Monday, 15 January 2007
Implementation of a Double Fourier Series (DFS) Dynamic Core in a Global Numerical Weather Prediction Model
Exhibit Hall C (Henry B. Gonzalez Convention Center)
The spectral method that incorporates the spherical harmonics (SPH) as an orthogonal basis function is widely used in the global models. As the model resolution increases, however, the computational cost for the spectral transform becomes dominant over other parts of computation because Legendre transform requires O(N3) operations, with N being the truncation wave number. The Double Fourier Series (DFS) was proposed as basis function of global model to reduce computational cost characterized by O(N2log2N) (Cheong 2006). The suggested DFS retains most of the advantages available for the traditional spectral method and provides an efficient and simple high-order filtering method. The DFS dynamic core is implemented in the YOnsei University Research Model System (YOURS) – global model, which was routed in a version of the NCEP global forecast model. The model with the DFS dynamic core is parallelized using Message Passing Interface (MPI) library routines. The performance of the new global model is evaluated for seasonal simulation, in comparison with the results with the SPH dynamic core.