Surprised by how tough this winter has been? You’re in good company: Last fall the Climate Prediction Center of the National Oceanic and Atmospheric Administration predicted that temperatures would be above normal from November through January across much of the Lower 48 states. This graphic shows just how wrong the official forecast of the U.S. government was:
The big red blotch in the top map represents parts of the country in which the Climate Prediction Center forecast above-average temperatures. The frigid-looking blue blotch in the bottom “verification” map shows areas where temperatures turned out to be below average.
“Not one of our better forecasts,” admits Mike Halpert, the Climate Prediction Center’s acting director. The center grades itself on what it calls the Heidke skill score, which ranges from 100 (perfection) to -50 (monkeys throwing darts would have done better). October’s forecast for the three-month period of November through January came in at -22. Truth be told, the September prediction for October-December was slightly worse, at -23. The main cause in both cases was the same: Underestimating the mammoth December cold wave, which brought snow to Dallas and chilled partiers in Times Square on New Year’s Eve.
The winter has stayed cold in 2014, and snowfall and snow cover are way above average. USA Today reported on Feb. 14 that there was snow on the ground in part of every state except Florida. That includes the peaks of Mauna Loa and Mauna Kea in otherwise balmy Hawaii. (The Climate Prediction Center predicts temperature and precipitation but not their spawn, snow.)
The cause of the tough winter, as is well known by now, is the polar vortex, which has been distorted by a strong and persistent ridge of high pressure over the eastern Pacific and the west coast of the United States. First, it’s causing California’s drought. Second, polar winds are flowing northward around the high-pressure ridge and then plunging down along the ridge’s backside.
What the Climate Prediction Center hasn’t been able to figure out is why that ridge has maintained its position so persistently. “Sometimes trying to figure out why something happened is as hard as making the forecast of what will happen,” Halpert said in a Feb. 14 interview.
The techniques that go into three-month climate forecasts are completely different from the ones used for daily weather forecasting. The daily stuff, says Halpert, is based on “initial conditions”—i.e., measuring as precisely as possible the current state of the atmosphere and predicting how it will evolve over the coming few days. Regardless of what Siri (AAPL) might say, weather forecasts have very little accuracy past a week.
Months-ahead climate forecasts are what Halpert calls “boundary-value problems.” Instead of taking a snapshot of the quickly changing atmosphere, climate forecasters focus on things that change more slowly, such as temperatures of the land and oceans. The concept is that these enduring conditions will “force” the daily weather in a certain direction. Climate forecasters can’t say when a storm might hit, but they try to say whether a given three-month period will be wetter, drier, hotter, or colder than average. (Here’s a good description of the difference between weather and climate forecasting from University of Toronto computer scientist Steve Easterbrook.)
Climatologists are trying to use their big miss this winter as a learning experience. What if anything does the high pressure ridge over the West Coast have to do with the pattern of sea-surface temperatures in the Pacific Ocean? It’s tempting to use statistics to solve the problem: If B is almost always preceded by A, then B must be caused by A. But the Climate Prediction Center is “looking for physical links, not statistical ones” to establish causality, says Jon Gottschalck, acting chief of the center’s operational prediction branch.
Meanwhile, as the hunt goes on, Gottschalck says, the Climate Prediction Center’s December-through-February temperature forecast so far is “floating toward zero-ish.” On a scale of -50 to 100, that has to be considered an improvement.