The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008 
142 
discovering and retrieval is made out. Finally, we make 
conclusions and discuss the future work need to do. 
2. RELATED WORKS 
Ontology based research for discovering and retrieval of 
geographic information has been booming. There are many 
research works done to make efforts to it. 
Klien et al (2004) presented an architecture for ontology-based 
discovery and retrieval of geographic information that can 
contribute to solving existing problems of semantic 
heterogeneity. So far they have defined components such as 
Enhanced Cascading Catalog Service and the Reasoner 
component and plan to develop a query scenario in which the 
user is able to formulate a question using terms from the 
familiar shared vocabularies. 
The research project SPIRIT (Spatially-Aware Information 
Retrieval on the Internet) developed tools and techniques to 
support spatial search on the internet based on ontology (Jones 
et ah, 2002). Geographical ontologies are constructed to assist 
spatial search (Fu, 2005a). An ontology-based spatial query 
expansion method is developed that supports retrieval of 
information relevant to space by trying to derive its 
geographical query footprint (Fu, 2005b). 
Hartwig H. Hochmair (2005) proposed a conceptual framework 
to overcome problems of semantic heterogeneity in keyword 
based retrieval of geographic information. In the architecture, 
the server-sided knowledge base including domain ontology 
and rules for query expansion is used to expand the keyword- 
based searches. 
Wiegand and Garcia (2007) developed a task-based ontology 
approach to automate geospatial data retrieval. In the approach, 
ontologies of task, data source, metadata and place, along with 
relationships between them are developed. With the ontology, 
reasoning can be done to infer various types of information 
including which data sources meet specific criteria for use in 
particular tasks. 
In this paper, we focus on uniform semantic descriptions of 
geographic information and approaches for discovery and 
retrieval based on semantic descriptions. 
3. ONTOLOGY-BASED SEMANTIC DESCRIPTION 
MODEL 
In this section, we set forth an ontology-based semantic 
description model that explicitly represents semantics of 
geographic information in uniform machine readable and 
understandable format. Thus, problems of semantic 
heterogeneity in description of geographic information are 
solved. 
3.1 Ontologies 
Ontologies are the key to semantic description of geographic 
information. In this paper, several ontologies are constructed to 
explicitly model knowledge for geographic information such as 
metadata ontology, fundamental geographic information 
ontology, spatial relation ontology, geometric ontology. They 
all belong to geographic information domain ontology and 
model non-spatial semantics and spatial semantics of 
geographic information. 
Metadata ontology explicitly represents the knowledge of 
geographic information metadata. It not only formalizes 
existing metadata specifications such as ISO-19115 (ISO 
TC/211, 2003) and CSDGM (FGDC, 1998) in Ontology Web 
Language (OWL) (W3C, 2004), but also builds mappings 
between them. Therefore, it makes various metadata according 
to different specifications be able to interoperation in semantic 
level. 
Fundamental geographic information ontology is created to 
describe non-spatial semantics of geographic features. It 
defines fundamental geographic feature concepts and 
relationships of them. 
Figure 1. A light ontology for example 
Spatial relation ontology and geometric ontology are built to 
describe spatial semantics of geographic features. Spatial 
relation ontology defines topological relations, direction 
relations and distance relations. Geometric ontology defines the 
concepts and relations of geometric point, line, polygon, 
surface, etc. 
These ontologies are built with a “down-up” abstracting 
method. First, key concepts in geographic information science 
domain are abstracted. Then, concepts relationships, geometric 
relationships, spatial relationships, location relationships are 
modeled and established. Concepts relationships include 
“subclass of’, “part-whole”, “member of’, “instance of’ and 
“dependency”, “reference”. Geometric relationships define the 
composite relationship among point, line, polygon and surface. 
Spatial relationships define topological relations, direction 
relations and distance relations. Finally, OWL DL (W3C, 2004) 
is adopted to describe these ontologies and test their 
consistency by semantic reasoning engine RACER (Racer 
Systems GmbH & Co. KG, 2005). 
3.2 Ontology-based semantic description model 
Metadata is an important part in SDI to describe geographic 
data for data discovering and sharing. It facilitates data sharing 
among Geographic Information Communities (GIC) in 
distributed environments. However, metadata is not enough to 
solve problems of semantic heterogeneity because different 
GICs publish metadata according various metadata 
specifications and various vocabularies. For solving problems 
of semantic heterogeneity, we proposed an ontology-based 
semantic description model to represent semantics of 
geographic data.