Abstract: The complexity and richness of spatial data models and geographic information systems (GISs) together with the popularity of WWW have led to an increasing number of heterogeneous and autonomous geo-referenced information sources that spread over Internet.To integrate and publish geographic information, a challenge is to match schemas between two arbitrary GML documents.In this paper, an algorithm for automatic matching of two GML schemas is proposed, in which document schema is first transformed to a tree, and then a similarity formula is introduced to measure the semantic and structural matching degrees of two documents based on the transformed trees.The proposed algorithm is implemented in a running GML-based GIS.