CSIRO - ACCESS-NRI standup minutes

CMIP7 CMIP7 development
83

24 March 2026

Paul, Anton, Jhan, Rachel, Spencer, Tilo, Claire, Josh, Martin, Siobhan, Pears, Dougie, Aidan, Lachlan

Runs

  • esm driven historical looks good (Atm CO2 looks good)
  • esm picontrol run > 600 years
  • runs paused (on P66) until new quarter
  • Min length for picontrol is 500 years, but generally aim for 1000 years

Nitrogen deposition

  • new version of forcing from upstream
  • Martin suggest we need consistent versions in PIControl and Historical
  • Inclined not to change this, but check for differences betwen old and new version first

Model documentation

  • Need to check on information on mass / velocity grids in ocean
  • Need DOIs for all references, Anton will email Elisabeth, Aidan will email Steve Griffies
  • Could make DOIs in ACCESS-NRI community with author support

Martin and Spencer have been looking at neaters ways to handle ensembles
Paul will look into scenario forcings soon

84

31 March 2026

Romain Anton Paul Dougie Martin Claire Spencer Lachlan Joshua Jhan Manodeep Aidan Rachel Tilo Matt Pearse Matt Siobhan Alex

Ocean diagnostics

  • Two core variables missing (2d daily surface salt and temperature)
  • Also a few other related changes

We save these monthly, and not super confident about use case for daily output

For the pi control - happy enough just to turn on from now

There are some idealised & historical runs which are already finished, these are possibly more relevant for reruns.

For now - well put the change in but not plan to re-run anything.

Atmosphere diagnostics

Some confusion about UM output having multiple time dimensions within one file. Are we saving instantaneous and average values correctly ?

Splitting into one file per variable seems like lots of work (this would be in post-processing). Some UM variables don’t have short names - so it’s hard to know what the naming scheme would be.
Spencer proposes splitting into one file for each time dimensions (as a simpler approach, possibly impacts UM output only ?)

Either approach, or the current output is probably ok for Moppy. So any changes are for future model users and to meet the access-nri data standard.

Josh and Spencer to spend some more time on post-processing (the work would be re-useable for AM3/ESM3) to meet access-nri data standard

Aidan wrote a tool called splitvar which may be a useful reference

CN and CNP

Happy to focus on CNP configs - we can archive CN configurations
It would be good if users / community can run in different modes
Some instructions for this could be good - one switch in UM namelist + some spinup

Payu clone

please test the interactive payu clone

module use /g/data/vk83/prerelease/modules/
module load payu/dev

Model registration

Rachel is progessing - slow on github reviews and unclear how to progress
Every grid + subgrid needs an issue and a review

Martin has piControl with v2 Nitrogen

will check results

85

7 April

Aidan, Anton, Paul, Jhan, Lachlan, Martin, Claire, Rachel, Dougie, Pearse

New control runs with time profiles updated

  • Restarted piControl+concentrations & emissions, from old year 200, and set back to 1. (So have upset the leap year sequenc)

  • Appears to have changed answers over restarts in piControl+emissions ?

  • Jhan to create an issue with exact configs compared

Some candidates:

  • possibly spack 0.22 to 1.1 wasn’t reproducible
  • check determinism over first 4 years - does this appear in 4th year? then it’s due to leap years

Nitrogen v2

No obvious impacts in Rachel’s analysis - we don’t know what we don’t know about it yet though (of course)

We like Nitrogen v1, but possibly this would be inconsistent with forcing in Scenarios - so as there apepars to be no impact, well update piControl+emissions to v2 (and all historical and all scenarios) - Does this make comparison between +emissions and +concentrations harder ?

Jhan will also start a parallel run with piControl+concentrations with Nitrogen v2 - spencer will make a restart for this.

Emissions driven historial

  • Trying to understant initial wood product pool size
  • There is some drift in CO2 in piControl, is this important for choosing start year for ensembles ?

Ensembles: - planned for 10-30 realisations of emissions-driven historical experiments and scenarios

Aidan stresses pushes payu runlogs to github.

See How to preserve reproducibility when applying perturbations in the event of a numerical instability crash - #8 by Aidan for committing when doing a pertubation.

86

14 April

Concentrations driven PI control has done 150 years with v2 nitrogen (after 100 years of Martin’s test run which is treated as spinup). Rachel’s first look at global means suggests things are ok, so this could be taken as official PI control. Other runs should start from year 100 of this control.

Idealised runs are highest priority - need to set up 4xCO2 etc with v2 nitrogen. Alex still looking at wood products for the historical.

There was a bug in weighting of ocean CO2 fluxes to the atmosphere at coastal points. Small but changes sign of the next flux. Has now been fixed but updated model release is delayed by spack build problems. The error does affect an atmosphere tracer field in concentration driven runs, but this is not submitted as a CMIP diagnostic, so those runs are ok.

Need to check that all recent runs are using the updated pool diagnostics.

Nitrogen, ozone and land use is now available for the VL and H scenarios. Paul to ask NCI to download. Need to check nitrogen continuity with the v2 historical.

Rachel: Some progress on model registration. Proper CMIP7 cmorisation depends on this. Romain said currently testing with the CMIP6 DRS.

Josh: progress on modifying um2nc to create single variable files.

87

21st April 2026

ESM1.6

Dougie, Anton, Paul, Martin, Siobhan, Claire, Romain, Rachel, Spencer, Jhan, Josh, Dave Bi, Aidan, Alex, Pearse, Matt

Ocean Age Tracer

  • We’ve lost history in the ocean age tracer in piControl, when changing the restart date to year 1. So tracer has been reset to zero, when in reality should be ~1600 years.

  • Although we have changed the forcings and surface conditions during the spinup, so maybe it’s not accurate to report these really old ages anyway.

  • We don’t think it’s valid to add the old age tracers to the current values (due to mixing etc)

  • Ocean age is a 3d field for CMIP - tentatively looks like High priority

  • We’ve agreed to continue to the existing piControl, we’ll need to fix it for future runs and document (e.g. in the paper) where we can.

  • Plan to not publish the age tracer, because it’s not accurate for the actual ocean state

  • Spencer will make the namelist change so that the tracer continues for runs which spinoff from piControl

Other

  • Land Use forcing for scenarios are available now on Gadi qv56 (Paul / Alex)
  • Rachel will update start dates with CMIP docs
  • Martin will email Manodeep about computer resource use questions from CMIP
  • Rachel is progressing Essential Model Documentation - grids are mostly done, model components are next
  • Workflow to use ILAMB for historical data is available (Romain)
88

28 Apr 2026

Dougie, Anton, Jhan, Rachel, Paul L, Spencer, Tilo, Josh, Siobhan, Claire, Martin

Run summary (Rachel):

esm-piControl looks good

esm-flat10 can use restart199 (200 years)

esm-flat10-zec (zero emissions - might be namelist only) and esm-flat10-cdr (prescribed by input file) can be started soo - these change CO2 emissions

Other idealised experiments are in progress / completed.

0.5xCO2 and 2xCO2 configs have been setup (Spencer)

Scenario forcings (Paul):

  • Paul is working through updating scenario forcings
  • Alex is working on processing land cover change for these scenarios

Historical (Rachel):

AMIP - will follow historical, using restart from historical

Post processing - one variable per file (Josh + Spencer):

  • single field per file changes pipeline are very close
  • BOM has um2nc code which may be useful (we may be saving variable which don’t have short names)
  • Adds 230 iNodes per year for atmosphere
  • hoping to post process the existing data to match, and update payu config to do this automatically.
89

I have a couple questions/comments, which you probably thought about already but:

  1. Did you consider using Anderson Acceleration to spin the age up fast to “catch up”? Or even transport matrices? We do have code for doing that from ESM1.5 outputs and I assume that adapting it to ESM1.6 should not be too hard? :person_shrugging:
  2. Did the age get restored to zero by this clamping issue (maximum age bug ¡ Issue #400 ¡ mom-ocean/MOM5 ¡ GitHub)?