Abstract: A hybrid spatial clustering method based on graph theory and spatial density(HGDSC) is developed.The HGDSC method employs Delaunay triangulation to model the spatial proximity relationships among spatial entities and the modified density-based clustering method,considering the similarity of both geometric distance and non-spatial attribute.Normally,the method can adapt to a spatial database which contains clusters of arbitrary shapes,non-homogeneous densities and/or large amount of noise.Only one input parameter is required.Experiments on both synthetic and real-world spatial dataset are utilized to demonstrate the effectiveness and advantages of the HGDSC method.