ACCESS OM3 Configurations
ACCESS-NRI has released four beta versions of ACCESS-OM3 global configurations. Of these four releases, two are updates to previously-released physics-only configurations and two are new configurations that include the WOMBATlite ocean BGC model.
All released configurations use ACCESS-OM3 2026.05.002, a global grid with a nominal resolution of 25km, GEBCO 2024-based bathymetry, and initial conditions generated from WOA2023.
The configuration releases are intended to serve two purposes:
- to assess scientific validity of the configuration;
- to seek feedback from Australian ocean, sea ice and climate researchers on model suitability, usability and documentation.
Feedback on these two points can be provided in the general feedback thread.
Updated physics-only configurations
Releases
release-MC_25km_jra_ryf-2.0-beta
release-MC_25km_jra_iaf-2.0-beta
These are the second beta releases of repeat-year forced (RYF) and inter-annually forced (IAF) configurations of ACCESS-OM3.
Key updates relative to the previous 1.0-beta version are:
Updated to recent version of all component models, including a number of bug fixes.
Updated liquid runoff remapping so that runoff is only added to cells that are directly adjacent to land. See here for details.
Added spatial spreading of frozen runoff according to the Mankoff et al. 2025 iceberg melt climatology. See here for details.
Implemented global freshwater balance in the mediator by scaling precipitation such that the sum of the global precipitation, runoff and evaporation is zero, and there are no constraints on net water flux into MOM6. See here for details.
Updated land mask to help address unrealistically thick sea-ice in fjords. The new mask eliminates B-grid non-advective channels in regions where sea-ice forms. See here for details.
Fixed bug in ocean mesh areas at 65°N. See here for details.
Artificially narrowed the width of the Dardanelles and Bosporus to reduce transport through these regions. See here for details.
Updated ocean Mesoscale Eddy Kinetic Energy Gent–McWilliams parameters. See here for details.
Now adjust net global ocean salt restoring to zero by scaling it, rather than applying a global offset. See here for details.
Additional ocean mixing is now applied wherever there is runoff, mixing it down to 40m depth where the ocean is deep enough.
Spatial regridding from the JRA55-do grid to the ice/ocean grid now uses conservative interpolation for fluxes and the patch method for all other fields, replacing the previous bilinear interpolation. See here for details.
Corrected the ocean bottom roughness used by the internal tidal dissipation parameterisation. See here for details.
Corrected the chlorophyll climatology used to calculate short-wave penetration into the ocean. See here for details.
Improved model efficiency and performance.
For a complete list of changes relative to the previously released versions, see:
Control experiments
No control experiments are included with these releases. However, the released WOMBATlite configurations do include control experiments with a physical ocean that is bitwise identical to the physics-only releases (see below).
Expected performance
| Walltime / model year | SU / model year |
|---|---|
| ~3 hr 6 min | ~16.5 kSU |
New WOMBATlite configurations
Releases
release-MC_25km_jra_ryf+wombatlite-1.0-beta
release-MC_25km_jra_iaf+wombatlite-1.0-beta
There is a new WOMBATlite-enabled configuration for each of the physics-only configurations described above (i.e. global 25km with both repeat-year and inter-annual JRA55-do forcing). These configurations follow the same naming scheme as the physics-only configurations, but with +wombatlite appended. E.g. the RYF configuration release branch is called release-MC_25km_jra_ryf+wombatlite.
Key features are:
New WOMBATlite model, version 2026.05.000. See the WOMBATlite 2026.05.000 documentation for features and details.
Based on the 2.0-beta physics-only configurations.
Control experiments
Control experiments using these configurations are included in the experiment list here. The experiment names are:
MC-25km_jra_ryf+wombatlite-test3-f4d79e82: 31 years of RYF dataMC_25km_jra_iaf+wombatlite-test4-d28e0359: one complete IAF cycle (1958-2023).
Note the IAF experiment was originally started with a tight limit on the strength of ocean surface salinity restoring (MAX_DELTA_SRESTORE = 0.5 [PSU]). However, this led to crashes and was subsequently relaxed, initially to MAX_DELTA_SRESTORE = 1.0 and finally to the value in the corresponding released configuration (MAX_DELTA_SRESTORE = 2.0).
A Jupyter notebook showing how to find/load the data using intake can be found here.
Expected performance
| Walltime / model year | SU / model year |
|---|---|
| ~3 hrs 50 min | ~20.5 kSU |
Known limitations
These configurations are still undergoing scientific evaluation.
Known limitations include:
There are unresolved surface salinity biases, particularly in the Arctic and North Atlantic Ocean and at large river mouths.
Liquid runoff is not spread horizontally. See here for more details.
Land mask requires some improvements to avoid ice buildup. See here for more details.
Latent heat due to basal melting is not accounted for. See here for more details.
There are two known problems with the new frozen runoff spreading scheme:
- Frozen runoff has a spurious column of zeros at −180° longitude. See here for more details.
- Frozen runoff has an unphysical seasonal cycle around Antarctica. See here for more details.
CICE6 still uses a B-grid. See here for more details.
Sea surface salinity restoring is incorrect along top and bottom rows of the grid. See here for more details.
Shortwave radiation forcing shows sawtoothing in its global sum due to issues with the “coszen” time interpolation in CDEPS. See here for more details.
Ice fields show unphysical patterns at and near the tripolar fold. See here for more details.
Outputting MOM6 diagnostics on density coordinates add ~30 min per model year to the runtime. These diagnostics can be removed if not needed (see here).
WOMBATlite diagnostic output can only be output on the native MOM6 vertical coordinate. See here for more details.
How to use
The released configurations can be run using the payu experiment manager. For instructions, see the access-hive Run ACCESS-OM3 documentation.
What we expect you’ll be able to do with these releases:
- Clone the config repo and run the configuration;
- Look at output from the model;
- Test sensitivity to model parameter choices;
- Discover how the model is configured through ACCESS-OM3 configuration documentation;
- Contribute to the ongoing ACCESS-OM3 development and evaluation effort by sharing or improving notebooks in the ACCESS-Community-Hub repository.
Importantly, while this configuration is considered stable in the software sense, the parameter choices, input files and software code will continue to evolve and improve for some time into the future. There will be newer versions of this configuration which change and improve the model’s results. Users should note that results and analysis using these configurations relate to the released versions only. Future releases will produce different results.
Credits
The ACCESS-OM3 software is built upon shared components of climate models, developed and supported across many organisations globally. This includes but is not limited to contributors and member organisations of the MOM consortium, the CICE consortium and the Earth System Modelling Framework organization. This configuration is developed from international and local expertise.
See the CITATION.cff file included with each configuration for a full list of contributors. Contributors to the configuration versions in this release announcement include:
Configuration development, testing and advice: @anton, @aekiss, @dougiesquire, @minghangli, @pearseb, @ezhilsabareesh8, @AndyHoggANU, @cbull, @adele-morrison, @Callum_Shakespeare, @Luwei, @sofarrell
Model development: @anton, @dougiesquire, @pearseb, @ezhilsabareesh8
Build infrastructure: @TommyGatti, @harshula, @anton, @dougiesquire
Documentation: @anton, @dougiesquire, @minghangli, @JuliaN, @cbull, @ezhilsabareesh8
Performance and optimization: @minghangli, @manodeep
Evaluation: @aekiss, @cbull, @AndyHoggANU, @ezhilsabareesh8, @pearseb, @minghangli, @dougiesquire, @anton, @claireyung, @wghuneke, @hrsdawson, @jemmajeffree, @Helen_Shea, @CharlesTurner, @NoahDay
Support
Replies to this topic are disabled. For feedback, use the General Feedback Thread.
For bugs/support requests, Create a topic on the ACCESS-Hive Forum in COSIMA and tag it with help. Support is provided by ACCESS-NRI for products it releases.