ZM_CONV IENTROPY: Tmix did not converge

[yingli@atmos_colostate_edu]

Hi,

I am trying to create a branch run using the restart files from an previous exiting run done by someone else.

I got this error message in the cesm.log.*

ERROR: **** ZM_CONV IENTROPY: Tmix did not converge ****
ERROR: 0031-300 Forcing all remote tasks to exit due to exit code 1 in task 5282

Does anyone know how to solve this problem?

My case directory is /glade/u/home/yingli/cesm_1_2_2/runs/f.FAMIPC5.ne120_ne120.test.004

Thanks,
Ying

 

[jedwards]

Hi Ying,This is an error that you typicaly see when something is wrong in the initialization.   Check that the restart files you are trying to use are appropriate for your model version and configuration.

 

[santos]

This signals something that could be a numerical stability problem in the Zhang-McFarlane convection scheme (ZM), but it can also be caused by bad inputs to ZM caused by problems with initialization, or instability in the dycore. You can try either of the following:1) Use the attached file in SourceMods to replace the version of zm_conv (I think it should work with CESM 1.2.2, but not older versions). This change makes ZM stable over a wider range of temperature/pressure conditions.2) Enable the option "state_debug_checks = .true." in the CAM namelist, to try to catch very bad values (e.g. infinity) more quickly. This may give you more information to find where the real problem is, if it is not actually in ZM.

 

[zhu467@purdue_edu]

Hi Santos,When I run cesm1.2.2 using B20TRCN, I want to setup different co2 level for land and atm component, so I set the namelist variable (co2vmr_rad) in atm_in, I leart from the guide which said this variable can rewrite the co2 senting to radiation part calculation while still sent transient co2 to land component. But I encountered the same error presented in the begining.  Then I use the zm_conv.F90 your provided, but it still crashed with the following error. I want to hear your suggestion. QNEG3 from TPHYSBCb:m=  3 lat/lchnk=    707 Min. mixing ratio violated at    1 points.  Reset to  0.0E+00 Worst =-8.0E-12 at i,k=   2  1 QNEG3 from TPHYSBCb:m=  3 lat/lchnk=    708 Min. mixing ratio violated at    1 points.  Reset to  0.0E+00 Worst =-3.5E-11 at i,k=   2  1 Lagrangian levels are crossing Run will ABORT! Suggest to increase NSPLTVRM(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping

 

[santos]

That message typically means that the model has become unstable. It may be that you have a mistake in your namelist settings, or you may simply need to run with a finer timestep. I would double-check your settings. If they seem correct, you can also try increasing the values of nsplit, nspltrac, and nspltvrm in your namelist (to see the current values, look in CaseDocs/atm_in, or search the atm.log).

 

[zhu467@purdue_edu]

Hi Santos,Thanks for your suggestion and I have reset the parameters nsplit as 2 but it still crashed. I just used the default configuration for B20TRCN with a resolution of f19_g16, except that I modified co2vmr_rad as a constant to overwrite the co2 which will be sent to radiation calculation. I hope you can make a test with this configuration. Best,Peng

 

[zhu467@purdue_edu]

Hi Santos,Have you checked the configuration I used? I have modified nsplit and nspltvrm but it will still crash.Thanks,Peng

 

[santos]

Peng, please see my recent reply here, since if you are increasing CO2 levels, it may be relevant to your case as well:https://bb.cgd.ucar.edu/waccmsc-4xco2preindustry-initial-condition

 

[zhu467@purdue_edu]

Hi, Santos Thanks for your help. Actually, I setup a decreased co2 level since I use the transient co2 for land component but a constant (285ppm) co2 level for atm radiation calculation.  Now the error becomes this: (shr_sys_abort) ERROR: Bad namelist settings for FV subcyclingBest,Peng Best,Peng

 

[santos]

Make sure that nsplit is a multiple of nspltrac, and nspltrac is a multiple of nspltvrm. If you don't use an exact multiple (e.g. nspltrac is 3 and nspltvrm is 4), the model will not run. 

 

[jie_wong]

This signals something that could be a numerical stability problem in the Zhang-McFarlane convection scheme (ZM), but it can also be caused by bad inputs to ZM caused by problems with initialization, or instability in the dycore. You can try either of the following:1) Use the attached file in SourceMods to replace the version of zm_conv (I think it should work with CESM 1.2.2, but not older versions). This change makes ZM stable over a wider range of temperature/pressure conditions.2) Enable the option "state_debug_checks = .true." in the CAM namelist, to try to catch very bad values (e.g. infinity) more quickly. This may give you more information to find where the real problem is, if it is not actually in ZM.

Hello Santos,
I encountered the same errors of

ZM_CONV IENTROPY: Tmix did not converge​

when I am trying to modify the model pressure. I agree that it is a problem of model unstable because in some word I want to add a unstable pressure in the model. You mentioned the 'attached file in SourceMods to replace the version of zm_conv ', but I didn't find any attached file in SourceMods, is there a link you can provided? or the place to find?

Thanks for help!
Jie

 

[rdeitrick]

Hi Jie,

I've been running into this issue as well in some paleo applications. I haven't solved the issue, but this thread might have the relevant file that Santos referred to (though it seems there is some uncertainty over whether any of the attachments are the exact same file). I have yet to test it.

I'd be interested to know if you have any luck solving the issue.

Cheers,
Russell

 

[jie_wong]

Hi Jie,

I've been running into this issue as well in some paleo applications. I haven't solved the issue, but this thread might have the relevant file that Santos referred to (though it seems there is some uncertainty over whether any of the attachments are the exact same file). I have yet to test it.

I'd be interested to know if you have any luck solving the issue.

Cheers,
Russell

Hi Russell,
Thanks for your help on the file. I have solved the problem by increasing the 'ATM_NCPL'. And in my case, when I modify the surface pressure, I should adjust the 3-d pressure by ak+bk*ps for each layer. After I modify these two steps, the errors were disappeared.
Maybe you can try some.

Good Luck
Jie

 

[rdeitrick]

Good to know, thanks Jie!