19th International CODATA Conference
Category: Knowledge Discovery

Information Integration through Events

Kate Beard (beard@spatial.maine.edu), Professor and Chair, Department of Spatial Information Science and Engineering, University of Maine, USA
Neal Pettigrew (nealp@maine.edu), Associate Professor and GoMOOS Chief Scientist, School of Marine Science, University of Maine, USA


Large amounts of environmental data are now being routinely collected and made available, often on-line. A substantial proportion of such collections can be regarded as spatio-temporal data that include recorded dates, times, and locations of observations on environmental variables. We can expect an increasing number of shared environmental data collections and a growing number of researchers interested in combining such data to interpret ecological phenomena and their interactions. In such a context there is a growing need for exploratory and analytical tools for investigating diverse collections of environmental data covering a range of spatial and temporal regimes. There are at least two significant challenges in this analytical context. One challenge is the extreme heterogeneity among data collections.  Environmental data are collected by a wide range of sensors and measurement protocols and thus differ with respect to measurement units and scales, spatial, temporal and thematic resolution, and formats and media as well as quality. The extreme heterogeneity of these data sets has tended to limit exploration and analysis to within individual observation data streams rather than across data streams and thus restricted opportunities for developing a systems level perspective on ecosystems.

The second challenge is the lack of tools for exploring and analyzing combined spatial and temporal dimensions.  Current information systems tend to handle one or the other dimensions but not both. Geographic information systems address spatial pattern analysis but the full potential of spatio-temporal analysis is limited by inadequate representation of the temporal dynamics.

This paper proposes an approach to overcoming environmental data heterogeneity by creation of a common data type: a spatial temporal event and presents a graphical exploratory framework for investigating the spatial temporal behaviors of such events. The approach to event detection in sensor data streams is outlined using
Gulf of Maine Ocean Observing System (GoMOOS) data. The exploratory framework shares properties of TRELLIS graphics (Becker and Cleveland 1996) and the measurement protocol described by Sinton (1976) where given the dimensions of space theme, and time, one dimension is held constant, a second is permitted to vary in a controlled manner and the third is measured for variation within the second level controlled attribute. The exploratory environment supports comparison of patterns of events in both time and space and investigation of event relations across space and time scales.