There are a few people developing their own regional MOM6 configurations and we (@mmr0 and I) thought it would be nice for us all to keep in regular contact for mutual support and advice (see below for details). If you are interested in joining us, can you please reply to this post (or message me) and fill out the poll to indicate availability.
Regional MOM(6)’s support group.
Why: Is your small model cranky? Throwing tantrums? Waking (crashing) often? Need constant minding? Are your friends sick of hearing about your model’s inputs and outputs?
What: A support group for asking questions, sharing information, and supporting one another!
When: Once a month for an hour at a day and time yet to be chosen. Vote here for your availability. Poll closes Wed 1st October
Format: Casual. There will be space to bring your own questions, issues and updates
Who: Anyone who is running regional MOM6 configurations. Anyone who is about to start running regional MOM6 configurations. Anyone who wishes they were running a regional MOM6 configuration but can’t because of cryptic error message.
Anyone interested in regional MOM6 is welcome to attend.
Meetings will then be the once every 4 weeks with a break over Dec/Jan.
For the first meeting, please come prepared to give a short (1-3 mins, slides optional) introduction on what you are working on, or are planning to work on. Don’t feel you need to spend too long preparing for this –off the cuff will work.
The current agenda is:
The support group aims (Madi and Helen)
Introduction to everyone’s project (everyone)
Regional MOM6 support structure and future plans (Madi and Helen)
ACCESS-NRI update (Helen)
Any issues/successes to share?
You are welcome to add agenda items – or to bring along a regional MOM6 related issue, question or success. Post here or email me if you want to add something
Structure to seek support, discuss experimental design, bugs etc
Awareness of experiments in progress & completed to enable collaboration
Roundtable - who are we / what are we working on (All)
@mmr0 - Southern Ocean / Tasman sea domains with a focus on resolution (10, 3, and 1km) & how it impacts the representation of surface ocean processes & air-sea fluxes.
@claireyung - Pan-Antarctic regional with ice shelves. Current focus on validation & tuning.
@Lizzie - Australian North West Shelf (NWS) domain encompassing parts of Indonesia, and including WOMBAT to look at the effect of tides on biogeochemistry
@janjaap.meijer - [planned] Meander region south of Aus/NZ. Regional model needed to resolve submesoscales & output high-frequency diagnostics not available in existing products
@cbull - Current ACCESS-NRI Ocean team lead, but also long-time regional modeller enthusiast using NEMO
@dougiesquire - ACCESS-NRI Ocean team, has been helping develop pan-An and NWS domains
Discussion points
Although we didn’t have too much time, several discussion points emerged from the roundtable.
Budgets and conservation - Claire is interested in heat/freshwater/volume budgets, mainly as a bug-finding tool. How to calculate for regional models? What extra diagnostics might be needed at the boundary (or for a sub-region?)?
Spin-up/equilibration times - Some discussion around what this means for regional models, which diagnostics to look at (e.g., KE and EKE), how this is affected by the change in resolution from forcing dataset → regional model, and how this is affected by flow conditions (e.g., coherent jet → chaotic eddy field) [madi - suggest this could be a good topic for a future meeting?]
Attendees shared some advice they had learned over the last month
Asking for too many CPUs (ncpus) can make the model run slowly. If the model is running slow then a simple optimisation test is to run the same short experiment on differing number of CPUs and compare the time taken.
If dye/passive tracers appear to leave out one boundary and then reappear in another then you need to explicitly set (in MOM_input) REENTRANT_X = False.
Auto tiling versus manual tiling
Auto tiling (where MOM6 automatically works out the layout of the model domain onto the requested CPUs) is easy to setup but some speedup can be gained if one manually optimises the tiling. Is it worth it? Suggestions were that it is not worth it in simple domains where there is only a single large land mass and only a few experiments being run off the domain.
Collaborations
@Lizzie knows of others who are running regional mom6 but may not have received invites to the meeting so will forward the invites
Discussions suggest that there are a few different groups who may be interested in developing a Southern Ocean domain. @janjaap.meijer to get in touch with them to see if they can coordinate efforts.
We need to have better visibility on who is working in the regional space to build collaborations, and it was suggested that it would be useful to compile a public list where we can share plans on models developed, under development or planned. Working groups such as COSIMA already have a concept for sharing plans – but this doesn’t have good visibility outside of COSIMA. @helen to start this conversation with the wider communities early next year.
ACCESS-NRI is developing a pathway whereby they can support the ongoing managements of a domain created by a researcher if there is enough community interest.
Sentonix
There is some community interest in getting models running on the Pawsey infrastructure. ACCESS-NRI is investigating and has gotten part of the way towards running an ACCESS-OM3 configuration on Sentonix. One issue is that files are deleted after 30 days so scripts are needed to move files on and off of scratch. Anyone interested in this problem is welcome to get involved, get in touch with @helen or @cbull.
Next meeting
Early Feb in 2026 works for everyone – date to be confirmed early next year.
@ashjbarnes has been developing an automated workflow to setup a nuopc/access-om3 based configuration. It run through the regional mom6 toolbox and from a user perspective, the notebook will be similar to exisiting regional mom6 notebooks. The regional mom6 toolbox will still retain the ability to create an FMS executable. The number of PEs used by the resulting configuration is hardwired and creating very small domains will produce an error. There is another pull request that needs to go through on the regional mom6 toolbox before this can be available, but early access can be provided upon request to @ashjbarnes
ERA5 winds
@mmr0 sees ERA5 wind forcing as useful for applications such as getting consistency between personal simulations and comparator, particularly when the boundary conditions come from a ERA5 forced simulation
At the moment ERA5 wind forcing is only available on the FMS executable. Work was started on implementing this through the NUOPC coupler last year, but this needed to be put on hold due to other priorities.
EAC eddies in regional models
@ashjbarnes noted that tides seem to have an impact on the direction on EAC eddies near Tasmania. Without tides, the eddies appear to travel too far south – the mechanism for this is uncertain.
Issues that came up
The SUs required to run a high-resolution domain means that rerunning a simulation can be cost prohibitive. Especially an issue where a new feature + executable is needed partway through a simulation which can create restart reproducibility issues.
Outputting subsets of data when the subset spans multiple PEs has been creating issues for a few people - the cause of the issue is not well understood but @Lizzie and @mmr0 have examples that work which could be shared.
Our next meeting is on Monday 16th at 11:30 am
The only agenda item is an around the room updates, successes and issues so please come ready to with items you wish to share / discuss.
Hi Helen, Madi and I won’t be able to go as this clashes with the ModSci meeting with the 21CW. Normally it’s on the opposite fortnight but it was pushed forward due to the public holiday last week!