Suite to run ACCESS-AM3

Atmosphere AM3
,
1

Dear Community

I am thinking of running a present-day global atmospheric simulation. I think it would be best to use a low-resolution (~N96) ACCESS-AM model first, as I will not dig into the ocean or biogeochemistry in the near future.

May I ask a few questions:

  1. I can find instructions to run ACCESS-CM and ESM, but not AM here. I have experience in running rAM3. Is there a standard or supported suite to run the latest ACCESS-AM?
  2. Would the community suggest going with ACCESS-CM or ESM directly? Are the UM versions used in CM and ESM still UM10.6 and UM7.3?
  3. I understand the Centre has a project on high-res AM modelling. For now I only need low-res, and I expect the high-res simulations may not be available very soon. Are there existing low-res AM simulations for me to start with? Shorter ones are also okay.

Thank you in advance.

Regards, Qinggang

2

@qinggangg to answer your questions:

  1. There is no released version of ACCESS-AM3 at this stage. We are working on an alpha release in the coming weeks. This release is intended to collect feedback on the model and will not provide any support. We expect a beta release to come in the first quarter of 2026 (dependant on the feedback on the alpha release obviously). This is why there is no documentation on AM on ACCESS-Hive as we do not support an AM model at this stage.
  2. ACCESS-CM2 is still using UM10.6 and ACCESS-ESM1.5 (and soon 1.6) are using UM7.3.
  3. With the alpha release of ACCESS-AM3, we will release a configuration based on the GAL9 configuration from the UK Met Office and ancillary files from CMIP6.

I will not comment on which model is best for your research project. However, it is possible to give you early access to ACCESS-AM3 if you wanted to. But keep in mind that, as with all early access software, support and stability are not guaranteed.

3

Hi Claire

Thank you. That sounds great. Could you please provide me access to ACCESS-AM3 and its ancillaries? Gadi ID qg8515 and Github: l975421700.

As long as it can run, I can use it to some degree for preliminary analysis and ideally also contribute to beta releases.

Thanks in advance.

4

Hi Qinggang,

I’m happy to help you configure your experiments with AM3. A few things to keep in mind:

But if you want to run a tests to, for example, assess the resources you will need let me know.

5

Hi Pao

Thank you. Could you please let me know where I should start to configure AM3?

I have been in touch with Shixue. I understand there is a one-month simulation for N512, and no comparison has been done between N96 and N512 because some configurations differ. I am happy to run the N96 simulation for one month as a test and compare it with N512, ideally with an identical configuration except for the resolution. Though, my own research focus will be solely on N96 for now.

Regards, QG

6

@qinggangg , I have sent invites to join the ACCESS-AM3 repositories you will need under the ACCESS-NRI github organisation:

  • access-am3-configs (private): contain the configuration (in the main branch). Please read the README for instructions. There were some recent modifications to the suite (note for @paocorrales …) so the steps to run the model have changed slightly. We have just deployed a version of ACCESS-AM3, so if you don’t need to modify the code base, I would suggest you use it (the configuration is set to pick it up by default).
  • UM and JULES (private): the code bases of the models used for ACCESS-AM3. This is just in case you need to modify the models or check a details in the models
  • GCOM4 (private): it’s needed to compile the UM if you ever make a change in the model and need to re-compile
  • ACCESS-AM3 (public): the deployment repository to manage and deploy the model builds.

Please remember that all the private repositories are covered under the UKMO licence and should not be shared with unauthorised people. In particular, if you clone one of these repositories locally it is recommended to clone them in your home directory or to remove group access permissions (chmod -R g-rwX <directory_name>).

To start, please read through the access-am3-configs’ README.

7

Thank you, Claire

I joined all the repositories and will have a look this afternoon. I may need to change the model source code later.

8

Hello. I managed to run ACCESS-AM3 for one month. And I am trying to configure the output variables and frequencies. I am just testing a variable Cloud Number now and I cannot find it in the NETCDF files.

Maybe an early feedback for the alpha phase: could you please provide an instruction to output desired variables?

Regards, QG

9

Hi Qinggang,

AM3 runs GAL9 science settings, which uses the Wilson and Ballard microphysics scheme. This is a single moment scheme. There is no predicted cloud droplet number.

10

Thank you Bethan. Yes, I thought the output maybe specified cloud droplet number. I can output this variable for rAM3 while using GAL9, which is 5 cm-3. I am very skeptical of this value/output. Do you maybe know what is the actual specified value?

11

It should be the same for all GAL9 but I’m highly sceptical of that number too! Where did you find it? Happy to help dig into what’s going on.

We should probably move this discussion to email / Teams / Cumulus / another Hive thread.

12

It is found in all my rAM3 simulations, e.g.: /g/data/gx60/qg8515/cylc-run/u-dq700/share/cycle/20200602T0000Z/Australia/d11km/GAL9/um/umnsaa_pa000.nc. You can use ncview to check the variable: STASH_m01s04i210:long_name = “CLOUD DROP NUMBER CONC. /m3”.

Yes we can move the discusssion on CDNC to emails as it is more on science part. So in this thread we can focus more on the general technical part of how to output variables that may be used for research.

1 Like