Weather Service upgrades its forecast model

The new version should help better forecast heavy snow events, downpours in Northland.

A Duluth resident snowblows their driveway after a 22-inch snowfall in December 2019. The National Weather Service this week upgraded its primary computer forecast model that officials say will better predict heavy snowfalls and downpours. (Tyler Schank / File / News Tribune)

The National Weather Service has upgraded its often-maligned computer forecast model to better predict extreme weather events such as winter storms, thunderstorm downpours and hurricanes as well as improve day-to-day forecasts.

The Weather Service update to its Global Forecast System is an effort to catch up with a European weather model that many forecasters consider more accurate.

The new Weather Service model has been under tests for two years and started full-time duty Monday. Officials said it has forecast heavy snows and rainfall 15% better five days out and improved hurricane and tropical storm tracks by more than 10%, better pinpointing storm formation up to a week in advance.

One of the key features is that the Weather Service model does not over-predict rain and snow amounts, which the previous version often did.


An all-wheel-drive Duluth city grader struggles to move snow from a heavy snowfall in December 2019. The National Weather Service says its new forecast model will help better predict big snowstorms, which may help municipalities better plan and deploy snow removal equipment. (Clint Austin / File / News Tribune)

“We’ve had the chance to monitor a parallel feed of the data, old and new, and we have definitely seen some improvements,’’ said Joe Moore, warning coordination meteorologist for the National Weather Service in Duluth.

Moore said the old model had a cold bias in winter and not only often over-estimated snowfall but would sometimes predict snow when rain would fall.

“The new model is much more accurate at deciphering temperatures at the lower levels, near the ground, and detecting correctly if it’s going to be rain or freezing rain or snow,’’ Moore noted. “No one is going to care if we nail the 10,000-foot winds dead-on. They want to know what’s happening on the ground because that’s where most of us live.”

Moore noted that each Weather Service office already uses an amalgam of forecast models, including the Global Forecast System, European Model, North American Model, Canadian models and a new blended analysis called the National Blend of Models.

“And the new GFS data was being fed into those blends starting at 7 a.m. Monday, so it’s already being used,’’ he said. “We also have people here with experience at interpreting the data, and nothing really beats that.”

Weather Service studies also showed the new model was generally more accurate earlier on at forecasting downpours in the Southeast last winter, Hurricane Dorian in 2019 and Hurricane Michael in 2018.

One key upgrade: The new model captures the atmosphere up to 50 miles high and has higher resolution than the old version. That helps forecasters better see the jet stream, which is how storm-producing weather systems move across the country.


“This substantial upgrade to the GFS, along with ongoing upgrades to our supercomputing capacity, demonstrates our commitment to advancing weather forecasting to fulfill our mission of protecting life and property,” said Louis W. Uccellini, director of the National Weather Service, in a statement. “Today’s upgrade also establishes a strong foundation for further planned enhancements that will allow for the assimilation of even more data into the model.”

Weather Service officials said that in the past month there was a stretch of eight straight days when the new version beat the highly touted European forecast model. But they noted that the European model still won more often.

John Myers reports on the outdoors, natural resources and the environment for the Duluth News Tribune. You can reach him at
What To Read Next
The system crashed earlier this month, grounding flights across the U.S.