Running HWRF with 2020 GDAS

Submitted by sim.aberson on Wed, 01/20/2021 - 11:29
Forum: Users | General
We're trying to run HWRF using the 2020 GDAS ensemble as background.
Each year, something changes with the GDAS or GFS making this difficult.

The output files are on jet at
/mnt/lfs4/HFIP/hfip-hda/Sim.Aberson/EXPERIMENTS_EREL_1KM/TEDDY20L.202009171200/MEMBER*.


In this case, WPS, PREP_HYBRID, and REAL all seem to be working fine,
and WRF blows up after just a few time steps.  I looked at the various
files, and al look fine except the wrfbdy files look very odd.  But,
REAL does not throw an error, and there isn't anything noteworthy in the
rsl files for real (in ATMOS/all_rsl_files).  So, I can't find a smoking
gun, and it's time to turn to you all for help.

Let me know if you have any questions.

Sim
Permalink

Submitted by sim.aberson on Thu, 01/21/2021 - 11:58

Hi Evan,

Jet is down now, so I can't check.  My recollection is that it was the last in the V3 series.  If necessary, we can upgrade for the 2020 runs, but we'd prefer to have a stable version to compare with other cases.

 

Thanks,

Sim

Permalink

Submitted by evan.kalina on Wed, 01/27/2021 - 14:46

Thanks Sim. Agreed that the wrfbdy files look odd – for example, some of the temperature values on the boundaries (T_BXS, T_BXE) are set to absolute zero. In addition, the O3RAD field, which contains the ozone concentration and is needed for radiative transfer calculations, is also set to zero in your wrfinput file.

So, a few things:

1) Can you point me to your input data files (i.e., GDAS ensemble) on Jet for this case?

2) In your WPS directory, there are a couple of broken links to the Vtable and METGRID.TBL files that are used in the ungrib and metgrid steps, respectively. Can you point me to where your copies of those files are stored on Jet? I would like to compare them to the ones that are currently in use.

3) Do you have a successful case (even if using an older version of the GDAS ensemble) that I can compare against?

Permalink

Submitted by sim.aberson on Thu, 01/28/2021 - 08:59

Hi Evan,

 

All the script output can be found in 

/mnt/lfs4/HFIP/hfip-hda/Sim.Aberson/EXPERIMENTS_EREL_1KM/TEDDY20L.202009171200/SCRIPTS/log*

I have set -x, so you can see exactly what all the scripts are doing, just in case you need more information.

 

The GFS ensemble files are at /mnt/lfs1/HFIP/hfip-hda/GFS_GRIB2NEMSIO_RETRO//experiments/2020091712/global-analysis/2020091706/ensemble/atmos/

 

The log files have for METGRID.TBL (for member9)

cp -f -r /mnt/lfs4/HFIP/hfip-hda/HEDAS_CODE/HWRF_2017/WPSV4/METGRID.TBL.NMM /mnt/lfs4/HFIP/hfip-hda/Sim.Aberson//EXPERIMENTS_EREL_1KM/TEDDY20L.202009171200/MEMBER9/WPS/metgrid/

ln -sf ./metgrid/METGRID.TBL.NMM ./METGRID.TBL

ln -sf ./METGRID.TBL.NMM ./metgrid/METGRID.TBL

which the output suggests works.

 

For Vtable, they have

cp -f -r /mnt/lfs4/HFIP/hfip-hda/HEDAS_CODE/HWRF_2017/WPSV4/Vtable.GFS2017 /mnt/lfs4/HFIP/hfip-hda/Sim.Aberson//EXPERIMENTS_EREL_1KM/TEDDY20L.202009171200/MEMBER9/WPS/ungrib/Variable_Tables/

ln -sf ./ungrib/Variable_Tables/Vtable.GFS2017 ./Vtable

ln -sf ./Vtable.GFS2017 ./ungrib/Variable_Tables/Vtable

which also works.  Do I need an upgraded Vtable, though?

 

I am currently running Hurricane Dorian (2019) which works fine.  You can find the runs that are going now at

/mnt/lfs4/HFIP/hfip-hda/Sim.Aberson/EXPERIMENTS_EREL_1KM/DORIAN05L.201908270600 and DORIAN05L.201908271200

 

Let me know if there is any other information you need.

 

Thanks,

Sim

Permalink

Submitted by evan.kalina on Mon, 02/01/2021 - 17:01

Sim, thanks for answering all of my questions. You can ignore what I said about the O3RAD field and the wrfbdy file. Although those fields look different to me than they do in the operational system, they are consistent with the fields in your successful run of Dorian, so I doubt they are the problem.

Instead, I think the issue is that there are some bad soil temperature data in the grib2 files, which WPS extracts and puts in the met_nmm.d01* files. If you look closely at the soil temperature fields in the met_nmm.d01* files (ST* 2D fields), you'll see that there are some extremely cold values over the island of Dominica (below 200 K). Other areas look reasonable.

I ran a test with your grb2 files, and I was able to fix this problem in METGRID.TBL by adding a line to each of the soil temperature variable stanzas setting the missing value to zero (missing_value=0.). You can see the modified METGRID.TBL here: /mnt/lfs4/HFIP/dtc-hurr/Evan.Kalina/WPS/METGRID.TBL. After making this change and rerunning metgrid.exe, the soil temperatures over Dominica are much more reasonable (~298 K). I did not test whether this resolves the issue with WRF, but I like its chances.

I cannot offer any explanation for why this change is needed for this particular case, when you have apparently ran many other cases without incident.

Please let us know if this resolves the issue.

Permalink

Submitted by sim.aberson on Mon, 02/08/2021 - 14:55

Thanks for the info.  I will test this out when the current 2019 jobs I am running are complete.

 

Will this change cause problems with other cases, like the 2019 cases, or can I substitute this new one for the old one I have been using?

 

Sim

Permalink

Submitted by evan.kalina on Sat, 02/13/2021 - 14:00

I do not think that this change should cause problems for other cases. If there aren't any bogus values in the soil temperature fields, none of the values will be flagged as missing and everything should proceed as before.

Sitemap

Contact

303.497.8921
dtcenter@ucar.edu

Privacy
Cookies
Accessibility
Terms of Use
Copyright Issues

Addresses

mailing:
PO Box 3000
Boulder, CO 80307-3000

Shipping:
3090 Center Green Dr.
Boulder, CO 80301

Follow Us

YouTube Icon
linkedin profile

Twitter Icon

Copyright © 2024. All rights reserved.