Remove redundant sea ice enthalpy flux term

[cbegeman]

The enthalpy flux associated with seaIceFreshWaterFlux is already included in seaIceHeatFlux which derives from fhocn in Icepack. Thus, it was incorrect to apply both seaIceHeatFlux and seaIceTemperatureFlux in MPAS-Ocean, as seaIceTemperatureFlux provides the thickness of seaIceFreshWaterFlux at the freezing point.

Here, we set seaIceTemperatureFlux to 0 always. The variable seaIceTemperatureFlux could be removed entirely if desired. I retain it here for clarity.

[cbegeman]

@vanroekel I wanted to check in with you partly to make sure that we don't need to make any changes to the surface buoyancy forcing used by KPP. I believe we do not need to make any changes there because seaIceHeatFlux is included in the surface temperature flux used to compute the buoyancy forcing.

[cbegeman]

Given that @proteanplanet, @alicebarthel and I feel reasonably confident in this change, I am going to proceed with testing this branch with a v3.LR.piControl simulation. For that test, I am using a branch based on the v3.0.2 code base: https://github.com/cbegeman/E3SM/tree/v3.0.2/fix-sea-ice-enthalpy-flux.

[alicebarthel]

Note that if ktherm != 2 (non-mushy thermo), emlt_ocn = 0 and no enthalpy/heat flux is passed.
Maybe the seaIceTemperatureFlux could be retained for cases where non-mushy is used?

[vanroekel]

@cbegeman I agree with your assessment and also that this is a double counting based on previous discussions.

[cbegeman]

Note that if ktherm != 2 (non-mushy thermo), emlt_ocn = 0 and no enthalpy/heat flux is passed. Maybe the seaIceTemperatureFlux could be retained for cases where non-mushy is used?

@alicebarthel That's worth considering. I think there is an argument to be had on the MPAS-Ocean side for forcing it to be 0 for now:

• non-mushy thermodynamics is not available in MPAS-Ocean (the ocn_freezing_temperature function called in that line previously is the mushy thermodynamics except when under land ice)

Note to anyone following the discussion: config_thermodynamics_type=ktherm is in the MPAS-Seaice namelist and not currently available to MPAS-Ocean. We could create a new config option in MPAS-Ocean and ensure consistency with this config option in MPAS-Seaice, though that's not elegant.

[alicebarthel]

I agree that we should set it to zero for now. If there is no need to retain compatibility with MPAS-SeaIce using non-mushy thermo then that's even cleaner. I just wanted to make sure we did not inadvertently remove something that is needed/wanted in a different (supported) configuration.

[alicebarthel]

Also FYI, the enthalpy/heat flux counted in both is quite different since seaIceTemperatureFlux calculates it based on ocean salinity, and emlt_ocn (which goes into seaIceHeatFlux) is based on freezing temperature at the sea ice salinity (which will be a higher freezing temperature).
Obviously it's second order compared to double counting but thought I'd document it here.
Interested to see what comes out of this, thanks for testing this @cbegeman!

[alicebarthel]

@cbegeman if we need to retain a melt heat flux when MPAS-SeaIce is non-mushy, it'd be easier in practice to add the flux in Icepack since we have ktherm and sss both available in there. But obviously, changing the default behavior of Icepack when ktherm != 2 is something that impacts other models that rely on Icepack so it's beyond just our own project decisions...

[alicebarthel]

Approving based on my understanding of the terms accounted for in emlt_ocn. My only caveat is on whether it would be preferable to remove seaIceTemperatureFlux entirely, or keep it as an option for non-mushy sea ice thermodynamics.

[cbegeman]

Testing

I have run through the end of year 69 of a v3.LR.piControl run branched from the v3.LR.spinup: https://e3sm.atlassian.net/wiki/spaces/pd/pages/5581930507/v3.LR.piControl-enthalpyfix

Key points:

• Differences appear to be non-climate changing (NCC) with internal variability seen in some sea ice regions and mixed layer depths

Analysis shown here is years 1-50, climo years 1-50:



SST diffs from Hadley ANN