A prototype implementation of the high resolution rapid refresh for Alaska

Motivated by the extension of the Rapid Refresh (RR) model into Alaska, its aggressive assimilation schemes, and its imminent use to initialize the High Resolution Rapid Refresh (HRRR) runs in the CONUS, the Arctic Region Supercomputing Center (ARSC) and the Alaska Region Headquarters (ARH) of NOAA's National Weather Service are experimenting with their own implementation of an ARSC HRRR. Initially providing four 24-hour forecasts daily on 256 processors, ARSC is making GRIB2 products available to NOAA and other interested parties via web and LDM servers. These products provide frequently updated short-term forecasts at 3km resolution over most of the Alaska Region and the Gulf of Alaska.

This presentation outlines the evolution of the prototype ARSC HRRR and its use by the NWS in Alaska. Originally initialized with the 12km NAM, the ARSC HRRR is now initialized with the 13km RR, opening the possibility to optimize for more frequent runs. Additionally, given some concerns over lengthy model spin-up times, experiments will be performed to gauge the efficacy of running the ARSC HRRR in a warm-start mode, initializing each forecast with the previous ARSC HRRR forecast, and then extending the experiments to assess the desirability of performing local data assimilation with each run. These activities are taking place during Summer and Fall 2010 and will be reported on during this presentation.