People love to hate weather forecasts, though it’s getting a lot harder to find fault. Forecasts gave plenty of advance warning that Chicago would see bitter highs of around minus 12 last week, and lows comparable to a bad day in Antarctica. Public officials closed schools and issued warnings, and police saved lives by combing the streets for homeless people before the worst hit.
Today, a five-day forecast is just as accurate as a one-day forecast was in 1980, giving us more time to prepare — or overreact and panic. Weather watchers in the Northeast saw last week’s cold snap coming for days. Storms are now forecast within a range of 50-some miles and timed to within a couple of hours. An icy storm that hit the Northeast on Jan. 20 was in the forecast before it even existed anywhere, said Richard Alley, a geosciences professor at Penn State University and co-author of a new paper in Science on weather prediction.
There’s even more room to improve, he said. Better forecasts save money — on plows and road salt, on crops, and on making sun and wind energy more predictable. Sometimes forecasts can have even bigger consequences.
Historians say D-Day might have gone very differently if weather had followed the German forecast for storms rather than the Allies’ more accurate forecast of fair skies and calm seas. Bad weather thwarted Napoleon and later Hitler in attempts to defeat Russia. Lost explorers over the ages might have planned differently. Ernest Shackleton might have avoided getting his ship locked into the Antarctic ice.
It makes sense that weather is inherently more predictable than, say, the economy. Ultimately, weather does reduce down to basic laws of physics — Newton’s laws, the Coriolis effect, and the like, said Alley.
The limiting factor is the complexity of the system: To build a good model you have to take into consideration where there’s ocean, where all the mountains are, and even where corn fields grow in the Midwest.
Alley said that indeed there’s a limit on the accuracy and range of weather forecasting, but since the 1970s, it went from mediocre to amazing. He chalks up the improvements to additional satellites and other data collecting devices, more powerful computers to crunch the data, and improved models.
He uses the game show “Wheel of Fortune” to show how forecasts work. The wheel is divided unevenly into segments, so right off you can calculate how often over many spins it will hit, say, the skinny million-dollar-jackpot segment. That’s why long-term climate patterns are easier to forecast than the weather on any given day.
A single spin, or a single day, is hard to predict, but as the wheel starts to slow down, you can start to guesstimate where it will land — just as the weather forecasts for a particular day are more accurate as the day comes closer.
In a time when forecasters of all types make overconfident proclamations about political, economic or natural events, uncertainty is a tough sell. It’s much easier to hawk overconfidence, no matter if it’s any good.
That’s a longstanding problem for many kinds of scientists. People might wrongly equate honesty about uncertainty with being wishy-washy. But there are objective tests of predictions. So-called “superforecasters” (who win forecasting tournaments) say that understanding their own uncertainty is critical.
People probably won’t stop complaining about the weather forecasts, said Alley, because even missing the mark by a degree or so can mean the difference between snow and rain. It’s also improved so slowly that it’s been hard to see. What’s all this worth in dollars and cents? Alley cited several papers, listing the benefits of weather forecasting and estimating that spending on weather forecasting technology can deliver a much greater return on investment. But to get specific there, you’ve got to factor in humans and the economy, which are not nearly so predictable.
Faye Flam is a Bloomberg Opinion columnist.