Building up spatial data infrastructures involves the task of dealing with heterogeneous data sources which often bear inconsistencies and contradictions, respectively. One main reason for those inconsistencies emerges from the fact that one and the same real world phenomenon is often stored in multiple representations within different databases. It is the special goal of this paper to describe how the problems arising from multiple representations can be dealt with in spatial data infrastructures, especially focusing on the concepts that have been developed within the Nexus project of the University of Stuttgart that is implementing an open, federated infrastructure for context-aware applications. A main part of this contribution consists of explaining the efforts which have been conducted in order to solve the conflicts that occur between multiple representations within conflation or merging processes to achieve consolidated views on the underlying data for the applications.