preindustrial+concentrations restart: don't change year

[blimlim]

1. Summary:
Following further discussion with @JhanSrbinovsky and @rml599gh, we would like to use the same restart from #699 but without the change to the restart year.

This PR updates the restart to use a new copy, without the date change.

2. Issues Addressed:

• Change preindustrial+concentrations restart #698

3. Dependencies (e.g. on payu, or model)

This change requires changes to (note pull request(s) where relevant):

• workflow manager (payu):
• model deployment (ACCESS-ESM1.6):
• model component or library dependency:
• input workflow:

4. Ad-hoc Testing

What ad-hoc testing was done? How are you convinced this change is correct (plots are good)?

5. CI Testing

• !test repro has been run

6. Reproducibility

Is this reproducible with the previous commit? (If not, why not?)

• Yes
• No - !test repro commit has been run.

7. Performance

Has the model performance (say, throughput of model-years/wall-day) changed?

• Yes
• No
• N/A (if selected, please add a brief explanation why performance testing is not necessary for this PR)

If yes, provide the numbers from your testing. Is the performance better or worse?

8. Manifests

Have you changed the executable, the input files and/or the restart files?

• Yes
• No

If yes, have you updated the manifests?

• Yes
• No

To update the manifests, run payu setup (in a cloned copy of your feature branch) with reproducibility tests turned off:

manifest:
  reproduce:
    exe: false
    input: false
    restart: false
runlog:
  enable: false

Then commit the newly created manifest files (under manifests/) only to the branch for this PR.

9. Documentation

Is the documentation updated?

• Yes
• N/A

10. Merge Strategy

• Merge commit
• Rebase and merge
• Squash

[blimlim]

!test repro

[github-actions]

❌ The Bitwise Reproducibility Check Failed ❌

When comparing:

• 698-update-pi-restart (checksums created using commit 5f5a698), against
• dev-preindustrial+concentrations (checksums in commit aa8e59a)

Further information

The experiment can be found on Gadi at /scratch/tm70/repro-ci/experiments/access-esm1.6-configs/pr700/698-update-pi-restart/5f5a69827e072e19d37026d57cacd4ec3e8ef42c, and the test results at https://github.com/ACCESS-NRI/access-esm1.6-configs/runs/71389841602.

The checksums generated by this !test command are found in the testing/checksum directory of https://github.com/ACCESS-NRI/access-esm1.6-configs/actions/runs/24435604462/artifacts/6442849560.

The checksums compared against are found here https://github.com/ACCESS-NRI/access-esm1.6-configs/tree/aa8e59a579adc78ab433c6b5d21ea03aebabeaf2/testing/checksum

Test summary:
❌ test_repro_historical
✅ test_repro_determinism

[blimlim]

❌ The Bitwise Reproducibility Check Failed ❌

Not expected. I'll try to see what happened

[JhanSrbinovsky]

Thanks Spencer

[JhanSrbinovsky]

what happened this was just the start date right? not even that, just a year number.

[blimlim]

Yeah, only the restart year has changed which is not expected to change results. The checksums are identical for all fields other than age_global:

diff -c 2 checksum1 checksum2

      ],
      "age_global": [
!       "-948087343509176042"
      ],
      "frazil": [
--- 218,222 ----
      ],
      "age_global": [
!       "-8010627089325603220"
      ],
      "frazil": [

[blimlim]

The following shows the column mean ocean age tracer for 1 month runs starting from each restart:

Non year 1 start:

Year 1 start:

MOM appears to completely reset the age tracer when starting from year 1. This might be something we don't want to ever happen, I'll make a separate issue for discussion.

[blimlim]

!test repro commit

[rml599gh]

I'm thinking that whether or not we change the date for this configuration depends on what we want it to represent if we are keeping the picontrol that does have the restart date change in.

[github-actions]

❌ The Bitwise Reproducibility Check Failed ❌

When comparing:

• 698-update-pi-restart (checksums created using commit 5f5a698), against
• dev-preindustrial+concentrations (checksums in commit aa8e59a)

🔧 The new checksums will be committed to this PR, if they differ from what is on this branch.

Further information

The experiment can be found on Gadi at /scratch/tm70/repro-ci/experiments/access-esm1.6-configs/pr700/698-update-pi-restart/5f5a69827e072e19d37026d57cacd4ec3e8ef42c, and the test results at https://github.com/ACCESS-NRI/access-esm1.6-configs/runs/72234131752.

The checksums generated by this !test command are found in the testing/checksum directory of https://github.com/ACCESS-NRI/access-esm1.6-configs/actions/runs/24697512767/artifacts/6544681374.

The checksums compared against are found here https://github.com/ACCESS-NRI/access-esm1.6-configs/tree/aa8e59a579adc78ab433c6b5d21ea03aebabeaf2/testing/checksum

Test summary:
❌ test_repro_historical
✅ test_repro_determinism
✅ test_repro_payu_setup

[blimlim]

@JhanSrbinovsky, @rml599gh, this PR swaps the restart for the preindustrial+concentrations run to a copy of /g/data/p73/archive/CMIP7/ACCESS-ESM1-6/production/Ndep2-PI-CNP-concentrations/restart170 without the date change. If you are happy to review this that would be great!

I'll add the MOM namelist change to a separate PR which will let the change be copied to all the configurations

[JhanSrbinovsky]

phew, I'm glad my memory wasn't playing tricks on me: That was the restart we used before from:

/g/data/vk83/prerelease/configurations/inputs/access-esm1p6/modern/pre-industrial/restart/2026.04.15/README 

Can you clarify, at the start of the ocean age tracer discussion - there was a comment that I read as "even though the age tracer changes, reproducibility shattered, all other physical characteristics are the same." Was that not the case?

[rml599gh]

OK - I think I've got this straight in my mind now. This takes the end of Jhan's picontrol run and puts it into the dev config instead of the test one - which is what we want.

[blimlim]

Thanks @rml599gh and @JhanSrbinovsky

"even though the age tracer changes, reproducibility shattered, all other physical characteristics are the same." Was that not the case?

Yeah that's correct – the ocean age tracer variable changed, but all other aspects of the physical state in the ocean, atmosphere and sea ice weren't affected.

This takes the end of Jhan's picontrol run and puts it into the dev config instead of the test one - which is what we want.

That's right, It's taking the restart from the end of Jhan's run and putting it into the dev config