Scott Singleton had just moved with his daughter from Ohio to a three-bedroom home in Lakeside Estates in Kissimmee, Fla. He had seen the tornado-watch warnings on television Feb. 22. But like most residents here, he didn't expect it would hit in the middle of the night - and certainly not his neighborhood.
As Mr. Singleton went to bed, he could hear the staccato of rain and the booming thunder. Then the rain turned into hail, hitting his house like "little baseballs." Still, he was not thinking tornado. But at about 1 a.m. he was awakened by the noise of shattering windows and a "moaning, chugging train" that shook his house to the foundation.
Singleton says he did not have time to think about what to do. "It was real quick," he says. "And then it was all over."
Quest for earlier warning
Unlike hurricanes, which build relatively slowly and cut a wide swath of devastation, twisters are complex, highly localized phenomena that spawn suddenly. Getting a timely and precise tornado warning to residents is one of the most difficult challenges faced by US weather forecasters.
As residents of five counties in central Florida pick up the pieces from this week's tornado outbreak, researchers are pushing powerful computers to their limits in an effort to improve the lead time for severe-weather warnings.
The goal is to develop forecasting tools that will allow the National Weather Service (NWS) to pinpoint where the most intense part of a storm will hit 6 hours before it arrives. In addition, researchers hope to predict the type, location, and intensity of new thunderstorms up to two hours before they form.
Although still experimental, the effort is "starting to show promise," says Gary Woodall, who coordinates regional warning efforts at the NWS southern regional headquarters in Ft. Worth, Texas.
The tornado outbreak in central Florida, which killed at least 38 people, served as a tragic beginning to "Florida Hazardous Weather Awareness Week," which was to include a statewide tornado drill.
Instead of a drill, residents and state and local emergency-management officials find themselves confronting a formidable cleanup job. Singleton's losses were relatively minor: a front door, a steel mesh fence, and some windows. But about 300 of his neighbors were left with little more than rubble where their homes once stood.
Compared with tornado-forecasting efforts along Tornado Alley in the Midwest, "in many respects, forecasting tornadoes is an easier problem with big storm systems" such as the one that moved through Florida, says Morris Weisman, a researcher at the National Center for Atmospheric Research in Boulder, Colo. Because the storm system is so intense, the conditions that foster tornadoes are more obvious.
"But 75 percent of tornadoes occur in environments where conditions are not as evident as they were in Florida," he says. Last May, a powerful tornado hit Jarrell, Texas. It came as a surprise because it came from a weak storm system.
The foundation for the outbreak in Florida was laid on the morning of Feb. 22, when a series of thunderstorms moved from west to east along Florida's panhandle, forecasters say. The storms left rain-cooled air in their wakes, forming a mini-cold front that eventually sagged into central Florida. The boundary line between the cold air and warmer air became the racetrack for the tornado-spawning thunderstorms, embedded in a larger storm system moving in from the Gulf of Mexico.
As the thunderheads grew, they began to rotate, triggering alarms at forecast offices when their radar picked up the rotation. A tornado watch was issued at 8:15 p.m., says Paul Janish, a meteorologist at the NWS Storm Prediction Center in Norman, Okla.
But like many Florida residents, Esther Bilone, a young mother who works at DisneyWorld in Orlando, Fla., had gone to bed early, oblivious of the warning. When the tornado hit, "It seemed like a huge explosion," she recalled. "I looked out the window and I saw cars upside down, people screaming, and screen doors flying by."
Although the National Weather Service issued warnings that in principle gave residents from 10 to 30 minutes' lead time, many people, like Ms. Bilone, had gone to bed by the time the tornado "watches" were upgraded to "warnings." The timing raises a difficult question, Mr. Janish says. "Many people on the East Coast watch the 11 o'clock news before going to bed. Yet the tornadoes began to touch down 30 minutes after the news ended. How do we emphasize the severity of the situation? How will people respond?"
Much of the challenge stems from the computer programs the NWS runs twice a day to help its national and regional forecast offices prepare their forecasts. These demanding forecast models treat the United States as if it were covered with a checkerboard of squares 30 kilometers (18.6 miles) on a side. The programs calculate weather conditions at each point where four corners meet and continue to crunch numbers to show how those conditions are likely to change over time. Yet many forms of severe weather - especially tornadoes - occur in areas much smaller than 30 km.
Shrink the checkerboard
When tornado watches are issued, the area covered by the advisory is large compared with the size of the expected weather disturbance. Until something shows up on radar, residents in the area are unsure about how concerned they should be.
One answer, researchers say, is to shrink the size of the squares on the forecasting checkerboard. Scientists with the Center for Analysis and Prediction of Storms at the University of Oklahoma at Norman, for example, are refining a computer model whose forecast points are as little as 1 km apart horizontally and 500 meters apart vertically. "The ultimate goal," Janish says, "is to be able to say that a line of storms will move through a particular county six hours before the storms even develop."