Add ocean option for Redi kappa to be set equal to GM (AMOC PR 3/6)

[mark-petersen]

add new option: config_Redi_closure = 'equalGM', sets RediKappa equal to gmBolusKappa everywhere. Default remains config_Redi_closure = 'constant'.

[NML]
[BFB]

[mark-petersen]

This PR was started with the commits in #4835 and #4864. To see the changes relevant to this PR, look at this last commit: 454cfd6

[vanroekel]

Assuming @jonbob is happy with the NL changes, the actually code mods look good to me. Approving by visual inspection. It would be good though if someone could run a really short test to verify in output rediKappa = gmBolusKappa. Maybe this has been done already though.

[mark-petersen]

@vanroekel I checked the output, and indeed RediKappa = gmBolusKappa when config_Redi_closure = 'equalGM'. But in the process, I realized that the ramp array, gmResolutionTaper, is only applied to GM, and not Redi (unless I missed it in the code). That would be a mistake, so I need to check carefully. It was my intention that the taper is applied uniformly to gmBolusKappa and RediKappa. Do you agree?

[vanroekel]

Yes you are totally right, good catch there. I forgot to mention that I passed gmResolutionTaper to redi in my implementation. This could be added here or in #4864 as it is odd to just taper GM

[darincomeau]

I've run two 5 year CRYO1850.ECwISC30to60E2r1 cases with this branch for comparison against the baseline:

Baseline with constant GM

Baseline with Visbeck GM

This branch with config_Redi_closure = 'equalGM'; default GM here is constant

This branch with config_Redi_closure = 'equalGM' and config_GM_closure = 'Visbeck', so GM kappa is not constant

[milenaveneziani]

I have run a 5-year test with this branch for the SORRM that is equivalent, in terms of settings, to Darin's third set mentioned above:
https://web.lcrc.anl.gov/public/e3sm/diagnostic_output/ac.milena/20220408.CRYO1850.ne30pg2_SOwISC12to60E2r4.chrysalis.CryoBranchRediEqGM/Years1-5/

[mark-petersen]

I realized that the ramp array, gmResolutionTaper, is only applied to GM, and not Redi (comment above)

Never mind. I did apply horizontal tapering to Redi this whole time. I just did it directly on init, rather than with a coefficient array.

E3SM/components/mpas-ocean/src/shared/mpas_ocn_tracer_hmix_redi.F

Lines 546 to 560 in 22db376

	! Add resolution taper
	if (config_eddying_resolution_taper == 'none') then
	! Nothing to do, as we just keep the same assignment as above.
	else if (config_eddying_resolution_taper == 'ramp') then
	!$omp parallel
	!$omp do schedule(runtime)
	do iEdge=1,nEdges
	if (dcEdge(iEdge) <= config_eddying_resolution_ramp_min) then
	RediKappa(iEdge) = 0.0_RKIND
	else if (dcEdge(iEdge) >= config_eddying_resolution_ramp_max) then
	! nothing to do, i.e. RediKappa(iCell) = RediKappa(iCell)
	else
	RediKappa(iEdge) = RediKappa(iEdge) &
	*(dcEdge(iEdge) - config_eddying_resolution_ramp_min) &
	/(config_eddying_resolution_ramp_max - config_eddying_resolution_ramp_min)

It’s only with the new ‘RossbyRadius’ tapering option in this PR that the taper changes in time, so changing RediKappa in init made sense.

[mark-petersen]

This PR requires the same init and additional tapering variables as #4864, so I am closing this one and adding this single flag addition (config_Redi_closure = 'equalGM') to that PR.