Difference Mapping Approach to Detecting the Cartographic Effects of the Modifiable Areal Unit Problem

Abstract Making sense of geographic phenomena across scale often requires the analysis and visualization of aggregated data. Aggregation can make spatial data more comprehensible and help to preserve the privacy of the individuals reflected in the data. However, the process is also prone to the modifiable areal unit problem (MAUP) that results from misalignment between the historical, political, and social factors used to derive geographic boundaries and the ways in which phenomena and processes operate and interact within those boundaries. The level of aggregation and the arbitrary sizes, shapes, and arrangements of zones contribute to statistical bias that affects the validity of results from aggregated data analysis. The modifiable areal unit problem is not just a statistical problem but also a cartographic one, impacting the reliability and usefulness of choropleth maps. Despite nearly a century of research on MAUP and the development of advanced analytical approaches to measuring its statistical effects, limited work has sought to characterize how it visually manifests across the surface of a map. This is problematic given the vital role choropleth maps play in supporting decision-making. To address this gap, a novel application of change detection analysis is proposed to expose the visual effects of MAUP. An unclassed choropleth difference mapping approach is employed to visualize nuanced changes in geographic phenomena representations across aggregation structures and quantify the distribution of those changes as a percent of the total mapped area. The utility of the approach is exemplified using three proof-of-concept analyses spanning seven U.S. Census-based enumeration unit types across political, health, and socioeconomic scenarios.