Texas tornadoes: How much warning time was possible? (+video)

The National Severe Storms Laboratory's Dr. Stensrud recalls a conversation with an emergency manager in Oklahoma in which Stenrud asked how much lead time the official would find useful.

“He actually gave me an answer of two hours,” Stensrud says. That lead time would allow officials to plan their responses in a more effective way, even if they ultimately don't act until much closer to a tornado's projected arrival time. Hospitals would have more time to move patients into hallways and away from windows. Sport stadiums would have more time to evacuate the tens of thousands of people filling the stands. And airports – no strangers to lots of large glass windows – would have more time to curtail operations and help passengers reach the safety of shelters.

Much of the technology such forecasts require already exists. Some of the biggest challenges involve integrating information from a range of sources – commercial aircraft, satellites, weather radar among them – and combining and displaying them in ways most meaningful to forecasters. Computer horsepower is another limiting factor, although Stensrud says researchers have had some success using computer processors designed for heavy duty, graphics-intensive use.

That computing power will become all the more important as a type of radar known as phased-array radar becomes a tool for weather forecasters.

Unlike current weather radar, which produces a complete sweep of the sky ever four to five minutes, phased-array radar completes its sweep in one minute – because it's radar beam is steered electronically from a fixed antenna array, rather than having an antenna turn mechanically.

When monitoring a storm, “a lot can happen in five minutes,” Stensrud says.

Testing the model

Up to now, the research team, which includes scientists from other National Oceanic and Atmospheric Administration labs and the University of Oklahoma, has tested its new approach against two major storms – a 2007 tornado that destroyed some 90 percent of Greensburg, Kan., and 2003 tornadoes that struck Oklahoma City.

The team fed the forecast model they developed with weather information available prior to the storms. And while the model can't reproduce features as small as a tornado itself, it did provide a forecast that yielded a high probability of a storm with rotation at low altitudes – often seen as a surrogate for a tornado.

For Greensburg, the modeled path the rotating segment took moved pretty much along the path the real tornado took, including its pass over Greensburg, Stensrud explains.

“That would have been on the order of a 40-minute advanced warning,” he says.

The team saw similar results for its “hindcast” of the Oklahoma events. The region of the modeled storm with the highest probability of triggering tornadoes closely matched the paths the real twisters took.

It would come as little surprise if Tuesday's twisters in the Dallas-Fort Worth area provide a further test for the new forecasting approach.

“Early indications are that it looks like we can do it,” he says of achieving the program's goals. But, he cautions, “I just don't know how reliably we can do it – for every event, every different type of event we'd like to handle – and make sure it's a robust system.”

It is likely to take a couple more years, he says, to reach a point where the team can use the model experimentally in real time, side-by-side with current forecasting approaches as severe-weather happens. That represents the final trial before a decision is made on whether to add the high-tech arrow to forecasters' severe-weather-prediction quiver.

Previous

1 | 2