Map Symbology & Classification: Renting in Brooklyn, 2000
A time-tested series of comparisons that help us think about the choices we make for map symbology, classification, normalization, and aggregation.
For almost twenty years, I have used a series of images (together referred to as Renting in Brooklyn, 2000) to walk through some of the many (carto-)graphic choices we make to symbolize quantitative values on a map. The series is usually presented in Week 2 or Week 3 of an introductory GIS lecture semester and referenced in other classes or shared as a slide deck with students in other classes when it is useful.
Below are those images and comparisons.
Start here: Renting in Brooklyn, 2000
All the maps in the series are formatted the same way: Centered on the borough of Brooklyn (Kings County, NY), with a pale-grey basemap. It has a combined title and legend in the top-right corner, a north arrow and scale bar in the lower left, and an abbreviated data source in the lower right.
All the maps in the series represent the some manipulation of the same data: renter-occupied housing units, from the 2000 US Census long form.
We start with renter-occupied housing units by block, and I ask whether there is any reason to distrust this map or the block-level pattern of renters across Brooklyn that it describes. (I assure them that there is no funny business, data-wise or legend-identifying-wise.) We agree that the map has all the basic elements that convey trustworthiness before moving on.
Classification Choice comparisons
Students have already been introduced to basic symbology approaches and classification methods by the time we walk through this series. Thus the images do not explain classification methods—opting instead for illustrating the effect that different methods can have on the interpretation of the mapped data and the spatial patterns they imply.
These two sets of three comparisons tell wildly different stories about the prevalence and pattern of renting in Brooklyn if the legends are not carefully examined by a reader. The point here, of course, is that the dataset does not tell the story
For the sake of comparability across slides, the Jenks Natural Breaks visualization is maintained across both slides. The final “Manual” map is the version that appears on the first slide of the series (above).
Symbology Choice comparisons
The “Manual — Jenks Approximation” map from the previous slide is carried forward for comparison into this next section. Here we discuss color choices (one-directional ramps, bidirectional ramps, and unramped or distinct-color symbols). We compare their legibility, value or attention-based implications, and use cases that might be most appropriate for different color ramps and directions.
We also deviate from the choropleth for a couple slides for a similar conversation about symbol types: classified (graduated) and unclassified (proportional ) sized symbols as well as dot density choices.
Representing Values & Normalization
We return to the choropleth map that has been carried forward from section to section thus far to discuss normalization as a means of providing context to the values represented on the map.
And then discuss ways to handle zeros, ending with a map of the percentage of occupied housing units that are renter-occupied, by block in 2000, that distinguishes blocks with a 0% rental rate from those with zero housing units. The result is already a markedly different geography of renting than where the series started.
Aggregation
Finally, we examine the effects of aggregation on the “storytelling” of the map, comparing the same rental rate maps by block, by block group, and by census tract. (And I usually give a heads up to my students here because the following week’s lecture introduces the modifiable areal unit problem.)
Ending here…
We conclude this discussion with a side-by-side comparison of the first map and the final map. We remind ourselves that the first map appeared clear, data-driven, accurate, and trustworthy. And yet we cannot help but notice how very different two readers’ understanding of the geography of renters would be if supplied these different maps.