COSIMA TWG Meeting Minutes 2025

COSIMA TWG
, , , , ,
17

Date: 23 July 2025
Chair: @anton
Minutes: @minghangli
Attendees: @AndyHoggANU @cbull @claireyung @sofarrell @dougiesquire @MartinDix @helen @aekiss

Minutes

Repository Strategy for region configurations

Currently there are +50 branches crowding in access-om3-configs. There will be several new panan and WOMBAT runs will add another 3-5 branches soon. Three options were proposed,

  1. create a separate regional-only repo to avoid clutter,
  2. use personal forks for experiment work, then merge back via PR,
  3. keep current repo for dev and release branches only,
  4. main CI wont run on external forks.

The current decision/action is to,

  1. continue to use access-om3-configs, everyone keeps an eye on deleting temporary branches once merged / obsolete.
  2. Most preferred the 2nd option, while @anton to test CI-on-forks feasibility then report next meeting.
  3. @helen will push two WOMBAT branches prefix branch names with helen/, and delete when work is merged.

Freezing temperature consistency between mom6 and cice

In the upcoming beta release of the 25km RYF configuration, MOM now uses TEOS-10 EOS with TEOS-Poly freezing form, while CICE continues to use a linear salinity-pressure relation. There are potential mismatch at the ocean-seaice interface. This might also create spurious sensisble heat fluxes or frazil artefects, especially under ice shelves.

Currently TEOS-10 in MOM6 provides 2 freezing forms, one is TEOS (exact) and the other is TEOS_POLY which is a 23-term polynomial fit (cheaper).

During coupling, MOM checks if the ocean cell is super-cooled. If so, it removes the latent heat of frazil ice and sends it as a flux to CICE, which at the same time, has its own routine to decide what the ice-ocean freezing point is. So if the two routines disagree, there will be a small mismatch.

For CICE, it uses a linear tfrz_option, which gives temperature only as a function of salinity. However for TEOS_poly in MOM, it evaluates the full TEOS-10 polynomial as a function of salinity and pressure and hence changes with pressure as well.

CICE is unaware of ocean pressure unless explictly passed. When near the sea surface, pressure is nearly 0 so pressure terms are negligible but inside deep cavities, pressure can be hundreds of bars which makes TEOS_poly more essential.

Some conclusions:

  1. In short term, make both components use the same simple linear freezing form.
  2. In long term, port the TEOS_POLY routine into CICE so both sides share the same formulation.

Sea-ice Initial condition

@aekiss found a bug in temperature and salinity initialisation, where there’s an unrealistic stripe at start-up. The reason behind is due to the default cice initial condition we are currently using. @anton will test using none instead of default to check if the simulation would crash.

For the 25km RYF beta release, it will retain current default initial ice for reproducibility. An issue will document the bias and the zero-ice spin-ups for research focuses (none initial ice).

COSIMA workshop talk (29 July)

@aekiss is going to extend AMOS talk: include config rationale, preliminary beta metrics, and (if ready) GPU performance notes from @edoyango.

25km RYF configuration beta release

@anton outlined the remaining tasks for the 25km RYF release as follows,

  • finalise citation.cff
  • merge to release branch,
  • update release note on github
  • finish/merge access-hive docs PR
  • make release note public on hive forum
  • make feedback thread public on hive forum.

We agreed to proceed with the current configuration without modifying the layout. The model will be run for 20–30 years, with a current performance of 5 model years per day. @minghangli is working on tuning the layout but these changes will be included in the next version, not the current release.

Last but not the least, a big thans to @anton for his leadership and dedication in driving the release process forward!