Assess the need for incl. topo vars in restart files

[adamrher]

Low priority, but I wanted to get it on the docket.

In the FV dycore, PHIS is both written and read in the restart files, but it is my understanding from Brian E. that for a restart run, this is over written by the value in the bnd_topo file, which is a namelist var. PHIS is not in SE restarts, but (like FV) SGH and SGH30 variables are in the restarts. It is not clear whether these SGH vars are being read in or used from the restarts. All the ridge topo variables for Julio's OGW are not in any of the restarts, and so are read from bnd_topo in all cases.

I'd like to have a discussion on whether the topo vars should be read in from restarts or read from namelist bnd_topo for restart runs. If it is decided that topo vars should be read from bnd_topo in all cases, then I would advocate for removing topo vars from the restarts.

This issue is inspired by confusion on how to couple CISM to CAM, in the scenario where we want changes in ice sheet topography to be reflected in the CAM topo fields.

@PeterHjortLauritzen
@Katetc

[adamrher]

Potentially relevant thread here before I forget:
https://bb.cgd.ucar.edu/cesm/threads/topography-change.5777/#post-38592

Merry Christmas!

[adamrher]

Summarizing the discussion we just had in the amp mtg regarding handling topo fields for restart runs. In particular, how should we update the topography when 2-way coupled CAM/CISM. I think it should be consistent across dycores.

Julio makes the point that modifying the state on restart files violates the definition of what a restart file is. Steve says they are checkpoints.

Peter wonders why not just read all topo fields (PHIS, SGH, SGH30) from the bnd_topo file for all dycores? All the ridge parameterization terms are already read off the bnd_topo file for all dycores.

Adam wonders how FV3 reads in topo for restart runs.

Pat wonders if changing topography for restart runs impacts in-line vertical regridding code.

Might be worth implementing a hydrostatic correction at initialization to adjust the model state if the topo is changed. This would also be helpful for spinning up new initial conditions for new grids. Colin has ncl code to do this, which on requires an old topo state and a new topo state: https://github.com/zarzycki/betacast/blob/76425a1e6d031badc4d53938b93c2eb16942aea2/atm_to_cam/atm_to_cam.ncl#L285

[adamrher]

I wanted to update this thread -- and state that it is frustrating that the FV dycore reads PHIS/SGH/SGH30 from restarts whereas the SE dycore reads these from bnd_topo. Hoping we can resolve this issue in 2023 with Peter's suggestion to always read from bnd_topo.