Developing error handling software for object-oriented geographical information

The inclusion of error handling capabilities within geographical information systems (GIS) is seen by many as crucial to the future commercial and legal stability of the technology. This thesis describes the analysis, design, implementation and use of a GIS able to handle both geographical information (GI) and the error associated with that GI. The first stage of this process is the development of an error-sensitive GIS, able to provide core error handling functionality in a form flexible enough to be widely applicable to error-prone GI. Object-oriented (OO) analysis, design and programming techniques, supported by recent developments in formal OO theory, are used to implement an error-sensitive GIS within Laser-Scan Gothic OOGIS software. The combination of formal theory andGIS software implementation suggests that error-sensitiveGIS are a practical possibility using OO technology. While the error-sensitive GIS is an important step toward full error handling systems, it is expected that most GIS users would require additional high level functionality before use of errorsensitive GIS could become commonplace. There is a clear need to provide error handling systems that actively assist non-expert users in assessing, using and understanding error in GI. To address this need, an error-awareGIS offering intelligent domain specific error handling software tools was developed, based on the core error-sensitive functionality. In order to provide a stable software bridge between the flexible error-sensitive GIS and specialised error-aware software tools, the error-aware GIS makes use of a distributed systems component architecture. The component architecture allows error-aware software tools that extend core error-sensitive functionality to be developed with minimal time and cost overheads. Based on a telecommunications application in Kingston-upon-Hull, UK, three error-aware tools were developed to address particular needs identified within the application. First, an intelligent hypertext system in combination with a conventional expert system was used to assist GIS users with error-sensitive database design. Second, an inductive learning algorithmwas used to automatically populate the error-sensitive database with information about error, based on a small pilot error assessment. Finally, a visualisation and data integration tool was developed to allow access to the error-sensitive database and error propagation routines to users across the Internet. While a number of important avenues of further work are implied by this research, the results of this research provide a blueprint for the development of practical error handling capa-