adcroft
commented
2 weeks ago The mask should be derivable from thkcello or volcello and would be "wet" for the partial cells.
I would thickness weight the velocities when interpolating to cell centers. With partial cells this will appropriately scale flow down near "walls".
If easy then yes, use conservative interpolation in the vertical taking into account the partial cells. However, since most of the action is in the upper ocean, I would not let this detail hold up getting something up and running.
FWIW, those partial cells at the bottom of the column technically should be handled differently but it makes things very cumbersome; the centers of those partial cells are raised above the nominal level and this turns an ostensibly a 2d lateral interpolation into a 3d interpolation. I think they are generally treated as if at the nominal depth purely for convenience even though wrong.
I think this is probably a question for @adcroft? but also pinging @LaureZanna @adcroft @adam-subel @IamShubhamGupto @suryadheeshjith!
Does the regridding for CM4/CMIP data in general need a mask for partial bottom cells? This will be important to conserve global integrals.
A confounding issue is that our current plan (#12) we plan to:
This makes me wonder what the recommended way of handling wet_masks for tracers and velocities is?
As an ad-hoc solution to not block progress I could simply mask out the cell closest to the bottom.