start date ensemble problems

[bnlawrence]

See CMIP6 experiment problem at issue 1 in the archive repo.

The problem at hand is that

• ensemble_axis: ensemble_requirement [1.N] in a multi_ensemble does not allow arbitrary numerical_requirements in the ensemble axis (the offending doc has a vanilla numerical_requirement, but it would be broken even if it had a start_date_ensemble (which is what it should be) since it is not a sub-class of ensemble_requirement.

This requires a modification to the CIM before a satisfactory serialisation can be provided.

[bnlawrence]

Notes on the issue:

• ensemble_requirement itself has attribute of ensemble_member: numerical_requirement [0.N]
• start_date_ensemble has attribute of ensemble_members: datetime_set [1.1]

For many ensembles, the overhead of having a full numerical_requirement document for each member seems excessive ... it might be better to have an array of simpler things within one document as an option (we still might want to, for example, have an array of different forcings which themselves are re-usable documents).

[bnlawrence]

We may need a "simple solution" which has minimal ramifications for CMIP6 tooling, and a "better solution" for CMI6 ...

[bnlawrence]

Existing multi-ensembles from Charlotte's spreadsheet sans straight-forward attributes:

[bnlawrence]

Observations:

• 10 multi-ensembles.
• 8 start time ensembles
• two forcing scenario ensembles, one of which is multi-resolution, one of which is
• A number of "x-member" ensemble documents.

[bnlawrence]

@charliepascoe A large number of the standard ensemble requirements seem to be of the form: "Continue from the ensemble in experiment X" with "initialisation method" as the ensemble type. Is this as simple as thinking if there are three ensemble members in experiment X, simply use their last fields as the initialisation for three ensemble members for ensemble requirement Y?

(e.g: G6solarInitialisation, G7cirrusInitialisation, SSP5-85Initialisation2020)

[charliepascoe]

Kind've. The description field gives more information. Some say simply that the initialisation is from the end of experiment X. In the case of the G6SolarInitialisation the instruction is that "Initialisation is made from the final year of the GeoMIP G6solar experiment".

[bnlawrence]

You chose not to use the "initialisation_requirement" document. Any particular reason (that you can recall)?

[bnlawrence]

(I am thinking there is a simplification available here, but need to decant your knowledge first to see if it really exists :-) ) (The ensemble_requirement points at lots of numerical_requirements, and is different from the initialisation_requirement for a starter ...)