Over the past few weeks, with the official start of winter still approaching but frigid weather already sweeping across much of the country, meteorologists everywhere from the National Oceanic and Atmospheric Administration to local television news to private services chimed in with their forecasts for the coming months. The people that understand forecasting best, however, don't take those long-range predictions as gospel. Far from it, in fact.
It wasn't that long ago that the most certain thing about the weather forecast was the widespread sense that it would be wrong. But over the past few decades, advances in computers and a better understanding of the science behind the weather have made short-term forecasts, defined roughly as 10 days or less, far more accurate than they used to be.
Peter Neilley, vice president for global forecasting at The Weather Co., the parent of The Weather Channel, notes that a five-day forecast issued today is as accurate as four-day forecast 10 years ago and a three-day forecast was 20 years ago. "It's been a subtle evolutionary trend and it continues to this day," said Neilley in phone interview from his office in Atlanta, adding that the change has been "slow enough that people may not notice it."
This improvement has saved countless lives by giving people advanced warning of serious weather events such as Hurricane Sandy. That makes the job of emergency managers, who need to prepare communities for disasters, easier. As Bryan Koon, head of Florida's Division of Emergency Management and president of the National Emergency Management Association, noted: "We save more lives than we lose them these days."
As short-term forecasts improve, though, demand is also rising for longer-term forecasts that remain far less accurate. "The market always wants more than the science can deliver," said Keith Seitter, the executive director of the American Meteorological Society, in an interview. "There is a lot of press on a day-to-day basis to push the forecast to the absolute limit of what that forecast can do."
Matt Rogers, the president and co-founder of the Commodity Weather Group, which provides forecasts to the energy and agriculture sectors, told The Fiscal Times that he offers clients 11-to-15-day forecasts even though he says his confidence in their accuracy is between 10 and 30 percent. That's far less than the 60 to 80 percent faith he has in his shorter-term forecast of about six to 10 days. What is especially problematic, according to Rogers and other meteorologists, is doing longer-term estimates of estimates of precipitation such as snowfall.
That demand comes from a variety of clients, such as governments looking to plan their road salt budget, energy companies trying to better anticipate additional stress on power plants, or businesses looking to manage their inventories. For example, Walmart, the world's largest retailer, counts on forecasts to make sure that its stores in a particular region are stocked with enough supplies to weather a particular storm. The company, though, has to make sure that it doesn't overreact and send too much seasonal merchandise such as shovels to a region where the weather is going to warm up.
"Our goal is to get the merchandise there within 24 to 48 hours," said Lucas McDonald, Wal-Mart's senior emergency operations manager, in an interview.
When longer-range forecasts started in the 1980s, AccuWeather founder and president Dr. Joel N. Myers likened them to "witchcraft" and refused to provide them to clients since they were so inaccurate. As technology improved, the company changed its tune. It now offers outlooks as far as 45 days out because they provide value to clients who know that they are not as reliable as shorter-term outlooks.
"Some clients are smart about these types of things and understand how they can use that type of information," said The Weather Co.'s Neilley. "There are others at the other end of the spectrum that require lots of hand holding."
Meteorologists try to account for the differences in reliability when crafting their forecasts over different time periods. The shorter-term forecasts take a more deterministic approach, widely used in the media, for estimating temperature and precipitation. Longer-term outlooks are probabilistic and speak of more vague concepts such as "above normal" and "below normal."
Even with those more nebulous parameters, the National Oceanographic and Atmospheric Administration wasn't able to make a winter forecast for a large portion of the country using those parameters, noting "that there is not a strong enough climate signal for these areas to make a prediction, so they have an equal chance for above-, near-, or below-normal temperatures and/or precipitation." A NOAA spokesperson couldn't be reached for comment.
"I don't personally put a whole lot of stock in them," said David W. Titley, a professor of meteorology at Penn State University and the director of the Solutions to Weather and Climate Risk, who likened them to a "slightly weighted coin flip." "This is science that has a long way to go."