A Global Version of the PSU–NCAR Mesoscale Model

Abstract

A global version of the fifth-generation Pennsylvania State University?National Center for Atmospheric Research Mesoscale Model (PSU?NCAR MM5) is described. The new model employs two polar stereographic projection domains centered on each pole. These domains interact at their equators, thereby eliminating the need for a lateral boundary condition file. This paper describes the method, and contrasts this fully compressible nonhydrostatic Eulerian global model with other global models. There are potential advantages over spherical polar grids in the resolution distribution and the treatment of curvature forces near the poles. The model also selectively damps acoustic modes, which appears to have some benefits in real-data initialization. The split-explicit time steps are different from the semi-implicit schemes used in several global nonhydrostatic models, and this localized scheme avoids the need for global elliptic solvers, making it particularly adept for distributed-memory platforms and the use of composite meshes. Tests of the model show that acoustic and gravity waves as well as advective features propagate across the equator without distortion. A trial 100-day perpetual January simulation shows realistic rain patterns as compared to climatology with no evidence of equatorial effects. Nesting is also available to focus on areas of interest, and this is demonstrated with a 72-h nested forecast over North America. While the time step is shorter than that typically used in semi-Lagrangian global models with a comparable resolution, the model is efficient enough to have allowed the running of daily 120-km grid forecasts on nondedicated computers as small as four-processor workstations since October 1999. Results from this real-time application of the model to 5-day forecasts are shown, and demonstrate that the model performs well at this scale. The model is consistent with the regular regional MM5 and shares dynamics and physics packages without modification. It can also make use of pre- and postprocessing packages developed for the MM5 system. This tight linkage between a regional and global model will have a clear benefit as future global models move toward higher resolutions. It allows current mesoscale numerical weather prediction research to directly feed into the next generation of global models.