Running ADAS: Difference between revisions

Fix the 2nd rename and add a sed command for AGCM.rc
Pchakrab (talk | contribs)
 
(107 intermediate revisions by 2 users not shown)
Line 1: Line 1:
First, we choose a DAS run to simulate. For a list of various runs
==ADAS run==


  $ ls ~dao_ops
===Build model===
First, choose a model tag to build. For this example, we choose GEOSadas-5_9_1_p7.


on NCCS discover. As an example, we choose the run 'e572p5_fp'.
$ cvs co -P -r GEOSadas-5_9_1_p7 -d g591p8 GEOSadas-5_9


1. Checkout and build the model:
Not the module name <tt>GEOSadas-5_9</tt>. Unlike AGCM, all ADAS tags have a corresponding module.


Check with Joe Stassi for model tag and module name.
$ cd g591p8/src
$ ./parallel_build.csh


  $ cvs co -r GEOSadas-5_7_2_p5_m1 GEOSadas-5_7
===Setup experiment (Run <tt>fvsetup</tt>)===
  $ cd GEOSadas/src
 
  $ source g5_modules
Run <tt>fvsetup</tt> from the bin directory (<tt>g591p8/Linux/bin</tt>) with the following options (default value chosen for ones not mentioned):
  $ gmake install
[we choose the default value for the initial conditions (FVICS) as well]
 
AGCM Resolution? [b72]
> C180
OGCM Resolution? [c34]
> f34
EXPID? [u000_C180]
> g591p8
OBSERVING SYSTEM CLASSES?
> "ncep_prep_bufr"
<!--
> "ncep_1bamua_bufr,ncep_1bamub_bufr,ncep_1bhrs2_bufr,ncep_1bhrs3_bufr,
    ncep_1bmsu_bufr,ncep_osbuv_bufr,ncep_prep_bufr,ncep_sptrmm_bufr,
    disc_airs_bufr,disc_amsua_bufr,ncep_mhs_bufr,ncep_1bhrs4_bufr,
    ncep_amsre_bufr,ncep_goesfv_bufr,ncep_mtiasi_bufr,ncep_gpsro_bufr,
    aura_omi_bufr,ncep_satwnd_bufr,ncep_atms_bufr,ncep_sevcsr_bufr"
-->
Do Aerosol Analysis (y/n)? [y]
> n
 
One can, of course, turn on Aerosol Analysis and AOD assimilation here.
 
A pdf of an example <tt>fvsetup</tt> session is provided here: [[Media:Fvsetup_session.pdf|fvsetup_session]].
 
====FVHOME====
FVHOME is the home directory for fvDAS experiment. Resource files, restarts and system output will be stored
under this directory. Usually it contains subdiretories:
 
ana/        first guess/analysis output
anasa/      job script to run stand-alone analysis
daotovs/    daotovs field output
diag/      diagnostic field output
etc/        listings and other odds & ends
fcst/      forecast run directory
fvInput/    process required inputs
obs/        post-analysis ODS files
prog/      prognostic field output
recycle/    latest restart files
rs/        restart files
run/        resource files
 
<tt>$FVHOME/.FVROOT</tt> gives the location of the installation <tt>bin</tt> directory.
 
Some of these directories (<tt>ana/</tt>, <tt>chem/</tt>, <tt>diag/</tt>, <tt>etc/</tt>, <tt>obs/</tt>, <tt>rs/</tt>) are stored in the archive. After completing the experiment, look in <tt>$ARCHIVE/g591p8</tt> (where <tt>g591p8</tt> is the experiment ID).
 
====Run directory====
$ cd $FVHOME/run
 
<tt>g5das.j</tt> is the job submission script. The number of cpus etc. can be changed here (if they haven't already been changed during setup). Also, if you want to run a single cycle of <tt>GSIsa.x</tt> & <tt>GEOSgcm.x</tt> add <tt>exit</tt> right before 'PART IV'.
<!--
To simulate the OPS run d591_fpit, we make a few changes:
# NX, NY in AGCM.rc.tmpl and AGCM.BOOTSTRAP.rc.tmpl have the same values (8, 48) as in the OPS run
# Copy HISTORY.rc.tmpl over from the OPS run
# Copy obsys.rc over from the OPT run
# NX, NY in GSI_GridComp.rc.tmpl should match those from the OPS run (6, 36)
# Set NUM_SGMT to 1 in CAP.rc.tmpl
# <tt>g5das.j</tt>
## PBS -l select=32
## <tt>setenv NCPUS    384</tt>
## <tt>setenv NCPUS_GSI 216</tt>
## '''One cycle''': To run one segment each of GSIsa.x and GEOSgcm.x add '<tt>exit</tt>' before 'PART IV'
-->
 
===Data Assimilation===
 
$ cd $FVHOME/run
$ qsub g5das.j
 
<tt>g5das</tt> creates a scratch directory <tt>$NOBACKUP/fvwork.xxxx</tt> and runs the experiment there. After the initial setup, <tt>GSIsa.x</tt> is run with output written to <tt>ana.log</tt> followed by <tt>GEOSgcm.x</tt> which writes its output to <tt>stdout</tt> and <tt>fvpsas.log</tt>.
 
===Forecast===
 
==Simulating OPS run==
<!--EVERYTHING BELOW IS COMMENTED OUT
 
==Simulating DAS forecast run==
 
The DAS forecast run can be simulated using the corresponding AGCM tag.


Or, one can checkout the corresponding AGCM model Fortuna-2_5_p6 (module
====Checkout and build the model====
Fortuna) and build it.


Note: For GEOSadas-5_7_2_p5_m1 in the example above, g5_modules will need to be edited in order to compile successfully on discover (due to SP1). You should change the discover section to be (where the '''bold''' items are changes, and mod5 is removed):
Grab the corresponding AGCM tag (Check out <tt>GEOSadas-5_9_1_p4</tt> and look for <tt>GCM</tt> in the file <tt>src/Applications/GEOSdas_App/ChangeLog</tt>).


  '''set basedir = /discover/nobackup/projects/gmao/share/gmao_ops/Baselibs/v3.2.0_buildSLES11SP1'''
  $ cvs co -r Ganymed-1_0_p3 Ganymed
  set mod1 = comp/intel-11.0.083
  $ cd GEOSagcm/src
  set mod2 = mpi/impi-3.2.2.006
  $ source g5_modules
  set mod3 = lib/mkl-10.0.3.020
  $ ./parallel_build.csh
  '''set mod4 = other/SIVO-PyD/spd_1.6.0_gcc-4.3.4-sp1'''
 
  '''set mods = ( $mod1 $mod2 $mod3 $mod4''' )
  set modinit = /usr/share/modules/init/csh
  set loadmodules = 0


2. Experiment setup:
====Experiment setup====


To simulate a forecast run we can use gcm_setup to setup our experiment
To simulate a forecast run we can use gcm_setup to setup our experiment
Line 36: Line 113:
   $ cd Applications/GEOSgcm_App/
   $ cd Applications/GEOSgcm_App/
   $ ./gcm_setup
   $ ./gcm_setup
        GOCART: yes
        GOCART Emission Files to use: OPS
        AERO_PROVIDER: GOCART
        HISTORY: Current


Cubed-sphere was not fully implemented in this model. Choose Lat/Lon with
=====Edit files=====
resolution 1152x721x72 for the 'e'-run. The other options are
======AGCM.rc, CAP.rc, HISTORY.rc======
 
  GOCART: yes
  GOCART Emission Files to use: OPS
  AERO_PROVIDER: GOCART
  HISTORY: Current
 
3. AGCM.rc, CAP.rc, HISTORY.rc


Edit these files to match with the corresponding .tmpl files in
Edit these files to match with the corresponding .tmpl files in
Line 54: Line 128:
   $ sed -i -e '/RESTART_TYPE:/ s/binary/pbinary/' AGCM.rc
   $ sed -i -e '/RESTART_TYPE:/ s/binary/pbinary/' AGCM.rc


4. gcm_run.j
======gcm_run.j======


Ensure that the correct boundary datasets are linked to. Compare the links in
Ensure that the correct boundary datasets are linked to. Compare the links in
Line 63: Line 137:
Also, be sure to add an "exit" after you run the executable (if, say, you want to run one segment).
Also, be sure to add an "exit" after you run the executable (if, say, you want to run one segment).


If you are running the test via qsub, you'll also want to change the select line of gcm_run.j to:
=====RC files=====
 
#PBS -l select=90:ncpus=8:proc=west:mpiprocs=8
 
5. RC files


Copy the .rc files over
Copy the .rc files over
Line 73: Line 143:
   $ find ~dao_ops/e572p5_fp/run -name "*.rc" -exec cp {} /path/to/home/dir/RC \;
   $ find ~dao_ops/e572p5_fp/run -name "*.rc" -exec cp {} /path/to/home/dir/RC \;


6. restart files
=====Restart files=====


Locate the dir containing restarts by querying FVARCH in FVDAS_Run_Config
Locate the dir containing restarts by querying FVARCH in FVDAS_Run_Config
Line 98: Line 168:
   $ echo '20120515 210000' > cap_restart
   $ echo '20120515 210000' > cap_restart


7. Run job
====Run job====
    
    
   $ qsub gcm_run.j
   $ qsub gcm_run.j
-->
[[Category:SI Team]]
[[Category:Running the Model]]