An Update on the 2018 HRRR Upgrade at NCEP and a Look at Future Plans (101)

Ed Szoke, CIRA & NOAA/ESRL/GSD, Boulder, CO

Stan Benjamin, NOAA/ESRL/GSD

Curtis Alexander, NOAA/ESRL/GSD


Trevor Alcott, NOAA/ESRL/GSD



The next version of the High Resolution Rapid Refresh (HRRR) model is on track for its next operational update at NCEP by the end of May 2018. This will be the third operational update for the HRRR and the fourth for its parent model, the Rapid Refresh Model (RAP), which will be updated at the same time. The HRRR became an operational model at NCEP on 30 September 2014, after being available to forecasters for several years previously on an experimental basis from the Global Systems Division (GSD) of ESRL/NOAA in Boulder. The HRRR is currently run once per hour at NCEP out to 18 h at a horizontal grid resolution of 3 km, allowing it to operate without convective parameterization. It is initialized with the RAP (Rapid Refresh model) model, which is run at NCEP at a resolution of 13 km once per hour out to 21 hours, but the HRRR assimilates radar and other data on the 3-km scale. The RAP model, also developed at NOAA/ESRL/GSD, became operational on 1 May 2012.

Considerable testing and evaluation was performed before the updates to the new versions were accepted into operations. In this presentation we will give an overview of the changes from the current operational HRRR (HRRRv2) and RAP (RAPv3), and discuss where forecasters should see improvement. Examples will be shown for a variety of weather scenarios, with comparisons shown between the current HRRR and the new version. One of the changes to the HRRR will be to extend the runs initialized at 00, 06, 12 and 18 UTC out to 36-h, and we will also show some of these forecasts. Even though both the HRRR and RAP are about to undergo an update at NCEP, we already have a list of model improvements that will be tested in experimental versions, which will be available at our website at, and which have also traditionally been available to forecasters over the LDM. The potential model improvements will be discussed, along with a schedule for future operational upgrades of the HRRR and RAP models. Other areas of focus with the HRRR will also be addressed, included experimental ensemble runs and the latest developments with the HRRR time-lagged ensemble (HRRRTLE), as well as the new operational domain in Alaska and experimental domain in the Caribbean.