Combining Process and Sensor Ontologies to Support Geo-Sensor Data Retrieval

The Sensor Web Enablement (SWE) framework specifies how to access web-enabled sensors and sensor data on a syntactical level. Nevertheless, it is difficult to discover a sensor that fulfills certain criteria from a huge set of available sensors; effort is required to retrieve, interpret, and combine sensor data based on heterogeneous schemas. Semantics-based approaches promise to overcome these challenges (Bermudez and Piasecki 2004, Babitski et al. 2009, Kuhn 2009, Janowicz et al. 2010), but so far most work has focused on ontologies for sensors and their observations. We propose a combined approach which relates a sensor ontology to a process-centric domain ontology and hence takes into account how observation data was created and what types of features it relates refers to. In accordance with the OGC specifications and related work (Compton et al. 2009, Kuhn 2009, Stasch et al. 2009), our sensor network ontology is not limited to physical devices but is constructed such that any entity (e.g., physical sensors, computational process) observing a phenomenon can be represented as a sensor. In contrast to existing hydrology ontologies (Bermudez and Piasecki 2004, Beran 2007, OrdnanceSurvey 2007, Vilches-Blázquez et al. 2009), we focus on a process-centric ontological approach. Process-centric means that the domain ontology is developed by first identifying geo-processes, entities (i.e., objects and matters), and their properties, followed by the relations between them. These relations are used as basis to handle naming heterogeneities and hence support sensor data retrieval. The following sections of the paper describe our ontologies. We will use the concept of Evapotranspiration (ET) as running example as it is a key component is in the Hydrological Sensor Web research by the CSIRO Water for a Healthy Country Flagship initiative. For details, see (Guru et al. 2008).