Purpose: Designing metadata ontology model for semantic representation of Theses by using the SPAR (Semantic Publishing and Referencing) Ontologies. Method: This study was an applied form and two methods were used, Content Analysis and mapping. The metadata of 69 theses and dissertations on the National Library and Archive of Iran in three Databases: 1) Digital Library of National Library and Archive of Iran. 2) Rasa Software and 3) Ganj in the Iranian Research Institute for Information Science and Technology were selected modified and completed by mapping. On the other hand, by analyzing the entities of each SPAR ontology and suggesting another entity to the researcher, the checklist was formed. This checklist included classes, properties, and individuals. At last, by entering them into Protégé software version 5.5, the model of metadata ontology, MdOntTDs, was drawn. Findings: Findings identified deficiencies in the existence of four important metadata elements (subject, supervisor, advisor, and abstract) in RASA and NLAI Digital Library. Among the 18 SPAR Ontologies, the most entities were selected from FaBiO, FRAPO, and CiTO respectively. All entities of BiDO, BiRO, C4O, Fivestar, FR, FRBR, PO, PRO, PSO, and PWO were suitable for theses. 195 individuals from 6 SPAR Ontologies, 292 individuals labeled with MdTDs from theses, and 100 individuals labeled with SUNMdTDs were selected by the researcher and entered into the software. 1558 entities categorized by class, Properties (object, data, and Annotation), and individuals along with the description and definition of each entity were placed in the software, in the form of hierarchical and determining axioms for classes. And specifying domain and range for relationships. Finally, the RDF graph was drawn using the OntoGraf plugin, and the final Model, MdOntTDs was developed. this research has proposed three new types of metadata: 1) Except for the existing keywords, topics have been categorized and modeled up to three levels including 4 main categories, 16 subcategories, and many units. Each of these final topics has been related to "hasSubject" and "isSubjectOf" properties. 2) The research methods of Theses that were connected with "hasMethod" and "used in" properties. 3) The papers taken from Theses were also searched, as far as possible, and were connected with "hasJournalArticle" and "journalArticleOf" properties. Conclusion: This model, if implemented, can overcome keyword search limitations, the problem of linking and Data sharing on the web, and the inconsistency of data. In the software, classes and their related individuals are visible in the form of a hierarchical network in RDF triples, and the connection between entities with increasing access points promises deeper semantic searches. However, due to the absence or lack of tagged and linked data, usage of the some of selected entities is not possible. [ABSTRACT FROM AUTHOR]