Washington D.C., Philadelphia and Baltimore were braced for big snow, but got only inches. Meanwhile, regions farther north and east where predictions were mild, are now being pummeled.
What happened? James Hoke, director of the National Oceanic and Atmospheric Administration's Hydrometeorological Prediction Center, explains that sometimes storms can stump forecasters — especially when it comes to snow. Still, in this case, he argues, forecasters were more or less on the mark.
"Snow forecasting is arguably the most difficult, most complicated thing for forecasters to predict," Hoke says.
To predict a snowstorm, Hoke says, forecasters need to answer endless questions, including when it will start, when it will stop, how hard it will snow, what temperatures will be on the ground and in the atmosphere and whether rain, sleet or snow falls from the sky.
"With so many different factors," says Hoke, "we can't expect to hit them all."
There are a number of possibly misleading factors. Weather data, for example, can be inaccurate or patchy. Using the wrong weather models can skew predictions. And, more commonly, unexpected factors can arise and effectively throw off predictions.
Weather forecasts are made by dividing the atmosphere into a three-dimensional grid and then inserting weather data into each square of the vast grid. Meteorologists then use powerful computers to crunch numbers within each grid by applying what Hoke calls "the most complicated equations in science."
These equations, or models, have been devised by generations of meteorologists who have watched the effects of factors like wind direction and strength, barometric pressure, evaporation rates, Earth temperature and temperatures above the Earth.
"It's like a giant math problem that's constantly changing," says Ralph Roskies, scientific director for the Pittsburgh Supercomputing Center. "And the computer does the arithmetic."
Once weather is calculated within each piece of the grid, the computers determine how one region will affect weather in neighboring grids. The smaller the grid size, the more detailed a forecast will be.
Today, thanks to more powerful computers that can make up to 140 billion calculations per second, the weather service can calculate weather data using grids only 7-8 miles apart, according to Kevin McCarthy, deputy director of the Hydrometeorological Prediction Center.
"The 5-day temperature forecasts we make now have the same accuracy our 3-day forecasts did 20 years ago," says Hoke.
Data has also gotten better. Nearly all large passenger airplanes, for example, are equipped with automated weather instruments. These instruments record information like wind and air pressure and temperature as they take off and land and relay the readings to national weather stations.
Despite the improvements, forecasts are never flawless. One reason is some weather equations must be simplified before computers may easily work with them. By simplifying equations, they're made less accurate.
Bad data can also throw off predictions. An incorrect temperature reading, for example, would not only skew a short-term forecast, it could also significantly alter long-term forecasts.
And even though weather services are gleaning more data from sources like airplane instruments, weather balloons, radar and satellites, there are still locations where data are spare.