The Fifth Risk(37)

The closest thing to an authority on the relative accuracy of various weather forecasts is a website called ForecastAdvisor. It began, as so much weather research seems to, almost by accident. Its founder, Eric Floehr, was managing a team of software developers and went looking for material on which to practice a new programming language. He stumbled upon weather forecasts—and a funny situation. All the forecasters were claiming to be better than each other: they couldn’t all be right. “When I started in 2003, the private weather companies—AccuWeather, for example—are saying,‘We’re the Number 1 forecast!’ So I called them and said,‘You make this claim that you are the most accurate forecast: what are you basing it on?’ They faxed me back an undergraduate paper written for a science fair that looked at forecasts for three months of one summer in Washington, DC. That was the best data they had to make that claim.”

Over the next thirteen years, Floehr collected eight hundred million weather forecasts. “I was curious. Was there really a difference?‘I live in Paducah, Kentucky. Should I look at AccuWeather or the Weather Channel?’” Lo and behold, there really was a difference. In the seemingly simple matter of predicting the high temperature for the day, some forecasters were better than others. None of them was consistently better all the time, however. Some were more accurate in some parts of the country than they were in others. Some were more accurate in some months of the year than in others. And there was no answering the question of who was better at tornado alerts or hurricane-track predictions or flood warnings, or at calling other life-threatening weather, because the private companies did not reveal their predictions of those events to anyone but their paying customers.

So Floehr analyzed everyone’s ability to predict the high temperature on any given day. From 2003 up until 2011, the National Weather Service’s forecasts had been as good as the most accurate private weather forecast, including AccuWeather’s. Since 2011, the private weather forecasters have been slightly more accurate than the National Weather Service. Still, says Floehr, “For sure I’m going to listen to the National Weather Service when they issue a tornado warning or a flash flood warning. I’m not going to trust right now AccuWeather or the Weather Channel.”

Floehr’s analysis uncovered two big trends in weather prediction. One was toward greater relative accuracy in the private sector—which of course was totally dependent on the National Weather Service data for its forecasts. The other was the astonishing improvement in all weather predictions. The five-day-out forecast in 2016 was as accurate as the one-day-out forecast had been in 2005. In just the last few years, for the first time in history, a meteorologist’s forecast of how hot it will be nine days from now is better than just guessing.

Barry Myers liked to say that he was in competition with the federal government. If so, the competition was bizarre: the U.S. Department of Commerce gave him, for free, most of the raw material he needed to create his product. Without the weather satellites, weather radar, weather buoys, and weather balloons, there would be no weather forecasting worth listening to, much less paying for. Whatever AccuWeather—and any other private weather forecaster—might be doing to refine the National Weather Service’s forecasts also depended on having those forecasts in the first place. “If the Weather Service forecast wasn’t there, all the private weather forecasts would get worse,” says David Kenny.

But the National Weather Service was forbidden by law from advertising the value of its services—and if it even hinted at doing so, Barry Myers could apply pressure on it in all manner of ways. AccuWeather might make any sort of wild boast it wanted to about the accuracy of its weather prediction. It might disparage the very people who supplied it with the information it had used to make that prediction. The meteorologists at the National Weather Service had no real ability or even inclination to respond. “We had to drag them kicking and screaming into defending themselves against false charges,” says a former Obama Commerce Department official. “They never claim credit. They always do these intensely self-critical how-can-we-do-better inquiries. It’s a public safety mentality: they do what they do because they really sincerely and since they were eight years old love the science and the service, not because they care at all about credit or glory.”

That was the sad truth—the public servants couldn’t or wouldn’t defend themselves, and few outside the U.S. government had a deep interest in sticking up for them. By the 1990s, Barry Myers was arguing with a straight face that the National Weather Service should be, with one exception, entirely forbidden from delivering any weather-related knowledge to any American who might otherwise wind up a paying customer of AccuWeather. The exception was when human life and property was at stake. Even here Myers hedged. “The National Weather Service does not need to have the final say on warnings,” he told the consulting firm McKinsey, which made a study of the strangely fraught relationship between the private weather sector and the government. “The customer and the private sector should be able to sort that out. The government should get out of the forecasting business.”

In 2005 Rick Santorum, a senator from AccuWeather’s home state of Pennsylvania and a recipient of Myers family campaign contributions, introduced a bill that would have written this idea into law. The bill was a little vague, but it appeared to eliminate the National Weather Service’s website or any other means of communication with the public. It allowed the Weather Service to warn people about the weather just before it was about to kill them, but at no other time—and exactly how anyone would be any good at predicting extreme weather if he or she wasn’t predicting all the other weather was left unclear.