U.S. upgrades its weather forecasting system, but still lags behind competitors

The National Weather Service (NWS) has completed a major upgrade to its main supercomputers, which are now processing weather data in four dimensions instead of three. The move will result in more accurate weather forecasts, the agency says.

However, the NWS still lags considerably behind several major weather centers abroad when it comes to its supercomputing capacity and the accuracy of its main weather forecasting model. These so-called weather "model wars" have pitted the NWS' Global Forecast System, also known as the GFS model, against its Canadian and European counterparts. 

The NWS improved the model last year by doubling its resolution, from breaking down the atmosphere into 27 kilometer-wide grid boxes to 13 kilometer grids. 

In general, the smaller the grid box is, the more likely it is for a model to capture small-scale weather features, such as thunderstorms. Models with smaller grids are said to have a higher resolution than those with larger grid boxes.

Here's how the NWS described the latest upgrade in a press release on Wednesday: 

Today’s shift to four-dimensional ensemble hybrid data assimilation takes into account how weather systems evolve on a 3-D spatial grid over time, with time now becoming the fourth dimension. This approach makes better use of earth observations used to initialize the model which gives forecasters a more accurate and timely picture of evolving weather situations.

The upgrade means that the GFS model, which is used by weather forecasters all over the world, will now deliver hourly forecast guidance out to five days in advance, instead of the three-hour increments it did before. 

Most importantly, the upgrade will allow the NWS to incorporate hundreds of thousands of additional weather observations from satellites and other sources, with the intention of utilizing these to refine forecasts. 

The NWS claims that when the new and improved computer model was tested, it improved forecasts across the croplands of the Great Plains, specifically for anticipating the development of severe thunderstorms. 

"The upgrade also improved precipitation forecasts for the continental U.S., and better predicted the development, track and intensity of tropical storms," the agency said.

“The GFS is the foundation for all of our weather and climate models, so today’s upgrade will add skill across all NOAA's forecast mission areas, including hurricanes and other high-impact weather,” NWS director Louis Uccellini said in a statement. 

Why this upgrade is not enough

However, the NWS still has a lot of catching up to do if it wants to win the model match-up. 

"Unfortunately, based on skill comparisons... the new GFS model has not demonstrated a statistically significant improvement," said Ryan Maue, a meteorologist with WeatherBell Analytics, in an email to Mashable. 

Maue closely monitors computer model skill scores via his company's website and his Twitter account, and while he praised the NWS for taking this step, he said more headway needs to be made to bring the GFS model up to par with other weather agencies around the world. 

Right now, data shows the GFS is still being beaten by the European Center for Medium Range Weather Forecasting, or ECMWF, located in Reading, England, as well as a model run by the Canadian government and another from the UK Met Office. 

"ECMWF recently upgraded their global model from 14-km to 9-km (compared to GFS current grid spacing of 13-km) and achieved a major skill improvement," Maue said. "The 'gap' between ECMWF and GFS widened significantly -- by almost 25 percent," using one widely cited metric. "Global weather models in general become more skillful as technology improves (bigger computers) and better software evolves (model upgrades). However, [the] ECMWF is now getting 'better faster' and GFS has continued to fall behind."  

"Indeed, GFS is now a 4th place model behind Canadian, UKMet, and ECMWF which is still the undisputed 'king' and continues to push the skill envelope to record high levels."

However, supercomputing power isn't the only factor when it comes to making an accurate forecast, Maue said. In fact, the model wars narrative is somewhat misleading, since forecasters at the NWS and agencies abroad typically have access to all the model data, regardless of what nation it originates in.

"While we focus on the model-to-model comparisons, it's important to note that GFS is a very good weather model in its own right and has scored major forecast 'wins' while other models have whiffed," Maue said. 

"We are experiencing a 'golden age' of computer weather forecasting and should marvel at the models' amazing capabilities. Europe currently has the bragging rights in the "weather model wars" but the competition is healthy and collaborative as research and experiment results are shared among agencies."