Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Error in wrf.exe when running with the Aerosol-Aware Thompson Scheme (mp_physics=28)

Irma R.S.

New member
Dear Colleges,
I'm trying to run WRF (v4.1.2) with the mp_physics=28 option activated, I have followed the instructions posted here:
WRF Model Users Site

I have included the QNWFA_QNIFA_SIGMA_MONTHLY.dat file in the WPS directory to run the metgrid.exe. Additionally, I have set the following options in the namelist.input:

&domains
wif_input_opt = 1
num_wif_levels = 30

&physics
mp_physics = 28, 28, 28,
use_aero_icbc = .true.
aer_opt = 3


The preprocessing part runs well and I obtain the variables W_WIF, I_WIF and P_WIF in the metgrid output files. However, the wrf.exe fails. No explanation other than "killed by signal 9" appears in the log file, so it looks like it's diverging.
Moreover, when I include this option "cu_rad_feedback = .true.,.false.,.false.", the execution seems to continue a little longer but eventually fails, but this time with a segmentation fault error.
I attach the namelist.input file and the rsl.error.0000 file with a high level of debug. I also attach the rsl.error.0000 when the cu_rad_feedback is activated.

Have you ever experience something similar? Are there any other values in the namelist that need to be changed?
Thanks,
Irma
 

Attachments

  • namelist.input
    9 KB · Views: 13
  • rsl.error.0000
    37.3 KB · Views: 3
  • rsl_cu_rad_feedback.error.0000
    21 KB · Views: 0
I have a few concerns on your namelist.input:
(1) for dx=9km, maximum time_step should be 54
(2) For ERA-I input data, num_metgrid_levels should not be 24. if I remember correctly, this number should be 38.
Please check your met_em file to determine the correct number
(3) why did you set io_form_* = 11?
(4) please turn off lightning_option, which doesn't work with mp_physics=28
 
Thanks for your quick response.

Commenting on what you have mentioned:
(1) This seems not to be the problem, I have tried lowering the time step and it still fails. Sometimes it runs a little bit longer, but it ends up diverging.
(2) We are using GFS input data
(3) io_form_* = 11 to use with parallel-netcdf
(4) I have switched off the lightning option but it keeps failing.

Any other idea?
 
GFS data should have pressure levels of 32. Where ddi you get GFS that has 24 pressure levels?

please set in_form* = 2, then try again.
 
yes, you're right. I am downloading the GFS data with all the pressure levels but I had included de "pmin=5000" parameter in the namelist.wps. So when ncdump command was executed to get the number of levels, it came up with 24.
I have removed this option and run wrf again with all pressure levels included and set io_form*=2, but it does not solve the problem.

One additional comment: I have found in the metgrid log a series of info messages that do not appear when I run wrf without the solar option. Something like this with several variables:
"INFORM: Couldn't find SW010200 at level 200100.000000 to fill level 2.000000 of SW."

Do you know if it can have any influence on the execution of wrf or should I ignore these messages?
 

Attachments

  • metgrid.log
    531.7 KB · Views: 2
No, I don't think the message in metgrid.log is an issue.

I just run a test case using this option and the same version of WRF, it works just fine. Thereby I am really perplexed by the weird model behavior in your case. Except the message ""killed by signal 9", are you able to find other error messages in your rsl files?

I would suggest that you recompile the code with the debug option (i.e., ./configure -D), then rerun this case. The output will tell you exactly where the code failed. That might've you some hints how to fix the problem.
 
No, I don't think the message in metgrid.log is an issue.

I just run a test case using this option and the same version of WRF, it works just fine. Thereby I am really perplexed by the weird model behavior in your case. Except the message ""killed by signal 9", are you able to find other error messages in your rsl files?

I would suggest that you recompile the code with the debug option (i.e., ./configure -D), then rerun this case. The output will tell you exactly where the code failed. That might've you some hints how to fix the problem.
Thank you very much. I have the same problem
 
Since I plan to recompile wrf to its latest version soon, I think I'll try that one.
I'll be back to let you know if I finally managed to run it with the latest version.

Thank you very much for your time.
 
Top