WRF observation nudging segmentation fault

[A WRF User]

I need some help figuring out a segmentation fault. I am running WRF (surprise) with 2 domains and observation nudging turned on. Domain 1 has 1-km grid spacing and runs for 6 hours before domain 2 (111-m grid spacing) is initialized. The simulation runs for 6 h, but after domain 2 is initialized the simulation segmentation faults. The namelist, rsl files and output are:/glade/scratch/ajensen/ncaruas/2018071904_fdda. The same simulation but with observation nudging turned off (which runs successfully) is: /glade/scratch/ajensen/ncaruas/2018071904_no_fdda. Thanks for the help. -Anders

 

[Ming Chen]

Please type ./clean -a and rebuild WRF in debug mode (i.e., ./configure -D), then rerun this case to find in which code/line the model crashed. This will give you some hints what is wrong.

Note that doing objective analysis on a nest may lead to inconsistency between the coarse domain and the nest. This may lead to failure in the model running.

 

[A WRF User]

I debugged WRF and found the issue in /frame/module_dm.F. There is a subroutine called get_full_obs_vector with two lines where the max number of processors is hard coded, and I was using more processors:

INTEGER IDISPLACEMENT(1024) ! HARD CODED MAX NUMBER OF PROCESSORS
INTEGER ICOUNT(1024) ! HARD CODED MAX NUMBER OF PROCESSORS

Thanks, Anders.

 

[Ming Chen]

Thank you for the kind message. We have modified the code based on your information. Please use the attached code (module_dm.F) to replace the original code (WRF/external/RSL_LITE/module_dm.F), then rebuild WRF and run your failed case. Please let me know whether this new code works.
Thanks!

Ming Chen

 

[A WRF User]

Ming, the new code works!

Thanks much.