Relationale und objektrelationale Datenbankkonzepte in Geoinformationssystemen

The management of spatial data in relational databases provides a number of advantages compared to traditional (separated) approaches where geometry and metadata are kept in file systems as opposed to alphanumeric data that reside in relational databases. Above all relational databases ensure efficient spatial data management by providing multi-user support and better control of spatial data quality. In addition relational database approaches lead to significant improvements for wide usage and availability of geographical information and are also better suited for maintenance and update processes. By extending Relational Database Management Systems (RDBMS) to support object-oriented concepts, reliable object-relational database management systems (ORDBMS) have become available. This development poses two interesting questions. What is the added value and what are the limitations for practical applications of Geographic Information Systems (GIS) and for geographic information science in general? Today special built-in extensions by leading ORDBMS vendors for storing and analysing spatial data can be combined with GIS products and these combinations already provide a reliable platform for efficient spatial data management. However a full integration of ORDBS and GIS which would allow for immediate utilization of additional modelling potential for GIS application development has not yet been thoroughly researched. Starting from an analysis of the conceptual requirements of GIS users this research investigates the potential for achieving added value through ORDBMS and it examines the remaining deficits of integrated geospatial data management concepts by comparing RDBS based approaches with ORDBS based strategies. Empirical research and selected GIS applications / analysis are exploited to evaluate and compare the capabilities of different data modelling and data management concepts to handle geometry data, alphanumerical data and metadata. This research describes the database structures resulting from the migration of real data into particular data (management) types. Concepts to optimize usage, maintenance and update processes of central spatial databases are developed and it is pointed out how geospatial objects can be entirely modelled to achieve complete integration. This research classifies GIS applications where the migration to more sophisticated data management concepts prevails over the costs being involved. Limitations which are presently remaining are identified. Based on the discrepancy between conceptual GIS requirements and their mappings into logical and physical database models of GIS and database vendors, the need for further research is outlined. In addition recommendations for future GIS and database development are given. The system platform used for this research is based on Oracle 9i, including the Oracle Spatial extension, ESRI’s middleware component ArcSDE and ArcInfo 8.3. The Bavarian Forest National Park was selected as the test area mainly because complex spatiotemporal processes such as the spread and distribution of deadwood after bark beetle attacks can be observed here. This condition makes the test area perfectly suited to meet the objectives of this research and it allows to put an emphasis on the comparison of the relational and object-relational database modelling approaches with respect to adaptation of user-driven conceptual requirements in general and the entire modelling process of geospatial objects in particular.