Tighten eps_imesh and eps_omesh

[anton-seaice]

Using the default value (0.1) for eps_imesh allows a very generous error in latitude and longitude between the cmeps mesh files and the cice components grid.

We can probably set this to 1e-7 or so? It is configurable through nuopc.runconfig

The eps_omesh default is better (0.0001) but could be stricter too?

The values seem to only be used to set the allowable error between lon/lat values read from the mesh vs those read from the component grid file.

[anton-seaice]

Strangely, making this change fails in cice at a collection of seemingly unrelated latitudes in 1 degree config:

ERROR: CICE n, mesh_lat , lat, diff_lat =        1    -73.7954776730756       -73.7937522141561             0.17255D-02
ERROR: CICE n, mesh_lat , lat, diff_lat =        1      4.5004574467160         4.5001706474547             0.28680D-03
ERROR: CICE n, mesh_lat , lat, diff_lat =        1    -77.8766233766234       -77.8773455120729             0.72214D-03
ERROR: CICE n, mesh_lat , lat, diff_lat =        1    -67.7330184754090       -67.7301835653587             0.28349D-02
ERROR: CICE n, mesh_lat , lat, diff_lat =        1    -73.7954776730756       -73.7937522141561             0.17255D-02
ERROR: CICE n, mesh_lat , lat, diff_lat =        1    -77.8766233766234       -77.8773455120729             0.72214D-03
ERROR: CICE n, mesh_lat , lat, diff_lat =        1     15.8505655907296        15.8595374685817             0.89719D-02

We should be doing better than that right @aekiss ?

EDIT:

For mom, this is set in MOM_input, and is accurate to 1e-13 (i.e. add EPS_OMESH = 1e-13 to MOM_input)

Extra note:

This is probably the same problem as #116, seems better if i rename angleT to anglet in the grid.nc file

[aekiss]

Hm, odd to have a relative difference of ~1e-5 if it's supposed to be 1e-13. The largest of these differences (0.89719D-02) is about the 1/10th the scale of the 0.1° grid cells, ie ~1km.

[anton-seaice]

Yeah thankyou - it's because CICE is trying to internally calculate anglet from angle(u), rather than reading from the file.

Once this is fixed (by renaming angleT to anglet in grid.nc) , we can enforce 1e-13 for both cice and mom :)

[anton-seaice]

Closed through linked PRs

EPS_OMESH and EPS_IMESH both set to 1e-13