A Double Fourier Series (DFS) Dynamic Core in a Global Atmospheric Model with Full Physics
Abstract This study describes an application of the double Fourier series (DFS) spectral method developed by Cheong (2006) as an alternative dynamic option in a model system that was ported in the Global/Regional Integrated Model system (GRIMs). A message-passing interface (MPI) for a massive parallel-processor cluster computer devised for the DFS dynamical core is also presented. The new dynamical core with full physics was evaluated against a conventional spherical harmonics (SPH) dynamical core in terms of short-range forecast capability for a heavy rainfall event and seasonal simulation framework. Comparison of the two dynamical cores demonstrates that the new DFS dynamical core exhibits performance comparable to the SPH in terms of simulated climatology accuracy and the forecast of a heavy rainfall event. Most importantly, the DFS algorithm guarantees improved computational efficiency in the cluster computer as the model resolution increases, which is consistent with theoretical values computed from the dry primitive equation model framework of Cheong. Our study shows that, at higher resolutions, the DFS approach can be a competitive dynamical core because the DFS algorithm provides the advantages of both the spectral method for high numerical accuracy and the grid-point method for high performance computing in the aspect of computational cost.