ACCESS-OM3 model evaluation team minutes

COSIMA Working Group
,
1

ACCESS-OM3 evaluation paper discussion

  • Global OM3 setup
  • Define priorities and who will be working on each config

Agenda (1 hr) (Chaired by Paul Spence, Notes by @fabiobdias ):
Participant list
@PSpence @MATTENGLAND @hrsdawson @edoddridge @pearseb @adele-morrison @AlbertoMeucci Alexander Babanin @fabiobdias @helen @LaurieM @lgbennetts @NoahDay @sofarrell @wghuneke @JuliaN @AndyHoggANU @MartinDix @ezhilsabareesh8 @anton @minghangli @cbull @dougiesquire @claireyung

  1. Understand NRI perspective (discussion led by Andy/Chris) (5 mins + 5 mins question)

Slides:
2025.05.28_evalpaperdiscussion.pptx (2.7 MB)

  • NRI workload;
  • NRI working plan/priorities
    • global ocean-sea ice model
      • Focused on 25km development
    • Some overlap with regional ocean mode
      • Parameter tuning, optimisation, WOMBAT, waves
  • Next steps:
    • Logistical steps (grid gen., IC, input files etc) - well understood
    • Optimisation/tuning
  • Questions / issues arising:
    • Priority on 25km - need to develop coarser configuration first (driven in part by development needs for full ACCESS coupled model)
    • Climate model would be ideal world - but COSIMA community still needs stand-alone ocean-sea-ice models (e.g. where high res. is needed but would be prohibitive in coupled mode with associated ensemble needs due to intrinsic variability triggered by atmospheric coupling).
    • Ocean-Wave coupled with WW3 was not done before at resolution finer than 25km
  1. Understand COSIMA perspective (discussion led by Adele/Julia) (5 mins + 5 mins questions)
  • 19 people said 8-km is mission critical
  • Ice shelf cavities - top request feature
  • 46 published papers used 0.1deg, 10 papers used 0.25deg.
  • Having an awesome model to do science is what the community wants
    • There’s lots of steps to get there (WOMBAT, waves, ice shelves, etc)
    • Set up priorities (trade off between coupled model development benefits from 25km; vs. needs of COSIMA community for 1/10deg version. Stay with OM2?)
  1. Discuss options for model dev pathway (discussion led by Andrew/Andy) (5 mins + 5 questions)
  • NRI workplan: workflow at 25km is easier to apply to 8km now (while it is still fresh on ppl minds)
    • OM2 evaluation paper - had big effort in the optimisation
    • Runtime issues will only appear when running the 8km model
    • How to organise workplan between 25km and 8km development?

1426×748 543 KB

  • Keep an eye on open ocean polynyas around Antarctica on the 25km
  1. How will process of model evaluation work (e.g role of NRI/Cosima)? (discussion led Andrew/Ed) (5 mins + 5 mins questions)
  • Fundamental challenge in the technical work required to update the cosima recipes:
    • Cosima-cookbook to Catalog-intake
    • MOM5 to MOM6
  • Challenge on transit to OM3 since OM2 already exists (from ECRs)
  • COSIMA vs ACCESS-NRI takeover on the Model Evaluation paper
    • More of a scientific task than technical (push for COSIMA led)
  1. Hypothetical paper ideas (discussion led by Adele/Wilma) (5 mins + 5 questions)

  • CORE model evaluation

    • Exclusive OM3-25km
    • Technical report with doi reference
    • CM3 evaluation (including OM3-25km)
    • Resolution comparison (25km Vs 8km)

  • Exploiting technical advances

    • Fast tracers:OM3-25km with different flavors of WOMBAT (Pearse et al)
    • Ice shelf cavities: OM3-8km (Claire et al)

  • Different resolutions likely to affect primary productivity - an idea for a paper publication

  • Most citations made were to refer to the model configuration (technical aspect)

    • Less citations related to the scientific findings?

Some extra items discussed:

  • OM3-100km development?

  • current CM3 testing and MOM6-CICE-WW3 development is being done at 1deg

  • Useful to speed up WOMBAT/BGC development

  • Tuning b/w different resolutions

  • Unclear if the tuning needs to be redone at a different resolution

  • BGC parameters in WOMBAT definitely needed a re-tuning (given the large-scale physics changes between ESM1.5 and ESM1.6)

  1. Closing summary/action items (discussion led by Paul/Andy) (5 mins)
  • Need to define the priority for the first evaluation paper
2

Hi @cbull - thanks for bringing this together! I am interested and able to help with the polynya/ice production distribution validation side of OM3 if that’s needed.

3

2026-06-25 Notes/minutes

ACCESS-OM3 evaluation paper discussion

Agenda (1 hr) (Chaired by Adele Morrison, Notes by ):
Participant list

1. Very brief overview of existing and planned runs (Andrew, 5 minutes)

2. Discussion of current plan for the paper scope / timeline (Andrew? 10 minutes)

  • two objectives:
    1. a collection of notebooks providing assessment of how configs perform with respect to things that are of interest to community - see GitHub - ACCESS-Community-Hub/access-om3-paper-1: Plotting notebooks and discussion for ACCESS-OM3 model description paper · GitHub
    2. a model announcement paper covering a subset of these configs and assessment metrics, perhaps along the lines of Kiss et al 2020.
  • paper scope:
    • ocean / ice / BGC / waves ?
      • TBD in the future, unclear yet
      • Currently waves are only in a 100km config, with ERA5 and KPP. We don’t (yet) have a config with both waves and BGC.
    • resolution: 100km / 25km / 8km ?
      • COSIMA community has more interest in the 8km, but for the coupled community.
      • The 25km is very valuable to compare along CM3.
      • OM3-25km has some improvements compared to OM2-025, which might make it a more attractive choice.
      • 100km has not been a focus for development (except with waves).
    • CM3 and ESM3 will also have announcement paper(s)
  • people are more motivated to contribute to assessment if it leads to a paper
  • more community interest in 8km than lower resolution

This discussion is unfinished, and will be continued in future meetings leading up to the COSIMA hackathon.

3. Overview of how to contribute (@cbull, 5 minutes)

A non-exhaustive (!) list:

How to help with paper writing

The paper is being written on Overleaf. There’s almost nothing written so far. Please ask @aekiss if you’d like edit access (the link is “view only”).

How to contribute to analysis and discussion of figures

All the below are guidelines, not requirements!

The steps:

  • Look at the experiment table to choose which experiment(s) you would like to focus on; it is recommended you choose the latest experiment, which will have the least issues. The older ones are only useful for comparing the impact of the configuration differences between them.
  • Have a look at the metrics “mega-issue” to determine what metric you’d like to contribute;
  • Get access to evaluation repository (open “user request” issue here, or send @cbull or @aekiss your GitHub username);
  • Modify the template to do your analysis and then upload back to GitHub;
    • the template is useful because it allows to re-run the notebook automatically with different experiments! (by injecting a different value for esm_file).
    • Your code can go in the polished directory or in the sandbox, for unpolished code
  • Share your figures in an issue related to your metric.
    • You can copy-paste figures in the issues as well without uploading the code. Encouraged to upload the code.

Please add yourself to the citation file, contributions that are “used” will lead to paper co-authorship

See the README. Help is available!

Slides:
2026.06.25_OM3_howtocontribute.pptx (1.6 MB)

4. Discussion of how people have been going so far with the mechanisms of contributing - do we need to lower the bar for contributions? (Adele, 5 minutes)

No comments

5. Allocate unclaimed / abandoned issues to people, and create new issues as desired (Adele, 5 minutes)

Some allocations have been made, but there are still issues unassigned, and always scope for more issues that are not currently listed in the mega-thread.

6. Open floor for people to share and discuss figures they have made (all, 15 minutes)

7. Dream list and priority of additional runs (sensitivity tests, 8km etc) that we would like to do by November COSIMA evaluation hackathon. (15 minutes)

4

2026-07-30 Notes/minutes

ACCESS-OM3 evaluation paper discussion

Agenda (1 hr) (Chaired by Adele Morrison, Notes by @cbull. Brownies (!) @wghuneke).
Participant list: @aekiss @adele-morrison @adfraser @helen @dougiesquire @ezhilsabareesh8 @hrsdawson @ZhiLiUNSW @Luwei @wghuneke @claireyung @KZCurtin @cbull Matthew England @janzika @pearseb @sofarrell
@hrsdawson
@JuliaN @NoahDay
@Zijin

Agenda:

@aekiss walked us through recent updates on the GitHub issues. Adele and Andrew have re-organised the mega-issue to be thematic.

GitHub issue links to OM3 evaluation:

Evaluation: mean sea surface temperature bias#9

Looks similar to last time in the latest run MC_25km_jra_iaf+wombatlite-test4-d28e0359. Note that it’s not a true “bias” as initialise from WOA. NH western boundary currents are not separating well so that’s driving strong biases. OM2 had pretty large biases at this resolution as well (noting this is WOA2013, top panel). Pearse suggestion: have you looked at the JRA TAS variable?

Evaluation: mean sea surface salinity bias#10

Latest two runs Evaluation: mean sea surface salinity bias · Issue #10 · ACCESS-Community-Hub/access-om3-paper-1 · GitHub are not looking great and worse than previous runs because we’ve turned down restoring. These runs don’t have any spreading of river run-off.

Comparing to OM2 and previous runs is here. Top panel is OM2 / OM3 has “limitless” restoring (so it’s very strong), OM2 had a relatively modest restoring.

OM3 has higher resolution at the surface and ePBL mixing.

Evaluation: zonally averaged temperature and salinity bias#11

See this figure: looks pretty similar globally to om2. Regional versions of these plots, see sections.

And now plots for the Mediterranean. Bias was sign opposite in OM2, currently have too much flow in Gibraltar. In the far north Atlantic (~OSNAP) large biases, need to dig more. Near A10, thermocline is too deep? Same thing happens in north Pacific (P02) and similar in P04, see this post. Similar issues are being seen in CM3.

Evaluation: Timeseries global averages#3

Latest run has more expected variability than previous runs. Adele: would be useful to add an obs line to this?

Evaluation: Mixed layer depth#38

Latest run is looking better than OM2.

OM2 run and obs (caveat that the metrics are calculated in different ways – reference heights/time sampling differ).

Query from Matt E: did these biases manifest in changes in the AMOC (will follow up later).

Adele: would be helpful to replot with biases from obs?

Evaluation: Meridional transects of mean planetary geostrophic potential vorticity, potential density anomaly and max/min monthly mean MLD#14

Dougie made a plot for an older run. @KZCurtin has been looking at similar things and can share them in the issue.

Evaluation: Mean dynamic sea level map#4

Hasn’t been updated for latest run, still needs a bug fix for the variability plot. Planning on looking at the Pacific/Indian SSH gradient going forward.

Evaluation: Time series of annual mean transport through the Indonesian straits.#16

Skip for now but latest plots do show an improvement.

Overturning

(GitHub issue). Looks like it’s improved in the latest run. Is the bottom cell extending too far north?

The AMOC looks weaker and trending down. It’s rather low – something we’d like to fix? Related to the fresh bias and the mixed layer biases? There’s also some Arctic sea ice bugs (rotation of the winds, pole seam) they’ve been fixed in cm3 but not yet om3?

Matt: strength of GM will affect strength of the overturning and could be useful to know how sensitive this model is.

Adele: worth looking at GFDL bathymetry edits to see if we want to carry across the Atlantic ones?

Jan: strength of restoring/surface salinity can be playing a big role. (0.11m day^-1 currently which is the same as what it was in OM2.)

Could be helpful to plot the restoring flux.

Evaluation: total meridional heat transport#12

Latest result is here. It’s particularly poor at the southern hemisphere low latitudes. Would be interesting to try @rmholmes suggested change in bulk formula.

Time series of running 12-month minimum, mean and maximum sea ice extent and volume#20

Latest run only goes to ~80m max (better than the old 10km problem!) in Greenland. Large values here could be due to polar seam bugs.

Evaluation: Sea ice area / extent#21

@hrsdawson has been comparing cycle 1 with test4. OM3 is generally looking better but their are spatial biases. (Anton has made some plots too)