Provenance tracking for Linked Data requires the identification of Linked Data resources. Annotating Linked Data on the level of single statements requires the identification of these statements. The concept of a Provenance Context is introduced as the basis for a consistent data model for Linked Data that incorporates current best-practices and creates identity for every published Linked Dataset. A comparison of this model with the Dublin Core Abstract Model is provided to gain further understanding, how Linked Data affects the traditional view on metadata and to what extent our approach could help to mediate. Finally, a linking mechanism based on RDF reification is developed to annotate single statements within a Provenance Context.