CORe¶

NOAA Conventional Observation Reanalysis (CORe)

Available

At a glance¶

Note on period: Starts at 00 UTC on 1 January 1950 (R1 and CDAS start in 1948). As of April 1, 2026, CPC reports that NODD holds 1950 through 2025, with operational output appended since then. NOMADS keeps only the last 7 days, so get the historical record from NODD or GDEX.

Overview¶

CORe is an atmospheric reanalysis from the NOAA Climate Prediction Center (CPC) covering 1950 to the present, built for climate monitoring. It replaced CDAS, the real-time extension of the NCEP/NCAR Reanalysis, in March 2026. It is not a rerun of R1 but a new reanalysis: a UFS-based FV3GFS model (C128, 64 layers) with an 80-member Ensemble Kalman Filter (EnKF). Unlike most modern reanalyses it assimilates no satellite data except Atmospheric Motion Vectors (AMV); the design emphasises consistent climate trends and anomalies over the best state for a forecast. Output is 3-hourly GRIB2, and the primary product is the mean of the 80 ensemble analyses.

Available variables¶

Main fields from the NCAR GDEX (d090004) THREDDS metadata and CPC documentation. Names are the GRIB2 abbreviations shown by wgrib2 inventories and .idx files; xarray (cfgrib) renames them to ecCodes short names (e.g. t, gh). All fields are means of the 80 ensemble members (member files are on NOMADS only). There is no PRMSL as in CDAS (R1): mean sea level pressure is PRES at mean sea level. Relative vorticity and divergence (RELV, RELD) must be computed from the winds.

instantaneous mean accumulated maximum minimum static — instantaneous: value at the time stamp · mean/accumulated/maximum/minimum: over the indicated interval · static: does not change with time

Pressure-level analyses (pgb, 512 × 256 Gaussian, 22 levels)¶

Sea level pressure (pgb)¶

Surface and near-surface state (flx)¶

Precipitation, fluxes and clouds — 3-hour averages (flx)¶

2 m maxima and minima — 3-hour window (flx)¶

Runoff and sunshine — 3-hour accumulations (flx)¶

Fixed surface fields (flx)¶

The full variable list is in the official documentation.

How to get the data¶

• NOAA NODD — Google Cloud bucket noaa-nws-ncep-core — Direct HTTPS download / Google Cloud (gsutil) / get_core.py No login needed
  • Full record from 1950 to near real time. Ensemble means only (no members or statistics), as 3-hourly, daily and monthly GRIB2 under grib/3hour|day|month/flx|pgb/, with simplified, unique names such as flx.YYYYMMDDHH.grb, pgb.YYYYMMDD.grb and pgb.YYYYMM.grb. Each file has an .idx inventory, so single fields can be fetched by byte range. The NOAA NODD listing shows Google Cloud only; there is no AWS or Azure copy.
• NCEP NOMADS — operational CORe output — Direct HTTPS download No login needed
  • Last 7 days only. Cycle directories (core.YYYYMMDD/HH/post/) contain flx (members 001–080 and ensemble mean), flx_stats (ensemble statistics), nceppost (pgb ensemble mean), and spost / spost_stats (reduced pressure-level files per member, mean and statistics). This is the only place to get ensemble members.
• NSF NCAR GDEX — NOAA CORe (d090004) — Web download / THREDDS (OPeNDAP) / subsetting and format conversion / Globus
  • 3-hourly, daily and monthly flx (2D) and pgb (3D pressure-level) ensemble means in GRIB2, CC BY 4.0, updated daily. On 2026-10-12 the THREDDS catalogue showed no folders for 2003–2025, so use NODD if you need the full record.
• CPC download script get_core.py — Python script (downloads from NODD by date range and field regex) No login needed
  • Usage is documented at https://ftp.cpc.ncep.noaa.gov/CORe/get_core/get_core.txt. Example — get_core.py flx 195001 12 1 PRATE .

Example code¶

Downloads one 3-hourly pgb file from NODD and reads 500 hPa temperature with xarray (cfgrib). Fields have different numbers of levels (VVEL has 14), so opening everything at once fails; filter by shortName.

import urllib.request
import xarray as xr

url = ("https://storage.googleapis.com/noaa-nws-ncep-core/"
       "grib/3hour/pgb/2025/01/pgb.2025011512.grb")
urllib.request.urlretrieve(url, "pgb.2025011512.grb")

ds = xr.open_dataset(
    "pgb.2025011512.grb", engine="cfgrib",
    backend_kwargs={"filter_by_keys": {"typeOfLevel": "isobaricInhPa", "shortName": "t"},
                    "indexpath": ""},
)
t500 = ds["t"].sel(isobaricInhPa=500)   # K, 512x256 Gaussian grid
print(t500)

Practical tips¶

• CORe is not an improved R1. The model (T62 28-level MRF → C128 64-layer FV3GFS), assimilation (3D-Var → 80-member EnKF), observations (no satellite temperature retrievals), boundary conditions (newer SST, sea-ice and snow analyses) and grid (2.5° → 512 × 256 Gaussian) all differ. Do not join R1 or CDAS and CORe into one series; compute anomalies against each dataset's own climatology.
• The standard product is the ensemble mean, so peak winds in intense systems such as tropical cyclones are damped by the analysis spread. For cyclone tracking CPC suggests running the tracker on each of the 80 analyses and averaging the results. Member files exist only in the 7-day NOMADS window.
• Files are GRIB2 (wgrib2 recommended). CDAS GRIB1 scripts will not work as-is, and names change, e.g. PRMSL → PRES (mean sea level) and TMAX/TMIN → TMP (3-hour max/min). Select records by field and level, not by record number: CPC warns that flx files will gain 11 GRIB messages, and be renumbered, once the real-time fix is in.
• To regrid, follow CPC's wgrib2 guide: -new_grid with budget interpolation for precipitation (PRATE, CPRAT), nearest neighbour for soil, vegetation and slope types, and bilinear for everything else. CORe assumes a spherical Earth of radius 6371.229 km.
• Using only satellite AMVs reduces artificial jumps at the start of the satellite era, but CORe is not fully free of satellite data. CPC's evaluation notes that no global snow analysis exists before 1979, so a simple correction was applied to the model snow.
• CPC's evaluation notes that absolute shortwave and longwave fluxes depend heavily on clouds and are hard to get right, so use them mainly as anomalies; it also reports a zonally varying artefact in the computed top-of-atmosphere downward shortwave (DSWRFtoa).
• CPC has offered non-operational, evaluation-only "CDAS look-alike" files mimicking the R1 format (2.5° pressure levels, 192 × 94 flux grid, 6-hourly, GRIB1) on its FTP (CDAS_clone_temporary). We could not confirm whether they are still updated; use the official GRIB2 archive for research.

Notices and change history¶

References for your paper¶

Dataset citation¶

NCAR GDEX copy (d090004)

@misc{core_d090004,
  author    = {{National Oceanic \& Atmospheric Administration}},
  title     = {{NOAA} - Conventional Observation Reanalysis ({CORe})},
  year      = {2026},
  publisher = {NSF National Center for Atmospheric Research},
  url       = {https://gdex.ucar.edu/datasets/d090004},
  note      = {Accessed DD Month YYYY}
}

Reference papers¶

Early CORe design (workshop digest; no journal paper found yet) — DOI 10.7289/V5JS9NH0

@inproceedings{ebisuzaki2017core,
  author    = {Ebisuzaki, W. and Kumar, A. and Whitaker, J. and Woollen, J. and Lee, H.-C. and Zhang, L.},
  title     = {A preliminary examination of a conventional {EnKF} atmospheric reanalysis},
  booktitle = {Climate Prediction S\&T Digest, 41st NOAA Climate Diagnostics and Prediction Workshop},
  pages     = {148--151},
  year      = {2017},
  doi       = {10.7289/V5JS9NH0}
}

Initial CORe assessment (same digest) — DOI 10.7289/V5JS9NH0

Acknowledgements¶

CPC and NCEP give no official citation. State that you used the NOAA NCEP/CPC CORe reanalysis, where you obtained it (NODD, NOMADS or GDEX) and the access date; if you used the GDEX copy, add the dataset citation above (the UCAR data catalog lists the dataset DOI as "Not Assigned"). The DOI above belongs to the whole 41st CDPW digest (Climate Prediction S&T Digest, 2017), not to the individual papers.

Information on the datasets used in this study was compiled with the help of the Korean Climate Data Navigator (KCDN) (Sim, 2026; https://doi.org/10.5281/zenodo.23300923).

Links¶

• CPC — CORe home page
• CPC — CORe archives (NOMADS vs NODD)
• CPC — CORe datasets (CORe vs CDAS, timing conventions)
• CPC — CORe evaluation materials
• CORe configuration (44th CDPW, Ebisuzaki)
• CORe overview talk (9th NOAA Ensemble Users Workshop, 2023)
• PSL — About NOAA CORe Reanalysis
• NWS SCN 26-12 Updated (CDAS → CORe)
• CPC — Regridding CORe files (wgrib2)
• CPC — Making CORe time averages