Home > HRRR documentation
The High-Resolution Rapid Refresh is a NOAA/NCEP operational weather prediction system
comprised of a numerical forecast model and an analysis/assimilation system to initialize the
model. It is run with a horizontal resolution of 3 km and has 50 vertical levels.
The HRRR was developed to serve users needing frequently updated high-resolution short-
range weather forecasts, especially those in the US severe weather forecasting community.
The model is run over the CONUS every hour of the day; it is integrated to 36 hours for the
00/06/12/18 UTC cycles and to 18 hours for every other cycle. The HRRR Alaska sector is
run every third hour; it is integrated to 36 hours for the 00/06/12/18 UTC cycles and to 18
hours for the 03/09/15/21 UTC cycles. The HRRR performs a 1-hour spinup using a first
guess from the RAP and then assimilates radar reflectivity data every 15 minutes during the
spinup forecast. This is followed by a Gridpoint Statistical Interpolation (GSI) analysis. The
assimilation of radar data combined with the GSI assimilation of cloud and hydrometeor data
provides significant skill in short-range cloud and precipitation forecasts. The forecast model
is the ARW core of the WRF model, and output is generated every 15 minutes during the
forecast for the first 18 hours. Hourly output is generated between forecast hours 18 and 36
for the extended cycles.
Full 3-dimensional upper level fields are not being provided for the HRRR, as there is little
relevant synoptic-scale information gained from the resolution difference between the HRRR
and its parent, the RAP. The HRRR should be primarily used to examine surface, near-
surface, and column parameters. For those interested in upper-level analyses and short-
range forecasts, we recommend using the RAP grids that are available on NOMADS.