There are many weather blunders—some are world-class! Here are three examples of how computers confused the forecasters and confounded the public.
Dozens of Tornadoes Take Oklahoma by Surprise
On May 3, 1999, forecasters expected damaging wind and hail over central Oklahoma. "But looking at the model forecast at 6:00 a.m., there was no indication that the event that would unfold 12 hours later would be as intense as it was," says Stephen Corfidi, a lead forecaster with the Storm Prediction Center in Norman, Oklahoma. "It wasn't recognized until relatively late, in mid-afternoon, that there was potential for a high-end tornado event." All told, 70 tornadoes left 40 dead and over $1 billion worth of damage.
Forecasters learned an important lesson from this—even without obvious red flags, tornadoes can form if the conditions are just right. In this case, one of these ingredients was contrasting wind speed: Gusts at 10,000 feet were at 45 knots, while lower winds were moving at 35 knots—not a lot of difference, but apparently enough. Corfidi says that the model data on that day certainly didn't fit a "blueprint." "But everything in unison added up to a sweet spot where it all came together."
When the profiler network of radars (these point directly skyward to measure wind speed and direction) revealed a strong frontal system coming from the southwest, tornado warnings were issued. "The radar told us the wind speed was stronger than had been originally forecast," says Bob Gall, a scientist with the U.S. Weather Research Program, based in Silver Springs, Maryland. Unfortunately, the warnings weren't issued soon enough.
Blizzard Predicted, Flurries Arrived
Some say that the much-ballyhooed blizzard predicted for the East Coast on December 30, 2000, was a prime example of over-reliance on technology-in this case, an "insistent" computer model.
On the eve of said blizzard, Storm Prediction Center forecaster Stephen Corfidi was visiting relatives in Baltimore, unaided (or as it turned out, unencumbered) by technology. "I was outside raking leaves, and it was clear from looking at the clouds that their motion was still westerly. Until you get easterly turning of the winds, you have nothing to worry about," he recalls. "I was telling folks, 'I just don't think it's going to snow.'"
Corfidi acknowledges that it was easy for him to buck the status quo on his day off. "If the computer says 'blizzard,' it takes guts to go against it. That subtle pressure takes the thinking out of the forecast situation, which isn't good," he says.
The egg-on-the-face outcome (there was nothing to shovel) underscores a simple truth: Personal observations and instincts can sometimes beat even the best technology.
River Crests Higher Than Predicted; Floods Result
In 1997, when forecasters using computer models predicted that the Red River, which flows north along the border of Minnesota and North Dakota, would crest at 49 feet, people believed them. "And it turned out to crest at 54 feet," says Roger Pielke Jr. of the University of Colorado in Boulder. "This is one of the flattest locations on the surface of the earth. River basins are like bowls, but the Red River basin is more like plates. So you had water flowing from one plate to another, which wasn't in the model." The devastating flood caused $2 billion in damage.
By scientific standards, the prediction was considered accurate (given a 10 percent error margin), but the public expected dead-on accuracy. "People are a little bit in awe of what scientific technology can do. They need to understand our limitations," says Pielke, pointing to the official "No Surprises" motto of the National Weather Service. "In my opinion, that sends exactly the wrong message," he says. "People should expect surprises."