Y2K Upgrade
                               --- -------

   During the latter part of 1997 and early 1998, the Reanalysis/CDAS
programs and scripts were made Y2K compliant.  One reason the Y2K
work was done in 1997 rather than in December 1999 was that we lost
the front-end which was responsible for processing the GTS data.  Another 
reason is that it is actually easier to fix the Y2K problems than to go 
through each piece of code, figure out what is not Y2K, document it and 
finally figure out how much time would be required to fix it (as required
by the local Y2K committee).


    Removal of HDS Mainframe:

        - went to the operational codes to decode the GTS data (March 1997)
        - a mistake was made in configuring the operational decoder,
             the India black-listed data was used causing an
             abnormally large Indian Monsoon
        - March 1997 to January 1998 data reprocessed

    RUC-II:

        - While the March 1997-January 1998 data were being reprocessed,
          the RUC-II model was being tested on Cray5.  This caused
          unpredictable failures in the Reanalysis codes whenever
          the RUC-II filled the process table.
        - Problems with the RUC-II were solved by recompiling the
          kernel with a larger process table, but apparently some
          of the post-processing jobs failed.  Failure to detect
          a problem may have been caused by the error-code failing
          because of an full process table.

    Changes to the GRIB files:

        - All Reanalysis GRIB files now use subcenter=1.  Previously
          some files did, and some didn't.
        - Previously the process id was set to 80 (T62 28 levels).
          Now the process id is a version number.  The first version
          is 80, the second version is 180, then 181, etc. Note,
          the reprocessed March 1997-January 1998 have a process id
          of 181 (version 2).
        - Monthly means now include SNOEV, SNOHF, SRWEQ.
        - see WGRIB (below)

    Change to the sigma spectral files:

        - The year code in the header was changed from 2 digits to 4.

    Y2K implementation:

        - The model and assimilation were converted to Y2K first.
        - The post-processing (monthly and daily means etc) were
          converted to Y2K a few months later.
        - Finally the theta post-processor was make Y2K.
        - Given the staggered Y2K implementation, some months may
          have been processed by Y2K and non-Y2K programs.  The
          routine CDAS for March 1998 was processed by the fully
          Y2K system.

    WGRIB problems:

        - with the change in the CDAS/Reanalysis GRIB files, wgrib
          incorrectly identified some of the GRIB files as being
          produced by NCEP operations rather than Reanalysis.
          As a result, the wrong GRIB table was used.  Wgrib needs
          to be upgraded to a version produced on or after 4/1/98
          (aka the April Fool's version).