Forecasting model variations add to this week's uncertainty
Creating a weather report is a scientific process that’s much more involved than just a series of colorful maps splashed across a television screen.
Take this weekend’s storm, blowing through Grand Forks with the potential to dump heavy snow across the region. Rain arrived Wednesday evening and lasted through Thursday, Oct. 10, as temperatures dipped into the mid-30s, according to the National Weather Service. Snow was expected to begin Thursday night and could continue in some places through Sunday, with the temperature remaining in the mid-30s through the weekend. And while nearly everyone could see that a storm was coming, weather professionals were being cautious about detailed predictions in the days before it hit.
"The knowledge that snow is a possibility is good to know," WDAY meteorologist John Wheeler tweeted Monday evening. "However, WDAY Weather WILL NOT be talking about snow or rain amounts or wind speeds of a potential late-week storm until it is appropriate to make a legitimate forecast and not a wild guess."
Looking at the forecast — steadily becoming more detailed and polished — it’s easy to miss how much goes into making it happen.
Various numbers, equations and data go into predicting the region’s weather. Brittany Peterson, a National Weather Service meteorologist based in Grand Forks, points out that part of what makes this weekend’s weather tougher to predict is the longevity of the system. It’s spread out over several days, which means potentially large temperature variations. Perhaps the most notable aspect of the forecast is that the National Weather Service wasn’t ready earlier this week to say exactly how many inches of rain or snow would fall this week — an offshoot of a cautious approach to forecasting that acknowledges its uncertainty.
Much of the forecasting that meteorologists rely on comes from predictive models, which are a big part of how weather is forecast around the world. There’s an alphabet soup of them — from the Global Forecast System, or GFS, which includes forecast modeling from the U.S. government, or the “European” modeling done on the other side of the Atlantic. Most writing on the topic suggests the models all provide valuable information, but that they differ in important ways.
Notably, this occurred in 2012, when Hurricane Sandy threatened the United States. American-model data suggested Sandy would sputter. A European model correctly predicted it would strike the East Coast. A report in Scientific American explains the difference between the two models as a difference in the amount of data collected, simulations run and the amount of detail that computer models process.
In Grand Forks, and with the latest storm, there are some similar disparities . A German weather model on Wednesday predicted it will be 33 degrees in Grand Forks at noon Friday. The American model said it will be 32 degrees; the European model, 35 degrees.
As Peterson explains, different models rely on different math that emphasizes different weather statistics — and when meteorologists are making predictions about local weather patterns, like this weekend’s storm, they’re looking at more than just one model.
“Not only do we look at what each one says, but we’re also looking at things probabilistically,” Peterson says. “So from those models, we can pull out the likelihood of, say, the snow falling, or the likelihood of things being below 32 degrees.”
Of course, there are still things that can’t be predicted with total confidence, Peterson said. But she reminds local residents that they’re in for a wet few days — and should prepare accordingly.
“Since we’re getting more rain, you’ve probably seen a lot of people’s sump pumps as well,” she said earlier this week.