Last year, criticism began to emerge concerning the inferior accuracy of the NWS’s Global Forecast System (GFS) model – run on earlier versions of the supercomputers – compared to the model run at the European Centre for Medium Range Forecasting (ECMWF) based in the United Kingdom. The GFS and ECMWF models are, by far, the most heavily relied on by meteorologists around the world for forecasting.
Cliff Mass, a professor of atmospheric sciences at the University of Washington and expert in numerical weather prediction, called the standing of the GFS model “third-rate” and “a national embarrassment” in a March 2012 blog post.
“It is a huge story, an important story, but one the media has not touched, probably from lack of familiarity with a highly technical subject,” Mass wrote.
But when the ECMWF model outclassed the GFS model in its long-range forecasts for Hurricane Sandy in late October, the accuracy discrepancy between the models caught the media’s attention.
“The American model is the basis for many forecasts, and its reliability problems beyond the short term suggest something major is amiss with its physics and inputs,” wrote USA Today in an editorial. “The European model’s embarrassing superiority on Sandy ought to accelerate efforts to identify and fix what’s wrong.”
The $23.7 million in improvements to NWS’s forecasting systems from the Sandy supplemental will facilitate a more than ten-fold increase in the capacity of the supercomputer running the GFS model.
“This is an extraordinarily positive development and will give the National Weather Service the potential to lead the world in numerical weather prediction,” Mass said in an email Tuesday.
In technical terms, the computing capacity will ramp up from 213 teraflops to 2,600 teraflops by the 2015 fiscal year according to the NWS. (Teraflops are simply a measure of the number of trillion calculations the computer can perform per second.) The NWS expects ECWMF’s supercomputer to have a capacity of 2,217 teraflops at that same time.
“By 2015, we will exceed the ECMWF in operational computing capacity with NOAA operational computers for the first time since the late 1980s or early 1990s,” Uccellini said.
The increased computing power will enable drastic improvements in the GFS model’s resolution Uccellini said. Higher resolution models pick up on weather features a lower resolution might miss, like some high altitude steering currents. (Model resolution defines the geographic size of grid boxes in which models perform calculations. The smaller the grid box, the better the model can simulate localized features.)
In the case of Sandy, the ECMWF’s significantly higher resolution at forecast times 8 days into the future helped it accurately forecast the storm’s unusual left turn towards the Northeast coast while the GFS incorrectly simulated the storm would track harmlessly out to sea (at that same forecast time).
The NWS projects the Sandy supplemental funds will help enhance the horizontal resolution of the GFS model by around a factor of 3 by FY2015, enough to rival the ECMWF.
To make additional gains in the model’s accuracy, Uccellini said some of the Sandy supplemental funds will help pay for contract scientists to improve the model physics and to enhance the systems that bring in (or assimilate) data from ground weather sensors, satellites and weather balloons.
“These upgrades and plans represent a substantial improvement of the primary modeling system used by the Weather Service,” said Ricky Rood, professor of meteorology at the University of Michigan. “With these developments, the U.S. forecast system will share many characteristics with that of ECMWF.
Uccellini stressed the Sandy supplemental funds will be distributed to not only advance the GFS model, but also specialized models for predicting short-range weather, hurricanes, and thunderstorms.
The large scope of the modeling efforts the Sandy supplemental funds support may make it difficult to sustain gains made relative to the ECMWF cautioned University of Washington’s Mass.
“[The Sandy supplemental] is huge, but is only a down payment on what is needed,” Mass said. “The ECMWF only has to do global weather prediction; in contrast, the NWS has many more responsibilities.”
Uccellini said the President’s FY14 budget proposal provides a $13.8 million increase in funding levels for operational computing on top of the Sandy supplemental which would provide the various modeling programs with long-term security.
“We now have a budget profile, if the President’s budget is passed, that will allow us to sustain and build off the [Sandy supplemental] increases,” Uccellini said.
Marshall Shepherd, president of the American Meteorological Society, called these investments in forecasting a “game-changer”.
“I applaud NWS, NOAA, and policymakers for stepping forth with investments,” Shepherd said. “Weather forecasting is as critical to U.S. citizen’s lives as many public safety and national security activities, in my view.”
(Note: The $23.4 million appropriated to the NWS for forecasting improvements is just a small fraction of the $309.7 million NOAA will receive as part of the Sandy relief legislation. For example, funds will support ocean observing and coastal monitoring, disaster assistance for fisheries, and upgrades to two NOAA hurricane hunter aircrafts among other initiatives.)
(CWG’s Steve Tracton contributed to this report. Tracton will post a forthcoming commentary assessing the caveats, details, and significance of this news.)