A global version of the Penn State / NCAR mesoscale model, MM5, will be presented. This model uses the fully compressible nonhydrostatic equations applied in two interacting hemispheric domains. The domains use polar stereographic projections centered on each pole and they interact at the equator. The model uses the same code as MM5 for the physics and dynamics, and so can easily make use of all MM5's options and nesting capabilities, as well as data assimilation and initialization techniques. The global version has also been parallelized to take advantage of high-power distributed computing platforms.

The primary roles for this model are research in medium-range predictability, data assimilation, and long-range transport studies. By eliminating the lateral boundary conditions, the model needs only an initial analysis input to run, and can be run indefinitely in principle. In practice, useful forecasts are limited to 5-10 days by predictability, and the sea-surface temperature specification.

The model has been run daily at NCAR since 1999 with a modest grid size corresponding to about one degree at mid-latitudes and half a degree nearer the equator. Recent verifications will be presented that focus on the model's capability at predicting the 500 hPa geopotential height field in the 5-day range. It will be shown that both for North America and the Northern Hemisphere there is skill beyond 5 days. Model biases in this field appear primarily in polar regions and point to areas where work is required to improve MM5 for global and regional applications, and for longer term forecasting.

The technique applied to MM5 can be applied to other regional models fairly easily. As global model resolutions become better with advances in computing power, global models will approach current-day mesoscale model resolutions, and will thus be using dynamical and physical packages similar to those developed by the mesoscale modeling community. The techniques presented here may provide a bridge for mesoscale research to feed directly into future global models.