New weather forecasting system goes into action; Wisconsin snowstorm is used to test it

When the National Oceanic and Atmospheric Administration began upgrading its weather forecasting system, some meteorologists criticized it for vastly over-predicting snowfall.

So NOAA scientists continued to refine and analyze the next-generation weather prediction model, called GFS-FV3, by testing it against actual storms to see how it performed.

Among those storms providing essential data was one of the late spring snowfalls that dumped a surfeit of flakes throughout much of central and northern Wisconsin.

So, in a way, one of Wisconsin's seemingly endless snowfalls this year helped improve weather forecasting for everybody.

An April 12 storm that hit much of the top half of Wisconsin was used to determine whether the new forecasting model was performing better at predicting snowfall and precipitation, Brian Gross, director of NOAA's Environmental Modeling Center, said Wednesday.

On Wednesday, forecasters throughout the U.S. began using the upgraded version of the country's flagship weather forecast model. It's the first major change in almost 40 years to the model's method of computing wind and air pressure.

Scientists say GFS-FV3 will provide more accurate forecasts of precipitation and temperature. Plus it's expected to be better at predicting heavy rainfall and tracking the path and intensity of tropical storm systems such as hurricanes.

Before rolling out the model, scientists were getting serious blowback this winter from meteorologists following the progress of testing the new model. They criticized it for inaccuracy, particularly because of a bias toward predicting more snow and colder temperatures than what actually happened.

"We used this particular (Wisconsin) case to evaluate upgrades to address some of those issues," Gross said in a phone interview with the Milwaukee Journal Sentinel.

The snow that fell in Wisconsin on April 12 was part of a storm that pummeled states in the Plains, with blizzard conditions before moving into Minnesota and then Wisconsin. Wausau got close to a foot of snow. Elsewhere, 7 inches piled up in Eau Claire, Wisconsin Rapids got around 8 inches and snow gauges in Rhinelander recorded 6 inches.

Two days later, 9 inches of snow fell in Union Grove, 7 in Caledonia and 6 in Franklin as another storm system blew through southeastern Wisconsin.

After the criticism, Wisconsin's April 12 storm was chosen as a case study because it was timely, said Gross.

Scientists checked meteorological data in Wisconsin on April 10, 11 and 12 in the run-up to the storm and the day of the snowfall and used the information to test changes being made to the model.

"In this case we saw an over-prediction of snow with the original upgrade, and we saw a more accurate prediction of snow with the changes," Gross said.

It wasn't just one Wisconsin snowstorm, though. The forecasting upgrade underwent rigorous testing, including real-time evaluations for a year of weather alongside the previous version. The model was also tested against historic weather dating back an additional three years.

For years, the American weather forecasting model has lagged in accuracy behind the European model. But that's changing with the model unveiled Wednesday.

"This is a major milestone in our ongoing effort to deliver the best modeling and forecasting products and services to the nation," NOAA Acting Administrator Neil Jacobs said in a conference call with journalists Wednesday.

The aim of the new model is to help meteorologists throughout Wisconsin and the rest of the nation create more accurate 1- to 2-day forecasts and 3- to 7-day forecasts.

"We're excited about it," said Mark Gehring, a National Weather Service meteorologist based in Sullivan, which handles forecasting for southeastern and south-central Wisconsin.

"It should result in improved forecasting. The physics of the model will be better and then they'll tweak it and improve it over time like all models. We're expecting it to help us out."