Ontologies: Semantic Nirvana for Earth Science Model Interoperability?
Ontologies: Semantic Nirvana for Earth Science Model Interoperability? (Invited)
J. Graybeal1
1. Marine Metadata Interoperability, Mountain View, CA, United States.
The Challenge: When we build a given model, we do so to meet today's needs. If the model is good, new people will want to use it in new ways. That tests how well the model can work in new contexts: new user groups, new science domains, or new data providers or data users. We can say a model is interoperable if it works well in each new case, with few or no changes. Here we deal with perhaps the least-addressed part of model interoperability: semantic interoperability, the ability of models to understand the meaning of each other's data.
The Scenario: A model has been built that uses observational data, and creates output data sets. In subsequent years, the model must (a) be connected to another model and exchange data with it; (b) be evaluated and used by a scientist in another domain; (c) document its outputs for two different repositories that use different keywords; and (d) identify and incorporate new observation streams as they come on-line. All these steps are mostly done manually today, and explanations about the data exchanged in similar form. Can we make them more efficient, or even automated, by leveraging good semantic practices? A problem in each case is the use of local or community naming conventions that are not known to all parties. How can this be improved?
The Reality: Many models use the standard name conventions and vocabularies specified by the netCDF COARDS Climate and Forecast conventions. These provide a good basic level of 'semantic interoperability', and for this reason alone Earth science models are semantically far ahead of most other Earth science data systems. Yet these conventions aren't always used, aren't always sufficient, and don't help us interoperate with lots of existing systems. What are the issues for semantic interoperability in modeling, how do ontologies and other semantic capabilities help us fix them, and are ontologies worth the trouble?
http://marinemetadata.org/semanticframework