Measures of pattern complexity for choroplethic maps

Cartographic and psychological literature both identify complexity as an important influence on visual pattern recognition. Two indexes already developed for measuring complexity in choroplethic maps, spatial autocorrelation and the degree of the best-fitting trend surface, are compared with three new indexes suggested by psychological research on pattern complexity. A computer algorithm is presented for the computation of these new metrics which are based on the number of contiguous areal units assigned to the same class. The fragmentation index is a standardized measure of the number of these map regions. The size disparity index and relative entropy both consider the variation in area of the regions on the map. Test data are generated so that these five complexity metrics can be examined for redundancy. The results suggest that there may be at least three dimensions of pattern complexity relevant to choroplethic maps: pattern fragmentation and autocorrelation, size inequality of map regions, and the general spatial trend and its autocorrelation or strength.