J.P. Schroeder

Middlebury College, Department of Geography, Middlebury, Vermont, USA


We typically conceptualize patterns in a geographic time series in one of two ways: as trends in spatial patterns or as spatial patterns in trends. For example, we might describe a trend in spatial patterns as “increasingly segregated” or “shifting from monocentric to polycentric”, whereas we might describe a pattern in trends as “rebounding in the center” or “fluctuating in the north”. In fact, most techniques of spatio-temporal data analysis—statistical, graphical, computational, or otherwise—tend to emphasize or facilitate comprehension of either trends in patterns or patterns in trends. This is also true of temporal mapping techniques; yet cartographic research has made little use of the distinction. This is unfortunate because using the distinction to categorize temporal mapping techniques can help us determine which mapping techniques are most appropriate in different settings.

This paper reviews standard temporal mapping approaches and demonstrates how different approaches emphasize either trends in patterns or patterns in trends. Chronological map series—either animations or small multiples of chronologically-sequenced maps—seem to have become the most common and well-researched approach to mapping geographic time series. I argue that chronological map series, however, are mainly useful for visualizing trends in patterns and are inadequate for visualizing patterns in trends. Meanwhile, a variety of “trend summary map” techniques that are useful for visualizing patterns in trends (e.g., time-series glyph maps, trend-cluster maps, and trend-component maps) have received relatively little attention, but could help us identify and illustrate many important spatio-temporal patterns in ways that chronological map series cannot. Furthermore, trend summary maps can summarize patterns in trends through a single graphic while chronological map series require us to view and compare multiple images. For this reason, trend summary maps should generally be a more effective way to compare spatio-temporal patterns in multiple regions than chronological map series.

To demonstrate the ways in which different mapping techniques emphasize different types of patterns, I provide illustrations of trends in the population densities of American cities using historical census tract data. I compare chronological map series with a novel approach—bi-component trend mapping—which illustrates two principle components of long-term trends simultaneously using a bivariate choropleth symbol scheme.