The paper presents a schema repository, an original repository containing different kinds of database schemas. The repository is part of a multidisciplinary approach for schema evolution called the predictive approach for database evolution. The schema repository has a dual role in the approach: (1) during the datamining process, the repository identifies and analyzes trends on collected schemas belonging to the same domain. (2) the repository is used in the building of the requirements ontology - a domain ontology that contributes in the database design and its evolution. This paper presents both the design and a heuristic-based method to populate such a repository.