Creating a subgrid

[mkbenneh]

I made a predefined subgrid over Florida using the settings below; I have also included it in the valid_param_vals.yaml file, however it still doesn't work for me and will like to understand how making a subgrid work.
"
GRID_GEN_METHOD: ESGgrid
ESGgrid_LON_CTR: -80.5.
ESGgrid_LAT_CTR: 26
ESGgrid_DELX: 3000.0
ESGgrid_DELY: 3000.0
ESGgrid_NX: 1820
ESGgrid_NY: 1092
ESGgrid_PAZI: 0.0
ESGgrid_WIDE_HALO_WIDTH: 6
DT_ATMOS: 36
LAYOUT_X: 28
LAYOUT_Y: 28
BLOCKSIZE: 28
QUILTING:
WRTCMP_write_groups: 1
WRTCMP_write_tasks_per_group: 4
WRTCMP_output_grid: lambert_conformal
WRTCMP_cen_lon: '{{ task_make_grid.ESGgrid_LON_CTR }}'
WRTCMP_cen_lat: '{{ task_make_grid.ESGgrid_LAT_CTR }}'
WRTCMP_lon_lwr_left: -81.5
WRTCMP_lat_lwr_left: 25.5
WRTCMP_stdlat1: '{{ task_make_grid.ESGgrid_LAT_CTR }}'
WRTCMP_stdlat2: '{{ task_make_grid.ESGgrid_LAT_CTR }}'
WRTCMP_nx: 198
WRTCMP_ny: 138
WRTCMP_dx: 3000.0
WRTCMP_dy: 3000.0
"

[MichaelLueken]

Hello @mkbenneh,

Thank you for reaching out with your question regarding how to create a subgrid in the SRW App! In order to assist with debugging your issue, there is some additional information that is required:

1. What version of the SRW App are you using (i.e., release/public-v2.2.0 or the latest develop)?
2. What is the error message that you are receiving and is the experiment workflow properly being generated?
3. If your experiment is failing, please provide the log files for the failed tasks.
4. If you are running the SRW App on a Tier-1 platform (i.e., Hera, Jet, etc.), please provide a path so that the changes can be checked.

Thanks again and I look forward to assisting you with this issue.

Mike Lueken

[MichaelLueken]

Hello, @mkbenneh,

EPIC only maintains data that is used for the workflow-end-to-end (WE2E) tests on Derecho. I am unable to provide extended datasets for verification outside of the WE2E test staged data. Unfortunately, you will need to stage any data that isn't currently available for your tests.

Mike Lueken

[mkbenneh]

Hello @MichaelLueken,

Got it. Thank you. I also noticed an issue with the SLP plot and wanted to find out what could be the possible issue here.

Here is a path to my results and working directory - "/glade/work/mbenneh/expt_dirs/term_project/2020062200/postprd"

Regards,
Michael

[MichaelLueken]

Hello @mkbenneh,

Unfortunately, I'm not sure what it happening in your SLP plots. I've built a copy of the release/public-v2.2.0 branch on Derecho and was able to replicate what you are encountering. Changing CCPP physics schemes didn't change the behavior encountered, not did altering your SEUS_12km grid values.

I was wondering, is the grid shown in the plot the grid that you were expecting? I saw in ush/predef_grid_params.yaml that the grid was supposed to include the SE US, Atlantic, and Sahara Desert. The above grid doesn't include all of that domain.

I'll reach out to my colleagues and see if they might know what is causing this.

Mike

[mkbenneh]

Hello @MichaelLueken,

I am currently working on just the Southeastern US and will later extend my domain over the Atlantic and Sahara.

I will be looking forward to your response.

Kind regards,
Michael

[mkbenneh]

Hello @MichaelLueken,

I have been able to resolve the issue from the image I sent earlier, attached here. However, I have a couple of questions regarding creating an ESGgrid, lambert comformal. I had to write a little script to do some math to calculate the WRTCMP_lon_lwr_left and WRTCMP_lat_lwr_left for the resolution I needed; otherwise, the forecast domain would be wrong. Which was pretty interesting. I would like to know the limit of WRTCMP_nx and WRTCMP_ny with respect to dy and dx.

Kind regards,
Michael

[MichaelLueken]

Hello @mkbenneh,

I'm glad that you were able to figure out the issue with the plotting!

There shouldn't be any limits imposed on WRTCMP_nx and WRTCMP_ny with respect to dx and dy.

Due to differences in the map projection type (the native that is being used for computational purposes is the extended schmidt gnomonic grid, while the write-component grid is lambert conformal), the grids don't correspond 1-to-1. So, you should ensure that the compute grid (ESGgrid_NX and ESGgrid_NY) is a few gridpoints larger than your write-component grid (WRTCMP_nx and WRTCMP_ny), otherwise you will likely see undefined data in your model output.

For more details on suggestions for setting up a custom regional ESG grid, please see this practical guide provided by DTC.

[mkbenneh]

Hello @MichaelLueken,

Alright, Thank you. I have further questions on the use of UPP and setting the wall time for longer simulations. Should I create a separate thread, or can I ask under here?

[MichaelLueken]

Hello @mkbenneh,

I would recommend creating a separate thread for questions related to UPP and setting the wall rime for longer simulations.

[mkbenneh]

@MichaelLueken

Alright. In the config file in this directory, "/work2/noaa/uswrp-hu/mbenneh/UFS-SRW/expt_dirs/dust", I am unable to go beyond 8 hours. If I set the walltime to anything beyond 8 hours, it stops and says something about resource limits. I am not sure if there are ways to go around this problem.

[MichaelLueken]

@mkbenneh

For Orion, the maximum limit for wall clock is 8 hours. This is a hard limit imposed by MSU and there isn't anyway to get around the issue.