ABSTRACT
The concept and effects of the modifiable areal unit problem (Openshaw and Taylor, 1981; Openshaw, 1984) have been well documented in quantitative geography (Fotheringham and Wong, 1991; Amrhein, 1995). Fundamentally, if areal units are imposed onto a discrete geographical distribution for the purpose of data aggregation, then the resulting areal values will be conditional on the locations of the boundaries. This situation frequently arises, for example, when administrative boundaries are imposed onto the distribution of a human population. The modifiable areal unit problem comprises two separate issues, known as the scale and aggregation problems: aggregated values will vary according to the scale of aggregation, and at any given scale, will vary according to the particular boundary configuration chosen. Even the simplest real-world areal aggregation schemes present a massive number of possible alternative configurations. In conventional choropleth (shaded area) mapping, the problem is one of map interpretation (Monmonier, 1996), but in GIS applications, the implications are more complex, due to the multiple and unpredictable uses to which the area data may be put. Recognition that the problem exists leads directly to the possibility of adapting the design of areal units to provide the best configuration for any particular application. Defining and obtaining the best configuration becomes computationally intensive as the number of areas rises above very small numbers, and this chapter presents a range of research which deals with automated zone design in a GIS context. Despite being both geographical and computational, zone design applications do not fit neatly within the definition of geocomputation expressed by Longley (1998), for example, who identifies ‘research-led applications which emphasize process over form, dynamics over statics, and interaction over passive response’. Nevertheless, these procedures are clearly in keeping with Openshaw and Alvanides’ (1999) view of geocomputation as ‘the application of a large-scale computationally intensive approach to the problems of physical and human geography’ (p. 270).
