Add diagnostic: thkcello

[jkrasting]

Add the 'thkcello' diagnostic. Note that the vertical remapping process will complicate this as thkcello will become time-dependent.

For now, this diagnostic should be based on 'h' and follow the convention:

cmor_field_name='thkcello', cmor_long_name='Ocean Model Cell Thickness', cmor_units='m', cmor_standard_name='cell_thickness'

[jkrasting]

In the event that h is not in m, such as in non-Boussinesq mode where h is in kg m-2, will need to calculate in in-situ density and divide.

[StephenGriffies]

The conversion factor, G%H_to_m, is a constant, 1/Rho0. In fact, we need to divide by the in situ density to get a proper accounting of the thickness of a grid cell for the non-Boussinesq model. So your placement of G%H_to_m does the proper unit conversion. Yet it will not provide a proper accounting of the steric changes in a fluid column.

That is, the vertical sum of G%H_to_m*h will not equal to the total ocean depth (bathymetry + sea surface height). Instead, we need to divide by the in situ density at each time step to properly compute the grid cell thickness in a non-Bouss model.

I propose we keep your code change as a placeholder. But if we move to non-Boussinesq for CMIP6, we need to address this issue regarding the need to divide by in situ density. I thus recommend we keep the smg: comment in the code, and keep this issue open on Github.

[adcroft]

Closed by commit ad5aab0 and PR #150.