(Possibility for) negative ice runoff in BLOM in NorESM runs with multiple Greenland and Antarctic ice sheet coupling

[mpetrini-norce]

I would like to report a model behaviour that can occur in NorESM (or CESM) simulations including interactive ice sheet coupling on multiple Greenland and Antarctic ice sheet domain. A 50 years-long test showed that, in some years, the ice runoff around Greenland is negative (see example in the image below, variable rfi (frozen runoff) in BLOM, blue is negative and red is positive).

Below, I am going to summarise a long email chain involving @billsacks, @mvertens, @Katetc, @hgoelzer, @gunterl, @whlipscomb and myself:

1. Why negative ice runoff is showing up: this is caused by dynbal fluxes corrections (see subsection 6.3.1 here https://escomp.github.io/cism-docs/cism-in-cesm/versions/master/html/clm-cism-coupling.html) during (relatively small) Greenland ice sheet margin oscillations affecting a few grid cells. This is consistent with further experiments where (a) SMB in Greenland was artificially set to zero in CISM, leading to no margin oscillation, but comparable freshwater fluxes: in this experiment dynbal fluxes were smaller and ice runoff was always positive, (b) SMB outside the observed Greenland ice sheet mask masked out and set to zero (using the force_retreat option, see Missing entry for 'force_retreat' namelist option in namelist_definition_cism.xml ESCOMP/CISM-wrapper#107), also leading to reduced margin oscillation. Also in this case, dynbal fluxes were smaller and ice runoff always positive.

2. Why negative ice runoff doesn't show up in NorESM (or CESM) fully coupled runs with Greenland-only: when Antarctica is not coupled, snow capping is generating quite a large amount of positive ice discharge (considering there is almost no melt over the Antarctic ice sheet under pre-industrial conditions). Under the Greenland-only coupling configuration, this positive ice discharge generated in CLM by snow capping is used to counterbalance the negative ice runoff over Greenland. In our NorESM experiments with Greenland-only coupling, ice runoff around Greenland always stays positive no matter of the oscillations happening at the ice margin.

3. What to conclude and what to do about this: If one wants to run NorESM (or CESM) coupled to CISM over Greenland and Antarctica masking out SMB outside present-day Greenland, it can be expected that this model behaviour will not be a problem. However, for more physical approaches (i.e., Greenland set free to evolve) negative ice runoff could happen. One code modification that @mvertens will implement in NorESM is to create separate output variables for physical and dynbal ice runoff fluxes also in the ocean component BLOM (variables are already separated in CLM and MOSART). This will facilitate (1) keeping dynbal fluxes under control, (2) diagnosing the correct amount of physical fluxes into the ocean, (3) future development on injecting ice fluxes (e.g., iceberg drifts) which would be done on physical ice fluxes only.