For the first time in nearly 40 years, the National Weather Service's flagship computer prediction model has received a major makeover, which its leadership says will pave the way for improved forecasts.
The Global Forecast System model, its primary tool for weather forecasting across the nation and around the world, has a new engine or what scientists call a "dynamic core," which the Weather Service says "will drive global numerical weather prediction into the future."
The upgraded model, known as the GFS-FV3 (FV3 stands for Finite Volume Cubed-Sphere dynamical core), became operational Wednesday, June 12.
The dynamic core underpinning the model has been under development for years, involving more than 100 scientists, engineers, technicians, and software developers.
The upgraded model was evaluated over three cold and warm seasons, as well as three hurricane seasons, and its forecasts have proven "equal to or better" than the legacy model, according to Neil Jacobs, acting director of the National Oceanic and Atmospheric Administration, which oversees the Weather Service.
The FV-3 based GFS combines the superior dynamics of global climate modeling with the reliability and speed of operational numerical weather prediction. We are excited about the advancements enabled by the new GFS dynamical core and its prospects for the future. https://t.co/tkPlAvgqme— Louis Uccellini (@NWSDirector) June 12, 2019
Brian Gross, director of the Weather Service's Environmental Modeling Center, told reporters that the model will instantly result in improved hurricane track and intensity forecasts, as well as forecasts for nor'easters and other types of mid-latitude storms.
"We're confident the upgrade will lead to overall improvement," he said.
The Weather Service offered several case studies which showed the GFS-FV3 outperformed the legacy GFS in high impact weather forecasts, including Hurricane Florence and the January 2018 "bomb cyclone."
But during the winter, forecasters testing the model expressed serious concerns that its predictions were too cold and snowy. Gross told The Washington Post in May that these biases were partially corrected but that the Weather Service continues working on them.
At Wednesday's call, June 12, with reporters, Gross added the model can also be "too progressive" with weather systems, meaning it predicts some storms to move along too quickly. He explained "every upgrade has its strengths and weaknesses," but that the GFS-FV3 "is serving as a foundation for future improvements."
The legacy GFS model will continue running until September, so forecasters can compare it with the upgraded GFS-FV3. In May, Gross encouraged model users to provide feedback on the new model so it can fix problems.
The upgraded model is a foundational step in NOAA's larger proposed initiative to establish the Earth Prediction Innovation Center (EPIC) which aims to advance U.S. weather modeling and reclaim international leadership in numerical weather prediction.
For decades, the GFS modeling system has trailed counterparts in Europe in overall accuracy. As part of the EPIC initiative, NOAA hopes to beef up modeling research efforts and build the framework for a cloud-based "community modeling effort," in which the code for the GFS-FV3 is made available to researchers and industry scientists for testing and improvements.
Meanwhile, the world's most accurate model at the European Centre for Medium-range Weather Forecasting received a major upgrade Tuesday, as institutions worldwide endeavor to advance weather prediction.
This article was written by Jason Samenow, a reporter for The Washington Post.