ACCESS-ESM1.6 development

CMIP7 CMIP7 development
62

Thanks @RachelLaw, I’ve sent an email to @pearseb @tiloz @dhb599 @matthew.chamberlain @harshula @Aidan regarding next steps for getting Dave’s changes into an ESM1.6 executable (happy to share these details if useful).

63

Reminder: stand up on shortly in our usual room at 9.45am.

Minutes now here, please amend as appropriate.

64

Hi everyone,
I’m just adding an update on the current state of the configurations in the esm1.6 configurations repository.

Available configurations

CABLE3 + WOMBATlite + iceberg fluxes

The branch dev-preindustrial+concentrations uses updated executables, namelists, input file paths, and restart paths for running with generic tracers WOMBATlite + CABLE3 + the new iceberg scheme.

The linked configuration uses the executables from PR 27-5 in the build repository. These executables use WOMBATlite + CABLE3 + the new iceberg scheme. However these don’t include the UM changes from https://github.com/ACCESS-NRI/UM7/pull/45, and or any pull requests that are currently open. Let me know if any of these changes need to be included in the default executables used by the payu configuration.

I have tried to run with this however have run into stability issues. The seventh year crashed due to a GCR convergence error in the UM. Perturbing the restart got past this, however the next year crashed from what looks like a segmentation fault coming from MOM in ocean_sbc_mod_mp_get_ocean_sbc_. Output from this simulation is available at /scratch/tm70/sw6175/access-esm/archive/wombatlite+cable3+icebergs-test-iceberg-flux-inputs+config-files-0bb1f2fc. Let me know if you are unable to access it.

CABLE3 + WOMBATlite

An second set of executables is available in PR26-4 which only use WOMBATlite + CABLE3 without the iceberg changes. A configuration temp-no-iceberg-preindustrial+concentrations is available for testing these executables. If everything works with the iceberg version, this second configuration could eventually be deleted.

Running instructions

For very brief quickstart instructions, you can run a one year simulation of this configuration using:

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

pay clone https://github.com/ACCESS-NRI/access-esm1.6-configs -B dev-preindustrial+concentrations <experiment-name>

cd <experiment-name>
payu setup
payu run -f

Please see the ESM1.5 instructions for more details on these steps, and on how to customise your experiment.

65

Thanks @spencerwong . This morning we had a CABLE meeting and discussed ESM1.6 configurations and model versions. We discussed the fact the configuration picks up the head of the dev-access-esm1.6 branch. This seems a good choice, it means it includes all changes that have been developed and reviewed and leaves out everything that is still on going.

We were ok with that choice to set what would be included in runs over the summer. This way we know we have to settle open PRs if we want them in the runs.

1 Like
66

@clairecarouge , @spencerwong , @Jhan - do we at least need https://github.com/ACCESS-NRI/UM7/pull/45 in since that one is done and the comment in the PR says “Even without them the “trunk” is basically broken until this goes in.”?
Could this be contributing to the instability - or does that look like and ocean problem? I have no experience with the ocean.

67

@RachelLaw Spencer told me I explained myself badly. What we discussed this morning is we want to run from the dev-access-esm1.6 branch but we will want to update the build “just before” the run is started. At that point, the build will pick up everything we can get ready on time.

So, yes we will want an updated build but @pearseb will likely need to do it before he starts the run to make sure he picks up all the latest rather than @spencerwong adding it to the dev-preindustrial+concentrations config on GitHub.

68

Thanks @clairecarouge, I think in that case the dev-preindustrial+concentrations should be ready to go!

69

Thanks. That was the message I left with @clairecarouge. It would ge good to start such a run Friday afternoon for quick review early next week - in exactly this workflow ^^

70

Nah - this was all so recent that there was not even a chance too see the instability. Its in now anyway :slight_smile:

71

I’ve put a new cable.nml in /scratch/public/rml599
This is what I’d suggest Pearse runs with.
It uses Medlyn photosynthesis, froot as ESM1.5 and keeps fwsoil_switch as standard. I have a 50 year run with similar settings (but pre-iceberg, CMT etc) so that would be good for comparing with.

For a 2nd test that hopefully Jhan can run, I’d suggest
fwsoil_switch=Haverd2013
turning on the soil_thermal_fix (should be enabled in the code now but not yet tested)
Changing snmin to 0.11 (pull request in progress: Make snmin a runtime parameter set through cable.nml. by rml599gh · Pull Request #51 · ACCESS-NRI/UM7)

Any thoughts on turning on hydraulic redistribution? My tests didn’t show much sensitivity to this, especially in its current form when it is only active for evergreen broadleaf and c4 grass. There was a little more impact if I switched it on everywhere.

72

Thanks Rachel, would you like the new cable.nml to be added into= the payu configuration, or is it easier for Pearse to copy it separately when he starts his runs?

73

@spencerwong - I think it would be good to add it to the payu configuration as it includes a couple of switches that aren’t in the current cable.nml and at least this version lets people know what they are and the choices for them (rather than picking up not so visible default values).

74

In /scratch/public/rml599 there is now also a cable.case2.nml. This should be the settings for what I described as a ‘2nd test’ in my earlier message.

75

Thanks @RachelLaw, I’ve just requested a review for this pull request to update the namelist file.

76

I had a look at STASHC_standard_concentrations in atmosphere/diagnostic_profiles.
I suggest deleting the timeseries output (lines 268-272) as we are not routinely looking at this output and are unlikely to for long simulations.

And thanks to whoever tidied up the STASHC file so that it was easy to find the timeseries entries.

77

Hi @RachelLaw, I’ve just added a pull request to remove the timeseries from the standard preset here.

78

Reminder: stand up tomorrow, in our usual room at 9.45am.

79

A few notes from the meeting today are here:

You should all be able to edit it for clarifications / extra detail if needed.

80

Notes from today’s stand-up are here:

Please edit fr clarification / extra detail if needed.

81

Hi everyone, @pearseb has added wombatlite fixes that should be included in the executables being used for testing.

Other than the main deployment here, are there any other deployment pull requests that should be updated with the changes?

Thanks!