Wiki to explore Ticket #921
How should we handle perturbed physics ensembles in the CMIP5 questionnaire and the CIM?
Model mods can only point to components and not to parameters... is this a problem for the CIM?
The solution will have implications for capturing information on configured and unconfigured models #888
When users describe a model mod used by a simulation they associate it with a component and not with a particular parameter. Apparently this is an issue for the CIM* because the CIM expects to be able to link a model mod with a parameter. But the CMIP5 questionnaire does not try to capture every component of a model, instead it captures only those parts that were of interest to scientists doing model inter-comparison work. So it is very likely there will be a mis-match between the broad information that we capture in model descriptions in the questionnaire and the fine detail that users give us when describing perturbed physics ensembles.
I see two options:
- adapt CIM to fit questionnaire - the questionnaire has the capacity to capture enough information about perturbed physics model mods to satisfy scientists. Can we adapt the CIM to allow mods to be associated with components?
- adapt questionnaire to fit CIM - add the ability to link model mods to parameters. Users will be required to add new parameters to their model descriptions, where necessary, so that they can be modified.
Option 1 means we must expect the perturbed physics mods to be about parameters that we have not asked users to describe in the questionnaire, users must specify the most appropriate component instead.
Option 2 puts a lot more work on the user - we are already getting push back from users so this will not be a popular solution.
Any solution must allow all the information we capture in the questionnaire to be displayed using tools that exploit the CIM
*Allyn wrote: It is not true that the CIM cannot cope with that. The CIM deals with mods by associating a change with the target of a conformance. This change can be of type ModelMod?, InputMod?. etc. And the target, or the change itself, has the reference stereotype, which means that it can point to anything: a component or a parameter
Charlotte replied: The thing is that the perturbed physics ensembles were not specified in the CMIP5 experiment design document and so are not captured as requirements in the questionnaire which means that they will not have conformances associated with them. Can the CIM cope with a change that is not the target of a conformance?