CAM5 crashes at high CO2 in iCESM

[mneedham]

This is a continuation of a previous post of mine here.

Essentially, I am performing a "F-Compset" simulation with iCESM forced by CMIP6 abrupt_4xCO2 SSTs and Sea Ice, and 1138ppm CO2 to the atmosphere and land models. I have made this change to the source mods in order to get to this point (otherwise, the model crashed at the first timestep). My simulation crashes after about 60 simulated days, which suggests that the model is initially setup correctly, but soon becomes unstable - perhaps I need to shorten the model timestep?

Also, I have looked through the error logs and the following line appears relevant:

103:(shr_sys_abort) ERROR: **** ZM_CONV IENTROPY: Tmix did not converge ****

Is there a known issue with either iCESM or CAM5 about instability of the Zhang-McFarlane convection scheme at high CO2 (4X pi → 1138 ppm)? This post on the bulletin board from 2014 appears relevant due to a discussion around ZM, although the zm_conv SourceMod file discussed by “Santos” is missing.

 

[cacraig]

The bulletin board migrated a number of years ago to a new platform and the attachment was a casualty of that move. Our sys admin was able to go back to the archives and found three files which were attachments with names similar to the name in the 4xCO2 experiment crashed during running...post. The zm_conv.F90.gz file has the exact same name as what was referenced in the original posting, so I believe it is the correct one. I am also attaching the two other files with zm_conv in their name, but suspect they are not the correct ones. You will need to keep this in mind when you look at these files.

 

[mneedham]

Thank you Cheryl. I will attempt to build and run with that zm_conv.F90 fil

 

[mneedham]

I was able to get around this issue. I am writing an update in case anyone ever has a similar issue:

The model crashing at high CO2 appears to be resolution dependent. I ran the exact same forcing at 2 degrees (1.9x2.5) and the model ran out to 6 years without issue. The model only crashed when run at 1 degree (0.9x1.25). It is unclear whether 2 degrees is the highest resolution that will run without crashing, or whether 1 degrees is a unique problem.

• Note: The files attached above for zm_conv.F90 made it so that the model would not build. It appears they were for a different version of CAM, as certain variable names are different to the iCESM source code (e.g., "qs" vs. "qsat" in the attached zm_conv.F90 vs the iCESM source code zm_conv.F90). Given enough time and patience, they could probably be made to work with this version of iCESM.