Results from SRW v2.2 with GFS_v17_p8 are problematic #1004
Comments
|
I check suite_FV3_GFS_v16.xml verus suite_FV3_GFS_v17_p8.xml. In v16, noah, cires_ugwp, and gfdl mp are used. In v17_p8, noahmp, unified_ugwp, and thompson mp are used. Perhaps static files, LBCs, or ICs are not staged properly in my v17_p8 experiments? LBCs/BCs are taken from GFS (gfs.t00z.pgrb2.0p25.f$FH) |
|
Hi UFS/SRW experts, I adopted the sample yaml (config.grid_RRFS_CONUS_25km_ics_FV3GFS_lbcs_FV3GFS_suite_GFS_v17_p8_plot.yaml) from I then modified this config to run different physics suites (rap, rrfs_v1beta and gfs_v16). The results (for 6-hour simulations) shows the runs with rap, rrfs_v1beta, and gfsv16 are similar but gfsv17_p7 is too cold.
|
|
@SarahLu-NOAA Thanks for bringing this to our attention. As you've pointed out, something is incorrect w/ the v17_p8 configuration. I will take a look at this later today and see what I can find. |
|
@SarahLu-NOAA thank you very much for detailing this issue. EPIC's software integration team is currently looking into this and attempting to re-create the issue for debugging/analysis. we will be in touch with updates! |
|
@SarahLu-NOAA @dustinswales @ligiabernardet we were able to reproduce the temp. discrepancy between the v16 and v17_p8 experiment configurations (i.e. the runs associated with these plots are located on hera in the following locations: gfs_v16: |
|
we took a look at @dustinswales @ligiabernardet based on these results, would someone be able to look further into the ccpp configuration for gfs_v17_p8? again, the runs used to make these plots are located on hera: gfs_v16:
|
|
@dustinswales @ligiabernardet fwiw, the differences between the gfs_v16 and gfs_v17_p8 regional (srw) namelists are where the differences between the global p8 namelist and the regional (srw) v17_p8 namelist are listed below ( |
|
Our DTC CCPP colleague Man Zhang is taking a look at this. |
|
I suggest using npz=65 to run SRW apps by changing npz=64 in ufs-srweather-app/parm/input.nml.FV3. Note that changing npz only will have a ptop~0.02hPa If you want to try RRFSL65 (with lower ptop~2hPa) , add the following two lines in input.nml.FV3:
npz=65 change only is sufficient to get reasonable P8 results. |
|
@mzhangw Thanks you for your suggestion. The run at Derecho failed with error msg like : FATAL from PE 8: FV3 top higher than NCEP/GFS The config.yaml is modified from ufs/ufs-srweather-app/tests/WE2E/test_configs/grids_extrn_mdls_suites_community/config.grid_RRFS_CONUS_25km_ics_FV3GFS_lbcs_FV3GFS_suite_GFS_v17_p8_plot.yaml The modifications include 1) not write out restart files, 2) change initial date to 2019-06-15, and 3) add the following line: FV3_NML_YAML_CONFIG_FN: /glade/work/clu/ufs/ufs-srweather-app/parm/FV3.npz.input.yml What other changes are needed for enable L65 run? Thanks. --Sarah |
|
try external_eta=.false., so ak, bk are read from the file
see my run on Hera
/scratch1/BMC/gmtb/Man.Zhang/_SRW2_2_P8/expt_dirs/test_GFS_v17_p8_3/2019061518
…On Wed, Jan 31, 2024 at 6:35 PM ch_sarah ***@***.***> wrote:
@mzhangw <https://github.com/mzhangw> Thanks you for your suggestion.
I conducted an experiment where input.yaml is set up as such:
npz = 65
npz_type = 'input'
fv_eta_file = 'global_hyblev_fcst_rrfsL65.txt'
The run at Derecho failed with error msg like : FATAL from PE 8: FV3 top
higher than NCEP/GFS
The config.yaml is modified from
ufs/ufs-srweather-app/tests/WE2E/test_configs/grids_extrn_mdls_suites_community/config.grid_RRFS_CONUS_25km_ics_FV3GFS_lbcs_FV3GFS_suite_GFS_v17_p8_plot.yaml
The modifications include 1) not write out restart files, 2) change
initial date to 2019-06-15, and 3) add the following line:
FV3_NML_YAML_CONFIG_FN:
/glade/work/clu/ufs/ufs-srweather-app/parm/FV3.npz.input.yml
(where the npz-related changes are specified)
What other changes are needed for enable L65 run? Thanks. --Sarah
—
Reply to this email directly, view it on GitHub
<#1004 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AG7TW2WYWQS6ESRXZU4AVOTYRLWMZAVCNFSM6AAAAABCCNSG7SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMRQGMYTQNBSGI>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
@SarahLu-NOAA Just wanted to check in to see how is your P8 with SRW v2.2 run going. |
|
@mzhangw Derecho is down for 3 days. Will check it once it's back on. Thanks. |
|
@mzhangw The run failed with error msg: FATAL from PE 1: check_nml_error in fms_mod: Unknown namelist, or mistyped namelist variable in namelist fv_core_nml, (IOSTAT = 64 ) I compare your input.nml versus my input.nml and found some differences such as agrid_vel_rst, d2_bg_k2, and delt_max. Since we ran the SRW at different platform (Hera vs Derecho), different job card related parameters are expected. But it seems that there are some difference in how the exp is set up. My config.yaml is modified from sample config: config.grid_RRFS_CONUS_25km_ics_FV3GFS_lbcs_FV3GFS_suite_GFS_v17_p8_plot.yaml from |
|
@SarahLu-NOAA |
|
@mzhangw Thanks. If you review earlier posts, the run with sample v17_p8 leads to unrealistic cold temp within hours. @ulmononian was able to reproduce the temp discrepancy. |
|
@SarahLu-NOAA Note that I use the default npz=64 instead of RRFSL65. It seems that the difference between FV3.p8.input.yml and FV3.input.yml is the key for reasonable P8 runs. |
|
@mzhangw Thanks, but I can't access your directory/yaml files. Could you please grant me to access it. |
|
PR #1055 has just been merged into the develop branch. This PR applies the recommended changes from @mzhangw to the Thank you very much. |
|
@MichaelLueken |
|
Completed with PR #1055. |
Expected behavior
I conducted 4 SRW v2.2 experiments at Derecho. The config is: 12 m over NE US, LBCs/ICs from GFS, DT_ATMOS set to 60 and FCST_LEN_HRS set to 84.
v16 uses out-of-the-box GFS_v16 suite
v17 uses out-of-the-box GFS_v17_p8 suite
v16x uses GFS_v16 with MERRA2 climo in RRTMg (iaer changed from 5111 to 1011)
v17x uses GFS_v17_p8 with dt_inner changed from 150 (the default) to 30 (half of DT_ATMO)
The 4 runs are expected to produce similar results for 24-hour simulations.
Current behavior
The two v17_p8 results are way too cold.
Steps To Reproduce
config.yaml for v17x can be found at
/glade/work/clu/ufs/expt_dirs/neus_12km_gfsv17_gfs4lbc_cldmp
Additional Information
While the SRWv2.2 with GFS_v17_p8 do not abort, the results are unrealistic. I understand that the public release only support GFS_v16. Just wonder whether the UFS/SRW experts can help with this issue. Thanks
The text was updated successfully, but these errors were encountered: