 |
Jiang Bin
Fuzzy Overlay Analysis and Visualization in Geographic Information Systems
Datum Date 23 februari 1996
Promotoren: F.J. Ormeling & W. Kainz (ITC)
|
English summary (chapter 9)
We have developed a fuzzy overlay model and visualization techniques aimed at improving the functionality of existing Gist. The purpose has been to find an appropriate approach for fuzzification and combination, and to develop visualization techniques that facilitate fuzzy overlay analysis. By coloring uncertainty, a modified HLS color system has been applied to the visualization of uncertainty. Perception investigations on the ways in which users interpret colored uncertainty information are discussed.
9.1 Major Contributions
The major contributions covered in this research are summarized below and
issues requiring future research are identified.
9.1.1 Contributions to the Fuzzy Overlay Analysis Model
Starting with fuzzification, a detailed fuzzy overlay analysis model has been developed. It has identified fuzzffication as a basis for fuzzy overlay analysis, and studied all possible membership functions for ftmification. An S-shaped membership function with two parameters to adjust the shape of membership curves is recommended.
Several combination solutions have been developed for the purpose of overlay operations. These include multiple variables membership function, use of interval analysis, and of fuzzy weighted average. Particularly, interval analysis provided good possibilities for finding combination solutions.
9.1.2 Contributions to the Visualization of Uncertainty
In contrast to the layer, the sublayer (as an uncertainty map) is defined and used for the visu@ation of unce@ty information. It describes the fuzzy aspect of reality, and is beneficial in fuzzy visual thinking.
A modified HLS color system for uncertainty representation has been developed. It provides advantages over other system in coloring uncertainty information. The basic principle for coloring uncertainty has been put forward, i.e., the variable hue used to represent the category and the variables saturation and lightness used to represent the uncertainty infonnation.
A system framework has been proposed. This is not only appropriate to fuzzy overlay analysis, but more general fuzzy spatial analysis as well. It provides the basic framework for doing fuzzy spatial analysis.
Akin to Bertin's visual variables, a set of exploratory acts (i.e., blink, highlight, zoom, pan, drag and click) for exploratory visualization has been drawn up, by which potential structures and pattems can be relatively easily explored. More importantly, these acts facilitate visual thinking.
9.1.3 Contributions to Perception Investigation in Coloring
Uncertainty Information
To color uncertainty information, a set of color variables has been suggested. It includes hue, saturation, lightness and random dots, as an extension of Bertin's system of visual variables. Moreover, lightness is further distinguished as lightness(+) and lightness(-) because one end is white and the other is black. The random dot scale was found to be the best for the construction of bivariate schemes.
To examine the users' response, a set of perception tests was conducted. 'Mis showed that, instead of the commonly accepted variable of saturation, lightness(+) is an intuitive color scale for uncertainty information. Rivariate map schemes were also considered for use in coloring combined uncertainty, and the randomly encoded scheme was found to be the optimum choice.
9.2 Future Research
In order to improve the functionality of Gis in spatial analysis and decisionmaking, fuzzy spatial analysis and visualization techniques need to be integrated. The design of such a system is a very complex undertaldng. As such, this thesis certainly cannot deal with all the problems that must be solved. Research topics that need further study are identified below.
9.2.1 Extension of Fuzzy Overlay Analysis
Although the fuzzy overlay model proposed is oriented towards linguistic notions, it is not limited to them. It can be extended to other cases in which fuzzy classification is used. For instance, if a fuzzy classification is conducted from remotely sensed images, a set of sublayers (e.g., on lakes, and vegetation) will be created. Overlaying these sublayers may generate a series of combinations. The results are useful for decision-makers. More importantly, once these sublayers are efficiently visualized, they provide more intuitive communication effectiveness for visual thinking. In contrast to spatial statistical analysis, fuzzy spatial analysis offers another tool for decision-making.
Next to the traditional map, the sublayer is an uncertainty map which reflects the fuzzy aspect of reality. It describes reality from another perspective, which may approach human thinking. This has been shown consistently in the research leading this thesis. A fuzzy classification system is developed and this results in sublayers. Measurements become fuzzy with a fuzzy buffer. In connectivity operations, more complicated fuzzy buffer zones are generated. Neighborhood operations may become fuzzy search operations.
In addition to the combination and semantic operation introduced in the fuzzy overlay analysis, some further spatial pattems can be recognized: spatial autocorrelation, spatial association, distance decay and social pattems as suggested by Openshaw (1994), although his original intention was for them to be applied to spatial statistical analysis. The sublayer can be viewed as an independent spatial pattem. Based on this some derived pattems can be worked out. In this context, a considerable number of statistical approaches can be applied to uncertainty (fuzzy numbers) to discover potential patterns and structures. Spatial autocorrelation and association leave more room for developing fuzzy spatial analysis.
9.2.2 Research on Multimedia Representations of Uncertainty Information
As mentioned in the preceding chapters, multimedia representations provide many possibilities to visualize uncertainty, but there has been no @er investigation on how human beings perceive these representations. Further research might be done in the following aspects: sensory variables for uncertainty visualization, comparative study on multimedia representations, visualization tools for uncertainty exploration.
9.2.2.1 Sensory Variables for Uncertainty Information
Color is considered as a visual variable for uncertainty inforfnation, but it has been recognized that sound and animation can be used for the visualization of uncertainty spatial information as well. Primary discussion on this matter was initiated by Fisher (1994).
Uncertainty can be sensed in the form of noise. Pitch is one of the sound variables (Krygier 1993), which can be used in 'visualizing' uncertainty. Pitch (the frequency of a sound), is analogous to the visual variable lightness, and as such can represent the level of uncertainty. As an important complement of visual variables, pitch provides an alternative method for uncertainty visualization, particularly in an already crowded display.
Frequency is a dynamic visual variable to control the speed of blinking, not included in Bertin's system of visual variables. The variation of frequency can be used to represent the level of uncertainty. Zero Frequency means static without blinking; any other magnitude of frequency has a ce@ unsteadiness. Compared with other visual variables, it can be regarded as dynamic and can be easily implemented in the computer environment.
Following up these studies, an extension of Bertin's system in the form of a set of sensory variables for uncertainty information might be extracted and applied to visualization.
9.2.2.2 Comparative Study of Visualization Schemes
Not only traditional cartography but also advanced multimedia representation
provides a wide range of seniiologies to represent uncertainty infonnation.
Consider, for instance, 3D representation, sound and animation. However, from the point of view of perception, it is impossible that all these schemes wffl always be obvious to everyone. The benefits of realism have been recognized in visualizing spatial data; the same considerations should be true for uncertainty infonnation. The difference of semiology in perception presents enough evidence to conduct comparative studies on different visualization schemes.
9.2.3 Visualization Tools for Uncertainty Exploration
From the Point of view of the coupling of fuzzy spatial analysis and visualization, flexible visualization tools are necessary to enable an analyst to form hypotheses and make decisions. Proposed exploratory acts need to be implemented searnlessly for uncertainty exploration.
Hyperstructure mechanisms should be considered in visualization tools. Different visualization schemes should be linked together for cross referencing. Thus, for the same set of uncertainty values, instead of an optimum scheme a range of schemes can be provided for visual interpretation.
Another significant aspect is the user interface for the visualization tools. Flexible user interfaces should, to a large extent, simulate the way in which geoscientists made spatial analysis in the pre-computer era. For instance, dragging two individual maps together leads to an overlapped map.
9.2.4 Application Studies in Environmental Monitoring
Uncertainty is ubiquitous in environni ental monitoring and resource management. Unlike sociopolitical phenomena where the boundary is clearcut and widthless, most natural-phenomena only exist as a fuzzy transition from non-membership to full membership. For instance, in the tropical regions, the boundaiy between a lake and a swainp is extremely fuzzy, and a lake is often surrounded by a swamp. We can not be certain that a given location belongs to a lake or a swainp; we can, however, ensure with a certain percentage of certainty that the location belongs, for instance, to a lake.
It is clear that fuzzy overlay analysis and visualization have great application prospects in environmental issues (Jiang 1995). However, further research is required.
