CFSv2

[emiliom]

Collecting relevant links and information for CFSv2 output data

• https://www.ncdc.noaa.gov/data-access/model-data/model-datasets/climate-forecast-system-version2-cfsv2
• https://rda.ucar.edu/datasets/ds094.1/
  • https://rda.ucar.edu/datasets/ds094.0/#docs/FAQs_6hrly.html
  • https://rda.ucar.edu/datasets/ds094.0/#docs/cfsv2_filenames.html
• https://www.ncei.noaa.gov/thredds/model/cfs.html
• https://cfs.ncep.noaa.gov/
  • https://cfs.ncep.noaa.gov/cfsv2/downloads.html
  • https://cfs.ncep.noaa.gov/cfsv2/docs.html
• https://catalog.data.gov/dataset/climate-forecast-system-version-2-cfsv2-operational-forecasts
• https://catalog.data.gov/dataset/climate-forecast-system-version-2-cfsv2-operational-analysis
• https://nomads.ncep.noaa.gov/pub/data/nccf/com/cfs/prod/

Google Earth Engine

• https://github.com/snowmodel-tools/preprocess_javascript/blob/master/define_SM_inputs.js#L283
• https://developers.google.com/earth-engine/datasets/catalog/NOAA_CFSV2_FOR6H
• https://explorer.earthengine.google.com/#detail/NOAA%2FCFSV2%2FFOR6H
• Refer also to the 10/27/2020 email from Dave, "Fwd: climate model time steps"

[emiliom]

It looks like the specific CFSv2 dataset that GEE is using ("We ingest only a subset of bands from files matching cdas1.t??z.sfluxgrbf06.grib2") is the CFSv2 Operational Forecasts - CFSv2 Operational Forecasts 6-Hourly Products, 9-Month Forecasts: 6-Hourly Surface and Radiative Fluxes (FLX); which can be accessed at this NCEI THREDDS server dataset: https://www.ncei.noaa.gov/thredds/catalog/model-cfs_v2_for_6h_flxf/catalog.html.

Here's one sample time-step dataset, for the NCSS endpoint page: https://www.ncei.noaa.gov/thredds/ncss/model-cfs_v2_for_6h_flxf/2020/202010/20201017/2020101718/flxf2020102012.01.2020101718.grb2/dataset.html

Note the file name and directory tree structure (where mr is model run):
mryyyy/mryyyymm/mryyyymmdd/mryyyymmddhh/flxyyyymmddhh.01.mryyyymmddhh.grb2
I don't know what the 01 is, but it seems to be fixed.

Update 12/20/2020: I've resolved that the GEE dataset is actually the analysis product, not the 9-month forecasts products. See below, and more information to come.

[emiliom]

GEE documentation states that the CFSv2 cell resolution is 0.2 deg. But it's not documented whether this is the actual model output resolution. I've concluded that this is indeed close (roughly) to the T574 Gaussian grid used in the CFSv2 operational analysis product, whereas the CFSv2 operational forecast product uses the T126 Gaussian grid (approx. 0.9 deg).

Resources:

• NCEP Grid identification and description: https://www.nco.ncep.noaa.gov/pmb/docs/on388/tableb.html
• http://www.ncl.ucar.edu/Document/Functions/sphpk_grids.shtml#GaussianGrids
• https://en.wikipedia.org/wiki/Gaussian_grid

[emiliom]

Notebook implementing ingest of the "CFSv2 Operational Analysis - 6-Hourly Surface and Radiative Fluxes (FLX)" product:
https://github.com/snowmodel-tools/preprocess_python/blob/master/CFSv2_fromNCEI_TDS.ipynb

[emiliom]

CFSv2 analysis product: CFSv2 Operational Analysis 6-Hourly Products

• Characteristics
  • Product: 6-Hourly Surface and Radiative Fluxes (FLX)
  • Grid/Scale: Gaussian: T574
  • Model Cycle: 4/day: 00, 06, 12, 18UTC
  • Output Timestamp: +00, +01, +02, +03, +04, +05, +06, +09
• From https://rda.ucar.edu/datasets/ds094.0/#docs/cfsv2_filenames.html
  • Inside of each tar file will be the model output files from the individual CFSV2 model initializations (the model is initialized four times per day). The file names will have the form cdas1.tHHz.XXXXXFF.grib2, where:
    • XXXXX is a file type (see above)
    • HH is the initialization time (00, 06, 12, or 18 UTC)
    • FF is the forecast hour (the time at which the data are valid)
  • Forecast hours:

FF	Description
anl	Analysis
01-09	Number of hours after initialization time at which the data are valid

• From GEE dataset description: "We ingest only a subset of bands from files matching cdas1.t??z.sfluxgrbf06.grib2"
• THREDDS catalog crawling
  • https://www.ncei.noaa.gov/thredds/catalog/model-cfs_v2_anl_6h_flxf/2020/202011/20201102/catalog.html
  • https://www.ncei.noaa.gov/thredds/catalog/model-cfs_v2_for_6h_flxf/2020/202010/20201017/2020101718/catalog.html
• See the differences in time stamps and list of variables between these two datasets/files that differ only in "f" "output timestep" (f01 vs f06)
  • https://www.ncei.noaa.gov/thredds/ncss/model-cfs_v2_anl_6h_flxf/2020/202010/20201023/cdas1.t18z.sfluxgrbf01.grib2/dataset.html
  • https://www.ncei.noaa.gov/thredds/ncss/model-cfs_v2_anl_6h_flxf/2020/202010/20201023/cdas1.t18z.sfluxgrbf06.grib2/dataset.html
• Generating a sequence of incremental dates and times: Use pd.date_range: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.date_range.html. Will use this sequence for the date, then cycle through the 4 "model cycles" for the time: t00z, t06z, t12z, t18z
• Other sample files on the THREDDS server, to compare content. The only differences are in the "t" and "f" file name tokens (eg, t00z and f05)
  • https://www.ncei.noaa.gov/thredds/catalog/model-cfs_v2_anl_6h_flxf/2020/202011/20201102/catalog.html?dataset=cfs_v2_anl_6h_flxf/2020/202011/20201102/cdas1.t00z.sfluxgrbf05.grib2
  • https://www.ncei.noaa.gov/thredds/catalog/model-cfs_v2_anl_6h_flxf/2020/202011/20201102/catalog.html?dataset=cfs_v2_anl_6h_flxf/2020/202011/20201102/cdas1.t06z.sfluxgrbf09.grib2
  • https://www.ncei.noaa.gov/thredds/catalog/model-cfs_v2_anl_6h_flxf/2020/202011/20201102/catalog.html?dataset=cfs_v2_anl_6h_flxf/2020/202011/20201102/cdas1.t12z.sfluxgrbf08.grib2