by the atmosphere`s own complexity: The smaller and briefer the event, the less possible it is to forecast it far into the future. And on a meteorological scale, events like tornadoes are pretty small stuff.
The new atmospheric sensors, especially NEXRAD and the wind profilers, should help forecasters issue more-focused severe-storm warnings, however. Currently, the Weather Service alerts whole counties or large parts of counties.
That would do much to improve public confidence in warnings and help to ensure that they actually achieve their intended effect of limiting damage and injuries.
As it stands now, people frequently ignore warnings altogether or run outside to see what is going on instead of taking shelter, meteorologists included.
”There`s no way I`d run my family into a shelter every time NOAA issued a warning for our county,” admits NSSL Director Maddox. The real sign that forecasting has come of age may be when the weathermen start heeding their own advice.
