Modelling a dataset (O&M, raster issue, ...)

[letmaik]

This subdiscussion of #3 shall be about how to model the actual data that is produced within Melodies. Of course every WP has different requirements and it is difficult to unify everything, but certain themes crop up again and again and this discussion should track best practices and how to solve certain issues in modelling. The goal is always to expose data in an as web-friendly and LOD way as possible, no matter if it's custom ontologies, "standard" O&M, raster data or something else.

As far as I see, ontologies shall always provide the glue and semantics. Even if custom data APIs or similar are used (possibly for raster data), they should always be woven together with LOD and form a big picture. Any existing experiences are very welcome here. Please also say things like "I'm lost, help me, we have this kind of data... any ideas?" just so that everyone is on the same page.

Some keywords:

• Observations & Measurements ontology (anyone tried that already?)
• other standard ontologies (which?)
• custom ontologies (who has done it? any issues? can we see it?)
• other developments

[jonblower]

We should take a look at the SmartOpenData modelling activity too. They are making heavy use of the new INSPIRE ontologies, partly developed by Andrea Perego, who is one of our PAB. Their data are very GIS-oriented (ours are perhaps more "scientific" in nature) but their work will still be relevant:

Here are the most relevant deliverables (go here for a full list):

• Initial data model
• Harmonization of data to the data model