| Spatial domain | Continental United States |
| Spatial resolution | 3 km |
| Time domain | 2014-10-01 00:00:00 UTC to Present |
| Time resolution | 1 hour |
⎘
The High-Resolution Rapid Refresh (HRRR) is a NOAA real-time 3-km resolution, hourly updated, cloud-resolving, convection-allowing atmospheric model, initialized by 3km grids with 3km radar assimilation. Radar data is assimilated in the HRRR every 15 min over a 1-h period adding further detail to that provided by the hourly data assimilation from the 13km radar-enhanced Rapid Refresh.
This analysis dataset is an archive of the model's best estimate of past weather. It is created by concatenating the first hour of each historical forecast to provide a dataset with dimensions time, x, and y.
This dataset uses the native HRRR Lambert Conformal Conic projection,
with spatial indexing along the x and y dimensions.
The example notebook shows how to use the embedded spatial reference to
select geographic areas of interest.
| Open notebook in github | |
| Open notebook in colab | |
| Icechunk example usage notebook *** |
***: Icechunk examples are pre-release for feedback. Subscribe to our newsletter to be notified about two small breaking changes that we will implement:
import xarray as xr # xarray>=2025.1.2 and zarr>=3.0.8 for zarr v3 support
ds = xr.open_zarr("https://data.dynamical.org/noaa/hrrr/analysis/latest.zarr")
ds["temperature_2m"].sel(time="2025-01-01T00", x=0, y=0, method="nearest").compute()
| min | max | units | |
|---|---|---|---|
| time | 2014-10-01T00:00:00 | Present | seconds since 1970-01-01 |
| x | -2700000 | 2700000 | m |
| y | -1600000 | 1600000 | m |
| units | dimensions | |
|---|---|---|
| categorical_freezing_rain_surface | 1 | time × y × x |
| categorical_ice_pellets_surface | 1 | time × y × x |
| categorical_rain_surface | 1 | time × y × x |
| categorical_snow_surface | 1 | time × y × x |
| composite_reflectivity | dBZ | time × y × x |
| dew_point_temperature_2m | degree_Celsius | time × y × x |
| downward_long_wave_radiation_flux_surface | W m-2 | time × y × x |
| downward_short_wave_radiation_flux_surface | W m-2 | time × y × x |
| geopotential_height_cloud_ceiling | m | time × y × x |
| latitude | degree_north | y × x |
| longitude | degree_east | y × x |
| percent_frozen_precipitation_surface | percent | time × y × x |
| precipitable_water_atmosphere | kg m-2 | time × y × x |
| precipitation_surface | kg m-2 s-1 | time × y × x |
| pressure_reduced_to_mean_sea_level | Pa | time × y × x |
| pressure_surface | Pa | time × y × x |
| relative_humidity_2m | percent | time × y × x |
| snow_area_fraction_surface | 1 | time × y × x |
| snow_thickness_surface | m | time × y × x |
| snow_water_equivalent_surface | m | time × y × x |
| snowfall_surface | m s-1 | time × y × x |
| temperature_2m | degree_Celsius | time × y × x |
| total_cloud_cover_atmosphere | percent | time × y × x |
| wind_gust_surface | m s-1 | time × y × x |
| wind_u_10m | m s-1 | time × y × x |
| wind_u_80m | m s-1 | time × y × x |
| wind_v_10m | m s-1 | time × y × x |
| wind_v_80m | m s-1 | time × y × x |
Dataset licensed under CC BY 4.0.
NOAA NWS NCEP HRRR data processed by dynamical.org from NOAA Open Data Dissemination archives.
Or NOAA HRRR from dynamical.org.
HRRR starts a new model run every hour and dynamical.org has created this analysis by concatenating the first step of each forecast along the time dimension. Accumulated variables (e.g. precipitation) are read from the second step of the previous hour's forecast.
There are a significant number of missing source files before August 2018 (HRRR v1 and v2 period), and a small number from August 2018 to December 2020 (HRRR v3 period).
downward_long_wave_radiation_flux_surface and relative_humidity_2m are
unavailable before August 2016 (HRRR v1 period).
This dataset has NaN values where source data are unavailable.
The source grib files this archive is constructed from are provided by NOAA Open Data Dissemination (NODD) and accessed from the AWS Open Data Registry. Operational data is additionally accessed from NOAA NOMADS.
Storage for this dataset is generously provided by Source Cooperative, a Radiant Earth initiative. Icechunk storage generously provided by AWS Open Data.
The data values in this dataset have been rounded in their binary floating point representation to improve compression. See Klöwer et al. 2021 for more information on this approach. The exact number of rounded bits can be found in our reformatting code.