Our work aims at developing an integration method for spatio-temporal data that focuses on conceptual level specifications. It’s position is somehow in between the highly abstract methods using formal ontologies to resolve data heterogeneity, as in [1], and the real-world instance based methods, as in [3]. The core part of our method is the set of semantic correspondences that are formulated for element pairs of the database schemas that model related real-world objects. As the common data model we employ the MADS [2] conceptual data model, which was designed to fulfill the requirements for modeling of spatial and temporal data.

Evolving from the relationships between real world sets of related objects our method takes into account the relativism of conceptual representation and employs the notion of the multiple instantiation class sets. Our method is intended for geodata, and it supports correspondences between objects’ spatial and temporal features. For the integrity issue of interoperable systems we propose an algorithm for consistency checking. Semantic correspondences that are established for the source data sets are checked for compatibility against the integrity constraints imposed on the same data. To ensure a meaningful integrated solution even for the cases of greatly diverse representations of related data, we employ a multi-representation solution that consistently preserves the initial representations on the integrated level.