EOMF-09. The impact of different configurations on forecasting extreme events with the Unified Forecast System

Abstract
We investigated the impact of different configurations on forecasting extreme events using a limited-area configuration of the Unified Forecast System (UFS). Ten cases of extreme events, provided by the UFS Case Studies project and including hurricanes, midlatitude storms, heat waves, cold blasts, and cold air damming, were used. The study is focused on the CONUS domain and uses the UFS Short-Range Weather (SRWeather) Application v1.0.0 released in March, 2021. The physics suites used in this study include GFS v15.2, GFS v16, GFSv17α, and RRFS v1 α. The results are compared against simulations previously made with the UFS global configurations, namely the UFS Medium-Range Weather (MRWeather) Application v1.0 and those obtained using the same workflow employed for the operational GFS v16 (EMC Global Workflow v16.0.10). The motivation for using a limited area configuration is that it requires less computer resources and allows running the model in high resolution (~ 3km horizontal grid spacing). The model forecasts are verified against station observations and analysis data. It is shown that the results are sensitive to the configurations used, such as the version of the code base, domain (SRWeather vs Global), model grid spacing (13km vs 3km, 64 levels vs 127 levels), and the physics suite employed. In particular, the 2-m temperature forecast for the cold air damming case obtained with the SRWeather App v1.0.0 is slightly better than the one obtained with the MRWeather App v1.0. Using the latest code and more vertical levels, the improvements are found for the cold air damming case with GFS v16. Results from 3-km horizontal grid spacing gives more detailed structures and better shape; however, the cold air area is shifted northward for GFS v15.2 and GFS v16 but not for GFS v17α. Finally, sensitivity tests show that the land surface model plays a role in the northward shift.