CSIRO - ACCESS-NRI standup minutes

CMIP7 CMIP7 development
62

Meeting Notes 07/11/2025

Participants: @anton @MartinDix @sofarrell @clairecarouge @pearseb @Jhan @Aidan @alexnorton @lachlanswhyborn @manodeep @paulleopardi @rbeucher @spencerwong @tiloz @RachelLaw @inh599

Chair: @clairecarouge

OASIS3-MCT version

Question from Manodeep whether it is ok to pick up a newer version of oasis3-mct for ESM1.6. Decided updated versions of dependencies is part of the optimisation work and there is no objections as long as the results do not change (tests suggest results will be identical).

DOI for models

Aidan: we publish our models on Zenodo to get a DOI. There is a draft Zenodo record. We would like input from the community, especially around proper acknowledgments. Please add any comment to the GitHub issue.

Frozen model

Build

Lachlan is working on on incorporating the latest UM and CABLE changes to the build. Will be merged in soon.

Anton working on CICE changes. To be incorporated when ready.

Configurations

PI-control

Temperature drift in the ocean: Martin did some comparisons with WOA23 dataset. Results show we are not worst than ESM1.5 compared to WOA23. Siobhan commented the ocean is globally cooler in ESM1.6 than ESM1.5 so globally, we have a less warm bias at the top of the ocean and a cooler bias at the bottom.

Decision was made at the latest NESP meeting to use “October B” test for the spinup. The tests Oct-A and Oct-C are now stopped.
Spencer to update the dev-preindustrial+concentrations with the October B configuration plus volcanic input changes.

Pearse is still working out the best sinking rates for WOMBAT. These will be integrated to the configuration when ready.

The land folks are testing a new vegetation distribution (to solve the “disappearance” of dust over Australia) and updated vegetation parameters for the Australian PFTs. To be integrated when ready.

Paul is working on the nitrogen and ozone forcings. To be integrated when ready.

AMIP

Spencer will update the AMIP configuration to correspond to the updated dev-preindustrial+concentrations.

Historical

Rachel asked about the status of this configuration. Martin says it’s best to redevelop from the updated PI-control and AMIP.

Outputs

Paul noticed the surface altitude field is not bitwise identical between output files. Martin to have a look, he suspects we output a monthly average which would give small differences for months of different lengths.

Optimisation

Manodeep says 5 nodes gives 25 years/day while currently we get 21 years/day, this would be at a similar cost. Should we do optimisation work at 5 nodes?
Tilo and Rachel say yes, we want the throughput considering we are already late on the schedule.

CMIP7

New round of survey from CMIP7 has been released. Meeting next week for a sub-group of people to add the required information.

*Next meeting: this coming Tuesday (11th November), on normal schedule.

ACCESS-ESM1.6 development
63

Meeting minutes 11/11/2025

Chair @clairecarouge
Minutes @cbull
Participants: @cbull @anton @clairecarouge @MartinDix @sofarrell @inh599 @spencerwong @Jhan @Aidan @alexnorton @lachlanswhyborn @manodeep @paulleopardi @RachelLaw @rbeucher @tiloz @pearseb

Land and atmosphere build changes

Lachlan: build for version 2025.11.000 should be ready, in final review stage.
Manodeep: oasis3-mct change to upstream might come in after that release as need more time for testing.

sea ice changes

Anton: still confused about the snow on sea ice variables and how they want them averaged
Emailed the sea-ice mip people (Patricia Derepentigny) – no response. Also emailed someone from NCAR.
Martin: different to cmip6? Anton: don’t think it’s different.

BGC update – change in sinking rate

@pearseb. Now run for 150 years each of the sensitivity experiments. Will use CO2 flux as a benchmark. Who I send the parameter changes to? Response: @spencerwong. @pearseb by end of today/tomorrow.

@RachelLaw: perhaps we want to consider tacking Pearse’s run onto the end of the existing candidate? @tiloz part of the issue here is how long the sinking rate changes take to equilibrate. @pearseb ~200+ years. @tiloz : sounds like it’s worth considering then.

@tiloz. Having said that, whilst we’re still deliberating, still worth continuing the October B run.

@clairecarouge: who’s creating the combined restart? @pearseb hasn’t done this recently, but can look into it. @RachelLaw just fudge the date perhaps? @pearseb will likely have a chat to Spencer once he has a plan.

latest vegetation distribution

@RachelLaw: struggling to get a stable run with latest setup. Wondering what’s wrong with the restart file. @MartinDix is happy to have a look at that.

I think (!) these are the details from @RachelLaw

Martin - in this directory /scratch/p66/rml599/payu-tests/test-restart-nov10/restart410/atmosphere, restart_dump.astart.orig works, restart_dump.astart.new doesn’t work. The README gives the location of the file that I copied to restart_dump.astart.new.

@inh599 do we need to provide a number of active tiles? @MartinDix no, it will work that out at run time. @inh599 what about when the um reads the restart file? @MartinDix No, the restart file includes all files and the restart fraction.

@clairecarouge what is the status of the updated vegetation parameters tests? Do we want to include them in the configuration? @alexnorton: where did @Jhan get up to on that? @Jhan: been wondering where to take that.
@RachelLaw : we’ve changed one more of the c4 crop parameters (moving from a hardwired value). The test is really to make sure it’s stable (?). @Jhan and @alexnorton will follow up offline.

forcing update

all following up from @paulleopardi

  • Nitrogen forcing for pre-industrial control
  • working on the ozone forcing (was trying to make do without the temperature information but that has since come through)
  • has a follow up meeting today

Misc

  • @manodeep would appreciate an amip configuration for testing UM7 compiler flags - easier to check for bitwise reproducibility and performance in the amip model (rather than a coupled model).
ACCESS-ESM1.6 development
65

Meeting minutes 18/11/2025

Chair @clairecarouge
Minutes @spencerwong
Participants: @harshula, @spencerwong, @RachelLaw, @paulleopardi, @MartinDix, @sofarrell, @lachlanswhyborn, @clairecarouge, @Jhan, @manodeep, @anton, @inh599, @rbeucher, @dhb599, @pearseb, @alexnorton, @dougiesquire

Configuration changes

  • The sea ice diagnostic changes have been added to the build and can now be brought into the configuration

  • A decision was made to update the sinking rate parameters without changing the restart file

  • @Manodeep will swap the configuration to 5 nodes, as this is known to boost throughput.

  • The pft parameter changes have not been analysed yet, however increasing the bare ground fraction over Australia has not worked to increase the dust. Given the timeline, it was decided to not try adjusting this any further and instead document that the PFT changes have caused a reduction in dust.

  • The plan is to consolidate the remaining land changes into this PR: Pre-industrial config for optimisation by blimlim ¡ Pull Request #268 ¡ ACCESS-NRI/access-esm1.6-configs ¡ GitHub
    @Jhan will send changes to @spencerwong or add them to the PR directly

New forcings

  • The updated nitrogen forcing should be ready today and @MartinDix notes that the patterns have changed. To understand its effect, we’ll set off runs with and without the new version after the above configuration changes are merged.
  • @paulleopardi is working to use the met office’s ozone processing code now that the required temperature data is available. It was noted that the forcing needs to be ready next week for the next stage of the runs.

Spinup table

  • @inh599 notes that the spin up table needs to be updated to make it clear which runs were kept in the main spinup, and which ones were dropped.

There was a brief note “This was started as a test configuration to investigate the bug.” already there, but this has been expanded upon - Jhan

AMIP and historical configurations

  • @RachelLaw and @tiloz noted that having a test historical configuration is urgent, even if it is not completely up to date with the latest preindustrial configuration as it will allow for tuning carbon
ACCESS-ESM1.6 development
66

Meeting notes 25/11/2025

Chair: @clairecarouge
Minutes: @cbull (started a few minutes late)

Participants: @paulleopardi @clairecarouge @lachlanswhyborn @pearseb @Jhan @MartinDix @spencerwong @sofarrell @dougiesquire @dhb599 @harshula @manodeep @cbull @anton (at end)

@Jhan reported that the November test runs are now working after some hiccups with the configuration and Gadi on Friday.

@Jhan AMIP config failing after first year due to old land use change data writing over new Aus PFTs vegetation, creating bad tiles.

Manodeep has been optimising the AMIP config by adding compiler flags to UM7 - gains have been minor. The best case has a throughput of 25.2 years/day vs the base throughput of 24.9 years/day. All finished 1-year runs with these updated compiler flags have been bitwise-identical. Details are here

@Jhan was interested in whether the AMIP configuration might scale (inefficiently) with more nodes? @manodeep looked into this a while ago with amip didn’t scale beyond a specific number of nodes but that was a while ago before quite a few changes were made. @manodeep will investigate.

@pearseb: looked at CO2 fluxes last night. Standard sinking parameter the values have came down to zero (good) – promising!

@MartinDix: nitrogen deposition change is in. What would we expect to change? Would be good to follow up with @inh599 or @RachelLaw @tiloz @alexnorton [please provide guidance here to know what to plot to compare the old vs new nitrogen deposition runs and reaching equilibrium!]

@paulleopardi gave a forcing update. [please put details here @paulleopardi]

@clairecarouge documentation work and release. Open to volunteers for people getting involved.

@manodeep, which configurations would be part of the release? @clairecarouge likely the picontrol config to start, then drip feeding the others as they are ready.

@manodeep has been looking into calculating the amount of energy used for reporting purposes. Query around which runs to count. @MartinDix clarified one just comes up with a number that the configuration costs.

@spencerwong other configurations update. prototype historical simulation seems to be running without issues but would like to include the cmip7 solar forcing (@paulleopardi has worked on this and there’s a pr in process). Will need some reviews soon.

ACCESS-ESM1.6 development
67

Meeting minutes 02/12/2025

Participants: @Jhan @clairecarouge @pearseb @dhb599 @dougiesquire @inh599 @lachlanswhyborn @manodeep @MartinDix @paulleopardi @RachelLaw @sofarrell @anton @Aidan
Chair: @clairecarouge
Minutes: @cbull

Off-topic but @Jhan thinks he doesn’t have write privileges to the am3 repo.

bgc update

@Jhan we have the two sinking rates tests going that have 130 and 150 years run so far (high and standard). @pearseb let’s go with the “standard” value.

ozone forcing

@MartinDix should be able to get it out in the next few days (pi control).
@paulleopardi pre-industrial ozone is currently being worked on
@paulleopardi: working on the preindustrial ozone forcing, but am not sure if the UK Met Office code will work in our context. @paulleopardi has a fallback that calculates the ozone in a different way, but not yet sure if that calculation is valid.

optimisation update

gadi’s problems have slowed work down
With the latest version we’re still getting 25 years per day with 5 nodes.

Manodeep is doing a load balancing analysis using oasis3mct, just started but this kind of work hasn’t been done in the past so maybe there is some scope for performance improvements.

Re: Amip performance. Last week there was a request to see if performance would be improved from increasing from 2 nodes to say 5 nodes. @MartinDix won’t be running amip runs immediately so you’ve got some time.

@anton is amip part of fast track? yes.

misc configuration discussion

@clairecarouge: what’s next for planning on the configurations.
@RachelLaw. We want to run:

  • pi control
  • emissions driven pi-control
  • any of the idealised run that branch off pi control (e.g. 1% CO2, 4xCO2)

@anton How long is @spencerwong away for? (tomorrow) – relevant as Spencer was working on keeping all the configs sync’d up. @MartinDix emissions driven is pretty simple to set up.

We would like to keep amip up to date too.

@MartinDix would like to do some configuration clean up. Getting rid of an unused namelist variable(s), configurations will need to be updated.

plan for a release?

@anton. Should we press on for multiple configurations releases. This year? Next year? Some of the things to do include: Creating a release branch / release tag / publicly, Final configuration, Forcing finalised?

@Jhan missing field in sea ice? Is this part of the release @anton?

@inh599 doesn’t think we can go to full. Haven’t checked the stash codes. Thinks a beta is realistic.

@cbull. What kind of release are we aiming for? Full, beta?

@Aidan does not have specific plans in terms of a release.

@clairecarouge thinks the documentation needs improvement before release. Some offline discussions have suggested that people don’t have time before January. Interested to hear best ways to coordinate this.

On config documentation, we have some notes here.

Perhaps sections could be assigned via sub-issues of this issue?

@anton project board could be used too.

Misc

@cbull will be stepping away from ESM1.6 coordination to work on OM3/CM3 evaluation. @anton will take on the role.

ACCESS-ESM1.6 development
68

9 Dec

Attendees:

Romain Anton Rach Claire Paul L Martin Lachlan Spencer Jhan Tilo Manodeep Siobhan Ian Dave Bi Aidan Alex N

Chair: Claire

Current spinup:

Jhan - 250 years

CMIP7 Pre-industrial Forcings

Paul L - Ozone forcing preperation code in final stages/review

Marting Did a short test run

This would make the Preindustril forcings complete for CMIP7

Historical forcings still need more work

Diagnostics - need to review for CMIP7 still

Clare: Need science review of data request to then match up with stash request / codes

Needs both science and technical input to complete this

Ian: might want options to low/high/medium amounts of output for different periods in the request

Too many configurations is a concern - we only want one configuration for one experiment

- Spencer - we are using symlinks at the moment to enable this, could work. It’s a little bit manual

- Martin - can we automate the switch (e.g. through payu or a user script ?)

Martin and Spender to investigate

Atmos/Land

Rachel: subdaily data only relevant for historical and scenarios - see [ #164]( Additional stash outputs for data request ¡ Issue #164 ¡ ACCESS-NRI/access-esm1.6-configs ¡ GitHub )

There is a list of mappings from data request → stash codes which might provide some good input

Need to confirm completeness withs needs in moppy (post-processing) too.

Martin / Spencer / Clare / Romain / Ian + - meeting at 4pm today

Ocean

There is the start of an issue for diag_table comments - Siobhan/Matt/Pearce/DaveBi have some feedback on diag_table - see [#303]( Changes to diag_table for the final configuration ¡ Issue #303 ¡ ACCESS-NRI/access-esm1.6-configs ¡ GitHub )

Dougie might be able to help with editting the diag table - Anton to ask

Optimisation - Manodeep

Using MOM collate changes from 25 years to 27 years per day

10% improvement in throughput

Non answer changing - could be implemented later without changing answers - Improve collation process ¡ Issue #311 ¡ ACCESS-NRI/access-esm1.6-configs ¡ GitHub

Also raised New `$NNOREST` flag in oasis3-mct v5.2 ¡ Issue #300 ¡ ACCESS-NRI/access-esm1.6-configs ¡ GitHub on missing coupler restart information - agreed to defer for now