NOAA NWS DOC Skip Naviagation Links

On July 30, 2013, NCEP will transition from the current operational Central Computing System (CCS) to the Weather and Climate Operational Supercomputing System (WCOSS). Live tests will be run with dissemination of data on June 25th and July 2nd. Due to the transition, there will be some product changes. They are listed below:

Product Timeliness Changes

  • Global Forecast System (GFS) - 4 minutes faster
  • Climate Forecast System (CFS) - 2/3 hours faster
  • Real Time Ocean Forecast System (RTOFS) Global days 3-6 - 6 hours faster
  • Real Time Ocean Forecast System (RTOFS) Atlantic - 3 hours faster
  • Air Quality Model (AQM) - 70 minutes faster

Product Output Changes

  • Global Ensemble Forecast System (GEFS)
  • -The calculation of CAPE/CIN has been changed to use virtual temperature
  • Climate Forecast System (CFS)
  • -The calculation of CAPE/CIN has been changed to use virtual temperature
    -The precision in the GRIB fields will change for the fluxes and WESD
  • Global Forecast System (GFS)
  • -The ordering of the fields in the GRIB files will change
      CCS - gfs.tCCz.pgrb2fFF
       gfs.tCCz.pgrb2fFF.1p0deg
       gfs.tCCz.pgrb2fFF.2p5deg
      ftpprd - gfs.tCCz.pgrb2fFF
       gfs.tCCz.pgrbfFF.2p5deg.grib2
       gfs.tCCz.pgrbfFF.grib2
      tggate - fh.xxxx_tl.press_gr.0p5deg
       fh.xxxx_tl.press_gr.1p0deg
       fh.xxxx_tl.press_gr.2p5deg
  • North American Model (NAM)
  • -The ordering of fields in the 5km CONUS nest output GRIB files will change
      nam.tCCz.conusnest.hiresfFF.tm00.grib2
    -The precipitation analysis output on the ftp server will be changing compression methods from "compress" to "gzip"; hence the names will change from suffix ".Z" to suffix ".gz"
      ST4.YYYYMMDDCC.HHh.Z > ST4.YYYYMMDDCC.HHh.gz
      ST2??YYYYMMDDCC.(Grb|HHh).Z > ST2??YYYYMMDDCC.(Grb|HHh).gz
      (gage|multi|rad|radunb)##.YYYYMMDDCC.HHh.Z > (gage|multi|rad|radunb)##.YYYYMMDDCC.HHh.gz
  • European Center for Medium Range Weather Forecasting (ECMWF) products will no longer be avaiable on the NCEP FTP server
  • -/mrf/prod/ecmwf.YYYYMMDD/nrecmgrb25.t00z.grib2
  • Global Real Time Ocean Forecast System (RTOFS)
  • -RTOFS Global was reconfigured to run on WCOSS for 8 forecast days as compared to only 6 forecast days on CCS. This was accomplished by increasing the forecast length for forecast steps 1 and 2 to 4 days each and eliminating forecast step 3.
    As a result, forecast steps 1 and 2 have double file files, and step 3 (rtofs_glo.fcst3* files) has been removed
    -Global NetCDF output will be increasing from a 6 day to an 8 day forecast and available from ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/
    -Native archive files will be made available on NCEP FTP server
  • Climate Data Assimilation System (CDAS)
  • -The following observation files will no longer be "blocked", and will be replaced by Endian-independent versions:
      cdas1.tCCz.cbathy.tm00.mbufr_d
      cdas1.tCCz.cdbuoy.tm00.mbufr_d
      cdas1.tCCz.ctesac.tm00.mbufr_d
    -Starting June 13th, both the blocked and the new unblocked files from the CCS will be made available at ftp://ftpprd.ncep.noaa.gov/pub/data/nccf/com/cfs/prod/. You will find the new unblocked with suffix .unblok. Starting July 30th the suffix will be removed.
  • Climate Data Center (CDC)
  • -This model is being discontinued, the following products will no longer be available from ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/:
      enspost.tCCz.prcp.grib2
      enspost.tCCz.t995.grib2
    -Link to Technical Information Notice
  • Short Range Ensemble Forecast (SREF) System
  • -Fixed incorrect vertical velocity field from NMMB and NMM members
    -Ensemble products (mean, spread and probability) of visibilty and ceiling are now based on all 21 SREF members.
  • Northern Gulf of Mexico Ocean Forecast System (NGOFS)
  • -The NGOFS output files are being broken into individual files by hour
      nos.ngofs.fields.[forecast|nowcast].YYYYMMDD.tCCz.nc > nos.ngofs.fields.[f|n]FFF.YYYYMMDD.tCCz.nc
  • National Ocean Services (NOS)
  • -Columbia Revier Estuary Ocean Forecast System (CREOFS) fields output file name is changing
      nos.creofs.fields.fFF.YYYYMMDD.tCCz.nc > nos.creofs.fields.f0FF.YYYYMMDD.tCCz.nc
    -Chesapeake Bay Ocean Forecast System (CBOFS) is replacing the large nowcast file with individual smallers ones
      nos.cbofs.fields.nowcast.YYYYMMDD.tCCz.nc > nos.cbofs.fields.nFFF.YYYYMMDD.tCCz.nc
    -CBOFS "roms" files are being renamed
      nos.cbofs.roms.nowcast.YYYYMMDD.tCCz.in > nos.cbofs.nowcast.YYYYMMDD.tCCz.in
    -Delaware Bay Ocean Forecast System (DBOFS) is replacing the large nowcast file with individual smallers ones
      nos.dbofs.fields.nowcast.YYYYMMDD.tCCz.nc > nos.dbofs.fields.nFFF.YYYYMMDD.tCCz.nc
    -DBOFS "roms" files are being renamed
      nos.dbofs.roms.nowcast.YYYYMMDD.tCCz.in > nos.dbofs.nowcast.YYYYMMDD.tCCz.in
    -Tampa Bay Ocean Forecast System (TBOFS) "roms" files are being renamed
      nos.tbofs.roms.nowcast.YYYYMMDD.tCCz.in > nos.tbofs.nowcast.YYYYMMDD.tCCz.in
  • WAVEWATCH III
  • -The spectral output files that are currently tarred together will now be gzipped prior to being placed on ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/
      glw.tCCz.spec_tar > glw.tCCz.spec_tar.gz
    -The following will have name changes on the ftp server:
      multi_1.icean_5m.tCCz.ice.v.3.0.6 > multi_1.icean_5m.tCCz.ice
      mens.tCCz.wind > mens.gfs_60m.tCCz.wind & mens.icean_5m.tCCz.ice
      (enp|nph).tCCz.wind > multi_2.henp_15m.tCCz.wind
      (nah|wna).tCCz.wind > multi_2.hwna_15m.tCCz.wind
      gep##.tCCz.wind > gep##.gfs_60m.tCCz.wind
      icean_5m.tCCz.ice > mens.icean_5m.tCCz.ice
    -These products are discontinued and the following output is no longer generated:
      nww3.tCCz.ice
      nww3.tCCz.next_restart
      nww3.tCCz.wind - A substitute will be multi_1.gfs_30m.tCCz.wind
      wna.tCCz.ice - A substitute will be multi_1.icean_5m.tCCz.ice
    -The spectral output files that are currently tarred together will now be gzipped prior to being placed on ftp://ftp.ncep.noaa.gov/pub/data/nccf/com/
      (nah|nph).tCCz.spec_tar > (nah|nph).tCCz.spec_tar.gz
  • NCEP FNMOC Combined Ensemble (NFCENS)
  • -The station output files placed on the ftp server will be changing compression methods from "compress" to "gzip"; hence the names will change from suffix ".Z" to suffix ".gz"
      HTSGW_station.tCCz.text.Z > HTSGW_station.tCCz.text.gz
  • Air Quality Model (AQM)
  • -Products for Hawaii and Alaska with headers like: L(NOUV)(RS)(ABZ).. will be missing for a period after the transition occurs from ftp://tgftp.nws.noaa.gov/SL.us008001/ST.opnl/DF.gr2/DC.ndgd/GT.aq/

Model System Upgrades

  • Blocked/Unblocked BUFR
  • -All BUFR files created on WCOSS are now unblocked. This decision was made back in September as a result of the Endianess changes on WCOSS (Little-Endian on WCOSS vs. Big-Endian on CCS) and its affect on the blocking control words. Users are recommended to recompile their codes to link to the latest NCEP BUFRLIB which can read unblocked BUFR files found here
  • North American Model (NAM)
  • -NAM Grid-scale Interpolation (GSI) analysis upgraded to a newer version with bug fixes for how satellite radiance data is thinned and binned for use in the analysis. This allows for as much as, or in some instances, more satellite radiance data to be assimilated in the WCOSS version.
    -Differences in the GDAS/GFS between CCS and WCOSS could affect the NAM, since the GDAS is used to start the NAM Data Assimilation System (NDAS) and the GFS is used to provide lateral boundary conditions to NDAS and NAM forecasts
  • Hybrid Single Particle Langrangian Integrated Trajectory (HYSPLIT)
  • -Upgraded to a newer version
    -Link to Technical Information Notice
  • Hurricane Weather Research and Forecast (HWRF) System
  • -Upgraded to a newer version
    -Link to Technical Information Notice
  • Short Range Ensemble Forecast (SREF) System
  • -All base models, NMMB, NMM and ARW are upgraded to a newer version but all models reproduce CCS results

Differences in Model System Solutions

We have made every effort to make the WCOSS and CCS solutions as close as possible to each other. However, there should not be an expectation that the WCOSS and CCS solutions will be the same. There are several contributors to differences in the solutions.

  1. The numerical round-off and order of calculation is different between the two machines. So when two numbers are multiplied together or any other arithmetic operation, the results will be slightly different. These differences can then grow non-linearly in the forecast and in the assimilation system.
  2. The input observation databases will not be identical since they are created independently. Due to small timing differences in the creation of the databases and communication speeds on the CCS and WCOSS machines, there may be slightly different data being used in the two systems.
  3. In the process of converting to the new computational infrastructure, some bugs were discovered. These bugs did not appear to have much impact (or in most cases any impact) on the CCS, but some had very severe impact on the stability or results on the WCOSS machine. Obviously, we had to fix these results which can result in small changes in the results.
  4. Some IBM power series functions were being used on the CCS (for example FFTs, and random number generators). These functions were not available on the WCOSS machine and substitutes were created. These substitutes produce similar results, but not identical.
For all of these reasons, the forecasts will not be identical and the differences have the potential to be more different the longer the assimilation systems are run independently and for longer term forecasts. We continue to monitor the systems to minimize the possibility that a real error in the transition to the new machine will occur, but since the results will naturally be different, it is not possible to perform a definitive test. I will note that there have been several problems with the NCO WCOSS parallel. These problems have occurred when there were machine problems, the parallel was switched from one machine to another and when files were not properly supplied to the systems. When the problems occur and are fixed NCO has been restarting the parallel from the CCS results. NCO is making every effort to eliminate these issues, but it is important that the evaluators do not waste their time on forecasts created when a problem exists. Because of the expected differences, we do not believe that you should be worried about changes in individual forecasts, but rather about systematic differences between the two systems which may indicate a problem in the conversion. We appreciate your evaluations of the two systems and appreciate notification of any systematic differences as soon as possible so that any potential problem can be diagnosed and if necessary fixed in the WCOSS parallel.

Page last modified: 27 April 2023.