Why Weather Predictions Will Never Be Perfect

The blizzard predicted for the East Coast was mostly spot on, except for one area: Manhattan.

|
Jan 27 2015, 6:40pm

​Image: ​NASA 

​The massive winter storm that caused the National Weather Service to issue a blizzard warning for a large swath of the east coast, including New York City, arrived overnight. It pummeled ​New England and while Manhattan didn't get the two feet of snow that was predicted, there was a foot of snow at La G​uardia airport and over two feet in Long Island. There was a blizzard, it just didn't come quite as far west as it might have. 

This forecast was far from a complete blunder, but it wasn't bang on either. We've all witnessed more obvious forecast errors over the years. It's 2015, the Earth is orbited by dozens of weather satell​ites constantly beaming down information and when it doesn't rain, we can shoot particles into the clou​ds to force a downpour. So why do meteorologists still get it wrong sometimes?

Over the past few decades, weather prediction has become increasingly precise. In the 1950s, computer models were little more​ than an experimental tool. Meteorologists had to analyze all of the weather data themselves to calculate their predictions. As computers became more powerful, the models began to play a bigger role, but they still lacked precision. In the early 70s, high temperature forecasts made three days in advance were off by an average of six degrees, accordi​ng to the New York Times. Now, it's accurate within three degrees.

Modern meteorologists use multiple computer models that ingest trillions of data points from across the globe to project how the weather will most likely change in the coming hours and days, Keith Seitter, the executive director of the American Meteorological Society, told me.

Weather prediction will always be 95 percent science and 5 percent art

Most of that data comes from satellites which can provide information like the temperature at different layers of the atmosphere or wind direction and speed. More data is collected from land instruments like thermometers, barometers, radar, and wind gauges. Commercial aircrafts are also equipped with instruments that funnel data to the computer models.

Each model takes that information and produces a slightly different prediction, based on different assumptions that the models make about the weather. The meteorologist's job is to look at all those simulations, consider their local knowledge of the weather, and make their best prediction.

The process produces pretty precise forecasts most of the time, but the atmosphere always has an edge of unknowability, Seitter said, especially when looking further into the future.

"There's always going to be a little bit of unpredictability because we can't get observations for everywhere. No matter how good we get, there's always going to be some little bit of unobserved activity. Beyond a week or so, the atmosphere itself has intrinsic unpredictability," he said.

He told me with winter storms like the blizzard that hit the east coast, the difference between a lot of snow and a moderate amount of snow is very narrow. A slight shift in the storm's trajectory can mean the difference between a two-foot blanketing and an annoying skiff.

Forecasters likely knew there was a chance the storm would skim Manhattan rather than bludgeoning it, but there was a chance it wouldn't, and the risk of over-pr​eparing is less than the risk of under-preparing. Still, the narrow difference meant New Jersey was more or less spared: despite being told to expect as much as 18 inches, many places saw as little as three inches of snow. That was enough to make one meteorologist beg forgiveness:

Gary Szatkowski, a meteorologist with the National Weather Service in New Jersey, ​tweeted his reparation

And the reality is that mistakes will continue to happen. Seitter told me the computer models are constantly being upgraded, but even with the software improving year over year, weather prediction will always be 95 percent science and 5 percent art.

"Forecasters who have a lot of experience predicting a certain kind of storm are going to do a better job than you or I would with the same information," he said.

"And then there is just a little bit of art form in there. Some of the best forecasters I know look at things and they just kind of get a gut feeling that things are going to happen a little differently."