CSIRO - ACCESS-NRI standup minutes

CMIP7 CMIP7 development
33

Meeting minutes

Date: 22/04/2025
Attended: @rbeucher, @MartinDix, @tiloz, @RachelLaw, @dougiesquire, @spencerwong, @paulleopardi, @dave, @Jhan

Cable workflow confusion

Rachel: stash c with esm1.6 wanting to add some extra fields: soil respiration on tiles. But why were there missing variables? Martin: currently comes from different files (https://github.com/ACCESS-NRI/access-esm1.6-configs/blob/dev-preindustrial%2Bconcentrations/atmosphere/prefix.PRESM_A). Would be good to consolidate a workflow into a single file.

New cmip7 forcings

Rachel: Might want to try 4x CO2 and 1% runs. Martin: should be straightforward. Paul: won’t be the official CMIP7 ones because we don’t have all the relevant files yet but can still give us an idea.

Chat update from Paul:

Only 7 out of 12 forcing datasets are currently available for CMIP7: Summary - input4MIPs Controlled Vocabularies (CVs) Even the testing phase only has 8 out of 12. The good news is that at least two forcings are not used by ESM1.6.

Mediterranean getting more saline like CM2?

Martin: Ian has highlighted some salinity issues, Mediterranean was getting more saline (ACCESS-ESM1.6 development - #183 by inh599) but CM2 was worse (~50 sos). Martin: suggested it remained a problem in CM2 but didn’t seem to create problems elsewhere so wasn’t fixed. If there was a known fix here it would be worth rolling in but perhaps not worth explicitly investigating. Arnold is not available to look this week but Tilo is having a look at a run he shared (not a totally clean comparison due to other changes).

Concern over Spack build using a branch to build from rather than a specific commit/tag. UM changes were rolled in when?

@Jhan: raised concern over Spack build using a branch to build from rather than a specific commit/tag?
@spencerwong: yes, that did occur but has since been updated to use specific tags/commits.

Note that the build system allows one to see the exact commit that was used for a PR. Example

image
image2059×946 87.5 KB

Chris: emphasized the importance of documenting experiments to mitigate this kind of confusion, especially those used for production runs in July.

Runlogs which are uploaded to the experiments repository also include a git history of the executables, restarts, and input files used during the experiments in the manifests directory.. This can be useful for finding what was used in a particular experiment, long after it has been run. For example, the git history for manifests/exe.yaml file for the first spinup experiment in the experiments repository shows that the same UM executable was used in all the years that have been uploaded to the repository.

Action items:

  • Please update this table and push relevant experiment information here – runlog needs to be set to True. Further details [0].
  • @dougiesquire to add details about how to find the file that tells you if/when the executable was changed. (exe manifest?)
  • @Jhan to check if he has missed pushing a repo, we looked at this page and he thought it looked like it was missing.
  • @pearseb (others?) needs to update the experiment table and push the updated runlog to this branch. Note this discussion for our suggestions of what to include in the table.
  • @spencerwong to put details here as to when the change from building from branches to tags/hashes occured:
    • The change to building from tags/hashes rather than branches occurred on February 10 for test PRs. This change was brought into the main branch on March 12.

Chat update from Paul:

From a branch you can work out a “no later than” date. This one is 22 January 2025
ACCESS-ESM1.6/spack.yaml at c214308600c3f545d48fc7076c7491eddbe1460f · ACCESS-NRI/ACCESS-ESM1.6 · GitHub
The last UM7 commit before 22 Jan was https://github.com/ACCESS-NRI/UM7/commit/1089e256dcec8ad3c900c294661b2cd5fef22565

land use change

Spencer: land use change ancillaries. Rachel: nothing to update at the moment.

cice5 updates

Dave: cice5 in place of cice4 is technically working in ESM1.6. Anton has been doing some tests, the results are not currently good enough to be integrated. Tried different initial condition strategies. Dave is still doing tests in esm1.5 but has other problems. Still needs input from Anton and Siobhan sea ice expertise to tune parameters. Anton is on leave this week.

Romain: no updates.

–

[0]
At a past ESM 1.6 standup [1], we showed how we’ve come up with a workflow for storing people’s experiments for provenance. The idea is that this will support the spin up documentation provided here.

In short, we’re asking if you could please have a look at the instructions here:

https://github.com/ACCESS-Community-Hub/access-esm1.6-dev-experiments

Then push your esm1.6 experiment run folder to that repo with this branch naming convention: -<dev/test>--

Then add a link to the relevant experiment commits here:

https://forum.access-hive.org.au/t/esm1-6-spin-up-experiments/4219

Experiment Documentation