In the field of meteorological prediction, “weather model” is a very important branch, using supercomputers to simulate the atmospheric environment and conditions, from the current atmospheric conditions to the future to calculate the possible changes in the weather. In addition to providing a week-long forecast, it is more important to prevent and control disaster weather, such as typhoon/hurricane path prediction, rainfall estimation, etc., is an indispensable tool for modern weather forecasting. Because the model’s deduction needs the support of the powerful computer computing power and theoretical basis, the development of meteorological model has long been regarded as a symbol of national strength, and all countries will try to develop their own system and compare the accuracy of the model. The two largest and most complex systems are the Global Forecasting System (GFS) of the National Weather Service (NWS) and the Integrated Forecast System (IFS) of the European Centre for Medium-Term Weather Forecasting (ECMWF), which arrives every year Typhoon season, two sets of systems who have more accurate control over routes and intensity are always a topic of comparison among weather enthusiasts.
Of the two, GFS is relatively an older system, with the core dynamic model not changing for nearly 40 years, and the prediction of a big loss to ECMWF in 2015 was a big blow to the United States. So in recent years NWS has greatly increased the computing power of the computer, from 776 TFlops to 8.4 PFlops, and began experimenting with a new set of core dynamic models called Finite-Volumeed Cubed-Sphere, or FV3. In addition to making more efficient use of computer resources, FV3 is also better able to simulate vertical air flow changes, giving it a better grasp of weather phenomena such as thunderstorms and tropical cyclones. After a period of trial operation, NWS today finally announced that FV3 will replace the old core and enable the new GFS system. However, not everyone is confident about the fV3’s performance. As things stand, FV3 is better than the old core in some places, but it is not enough in others, especially for the forecast of tropical cyclone paths and intensity, and although there is a special view of vertical airflow, the accuracy is still roughly in line with the old system. So while FV3 is on the eve of the 2019 hurricane season, it’s unclear whether it can really compete with ECMWF. It’s just that the old GFS model has almost reached its limit after decades of evolution, and if the U.S. wants to further improve forecast accuracy, it’s imperative to switch to a new model. I can only hope that FV3 will continue to improve over the next few years to give us a more accurate forecast!