Guest Editorial Semantic reference systems

Four centuries after René Descartes watched a fly walk across his ceiling and wondered how to capture its position (Gribbin 2002), we use Cartesian coordinates routinely to describe locations. We identify the positions of entities in the real world, transform their GIS representations from one coordinate system to another, and integrate spatially referenced data across multiple coordinate systems. A theory of spatial reference systems standardises the notions of geodetic datum, map projections, and coordinate transformations (ISO 2002). Similarly, our temporal data refer unambiguously to temporal reference systems, such as calendars, and can be transformed from one to another. Geographical information systems contain spatial, temporal, and thematic data. The first two kinds are firmly tied to reference system theories and tools. We now need to produce the third component—semantic reference systems. Descartes might wonder today how to establish common frames of reference for, say, a geneticist and an entomologist to talk about that fly. They would need methods to explain the meaning of their specialized vocabularies to each other, to detect synonyms and homonyms, and to translate expressions. Accordingly, a theory of semantic reference systems will enable producers and users of geographical information to explain the meaning of thematic data, to translate this meaning from one information community to another (OGC 1998), and to integrate data across differing semantics. Reaching this goal may not take centuries again, but a decade or two could pass until we see the kind of on-the-fly (so to speak) integration that is now possible across different spatial reference systems in GIS (Lutz 2003). There is no need for universal geographical information semantics, as long as we can provide the means for any pair of information communities to define their concepts and translate between them. All the same, work on semantic primitives and universals (Wierzbicka 1996) as well as on top-level ontologies (Sowa 2000) strongly suggests that a common core of concepts exists and can be defined. Formalizing the semantics of geographical information communities is, in any case, much simpler than defining natural language terms. The reason is that every information community agrees, per definition, on a shared set of concepts, expressed in conceptual models, feature-attribute catalogues, legal documents, work practices,