ABSTRACT: The problem of generalizing spatial data sets has been studied in the context of categorical data. Map Algebra, a deterministic modeling language, has been chosen as the modeling language for the generalization transformations. The methodology is based on simple basic Map Algebra functions, which are brought together and run as a sequenced batch file. From the basic Map Algebra functions several individual generalization transformations have been composed; the three most important are described extensively in the paper. The two major practical projects implemented with the automatic generalization transformations are presented. Finally, the results of the methodological development for generalizing categorical databases are presented and the prospects for extending the methodology are discussed.
KEYWORDS: Generalization, multi-scale, Map Algebra, modeling, quality
Motto: The world should be filled with all kinds of multi-scale databases.
A remarkable change during our lifetime has been the diffusion of computers and computerized processes and methods. Computers are entering nearly every aspect of our lives, and mapping and cartography have, naturally, not been an exception in this diffusion process. The effect of computers on the analyses of map data and the ease of bringing together different data sources has weakened our awareness of some of the basic properties of spatial data and processes. When geographical information systems (GIS) were introduced in mapping the importance of map generalization was undervalued, GIS users ignored the concept of scale to some extent (see Aspinall 1995), and thus the contention arose that totally scale-free data were possible within GIS. However, the awareness and consequences of misusing map data have steadily increased and there is now a tendency to study and develop methods for integrating spatial data sets at different scales and geometric qualities. The need for generalization as a task has become more and more evident, and the need for automating this task has been understood. Until now, the national mapping agencies in most countries have been forced to produce and store multiple-scale versions of topographical map databases. This is because there has been no production tool for generalizing the multiple data sets and no propagation tool for updating derived data sets, and the processes for regenerating data sets have been expensive and time consuming (see Muller et al. 1995). In this research, the aim has been to develop a methodology for fulfilling some of these needs, especially for the generalization of area features in a nominal categorical coverage.
Maps represent spatial phenomena in units of supposedly uniform content and dear boundaries. One of the key processes in map making is generalization, which is a process that tries to retain the most important geographical characteristics while at the same time reducing the complexity of the world phenomena represented in a map. Another generalization objective is to keep the cartographic portrayal of the mapped data consistent with the chosen purpose and intended audience (see Joao 1991).
Cartographic generalization can be used for producing either graphical outputs, i.e. maps, or for producing spatial data sets for geographical or statistical analysis. In both cases we may use similar generalization transformations, although we cannot optimize the data set for all purposes. It is therefore good to keep in mind that generalized maps are not always good input data for certain types of spatial analysis, a fact that is also related to the claim that we should always use the most accurate and ungeneralized data available in the spatial analysis. In practice, the difficulty in using accurate large-scale data sets is that the data are often not available, they are too large in volume, or else the different data sets used in spatial analysis are at different levels of accuracy and generalization. …