ACCESS-OM3 development highlights/updates

COSIMA Working Group
1

These are presented fortnightly at COSIMA meetings.

First, some context

[This is a wiki post - free free to edit.]

2

2026-02-12

Working towards next IAF 25km test configuration and 66-year run

Also working on (maybe not ready for the next run)

Call out for missing diagnostics:

3

2026-03-05

Continuing work towards next ACCESS-OM3 IAF 25km test configuration and 66-year run:

Request diagnostics for new run:

Other ACCESS-OM3 development

ACCESS-OM2

  • new release soon: WOMBAT-lite and GCC compiler support
5

2026-04-02

ACCESS-OM3 evaluation update

6

2026-04-16

3rd ACCESS-OM3 25km IAF 1958-2023 test run

  • currently underway; details here
  • currently simulating 2008; should be finished (to end of 2023) by early next week
  • data: /g/data/ol01/outputs/access-om3-25km/MC_25km_jra_iaf+wombatlite-test3v2-00532b88

many of the problems highlighted in the last update are now resolved or much improved in the new run :tada:

  • reverted to previous MOM6 and CICE6 code to sidestep still-unresolved mediator NaN bug
  • SSS restoring limited by MAX_DELTA_SRESTORE = 1.0 psu (rather than 0.5 psu, to avoid error)
  • no crashes so far :crossed_fingers:
  • B-grid compatible land mask in polar regions has prevented unbounded ice and snow thickness growth. Thickest ice is <50m thick rather than 10km. Still thicker than OM2, but this would probably be reduced further by spreading solid runoff and associated latent cooling of the ocean.
  • Improved GM parameters have significantly reduced abyssal temperature drift
  • Indonesian throughflow has been improved by topography edits
  • Constricting Bosporus has greatly reduced Black Sea and eastern Mediterranean SSS bias
  • WOMBAT-lite BGC is looking good - @pearseb tells me
    • Chlorophyll looks amazing
    • CO2 flux and iron limitation look great
    • Surface nitrate looks great, but surface iron is too high in coastal areas. Not sure why it’s not being removed rapidly at those concentrations.
    • Vert integrated NPP much too high in a weird place in the Amazon outflow. Otherwise it looks great. However, the weird high values are causing globally integrated rates of NPP that are 3x higher than the median estimate from observations. And much higher than what is considered reasonable.
    • Zooplankton is too low.
7

2026-04-30

3rd ACCESS-OM3 25km IAF 1958-2023 test run

  • 66-year IAF cycle completed
  • includes WOMBAT-lite BGC
  • details here
  • data here: /g/data/ol01/outputs/access-om3-25km/MC_25km_jra_iaf+wombatlite-test3v2-00532b88
  • we welcome analysis/scrutiny of this data - especially the BGC output

Some current ACCESS-OM3 development issues

8

2026-05-14

Good news :tada:

9

2026-05-28

10

2026-06-11

11

2026-06-24

12

2026-07-09

13

2026-07-23

14

2026-08-06

  • 100km MOM6-CICE6-WW3 OM3 nearing release
  • 100km MOM6-CICE6-WW3 OM3 load balancing complete
  • 25km + 100km OM3 performance optimisation underway - gains of ~10% with MOM6 code changes and compiler directives
  • implementation of river runoff spreading nearing completion
  • code changes underway (splitting JRA55 runoff into river and basal melt) to allow basal melt to be input at depth, also accounting for latent heat
  • 25km OM3 test4 assessment (all the evaluation for the latest ACCESS-OM3-25km test4 run here) highlights areas that need improvement (details here)
    • thermocline too deep,
    • weak southern hemisphere heat transport,
    • SSS too fresh,
    • poor Kuroshio and Gulf Stream separation and variability distribution,
    • incorrect North Atlantic deep convection locations,
    • weak and declining NADW,
    • excessive Mediterranean outflow.
15

2026-08-20

  • river runoff spreading implemented
  • basal melt latent heat being implemented, including distribution of meltwater/cooling at depth; distribution of frazil brine will be considered as a later step.
  • 25km OM3 test4 assessment continuing - please join in to ensure that the model will work well for your areas of interest!
  • 25km OM3 test5 in preparation
    • C-grid CICE
    • horizontal runoff spreading
    • basal melt latent heat
    • numerical mixing diagnostic
    • bathymetry fixes
    • tripole fixes
    • code updates, bug fixes, performance optimisations, etc.
  • WOMBATmid code and documentation complete; parameter optimisation underway
  • Method for cycling OM3 runs under development
  • MCW 100 km ERA5 config: aiming for a beta release; long-run underway, to be presented at AARC next week (slide).
16

2026-09-03

17

2026-09-14

Development

Evaluation

3 evaluation tasks open to being picked up:

  • Add lines for ACCESS-OM2-01 (IAF cycle 1) to time series of global averages: temperature, salinity, sea surface temperature, sea surface salinity. Modify code here. GH Issue
  • Arctic sea ice analysis (see sea ice plots below mentioning Arctic).
  • Arctic transects in Transects.ipynb to explore the vertical extent of the SSS biases. Just use grid-aligned transects (which won’t be zonal or meridional, due to tripolar coordinates).

Feel free to assign yourself in this post and ask @aekiss or @cbull for access to repository (you’ll need to do this first).

18

2026-10-01

Development

ACCESS-OM3

  • 8km RYF config dev-MC_8km_jra_ryf is now a dev branch :tada:
  • MCW 100km ERA5 IAF config (dev-MCW_100km_era_iaf) and docs being prepared for beta release, currently not reproducible with c-grid CICE.
  • 100km MC_100km_jra_iaf+wombatlite test 1 underway
  • new ACCESS-OM3 release with MOM6 and WOMBATlite bug fixes and new features
  • MOM6 fix to correct the mass of ocean (masso) diagnostic by excluding masked land, making it independent of processor decomposition
  • regional-mom6 support for ACCESS-OM3 announced
  • Spack develop help page being updated to focus on ACCESS-OM3. Call for testers!

ACCESS-OM2

  • full suite of configurations released, including new ocean/ice diagnostics and WOMBATlite bug fixes
    • includes @rmholmes et al. mixed layer heat budget diagnostics
  • 2000-year 1° spin-up with WOMBATlite restarted

WOMBAT

  • first version of WOMBATfull released

Evaluation

Next evaluation meeting: 5 November 2026. Note

Call for Expressions of Interest (EoI) in ACCESS-ESM3 Model Evaluation & Development. Due: 31 October 2026
LINK.

Merged PRs (hooray!):

  • #139 Update MHT evaluation notebook (Claire Yung, 29 Sep). Splits meridional heat transport into the Atlantic and Indo-Pacific basins and adds observations for each basin, including Macdonald & Baringer 2013 values read off the plots in Tsujino et al. 2020.
  • #131 Manually calculated mixed layer depths (Katja Curtin, 27 Sep). Adds notebooks for mixed layer depths calculated by hand.
  • #130 Add floe size distribution analysis (Noah Day, 17 Sep)

Outstanding tasks for anyone interested:

  • Arctic sea ice analysis (see sea ice plots below mentioning Arctic).
  • Arctic transects in Transects.ipynb to explore the vertical extent of the SSS biases. GH issue
  • Do some research on what other modelling centres have done to fix their low AMOCs. See related issue.
  • AMOC bias: zonal mean latitude-depth section of model biases in Atlantic T and S (suggested here).

Assign yourself here or get in touch with @cbull / @aekiss

19

2026-10-15

Development

ACCESS-OM3

  • 8km RYF config dev-MC_8km_jra_ryf test 2 underway with reduced Arctic resolution in OM3-8km - improved stability, allowing dt=400s, increasing run speed and decreasing SU cost
    • C-grid CICE is currently the main performance bottleneck due to uneven ice workloads across MPI ranks and frequent MPI communication. Initial tests with optimised block sizes and more cores show ~6% faster runtime and ~3% lower SU cost, bringing the projected cost to ~235 kSU/year. This is still ~45% higher than ACCESS-OM2 0.1° (162 kSU/year), although the configurations are not directly comparable.
  • MC_100km_jra_iaf+wombatlite test 1 completed. Analysis is underway, with a few minor issues in the evaluation notebooks being resolved before the dataset is published.
  • New option under development to output CICE6 diagnostics as a separate file for each variable
  • 25km IAF test5.1 to start soon
  • First ACCESS-NRI pull request to shared MOM6 repo ACCESS-NRI to main (2026-10) by dougiesquire · Pull Request #100 · ACCESS-NRI/MOM6 · GitHub

ACCESS-CM2 data availability

  • Some ACCESS-CM2 datasets (e.g., cj877, by647) are no longer accessible under /g/data/p73/archive/non-CMIP, causing catalogue errors. Data may have been moved or removed. See issue #693.

Other

  • cosima.org.au website infrastructure upgrade - may be unavailable for a day or so this week or next. Please don’t edit the website until the upgrade is completed.
20

2026-10-29

ACCESS-OM3

Development

Evaluation

Next evaluation meeting is next week, 4pm 5th Nov. Lots of things to do - assign yourself here or get in touch with @cbull / @aekiss if you’d like to contribute.

Running all the notebooks (mkfigs.sh) on: MC_8km_jra_ryf-test2-expt-c349f800, MC_100km_era_iaf-test1-1d86aa55 and MC_100km_jra_iaf+wombatlite-test1-e1de2a94 is currently underway ahead of the hackathon.

ol01 storage running out..

Anyone have any ol01 datasets they want to publish or give a more permanent home to? Please get in touch with Chris if you have any candidates for the Datapool or Curated collections (e.g. cj50).