Linked data provide an alternative route for dissemination of spatial information as compared to the traditional SOA-based SDI approach. Where the latter is built on predefined structuring of semantics within domains, linked data is open to linking information to any data over the Web. In this respect both are complementary. The traditional approach providing a mechanism for a basis of standardized and structured data within domains, and linked data providing an open mechanism for sharing and combining. GML as the ISO standard for exchange of service based spatial data and RDF as the linked data format are therefore related. GML provides the format in which many spatial datasets are available and exchanged. This standardization process and effort has been realized on a large scale. Why not let the web of linked data take advantage of this effort? This article will focus on the use of GML structured data as a source for deriving RDF structured data. The first part of the paper focusses on deriving linked data from GML data. The first version of GML, v1.0, was based on RDF. From version 2.0 onwards GML was based on XML and XML Schema, but the object-property structure was retained. We describe a transformation for translating any correctly structured GML to RDFS/OWL automatically, using XSLT. Because GML’s object-property structure translates very well to triples, the transformation is straightforward. Well-known GML content elements such as names and descriptions are mapped to their RDF equivalent. However, any semantics specific to the input GML data (a.k.a. the application schema) are ignored in this translation. In the second part, we study how more meaningful RDF can be created from GML, given the underlying information model, by transforming it from UML to RDFS/OWL. There exists a straightforward mapping to convert a UML model into a RDFS/OWL vocabulary. However, the re-use of existing concepts in vocabularies takes a central role in RDFS/OWL while in UML the use of vocabularies is not supported. We describe how annotating the UML model could improve this translation.