-
Notifications
You must be signed in to change notification settings - Fork 168
Prepare initial conditions
Sections:
There are two types of initial conditions for the global-workflow:
- Warm start: these ICs are taken directly from either the GFS in production or an experiment "warmed" up (at least one cycle in).
- Cold start: any ICs converted to a new resolution or grid (e.g. GSM-GFS -> FV3GFS). These ICs are often prepared by chgres_cube (change resolution utility).
Most users will initiate their experiments with cold start ICs unless running high resolution (C768 deterministic with C384 EnKF) for a date with warm starts available. It is not recommended to run high resolution unless required or as part of final testing.
Resolutions:
- C48 = 2 degree ≈ 200km
- C96 = 1 degree ≈ 100km
- C192 = 1/2 degree ≈ 50km
- C384 = 1/4 degree ≈ 25km
- C768 = 1/8th degree ≈ 13km
- C1152 ≈ 9km
- C3072 ≈ 3km
Supported resolutions in global-workflow: C48, C96, C192, C384, C768
Not yet supported. See Manual Generation section below for how to create your ICs yourself (outside of workflow).
Free-forecast mode in global-workflow includes getic and init jobs for the gfs suite. The getic job pulls inputs for chgres_cube (init job) or warm start ICs into your ROTDIR/COMROT. The init job then ingests those files to produce initial conditions for your experiment.
Users on machines without HPSS access (e.g. Orion) need to perform the getic step manually and stage inputs for the init job. The table below lists the needed files for init and where to place them in your ROTDIR.
Note for table: yyyy=year; mm=month; dd=day; hh=cycle
Operations/production output location on HPSS: /NCEPPROD/hpssprod/runhistory/rhyyyy/yyyymm/yyyymmdd/
| Source | Files | Tarball name | Where in ROTDIR |
|---|---|---|---|
| v12 ops | gfs.thhz.sanlgfs.t hhz.sfcanl |
com_gfs_prod_gfs.yyyymmddhh.anl.tar |
gfs.yyyymmdd/hh
|
| v13 ops | gfs.thhz.sanlgfs.t hhz.sfcanl |
com2_gfs_prod_gfs.yyyymmddhh.anl.tar |
gfs.yyyymmdd/hh
|
| v14 ops | gfs.thhz.atmanl.nemsiogfs.t hhz.sfcanl.nemsio |
gpfs_hps_nco_ops_com_gfs_prod_gfs.yyyymmddhh.anl.tar |
gfs.yyyymmdd/hh
|
| v15 ops pre-2020022600 | gfs.thhz.atmanl.nemsiogfs.t hhz.sfcanl.nemsio |
gpfs_dell1_nco_ops_com_gfs_prod_gfs.yyyymmdd_hh.gfs_nemsioa.tar |
gfs.yyyymmdd/hh
|
| v15 ops | gfs.thhz.atmanl.nemsiogfs.t hhz.sfcanl.nemsio |
com_gfs_prod_gfs.yyyymmdd_hh.gfs_nemsioa.tar |
gfs.yyyymmdd/hh
|
| v16 ops | gfs.thhz.atmanl.ncgfs.t hhz.sfcanl.nc |
com_gfs_prod_gfs.yyyymmdd_hh.gfs_nca.tar |
gfs.yyyymmdd/hh/atmos |
| v16 retro | gfs.thhz.atmanl.ncgfs.t hhz.sfcanl.nc |
gfs_netcdfa.tar* | gfs.yyyymmdd/hh/atmos |
*For HPSS path, see retrospective table in pre-production parallel section below
The following information is for users needing to generate initial conditions for a cycled experiment that will run at a different resolution or layer amount than the operational GFS (C768C384L127).
The new chgres_cube code is available from the UFS_UTILS repository on GitHub (maintained by George Gayno) and can be used to convert GFS ICs to a different resolution or number of layers. Users may clone the develop/HEAD branch or the same version used by global-workflow develop (found in sorc/checkout.sh). The chgres_cube code/scripts currently support the following GFS inputs:
- pre-GFSv14 (GFS-GSM)
- GFSv14 (GFS-GSM)
- GFSv15 (FV3GFS)
git clone --recursive https://github.com/NOAA-EMC/UFS_UTILS.git
Then switch to a different tag or use the default branch (develop).
sh build_all.sh
cd fix
sh link_fixdirs.sh emc $MACHINE
...where $MACHINE is "cray", "dell", "hera", "jet", or "orion".
cd util/gdas_init
vi config
Read the doc block at the top of the config and adjust the variables to meet you needs (e.g. yy, mm, dd, hh for SDATE).
./driver.$MACHINE.sh
...where $MACHINE is currently "dell" or "cray" or "hera" or "jet" or "orion". Additional options will be available as support for other machines expands.
90 small jobs will be submitted:
- 9 jobs to pull inputs off HPSS (1 for deterministic and 8 for the EnKF ensemble members)
- 81 jobs to run chgres (1 for deterministic/hires and 80 for each EnKF ensemble member)
The chgres jobs will have a dependency on the data-pull jobs and will wait to run until all data-pull jobs have completed.
In the config you will have defined an output folder called $OUTDIR. The converted/chgres'd output will be found there, including the needed abias and radstat initial condition files. The files will be in the needed directory structure for the global-workflow system, therefore a user can move the contents of their $OUTDIR directly into their $ROTDIR/$COMROT.
Please report bugs to George Gayno ([email protected]) and Kate Friedman ([email protected]).
The GFSv15 was implemented into production on June 12th, 2019 at 12z. The GFS was spun up ahead of that cycle and thus production output for the system is available from the 00z cycle (2019061200) and later. Production output tarballs from the prior GFSv14 system are located in the same location on HPSS but have "hps" in the name to represent that it was run on the Cray, where as the GFS now runs in production on the Dell and has "dell1" in the tarball name.
See production output in the following location on HPSS:
/NCEPPROD/hpssprod/runhistory/rhYYYY/YYYYMM/YYYYMMDD
Example location:
/NCEPPROD/hpssprod/runhistory/rh2021/202104/20210420
Example listing for 2021042000 production tarballs:
[Kate.Friedman@m72a2 ~]$ hpsstar dir /NCEPPROD/hpssprod/runhistory/rh2021/202104/20210420 | grep gfs | grep _00. | grep -v idx
[connecting to hpsscore1.fairmont.rdhpcs.noaa.gov/1217]
******************************************************************
* Welcome to the NESCC High Performance Storage System *
* *
* Current HPSS version: 7.5.3 *
* *
* *
* Please Submit Helpdesk Request to *
* [email protected] *
* *
* Announcements: *
******************************************************************
Username: Kate.Friedman UID: 2391 Acct: 2391(2391) Copies: 1 COS: 0 Firewall: off [hsi.6.3.0.p1-hgs Thu May 7 09:16:23 UTC 2020]
/NCEPPROD/hpssprod/runhistory/rh2021/202104:
drwxr-xr-x 2 nwprod prod 11776 Apr 19 23:44 20210420
[connecting to hpsscore1.fairmont.rdhpcs.noaa.gov/1217]
-rw-r----- 1 nwprod rstprod 51268255744 Apr 22 05:29 com_gfs_prod_enkfgdas.20210420_00.enkfgdas.tar
-rw-r--r-- 1 nwprod prod 220121310720 Apr 22 06:42 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp1.tar
-rw-r--r-- 1 nwprod prod 220124178944 Apr 22 07:04 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp2.tar
-rw-r--r-- 1 nwprod prod 220120305664 Apr 22 07:24 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp3.tar
-rw-r--r-- 1 nwprod prod 220116934656 Apr 22 07:38 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp4.tar
-rw-r--r-- 1 nwprod prod 220121547776 Apr 22 07:56 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp5.tar
-rw-r--r-- 1 nwprod prod 220125794816 Apr 22 08:09 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp6.tar
-rw-r--r-- 1 nwprod prod 220117037568 Apr 22 08:23 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp7.tar
-rw-r--r-- 1 nwprod prod 220117203968 Apr 22 08:33 com_gfs_prod_enkfgdas.20210420_00.enkfgdas_restart_grp8.tar
-rw-r----- 1 nwprod rstprod 9573153280 Apr 22 02:49 com_gfs_prod_gdas.20210420_00.gdas.tar
-rw-r--r-- 1 nwprod prod 1020249088 Apr 22 02:49 com_gfs_prod_gdas.20210420_00.gdas_flux.tar
-rw-r--r-- 1 nwprod prod 92950728704 Apr 22 03:05 com_gfs_prod_gdas.20210420_00.gdas_nc.tar
-rw-r--r-- 1 nwprod prod 10647806464 Apr 22 02:50 com_gfs_prod_gdas.20210420_00.gdas_pgrb2.tar
-rw-r----- 1 nwprod rstprod 65121796608 Apr 22 02:56 com_gfs_prod_gdas.20210420_00.gdas_restart.tar
-rw-r--r-- 1 nwprod prod 18200814080 Apr 22 03:06 com_gfs_prod_gdas.20210420_00.gdaswave_keep.tar
-rw-r----- 1 nwprod rstprod 13013076992 Apr 22 03:08 com_gfs_prod_gfs.20210420_00.gfs.tar
-rw-r--r-- 1 nwprod prod 62663230976 Apr 22 03:13 com_gfs_prod_gfs.20210420_00.gfs_flux.tar
-rw-r--r-- 1 nwprod prod 127932879360 Apr 22 03:47 com_gfs_prod_gfs.20210420_00.gfs_nca.tar
-rw-r--r-- 1 nwprod prod 138633526272 Apr 22 04:00 com_gfs_prod_gfs.20210420_00.gfs_ncb.tar
-rw-r--r-- 1 nwprod prod 140773240832 Apr 22 03:27 com_gfs_prod_gfs.20210420_00.gfs_pgrb2.tar
-rw-r--r-- 1 nwprod prod 61253672960 Apr 22 03:32 com_gfs_prod_gfs.20210420_00.gfs_pgrb2b.tar
-rw-r--r-- 1 nwprod prod 19702107136 Apr 22 03:34 com_gfs_prod_gfs.20210420_00.gfs_restart.tar
-rw-r--r-- 1 nwprod prod 18617610240 Apr 22 04:02 com_gfs_prod_gfs.20210420_00.gfswave_output.tar
-rw-r--r-- 1 nwprod prod 30737774592 Apr 22 04:05 com_gfs_prod_gfs.20210420_00.gfswave_raw.tar
The warm starts and other output from production are at C768 deterministic and C384 EnKF. The warm start files must be converted to your desired resolution(s) using chgres_cube if you wish to run a different resolution. If you are running a C768/C384 experiment you can use them as is.
That depends on what mode you want to run...free-forecast or cycled. Whichever mode navigate to the top of your COMROT and pull the entirety of the tarball(s) listed below for your mode. The files within the tarball are already in the $CDUMP.$PDY/$CYC folder format expected by the system.
Two tarballs to pull:
File #1 (for starting cycle SDATE):
/NCEPPROD/hpssprod/runhistory/rhYYYY/YYYYMM/YYYYMMDD/com_gfs_prod_gfs.YYYYMMDD_CC.gfs_restart.tar
File #2 (for prior cycle GDATE=SDATE-06):
/NCEPPROD/hpssprod/runhistory/rhYYYY/YYYYMM/YYYYMMDD/com_gfs_prod_gdas.YYYYMMDD_CC.gdas_restart.tar
There are 18 tarballs to pull (9 for SDATE and 9 for GDATE (SDATE-06)):
HPSS path: /NCEPPROD/hpssprod/runhistory/rhYYYY/YYYYMM/YYYYMMDD/
Tarballs per cycle:
com_gfs_prod_gdas.YYYYMMDD_CC.gdas_restart.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp1.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp2.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp3.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp4.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp5.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp6.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp7.tar
com_gfs_prod_enkfgdas.YYYYMMDD_CC.enkfgdas_restart_grp8.tar
Go to the top of your COMROT/ROTDIR and pull the contents of all tarballs there. The tarballs already contain the needed directory structure.
Recent pre-implementation parallel series was for GFS v16 (implemented March 2021). For the prior v15 (Q2FY19) see an additional table below.
- What resolution are warm-starts available for? Warm-start ICs are saved at the resolution the model was run at (C768/C384) and can only be used to run at the same resolution combination. If you need to run a different resolution you will need to make your own cold-start ICs. See cold start section above.
- What dates have warm-start files saved? Unfortunately the frequency changed enough during the runs that it’s not easy to provide a definitive list easily.
- What files? All warm-starts are saved in separate tarballs which include “restart” in the name. You need to pull the entirety of each tarball, all files included in the restart tarballs are needed.
- Where are these tarballs? See below for the location on HPSS for each v16 pre-implementation parallel.
-
What tarballs do I need to grab for my experiment? Tarballs from two cycles are required. The tarballs are listed below, where $CDATE is your starting cycle and $GDATE is one cycle prior.
- Free-forecast:
- ../$CDATE/gfs_restarta.tar
- ../$GDATE/gdas_restartb.tar
- Cycled w/EnKF:
- ../$CDATE/gdas_restarta.tar
- ../$CDATE/enkfgdas_restarta_grp##.tar (where ## is 01 through 08) (note, older tarballs may include a period between enkf and gdas: "enkf.gdas")
- ../$GDATE/gdas_restartb.tar
- ../$GDATE/enkfgdas_restartb_grp##.tar (where ## is 01 through 08) (note, older tarballs may include a period between enkf and gdas: "enkf.gdas")
- Free-forecast:
- Where do I put the warm-start initial conditions? Extraction should occur right inside your COMROT. You may need to rename the enkf folder (enkf.gdas.$PDY -> enkfgdas.$PDY).
| Time Period | Parallel Name | Archive Location on HPSS (/NCEPDEV/emc-global/5year/emc.glopara/...) |
|---|---|---|
| 2019050106 ~ 2019060100 | v16retro0e | .../WCOSS_D/gfsv16/v16retro0e/yyyymmddhh
|
| 2019060106 ~ 2019083118 | v16retro1e | .../WCOSS_D/gfsv16/v16retro1e/yyyymmddhh
|
| 2019090100 ~ 2019110918 | v16retro2e | .../WCOSS_D/gfsv16/v16retro2e/yyyymmddhh
|
| 2019111000 ~ 2020122200 | v16rt2 | .../WCOSS_D/gfsv16/v16rt2/yyyymmddhh
|
| 2020122206 ~ implementation | v16rt2n | .../WCOSS_D/gfsv16/v16rt2n/yyyymmddhh
|
| Time Period | Parallel Name | Archive Location on HPSS (/NCEPDEV/emc-global/5year/...) |
|---|---|---|
| Real-time (05/25/2018 ~ 06/12/2019) |
prfv3rt1 | .../emc.glopara/WCOSS_C/Q2FY19/prfv3rt1 |
| 2017/2018 Winter/Spring (11/25/2017 ~ 05/31/2018) |
fv3q2fy19retro1 | .../Fanglin.Yang/WCOSS_DELL_P3/Q2FY19/fv3q2fy19retro1 |
| 2017 Summer/Fall Part 1 (05/25//2017 ~ 08/31/2017) |
fv3q2fy19retro2 | .../emc.glopara/WCOSS_C/Q2FY19/fv3q2fy19retro2 |
| 2017 Summer/Fall Part 2 (08/02//2017 ~ 11/30/2017) |
fv3q2fy19retro2 | .../Fanglin.Yang/WCOSS_DELL_P3/Q2FY19/fv3q2fy19retro2 |
| 2016/2017 Winter/Spring (11/25/2016 ~ 05/31/2017) |
fv3q2fy19retro3 | .../Fanglin.Yang/WCOSS_DELL_P3/Q2FY19/fv3q2fy19retro3 |
| 2016 Summer/Fall Part 1 (5/22/2016 ~ 08/25/2016) |
fv3q2fy19retro4 | .../emc.glopara/WCOSS_C/Q2FY19/fv3q2fy19retro4 |
| 2016 Summer/Fall Part 2 (08/17//2016 ~ 11/30/2016) |
fv3q2fy19retro4 | .../emc.glopara/WCOSS_DELL_P3/Q2FY19/fv3q2fy19retro4 |
| 2015/2016 Winter/Spring (11/25/2015 ~ 05/31/2016) |
fv3q2fy19retro5 | .../emc.glopara/WCOSS_DELL_P3/Q2FY19/fv3q2fy19retro5 |
| 2015 Summer/Fall (5/03/2015 ~ 11/30/2015) |
fv3q2fy19retro6 | .../emc.glopara/WCOSS_DELL_P3/Q2FY19/fv3q2fy19retro6 |