Current methods for specifying data models lack well-defined descriptions of the expressions, which represent data types, attributes, attribute values, and operations. It is impossible to define clear translation rules for data models, because the relation between source and target data model elements is not computable. Furthermore, translation has to account for imprecision caused by conceptual heterogeneities and measurement error. In this thesis we use the DOLCE foundational ontology to provide a semantic reference frame for geospatial data. Available extensions to DOLCE are profiled and additional geospatial characteristics, such as topological relations, are included. Annotating (or semantically referencing) expressions, which are used to define geospatial data models, with this frame supports computability and allows for selecting appropriate translation rules on the attribute level. Using a logic-based approach, semantically referenced data models allow for inferring relations between source and target attributes. This includes inference on applicable translation operations and the detection of match types (exact, upper bound, and lower bound) between attributes. If detected match types fit the user’s purpose, translation scripts are extracted. The scripts are executed using an algebraic theory, which includes propagation of measurement errors. The approach allows for specifying data model semantics and imprecision. A demonstrator is provided as proof of concept. Our research is guided by an example of translating information about road width from a national data model (ATKIS road data model) to an international one (INSPIRE Data Specification for Transport Networks).