Why do I have to care about CISM grid when trying to build CAM

[martina]

Hello!
I installed the latest version of CAM from GitHub - ESCOMP/CAM: Community Atmosphere Model, but I can't build it right now as the CISM module needs a specific grid (gland20 or gland4 instead of gris4). I don't understand why I have to worry about the ice grid when I'm using CAM. Could you please help me understand this? Also I'm not sure which one of these options (gland4 or gland20) are better to use.
The command I used to create is: ./create_newcase --case ../../cases/ERAi_ndg_movmtn_on --compset FWscHIST --res f09_f09_mg17 --walltime 10:00:00 --mach derecho --project XXXX --compiler intel

I'm grateful for any kind of insight.
Thank you,
Martina

 

[sacks]

Very good question. Indeed, in recent development versions the various CAM compsets should not be using CISM. Perhaps you're using a release version of CAM where CISM was still included in this and other compsets?

You have two options: (1) Use the gland4 grid for CISM; this is probably easiest to just get going, though note that this doesn't work if you are running with a Gregorian calendar that has leap years (most configurations use a NOLEAP calendar, but some configurations require a calendar with leap years, which currently doesn't work with CISM). (2) Use a compset long name (rather than alias), and replace "CISM2%NOEVOLVE" in the compset long name from FWscHist with "SGLC" to use a stub glacier model.

To answer your broader question about why this is needed: In versions of the code that were close to those used in the CMIP6 experiments, we included an active ice sheet component over Greenland in most configurations; this serves two main purposes: (1) it uses the most accurate ice sheet coverage over Greenland in the land model coupling to the atmosphere; (2) it provides a grid onto which we can downscale ice sheet surface mass balance for later analysis. However, in recent development versions, we decided to remove these capabilities from most default configurations because, as you have stumbled across, it can sometimes cause extra confusion.

 

[martina]

Very good question. Indeed, in recent development versions the various CAM compsets should not be using CISM. Perhaps you're using a release version of CAM where CISM was still included in this and other compsets?

You have two options: (1) Use the gland4 grid for CISM; this is probably easiest to just get going, though note that this doesn't work if you are running with a Gregorian calendar that has leap years (most configurations use a NOLEAP calendar, but some configurations require a calendar with leap years, which currently doesn't work with CISM). (2) Use a compset long name (rather than alias), and replace "CISM2%NOEVOLVE" in the compset long name from FWscHist with "SGLC" to use a stub glacier model.

To answer your broader question about why this is needed: In versions of the code that were close to those used in the CMIP6 experiments, we included an active ice sheet component over Greenland in most configurations; this serves two main purposes: (1) it uses the most accurate ice sheet coverage over Greenland in the land model coupling to the atmosphere; (2) it provides a grid onto which we can downscale ice sheet surface mass balance for later analysis. However, in recent development versions, we decided to remove these capabilities from most default configurations because, as you have stumbled across, it can sometimes cause extra confusion.

Thank you Bill for all your information, I appreciate it! Now I have more questions :) (my apologies for all the beginner questions)
1.) What is the proper compset long name for FWscHist?
2.) what is actually the proper procedure to get WACCM or CAM running on Derecho?
What I did was:
git clone GitHub - ESCOMP/CAM: Community Atmosphere Model
cd CAM
git checkout cam_cesm2_2_rel_02
./manage_externals/checkout_externals
cd cime
git checkout remotes/origin/maint-5.8
Please note that I never had to perform the two bold steps on Cheyenne, but it was the only way I could create a new case on Derecho. Otherwise the ./newcase command was not working as it wouldn't recognize derecho as a valid machine. What does checkout remotes do? Is this the proper workflow to get CAM installed and running on Derecho?

Thank you!
Martina

 

[dbailey]

Are you sure it is CISM and not CICE?

 

[sacks]

Sorry for the delayed reply. And no problem for the beginner questions - you have to start somewhere! And actually, many of your questions are beyond beginner, anyway, including your latest ones, since we're still in the midst of the cheyenne-to-derecho transition.

To see the long name for a compset, you can run the query_config script from cime/scripts: ./query_config --compsets | grep -i FWscHist. With the latest development code, this gives FWscHIST : HIST_CAM60%WCSC_CLM50%SP_CICE%PRES_DOCN%DOM_MOSART_SGLC_SWAV (with the release code you're using, it will probably have CISM2%NOEVOLVE instead of SGLC). (Another way to see the compset long name is to create a case and then run ./xmlquery COMPSET.)

Regarding getting the code running on derecho: As you have found, the 2.2 release code doesn't work out-of-the-box on derecho, because derecho didn't exist when this release was created. We are in the process of transitioning from cheyenne to derecho, and I recommend posting in the CAM forums for guidance on the best code base to use right now for running on derecho. In particular, I'm not sure if there's a version of the 2.2 CAM release that has been ported to derecho, or if the guidance would be to use the CAM development code for now. In your post, you could describe your reasons for wanting to use the release code over the latest development version.

The git checkout command you ran gets the latest version of the CIME maint-5.8 branch. In general, I wouldn't recommend this unless someone has guided you in that direction: We run testing with specific versions of CIME and other repositories, and you can run into trouble mixing and matching versions that may not have been fully tested together.

 

[martina]

Thank you once again Bill for your insightful answers! I'll reach out to CAM forums.