Date: 19 Nov 2025
Attendees: @AndyHoggANU, @anton, @aekiss, @cbull, @ezhilsabareesh8, @jbisits, @MartinDix, @NoahDay, @Paul.Gregory, @sofarrell
Chair: @cbull
Minutes: @dougiesquire
Wave-ice coupling update from Noah
Now have waves making it into the ice. There was an issue with how the coupling was configured - mismatch between the field coming in and what WW3 thought it should be.
Antarctic sea-ice extents are about 1/3 of what they should be (Arctic is less affected). Still working out why.
MOM6 numerical mixing diagnostic
Joey is adding a diagnostic for numerical mixing developed by @janzika into MOM6.
He has an offline version of the diagnostic and has added it online into MOM6. He is trying to run it in the OM3 25km IAF config.
He is noticing that some existing diagnostics that he uses in the calculation of his diagnostic (but shouldn’t modify) have changed after his code modifications. Currently tracking down.
25km IAF release
Discussed Project Board release section
Bathymetry edits are close. Still tweaking a few things in Indo straits.
- Lombok needs to be narrowed - Already done by Ezhil.
- Leti sill needs to let more flow into Timor - Already done by Ezhil, though possibly widened too much, maybe try just 1 or 2 cells.
- Makassar needs to be opened up - Still to do.
Bosphorous needs restricting but is already one cell wide. Do we need a porous barrier? No one has the expertise on how to do this, perhaps it was done at a config we imported but has been turned off. This can probably wait until after the release.
Chlorophyll for SW penetration - Very nearly done by Dougie. Should be done today or tomorrow.
Documentation needs to be reviewed by the ocean team.
@cbull will coordinate config doc review focusing on the “old stub” that @minghangli wrote but never got fully reviewed.
@ezhilsabareesh8 and @aekiss will try to finalise the bathymetry by Friday. We’re aiming for a release early next week.
MOM6 dev meeting
TURBO
John Dennis (NCAR) presented an overview of the TURBO project (something like “Towards ultra high resolution CESM with MOM6 and ocean BGC”). The project is to allow MOM6 to run on GPUs. This is separate from the ACCESS-NRI/GFDL effort to port MOM6 to GPUs.
After some initial scoping of various technologies, have decided to use AMReX:
- C++ productivity framework - framework for block structured AMR (LBNL)
- Supports CPU and GPU
- Assumes logically rectangular grid
- Includes support for IO distributed memory - don’t need to write your own message passing code
- Can mix Fortran and C++ quite easily. Can write individual subroutines in C++ or Fortran, adding OpenACC directives to fortran subroutines
- Good CPU performance - comparable to optimized Fortran and C
- Good GPU performance
Anticipating non-b4b changes:
- Assess with Ensemble Consistency Test - GMD paper (2016) Evaluating Statistical Consistency in the Ocean Model Component
- Want to create ECTs for MOM6. Call for help to develop these tests.
Loose timeline for TURBO:
- Starting with FMS infra layer. Goal is not to provide an complete alternative to FMS, but to replace pieces at a time.
- Within 6 months hoping to have unit tests for the FMS infra rewrite and a prototype of a more-extensive rewrite
- Within 2 years hoping to complete both C++/AMReX-based and Fortran+directive-based implementations so that comparisons can be made.
Plan to collaboration with GFDL/ACCESS-NRI effort as thought partners for algo dev.
ACCESS-NRI/GFDL GPU port update
Dynamic core is running. Seeing encouraging performance numbers.
In particular, finding that loop orderings that are fast for CPU are not good for GPU and vice versa.
Double-gyre test case is now running really well on GPUs. Diagnostics are difficult (GPU→ CPU→disk)
COSIMA/Cryosphere workshop
There will be a combined COSIMA+Cryosphere WG workshop 16th-20th Nov in Hobart
Still working out the venue. Inputs on venues welcome
Sessions on understanding MOM6
There is a plan to start open sessions on understanding the MOM6 codebase, possibly as part of the COSIMA WG meeting. Stay tuned. Get in touch with chris/adele to be involved in the scoping meeting.
Compute Usage
Is anyone planning any large runs in the near future?
Andy would like to run GM/Redi testing on ACCESS-OM3, once the new release is out.
CM3 will also be doing a long run
Storing intermediate products from inputs generation
Some of our workflows for generating configuration inputs require generating intermediate products that can be useful more broadly (e.g. for evaluation). Currently these are being stored in an ad hoc way in ik11.
Agreed to create a /g/data/vk83/data directory as previously suggested by @Aidan.
CICE C-grid
Siobhan asked for an update.
Anton: hasn’t progressed that much. Still have same instability, probably related to ice strength or ridging. Could also be related to coupling - stand-alone case didn’t crash. Still on the todo list. Andy suggested it would be good to have in the full release of the model but how it’s completed is yet to be decided.
new iaf run after the release is out
@cbull suggested it would be best to do the new iaf run (after the release) from a cold start, this will make it more readily comparable with cm3 whom are planning to another run. @MartinDix and @aekiss confirmed.