Confused on running a pre-industrial test of ACCESS-ESM1.5 with payu

Hi all - likely @Aidan or @holger
I’m trying to run some ESM1.5 tests using payu.
Previously I’ve used GitHub - coecms/esm-pre-industrial but this is now saying to use this instead: GitHub - coecms/access-esm: Main Repository for ACCESS-ESM configurations. This would be OK but the setup instructions on that page include the line

git co pre-industrial

which reports it is not a git command.
I thought I’d go back to the Hive and look at the instructions there for running ACCESS-ESM but it is still pointing to the older esm-pre-industrial GitHub.
My git knowledge is almost non-existant and my payu knowledge isn’t much better so I really struggle when the instructions don’t make sense.

Hi Rachel,

git co would be an alias for git checkout I believe. I’ll propose the change on the GitHub instructions.

1 Like

Hi Claire,
Yes - that seemed to make sense. Still getting my head around how git does branches. Ben’s tutorial on YouTube is helping but it took me a while to find it in the transitioning cable to github forum topic. Perhaps it is worth adding to the training section of the Hive.
Having got the payu run going, I think there may be an issue with the carbon pools in the start dump. Is that best reported as an issue on the github (coecms/access-esm) repository?

Excellent idea! Going to do that.

For the issue of the carbon pools, as you know the support for the esm will be transitioning. So it isn’t clear to me right now. I would suggest you put it as an issue on the repo. I’m discussing with Paola this afternoon and I’ll discuss internally as well. But we’ll know it’s there and it will be handled.

My fault, my fault!

The command is

git checkout pre-industrial
1 Like

README is fixed

Hi @holger ,
Thanks for fixing. Further down in the Readme, when it steps through how to set up and run the model, I think it still refers to the old esm-pre-industrial version. If so, it would be good to update that too.

On the carbon pool issue that I also mentioned in this topic. Here are a couple of example figures from xconv. The first one is from /g/data/access/payu/access-esm/input/pre-industrial/start_dump/pre-industrial.astart.
The second one is from /g/data/access/payu/access-esm/restart/pre-industrial/atmosphere/restart_dump.astart.
It is hard to spot but there is one dodgy grid-cell on the coast of Norway. Other fields in the file also seem to be corrupted in a similar way. It always seems to be coastal points so perhaps this file has somehow got mixed up with one with an incorrect land-sea mask. I would have thought this file should have been exactly the same as the pre-industrial.astart which looks OK.

1 Like

@RachelLaw I see you haven’t put the issue on GitHub. Also, following our conversation yesterday, do you still have a crash happening when replacing the start dump?

Hi @clairecarouge ,
Thanks for checking. No, I haven’t put anything onto github - another new thing to try! Arnold has given me some suggestions on how to run with an alternate start dump so I will try and get to that later today.

Almost certain they mean git checkout (svn uses shorthand co)


@RachelLaw have you had any resolution of this?

No resolution. I never really got to the bottom of where/how to reliably change the restart file and reverted to script-based running of the model.
This topic, Porting CSIRO/UMUI ACCESS-ESM1.5 ksh run script to payu - Earth System - ACCESS Hive Community Forum ( also seems relevant as it notes some challenges with multiple namelists etc.

@paulleopardi has also been running into difficulties with the ACCESS-ESM payu configs

The intention is for ACCESS-NRI to release supported versions of ACCESS-ESM1.5 in March, which will require any outstanding issues with these configurations be fixed in the versions we release.

The pre-industrial experiment actually uses the file from /g/data/access/payu/access-esm/restart/pre-industrial/atmosphere/restart_dump.astart and confusingly ignores /g/data/access/payu/access-esm/input/pre-industrial/start_dump/pre-industrial.astart. It crashes almost immediately.

I’ve got a test version in GitHub - MartinDix/access-esm at pre-industrial that uses the year 101 restart files from PI-02 and reproduces PI-02 exactly. In the process of moving all the inputs to the vk83 project.

Hi Rachel,

Are you still looking into this? Did it eventually work or did you move on to other things?

As noted above (1 Feb), I reverted to using a script-based run rather than payu.

Work is underway to re-release the ESM1.5 configurations with payu which should fix this. As a consequence, I will mark this as wontfix.