PrintMap Generalization
In Lesson 7, we discussed a common theme in cartography: uncertainty visualization. In Lesson 8, we focus on how the complexities of our world are reduced into a visualization via a map. Specifically, when we reduce the complexities of Earth’s geography into a form more appropriate for a map’s given scale and purpose, this is called cartographic generalization. Thorough understanding of generalization and the related concept of scale is—and has always been—essential for creating high quality maps. The increased prevalence of web-maps, which simultaneous and seamless zooming and panning across multiple extents and scales, has encouraged increased research in the process of generalization. In Lesson 8, we discuss generalization, both in general, and in the context of multi-scale and interactive web maps.
Maps of Earth or other terrestrial bodies are set to some scale which reduces the complexities of Earth's features and their true size. As a result, all maps contain some level of generalization—maps would be unusable otherwise. A map at a scale of 1:1 is rather pointless. Representing every element of the real world on a map is not feasible, nor would such a map be interpretable by readers. Generalization permits cartographers to construct maps with an appropriate level of detail while preserving spatial relationships, feature density, and complexity. In Lesson 4, we discussed the necessity of using the correct resolution of (raster) digital elevation data to create terrain visualizations of a large scale map. In Lesson 8, we focus primarily on the generalization of vector data, such as road networks, hydrologic features, and political boundaries.
When considering what level of detail is appropriate, it is important to consider your map's location, scale, and geographic extent. A map of seaside hotel locations in Massachusetts would, for example, show a much more detailed coastline of Cape Cod than would a map of the entire United States.
It’s also worth quickly reviewing the difference between small-scale and large-scale maps: small-scale maps represent a considerable extent of Earth's surface, while large-scale maps represent a reduced extent of Earth's surface. In terms of representative fractions, a 1:2,000,000 map scale shows a greater extent of Earth's surface than a 1:24,000 map scale. Whereas the former scale would be appropriate to map a state the latter scale would be more appropriate to map a limited portion of a city. Even professional cartographers mix these two up on occasion, so just do your best commiting this to memory.
It should also be noted that there is no hard-and-fast definition of what contitutes “small” and “large” scale, i.e. there is no quantitative agreement on the scale at which, for example, a map goes from small- to large-scale. Added to this confusion is the use of "medium" to indicate a scale. General, maps referred to as “small-scale” tend to depict large areas, such as regions, states, or continents. Large-scale maps tend to depict cities, neighborhoods, streets, and so on. But again, these are not immutable declarations. Instead, when we compare small- and large-scale maps, we are doing so in relative terms. However, small-scale maps nearly always benefit from increased generalization, and large-scale maps benefit from greater resolution or detail. Here is a brief overview(link is external) of map scale by the USGS.
Natural Earth(link is external) is a source of boundary data that we have used extensively in this course. Figure 8.1.1 below demonstrates the differences in level of detail between different boundary datasets that Natural Earth offers. Natural Earth offers various categories of data and for each category at different scales. For example, the purple boundaries (left) show the most detail. Such data are appropriate for large-scale maps (scale = 1:10,000,000). Here, the level of detail shown in this large-scale vector linework is high. At this map scale, the intent is to present a dataset with a greater amount of detail compared to the other smaller scales. The pink (center) boundaries would be better suited for smaller-scale maps of countries or continents (scale = 1:50,000,000). The blue (right) boundaries are highly generalized, and would be best suited for very small-scale maps such as the globe or advertizing maps that use heavily stylized data (scale = 1:110,000,000).
Data Source: Natural Earth, Esri (basemap from ArcGIS).
Figure 8.1.2 shows each of the above boundary files at an appropriate scale given their level of detail. The extent of the largest-scale map (black rectangle of the frameline - left) is shown by the black rectangle extent indicator in the center and right maps. Note that as the scale becomes smaller (moving from the maps left to right in Figure 8.1.2, we see a diminished level of complexity with the administrative boundaries. Try to think of mapping purposes for which each level of boundary detail would be appropriate.
Data Source: Natural Earth, Esri (basemap from ArcGIS).
Student Reflection
For an interactive experience with generalization, try uploading a shapefile from NaturalEarth(link is external) to the interactive tool MapShaper(link is external).
So far, we have talked about the overall idea of generalization – using data that is the correct level of detail for your map’s scale. A general-purpose map of a small town, for example, would likely show lakes, ponds, and reservoirs, while a small-scale map of a large region would show only the largest waterbodies (e.g., rivers, large lakes, and oceans). Often, rules decided upon by the cartographer are used to determine what elements are displayed on a map (e.g., “only show lakes that cover more than five square miles”). However, due to the uneven distribution of features across the landscape, cartographers also have to make some generalization decisions that are complex, subjective, and specific.
An example of this is demonstrated by Figure 8.1.3. Some cities are labeled, and some are not. At first, it may appear that the largest cities are labeled, and to some extent this is true. New York, NY is labeled, as well as Washington, DC. However, you may notice some cities that are absent—most notably Philadelphia, PA. A city with 1.5 million people is left off the map, while Reading, PA—a city of about 94,000—is included. Why?
Philadelphia is located in a densely-populated region, with many nearby cities, such as Trenton, Baltimore, and Washington, D.C. By contrast, Reading, PA is surrounded only by smaller towns. Web-maps are designed to display—or not display—city labels based on a number of factors. These include population and general importance, but also design-relevant factors, such as the density of labels on the map. In the case of tools like Google Maps and Apple Maps, the features included on the map are also influenced by the user’s search history and/or other digital activity. Look at several web maps of the same location at the same scale (or zoom level) to see what similarities and differences in labels, road network, and hydrology features are apparent.
Recommended Reading
Chapter 3: Map Generalization: Little White Lies and Lots of Them. Monmonier, Mark. 2018. How to Lie with Maps. 3rd ed. The University of Chicago Press.