How to Fix WRF Run Crash (Error related to module_ra_rrtm.rrtm_lookuptable)

[kinguEnt]

I'm encountering an issue while running WRF. The model crashes shortly after starting, and I've attached the relevant files (namelist.input, rsl.error.0000, and part of the job script) for reference.
=========================================================================
jwe0021i-s line 6925 An endfile record was detected in a READ statement (unit=10).
error occurs at module_ra_rrtm.rrtm_lookuptable_ line 6925 loc 0000000002454708 offset 0000000000000198
module_ra_rrtm.rrtm_lookuptable_ at loc 0000000002454570 called from loc 0000000002454564 in module_ra_rrtm.rrtminit_ line 6806
module_ra_rrtm.rrtminit_ at loc 00000000024544f8 called from loc 00000000026e224c in module_physics_init.ra_init_ line 2131
module_physics_init.ra_init_ at loc 00000000026e0ad4 called from loc 00000000026d99b0 in module_physics_init.phy_init_ line 1337
module_physics_init.phy_init_ at loc 00000000026caef8 called from loc 000000000196a628 in start_domain_em_ line 1089
start_domain_em_ at loc 000000000191d160 called from loc 00000000015d90d8 in start_domain_ line 133
start_domain_ at loc 00000000015d4608 called from loc 0000000000e7652c in med_initialdata_input_ line 251
med_initialdata_input_ at loc 0000000000e750e8 called from loc 0000000000405bf4 in module_wrf_top.wrf_init_ line 294
module_wrf_top.wrf_init_ at loc 00000000004058c4 called from loc 0000000000405890 in MAIN__ line 44
MAIN__ at loc 0000000000405864 called from o.s.
jwe0903i-u Error number 0021 was detected. Maximum error count exceeded.
error summary (Fortran)
error number error level error count
jwe0021i s 1
total error count = 1
=========================================================================

Could anyone suggest how to resolve this issue?

 

[kwerner]

Hi,
I'm not sure if this is causing the problem, but d01 is MUCH larger than d02. If d02 is the area of interest, you can reduce the size of d01 drastically. Typically, due to the resolution differences, d02 has more grid points than d01, and as a rule of thumb, d02 takes up about 1/3 of the space of d01. Can you try to reduce the size of d01 and run this again? This means you'll need to rerun geogrid, metgrid, and real again, prior to running wrf.

If it still fails, please package all of your rsl.* files together into a single *.tar file and attach that, along with the modified namelist.input file (and your namelist.wps file). Thanks!

 

[kinguEnt]

Hi,
I'm not sure if this is causing the problem, but d01 is MUCH larger than d02. If d02 is the area of interest, you can reduce the size of d01 drastically. Typically, due to the resolution differences, d02 has more grid points than d01, and as a rule of thumb, d02 takes up about 1/3 of the space of d01. Can you try to reduce the size of d01 and run this again? This means you'll need to rerun geogrid, metgrid, and real again, prior to running wrf.

If it still fails, please package all of your rsl.* files together into a single *.tar file and attach that, along with the modified namelist.input file (and your namelist.wps file). Thanks!

Hi kwerner,

Thank you for your quick and helpful response.

The reason I initially made the parent domain (d01) a bit large was to make sure it captured key water bodies that are known to influence the nested domain (d02). Specifically, I included the Mediterranean Sea, the Atlantic Ocean—especially the Gulf of Guinea for moisture transport—the Red Sea, the Indian Ocean, and the African Great Lakes region. These areas can have a significant impact on moisture availability within d02.

Do you think reducing the size of the parent domain would still allow d02 to capture the effects from those water bodies?

Thanks again for your support!

 

[kwerner]

Hi,
Yes. Regardless of whether your domain includes the oceans, the model interpolates from the global input that includes the oceans, and the WRF physics are written to consider ocean data when it does its calculations.

 

[kinguEnt]

Thanks so much!
I’ll include the rsl.* files and the namelists, just in case any issues come up after the changes.

 

[kinguEnt]

Hi kwerner,

I modified the domains; however, the same error occurred.
I’ve attached the rsl and other relevant files for your review.

Thank you.

 

[kwerner]

Hi,
Apologies for the delay in response. I've finally been able to run some tests and have determined that I think it's the cu_physics option you're using that is causing the issue. Can you run a test with your same setup, but change cu_physics to = 6, 0 (instead of 3, 0) and see if you still get the seg-fault? For reference, I used 196 processors to run this.