Add scaling of DIB and DISMF by running mean of removed Antarctic ice runoff

[xylar]

This merge adds a capability by which data iceberg (DIB) freshwater fluxes and data ice-shelf melt fluxes (DISMF) can be scaled by the running mean of the the amount of ice (solid) runoff coming from Antarctica (which is removed from the system by MPAS-Ocean).

This merge includes:

• A capability in MPAS-Ocean to compute daily averages of the area integrated removed ice runoff
• A capability in MPAS-Ocean to compute the running mean of these daily means over a given period of time (currently 20 years = 7300 days)
• A capability to pass this running mean through the coupler (via the field rmean_rmv_ice_runoff in the infodata datastructure) from MPAS-Ocean to MPAS-Seaice
• A requirement that the the total annual mean DIB and DISMF fluxes be computed and stored in the DIB and DISMF input files (fulfilled by Add area integrated annual mean to data iceberg and ice-shelf flux files MPAS-Dev/compass#836). These files are currently only available for the IcoswISC30E3r5 mesh.
• Capabilities to compute a scale factor between the annual mean total data freshwater flux and the running mean of the removed ice runoff in both MPAS-Ocean and MPAS-Seaice
• A capability to scale landIceFreshwaterFlux and landIceHeatFlux in MPAS-Ocean by the scale factor
• A capability to scale bergFreshwaterFlux and bergLatentHeatFlux in MPAS-Seaice by the scale factor

The capability is described in more detail in this design document
https://acme-climate.atlassian.net/wiki/spaces/PSC/pages/4210098268/Design+Document+Data+iceberg+and+ice-shelf+melt+flux+patterns+for+E3SM+runs

Relevant discussion can be seen at E3SM-Ocean-Discussion#109

[cbegeman]

@xylar Is there any testing that would be helpful for me to do?

[xylar]

@cbegeman, I think we're mostly waiting on the longer run that @darincomeau is doing. I can't think of other testing that I didn't already report on E3SM-Ocean-Discussion#109, at least not anything that would be a good use of your time.

[darincomeau]

Update, longer test run is in year 50 now: https://web.lcrc.anl.gov/public/e3sm/diagnostic_output/ac.dcomeau/E3SMv3_dev/20241018.v3.SORRME3r3.piControl.scaledDISMF.chrysalis/mpas_analysis/ts_0001-0050_climo_0041-0050/

Note ISMF is not plotting due to #6719, fixed in #6729
Plan is to run it 100 years to see longer trends in sea level/AIS mass imbalance, but looking promising so far.

[xylar]

@jonbob, if you look at the surface fluxes, a lot of them show atmosphere blockiness. Is that expected? Here's an example:

Less so here:

Not related to this PR but showing up in the analysis above.

[xylar]

In terms of mass imbalance and ssh, no sign of trouble:

But we need more years to be sure, so, yes, lets do the full 100 years.

[jonbob]

@xylar -- @cbegeman and I were looking at this last week. We rely on the nonlinear CAAS mapping files to smooth out some of the grid imprinting, but we found that snow and rain are removed from the list of fluxes that get this treatment. It's easily changed in user_nl_cpl if you want to test the impact.

[jonbob]

And I guess the first thing to do is to make sure we have the nonlinear mapping file for atm2ocn specified for this run

[cbegeman]

@xylar and @jonbob I discussed this with @proteanplanet and we will not be changing the atmosphere mapping files because it would compromise conservation accuracy across the coupled system, which is a higher priority. Thus, we will be retaining the blockiness moving forward.

[proteanplanet]

Added information: The decision was made to switch off non-linear remapping following testing in the lead-up to v3, but just for mass fluxes as documented by Andrew Bradley in the Critical Path meeting on December 4, 2023:
https://acme-climate.atlassian.net/wiki/spaces/CM/pages/4021780481/2023-12-04+v3+Critical+Path+Meeting
The subsequent pull request reverting non-linear mass flux remapping is here: #6094

[rljacob]

which grid set is this? I think we only automatically turn on CAAS for the v3 grid set.

[jonbob]

@rljacob -- it's using the new SOwISC12to30E3r3 grid

[xylar]

@jonbob (and anyone else who wants to chime in), is there a reason we can't use higher order, conservative (linear) mapping files from TempestRemap for these fields? It would seem to me that would reduce blockiness without compromising conservation. I'm mostly just seeking to understand why some flux fields are blocky and others don't seem to be.

[jonbob]

@xylar -- we do use conservative mapping files from tempest for all flux fields. The nonlinear CAAS mapping files are also used for some, but not all, of those fluxes. We would probably see less blockiness by also using the nonlinear mapping files for snow and rain -- but apparently there has been a decision not to do that

[rljacob]

Some discussion here: https://acme-climate.atlassian.net/wiki/spaces/CM/pages/4016177153/2023-12-01+v3+Critical+Path+Meeting#Discussion-on-coupler-water-budgets
The nonlinear maps plus the way the coupler does fractions leads to water budget issues for tri-grid.

[proteanplanet]

The links @rljacob and I have provided above summarize the issue, but ultimately it boils down to several decimal places of accuracy in the global water budget; non-linear remapping leaks mass more than linear remapping.

To answer your questions @xylar: A reason why certain fields in sea ice show the aliasing, but others do not, is that sea ice growth is particularly sensitive to snow conductivity. There are other reasons associated with the aging of snow, too, but rain does not play a role because it just runs off.

[xylar]

Okay, sorry for the noise about this. It doesn't relate to this PR in any case, I just noticed it and it looked pretty ugly.

[proteanplanet]

Okay, sorry for the noise about this. It doesn't relate to this PR in any case, I just noticed it and it looked pretty ugly.

I'm very glad you noticed it and questioned it. The more scrutiny E3SM receives, the better it gets.

[xylar]

@cbegeman and I feel uncomfortable with the amount of long-period (~20 year) oscillation in the cumulative land-ice mass removed:

@cbegeman suggested reducing the duration of the running mean to 2 years instead of 20, so we are going to start a new simulation with that change.

I have also rebased to pick up recent changes (including the fix to total land ice fluxes #6729) and cherry-picked the commits from #6501 so we should no longer have redundant frazil fluxes under ice shelves.

[xylar]

@mahf708, do you have any intuition about why my rebase caused a bunch of CI failures in tests that seem to have nothing to do with the changes here?

[xylar]

@mahf708, ignore that. I committed some submodule changes by mistake and that's almost certainly what's wrong.