FATAL ERROR real.exe : Could not find level above ground

[delmoral]

Hi all,

I was running a simulation with WRF 4.1.5 using the aerosol-aware Thompson scheme but realized that something was wrong with real because I was not getting the QNWFA2D variable in the wrf boundary files, and then it was always 0 in the output.

Before testing again real.exe, I changed two constants regarding that variable in the dyn_em/module_initialize_real.F and in the phys/module_mp_thompson.F, that was automatically fixed in WRF version 4.3: https://github.com/wrf-model/WRF/pull/1507
After changing those constants, I recompile my wrf and all the log files indicate that it was done successfully. Attached are my modules files, originally from version 4.1.5, and where I changed just a constant in a line of the code (repeated in both files).

However, now I am unable to succeed in running real.exe. This is the message that I am getting at the very beginning, with the first met_em file:

taskid: 0 hostname: r036c01s01
module_io_quilt_old.F 2931 T
Ntasks in X 12 , ntasks in Y 15
REAL_EM V4.1.5 PREPROCESSOR
*************************************
Parent domain
ids,ide,jds,jde 1 443 1 313
ims,ime,jms,jme -4 44 -4 28
ips,ipe,jps,jpe 1 37 1 21
*************************************
DYNAMICS OPTION: Eulerian Mass Coordinate
alloc_space_field: domain 1 , 106637928 bytes allocated
d01 2020-01-01_00:00:00 Yes, this special data is acceptable to use: OUTPUT FROM METGRID V4.1
d01 2020-01-01_00:00:00 Input data is acceptable to use: met_em.d01.2020-01-01_00:00:00.nc
metgrid input_wrf.F first_date_input = 2020-01-01_00:00:00
metgrid input_wrf.F first_date_nml = 2020-01-01_00:00:00
d01 2020-01-01_00:00:00 Timing for input 4 s.
d01 2020-01-01_00:00:00 flag_soil_layers read from met_em file is 1
Max map factor in domain 1 = 1.06. Scale the dt in the model accordingly.
d01 2020-01-01_00:00:00 Using Thompson_aerosol_aware so using QNWFA monthly climo arrays to create QNWFA_now
Using sfcprs to compute psfc
-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE: <stdin> LINE: 6441
-------------------------------------------
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0

Following some forum posts, I have checked my met_em files, the PSFC data, and I even download again ERA5 data, ungrib (separately SL and PL) and then metgrid again to compare and give it a try, but it does not work.

Although having the missing QNWFA2D variable in my wrf boundary files, I was able to run real with no problem, and it never gave me an error. Now it seems something is messed up after the new compilation.

I attach here my namelist.input, and the modified Fortran modules (I have not succeeded in uploading a met_em zip file).

I would appreciate any help.

Thank you so much in advance,

Anna

 

[Ming Chen]

Anna,
I don't think anything is wrong in your modified code (module_initialize_real.F). This is also the code that may affect real.exe. The other you modified, module_mp_thmpson.F, only affects wrf.exe and it has no impact on real.exe.
Based on the above thinking, I am suspicious that this is an issue related to your machine environment settings. To do a quick test, you may rerun the codes that have no modification and see how it works.
Anyway, about the aerosol-aware Thompson scheme, as I told you in my previous post, we are working to fix the problems in this scheme. At present, please try to avoid using it.

 

[delmoral]

Ming Chen,

You are right, there was a problem with the machine.
Now, this is solved and I finally was able to run real.exe successfully. The confusion came through the error message that the log files displayed, which indicated that the met_em files were probably wrong or corrupted, but they are correct.

Thank you so much.

Anna

 

[Ming Chen]

Anna,
Thanks for the update. I am glad to know this is a machine issue instead of a code issue.