Running NDOWN for a lower domain (WRF3.9.1.1)

[jamesrup]

Hi WRF Help,

I'm trying to figure out a way to trick ndown into generating the boundary conditions from my wrfout_d02* files to use as forcing for a new simulation corresponding to that run's third domain, d03.

I successfully ran ndown with my code on this model output going from d01 --> d02, following the instructions in the WRF User's Guide, but it's crashing without any clear error messages* when I try to do it for the next level down. (*e.g.: forrtl: severe (174): SIGSEGV, segmentation fault occurred).

The method I've tried is copying my wrfout_d02* files and renaming them to wrfout_d01*, and doing similar for the initial files: wrfinput_d02 --> wrfinput_d01, and wrfinput_d03 --> wrfndi_d02. And similarly, I've adapted the appropriate columns of my namelist.input to be consistent with this change.

The initial and modified namelist files are attached, along with one example rsl.error file from the crash that shows some MPI errors.

Thank you for the help!
James

 

[jamesrup]

Small update - I managed by now to scroll down slightly in the User's Guide to see there that the procedure for generating boundary conditions for the third domain is described there. ;-)

Nevertheless, the procedure described is nearly identical to what I've tried anyway, which is failing. The only difference I can see from the instructions is that I don't actually run a simulation after ndown for my d02 renamed to d01 - I'm simply trying to use the wrfout_d02* files I already have from the first simulation with three domains, and have just renamed those files from wrfout_d02* --> wrfout_d01*. Would that cause an issue? I would greatly prefer to avoid having to rerun the d02 simulation.

Thank you,
James

 

[jamesrup]

A final update - I got it to work for d02 --> d03. I did simply follow the procedure described in the User's Guide. It seems I had an issue with my WRF build. When I recompiled and changed some compiler options it worked. But it seems to be sensitive to even just to the parallelization when compiling. When I compile with ./compile -j 2 em_real ndown works to completion, but when I instead compile with ./compile -j 6 em_real it hangs up early in the process and never completes. For my purposes for the time being, problem solved!

James

 

[kwerner]

Hi James,
That is odd regarding the difference in whether the code was built with 2 vs. 6 compilers; however, I am very glad to hear that you were able to get past the problem. Thank you for letting us know!