pkg/shelfice and kpp-mixing underneath ice shelves

[mjlosch]

Non-local vertical mixing models like kpp and ggl90 do not work properly underneath ice shelves.

• For pkg/kpp, the surface flux are computed always for k=1, so that there is no forcing for KPP underneath ice shelves and only internal Ri-number based mixing is in effect, see related comments in pkg/kpp/kpp_calc.F
• For pkg/ggl90, nothing is done to use it with pkg/shelfice and one should not expect sensible results. The minimum changes required are sketched here: https://github.com/mjlosch/MITgcm/blob/ggl90_shelfice/pkg/ggl90/ggl90_calc.F With proper masking with hFac's this may work but needs to be tested.
• Update: PR #597 fixes the issue for pkg/ggl90, and pkg/ggl90 should now work properly with pkg/shelfice.

[mjlosch]

I think that the changes in #590 are not a good way of making 'pkg/ggl90' work with 'pkg/shelfice'. It also mixes two packages unnecessarily. @owang01 If I create a PR from this: https://github.com/mjlosch/MITgcm/blob/ggl90_shelfice/pkg/ggl90/ggl90_calc.F would you be willing to test this in your ecco4 setup? Alternatively, you could use my branch as a starting point for a new PR of yours and I can help. I currently do not have a useful test for 'shelfice+ggl90'. E.g., without any changes to the code, nothing happens when I turn on ggl90 in isomip. With my changes there is a lot of mixing, but then again, there is no stratification in the initial conditions of that experiment, so I expect a lot of mixing just to get a stable water column, but the strong mixing also leads to some noise in the diffusivity coefficients. We can also wait until @jm-c is done with #593, if there will be a useful test within that PR.

[owang01]

@mjlosch If you create a PR, I'd be happy to test it out in the ECCO V4r5 setup.

[jm-c]

@mjlosch when you have time, may be you could update this issue since: a) regarding pkg/ggl90 and ice-shelf, things got fixed in PR #597 ; and one remaining little (potential) improvement is described in issue #627. b) so the only reamaining issue is with pkg/kpp (nothing has been done there). What do you think ? May be even update this issue title ?