Josh Lieberman, Will Pozzi, Mattia Santoro, Brad Lee
Josh: Discussion with Gulf of Maine and NH people at NECODP:
Shellfish bed closures are based on 2-3 year correlations between rainfall and microbial concentration in water column and organism.Will: Drought scenario can have both work on drought ontologies and drought prediction components.
Beach advisories are not based on formal correlations, because beaches are not actually closed. So data are collected, but correlations remain to be done
Proposal to branch the scenario into decision support for shellfish and statistical analysis for beaches.
Mattia: How shall ontologies be published, in what form?
Will: CUAHSI ontology is currently in OWL modules. Intend to add drought components in Protegé (from Princeton US and Global as well as European Drought Observatory terms) and re-export as OWL.
Mattia: should consult with Cristiano on what ontology structures can be accommodated. Initial structure seems to be SKOS entities, but unknown whether a more general structure (OWL Lite, OWL DL) is tractable.
Will: I have uploaded two presentations that demonstrate drought systems for Spain. The Spain "indicators" system is based upon monitoring groundwater levels and reservoirs levels, water usage, and precipitation levels (sensor web of sorts) so that the ability to withstand a drought can be assessed in terms of having supplemental water form reservoirs or groundwater. If supplemental water sources are available, then the drought response might be downgraded from an emergency to simply an alert. Setting up an actual system (drought alert system) will be complicated; here in AIP we will offer a more simplified case. However, the EDO scenario is expressed as decision makers being presented with a drought index (areas colored as red or yellow or green), and the decision maker has to evaluate the severity based upon mitigating factors, i.e., reservoirs and groundwater sources (or the ability to substitute treated wastewater for fresh irrigation water in agricultural operations). The decision maker can query datasets using EuroGEOSS broker to search for this level of detail of information. To support this scenario, the ontology has to include these components, reservoirs, groundwater, as well as soil moisture and drought variables as used by both the EDO and Princeton case studies. But work is not limited to expanding the ontoloogy; we also have to select a localized enough area as a test grounds to ensure that datasets covering these indicator components will be registered in the EuroGEOSS GI-cat catalog (and GEOSS common registry).
What steps are missing to leverage the semantic capabilities and merge drought scenarios?
Josh: But we can't be limited to an area; the methodology has to be scalable and extensible to other areas.
Will: That is the reason for using the upper level semantics: to permit the datasets for each river basin to be integratable together (and retrievable through) the appropriate terms in the ontology.
Josh: Ontologies could be simply utilized in two stages: query expansion to find data sources for evaluating a particular area (same-as and broader-term / narrower-term relationships), then result translation in to multiple threat notification / categorization schemes (may need a category calculation as well as mapping).
Mattia: essential role of broker is to expand keyword searches to find related datasets.
Will: interest in pulling in other data related to or mitigating drought hazard.
Josh: Three elements to enabling this, probably requires some iteration:
Mattia: First steps should be to expand queries by one relationship level, then investigate what other relationships or patterns may be important for later steps.
Will: Worth looking at Noesis for ideas. Need at this point to try out some concrete cases with real datasets.
Will: Next steps?
Josh: Consider starting with a "canonical" drought estimation workflow, including discovery steps. Operate it in one region. Then move to another region and utilize ontology mappings to discover the right datasets to carry out the workflow.
Will: The right datasets need to be registered in order to do this.
Will: Are these problems arising within the other SBA working groups in AIP needing term mapping / expansion?
Josh: should be a question in the plenary as to where the other groups are with needing semantic integration.
Brad: every group will face this, whether they are prepared to grapple with it now or not. Registration and cataloguing is a major effort along, though
Josh: This is a potential value of catalog federation, so that the registration task becomes one of hooking up existing catalogs according to a particular interface, not starting from scratch.
Brad: have been using SKOS as basis for water ontologies for now since the relationships are more traversible than those of more complex ontologies. What we use is largely homegrown, but a student at Muenster is working on expanding that.
Document WQ scenario.
Describe drought estimation workflow, so that requirements for broker functionality can be expressed and reviewed.
Look at BDCC group semantic processing requirements (telecon next week).