Scale as artifact: GIS, ecological fallacy, and archaeological analysis

Abstract

The concept of scale is confusing, frustrating, little understood, and yet . . . intriguing. Geography, as the discipline with spatial pattern and relationships as its core focus, has traditionally embraced and confronted the meaning, interpretation, explanatory power, and some would suggest, intractability, that scale brings to studies. There are many other spatially related disciplines such as archaeology that share with geography the need to operate across a scale continuum ranging from the local to the regional, national, and international levels. Indeed, as Mcmaster and Sheppard (2004) point out, it is difficult to identify a completely ''scaleless'' discipline. Our understanding of geographical and archaeological patterns, societal processes, and spatial heterogeneity, are highly dependent on scale. Because of the extent and complexity of the Earth's surface, researchers must invariably sample, generalize, or aggregate in order to comprehend reality. Identifying an appropriate scale of data capture and analysis to use in a study, and managing the trade-offs that must occur in matching resources with data capture, have long been acknowledged. But while scale is readily associated with the level of detail involved with geographical description, what is missing from most studies is any discussion of how representative that scale of analysis is, given that there are no standard measures of uncertainty related to particular scales of analysis (Tate and Atkinson 2001). Indeed, Meentemeyer (1989) suggests that it seems that study scales are selected unconsciously and therefore may seem to be completely arbitrary. Watson (1978) observes that we tend to work at one analytical level exclusively and implicitly, without considering other alternatives, almost as an act of faith. Scale issues associated with line generalization, scale of data capture and data display, and with the even less well-known issues of ecological fallacy and its derivative, the Modifiable Areal Unit Problem, are only rarely acknowledged. Geographic Information Systems, GIS, facilitate an almost effortless integration and display of multi-scale data and aggregation of areal units. Generating data for GIS analysis invariably requires extensive effort and introduces a slew of issues associated with data capture, data accuracy, data resolution, error estimation, metadata, data structure, storage and compression, and data sharing to name but a few. The problem of identifying an appropriate scale of study when using GIS is compounded by the need to utilize, wherever possible, data created by others in order to reduce data duplication and replication and to minimize the cost of populating the GIS database. As a result, additional problems arise because of the need to work with digital data inherited from elsewhere and over which an investigator had very little, if any, control in its creation. The ease of digital data use in GIS thus raises concern not only about how scale issues are handled explicitly in disciplinary studies but implicitly through data acquisition and the application of a range of methods and tools. The identification of an optimal scale, the sampling and generalization of reality that underpins ''data scale'', multi-scalar data issues, and the desire to ''jump'' scale from local to global and the interconnections that lie between them, is methodologically highly problematic. In the face of these many scale-based issues, Gregory (1994:545) contends that ''one might expect a renewed and conjoint interest in both the technical and the theoretical issues raised by the question of scale. . . . '' Given the multiple meanings of the term scale, this chapter explores the conjoint implications and meanings of scale by focusing on spatial scale, and the implications arising from aggregation issues and inferences that contribute to the fundamental problem of ecological fallacy. If, as Bird (1989:22) suggests, different scales of approach may eventuate in different results and study outcomes, then it is important to understand scale not solely as a ''mechanism'' or scalar measure of data accuracy or cartographic display, but as a conceptual primateur in the production and reproduction of space. This paper focuses on the theoretical link between data scale and the inferences that can be drawn from scale-conditioned studies. © 2006 Springer Science+Business Media, LLC.