Should the VR grids continue to use the POP t12 land/ocean mask or be updated to the MOM t232 mask?

[ekluzek]

What is the feature/what would you like to discuss?

All of the testing for the VR grids in CAM has them using the POP t12 land/ocean mask. The only MOM land/ocean mask is t232, so to be consistent with the ocean model it makes sense for us to remove the use of the POP mask and use the MOM mask. At least in order for our simulations to be the most consistent with the fully coupled simulations and for them to make the most sense when they run fully coupled. Near the genesis of CSM I know there were science issues discovered when components did standalone simulations on a land/ocean mask different than the grid used for fully coupled simulations. So it became a convention to run our standalone simulations on the same land/ocean mask as the ocean model that was being used. Doing this has served us well. I'm concerned about those problems coming back to haunt us if we continue to run our standalone and testing simulations at an ocean/land mask that isn't even available in the fully coupled system anymore.

However, the impact to these high resolution grids is that the t232 mask is pretty course and t12 is much higher and when run with a higher resolution atmosphere -- you may purposely WANT that higher resolution land mask. Especially if it's thought that running the VR grids fully coupled is NOT something that will be done often.

If we do want to run the VR grids at the MOM t232 mask we will need to add them to ccs_config. So part of what I'm asking here is -- should we do that?

Here are the VR grids that are currently in the CAM test list:

  <test compset="FCnudged" grid="ne0CONUSne30x8_ne0CONUSne30x8_mt12" name="SMS_D_Ln9_P1280x1" testmods="cam/outfrq9s">
  <test compset="FHIST" grid="ne0CONUSne30x8_ne0CONUSne30x8_mt12" name="SMS_D_Ln9_P1280x1" testmods="cam/outfrq9s" supported="false">
  <test compset="FHIST" grid="ne0ARCTICne30x4_ne0ARCTICne30x4_mt12" name="SMS_D_Ln9_P1280x1" testmods="cam/outfrq9s" supported="false">
  <test compset="FHIST" grid="ne0ARCTICGRISne30x8_ne0ARCTICGRISne30x8_mt12" name="SMS_D_Ln9_P1280x1" testmods="cam/outfrq9s" supported="false">
  <test compset="FCHIST" grid="ne0CONUSne30x8_ne0CONUSne30x8_mt12" name="SMS_D_Ln9_P1280x1" testmods="cam/outfrq9s">
  <test compset="FCHIST" grid="ne0CONUSne30x8_ne0CONUSne30x8_mt12" name="ERS_Lh3_P1280x1" testmods="cam/outfrq3h">

Is there anyone in particular you want to be part of this conversation?

@adamrher @patcal @PeterHjortLauritzen @JulioTBacmeister

Will this change (regression test) answers?

No

Will you be implementing this enhancement yourself?

Any CAM SE can do this

[ekluzek]

I did also just realize that there is a quarter degree MOM grid/mask that's at least recognized as experimental and has one test for it in MOM. I'll be asking CSEG if that's something that should/could be used for high resolutions for components. But, this does seem like the MOM mask that makes the most sense for this case.

[adamrher]

Hi Erik. I just saw that 1/4deg MOM6 grid as well and that might be sufficient for replacing the t12 and t13 ocean masks in the VR grids. I would like to know if more MOM6 grids are in the pipeline as I know the ocean folks have been experimenting with a 1/12deg grid, which would then be my preference.

You are correct that we don't support fully coupled configurations for VR and so I think there is less urgency to resolve this than the CESM3 workhorse grids aliases, which do currently use different ocean masks (ne30pg3_ne30pg3_mg17 vs ne30pg3_t232) as you've pointed out.