Extend axis identifiers to include probabilistic dimensions

[fionaRust]

As developers we want to be able to use a short hand way of accessing the probabilistic dimensions on a cube without searching for a coordinate name.

Within Iris we can access spatial and time dimension using axis="T" or axis="X" etc. We would like to do the same for ensemble, percentiles and probabilities.

Note this may not be the best place for this ticket

[bayliffe]

See CF proposals for an E axis identifier to denote an ensemble/realization axis. It is in this vein that we think additional identifiers would be extremely useful for denoting types of probabilistic data.

[bayliffe]

As an alternative we could decide to use the 'var_name' to store the barest description of the data being stored, thus eliminating the need to pass the more complicated 'name' element. In this situation the 'percentiles_over_realizations' coordinate would have 'var_name' of just 'percentiles', functioning as an easy identifier of the different probabilistic data axes.

[marqh]

See CF proposals for an E axis identifier to denote an ensemble/realization axis. It is in this vein that we think additional identifiers would be extremely useful for denoting types of probabilistic data.

there was a varied discussion about adding new axis labels within CF but it did not reach a positive conclusion, there were concerns about expanding this capability beyond X, Y, Z or T.

I think that axis label metadata will struggle to live up to your requirements

[marqh]

I think that the use of controlled var_name within software is a common pattern and can be made to be robust through good design.

I think that this could provide a useful pattern for you, without being overly constraining.

[bayliffe]

@marqh we are closing our local copy of this ticket as we are unlikely to return to this issue any time soon. I do not have permission to close tickets in this repository, so please close it when possible. Thanks.