Welcome to Lesson 5! Last lesson, we learned about techniques that cartographers employ to visualize Earth’s terrain. This week, we begin to focus on a more statistically driven type of thematic map: choropleth maps. Choropleth maps are a very common thematic map type. To design them properly, an adequate understanding of other important topics in cartography, such as data standardization and classification methods are needed. Choropleth maps also typically employ color in their design: in this lesson, we discuss color in-depth. You will learn about the different ways in which we can model color space, and how visual perception constraints - both in the general population, and in those with color-vision impairments - influence map perception.

In Lab 5, we'll explore how choosing a different color scheme and data classification method can alter the way the information is presented and how readers interpret that information. We’ll also learn how to make maps that work well in pairs—a common task that is often significantly more challenging than making one map that stands alone.

Learning Outcomes

By the end of this lesson, you should be able to:

• match the most fitting type of color scheme (e.g., sequential; diverging; qualitative) to specific data sets;
• demonstrate how to identify and specify colors using the three perceptual dimensions of hue, saturation, and lightness;
• integrate knowledge of color perception and human visual limitations (including color-vision impairment) into map color decision-making;
• standardize and classify data appropriately for use on choropleth maps;
• select an appropriate color scheme for a map based on probable perceived connotations of those colors as they relate to the map's data.

Lesson Roadmap

Questions?

If you have questions, please feel free to post them to the Lesson 5 Discussion Forum. While you are there, feel free to post your own responses if you, too, are able to help a colleague.

Color is frequently used to symbolize information on maps. In recent years, cartographers have begun to employ color more frequently. in a study of map-color use in scientific journals, White et al., (2017) found that the use of color in published map figures increased from 18.4% in 2004 to 69.9% in 2013. This trend can primarily be attributed to the expansion of practical map production technologies. The cost of color printing, for example, is no longer prohibitory. Additionally, the increasing popularity of web-based dissemination of maps and other visual graphics makes such color production costs irrelevant. Tools such as ColorBrewer Colorbox, and Colorgorical have also made color selection easier; the first of these is now integrated into the color selection tools in ArcGIS Pro and a separate package in R (RColorBrewer).

Figure 5.1.1: ColorBrewer (left), and Color Schemes in ArcGIS Pro (right).

Credit: Cynthia Brewer, Mark Harrower and The Pennsylvania State University: ColorBrewer2.org; Screen capture from ArcGIS Pro.

In this lesson, we will explore the basics of specifying, mixing, and selecting colors for choropleth maps. You should aim to understand and properly apply the color schemes available in GIS software, and alter them as appropriate based on your maps’ audience, medium, and purpose. Eventually, you might even design your own color schemes from scratch.

Figure 5.1.2:: A choropleth thematic map from the Census.

Credit: United States Census Bureau

You may remember the map in Figure 5.1.2 from Lesson 1. This map is a thematic map, and more specifically, a choropleth map. Discussions of color in mapping often focus on choropleth maps. This is for good reason—choropleth mapping is the most common thematic mapping technique, and its employment typically requires thoughtful analytical use of color. We will discuss the details of choropleth mapping later in this lesson. However, note that color is also frequently used on other types of maps. General purpose maps often employ color to delineate between different kinds of features, and maps that focus on other symbolization types (e.g., proportional symbol maps) often also use color to encode an additional variable, or to add visual interest.

When you hear the word "color," words such as blue, red, and green likely spring to mind. Though these are colors in the colloquial sense, these are better described as color hues. Color has more dimensionality than just the color name. In fact, when thinking about color as a visual variable, each color is specified not just by hue but by three dimensions: hue, lightness (also “brightness” or “value”), and saturation (also “chroma” or “intensity”) (Figure 5.2.1). Some people regard these “alternative terms” as completely synonymous with each other, while others argue that they each refer to something specific. For now, just know that the synonymous terms refer to roughly the same properties.

Figure 5.2.1: Lightness, Hue, and Saturation.

Credit: NASA Earth Observatory

Color is produced when light is either reflected off of (e.g., a car; a printed map) or emitted by (e.g., a computer screen) an object. Hue refers to the portion of the electromagnetic spectrum where human vision is sensitive. We can discuss color falling along that spectrum in terms of its wavelength of light, from longest (oranges and reds), to shortest (blues and violets). Figure 5.2.2 shows nine swatches of color with different hues, in the order of the rainbow spectrum. It is important to understand that the electromagnetic spectrum offers a vast range of wavelengths, and the human visual system can only perceive a relative tiny portion of that range. For an overview of the electromagnetic spectrum, NASA has a useful website (https://science.nasa.gov/ems/01_intro/).

Figure 5.2.2: A set of color hues in spectral order.

Credit: Cary Anderson, Penn State University.

In mapping contexts, hue is typically used to differentiate between features. In general purpose maps, for example, the use of different hues creates different categories, and helps the reader identify different features as belonging to a particular group. In Figure 5.2.3, for example, the color choices are visually distinguishable, and improves the legibility and aesthetics of the map. Though multiple types of roads are shown, all roads are shown in red. Similarly, all hydrologic features and labels are shown in blue - a familiar color easily recognizable by map readers as associated with water. Furthermore, features and their labels that are shown in green, map readers conceptually associate with vegetation.

Figure 5.2.3: A general purpose map that uses color hue.

Credit: National Institute of Standards and Technology

Lightness is another dimension of color; it describes how perceptually close a color appears to a pure white object. Lightness is also commonly called value, though cartographers sometimes avoid that term, as value is also used to describe data values—using the same word for both items can cause confusion. Another alternative word, brightness, might sound like you’re referring to the brightness of a screen on which a map is being displayed, so use of that word is not recommended either. Lightness works well for visually encoding the order and/or magnitude of thematic data values—typically, lighter colors signify lower data values (i.e., less implies less), and darker, more visually-prominent features implies higher data values.

The third dimension of color is saturation. Saturation is also sometimes called chroma or intensity. Highly saturated colors are particularly useful for calling attention to small but important map elements that would otherwise be lost (Figure 5.2.4). Caution should be used when using saturation in this way, however—the use of too many highly saturated colors, particularly over large areas, may be distracting or accidentally overemphasize unintended features. An effective alternative approach is to desaturate your basemap/background so that your most important features can remain at a reasonable saturation level, but still stand out. If you look at maps in popular media outlets, such as the New York Times or National Geographic, you’ll notice that this approach is extremely common.

Figure 5.2.4: Using saturation to call out important elements on the map. Bright orange is used to highlight areas that are rather small in areal extent with zone classifications important given the map's purpose.

Credit: Cary Anderson, Penn State University; data source: NYC GIS

The three color dimensions (hue; lightness; saturation) were originally identified by Dr. Albert H. Munsell in the early 20th century. Munsell’s first color model, a color sphere, was an attempt to fit these three dimensions of color into a regular shape. Though this model was still a breakthrough, Munsell realized that it was quite insufficient, as human color perception is not linear and cannot be accurately modeled by a regular shape. The final shape he landed on looks more like a lopsided ellipsoid. The Podcast 99% Invisible has written an excellent short piece on the origins and specifics of the Munsell's color system, with helpful explanatory graphics. Read it here: The Color Sphere: A Professor's Pivotal "Color Space" Numbering System.

Figure 5.2.5 below takes a top-down approach to visualizing this color space: each of the four graphics demonstrates what is, in essence, a slice of the Munsell model, with increasing lightness from left to right. As shown, the colors that the human eye can perceive do not change linearly through color space—note that there is a greater range of red hues than blue hues. This non-linearity makes color specification and design a challenging task.

Figure 5.2.5: Colors perceived by humans at different “slices of lightness”, as shown by the NASA Ames Color Tool.

Credit: NASA Earth Observatory

Though Munsell’s model is helpful for understanding color perception, and perhaps for sharing color specifications with others, a working knowledge of other models is required for building color schemes in GIS and graphic design software. When specifying colors, it is important to consider the display medium that you are using to create them. When mixing paint, cyan, magenta, yellow, and black are used (CMYK [“K” stands for black because it used to refer to the “key plate” in printing and that mixing CMY does not produce a true black, which had the most detail and was usually black]). As mixing paint (or laser printing toner) results in less light being reflected from the color surface, this is called subtractive mixing. The opposite occurs on digital display screens, which create colors by mixing red, blue, and green (RGB) light. Mixing these primaries is called additive mixing.

Figure 5.2.6: Additive and subtractive primaries (color models)

Credit: Lisa Cianci CC BY-NC-SA 4.0

Figure 5.2.7: Specifying a color in ArcGIS Pro.

Credit: Screenshot from ArcGIS Pro.

ArcGIS offers a wide selection of color model choices for specifying colors, including RGB, HSV, and CMYK. RGB and CMYK color models refer to the aforementioned models for mixing additive and subtractive primaries, respectively. RGB is useful for digital media, and CMYK is the color language typically used by graphic artists, largely for print media. Another popular model is hue, saturation, and value (HSV). HSV is reminiscent of the Munsell model (see Figure 5.2.8), but with much greater symmetry—recall the oddly-shaped structure of Munsell’s model.

Figure 5.2.8: A visual model of the HSV color system.

Credit: Datumizer via Wikimedia Commons (CC BY-SA 3.0)

The symmetry of HSV makes it fit much better into the language of computers, but as human color perception is not linear (recall Figure 5.2.5), using HSV can cause problems unless you remain cognizant of this shortcoming.

Additional color models, including hue, saturation, and lightness (HSL) and Commission internationale de l'éclairage that expresses color as three values: L* (perceptual lightness) and a* and b* (red, green, blue and yellow) (CIEL*a*b* or CIELAB), offer other ways of specifying colors. We will not go further into the details of color specification here, but you are encouraged to explore the recommended readings for more information.

Types of color schemes

When applying color schemes to maps, there are many factors to consider. First and foremost, keep this rule in mind: the perceptual structure of the color scheme should match the perceptual structure of the data. For example, if your data go from high to low (sequential data), you should use a color scheme that demonstrates this quantitative order, as shown in the map in Figure 5.3.1. Note also that the primary color hue, green, was selected due to its cultural association with the mapped theme.

Figure 5.3.1: Color lightness demonstrating data order.

Credit: US Census Bureau, Census.gov

There are three main types of color schemes: sequential, diverging, and qualitative. We will discuss what these mean below, but you may find it helpful to augment our discussion by visiting ColorBrewer, a popular tool for choosing color schemes on maps. This tool was designed by Dr. Cynthia A. Brewer at Penn State. ColorBrewer’s interface is shown in Figure 5.3.2. Feel free to explore the many color schemes available on the site as you read more about types of color schemes in this lesson and consider how you might apply them to your maps.

Figure 5.3.2: A Screenshot from Colorbrewer.org.

Credit: Cynthia Brewer, Mark Harrower, and The Pennsylvania State University, ColorBrewer2.org.

Sequential color schemes are one of the most popular color schemes used in thematic mapping, as they intuitively communicate the quantitative order of data values. If you are attempting to visually contrast the numerical arrangement of values of a particular dataset, then a sequential color scheme is probably an appropriate choice. Several examples of sequential color schemes are shown in Figure 5.3.3.

Figure 5.3.3: Sequential color schemes using color lightness.

Credit: Cary Anderson, Penn State University; color specifications via ColorBrewer2.org.

Though color lightness is effective on its own, sequential color schemes are also often designed with multiple harmonious hues, such as in the color schemes shown in Figure 5.3.4. The multi-hued nature of these color schemes can make it easier for viewers to discriminate between all data classes on the map. They also often create more aesthetically-pleasing visualizations. As long as it doesn't take away from readers' comprehension of your data, why not make a better-looking map?

Figure 5.3.4: Sequential color schemes using color lightness and color hue.

Credit: Cary Anderson, Penn State University; color specifications via ColorBrewer2.org.

As shown in Figure 5.3.5, when hue is paired with lightness it can create dramatic contrast in a sequential color scheme. When adding sequential color schemes to such maps, ensure that the chosen scheme accurately reflects the progression of your data—it is challenging to create an effective sequential color scheme that relies heavily on hue.

Figure 5.3.5: A map that uses a sequential color scheme with changing lightness, hue, and saturation.

Credit: NASA Earth Observatory

Diverging color schemes are similar to sequential color schemes, as they also demonstrate order. However, instead of showing a single progression, they visualize the distance of all values from a meaningful midpoint, usually an average or median using two contrasting color hues. For example, a map that shows percent change with red hues showing increases and blue hues showing decreases. This middle value or class is often represented using white or a light grey representing a neutral position in the data. Diverging schemes are typically limited to two color hues. Using a third color hue may cause readers to assume that the color scheme is qualitative (more on that later).

Figure 5.3.6: Diverging Color Schemes.

Credit: Cary Anderson, Penn State University; color specifications via ColorBrewer2.org.

If your data has a natural midpoint—such as a 0% change in some phenomenon— a diverging color scheme works well, as it permits the reader to easily identify values on the map as either above or below that value. An example of this is shown in Figure 5.3.7 below.

Figure 5.3.7: A map with a diverging color scheme, showing a critical midpoint of 0% change.

Credit: GlobalChange

Other values can also serve as helpful midpoints in mapped data. For example, a map might use a diverging color scheme to demonstrate values that fall above or below the data’s mean, or perhaps some external value (e.g., a choropleth map of median income where a diverging color scheme is centered around a calculated national level value of a living wage).

An important consideration when applying a diverging color scheme is whether your data has a critical class or a critical break (Figure 5.3.8). Using a diverging scheme with a critical class will highlight a critical group of areas on your map, as well as those above and below. A critical break will show all areas as either above or below a critical value—there is no “neutral” color class in this scheme. Diverging schemes also do not always have to be symmetrical. Your critical class/break will often be near the center of your data range, but it in no way needs to be.

Keep the divergent schemes shown in Figure 5.3.8 below in mind as we discuss data classification for choropleth mapping later in the lesson.

Figure 5.3.8: Diverging schemes with a critical class (left), and critical break (right).

Credit: Cary Anderson, Penn State University; color specifications via colorbrew2.

Figure 5.3.9: A diverging color scheme used on a map about temperature.

Credit: NOAA

The third type of color scheme is the qualitative color scheme. These schemes are used to demonstrate differences—but not numerical order—between map features. Several examples are shown in Figure 5.3.10 below.

Figure 5.3.10: Qualitative Color Schemes.

Credit: Cary Anderson, Penn State University; color specifications via ColorBrewer2.org.

Qualitative color schemes are often used when creating maps of political boundaries, or to create categorical choropleth maps, such as the one in Figure 5.3.11. As the term choropleth is composed of the Greek words for “area/region” (khṓra) and “multitude” (plēthos), it is technically incorrect to refer to a map of nominal values as a choropleth map, despite the characteristic enumeration-unit shading such maps employ. These maps should instead be called chorochromatic maps. That being said, it’s unlikely that you’ll hear even GIS or most cartographer professionals use that term. But hey, be the change you wish to see in the world, right?

Figure 5.3.11: A Chorochromatic Map.

Credit: Map by Cary Anderson, The Pennsylvania State University, Data Source: Boundaries, US Census. Thematic data are invented.

Perhaps the most common use of qualitative color schemes in mapping is in land use/land cover (LULC) maps. These maps seek to demonstrate category (e.g., residential vs. commercial) but not to demonstrate order. An example of a land cover map is shown in Figure 5.3.12.

Figure 5.3.12: A map of land cover in Des Moines, Iowa, using a qualitative color scheme with additional variation added in lightness and saturation.

Credit: Cary Anderson, Penn State University; Data source: USGS, The National Map.

The (color vision unimpaired) human eye can discriminate between about twelve different hues in the same image, and, dependent on the reader and the design of the map, often even less. Many maps, and LULC maps in particular, contain more than this number of categories. A frequent strategy is to group categories into hue classes (e.g., green for vegetation) and then to use lightness and saturation to create intra-class differences. In Figure 5.3.12, for example, green hue is used for forest, and lightness variations are used to differentiate between forest types. When designing a color scheme for land classification-land cover maps, one must be careful to choose color variations that are visually perceptible from others (i.e., too many similar green hues may not be visually perceptual).

So far in this lesson, we have talked about multiple ways to specify colors, and how we might apply them to maps. As we discuss color, however, we also need to discuss color vision deficiency—the inability to discriminate between certain (or occasionally, all) colors. Though color blindness varies by gender and ethnicity, you can generally expect that between five and ten percent of your map readers will have some form of color deficiency. You may even have some form of color vision deficiency yourself.

The good news is that several web tools exist to help you design more accessible maps. Viz Palette, developed by Elijah Meeks and Susie Lu, is one useful example. It permits you to import your own color schemes from popular color-picking tools such as ColorBrewer and view their appearance through the eyes of those with different types of color vision deficiencies. Vizcheck is an application that allows an image or map to be uploaded and view how the colors on that image or map appear according to different color vision impairments.

Figure 5.4.1: Viz Palette, using colors from ColorBrewer’s qualitative color palette "Set2."

Credit: Elijah Meeks and Susie Lu, Viz-Palette (click the link to try it out!)

Figure 5.4.2: Viz Palette: The same ColorBrewer scheme as used in Figure 5.4.1 above, as seen through the eyes of someone with deuteranomaly—the most common form of colorblindness

Credit: Elijah Meeks and Susie Lu, Viz-Palette.

Tools such as Viz Palette are useful for understanding how different people might view your data visualizations and maps. You can then decide for yourself whether your chosen palette is acceptable. ColorBrewer also lets you select from among only color schemes that have been empirically-verified as colorblind friendly its interface includes an option to show only “colorblind safe” color schemes. Unsurprisingly, the scheme in Figure 5.4.1(2) does not appear.

How much you factor color accessibility into your map design will depend greatly on its audience, medium, and purpose. Color discriminability is affected by many factors outside of genetics, including reader age, lighting conditions, and map resolution. It is also more crucial in some mapping contexts than in others. A map for entertainment, for example, may sacrifice accessibility for increased aesthetics and visual interest among the not color-vision impaired. When a map’s purpose is emergency management or vehicle routing, however, the cartographer may place a greater value on ensuring readability for all map users.

Even among those without color vision impairments, human color perception does not come without flaws. View the squares labeled A and B in Figure 5.4.3—do they look the same to you?

Figure 5.4.3: The Checker Shadow Illusion

Credit: Edward H. Adelson, available under the Open Database License (CC BY-SA 4.0)

You likely perceive squares A and B as different shades of grey, but, as you may have guessed, your eyes are deceiving you—these two squares are exactly the same shade of grey. (If you don't believe it, check out the interactive version of this graphic at illusionsindex.org). This is the result of a principle of color interpretation called simultaneous contrast, or induction—colors appear differently, dependent on the backdrop against which they appear.

Figure 5.4.4: A set of choropleth maps demonstrating the principle of induction.

Credit: Cary Anderson, Penn State University; data boundary source: US Census Bureau, thematic data invented.

To date, little empirical research in cartography has evaluated the influence of induction on map interpretation, and, thus, few suggestions exist for minimizing its effects in practice. You should, however, anticipate the effects that varied backgrounds will have on the interpretation of your map symbol colors, particularly for maps in which such comparisons are common and/or critical.

So far in this lesson, many of our examples have been choropleth maps—the most common thematic mapping technique, and one which typically makes extensive use of color as a visual variable. In the next section, we will focus on other aspects of choropleth mapping, including data standardization and classification, as a deeper understanding of how these maps are built using data is required for selecting an effective color scheme.

The choropleth mapping technique should be used on standardized data such as rates and percentages—rather than on totals or counts—which are better represented by point symbol maps.

There is almost never a good reason to make a choropleth map without standardizing your data. Why? Because if you don’t standardize your data, then you are inadvertently creating a map of the underlying population. For example, you could create a choropleth map of the United States showing counts of, say, gas stations in each state. Texas and California have the largest populations of any state in the US, so they would likely have more gas stations and show primacy in this count. The result is a map without much useful information—California and Texas have more people and things because simply because they have more people and things. The map would tell us nothing interesting about each state’s respective consumption of gas or transportation infrastructure in relation to the underlying population. However, if you were to map gas stations per capita (i.e., if you standardized your data), then we would be able to meaningfully compare rates, and a choropleth map would be an appropriate method.

If you’re lucky (really, really lucky), your data will be delivered in the proper standardized format. For example, for each enumeration unit in your data, you might have a rate, density, or index value. All of these are appropriate standardized data for choropleth mapping. Oftentimes, however, you will need to calculate these values yourself. Data from the US Census, for example, is often delivered as count data by enumeration unit but includes a population field which can be used for standardization.

Figure 5.5.1 Census data from the American Community Survey.

Credit: Health Insurance data from North Carolina, US Census Bureau. Screenshot by Cary Anderson, Penn State University.

Using the example data in Figure 5.5.1 above, imagine we wanted to map the number of people in each county who are under 18 years old AND have one type of health insurance coverage (Column F). And imagine we created a county-level choropleth map using those Column F values. What would this map tell us? It might tell us a little something about geographic health insurance trends in North Carolina, but mostly it would just show us in which counties more people live.

Remember the importance of map purpose: rather than just making a population map, we want to understand the geography of health insurance coverage for young people. For this, we need to map standardized values. To do so, we can divide the number of under 18-year-olds with one type of health insurance (Column F) by the appropriate universe: the count of items (here, people) that could possibly fall into this category. Since our data value of interest only applies to a specific age group, our universe, in this case, is not all people (Column D), but all people under 18 (Column E).

Some texts and software programs, including ArcGIS, call this process normalization rather than standardization. As suggested by (Slocum et al. 2009) we use the term standardization, as normalization has a more specific meaning in statistics with which we do not want this process to be confused.

Figure 5.6.1: A choropleth map from the Census (posted earlier as Figure 5.1.2)

Credit: US Census Bureau

Let’s return again to a map that should be becoming familiar, posted now as Figure 5.6.1. Median income is visually encoded in each state as belonging to one of four classes: (1) less than $45,000; (2) $45,000 to $49,999; (3) $50,000 to $59,999, and (4) $60,0000 and more. How were these classes chosen?

When the map in Figure 5.6.1 was being designed, the aforementioned classes had to be decided upon – and there are many different ways in which class breaks in median income could have been drawn. So, how do you choose? Rather than simply choosing the default classification scheme that your GIS software suggests, you should think critically about how your data classes are defined. Before you decide how to class your data, the first decision you should make, however, is not how, but whether to class your data.

Figure 5.6.2: An unclassed (top) and classed (bottom) choropleth map.

Credit: Maps by Cary Anderson, Penn State University; Data Source: United States Census Bureau, American Community Survey.

Figure 5.6.2 shows an example of two maps—one unclassed and one classed. Unclassed maps (sometimes called N-classed, where the N represents the number of enumeration units, or "class-less" map) encode color (usually with lightness) based on the specific value within each enumeration unit, rather than based on a pre-defined class within which the data value falls. These maps are useful as—if designed properly—they may more accurately reflect the ordinal nuances in the distribution of the data as map readers can see the differences between the color lightness (a given color lightness is more or less light than its neighbors). However, unclassed maps should not be considered an easy solution to the problem of data classification. They have their own disadvantages, for example, they make it challenging for the reader to match the value encoded in an enumeration unit to its location on the legend.

Before modern GIS software, unclassed maps were quite difficult to create, but new technology has made their design quite simple. Unclassed maps show a visualization of the data that respects the inherent numeric distribution of data values, while classifying maps gives you more control over the final map. It will be up to you as the map designer to decide whether to class your map; however, many map readers—and cartographers—still prefer classed maps.

As you will likely be classifying your maps, it is important to understand how this process can influence your final map design. Most of the commonly-used classification methods are available in ArcGIS, and the software interface gives a simple explanation of each of these methods (Figure 5.6.3). We will not discuss the mathematical details of each of these classification methods here—it is recommended that you explore the recommended readings or do your own research on the web to learn more.

Figure 5.6.3: Classification options available in ArcGIS Pro.

Credit: Screenshot from ArcGIS Pro.

Click here for a text alternative to this screenshot of classification options available in ArcGIS Pro.

Natural Breaks (Jenks): Numerical values of ranked data are examined to account for non-uniform distributions, giving an unequal class width with varying frequency of observations per class.

Quantile: Distributes the observations equally across the class interval, giving unequal class width but the same frequency of observations per class.

Equal Interval: The data range of each class is held constant, giving an equal class width with varying frequency of observations per class.

Defined Interval: Specify an interval size to define equal class widths with varying frequency of observations per class.

Manual Interval: Create class breaks manually or modify one of the present classification methods appropriate for your data.

Geometric Interval: Mathematically defined class widths based on a geometric series, giving an approximately equal class width and consistent frequency of observations per class.

Standard Deviation: For normally distributed data, class widths are defined using standard deviations from the mean of the data array, giving an equal class width and varying frequency of observations per class.

Though Figure 5.6.3 gives brief descriptions of each classification method, it offers little advice as to when to use them. A good way to approach this question is to view your data along the number line. You can use histograms (for large data sets) or dot plots (for small data sets) to visualize how your data is distributed, and to select class breaks accordingly. The following suggestions are given by Penn State cartographer Dr. Cynthia Brewer.

1. For data with near-normal distributions, consider classifying your data based on the mean and standard deviation.
2. For skewed distributions, consider systematically increasing classes, such as arithmetic and geometric classing methods.
3. If your data are evenly distributed, equal interval and quantile classing methods work well. These methods are also best for ranked data.
4. Natural breaks, created using Jenks classing method or in selecting breaks by eye, work best for data that shows obvious groupings through the range. The natural breaks method highlights the numeric relationships in the data values.

We will look at data using dot plots during this lab associated with this lesson. When you make maps, unless you are working with a very large data set, this will often be the most effective way to visually investigate the distribution of your dataset in order to choose a classification method or visually/manually place your own breaks. ArcGIS, however, creates histograms of your data that you can also use to understand how the breaks you have chosen to relate to the spread of your data.

Figure 5.6.4: Two maps created using the quantiles classification method (top), and natural breaks (Jenks) method (bottom). Note the histograms of each dataset shown to the right of each map.

Credit: Maps by Cary Anderson, Penn State University, Data Source: US Census.

Note that the spread of your data is only one of multiple elements you should consider when choosing how to classify your data. As with other map design choices, your map's intended audience, medium, and purpose are also of vital importance here.

In addition to choosing a classification method for your maps, you also must decide how many classes to create. It may be tempting to create a large number of classes, as more classes means less simplification of your data, and thus more information conveyed to the map viewer. Unfortunately, the human eye can only differentiate between so many colors. There are recommendations for the maximum number of color classes on a map, generally ranging from about 5 to 12. But a good rule of thumb is that the fewer classes your reader has to remember, the better.

Figure 5.6.5: A map with many classes (left) makes it more difficult to interpret than a simpler map (right).

Credit: Maps by Cary Anderson, Penn State University, Data Source: US Census.

Finally, when classifying your map data, you will have to contend with outliers in your dataset. Consider a county-level map, where one county has double the rate (for example, of people with graduate-level degrees) of any other county in your data. Some classification methods, such as natural breaks or equal intervals, will most likely group this outlier into a class of its own. Other methods, such as quartiles, will simply place it into a group with all the next-highest counties.

There is no rule for which method is best, except that context matters. Is the rate high because that county contains the most prestigious university in the state? In that case, you probably want it to be highlighted on your map. If, instead, it is the highest because only five people live there—and two are college professors—you probably don’t. In general, the more data you have, the less likely an outlier is to be noise: this is called the law of large numbers. Whenever possible, however, you should investigate the possible causes of an outlier; there is no substitute for contextual clues.

There are additional ways to classify your data, including by combining methods; for example, using equal intervals for most of the range, and then switching to natural breaks. Methods also exist that consider not just the distribution of data along the number line, but its distribution through geographic space as well. These are beyond the scope and intent of this lesson, but be aware that you may encounter them in the future.

By now, you should feel pretty good about creating a single choropleth map. But while we frequently encounter choropleth maps in the singular, the power of maps often comes from our ability to compare them. Static maps—all of the maps we’ve discussed thus far—typically only represent one snapshot in time. What if we are interested in how a phenomenon has changed over time, or how it varies between two disparate locations?

View the two maps below in Figure 5.7.1. They are both maps of population density from New Jersey and Vermont and are shown using the same scale. A casual inspection of the maps (to non-US residents, perhaps), the vibrant colors appearing on the Vermont map suggest that this state may have a higher level of population density. But take a closer look at the legends.

Figure 5.7.1: A map of population density (2010) in NJ and VT, each designed with the default ArcGIS classification scheme.

Credit: Cary Anderson, Penn State University, Data Source: US Census Bureau.

The legends in the maps in Figure 5.7.1 don’t match. The darkest color, for example, represents a vastly higher level of population between the two maps. How much does population density differ between New Jersey and Vermont? Due to the unmatched legends, it’s almost impossible to tell.

Using the same data classification scheme for a set of maps whose purpose is to compare a dataset is necessary. For example, the maps in Figure 5.7.2 use the same data, but this time, both legends are equivalent.

Figure 5.7.2: A map of population density (2010) in NJ and VT, with matching classification schemes.

Credit: Cary Anderson, Penn State University, Data Source: US Census Bureau.

This gives us an entirely different view of the data: New Jersey is now represented as obviously more densely populated. Note, however, that this map just took New Jersey’s classification scheme and applied it to Vermont, which is still not a good solution. Though it is now easy to compare these states, we are unable to discern which areas of Vermont are more populated than others: they are all simply classified as "less than 562 residents per square mile." Making maps that work well both independently and when compared is a challenging task, and one which we will contend with in Lab 5.

Another important aspect of choropleth—and any—map design is making sure that marginal elements such as legends and labels are well-crafted to support reader comprehension of your map. For example, see Figure 5.7.3. It may seem at first that this legend is too text-heavy at the expense of the geography mapped: you don’t generally create visual graphics with the intention of asking people to read. However, without necessary information being conveyed through the text, the content of the map would be confusing, and many readers would likely misinterpret it.

Figure 5.7.3: A map legend with thorough annotations. Note that the legend size is exaggerated on this map to show effect.

Credit: Cynthia A. Brewer. US Census Bureau. Made with Natural Earth. Map updated by P. Limpisathian, Penn State Geography.

This map also purposefully places breaks in the data; for example, one break is placed at 24 percent, which is the percentage of all people in the US who are under 18 years old. The break is annotated to inform the reader of this fact; without this annotation, the use of this specific break would not be useful. Additional legend annotations (e.g., “High proportion of AIAN are young”) serve to clarify the map.

Figure 5.7.4 below similarly uses a text explanation to clarify the data mapped. Due to the classification scheme used, the location indicated by the leader line and Prisons* note does not immediately stand out as an outlier. However, given the topic of the map, this explanation is important. We discussed dealing with outliers earlier in the lesson—one option for dealing with a relevant outlier is simply to point it out to your readers via explanatory text. Mapping is all about graphic presentation, but sometimes the best solution is a simple, concise, text explanation.

Figure 5.7.4: A map showing HIV/AIDS causes with a note attached that explains an outlier.

Credit: City of Philadelphia Department of Public Health

When using color as a symbol on your maps, your first priority should be to apply it analytically. As stated before: the perceptual structure of your color scheme should match the perceptual structure of your data. You should apply color based on the guidelines previously discussed in this lesson before worrying about choosing aesthetically-pleasing colors, or your audiences’ likely favorite colors, or colors that correspond to the context of the data (e.g., using a green color scheme to create a map about sustainability).

However—when appropriate—adding context to colors in your maps can benefit your readers. See the map in Figure 5.8.1 below. Rather than choosing a traditional sequential color scheme, this cartographer chose to match the map’s colors to colors of tree leaves as they turn in autumn.

Figure 5.8.1: Map predicting fall foliage colors for 2024. It’s interactive– follow the link to use it yourself.

Credit: SmokyMountains.com

This approach may not always work to best represent the mathematical order of your data classes. But your maps aren’t always about dots along a number line—they represent real-world phenomena. Using color assignments that make sense (e.g., red for negative values), or are customary (e.g., yellow for residential in zoning maps) can improve the clarity and comprehensibility of your maps.

Critique #3 will be your second critique involving a peer review of a map created by someone in this class. In this activity, you will be assigned a colleague's map from this class to critique from Lab 4: Terrain Mapping.

Your peer review assignment includes writing up a 300+ word critique of one of your colleague's Lesson 4 Lab.

In your written critique please describe:

• three (3) things about the map design that you think works well and why.
• three (3) suggestions you have for improvement of the map design and why these improvements would be helpful.

According to the two prompts above, a map critique is not just about finding problems, but about reflecting on a map in an overall context. Your critique should focus on the map design that works well as much as it does on suggestions for design improvements. In your discussion, you should connect your ideas back to what we learned in the previous lessons.

Remember, your critique should be as much about reflecting upon design ideas well-done as it is about suggesting improvements to the design. In your discussion, connect your ideas to concepts from previous lessons where relevant.

Grading Criteria

Registered students can view a rubric for this assignment in Canvas.

Color and Choropleth Mapping in Series

In Lab 5, we will explore different ways of choosing data classification and color schemes for choropleth maps. As a cartographer, you will often have to choose between several of these options, many of which may seem at first glance to be equally appropriate. In this lab, we will utilize data from the American Community Survey, provided by the U.S. Census—a commonly used source of data for statistical maps. From this data source, we will focus on a specific variable frequently in focus during public policy debates: health insurance.

The first part of Lab 5 will focus on data classification. There are many ways to classify statistical data on maps, and it is important that you understand them, and be able to defend your choice of classification scheme to others. As we will be not only be classifying data but also adding that data to maps, this lab will also focus on the use of color on maps. Finally, as suggested in the lesson content, we will explore ways of making comparable maps - in this lab, we will be making three pairs of maps.

Lab Objectives

• Create three pairs of county-level choropleth maps describing health insurance in New England.
• Utilize shared or similar legends to help readers understand the relationships between pairs of maps.
• Use information about data distributions and health insurance rates in New England and the US overall to plan shared data classification breaks.
• Understand the impact of different color schemes and classification methods; be able to reflect upon and write about these decisions.

Overall Lab Requirements

For Lab 5, you will create three pairs of maps, each pair as its own full-page map layout. In total, you will have three separate pages. Two maps will appear on each page. You will also write a short reflection statement about each pair of maps.

• For each pair, use the same map positioning and scale within each frame; one scale bar for both maps.
• Prepare balanced page layouts with all elements suitably sized and balanced negative space—no pinched elements or visual collisions.
• Attend to text hierarchy: overall title, subtitles, legend title(s), legend class labels, scale, data source, and name. Use thoughtful and efficient wording when labeling map elements.

Map Requirements

Map Pair One: Use a Sequential Color Scheme

• Choose two related variables to map from the provided American Community Survey (ACS) data.
  • Do not just choose two age groups (e.g., 18-under; 19-25 years).
  • The mapped data must be two related variables.
• Select class breaks manually
  • Create dot plots in Microsoft Excel
  • Draw appropriate breaks using your eye to judge the data
  • Enter these values as manual breaks in ArcGIS Pro.
• Use a sequential color scheme and a single shared legend for both maps.
• Include a short write-up (100+ words) which includes a screenshot of your dot plot with lines drawn to demonstrate the breaks you chose, as well as a short description of how you selected these breaks. Also, include a screenshot of the symbology pane for both maps.

Map Pair Two: Use a Diverging Color Scheme

• Re-create your maps from map pair #1; using a diverging color scheme.
• Choose a critical break or class using external information using either of the approaches listed
  • Use a value that is directly derived from your chosen data set (e.g., the mean of the data)
  • Any logical dividing point that is calculated from an external source (e.g., the U.S. national average)
  • Adjust other class breaks accordingly.
• Use a single well-designed shared legend for both maps.
• Include a short write-up (100+ words) describing the critical break or class you chose and why. You may also discuss why you selected this particular color scheme.

Map Pair Three: Unclassed vs. Classed Maps (Choose your own appropriate color scheme)

• Choose one of the maps from map pairs #1 and #2 and create two more maps of this data—unlike in the previous layouts you made, these two maps will show the same data/topic.
• One of the maps should be an unclassed map; one should be classed.
• For the classed map, choose a classification method available in ArcGIS Pro—do not manually adjust the class breaks created, but ensure that this method is appropriate for the data you are mapping.
• Include a well-designed legend for each map.
• Include a short write-up (100+ words) that describes why you chose the classification method you did, and how you think its effectiveness compares to that of the unclassed map.

Lab Instructions

1. Download the Lab 5 zipped file (43.2 MB). It contains:
   • a project (.aprx) file to be opened in ArcGIS Pro;
   • a database that includes the spatial boundary and health insurance data needed to start this lab;
   • a spreadsheet containing New England health insurance data.
     • Data source: US Census Bureau - TIGER boundary files and American Community Survey (ACS) S2701 (Health Insurance Coverage Status) 5-year estimates for 2016.
     • For the purposes of this lab, New England is defined as the following states: Massachusetts, Connecticut, Rhode Island, Vermont, New Hampshire, and Maine.
2. Extract the zipped folder, and double-click the blue (.aprx) file to open ArcGIS Pro.
   • In addition to the ArcGIS Pro file, you will also be using the ACS_2016_NewEngland_HealthInsurance.xlsx file to explore New England health insurance data.
   • Note that you will not need to import any data into ArcGIS Pro - all data is included and ready to map. The Excel file is only for visually exploring the data in order to select class breaks for your maps.

Grading Criteria

Registered students can view a rubric for this assignment in Canvas.

Ready to Begin?

More instructions are available in the Lesson 5 Lab Visual Guide.

Summary

Congrats on making it to the end of Lesson 5! In this lesson, we learned about color, data classification, and choropleth maps - three topics that are quite inter-related. During our discussion on color models and human color vision, we talked about how to select appropriate color schemes to choropleth maps that represent quantitative data. We learned how to choose a color scheme for a map based on the perceptual progression of our data, as well as how to consider other factors such as map purpose, color accessibility, and data context. We also explored ideas related to data classification. We specifically focused our attention on how to choose a classification method and how that choice can affect the information presented on the map. Choropleth symbolization, while commonly used to map quantitative data, does present limitations in that data are aggregated to an enumeration unit and are assumed to be continuous across that unit which may not be how the data truly are distributed.

In Lab 5, we made pairs of choropleth maps. In doing so, we took on the challenge of making maps that work well both independently and when viewed together. We also compared the visual effect of classed vs. unclassed maps, and considered the impact of each method on reader perception of our maps. In building our final map layouts, we utilized knowledge from earlier lessons, such as legend and layout design. As we move forward with the course, the skills we learn will continue to build upon each other. We will design some more interesting map layouts in Lab 6!

Reminder - Complete all of the Lesson 5 tasks!

You have reached the end of Lesson 5! Double-check the to-do list on the Lesson 5 Overview page to make sure you have completed all of the activities listed there before you begin Lesson 5.