Some recipes, and in particular those ported from AutoAssess, will expect an "test" (the new simulation to evaluate) and a "reference" (a baseline or control run against which to compare). Initial development of CMEW has standardised a single model development run as the experiment, and used CMIP simulations as the control.

As a precursor to #130, we want to be able to standardise and compare two model development runs. This will require two separate calls to cdds_convert, one for each model development run.

The following variables will need to be duplicated across the reference and test (the user will need to specify values of these for each run:

• model_id
• suite_id
• calendar
• variant_label