GSI On-Line tutorial

Download GSI system and compile (version 1.0)

After this section, user should get gsi executible

  1. Goto the follow page and download GSI tar file to local machine:
    download GSI

  2. move tar file to work directory and untar the file :
  3. 	 tar -zxvf comGSI_v1.tar.gz  

  4. You will see directory comGSI_v1. cd into it
  5.            cd comGSI_v1  

  6. set environment
    • after compiling the WRF (either EM or NMM core)
           setenv WRF_DIR "path for WRF"  
    • Check $NETCDF using:
             echo $NETCDF 

      if it's empty or doesn't link to the libraries that were installed using compile you are using, set it using:
            setenv NETCDF  "path for netCDF" 
  7. create GSI configuration
  8.          ./configure

    Pick the compiler you are using. Right now, the follow platforms and compilers have been tested:
    • AIX
    • LINUX pgi(7)
    • LINUX ifort(10.1)
    Will see a new file configure.gsi. Edit it if necessory.

  9. compile
  10.           ./compile >& compile_gsi.log 

  11. check executible
  12. 
                  cd ./run  
                  ls gsi.exe
                  

    if no gsi.exe under directory run, check file compile_gsi.log for errors.

practice case one: May 11th, 2008 east US cyclone with ARW background

After successfuly compiling GSI, user can follow the following steps to run GSI:

  1. create a directory (for example: casedata) to save data (anywhere you have write permit)
  2. 
            mkdir casedata 
            cd casedata
                
  3. download the data
  4. goto Download/Practical Case page (link)
    and download background, observation data to casedata directory.


  5. untar observation data file and unzip background files
  6.  
       tar -zxvf obs.tar.gz 
       gzip -d wrfout_d01_ARW_2008-05-11_12:00:00.gz   (ARW background)
       gzip -d wrfinput_d01_NMM_2008-05-11_12:00:00.gz (NMM background)
             
  7. create a work directory (for example: testgsi)
  8.  
          mkdir testgsi 
          cd testgsi 
             
  9. copy running script to work directory
  10. Example running script (run_gsi_v1.ksh) is in unloaded GSI system under ./run directory.

  11. edit running scripts: run_gsi_v1.ksh
    • depends on the computer system, the head of the script should be changed. This is the same as you run other jobs
    •  
       For example:
        IBM:
         #BSUB -a poe                   
         #BSUB -x                   # exlusive use of node (not_shared)
         #BSUB -n   12              # number of total tasks
         #BSUB -R "span[ptile=2]"   # how many tasks per node (up to 8)
         #BSUB -J gsi               # job name
         #BSUB -o gsi.out           # output filename (%J to add job id)
         #BSUB -e gsi.err           # error filename
         #BSUB -W 00:02
         #BSUB -q regular           # queue
      
        Linux Cluster (with ifort compile):
         #$ -S /bin/ksh
         #$ -N GSI
         #$ -cwd
         #$ -r y
         #$ -pe comp 8
         #$ -l h_rt=0:20:00
         #$ -A wrfruc
      
      
         ###big endian conversion options for use with Intel compilers
             export F_UFMTENDIAN="big;little:10,13,15,66"
             export GMPIENVVAR=F_UFMTENDIAN
      
             export MV2_ON_DEMAND_THRESHOLD=256
      
        Linux workstation:
             no head, delete other heads 
               
    • define machine type (pick one close to your computer)
    •   For IBM:
           machine=BLUEFIRE
        For Linux cluster with ifort compile:
           machine=LNXC
        For Linux worstation with pgi compile:
           machine=LINDEN
               
    • define the following variables in run_gsi_v1.ksh based on previous setup:

    ANAL_TIME 2008051112 analysis time (YYYYMMDDHH)
    WORK_ROOT path/testgsi working direcotry, where GSI runs
    BK_FILE path/wrfout_d01_ARW_2008-05-11_12:00:00 background file
    OBS_ROOT path/casedata path of observations
    PREPBUFR ${OBS_ROOT}/newgblav.gdas1.t12z.prepbufr.nr conventional observation
    FIX_ROOT path/comGSI_v1/fix path of fix files
    GSI_EXE path/comGSI_v1/run/gsi.exe gsi executable

    NOTE: for IBM: please link the observation file : gdas1.t12z.prepbufr.nr

  12. run gsi
  13. The example script for linux workstation (Linden) can be found in: (ARW)
    Dependes on computer system, here are some examples:
       run_gsi_v1.ksh  (linux workstation)
       qsub run_gsi_v1.ksh (linux cluster)
       bsub < run_gsi_v1.ksh    (IBM)
             
  14. check the results
  15. cd gsiprd

    check stdout and compare to reference stdout (ARW)
    check files in gsiprd and compare to reference file list (ARW)

single observation test

The single observation test can be done by editing the above runnig script:

  1. change the follow variable in run_gsi_v1.ksh
  2.     &SETUP
          oneobtest=.true.,
    
       &SINGLEOB_TEST
          maginnov=1.0,magoberr=0.8,oneob_type='t',
          oblat=38.,oblon=279.,obpres=500.,obdattim=${ANAL_TIME},
          obhourset=0.,
       $SINGLEOB
             
  3. run gsi
  4. the same as step 7 in above case The example script for linux workstation (Linden) can be found in: (single example)

  5. check results
    • check stdout and compare to reference stdout
    • (ARW)
    • Check the analysis increment plots
    • XY and XZ section through the maximum analysis increment point: (pdf)
      NCL script to make the plot can be find here : (ncl script)