Initializing using RR or NAM?

Submitted by leuthold on Thu, 03/10/2022 - 16:25
Forums
Short-Range Weather/Convection Allowing Application

I have UFS-SRW working when using the GFS for initialization.  I've tried both the RAP and NAM, and they both generate input files for UFS with no errors, and they do run, there are problems with the forecasts.  Both do not initialize or forecast clouds correctly.  I attached both lw and sw images.  I've tried both the RRFS and GFSv15p2 configurations.  I've not found any documentation on initializing using the NAM or RAP, just that it can be done.  Any clues about what I'm doing wrong?  Thanks.

Mike

 

 

Permalink

Submitted by larissa.reames on Thu, 03/10/2022 - 17:41

Hi Mike,

I can't be sure exactly what could be wrong here, if anything, without some more information.

First, it would be most helpful to have high-quality plots of a the same field from all three runs to illustrate which problems you're seeing in the clouds. Perhaps you could plot something like vertical sum of cloud water mixing ratio from both the input files (INPUT/gfs_data.nc) and from the first forecast output time. It would also be helpful to see how your target domain compares to the grids of the NAM and RAP data just to make sure your target grid is well-formulated. Pre-processing should have failed if it wasn't, but I don't want to rule out that something has gone awry in that process. 

Could you also provide the grib2 files you used (either one will do) and the configuration you used to run with it? The app-level config.sh script should be in the top experiment directory (${EXPTDIR}), the preprocessing namelist should be fort.41 in EXPTDIR/YYYYMMDDHH/INPUT/tmp_ICS, and your weather model namelist should be input.nml in EXPTDIR/YYYYMMDDHH/. If you're running on a system with outside access, like Jet or Hera for example, a directory where your forecast is located would work for that. 

Larissa

Permalink

Submitted by larissa.reames on Mon, 03/14/2022 - 17:19

Mike,

Thanks for providing all of those plots and files. They're super helpful.

The input data (your plots from gfs_data.nc) definitely don't look correct, so something is going wrong in the pre=processing step. When you processed RAP and NAM data, what did the fort.41 file for the pre-processing look like? Was change external_model RAP or NAM? I ran a quick test with a previously successful fort.41 file for NAM data on a 13km conus output grid, but set external_model to GFS and I got weird E-W streaks in my data like yours.

Larissa

Permalink

Submitted by leuthold on Tue, 03/15/2022 - 17:39

Yes, it's set to GFS.  I'll change it and I would assume this will fix the problem, right? I'll try it later to be sure.  Is there somewhere else that I need to change "external_model" to either RAP or NAM?  Maybe EXTRN_MDL_NAME_ICS/LBCS=RAP/NAM? Thanks!

Mike

 

Permalink

Submitted by larissa.reames on Tue, 03/15/2022 - 17:54

Yep, that should solve your problem. That's the only place that the input data source matters, so you won't need to change it anywhere else. 

Good luck!

Larissa

Permalink

Submitted by leuthold on Tue, 03/15/2022 - 18:35

Wouldn't it be better to change EXTERN_MDL_NAME_ICS/LBCS in var_defns.sh/config.sh?  Also, consider adding some documentation regarding other initializations besides GFS.

thanks again.

Mike

 

Permalink

Submitted by leuthold on Tue, 03/15/2022 - 19:18

I missed that, sorry.  I was mainly using https://ufs-srweather-app.readthedocs.io/en/latest/SRWAppOverview.html for documentation.

While I have your attention, how can I include the variables geolon and geolat in gfs_data.nc into UFS output?  I apologize if this is a simple question.  I've struggled to try to understand UFS output variable and the diag_table file.

Mike

 

  • 50 views
Attach Files
rr_0.PNG
rr-ir.PNG

Copyright © 2023. All rights reserved.