Automated interpretation of digital landscape models

The purpose of this paper is to provide an overall view of the methods of spatial data mining and its applications to digital landscape models. Spatial data mining can be defined as the deduction of information that is not explicitly stored in a given spatial data model. The subject spatial data mining represents the integration of several fields, including machine learning, database systems, data visualization, statistics, information theory and computational geometry. The automation of spatial analysis functions has two main aspects. On the one hand it deals with the automation of spatial operators conventionally used in a GIS-program for complex analysis applications, e.g. site planning. Such an application usually involves a sequence of operations, e.g. classification, buffering, selection, etc. This process is controlled by the human operator according to a ''model'' he has in mind. Automation of such a process requires to make explicit this model and apply it to the data. In this way, e.g. a model for site planning can be created. The second aspect concerns data mining. This approach is used to find connections in the data which are not known in advance therefore no model exists which are however implicit in the data.