For years, there’s been broad agreement that the U.S. weather forecasting system needed an upgrade. “Over the years, the National Weather Service has fallen behind its European counterparts in the accuracy of its forecasts,” meteorologist Jason Samenow told NPR in 2015.
That was particularly apparent in the lead-up to Hurricane Sandy in 2012. “The European model was the first to accurately predict that Sandy, rather than hooking out to sea, would actually strike the U.S.,” Samenow said.
After that devastating storm, Congress approved NOAA funding to, as NOAA put it, “dramatically improve U.S. hurricane forecasting and develop a forecast system that would be second to none.”
NOAA has been running the new engine for the past three years of weather to compare its forecasting power to the model it had been using.
The improved accuracy is clear in several side-by side examples provided by NOAA. For example, during the powerful winter storm in January 2018 that became known as a “bomb cyclone,” the model in use at the time indicated a much lighter day of snowfall in the upper Northeast U.S. than what actually happened. The updated model was more accurate — it showed heavy snowfall.
The side-by-side examples also show that the old model inaccurately predicted “extreme strengthening” of Hurricane Florence last year, while the updated model was much more accurate at predicting its path and intensity.
The old model will continue to run in parallel through September to compare the predictions.
This new model has faced delays in its rollout. As The Washington Post reported, it had been planned for early this year, but the National Weather Service said the model had a few technical issues. And the partial government shutdown also slowed things down.
The information supplied by the new forecast model will be used in local forecasts across the United States.
“The weather forecasts that you hear are provided by National Weather Service forecasters, or folks that are using the numerical guidance that our numerical models provide to them,” Gross said. “So we’re really looking at improvements in this foundational information that the individuals that forecast the weather use on a day-to-day basis.”