National Scale Coastal COMPAS configurations

Hi All,

Following up on from @helen’s recent email, I wanted to open up a conversation around how we might align modelling strategies at a national coastal scale with the proposed development of Coastal Modelling Commons and the Super-Regional MOM6 being suggested in the 25/26 ACCESS-NRI workplace.

At CSIRO, we have a set of mature national scale coastal unstructured grid models (using COMPAS) with coarse, intermediate and fine resolutions. These configurations are stable and and will be continually developed over the coming years. We will be progressively including rivers, sediment, carbon and BGC processes. We’d be happy to contribute them to the broader ACCESS-NRI effort—either as part of a downscaling pathway into the proposed super-regional MoM6 model or potentially directly within parent ACCESS-NRI models. We’d like to adopt open-development principals and would invite community members to participate in the development (code through to configurations), if this is of interest?

I’d be keen to start the conversation about how we might submit these configurations as a community contribution?

We see this as a valuable opportunity to build toward future collaborations and start realising further components of the ANCOMS vision in a “no regrets” manner.

I should also point out there are national scale SCHISM configurations and datasets … perhaps @VanessaHernaman you could comment on these configurations?

I’d be keen to hear thoughts from others on how we might best align coastal efforts—particularly in terms of domain boundaries, resolution strategies, and integration with the coastal commons.

Looking forward to your feedback and suggestions!

Cheers,
Emlyn Jones
CSIRO Environment, Hobart

3 Likes

Hi Emlyn,

Thanks for your suggestions – I think it is a good idea to work together to streamline approach for modelling higher-resolution scales. What resolution are the 3 COMPAS models (perhaps report mean and range in the resolution?)? I will describe mine in a post below (about to board a flight so this may be later today).

Aside from the regional domains, there are also plans to start developing an 8km global domain. So the global domains available will be 100 km, 25 km and 8km resolution (with BGC at the coarse resolutions) with the addition of the regional Aus-wide MOM6 domain at 4km. There is still a coverage gap at higher resolutions and I think that sharing some of the regional domains would be a great way to fill this gap (albeit with differing levels of support). We have shared a few MOM configurations at ACCESS-NRI so if we are keen, I can share some ideas on how we can share some community configuration in such a way to make it easier to collaborate across the community.

We are also developing a coastal commons github repository (likely we will put it here )and UNSW (@ckerry, @pcouto, @fsobral) are looking to upload scripts for model/data assessment. Initially the scripts will only work for ROMS but we are looking to make them code agnostic. I was thinking that a good first step for collaboration could be to or add some scripts that could use outputs from one model to force another model (e.g. some offline nesting between the different models). I would love to hear your (or anyone else’s) opinion on these ideas.

2 Likes

The regional MOM6 domain that I am developing is still early stages so there is room the adjust the domain size and coverage to better align with yours and other plans.
I am planning to initially develop a domain at 4 km resolution. I have a picture of the nominal domain below – but boundaries may need to move a bit for stability and run-performance. The idea is that we want to make the domain large enough to include remote tidally generated forcing, so it will include tides. This will make it harder to output at a useful frequency for downscaling. I have some ideas on how we can do this, but I think it may need to be a wider discussion.

The plan for this year is to run the domain for physics-only (inc tides). There are some MOM6 developments occurring at ACCESS-NRI, the wider Australian community and international collaborators (e.g. waves, WOMBAT BGC, Data assimilation). If there is interest and funding, we might be able to incorporate some of these in future versions of the configuration. However, we may need to also consider running a second coarser resolution model if we incorporate too much!

Was there a plan for the COMPAS configurations to include some of the EReefs functionality?

2 Likes

Hi Helen - sounds good - it’ll be interesting to do some experiments nesting into the MOM regional domain to see how the explicit propagation of tides propagates into the hi-res nested model - some early experiments suggested we needed output at fairly high frequency to avoid under-specification of the high-frequency sea level and currents. I’ll send a link to a description of the National Scale Coastal COMPAS configurations. We’ll likely start with the coarse and intermediate resolution grids first, as the very high resolution grid is still being optimised.

Link to the Coastal and Continental Ocean Model of Australia (CoCOMA) project webpage which gives a description of the configurations, links to some example output and configuration files:

The system is being run on NCI, so it’d be good to work with ACCESS-NRI over the coming months to expose the configuration (via the ACCESS-NRI Github if appropriate?), forcing and output.

@frizwi - can provide further info on the technical side.

2 Likes

@Emlyn I think that starting up an ACCESS-NRI community github repository for configuration files would work well. If you like, I can start getting the ball rolling on this.

Over the next few months, I am going to try to get some example output for the MOM6 configuration. I will do a short run with high resolution outputs to start playing with. How frequently did you find you needed the outputs? Would hourly temperature, salinity, 3d velocity and zeta do? Another way to get high frequency outputs (whilst not blowing out storage budgets!) is to output at only the location of the compas boundaries – but the downside here is that we would need to know in advance of any production run where the boundaries are for all ANCOMS models that might want to use this.

We have been playing with setting up a coastal commons github repository – it is early days and very skeleton (and not yet model agnostic) but there is more to come. This repo can be used if needed for adding/developing some scripts for doing the downscaling.

At ACCESS-NRI we have a workflow whereby a user or developer just runs a few lines of code to download a configuration from a githib repository and start running it:

payu clone -b expt -B release-MC_25km_jra_ryf https://github.com/ACCESS-NRI/access-om3-configs 25km_jra_ryf
cd 25km_jra_ryf
payu run

Payu helps to manage version control of experiments and tracking of input + restart files.

Would this be useful? There would be some effort to get this working as you would need to write a compas driver for Payu and compile with spack (I believe that @frizwi has previously compiled compas with spack). @pcouto has been visiting us this week and setting this up for ROMS. He, @jo-basevi and @Aidan managed to setup a very basic payu driver for ROMS in about an hour but time taken to set this up will vary depending on the workflow needed for compas

1 Like