80 likes | 92 Views
Agenda. expectations & goals what do we want ontologies to do for us? are there common measurements within lter? what information is needed to describe an attribute? Implementation issues common framework user interfaces both for query and submisssion EML integration. What is an ontology.
E N D
Agenda • expectations & goals • what do we want ontologies to do for us? • are there common measurements within lter? • what information is needed to describe an attribute? • Implementation issues • common framework • user interfaces both for query and submisssion • EML integration
What is an ontology • set of related concepts or classes • classes can be inherited from other classes • classes can have properties • Ontologies can be very deep or very shallow • Tools exist to traverse and query ontologies • eg. find sibling concepts of a class through their common parent class
Building ontologies • Top down – start with very abstract concepts and fill in detail • Bottom up – define lowest-order concepts, then start looking for grouping concepts. • Middle out – define reasonably low-order concepts but avoid extreme detail introduced terms that merely mix concepts
What goes in our ontology? • To fully evaluate, use, or rescale an ecological observation one must understand its properties: • Measurement type • Context: Space, time, culture, deposition, substrate • Methods • Units • Accuracy
Measurement Type • formally register the concept for what is being observed (could be classifications, measurements, presence/absence) • ideally divorced from other properties such as context methods, units (contrasts with EPA Storet which defines a unique concept for each combination of measuremnt, units, methods) • measurement scale& domain information • however, domain within given instance may be further restricted by context – body length vs body length of voles
What do we expect from this • simplify or reduce work involved in filling out eml-attribute • reusing information, eliminating redundancy • enabling templates in metadata creation • Improving applications • using attribute ontologies as indexing and keywords • linking attribute ontologies to display rules • Support for data integration • linking attributes to rules for data aggregation, reclassification, calculations
Semantic Issues • Dealing with attribute descriptions that embedd some contextual constraint • eg soil temp, air temp, min temp are all conextualized observations of temp • virtually all count and proportion based measures are context laden – taxonomy or chemistry • not clear rules yet • guided by practical limits (taxonomic abundance) • what resonates with scientists (chemistry)
Pilot pilot-studies • water chemistry • found common variables eg DOC. • found many qualified by context (GW, stream) • biodiversity • konza datasets five biomass columns that were subclasses of live biomass restricted by woody, annuals, etc. • sampling was an important but difficult concept to evaluate. • ANPP