CAM4 initial crash due to ZM_CONV under Permian geography

[Martin Renoult]

Hello all,

I am a paleoclimate modeller at Stockholm University and I have been using CESM1.2-CAM4 with recent (~Pliocene - Miocene) and older (Cretaceous) geographies with success over the last two years. Recently, I have prepared a late Permian simulation (~250 million years ago, so a tremendous change of land-sea mask), but the model crashes at initialisation with an error I have not seen before, and for once I cannot find any obvious problem in my boundary files.

The error is the following:
ERROR: **** ZM_CONV IENTROPY: Tmix did not converge ****
ZM_CONV: IENTROPY. Details: call#,lchnk,icol= 2 329 8 lat: 54.00 lon: 62.50 P(mb)= 37.23 Tfg(K)= 180.10 qt(g/kg) = 0.00 qst(g/kg) = NaN, s(J/kg) = Inf

Since my model does not even start, I assume the issues is somehow related to the boundary conditions, also since I did not experience problems before even on early Cretaceous geographies, which are also quite a big change in land-sea mask. The error seems to indicate a point in latitude = 54° and longitude = 62.5°, which is the location of 5-km high mountains in my geography.

I have tried the following change without any impact on the problem:
- Decreasing the atmospheric time step of CAM4
- Lowering the initial, non-continental forcing (solar, CO2)
- Lowering all mountains from 5 km to 4.5 km

Thanks for helping!

 

[jiangzhu]

Hi Martin,

Since you have tried decreasing timestep of CAM4 (I assume its the dynamical timestep through changing nsplit), you may want to try some "fixes" in deep-time CESM: https://svn-ccsm-models.cgd.ucar.edu/cesm1/exp_tags/pcesm_cesm1_2_2_tags/cesm-dt2.0_cesm1_2_2_1/
Specifically, the following updates may be relevant:
- In the ZM deep convection scheme, use Brent's method to invert the entropy equation, as it converges faster. (cam5_3_24)
- In the ZM scheme, for the mixing ratio of water vapor at saturation, use a ratio over total mass rather than over dry mass, to avoid numerical overflow that can occur at high temperature. (cam5_3_24)

Best,
Jiang

 

[Martin Renoult]

Hi Jiang,

Thank you for your answer! I have managed to find a temporary solution before getting your message where I slightly lowered the height of those particular mountains and the model is now running fine for two stages of the Permian. Since I managed to keep them quite high, I think it is a good enough solution for now since the paleogeography is quite uncertain anyway. I will try your solution later on when I have more time. I cannot svn the version of the model you are providing (at least not entirely, some links seem to be problematic), but I managed to get the ZM F90 file. Can I just replace the ZM source file of my version of the model by this one perhaps?

Martin

 

[jiangzhu]

Yes, you can simply replace the changed files. Please compare and figure out which files are changed.