As the summer heat waves dissipate and tropical storms start rolling in, the nation turns its eager eyes to the weather forecasters. Polls show that an accurate local forecast is consistently the second most cited reason for watching television news, even during the summer doldrums. Impending snowstorms and other types of “extreme weather events” bump the rating to No. 1. There isn’t a politician or entertainer or athlete in the world who wouldn’t kill for a fraction of this power to command the public’s attention. With the storm season upon us, it’s worth asking: Who are these weather forecasters? And how good are they, really?
The National Weather Service (NWS) and its counterparts in Japan, Europe and Canada use supercomputers to run “atmospheric models” that churn out data like temperature, humidity, wind speed and barometric pressure. At AccuWeather’s glass-box headquarters in rural State College, Pa., dozens of satellite-dish antennas on the roof and lawn silently take in these billions of numbers and pipe them to a cavernous room like something out of “Dr. Strangelove.” That’s where Abrams and his colleagues meet up each morning, “dark and early” as he puts it, to discuss what the day’s models are saying and how to translate them into forecasts that ordinary people can understand.
Joel Myers founded AccuWeather back in 1962, when he was a Ph.D. candidate in meteorology at Penn State, on the proposition that weather prediction is an art, not a science. He believed the output of government meteorologists could be improved upon. “Warren Buffett beats the S&P virtually every year,” says Myers. “It’s the same thing with weather forecasters. You have some who can and some who can’t. It’s a matter of skill and talent.” And picking the right weather model. An NWS model might nail hurricanes coming up the Gulf of Mexico, while a Japanese model better handles thunderstorms in California. Abrams and his colleagues still crow about their deft juggling of the models during Hurricane Floyd in 1999, when they used Europe’s data to forecast the storm’s trajectory more accurately –than the Weather Channel, their archrival.
AccuWeather meteorologists serve as weather personalities for hundreds of radio stations across the country. The firm supplies predictions and graphics to television stations and newspapers. Beating the competition is paramount. Perhaps that’s why meteorologists at commercial firms like AccuWeather are so prickly about their public image. Everybody listens to long-range weather forecasts, but few believe in them, and fewer still keep score on who’s most accurate. “I think there’s a mass case of attention- deficit disorder,” says AccuWeather meteorologist Joe Bastardi. “Every single day somebody will say, ‘You guys got it wrong again.’ I want to say, ‘Go back and look at what we said’.”
Bastardi’s job is to look ahead of the computer models, probing as far away as Asia for patterns that will inform events weeks away in North America. “I’m always thinking about the weather,” he says. An innocent question can elicit a 30- minute torrent of prognostication. “This warm water sitting in the northwest Atlantic is very similar to what we saw in ‘99,” he says. “Not exactly the same, but similar. With warm water here”–he points to a world weather map–“an intensification of the drought in this area, and we’re getting an overriding signal… "
“We’re getting off the point,” interrupts Myers, who’s been looking on like a benevolent raptor. “The thing is, a lot of meteorologists are lazy. They rely too much on the models. Very few look far beyond the United States like Joe does.”
Back in August, Bastardi was flagging developing storms in the Atlantic that gave rise to rainfall in the Northeast in early September, and more recently to Hurricane Isidore, which late last week was heading into the Gulf of Mexico. Nobody else, he says, had seen it so early. If true, it would lend credence to Myers’s claim that AccuWeather can forecast 15 days away “with some skill.” But storms can be tricky. Last week Bastardi was unable to say what Isidore was likely to do in the next few days–move north into Texas or west to Mexico, or stall in the gulf? “The whole thing can be blown with one bad forecast in the end despite 14 good ones before,” he writes in an e-mail. “The sad thing is that the public only comes in at the endgame.”
So how accurate are the forecasts anyway? That depends on whom you ask. “Our customer-renewal rate is 98 percent, so we must be doing something right,” says Myers, who is, after all, running a business. The NWS uses a different metric. Every so often its meteorologists take a map and color in those areas where they forecast heavy rains. Then they take another map and color in the areas where those rains actually occurred. Comparing the two, they find that they’re correct about a third of the time. Beyond six or seven days, they don’t even bother. Bottom line: the fastest computers and the smartest guys still can’t see more than about a week ahead. Sorry, Joe.
