Realtime high-resolution mesoscale modeling for the Vancouver Olympics

Penn State University ran its MM5 mesoscale modeling system in realtime for the entire 16-day Vancouver Olympics as R&D support for Defense Threat Reduction Agency (DTRA) Reachback. The MM5 configuration consisted of a set of nested grids with 36-/12-/4- and 1.3-km resolutions. MM5 forecasts up to 36 h into the future from the initial time (either 00 or 12 UTC daily) were supplied to Hazard Prediction and Assessment Capability (HPAC) that runs SCIPUFF as the atmospheric transport and dispersion (AT&D) model. The NCEP Global Forecast System (GFS) data was used for baseline cold-start initial conditions and the gridded lateral boundary conditions. To achieve better initial conditions and improved short-term forecasts, the MM5 system ran a dynamic initialization (DI) using multiscale four dimensional data assimilation (FDDA) (both analysis nudging and observation nudging during a pre-forecast period). During the Olympics, the meteorological files to run SCIPUFF (called MEDOC files) were created by DTRA on the Meteorological Data Server (MDS). Penn State also produced its own MEDOC files using the new unified MEDOC converter developed by Sage Management Inc. and Penn State to provide a tighter coupling between MM5 and SCIPUFF. This unified MEDOC converter allows a native grid option that minimizes the interpolations from the MM5 to the SCIPUFF grids.

Similar and expanded FDDA capabilities have been developed for WRF-ARW system under the support of DTRA. To evaluate the performance of the FDDA capabilities in WRF, a few representative cases are selected for in-depth analysis, and the model results are compared between the two modeling systems. In this presentation, the effects of model resolution and FDDA on the MM5/WRF and SCIPUFF predictions are investigated to show the effects of the complex terrain on the meteorological and AT&D solutions. Some SCIPUFF results are evaluated using both the MDS version and the new R&D version of the MEDOC converter to show the benefit of improved model coupling.