ABSTRACT
This chapter proposes spatiotemporal constructs and a conceptual framework to lead geospatial knowledge discovery beyond what is directly recorded in databases. While knowledge discovery is fundamentally a data-driven approach to elicit novel, previously unknown patterns from massive, heterogeneous data, what we can discover from a database is constrained by what can be conceptualized and therefore represented in the database. Just as analytical possibilities for the data depend on the chosen representation schemes (Miller and Wentz 2003), the knowledge that can be discovered from data records is limited to patterns and rules of the data objects represented in the employed data models. Conventional geospatial data models adopt space-centric representations. Geospatial facts are recorded based on geometry and location, and, therefore, the knowledge that can be discovered is constrained to patterns and relationships derived from geometry and location.
