August 21, 2001

MEMORANDUM TO:  Record
FROM:           Dennis A. Keyser -- NP22
Subject:        Changes to IBM-SP Data Preprocessor (PREPDATA)
                (February 17, 2000 Version -- UPDATE # 6)


The February 17, 2000 data preprocessor program (PREPDATA) was
updated for the sixth time.  UPDATE #6 was implemented on the
IBM-SP machine at 12Z on August 21, 2001.  Except where noted,
the changes below affect all five versions: ETA, AVN, FNL, RUC
and CDAS.

**************** I B M - S P   U P D A T E   # 6  ***************

I.  GENERAL CHANGES

     1) CDAS only: Namelist switch IRNMRK is now set to 7 for
CDAS-Reanalysis/SSI runs (prior to this it had been lumped
together with FNL-GDAS/SSI runs as 5). [MAIN, CDAS DATACARDS]

     2) The standard output "Welcome" now prints the particular
run type rather than the generic print of all run types, as
before. [MAIN]

     3) Namelist switch IWINDO was dissolved and two new switches
(with the same array sizes) were created called IWINDO_e (holding
the time window for the earliest satellite sounding/retrieval/
radiance reports) and IWINDO_l (holding the time window for the
latest reports).  In both cases, a "-" time means before cycle
time and a "+" time means after cycle time.  This allows the time
window to be asymmetric about the cycle time (see III.A.1 and
III.A.2). [MAIN, SATEDS, SNDTBL, SATBFR, GOESDG, GOESCHK,
DATACARDS]

     4) The BUFR mnemonic table was modified to correct some
errors in the description of some data types.  This change has no
effect on the structure of the PREPBUFR file. [BUFRTABLE]

     5) Subroutine W3ATOVSUNP modified as follows:
          a) Checks value of argument IERR on input.  If it is
not equal to 1 (meaning most likely 0), the assumption is made
that the input BUFR file is from a data dump (as before).  If it
is equal to 1, the assumption is made that the input BUFR file is
from a tank on /dcom.  In the case of the latter, no attempt is
made to obtain a center time or dump time in messages 1 and 2
since they do not exist, and there is a 250 line limit on
diagnostic warning prints about reports failing a surface
pressure d-val check and reports with one or more missing
temperatures in their profile (both of which are tossed).
          b) Now closes the input BUFR file after all reports are
processed (via call to BUFRLIB routine CLOSBF).
          c) Upon each entry into subroutine, it checks to see if
the input BUFR file is closed.  If this is the case, the
subroutine processes this entry as if it were the first entry
into the subroutine.  This allows a single program to read
through and process the same input BUFR file more than once or to
read through and process several different input BUFR files.



II. OPERATIONAL CHANGES SPECIFIC TO A DATA TYPE

  A. CHANGES TO GOES SFOV CLOUD TOP DATA PROCESSING

     1) Subroutine W3UNPKB7 now tallies separately a count of all
reports read from the "goesnd" dumnp file with a missing field-
of-view value (e.g., the SFOV cloud top reports).  Prior to this,
the count of reports with missing f-o-v was incorrectly included
with the count of reports with f-o-v of 10 to 25. [W3UNPKB7]



III. NON-OPERATIONAL CHANGES SPECIFIC TO A DATA TYPE

  A. CHANGES TO GOES SFOV CLOUD TOP DATA PROCESSING

     1) ETA only: The time window for processing GOES-08 and
GOES-10 cloud top reports was changed from +/- 3 hours about
cycle time (effectively -3 to -0.01 hours since the dump window
for these data was set to this) to -3.25 to -0.25 hours about
cycle time for GOES-08 and -3.00 to -0.01 hours about cycle time
for GOES-10 [possible now with new switches IWINDO_e and IWINDO_l
(see I.3)] (the dump window for both satellites is now -3.25 to -
0.01 hours).  This was done so that all of the cloud top data
from a particular satellite would be from the same retrieval time
sets.  These data are written into the PREPBUFR files but are not
yet used by the Eta model. [DATACARDS] 

     2) ETA only: Restored the processing of GOES SFOV cloud top
data into the PREPBUFR files (Namelist switch GOESCT(2) changed
from FALSE to TRUE.) (The BUFRLIB has been modified to increase
the array size for MAXMEM from 8 MB to 16 MB.) [DATACARDS]

     3) RUC only: The time window for processing GOES-08 and
GOES-10 cloud top reports was changed from +/- 1 hour about cycle
time (effectively -1 to +0.43 hours since the dump cutoff time is
T+0:26) to -1.25 to -0.25 hours about cycle time for GOES-08 and
-1.00 to -0.01 hours about cycle time for GOES-10 [possible now
with new switches IWINDO_e and IWINDO_l (see I.3)] (the dump
window for both satellites is now -1.25 to -0.01 hours).  This
was done so that all of the cloud top data from a particular
satellite would be from the same retrieval time set.  These data
are written into the PREPBUFR files but are not yet used by the
RUC. [DATACARDS]

  B. CHANGES TO ATOVS 40-LEVEL SOUNDING DATA PROCESSING

     1) In response to change to subroutine W3ATOVSUNP (see
I.5.a), now passes the value IER=0 into this subroutine upon
first call. [MAIN, SATBFR]