Observations are fed into the Sensor Web through a growing number of environmental sensors, including technical and human observers. While a wealth of observations is now accessible, there is still a gap between low-level observations and the high-level descriptive information they reflect. For example, we may ask what the measurements mean when a weather buoy provides a temperature time series. The challenge is not to gather a vast number of observations, but rather to make sense of them in environmental monitoring and decision making.
In order to infer meaningful information about occurrences from observations, a description of how one gets from the former to information about the latter must be expressed. This thesis develops an ontology to formally capture the relationships between geographic occurrences and the properties observed by in situ sensors. Building upon the existing positions on experiential and historical perspectives, stimulus-centric sensing, event-process algebra and thematic roles, the ontology elucidates the key concepts associated with geographic occurrences that are particularly significant from a sensing point of view. A use case for reasoning about blizzards and their temporal parts from real time series supplied by the Environment Canada illustrates the ontological approach. This thesis evaluates its findings on the basis of a comparison with an alternative approach in the Sensor Web, a verification of the use case results using an official event report published by the weather agency and an analytical assessment approached from the system development perspective.
The theoretical contribution of the thesis lies in the development of a formal model, which constitutes common building blocks for constructing application ontologies that account for inferences of geographic events from observations. With regards to its practical contribution, the thesis has demonstrated how ontological vocabularies are exploited with reasoning mechanisms to infer information about events, and to formulate symbolic spatio-temporal queries.