Pass new seaice_melt field from SIS2 to MOM6

[hdrake]

See companion PRs NOAA-GFDL/SIS2#214 and NOAA-GFDL/FMScoupler#145

[theresa-morrison]

@marshallward Is this sufficient to make sure that fluxes%seaice_melt is not filled with nonsense if IOB%seaice_melt is not allocated or used by the coupler?

[marshallward]

I wouldn't say prevent, it only ensures that the potential nonsense from IOB%seaice_melt is transferred to fluxes%seaice_melt.

The only solution I can see is for net_FW to have some awareness of the state of IOB%seaice_melt. If the coupler fills IOB%seaice_melt, then it can be used to compute new_FW (via fluxes%seaice_melt). Otherwise, it needs to be filled some other way (if at all).

Zero-initialization might get you out of this, although generally we try to avoid that solution where possible.

[theresa-morrison]

I am guessing there is no way to check if IOB%seaice_melt has been allocated with actual values?

What if we add use_seaice_melt to IOB as a logical, default it to false and then set it to true in the coupler where IOB%seaice_melt was allocated? Then use that here to see if seaice_melt should be used or not in this routine.

[theresa-morrison]

After some in preson discussion @marshallward confirmed the if associated should be enough to make sure the pointer was allocated in the coupler. There should be no need to change anything to get this working with older versions of the coupler.

[theresa-morrison]

Were you able to test the changes to the non-FMS caps with the associated couplers?

[theresa-morrison]

Is the expected behavior here that lprec_new + seaice_melt = lprec_old? And if so, do you find that to be the case?
lprec_new being the the liquid precititation into the ocean after these changes and lprec_old being the liquid precititation before these changes.

[hdrake]

Is the expected behavior here that lprec_new + seaice_melt = lprec_old? And if so, do you find that to be the case?
lprec_new being the the liquid precititation into the ocean after these changes and lprec_old being the liquid precititation before these changes.

Yes, that is the expected behavior. I have not done a careful side by side comparison to verify they are the same, but visually they look like it.

What kind of workflow do you use to directly compare two configurations? Checkout the different commits and manually run different configurations and compare the outputs?

[hdrake]

Were you able to test the changes to the non-FMS caps with the associated couplers?

No, I've only tested with the Baltic_OM4_05 configuration that I believe uses the FMS cap.

[theresa-morrison]

What kind of workflow do you use to directly compare two configurations? Checkout the different commits and manually run different configurations and compare the outputs?

Yes. Checkout and compile the dev/gfdl code, run and save lprec. Then recompile with all changes are rerun the experiment saving lprec and seaice_melt. I think it would be sufficient to test the final versions of MOM6, SIS2, and the coupler and not every commmit along the way (assuming of course that it works).

[hdrake]

Were you able to test the changes to the non-FMS caps with the associated couplers?

@theresa-morrison, are these particular experiments I should test with? I guess all of the ones named SIS2_* in the MOM6-examples/ice_ocean_SIS2?

[hdrake]

Is the expected behavior here that lprec_new + seaice_melt = lprec_old? And if so, do you find that to be the case?
lprec_new being the the liquid precititation into the ocean after these changes and lprec_old being the liquid precititation before these changes.

Below is a comparison between the current MOM6-examples/dev/gfdl and my three PRs.

I do not have any idea why the two are different. I don't see how I could have gone wrong in separating the diagnostics in the source code...

[hdrake]

Is it possible that the answers with default parameter values have changed this much?

The boundary mass flux diagnostics seem internally consistent within each of the simulations:

(These residuals are about 1e-5 each of the individual terms in the budget, so indistinguishable from zero at single precision.)

[theresa-morrison]

There should be no parameter changes between the two experiments!
If you are using the version of MOM6 in MOM6 examples for the "old" experiment, you should instead use the version of dev/gfdl that this was branched from to make sure you have all the same defaults.

[hdrake]

If you are using the version of MOM6 in MOM6 examples for the "old" experiment, you should instead use the version of dev/gfdl that this was branched from to make sure you have all the same defaults.

Okay, yeah that makes sense. I'll figure out how to do that and regenerate these plots.

[marshallward]

Some additional comments

• The changes to STALE_mct_cap and nuopc_cap should be removed. These caps are not meant to be used with the FMS coupler, and we don't maintain them. They already have their methods for handling ice-ocean boundaries.

  Additionally, STALE_mct_cap is an archived copy of an unmaintained MCT cap, so it should never be changed.

• Are modifications to MESO permitted? I am unsure if it's meant to be an archive of the original experiments, or if we should extend its support? (I suppose it should give identical answers either way? Maybe this is OK. I will confirm.)

  Similar comments for BFB and dumbbell, although they are probably OK.

[hdrake]

There should be no parameter changes between the two experiments!
If you are using the version of MOM6 in MOM6 examples for the "old" experiment, you should instead use the version of dev/gfdl that this was branched from to make sure you have all the same defaults.

Okay, I did a clean comparison where the only changes were those in my three branches, and I confirmed there were no differences between the MOM_parameter_doc.all and SIS_parameter_doc.all files. However, the differences in the outputs shown above persisted. So it seems I must be doing something wrong in one of these PRs.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 35.91%. Comparing base (9b45087) to head (13b9960).

Additional details and impacted files

@@             Coverage Diff              @@
##           dev/gfdl     #710      +/-   ##
============================================
- Coverage     37.03%   35.91%   -1.13%     
============================================
  Files           273      270       -3     
  Lines         82538    81496    -1042     
  Branches      15441    15427      -14     
============================================
- Hits          30568    29269    -1299     
- Misses        46253    46458     +205     
- Partials       5717     5769      +52     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.