NOAA GEFS analysis

Spatial domain	Global
Spatial resolution	0.25 degrees (~20km)
Time domain	2000-01-01 00:00:00 UTC to Present
Time resolution	3.0 hours

⎘

The Global Ensemble Forecast System (GEFS) is a National Oceanic and Atmospheric Administration (NOAA) National Centers for Environmental Prediction (NCEP) weather forecast model. GEFS creates 31 separate forecasts (ensemble members) to describe the range of forecast uncertainty.

This analysis dataset is an archive of the model's best estimate of past weather. It is created by concatenating the first few hours of each historical forecast to provide a dataset with dimensions time, latitude, and longitude.

Related Datasets

NOAA GEFS forecast, 35 day - Weather forecasts from the Global Ensemble Forecast System (GEFS) operated by NOAA NWS NCEP.

Examples

	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:

URLs will be updated to stable long-term URLs
Users will need to update Icechunk client to version 2 when it is released.

dynamical.org - NOAA GEFS analysis

Temperature at a specific place and time

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/gefs/analysis/latest.zarr")
ds["temperature_2m"].sel(time="2025-01-01T00", latitude=0, longitude=0).compute()

Dimensions

	min	max	units
latitude	-90	90	degree_north

longitude	-180	179.75	degree_east

time	2000-01-01T00:00:00	Present	seconds since 1970-01-01

Variables

	units	dimensions
categorical_freezing_rain_surface Categorical freezing rain 0=no; 1=yes	1	time × latitude × longitude

categorical_ice_pellets_surface Categorical ice pellets 0=no; 1=yes	1	time × latitude × longitude

categorical_rain_surface Categorical rain 0=no; 1=yes	1	time × latitude × longitude

categorical_snow_surface Categorical snow 0=no; 1=yes	1	time × latitude × longitude

downward_long_wave_radiation_flux_surface Surface downward long-wave radiation flux Average value in the last 6 hour period (00, 06, 12, 18 UTC) or 3 hour period (03, 09, 15, 21 UTC).	W m-2	time × latitude × longitude

downward_short_wave_radiation_flux_surface Surface downward short-wave radiation flux Average value in the last 6 hour period (00, 06, 12, 18 UTC) or 3 hour period (03, 09, 15, 21 UTC).	W m-2	time × latitude × longitude

geopotential_height_cloud_ceiling Geopotential height	m	time × latitude × longitude

maximum_temperature_2m Maximum temperature	degree_Celsius	time × latitude × longitude

minimum_temperature_2m Minimum temperature	degree_Celsius	time × latitude × longitude

percent_frozen_precipitation_surface Percent frozen precipitation Contains the value -50 when there is no precipitation.	percent	time × latitude × longitude

precipitable_water_atmosphere Precipitable water	kg m-2	time × latitude × longitude

precipitation_surface Precipitation rate Average precipitation rate since the previous forecast step. Units equivalent to mm/s.	kg m-2 s-1	time × latitude × longitude

pressure_reduced_to_mean_sea_level Pressure reduced to MSL	Pa	time × latitude × longitude

pressure_surface Surface pressure	Pa	time × latitude × longitude

relative_humidity_2m 2 metre relative humidity	percent	time × latitude × longitude

temperature_2m 2 metre temperature	degree_Celsius	time × latitude × longitude

total_cloud_cover_atmosphere Total Cloud Cover Average value in the last 6 hour period (00, 06, 12, 18 UTC) or 3 hour period (03, 09, 15, 21 UTC).	percent	time × latitude × longitude

wind_u_100m 100 metre U wind component	m s-1	time × latitude × longitude

wind_u_10m 10 metre U wind component	m s-1	time × latitude × longitude

wind_v_100m 100 metre V wind component	m s-1	time × latitude × longitude

wind_v_10m 10 metre V wind component	m s-1	time × latitude × longitude

Don't see what you're looking for? Let us know at [email protected].

Details

License

Dataset licensed under CC BY 4.0. Attribution can be found in the dataset's metadata, e.g. ds.attrs["attribution"].

Sources

To provide the longest possible historical record, this dataset in constructed from three distinct GEFS forecast archives.

From 2000-01-01 to 2019-12-31 we use the GEFS reforecast.
From 2020-01-01 to 2020-09-23 we use GEFS forecast archive data which has a lower spatial and temporal resolution.
From 2020-09-23 to Present we use GEFS operational forecast archives.

Variable availability

Data is available for all variables at all times with the following exceptions.

Unavailable before 2020-01-01: relative_humidity_2m, percent_frozen_precipitation_surface, categorical_freezing_rain_surface, categorical_ice_pellets_surface, categorical_rain_surface, categorical_snow_surface
Unavailable 2020-01-01T00 to 2020-09-22T21: geopotential_height_cloud_ceiling

Construction

To create a single time dimension we concatenate the first few hours of each forecast. From 2000-01-01 to 2019-12-31 reforecasts are available once per day and this dataset uses the first 21 or 24 hours of each forecast. From 2020-01-01 to present forecasts are available every 6 hours and this dataset uses the first 3 or 6 hours of each forecast. Variables with an instantaneous step_type use the shortest possible lead times (e.g. 0 and 3 hours) while accumulated variables must use one additional forecast step (e.g. 3 and 6 hours) because they do not have an hour 0 forecast value.

Interpolation

For most of the time range of the archive the source data is available at 0.25-degree resolution and a 3 hourly time step and we perform no interpolation. There are two exceptions to this. 1) From 2020-01-01 to 2020-09-23 the source data has a 1.0-degree spatial resolution and a 6 hourly time step. 2) From 2020-09-23 to present the 100m wind components have a 0.5-degree spatial resolution in the source data. To provide a consistent archive in the above two cases we first perform bilinear interpolation in space to 0.25-degree resolution followed by linear interpolation in time to a 3-hourly timestep if necessary. The original, uninterpolated data can be obtained by selecting latitudes and longitudes evenly divisible by 1 and, in case 1), time steps whose hour is divisible by 6.

Storage

Storage for this dataset is generously provided by Source Cooperative, a Radiant Earth initiative. Icechunk storage generously provided by AWS Open Data.

Compression

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.