Note, these configurations are still being actively developed and are not intended for general use
This repo will contain standard global configurations for ACCESS-ESM1.6, the ACCESS Coupled Earth System Model.
This is an "omnibus repository": it contains multiple related configurations, and each configuration is stored in a separate branch.
Branches utilise a simple naming scheme:
{release}-{scenario}[+{modifier}]
where release signifies this is the release branch that is tested, versioned and ready for use, scenario is the base experimental design with optional modifiers. All configurations are assumed to be global extent with nominal 1 degree resolution.
Some examples of possible values of the specifiers:
- scenario:
historical,preindustrial,ssp126 - modifier:
concentration,interactiveC,noLUC
where scenario is typically a CMIP experiment identifier, concentration and interactiveC describe the CO2 cycling protocol, and noLUC is no land-use change.
All available configurations are browsable under the list of release branches. There are currently no released configurations.
All configurations use payu to run the model.
This repository contains many related experimental configurations to make support and discovery easier. As a user it does not necessarily make sense to clone all the configurations at once.
In most cases only a single experiment is required. If that is the case, choose which experiment and then run
git clone -b <experiment> https://github.com/ACCESS-NRI/access-esm1.6-configs <experiment>and replace <experiment> with the branch name or tag of the experiment you wish to run.
ACCESS-Hive contains detailed instructions for how to configure and run ACCESS models with payu.
This repository makes use of GitHub Actions to perform reproducibility checks on model config branches.
Config branches are branches that store model configurations of the form: release-<config> or dev-<config>, for example: release-historical+concentration. For more information on creating your own config branches, or for understanding the PR process in this repository, see the CONTRIBUTING.md.
Config tags are specific tags on config branches, whose MAJOR.MINOR version compares the reproducibility of the configurations. Major version changes denote that a particular config tag breaks reproducibility with tags before it, and a minor version change does not. These have the form: release-<config>-<tag>, such as release-historical+concentration-1.2.
So for example, say we have the following config tags:
release-historical+concentration-1.0release-historical+concentration-1.1release-historical+concentration-2.0release-historical+concentration-3.0
This means that *-1.0 and *-1.1 are configurations for that particular experiment type that are reproducible with each other, but not any others (namely, *-2.0 or *-3.0).
*-2.0 is not reproducible with *-1.0, *.1.1 or *-3.0 configurations.
Similarly, *-3.0 is not reproducible with *-1.0, *-1.1 or *-2.0.
These checks are in the context of:
- PR checks: In which a PR creator can modify a config branch, create a pull request, and have their config run and checked for reproducibility against a 'ground truth' version of the config.
- Scheduled checks: In which config branches and config tags that are deemed especially important are self-tested monthly against their own checksums.
More information on submitting a Pull Request and on the specifics of this pipeline can be found in the CONTRIBUTING.md and README-DEV.md respectively.
For more information on the manually running the pytests that are run as part of the reproducibility CI checks, see model-config-tests.
<TO DO>