New method added that allows users to fill values directly from a pandas DataFrame.

v2.2.2 fixes a bug in the direct aerosol scaling that was picking up the scale factor of precursor species, and turns off the warning and debug messages from the logging module.

This release formally adds support for python 3.13 on pip and conda.

v2.2 introduces new functionality so is a new minor release of fair - the first for a while!

The biggest improvement is the ability to define emissions scenarios and parameter sets easily in importable CSV files, and makes defining custom emissions scenarios much more straightforward. This is handled by the new fill_from_csv() method.

Large probabilistic ensemble parameter sets can be straightforwardly included thanks to the new override_defaults() method.

An example that shows these features in action is at https://docs.fairmodel.net/en/latest/examples/calibrated_constrained_ensemble.html. There are many fewer lines of code required to run fair than previously, and hopefully this makes the model a little more intuitive.

While it almost certainly worked before, python 3.12 is now officially supported, and code coverage has been brought up to 100%.

This version fixes an error that occurred if either CH4 or Aerosol-cloud interactions were missing from the list of species defined and the list of species configs in f.fill_species_configs() was read in from a CSV file.

An inconsistency with ocean heat content before and after a restart has now been fixed. It was not affecting results since OHC is a post-processed quantity, only reporting.

One module dependency was failing tests with python 3.7, so support for python 3.7 has been dropped.

The move to the new Zenodo API made pooch fail in the fill_from_rcmip() convenience function. The path to the emissions files from RCMIP have been updated to directly link to URLs rather than DOIs.

xarray 2023.9.0 introduced a change that broke FaIR with python 3.9+. Following user feedback, documentation has been updated, including, most importantly, an example using the latest available calibration. The positive semi-definite RuntimeWarnings should have also disappeared when running stochastically, or at least can be disabled.

Support for python 3.6 is removed due to dependency issues, and 3.11 added.

v2.1.0 is an overhaul of FaIR, and includes the following:

• object-oriented design
• interactive methane lifetime
• optional internal variability
• mass parallelisation

v2.1.0 is an overhaul of FaIR, and includes the following:

• object-oriented design
• interactive methane lifetime
• optional internal variability