FV3

FV3, short for Finite-Volume on a Cubed-Sphere, is a numerical weather prediction dynamical core used for global atmospheric modeling. It solves the primitive equations on a cubed-sphere grid using finite-volume discretization, with the goal of improving conservation properties, scalability on modern high-performance computers, and reduced grid imprinting near the poles. The core is designed to be flexible with respect to physics packages and data assimilation components, enabling a range of configurations for global forecasts.

The cubed-sphere grid partitions the sphere into six faces, mapped to a cube, which helps avoid singularities at the poles that complicate traditional latitude-longitude grids. The finite-volume approach emphasizes the conservation of key quantities such as mass, momentum, and energy across grid cells, contributing to robustness in long-range forecasts and in multiyear climate simulations. FV3 is built to support high-resolution configurations and to operate efficiently on contemporary parallel computing architectures.

FV3 is associated with NOAA’s Unified Forecast System and has been deployed as the dynamical core for the Global Forecast System (GFS) in modernization efforts aimed at improving forecast accuracy and computational efficiency. In practice, the FV3 core is used in conjunction with a suite of physics parameterizations and data assimilation processes to produce daily global weather forecasts and climate-model runs. The model’s design emphasizes portability across HPC platforms and adaptability to evolving forecast requirements.

See also: Global Forecast System, Unified Forecast System, finite-volume methods, cubed-sphere grids, numerical weather prediction.