Help required regarding a CLM5 simulation using preindustrial CO2 levels

[knreddy]

Hi,
I am trying to analyze the impact of CO2 concentrations on the carbon fluxes in Indian agroecosystems. I am running the control simulation for a span of 20 years, which usually takes me around 70 hours to complete (using 196 processors). The resolution of the model is 0.5x0.5 deg, and it is a regional run covering the Indian subcontinent region. The compset I am using for the simulation is HIST_DATM%GSWP3v1_CLM50%BGC-CROP_SICE_SOCN_SROF_SGLC_SWAV.
I am trying to run a CLM5 simulation using constant pre-industrial CO2 concentrations.
The issue arises when I try to run a CLM5 simulation using constant CO2 levels by using the CLM_CO2_TYPE = constant in the env_run.xml file. There is no other difference between the control run (using CLM_CO2_TYPE = diagnostic) and the reduced CO2 concentration run. But when I use constant CO2 levels, the model simulation is taking longer to complete. It is taking around 100 hours to complete 5 years of simulation (using the same number of processors as the control run).

Has anyone faced a similar issue, or is it a common thing expected from CLM5? Please help me rectify this and perform the simulations effectively.

Thanks and regards
Narender

 

[oleson]

There shouldn't be any reason why such a simulation would be so much slower than your historical. You could setup an 1850 compset to see what the CO2 setting in that are and compare to your constant simulation setup. For an 1850 simulation and your simulation I would expect to see the following after executing ./xmlquery -p CO2 in your case directory:

Results in group run_co2
CCSM_CO2_PPMV: 284.7

Results in group run_component_ctsm
CLM_CO2_TYPE: constant

Results in group run_component_datm
DATM_CO2_TSERIES: none

Otherwise, the only thing I can think of is that you are running your constant CO2 simulation in DEBUG mode for some reason?

 

[knreddy]

Hi Oleson,
Thanks for this quick reply.
How may I check if the current simulation is running in debug mode??

And if it is running in debug mode how can I change it?

Thanks in advance

 

[oleson]

./xmlquery DEBUG

./xmlchange DEBUG=FALSE

If you are not aware of changing it, then it is probably FALSE already.

 

[knreddy]

Yes, I haven't changed the settings. and it shows DEBUG=FALSE when xmlquery is used
What else could cause this issue?
I've tried building the case from scratch, and still the same issue persists.

 

[oleson]

You could double-check that your pe layouts are the same in both cases by running ./pelayout in your case directories.

 

[knreddy]

Yes, both the pe layouts are same.
./pelayout % for the CO2 case
Comp NTASKS NTHRDS ROOTPE
CPL : 24/ 1; 0
ATM : 24/ 1; 0
LND : 24/ 1; 0
ICE : 24/ 1; 0
OCN : 24/ 1; 0
ROF : 24/ 1; 0
GLC : 24/ 1; 0
WAV : 24/ 1; 0
ESP : 1/ 1; 0

./../CLM5_CTRL/pelayout
Comp NTASKS NTHRDS ROOTPE
CPL : 24/ 1; 0
ATM : 24/ 1; 0
LND : 24/ 1; 0
ICE : 24/ 1; 0
OCN : 24/ 1; 0
ROF : 24/ 1; 0
GLC : 24/ 1; 0
WAV : 24/ 1; 0
ESP : 1/ 1; 0

 

[knreddy]

The issue is rectified when I build the model using the compset 1850_DATM%GSWP3v1_CLM50%BGC-CROP_SICE_SOCN_SROF_SGLC_SWAV.
The reason for the slow simulation might be the way the model was built in my case, using a historical DATM and stating CO2 as a constant.

It might be a helpful remark for others not to do this.

 

[knreddy]

But then I cannot make the comparison between the control and the Limited CO2 simulations because the historical land use timeseries also varies?

 

[slevis]

What if you make a copy of the case that worked for you using ./create_clone or ./create_newcase (either should be fine). Then modify the new case manually to make it do what you had intended with your "hist" case. I usually do this with the unix "diff" command on the directories of the two cases. Something like this, though I don't know the actual names of your cases: "diff clone hist". This way you see what you need to get from your hist case into your cloned case.

(I hope this helps with your last question, but I may have misunderstood what you were asking.)