A Comparison of High-Resolution Mesoscale Forecasts using MM5 and WRF-ARW

The Weather Research and Forecast (WRF) model is a new state-of-the-science mesoscale model framework under development for many years by UCAR and NCEP, and specifically designed for the 1-10 km grid length scales. The advanced research WRF (WRF-ARW) uses more advanced, higher-order spatial and temporal finite differencing schemes than MM5 and also contains cloud-scale 3D sub-grid turbulence closures in addition to many of MM5's model physics packages for the planetary boundary layer, microsphysics, land-surface, convective parameterization, etc. As WRF matures to include more of the important options from MM5 (e.g., data assimilation), it is becoming more attractive for fine-resolution numerical weather prediction (NWP) due to its mass conservation, improved numerics and expanding physics. To help determine the potential added value of transitioning from MM5 to WRF, direct comparisons between MM5 and WRF are presented here for different terrain and weather conditions using the exact same model grids, initial conditions, and lateral boundary conditions, and comparable model physics options

The February 2006 Torino Winter Olympics over northern Italy and the September 1983 Cross Appalachian Terrain Experiment (CAPTEX) study over the Northeast United States are used in the current evaluations. Penn State and the Defense Threat Reduction Agency (DTRA) used its on-demand MM5 realtime systems to support hazard prediction and consequence assessment over the complex terrain of the Alps and the Torino plains during the Winter Olympics. Four nested grids of 36-km, 12-km, 4-km and 1.3-km resolutions were used to produce 24h forecasts over all the Olympics venues. After the games, six cases representing the great variety of weather scenarios during the 16-day period were chosen for a retrospective comparison study between MM5 and WRF-ARW. In order to make a fair comparison, both MM5 and WRF used the same grid configurations and similar physics (e.g. turbulent kinetic energy (TKE) PBL parameterization, explicit-moisture microphysics with Kain-Fritsch cumulus parameterization on the outer two domains, force-restore land-surface models using satellite snowcover, etc.)

The WRF runs produced shortly after the games using WRF-ARW 2.1 produced some lateral boundary noise problems and missed some important precipitation over the Olympics venues for some of the cases, as compared to the MM5 runs. Nonetheless, statistical differences between these two models without FDDA over the special Italian mesonet data network and during CAPTEX83 over the northeast United States were relatively small and suggest that there was no clear statistical advantage of one model over the other for wind, temperature and moisture. Updated and more detailed MM5-WRF comparison results using the latest WRF version, WRF 2.2, will be presented at the conference for the six Olympics cases and CAPTEX83.