wiki:CMIP5/Storyline

This page describes the plan for the overall path through the CMIP5 metadata questionnaire. It provides documentation for ticket:195.

Initial assumptions

  • The experiments are fully defined a priori, and so questions can use that knowledge.
  • There is a list (controlled vocab) of modelling centres available from PCMDI to use.
  • There are (or will be) some forcing files and descriptions available from PCMDI so we don't need to prompt for details of those.
  • That the model grid information is coming from a gridspec file and grid information can be linked into the metadata in a different exercise.
  • The the output data description is coming from the data files and doesn't need to be described and can be linked to the metadata in a different exercise.
  1. User starts by choosing their modelling centre
    1. If not already filled in, modelling centre details should be filled in.
  2. Prompt for experiement from the experiment list.
    1. The centre may already have done something for this experiment, so discriminate between edit and add.
    2. If desired, walk them through the experiments in "our" order (starting with control) so as to minimise duplication as layers of info are added within the experiment families.
  3. Identify the model for this experiment.
    1. There will be a model-id-name in the filename and files, but that's a model family name, not the specific "configured model" used for this experiment, so we need to discriminate those concepts even if we don't use those names. And prompt appropriately.
    2. May involve adding a new model, or copy an existing model.
    3. Describe model in detail (either from scratch or from what you've done so far)
    4. At this point we need to handle ensembles of different types, so modellers can enter the various different numbers/dates/whatevers along the axis of ensemble ... and then replicate model/run descriptions ... Use “edit mode”, but avoid 56 clicks for an edit that needs to be repeated for each ensemble member .... (i.e. we will know for CMIP5 the "ensemble axes" so we ought to be able in the questionaire offer a drop down list of ensemble types for model replication rather than editing each one individually).
      • This avoids the need for the metafor "change" property within CMIP5.
    5. Model descriptions involve full descriptions of appropriate subcomponents, the list of appropriate subcomponents will come via the controlled vocab work. Components will need to be reusable particularly as work progresses (e.g. an AMIP model will be different enough from it's parent coupled model to start from scratch, but it would be nice to just "use" the atmosphere component already described.
      1. The use of (and descriptions of) post processors may be important
  4. describe run context (configuration)
  5. describe conformance (ie link numerical requirements to data or software)
    1. if possible exploit the preloaded information about forcing datasets
  6. describe platform
  7. ensure all official experiment diagnostics are associated with some component
  8. link to atomic datasets
    • not sure about when and how this should be done. (we need a list mapping the diagnostic names to the names atomic datasets so we can pre-fill all the obvious ones, maybe all of them in which case this step isn't necessary)
    • Need to link to information harvested from NetCDF files and URI.
  9. Have you finished? No, go back to an earlier step (maybe offer jump back to experiment start, maybe elsewhere).