GFS Archive

Archived forecasts from NOAA’s Global Forecast System (GFS) are available via NCAR’s Research Data Archive (specifically from the primary variable set and the secondary variable set.

Currently, data from the following grib typeOfLevel filters are available:

• isobaricInhPa see available pressure levels below

• surface, where variables with stepType of accum or avg are prefixed in ufs2arco with those labels (e.g., instead of tp for total precipitation, look for accum_tp).

• heightAboveGround, where we append the height to any variables that do not have the height in their name (e.g., u at level=80 gets renamed to u80)

Available Pressure Levels

Warning

Not all variables are available at all of these levels. Eventually, we hope to document what’s available for each variable, but until then, go for trial and error (unavailable levels will be filled with NaNs), or refer to the original data source links above.

Available Pressure Levels (hPa)

1	2	3	5	7
10	15	20	30	40
50	70	100	125	150
175	200	225	250	275
300	325	350	375	400
425	450	475	500	525
550	575	600	625	650
675	700	725	750	775
800	825	850	875	900
925	950	975	1000

Available Variables

Note

There are some variables are available during some years but not others. For now, only variables that are available during the entirety of 2015-2024 are available.

Variables from GFS available in ufs2arco

Variable	Long Name	Forecast Only
SUNSD	Sunshine Duration	False
accum_acpcp	Convective precipitation (water) accumulated over forecast	True
accum_tp	Total Precipitation accumulated over forecast	True
accum_watr	Water runoff accumulated over forecast	True
aptmp2	2 metre Apparent temperature	False
avg_al	Time-mean forecast albedo	True
avg_cduvb	Time-mean Clear sky UV-B downward solar flux	True
avg_cfrzr	Time-mean Categorical freezing rain	True
avg_cicep	Time-mean Categorical ice pellets	True
avg_cpr	Time-mean Convective precipitation rate	True
avg_crain	Time-mean Categorical rain	True
avg_csnow	Time-mean Categorical snow	True
avg_duvb	Time-mean UV-B downward solar flux	True
avg_gflux	Time-mean Ground heat flux	True
avg_iegwss	Time-mean Instantaneous eastward gravity wave surface stress	True
avg_ingwss	Time-mean Instantaneous northward gravity wave surface stress	True
avg_ishf	Time-mean surface sensible heat flux	True
avg_prate	Time-mean Precipitation rate	True
avg_sdlwrf	Time-mean Surface downward long-wave radiation flux	True
avg_sdswrf	Time-mean Surface downward short-wave radiation flux	True
avg_slhtf	Time-mean surface latent heat flux	True
avg_sulwrf	Time-mean Surface upward long-wave radiation flux	True
avg_suswrf	Time-mean Surface upward short-wave radiation flux	True
avg_utaua	Time-mean U-component of atmospheric surface momentum flux	True
avg_vtaua	Time-mean V-component of atmospheric surface momentum flux	True
cape	Convective available potential energy	False
cin	Convective inhibition	False
cnwat	Plant canopy surface water	False
cpofp	Percent frozen precipitation	False
d2m	2 metre dewpoint temperature	False
fldcp	Field Capacity	False
gh	Geopotential height	False
gust	Wind speed (gust)	False
hindex	Haines Index	False
lftx	Surface lifted index	False
lftx4	Best (4-layer) lifted index	False
lsm	Land-sea mask	False
mslet	MSLP (Eta model reduction)	False
orog	Orography	False
pres80	80 metre Pressure	False
prmsl	Pressure reduced to MSL	False
q	Specific humidity	False
q80	80 metre Specific humidity	False
r2	2 metre relative humidity	False
sde	Snow depth	False
sdwe	Water equivalent of accumulated snow depth (deprecated)	False
sh2	2 metre specific humidity	False
siconc	Sea ice area fraction	False
sithick	Sea ice thickness	False
sp	Surface pressure	False
t	Temperature	False
t100	100 metre Temperature	False
t2m	2 metre temperature	False
t80	80 metre Temperature	False
t_surface	Temperature at surface	False
tmax2	2 metre Maximum temperature	True
tmin2	2 metre Minimum temperature	True
u	U component of wind	False
u10	10 metre U wind component	False
u100	100 metre U wind component	False
u80	80 metre U component of wind	False
v	V component of wind	False
v10	10 metre V wind component	False
v100	100 metre V wind component	False
v80	80 metre V component of wind	False
w	Vertical velocity	False
wilt	Wilting Point	False

Temporally Accumulated Variables

Variables that are accumulated over a specific period have been prepended with "accum_". To be specific, reading these fields with xarray + cfgrib might look something like this:

import xarray as xr
ds = xr.open_dataset(
    "mygribfile.grib2",
    engine="cfgrib",
    decode_timedelta=True,
    filter_by_keys={
        "typeOfLevel": "surface",
        "stepType": "accum",
    },
)

All variable names with this stepType are prepended with "accum_" within ufs2arco (see e.g., "accum_tp" in the table above).

For these variables, an additional option can be provided to read different accumulation periods. For example, if we were reading data from forecast hour 6, we could modify the filter_by_keys option above to read surface variables accumulated from the forecast initialization to hour 6 by providing the stepRange option:

filter_by_keys={
    "typeOfLevel": "surface",
    "stepType": "accum",
    "stepRange": "0-6"
}

or, for example when reading HRRR data we could read the accumulation over the previous hour as so,

filter_by_keys={
    "typeOfLevel": "surface",
    "stepType": "accum",
    "stepRange": "5-6"
}

The default behavior for xarray+cfgrib (which ufs2arco uses internally) appears to grab the accumulation over the full forecast period. However, to provide a different accumulation period for the stepRange argument, add the accum_hrs to the yaml recipe. For example with accumulated total precipitation,

accum_hrs:
  accum_tp: 1

This would grab the total precipitation averaged over the most recent forecast hour. So, at forecast hour 6, this is the same as providing {"stepRange":"5-6"} when reading a single grib file.