NOAA is developing its next generation global model prediction system, and at its heart is the Finite Volume Cubed-Sphere dynamical core (FV3) modernizing the National Weather Service’s approach to weather modelling.
A dynamical core takes equations describing movement in the atmosphere, such as moisture traveling through the water cycle, and translates them into computer-solvable language. It’s the engine of a weather forecast model, tracking how the Earth’s atmosphere is changing and what weather might develop as a result. It doesn’t have all the parts needed to make a forecast. Every model needs three fundamental pieces: a dynamical core, a set of physics equations representing weather processes, and data about the real atmospheric conditions before forecasting.
The complexity and scope of numerical weather prediction means supercomputers are also essential to forecasting. As computer power increases with each new generation, weather models improve exponentially alongside them.
[FV3 is able to resolve the path of storm systems on the scale of counties]
The National Weather Service’s current Global Forecasting System (GFS) has been the foundation of NOAA’s suite of weather models for over 30 years. But throughout that time, technology has improved and scientists have found new approaches to processing data. NOAA Research has taken the steps to replace the GFS’s components with newer, more efficient and more accurate processes.
FV3 is the first step. It was developed in NOAA Research’s Geophysical Fluid Dynamics Labratory initially to power climate models and was then adapted for detailed global weather prediction. The NWS chose FV3 as the new GFS’s dynamical core in part because it uses less computer resources than other options. FV3 brings unprecedented accuracy to forecasts in three important ways:
- Computer Usage — FV3 is designed to efficiently scale to the available resources on any supercomputer for faster, higher resolution images. The current GFS, developed before the age of high speed computers, is not able to provide such highly detailed information. Even if it ran on a computer with more processing power, it would not work faster.
- Vertical Equations — FV3 uses vertical equations to limitlessly zoom down to local scales and provide images of up-down air fluctuations, allowing us to resolve thunderstorms and their updraft winds. Older models assume the atmosphere experiences equal forces from above and below. This assumption can provide accurate prediction over large areas, but is unable to see the small-scale fluctuating winds that can lead to severe weather.
- Representation — FV3 represents weather through points in connected grid cells, so it can resolve weather that comes in irregular shapes. The current GFS represents all weather as waves. It’s been successful in large-scale modelling, but weather phenomena do not always follow wave patterns on the local level. For example, thunderstorms and cold fronts have sharp edges that a wave shape can not fully capture.
[Above image is a comparison between the current GFS and FV3 modelling annual mean rainfall across South America. The results from the Tropical Rainfall Measuring Mission (TRMM) show what actual values were. FV3 can resolve small-scale features without the dot-like distortion current GFS shows, which represents false storms.]
NWS and NOAA Research scientists are phasing in the GFS with FV3; it is being run experimentally with the target of going operational in late 2019. Our new global model aims to deliver better, more timely forecasts to serve the growing needs of our forecasters and the weather enterprise.
For WeatherNation: Meteorologist Mace Michaels