Semantic Frameworks for hydrological Sensor Webs

The Semantic Framework for Hydrological Sensor Webs (SFHSW) project is developing methods to combine semantically rich descriptions of hydrological sensors data and models with logic reasoner based techniques to partially automate the development of complex scientific worklfows for predictive hydrological modelling systems. There are three aspects to this work:

• Develop rich descriptions of sensors and observations
• Develop  techniques to relate (or map) existing data to comunity conceptualisations (ontologies)
• Develop technique to utilise both the ontologies and the mappings to automate aspects of model data integration.

The project has three work packages:

Work package 1: Development of an ontology for sensors and observations

This work makign major contribution to a W3C Incubator project. Project staff co-chair the incubator group. The staring point for work in 2009 was an ontology that provides coverage of:

• Abstract characteristics of sensors - including types of measurement, accuracy, resolution and method of measurement
• Concrete characteristics of sensors - including location, power, associated documentation and some contact and control information
• Derived sensors - ie a 'sensor' that does not physcially exist, but whoose measurement is obtained by mathematical combination of measurements from 'real' sesnors. An example is a sensor that measure wind chill, derived from air temperature and wind speed.

In 2009-2010 this ontology will be extended to include:

• A conceptualistaion of 'observation'. The conceptualisation will be derived from the OGC's Observation and Measurements standard
• Conceptualisation of derived types of observations (eg a timeseries) and links to units of measure

Work package 2: Tools and techniques to map existing data to community ontologies

Tools and techniques will be developed to support the creation of mappings between private data schemas and the conceptulisation embodied in community ontologies. This work will be focussed on the needs to the Bureau of Meteorology, Water Dvisision.

The BoM receives a large amount of data relating to water from orgnaisations named in the Water Regulations.  The current regulations do not specify the schema or the representation for data supplied. The only requirement is that the information is supplied in an machine readable electronic format.A standardisd format for the exchange of water information - WDTF - is being developed.

In order for the BoM to utilise this data an mostly automatic process that resolves the heterogenity of the inputs schemas and formats must be developed. This has become know as the 'data ingestion' problem.

The data ingestion problem is an example of a general need to map from exist data with a private schema and a community ontology as represented by WDTF.

Work package 3: Semantic brokering for Model Data integration

A  significant part of the model data integration puzzle is developing the software 'glue' that allows the different steps in the 'workflow' (that makes up the model data integration) to overcome various sorts of incompatabilities.
From a high level view, these information exchanges must be correct otherwise the entire model data integration process is flawed. However at the detail level a range of incompatabilities occur

Some examples of these incompatabilities are:

• The output from step (n-1) is presented in CSV format, but step n wants netCDF
  This has been described as a syntactic or structual difference.
• The output from a step (n-1) is measured in Deg C, but step n wants requires Deg F.
  This is a simple example of what might be termed semantic incompatability.
  It is simple because the conversion process for units is a simple mathematical relationship.
• The output from step (n-1) is a timeseries of temperature measurements in Deg C, with a time step 10 minutes, but step n wants a timeseries of temperature measurements in Deg C but with a time step of 1 day.
  This is a more complex example, because there are a choices about how the conversion is performed, and the 'correct way' (or more probably the 'best way') can only be determined by knowing more about characteristics of the data source and the data sink as well as the actual data being exchanged. That is the conversion needs information about the context within which the conversion occurs.

We consider all of these to be examples of 'semantic brokering'

Our approach is to use ontologies to provide a sematically rich, machine readble description of the requirements of each mediation step, and sematically rich description of available mediation operations; and then use reasoners to discover a combination of mediation operations that satisfies the mediation requirements for each step.