Diverse GIS applications in environment, natural resource, agriculture, landscape transportation, landuse, and urban planning require varying degrees of detail for spatial analysis according to the extent. Multi-level geographic data abstraction to implement diverse spatial analyses is very important, and is a major research issue in the field of GIS.
Multi-level data abstraction in GIS has its origin in multi-scale mapmaking, a task previously undertaken manually by cartographers, following theoretical and practical concepts. Investigations of simplification algorithms for digital cartography began in the 1960s, but it was not until the 1980s that research and development of automated map generalization systems using such algorithms began. Research on multi-scale GIS and database integration built with geographic data of varying scales began in the 1990s.
These studies focused on the geometric generalization of geographic data to change the degree of detail. However, multi-level geographic data abstraction must deal with geometric generalization of features, process accompanied with modification of spatial and aspatial data, topology, and relationships between classes. Moreover, the processes used in abstraction have. a preferential ordering that increases efficiency and, reduces loss or distortion of information.
This study focuses on spatial database whereas previous studies for generalization deal mostly with geometrical presentation or topology model. The purpose of this research is to find methods for the generation of a high level database by abstracting a more detailed spatial database. For this purpose, the following issues were studied.
Firstly, the changes that take place on a procedure of spatial database abstraction were systemized into spatial database structure and its' contents. A conceptual framework of spatial database abstraction was developed based on it. The framework is composed of schema evolution, spatial data generalization, and aspatial data derivation. The difference from traditional map generalization concepts is that schema changes as an abstraction factor were evolved. To abstract original spatial database into target database not only simplification of geometrical representation of object but also schema changes that include changes of class, object, spatial and aspatial data, topology, semantic and spatial relation are needed.
Secondly, applying this concept, operators for spatial database abstraction were defined such as Preselection, Elimination, Collapse, Aggregation, Classification, and Simplification. Preselection is an operator that defines schema using correspondent structures between original database and target database on a class level. Elimination is an operator that removes features unsuitable to get visual effect desired and analytical purpose according to contents of feature abstraction on an instance level. Collapse is an operator that reduces a level of dimension to transform object types represented. Aggregation and Classification are operators that resolve several spatial features into a feature or a class. Simplification is an operator for simplifying geometrical shape of lines of an object. The most important issue on the process of change of spatial database abstraction is to maintain consistency in topological, spatial, and semantic relationships with original database. The principle of maintenance of consistency was applied to the definition of operators, and to the implementation process of operators, too. It is an important difference of this study from other studies.
Thirdly, basic workflow for individual object was suggested according to the criteria (object type and properties) on operator selection that is applied to a target object of abstraction, and the application criteria (minimization of data loss and distortion, and maximization of efficiency of data processing) of operator selected.
Fourthly, prototype system was developed to confirm the validity of the suggested method in this study, and the results of experiments were evaluated. Spatial databases using for experiment were 1/1000 scale digital topographic map and 1/25000 scale digital topographic map provided by National Geography Institute, and cadastral map. Prototype system was developed using spatial data operation functions in Arcinfo8 GIS. Evaluation was carved out by comparison of the validities of the results from basic workflows and alternative workflows. The experimental results showed that the possibility of data loss and distortion for alternative workflows that has different sequences of operators from basic workflows is higher than for basic workflows. This means that basic workflows are superior to alternative workflows.
This study is expected to contribute for several aspects as follows:
First, this approach allows the multi-level spatial database abstraction to generate spatial databases suitable for diverse GIS applications according to the users requirements. Spatial database is generally built using data from various maps. During the procedure, maps that have similar contents but different scales are often included. It is time consuming and costly works to input diverse scales of maps and their attribute data. Various sources of map data also provoke inconsistency between them. Multi-level spatial database abstraction can save lots of time and cost, supply spatial database wanted by users with timely manner, prevent users from inputting similar in content but different scale maps in duplicate.
Second, continuous changes of geographic information should be reflected to various scales of maps or spatial database with almost real time. Current methods of individual update make inconsistency between data, are time consuming, and require lots of cost. If we more study on the automatic update of the target database based on the defined processes of the spatial database abstraction, it will allow the user to reduce database maintenance costs, and to keep data consistency.
Third, the findings of this research are expected to provide important specifications for the development of CASE tools to produce multi--level spatial databases suitable for diverse GIS application. Specifically, this study will contribute to the development of efficient methods for mapping diverse scales of thematic maps.
