Replace ne16np4 grid for WACCM HIST test with ne16np4.pg3

[ekluzek]

What happened?

The test SMS_D_Ln9.ne16_ne16_mg17.FX2000.derecho_intel.cam-outfrq9 fails because of a missing transient dataset for ne16np4, while there is one for ne16np5.pg3.

What are the steps to reproduce the bug?

Run SMS_D_Ln9.ne16_ne16_mg17.FX2000.derecho_intel.cam-outfrq9

What CAM tag were you using?

cam6_3_161

What machine were you running CAM on?

CISL machine (e.g. cheyenne)

What compiler were you using?

Intel

Path to a case directory, if applicable

No response

Will you be addressing this bug yourself?

No

Extra info

See the CTSM issue on this:

ESCOMP/CTSM#2547

There it was decided that CAM should address this and change the test to ne30np4.pg3 rather than adding a new dataset in CTSM.

[adamrher]

I just want to double check with @PeterHjortLauritzen here. In the grids doc, we have no land support for ne16np4. I had originally thought that this issue was about missing transient support, but there's actually not even land support for non-transient cases (year 1850, 2000). Are we OK to not support land for ne16np4? We had decided to have some land support (1850, 2000) for ne30np4, but not transient support.

[brian-eaton]

@fvitt, I just tried running FX2000 on the ne16pg3 grid and got the following error:

set_geogrid: npes,nlon_g,nlat_g:    512   144    96
set_geogrid: ntasks_lon,ntasks_lat (oplus and edyn grids):     16    32
Enter edynamo_init:
Grid: gmag_grid, ID =  111, lat coord  = mlat, lon coord  = mlon, unstruct   =  F, block_ind  =  F, zonal_grid =  F
Grid: geo_grid, ID =  112, lat coord  = glat, lon coord  = glon, unstruct   =  F, block_ind  =  F, zonal_grid =  F
 edyn_gcomp_init: coord mismatch... n, lon(n), lonmesh(n), diff_lon =      1    
 321.5625000000000       354.3750000000000             0.32812D+02
 edyn_gcomp_init: coord mismatch... n, lat(n), latmesh(n), diff_lat =      1    
  28.4630367524489        33.6223637109358             0.51593D+01
...
 edyn_gcomp_init: coord mismatch... n, lon(n), lonmesh(n), diff_lon =     27    
 325.3125000000000       358.4452882373436             0.33133D+02
 edyn_gcomp_init: coord mismatch... n, lat(n), latmesh(n), diff_lat =     27    
  38.5124386385902        37.8100702260748             0.70237D+00
 ERROR: edyn_gcomp_init: physics mesh coords do not match model coords

Can FX2000 be run on a CSLAM grid?

[fvitt]

Yes, FX2000 can be run on a CSLAM grid. cam_physics_mesh needs to point to a valid EMSF mesh file for the CSLAM grid. This should work for the ne16pg3 grid:

cam_physics_mesh = '/glade/campaign/cesm/cesmdata/inputdata/share/meshes/ne16pg3_ESMFmesh_cdf5_c20211018.nc'