Note for gdas:

The analysis system is based on rather old NCEP/DOE Reanalysis, but the forecast model and the treatment of surface fields are from the latest CVS version of GSM. Note that all the analysis components, including Statistical Spectral Interpolation Scheme (SSI) are NOT parallelized. The code was tested on Linux and IBM-SP machines.

The "gdas" script utilizes input observation and analysis guess stored in the libs/con directory. The date of the analysis is December 1, 1978. If you want to run other dates, you need to edit gdas script (see the detail in the last part of this page). Limited number of observation files and initial guess files are found in the ECPC data server.

The analysis scheme utilize satellite retrievals of temperature, rather than the radiance itself. The NCEP analysis scheme went through many improvements and changes since this version has been developed. See the recent status of NCEP global analysis system here.

Go to installation for the procedure for using CVS to install, compile and execute.

Some more nice document written by Seung-On Hwang is available here .

Observation BUFR file, guess sigma file, guess sfc file, 6-hour earlier analysis sigma file (if available):

Note: For more general specification, 78120100 can be replaced by \$YYYY\$MM\$DD\$HH
e.g. INPUT_BUFR_FILE_NAME=/rokka3/kana/model/gdas/libs/con/prepbufr\$YYYY$MM$DD$HH.bufr
The script automatically identifies file format by the last file identifier (asc, ieee, etc...).
SIG_ANL_PREV can be blank. It is only used in CQC for time continuity check.
Makef sure that SFCFTYP is properly specified for given SFC_GUESS file (osu1, osu2, noa1, ...)
Set INPUT_BUFR_BLOCKING to no most of the cases. If progam fails at the bufrblk step try cray.

Observation SST, snow cover or snow depth analysis and ice cover/concentration analysis grib files:

Note: Again, 78120100 can be replaced by \$YYYY\$MM\$DD\$HH for more general case.
Provide either SCV_GRB (Snow cover) or SNO_GRB (snow depth). If not available, leave them blank.
ICE_GRB should be ice cover mask or ice cover percentage. The grib header should tell which.

Starting and ending dates:

Soil moisture adjuestment:
If soil moisture adjustment using observed pentad precipitation, set PRECIPITATION_SOIL_ADJUSTMENT=yes
Make sure that you have observed pentad precipitation grib file properly specified. OBS_PREC_FILE=/rokka3/kana/model/gdas/libs/con/p_pingrainT62_mask1.grb

Explanation of program execution steps:
  1. 0. block -- blocks/unblocks bufr file for efficiency
  2. 1. grepsat -- remove all sat data (from independent sat qc) if switched on.
  3. 2. edupat -- remove all radiosondes (from independent obs qc) if switched on.
  4. 3. fixpaobs -- fix paobs 180 degree location error (no longer needed)
  5. 4. edsatw -- remove satwind (optional)
  6. 5. edswnd -- remove TIROS sat wind if switched on.
  7. 6. eddrib -- remove drifting buoys (optional)
  8. 7. ednosat -- remove space based observing system (optional)
  9. 8. edbuoy -- remove buoy using ECMWF blacklist
  10. 9. check -- check to see whether all the necessary files for gdas exist
  11. 10. prevents -- compute increments and split the bufr file into two parts for cqc and acqc.
  12. 11. cqc -- complex quality control
  13. 12. acqc -- aircraft quality control
  14. 13. combbufr -- combine qc'ed radiosonde and aircraft bufr files
  15. 14. oiqc -- optimum interpolation quality control
  16. 15. ssi -- spectral statistical interpolation analysis
  17. 16. sfc -- set up surface conditions
  18. 17. precipadj -- adjust soil mositure using observed precipitation
  19. 18. fcst -- atmospheric model integration
  20. 19. pgb -- sigma to pressure interpolation of analysis field
  21. 20. copy -- copy/ftp created files to the archive directory
  22. 21. postvents -- create final bufr file with meta data

When the execution is successful, you should see the following files in the G_000 directory:

0LOG.bindates cqc.x guessig sdmstac
acqc.out cqe.anl.ascii nmcdate sfc0.out
acqc.parm cqt.anl.ascii obcbt.out sfc0.parm
acqc.x cwords.x obcycle.anl.1978120100.bufr sfc0.x
bufr_adpupa debugout obogram.bin sfc.anl
bufr_adpupa_cqc_output edbuoy.out obogram.out sfc.ft06
bufr_aircft edbuoy.x obprt.wrk sfci
bufr_aircft_acqc_output empty1 oiqc.anl.bufr sfco
bufr_sfcsat empty2 oiqc.out sig.anl
ccwords.c empty3 oiqc.x sig.ft00
ccwords.o empty4 out.bufr sig.ft06
combbufr_bufr_output ERROR.out pgb.ctl sigit
combbufr.out fcstout.ft00 pgb.ft00 sigitdt
combbufr.parm fcst.parm pgb.ft06 sigp.ft06
combbufr.x fcst.x pgb.idx ssi.out
cqb.anl.ascii fcwords.f pgf90.out ssi.parm
cqc11 fcwords.o prepbuoy ssi.x
cqc13 flx.ft00 prevents.out tolls.wrk
cqc16 flx.ft06 prevents.x tosslist
cqc61 grmap.out print.out toss.sat
cqc62 grmap.parm toss.sfz
cqc65 grmap.x scnvrt.out toss.upa
cqc.out guessfc sdmacqc

For debugging purpose, you may specify the step number to skip program executions:
"gdas 14" starts from ioqc. See the comment at the beginning of the gdas script.

Sample output files:

Machine dependency:
  • for Linux with PGI6 or higher, "-Msave" option should be added to opt/options-linux-single. Otherwise it stops in "oiqc".

How to prepare and change the gdas script to make multiple cycles

You need the following files:
1. prepbufr observation files (6-hourly).
2. SST analysis file (daily).
3. Sea ice analysis file (daily or weekly).
4. Snow analysis file (daily or weekly).
5. For cold start only (very beginning of the analysis cycle)
5.1 6-hour forecast guess sigma file valid at analysis time
5.2 Same as 5.1 but guess sfc file.
A) How to obtain those files and how to specify them in the gdas script.
1. Prepbufr file.
cd /home/afu/kana/prepbufr_new/yyyymm
I found that currently files are available from 199601 to 200109 but I can add more later. Also pre 1996 files are available in vsn01 directory.
2. SST, Sea-ice, snow files (GRIB)
Currently, our data server is down. Please try:
cd /home/afu/kana/sfcanl_new/yyyy
get sst.grib
get ice.grib
get snow.grib
3. Guess files
cd /home/afu/kana/rerunarch/yyyymm/prod_v02.A.vsn02/guess
Chose yyyymmddhh in such a way that the 6-hour forecast starting from yyyymmddhh is valid at the very initial analysis time. For example, if you are cold starting your analysis from 1999010100, yyyymmddhh should be 1998123118.
1-3 files should also be available from Detail will be posted later.
B) How to change the gdas script.
Provide desired starting and ending year, month, day, hour of the data assimilation cycle.
Then use the following definitions for the variables:
Please provide yyyymmddhh appropriately.
Please note that the variables $YYYY, $YY, $MM, $DD, $HH are available and properly renewed for each cycle. Watch for the statement with `eval echo` in the grads script.