With less than 48 hours left until the start of a potential major storm, what should meteorologists do when the computer models they rely on are wildly inconsistent?
Forecasters are in the most uncomfortable position Friday as two of the leading computer models, the U.S. Global Forecast System (GFS) and the European model, as well as others, continue to debate the strength and path of the mid-Atlantic and New England winter storm that begins Sunday.
The difference between the two major models and other forecast guidance is the difference between a paralyzing snowstorm along I-95 and a nearly forgettable light snowfall.
Given the devastating snowstorm damage in major cities between Washington, D.C. and Boston, accurate predictions could be worth hundreds of millions of dollars or more in this situation. Not being prepared for a major storm can be even more costly.
In this case, the reason the computer models’ solutions are so different can be traced back to subtle changes in how each handles the portion of atmospheric energy that dives south from Canada and ripples along the southern jet stream near the Gulf Coast, into areas of high pressure over the Rocky Mountains, or along jet stream bulges.
In the case of weather, the most subtle differences in the depiction of the same atmospheric features can mean the difference between an all-out snowstorm and light to moderate snowfall in populated areas from Washington, D.C., to Boston. Once you have all the ingredients in place, boom! Big storm on the east coast. If only one or two of the features are off in timing or location, a major storm is much less likely to occur.
Over the past few days, GFS has been claiming a near-record snowstorm in Washington, DC and other parts of the mid-Atlantic, with a veritable bomb cyclone tracking north-northeast near the coast.
Euromodels, on the other hand, consistently show that storms have a much weaker impact farther from the coastline, and that much less snow falls in large cities.
This difference could not be more obvious and is unlikely to be explained by National Weather Service staffing reductions associated with DOGE or the resulting lack of weather balloon data. After all, the GFS and the euro are making predictions based on the same observations, yet they are predicting fundamentally different outcomes.
Early Friday morning, Euromodels predicted the event would bring about 4 inches of snow to Washington, D.C., while the GFS model predicted 18 inches of snow.
The forecast nightmare is even worse in Philadelphia, where the GFS shows a whopping 29.2 inches of snow, while the Euromodel shows just 1.7 inches of snow.
The advent of AI models gives forecasters more tools to make accurate predictions, but this forecasting challenge remains. Prediction is much more complex than simply choosing a model and spitting out what the model says. If that were the case, there would be many more forecasters and more inaccurate predictions.
To create accurate forecasts, meteorologists need to understand why a model is showing the way it does, and they need to use their unique expertise to consider far more factors than just these two competing models.
Cody Snell, a meteorologist with the NOAA Weather Prediction Center, said experts are looking at all possible data and focusing their messages on the storm’s potential impacts rather than specific snowfall amounts. Many good meteorologists do this, so forecasts often show a potential range of snowfall amounts rather than specific numbers. A good forecaster will tell you what the uncertainties are and how confident they are in their predictions.
“We look at every model available, every ensemble system, and the answer is usually somewhere in the middle,” Snell says. “To create a good message about a storm, we need to leverage all the guidance available. And that’s the important thing: create the right message for a specific snowfall amount.”
He said he and his colleagues avoid getting caught up in computer model battles and keep the big picture in mind.
“Obviously we need certainty. We want all the models to agree and we can say with confidence that this is the scenario that’s going to happen,” Snell said. “But any time there is a difference, especially with this population along the Northeast, you have to consider all possibilities.”
Also taking into account the current situation is the track record of two major war models. Euro and its AI-powered model have performed much better than GFS this winter, leaving many with deep doubts about the blockbuster scenario of the US model.
Typically, when GFS and the euro disagree, forecasters favor the euro due to past performance.
However, despite these questions, no other recent storm better represents model differences that can persist for almost the entire storm onset time.
In situations like this, meteorologists tend to split the difference between the two modeling camps, calling for light to moderate snowfall, but with more snowfall possible. But in such a situation, the public message could be garbled by people depriving themselves of large areas of snowfall, say 4 to 18 inches, even though experts view the 18-inch cap as a low-probability, high-impact outcome.
