cesm2.1.2 in derecho

[m_valmartin@sheffield_ac_uk]

Hello, I'm trying to get cesm2.1.2 running in derecho. I understand that the latest version supported is cesm2.1.5 but my cesm2.1.2 version has important implementations for ongoing research projects and publications! My version builds, compiles, and works fine with a spin up CLM compset, for example. However, It gives an error when running the model on other compsets within the subgridAveMod.F90 module.

Has anyone successfully made an older version like cesm2.1.2 from Cheyenne work in Derecho? If so, did you encounter a subgridAveMod.F90 error?
Thanks a lot!
Maria

 

[slevis]

I have a question that may help troubleshoot this, though I do not have experience with the specific problem:
Could you list the exact compsets that result in the error and the exact compset that works?

 

[m_valmartin@sheffield_ac_uk]

Hi Sam, Thanks for getting back to me!

For example, the case works fine for a CLM spin-up with a regional domain: /glade/work/mariavm/derecho_cases/i.clm5.AustSpin.001

But it gives the error in the subgrid module with the following compset (2000_DATM%GSWP3v1_CLM50%BGC-CROP_SICE_SOCN_MOSART_SGLC_SWAV): /glade/work/mariavm/derecho_cases/i.clm5.AustGSWP3.002

The error is within the same subgrid module when running the compset (2000_DATM%GSWP3v1_CLM50%BGC-CROP_SICE_SOCN_SROF_SGLC_SWAV) globally: /glade/work/mariavm/derecho_cases/i.clm5.Cont.001

Thanks again!

Maria

 

[oleson]

The history variable lists in your case that works and the cases that don't are different. I think I would try running one of your failed cases, the global one, with the same history variable lists that you have in the case that is working. Maybe there is a problem with one of the history variables in your failed case. The relevant line in subgridAveMod.F90 is:

SHR_ASSERT_ALL((ubound(carr) == (/bounds%endc, num2d/)), errMsg(sourcefile, __LINE__))

Maybe one of the history variables is not defined correctly.

 

[m_valmartin@sheffield_ac_uk]

Hi Keith and Sam, Thanks a lot for your input! I did more testing running the global configuration and modifying the history field list as suggested. It turns out the issue is with two history variables that I added in my version. I used to ouput them when running the version in cheyenne but it seems the issue is detected in derecho. All fixed now. Thanks again! Maria