types of color schemes for maps, including sequential, diverging, and qualitativeMANY FACTORS AFFECT THE COLORS YOU CHOOSE FOR map symbols. The perceptual structuring of the colors should correspond with the logical structuring in the data. When designing maps, remember that datasets have sequential, diverging, or qualitative arrangements. You can reflect these arrangements—and make your maps easier to read—by ensuring that the character and organization of the colors on your map match the logic in your data.
When choosing map colors, you should not be overly concerned about which colors your audience likes. Everyone has an opinion about color aesthetics, and members of your audience undoubtedly have differing opinions based on their own color preferences. There has been a substantial amount of loosely structured research on color preferences. Regardless of context, it seems that most people like blue and do not like yellow, but that is an overly simplistic guideline for multicolor contexts. People also like maps with many colors, so focus your attention on presenting your data clearly and do not worry about whether you have picked everyone’s favorite colors.
Map designers who want to build effective maps consider the following:
types of color schemes for maps, including sequential, diverging, and qualitative
combining scheme types for two-variable mapping
adjusting color selections for simultaneous contrast and color-blind readers
using custom color ramps
When people read your color maps, they use the perceptual dimensions of color, even though you may have specified colors in a mixture system like RGB (red, green, and blue). Your readers are seeing and thinking about color as “light desaturated blues,” “dark saturated oranges,” or “dark grays,” not as strings of numbers. By using perceptual dimensions in ways that parallel the logical structures in a map’s data, you make it easier for your readers to understand the way the information is organized.
Sequential, diverging, and qualitative schemes are used to structure color symbols to correspond with simple data arrangements. More complex data can be mapped by overlaying and combining these schemes.
The most basic guidance for color use on maps is to use lightness to represent ordered data (figure 8.1). Ordered data may be numerical data or ranked data. For example, two city populations are ordered whether listed as “50,000 and 650,000 people” or simply as “small and medium sized.” Generally, darker colors are used to represent higher data values, and lighter colors represent lower values.
The forest fire map in figure 8.2 shows areas of high vegetation mortality in a dark color. The colors get progressively lighter as fire severity decreases. This progression is referred to as a sequential color scheme.
Figure 8.1 A set of simpler, single-hue, sequential color schemes from ColorBrewer2.org.
Sequential color schemes may include hue variation, but they should rely most on variations in lightness. The fire severity map (figure 8.2) has symbols that progress from light to dark within one hue (pink to red to black). In contrast, the map of high-speed Internet (figure 8.3) ranges from dark blue for low access to light yellow for high access. The hue change of blue-green-yellow supports the steady progression from dark to light and emphasizes low Internet access.
Figure 8.2 Map of fire severity demonstrates light-to-dark colors that parallel low-to-high data ordering. Data sources: US Geological Survey (USGS), US Forest Service. Map by A. Dennis, Department of Geography, The Pennsylvania State University (Penn State Geography).
Figure 8.3 Some counties had almost no farms with access to high-speed Internet in 2007. The color sequence is reversed from the standard “dark represents high” convention to emphasize lack of access. Data source: US Department of Agriculture (USDA). Map by A. Dennis, Penn State Geography.
Including hue differences with lightness differences while controlling saturation provides contrast between colors, helping your reader to tell them apart. Figure 8.4 shows a selection of example schemes that include hue transitions.
Figure 8.4 Sequential color schemes included on ColorBrewer2.org (see appendix). Each scheme includes a hue transition to complement a lightness sequence. The labels above each scheme indicate the hues used (BuPu indicates blue to purple).
The color schemes shown here are from ColorBrewer (http://www.ColorBrewer2.org), a web application that can guide you in selecting map colors. ColorBrewer adapts a selected scheme to the number of classes required for your dataset, then displays it on a sample map. It reports a variety of specifications for each color, including RGB, to allow you to then use the scheme in your mapmaking software. The appendix provides CMYK (cyan, magenta, yellow, and black) versions of the ColorBrewer schemes for use on maps that will be printed. RGB conversions of these colors (from Adobe Illustrator) are also listed in the appendix.
Each scheme has been designed to adapt to datasets with three-to-nine color classes. Shown here is the purple-blue-green scheme applied to each of these cases (figure 8.5). As the number of classes increases, it becomes more likely that adjacent colors in a sequence will be too similar to be discernable in some media, such as a projected image.
Figure 8.5 The PuBuGn family of sequential color schemes with an extended hue transition (purple pink, purple, blue, green blue) shown with three to nine classes.
Careful use of hue and lightness can make maps showing change and difference easy to understand. For example, differences in median home value may be represented by emphasizing both highs and lows. The cheapest and priciest places to live in central Atlanta are shown in figure 8.6. Meaningful midpoints in this data range exist at overall median values. Although values of homes increase sequentially, it can be more helpful to think of this dataset as values higher or lower than overall medians. A color scheme that emphasizes this midrange with a light color and the extremes with contrasting dark hues, and runs through two lightness sequences (dark to light to dark), reflects the nature of the dataset.
Figure 8.6 This map shows median house value by census block group in Atlanta, Georgia. It uses a diverging color scheme with two hues (magenta and green) and a light-gray middle class for block groups between overall median values for Atlanta and the United States. Data sources: US Census Bureau, made with Natural Earth. Map by A. Dennis, Penn State Geography.
Emphasizing the two extremes with darker colors and the median with the lightest color clearly parallels the structure within this Atlanta dataset. In addition to representing a median or average, this technique can be used to highlight a no difference break, no change, zero, a national rate, or a threshold value.
If the critical value is midway through a class, then that class is represented with the lightest color (figure 8.7). An odd number of classes will result when the critical class is right in the middle of the data range. But, the critical class or break need not be in the middle.
If the critical value is used as a break between classes, then the scheme will have two light colors of different hues straddling that break and no neutral color. The set of diverging schemes in figure 8.8 shows an even number of classes straddling a critical class break.
Figure 8.7 Diverging purple-green color schemes with odd numbers of classes. Each scheme is symmetrically organized around a middle light class that corresponds to a critical range in mapped data.
Figure 8.8 Diverging purple-green color schemes with even numbers of classes. Each scheme is symmetrically organized around a middle class break that corresponds to a critical data value.
A critical class may also sit between two critical values. The home values map has an upper break equal to the median home value for the United States and the lower break equal to the city median. The magenta classes are above both values and more expensive than both medians (figure 8.6).
Spectral or rainbow schemes are popular in scientific visualizations and news media graphics like daily weather maps. The full sequence is familiar: dark red, red orange, orange yellow, yellow, yellow green, green blue, and dark blue. Unfortunately, this scheme is often misused as a sequential scheme rather than as a diverging scheme. Using it to display a sequential dataset, such as temperature, places unintended emphasis on arbitrary midrange values by representing them with the lightest and most vivid color: yellow. The dark red and dark blue endpoints of the scheme should mark the extremes in the data, and light yellow should emphasize a meaningful midrange, such as zero change in population or average stream flow.
The most informative use of a spectral scheme is as a diverging scheme. People like the multihued character of the scheme, and the variety of hues helps distinguish symbol categories. Structuring the lightness sequence to parallel the characteristics of the data produces an informative map (figure 8.9).
A dataset can often be examined as both sequential and diverging. Regarding data as high-to-low and as deviations from a median may both be equally meaningful. Diverging and sequential are conceptualizations rather than absolute properties of data.
Figure 8.9 Atlanta home values shown with a diverging spectral color scheme. The lightest color, yellow, is used to represent the between-medians class. Data sources: US Census Bureau, made with Natural Earth. Map by A. Dennis, Penn State Geography.
Figure 8.10 Atlanta home values shown with a sequential (light to dark) scheme, rather than a diverging scheme. The between-median midrange is subtly emphasized using higher saturation. Data sources: US Census Bureau, made with Natural Earth. Map by A. Dennis, Penn State Geography.
Figure 8.10 shows the same home value data using a sequential scheme that emphasizes the range of home values, rather than how they differ from a middle value. An advantage of this rendition is that it could be adapted to print in black and white. This sequential scheme retains some emphasis on the class between medians by using a higher saturation color than others in the scheme. Thus, the scheme includes sequential lightness (light-medium-dark) and diverging saturation (low-high-low saturation).
Figure 8.11 The selection of diverging schemes available in ColorBrewer. Most offer two hues in lightness ramps joined at a shared light color. The last three include a wider range of hues.
Qualitative schemes represent different kinds of map features or categories that are not ordered. Categorical differences in data are usually represented with differences in hue. For example, types of government spending, such as military, education, and healthcare, are categories that could be shown with unique hues like red, green, and blue. A well-designed qualitative scheme will not suggest that data is ordered or that one category is erroneously more important than another. To maintain this sense of similar importance, ensure that hues in a qualitative color scheme maintain similar contrast with the background of the map by controlling the lightness and saturation of each color. For example, yellow will not be as visible on a white background as red, green, or blue, so the class you assign to yellow may be perceived as having a different importance than the others.
Figure 8.12 A comparison of the prevalent minority groups by county is an example of qualitative data. The color scheme maintains fairly constant lightness and saturation, varying only hue among the five classes. The minimal contrast in this scheme diminishes the perception of groupings in the data.
Source: US Census Bureau. Maps updated by P. Limpisathian, Penn State Geography.
Maps with small mixed color patches and more than about five qualitative categories are difficult to symbolize. You will need to make use of all of the perceptual dimensions of color, varying lightness and saturation regardless of the implied magnitude differences these color differences will suggest (compare figures 8.12 and 8.13). Make these adjustments as intelligently as you are able given the map topic.
Figure 8.13 The same map shown with more lightness and saturation contrast between colors. Overall, the colors are not markedly light or dark, but the addition of some lightness contrast assists the reader in identifying county classes and seeing groups. Categories with fewer counties have more contrast: “Asian” is darker than the others and “Two or more races” is more saturated than the others.
If logical relationships exist between categories, echoing those relationships with related colors within the qualitative scheme improves the map (figure 8.14). For example, use three different greens for three types of forest on a land-cover map. Groupings of color allow the map to be read for a general overview as well as category by category. When you have small areas on a map, choose high-saturation colors to emphasize them and to make sure the colors can be identified. The open water and developed land-cover types are shown in more saturated colors than those of larger areas to make them more visible (figure 8.14).
One way to choose colors for qualitative categories is to use hues already associated with each category, for example, red for tomatoes and yellow for corn. But do not become overly concerned about figuring out the perfect associations between map categories and hues. More often, there are no obvious hue relationships for the abstract topics common to thematic maps. Slight associations will be understood by only some of your map readers, so they are not worth laboring over. Use hue as an abstract symbol and focus your energies on making the colors easy to differentiate on the map.
A reason not to be adamant about hue associations is they may not be understood by everyone who reads the map. Sometimes a seemingly obvious hue association for you may not be so obvious to other groups. For example, red may seem just right for a hot spot on your map, but red may indicate loss to a more accounting-oriented audience (“in the red”), producing an unintended association. Green might represent money to a US mapmaker, but that link may not be meaningful to an audience whose paper currency is not green.
Figure 8.14 Qualitative land-cover map with varied lightness and saturation for map symbols. Greens, browns, and blues are used to group related classes. High-saturation cyan and magenta colors are used to emphasize small and important land-cover classes. Source: USGS National Land Cover Database (NLCD). Map by A. Dennis, Penn State Geography.
Be alert for color associations that may be offensive. For example, exercise caution if a particular hue is associated with a controversial political party in the region mapped. Also, take care with some literal uses of color, such as black for people who are African American, yellow for Asian Americans, and red for American Indians. The superficial and exaggerated emphasis on skin color associations for groups is likely to offend your readers. Use a purposely abstract set of hues for mapping race groups instead. Work to develop a set of easily distinguished hues to symbolize your data, but also be astute about unintended meanings individual hues may have for the topic mapped.
Figure 8.15 Do not use stereotypical skin colors for race map symbols. Overemphasis on superficial characteristics detracts from the map message. Source: US Census Bureau. Map updated by P. Limpisathian, Penn State Geography.
The qualitative schemes in ColorBrewer range from three to twelve colors and vary in lightness. The second-to-last scheme (figure 8.16) includes lightness pairs within hues to use for qualitative data that includes related categories.
Figure 8.16 A series of five-class qualitative schemes from ColorBrewer. Families of schemes with up to twelve classes are offered by the online color tool.
Sometimes it is useful to represent more than one variable on a map at the same time. A bivariate color scheme uses a matrix of colors to represent two variables on the same map. Each variable can be thought of as sequential, diverging, or qualitative in nature, but symbolizing both variables within the same areal units requires careful decisions about color. Bivariate maps are not always a good alternative to separate maps for each variable, but when space is limited or patterns in each variable are particularly complicated, a bivariate representation can be valuable and make otherwise unseen patterns readily apparent.
A critical part of creating bivariate maps is legend construction. Many bivariate maps will use a square-shaped legend split into nine sections. You can think of this legend as a scatter plot, with the first variable on the horizontal access and the second variable on the vertical access. Each corner of the square then represents an extreme of the two-variable combination (for example, figures 8.17 and 8.18). The top-right corner would represent the highest values of both variables and the bottom-left corner the lowest. The bottom-right corner would be high values for the variable on the horizontal axis but low values for the variable on the vertical axis, and the opposite is true in the top-left corner. The center of the legend represents middle values for both variables.
Perhaps the easiest way to create a bivariate map and color scheme is overlaying semi-transparent data layers. Figure 8.17 makes use of this technique by overlaying a map of global January land- surface temperatures on temperature anomalies. The blending of the color schemes from each map creates a new color scheme with different hues at each extreme. For example, places that had high January temperatures and high temperature anomaly (such as an abnormally warm summer) are represented with brown, resulting from a blend of green and red from the single variable schemes (see browns in northern Australia and central Argentina). Places that were in winter (low temperature) and were experiencing abnormally cold temperatures are in purple, a blending of blue and magenta from the single-variable schemes (see purples in northern Russia and the eastern United States).
Figure 8.17 Top left: A map of January 2014 land-surface temperatures with red representing hot (summer) and blue representing cold (winter). Top right: A map of January 2014 land-surface temperature anomaly with magenta representing abnormally cold and green representing abnormally warm. Bottom: A bivariate map of land-surface temperature and temperature anomaly created by overlaying transparent, single-variable maps. Data source: National Aeronautics and Space Administration (NASA) Earth Observatory. Map by A. Dennis, Penn State Geography.
Depending on the color-blending capabilities of your design platform and the two single-variable color schemes used, transparency may or may not be an effective method for creating a bivariate map. Using transparency to create the color scheme puts you at the mercy of how the single-variable schemes blend, which is hard to predict and usually produces less vivid colors. It is often more effective to create nine separate map classes and carefully specify a color for each class.
Figure 8.18 An example of a sequential-sequential color scheme showing density of vacant lots and crime density. Legend annotation lists hue meanings at corner extremes. Lightness progresses diagonally from low-low to high-high. Data sources: City of Baltimore, data.baltimorecity.gov; ©OpenStreetMap contributors, http://www.openstreetmap.org/copyright; CC BY 3.0, http://creativecommons.org/licenses/by/3.0/. Map by A. Dennis, Penn State Geography.
Visualizing two possibly related quantitative variables using a nine-class bivariate choropleth map can reveal important patterns in a dataset. Such a map would reveal areas with extreme values of both variables as well as areas that are characterized more by one distribution than another.
For example, a city official in Baltimore might be interested in crime and property vacancy patterns to better understand how neighborhoods are affected by these conditions or find clues about how they are related. Figure 8.18 shows a map designed for this purpose.
This map uses a carefully designed sequential-sequential color scheme to make the map reader’s experience as valuable as possible. In general, the color scheme progresses from light to dark along the bottom-left to top-right diagonal. Areas with the lowest values are lighter than areas with the highest values. This general pattern has the same logic as a single-variable sequential choropleth map. Figure 8.19 provides generalized explanations of the diagonal lightness change across the legend and hue patterns.
Figure 8.19 Basic patterns useful for sequential-sequential color schemes. Lightness progresses diagonally from lower left (low-low) to upper right (high-high). Hues (p and s) for each ordered data variable are mixed systematically.
Breaking down the legend further, the left column (along the y-axis) uses a sequential scheme, moving from a light gray to an increasingly saturated green. Similarly, the bottom row (along the x-axis) moves from gray to a mid-saturation magenta. The saturated magenta and green represent the highest values of one variable co-occurring with the lowest values of the other variable. These areas are important because they represent extremes in the data, and the saturated colors do well to make them easily noticed on the map.
Besides the green and magenta, a darker blue hue is used to represent areas with high values of both variables. The top row and right column move sequentially from the mid-lightness green and magenta toward the darker blue as data values increase. The middle classes along the legend edges are slightly less saturated because they are not as interesting as the extreme values at each corner. Their similarity in hue to the green or magenta colors is key to making the map readable. These colors still represent values higher on one variable than another. Looking at the part of Baltimore shown on the map (figure 8.18), pink and purple color areas to the west (left) with relatively higher densities of vacant lots than crime, and green or teal colors show areas with relatively higher density of crime than vacant lots. The maps are classed into quartiles, so even though these topics are measured in very different units, both are calculated by area (instances per square mile) and the map allows us to compare relative prevalence across the city.
When one or more of the variables in a bivariate map is thought of as diverging in character, the color scheme is designed differently. Figure 8.20 shows a bivariate map with the data for both variables diverging; values are increasing from a middle value of interest toward one end and decreasing from that value in the opposite direction. The middle square in the legend shows the center of both diverging variables, areas where house values and asking prices remain relatively unchanged. From this light gray color, the map uses lightness and saturation to represent areas with more significant changes in house values and asking prices.
Figure 8.20 Example diverging-diverging color scheme showing changes in home values and asking prices from 2010 to 2012, each diverging from city medians. Legend annotation lists hue meanings at corner extremes. Lightness progresses outward from center to darker hues at each corner. Data sources: US Census Bureau, made with Natural Earth. Map by A. Dennis, Penn State Geography.
The four corners, the extremes for both variables, use distinctive hues that are also the darkest and most saturated colors on the map. Between these corners, a lighter and less saturated color with a hue somewhere between that of the adjacent corners is used. The four edges of the legend are color schemes diverging in hue, saturation, and lightness. A bivariate diverging-diverging color scheme takes the idea of a central light class and moves out in all directions toward four extremes rather than two. Figure 8.21 describes the general logic of diverging-diverging schemes.
Figure 8.21 Basic patterns useful for diverging-diverging color schemes. Lightness progresses from a center light color outward to dark corners. Hues (abstracted to p, q, s, and t) are most saturated at corners and combine for classes between them.
Figure 8.22 The 2013 Yosemite Rim Fire is shown with a qualitative-sequential bivariate scheme. The two variables are land-cover type and vegetation mortality. Data sources: USGS, US Forest Service; http://www.mtbs.gov. Map by A. Dennis, Penn State Geography.
Creating bivariate maps that incorporate qualitative variables requires slightly different design considerations.
The map in figure 8.22 uses a qualitative-sequential bivariate scheme. The two variables are land cover and vegetation mortality. Land cover is divided into three classes using three distinct hues. As vegetation mortality increases, the colors become increasingly dark but remain distinguishable. The bivariate nature of this map provides more information than would separate maps of fire severity and land cover. Figure 8.23 describes the overall logic of this qualitative-sequential scheme.
Figure 8.23 Basic patterns useful for qualitative-sequential color schemes. Lightness progresses by rows, and each column is a different hue.
There are varied reasons to fine-tune color selections to accommodate map readers. The apparent lightness, hue, and saturation of a color can be altered by surrounding colors, often producing unexpected color changes. It is important to adjust your design to anticipate these changes. Additionally, color maps sometimes need to be read without their intended hues. If your map readers may be color-blind or your map is photocopied in black and white for inexpensive distribution, lightness differences are the key to retaining as much map information as possible. GIS color ramps can be improved by using perceptual color design skills and customizing the match between ramp characteristics and data values.
Color appearance is affected by context. Small colored objects on a map are more difficult to identify than large colored areas—you will be able to distinguish fewer colors when the map includes small point symbols or thin lines. Different surroundings can also change the appearance of a color.
The contrast between a patch of color and its surrounding color enhances the difference between the two colors in a process called simultaneous contrast. For example, a color of medium lightness will look darker on a light background and lighter on a dark background. Hue changes are toward opponent complements: red-green and yellow-blue. A gray will look greenish on a red background. A green will look lighter and yellower on a dark blue background. The saturation component of color is particularly susceptible to being changed by simultaneous contrast.
Figure 8.24 Two isolated colors have dark surroundings; they are identified as X and Y above the map. Which legend color do they match? Which of the four colors A to D in figure 8.25 do X and Y match? The map shows the percent of total population who indicated non-Hispanic White race/ethnicity for Census 2000. Source: US Census Bureau.
A series of examples based on simple diagrams and two maps (figures 8.24 and 8.25) demonstrate the relative nature of color and how it changes with differences in background. The maps and diagram (figure 8.26) all share the same set of colors: a sequential scheme that ranges from light yellow to dark blue. The map scheme has eight steps, which is a lot, so problems with simultaneous contrast effects are expected.
Figure 8.25 These counties have light surroundings, which make them look darker. Can you match them with X and Y in figure 8.24? (See answer in figure 8.27.) The map shows the percent of total population who indicated Black or African-American race for Census 2000. Source: US Census Bureau.
Observe how colors are affected by their surroundings, and test the map schemes by seeing whether you can match the identified colors to the legend despite their varied backgrounds. These sorts of changes also affect color comparisons between regions on a map and between maps in a series. Can you decide which counties represent the same data range on both maps? (See figure 8.27.)
Figure 8.26 Simplified simultaneous contrast diagrams show how map colors are made to look different (left) and how different colors are made to look the same (right) by their surroundings. Colors X and D are the same, but the color surrounding X makes it look lighter and more saturated with yellow than D. Color Y is darker and greener than A (see repeated pair below the diagram), but their surroundings make them look similar. The same colors are compared between maps in figures 8.24 and 8.25.
The same colors are shown in a simpler arrangement (figure 8.26), with and without homogeneous colorful surroundings, so you can see which ones are made to look similar and how matched colors look different depending on surround effects.
Figure 8.27 Color X matches D, though its dark surroundings on the first map make it look like C, a lighter color. Color Y matches B, though simultaneous contrast makes it look like A on the map (compare figures 8.24 and 8.25).
Simultaneous contrast makes a mess of the map in figure 8.28. The gray local roads shift toward blue because they are surrounded by yellow. The blue ferry routes, surrounded by a darker blue, shift lighter and more yellow to look grayish. The yellow highways shift lighter and bluer, so they look like a very light gray. In the end, ferry routes look like local roads in the legend, local roads look like ferry routes, and highways do not seem to match any legend symbol. The different surroundings for symbols on the map and in the legend make the legend almost worthless (figure 8.29).
Figure 8.28 Varying backgrounds shift the appearance of desaturated line colors. Data source: ©OpenStreetMap contributors, http://www.openstreetmap.org/copyright; Geofabrik (server hosting OSM extracts), http://download.geofabrik.de/. Map by E. Guidero, Penn State Geography.
Figure 8.29 A schematic diagram shows the changes in line colors on different backgrounds for the map in figure 8.28. From left to right, the yellow line looks grayer and then lighter on the two backgrounds. The gray line looks bluer, then lighter. The blue line looks bluer, then grayer.
There is little that can be done to prevent contrast effects from occurring because the distribution of data controls the positions of the symbols. The best approach is to look carefully at the colors you are planning for a map. Do not select colors by comparing them only in the legend where they are seen in one order on a uniform background. Look at your selection of colors in the final map pattern and in the final media (on screen or as a projected image, inkjet print, lithographic proof, color photocopy, and so forth). Be sure you can identify examples of each color symbol and can tell colors apart throughout the map. Look for isolated areas where one color is surrounded by contrasting colors; these situations truly test a scheme (figure 8.30). A critical eye and good color contrast counteract simultaneous contrast on maps.
Figure 8.30 The diagnostic map display used in ColorBrewer, shown here with a five-color sequential scheme. Each color is isolated in a band of each of the other colors. If the scheme is well-designed, you should be able to easily find the four other colors within each band. Source: ColorBrewer2.org.
Approximately 4 percent of the population has some degree of color vision impairment (approximately 8 percent of men and less than 1 percent of women). Color-blind people can see lightness differences and a range of hue differences.
Red-green color-blindness is the most common impairment, but it includes confusions between other hue combinations as well, such as magenta and cyan. The severity of color-blindness varies from person to person. Desaturated colors, like rust and olive, are more difficult for people with milder color vision impairments to distinguish than saturated red and green.
Expected color confusions can be rigorously modeled in color order systems, but these results are difficult to apply without color measurement equipment. Guidelines specified by color names, rather than color measurements, eliminate a wider range of color combinations than necessary, but they are useful in the context of designing maps. The following pairs of hues are not confused by people with the most common types of color vision impairments:
• red and blue
• red and purple
• orange and blue
• orange and purple
• brown and blue
• brown and purple
• yellow and blue
• yellow and purple
• yellow and gray
• blue and gray
These ten color pairs are from a total of thirty-six pairs of basic color names, so many other hue pairs are confusing. For example, any combination of red, orange, brown, yellow, and green is potentially confusing if the colors have similar lightness. Any combination of magenta, gray, and cyan is also likely to be indistinguishable to color-blind people. Blue and purple can also be difficult.
These guidelines are intended to be used to design a map by simply naming colors, rather than specifying positions in a color space. For example, choosing a red-blue pairing includes any colors that you, or your client, would call red and blue, without regard to what sort of red and blue is chosen (yellowish red or bluish red, for example).
The map in figure 8.31 includes many hue combinations that will be difficult for color-blind map readers to discern. The magenta ferry routes may be hard to see against the cyan water. The orange and red roads may be hard to tell apart and hard to see on the green land. The brown town symbols may also be hard to see against green. Poor legibility for any reader is exaggerated by the lack of lightness contrast between symbols and their backgrounds, but these symbols will be nearly invisible for color-blind people.
Figure 8.31 A portion of a map of Seattle demonstrates poor choices in colors because many will look the same to color-blind map readers. The highway, local road, and populated place colors will all look like each other and will be difficult to distinguish from the background. The pink for hypothetical ferry routes is likely to look the same as the surrounding cyan water. Data source: ©OpenStreetMap contributors, http://www.openstreetmap.org/copyright; Geofabrik (server hosting OSM extracts), http://download.geofabrik.de/. Map by E. Guidero, Penn State Geography.
Guidance more specific than general color name pairs can be derived from a variation on the color circle you were using in the previous chapter (figure 8.32).
Figure 8.32 The CMY (cyan, magenta, and yellow) color cube with the white corner centered. Hues are arranged in spectral order around the edge of the cube in this view. Figure 8.33 shows a sampling of colors across the visible faces of this cube.
Flattening and fleshing out the cube to show colors on its three visible surfaces presents a hue circle with a white center, shown in figures 8.33 and 8.34. I built an approximate set of confusion zones across this color diagram based on color-blind confusion lines through a more technical color space called CIE xyY, which is difficult to use for map color specification.
Colors with similar lightness that are in the same zone or adjacent zones of figure 8.33 are likely to look the same to a color-blind person. Colors chosen two or more zones apart will be easier to discern. Figure 8.34 shows how to use the diagram. Colors from the purple-blue zone will not look the same as colors from the red-yellow-green zone. Colors that are different in lightness within either zone—or adjacent zones—will be visibly different as well.
Figure 8.33 The black lines across the diagram outline color-blind confusion zones. The underlying arrangement of colors is built from all combinations of 0, 10, 30, 60, and 100 percent of CMY inks (see the appendix for percentages for individual colors and conversions to RGB).
The selection of six colors, with black outlines in figure 8.34, forms a workable diverging scheme for both color-blind readers and readers with normal color vision. The red-orange-yellow colors differ in lightness, so they can be used even though they are from the same zone. Likewise, the blue-purple set includes lightness differences. Colors with similar lightness (orange and blue, for example), are from separated zones. This scheme is a good, colorful replacement for a spectral scheme because it skips over the greens.
Figure 8.34 To use the color-blind confusion diagram, choose colors separated by at least one zone, or colors that are different in lightness. The selected red-orange-yellow colors (with black outlines) are from the same zone but differ in lightness. Selected colors with similar lightness are two or more zones apart. A diverging scheme with six colors was constructed using the diagram (arranged at left).
All types of color schemes—sequential, diverging, qualitative—can be adjusted so they can be read by the color-blind. Almost any well-designed sequential scheme that has good lightness contrast between colors will accommodate color-blind readers. Beyond visualizing the logical structure of the data, accommodating map readers with color vision impairments is another reason to use systematic lightness steps when representing ordered data.
Figure 8.35 Diverging schemes built from hue pairs, or gray paired with a hue, that are distinguishable to color-blind map readers.
A basic strategy for designing diverging schemes is to choose pairs of hues from the list earlier in this section and build a lightness sequence within each hue. Saturated yellow does not include dark colors, so it offers few lightness steps. The small legends shown in figure 8.35 are good examples of hue pairs for diverging schemes. Diverging spectral schemes that skip greens are also good for color-blind readers (figure 8.36).
Figure 8.36 The Atlanta home values map with a near-spectral diverging scheme that skips over greens to accommodate color-blind readers. Data sources: US Census Bureau; made with Natural Earth. Map by A. Dennis, Penn State Geography.
Designing qualitative schemes for people with color vision impairments is difficult because these schemes require many different hues (figure 8.37 is confusing to color-blind readers). Careful use of lightness differences and separation among confusion zones can produce a qualitative scheme that accommodates most color-blind readers (figure 8.38). For example, orange and blue in a scheme can have the same lightness and still be legible to color-blind readers. But if green is also used, it should be either lighter or darker than the orange.
Finding a color-blind person to look at your maps is helpful. With 8 percent of men affected, a group of twelve fellows is likely to include one who is color-blind. You will have a much harder time finding a color-blind helper in a group of women given the genetics of congenital color-blindness. Another way to evaluate colors is to use the online utilities at the Vischeck website (http://www.vischeck.com) or other diagnostic apps to simulate what color-blind people will see when looking at your map.
Figure 8.37 This qualitative scheme does not accommodate color-blind map readers. The red, orange, and green are likely to look the same as well as blue and purple.
Source: US Census Bureau. Maps updated by P. Limpisathian, Penn State Geography.
The severity of color-blindness varies widely, and there are a few different types of color-blindness. Thus, you do not want to lull yourself into a false sense that you have designed for all color-blind people when you have accommodated a single color-blind person. Having someone around to vet the schemes is helpful, but it is not complete assurance that the map will work for all color-blind readers.
Many people with mild color-blindness do not know they are color-blind. You may find yourself making adjustments to seemingly good schemes to accommodate a difficult colleague, when in reality, you are accommodating his vision impairments. Watch which colors the person is objecting to and you may be able to make an amateur diagnosis (though perhaps you will keep it to yourself). You can be proud of your ability to make color maps that the colleague and a wider audience of color-blind map users are able to read using the guidance presented here.
Figure 8.38 Here, the qualitative scheme has been redesigned with greater lightness differences and careful hue choices. Yellow-green-brown would normally be confusing, but here they differ in lightness, as do the blue and purple. Color pairs with similar lightness—purple-green and blue-brown (brown is a desaturated orange)—are well separated on the confusion diagram.
The example maps earlier in this chapter present sets of colors with discrete classes, such as a five-class sequence ranging from yellow to blue. An alternative strategy for designing a color scheme using GIS is to select endpoint colors and automatically “ramp” between them. The ramp may be presented as a continuous color gradation on a map or may be split into a series of discrete colors for map classes.
Because of problems with routes through color space selected by the software, ramping does not always produce quite what you intend. For example, a ramp from yellow to blue cuts through color space to desaturated colors midway through the ramp. This is a logical straight-line route through the color space, but may not produce an appealing map (figure 8.39A).
Figure 8.39 A yellow-blue ramp (A) is improved by fixing a green midpoint. Yellow and blue are set for the end colors in both ramps, but the second ramp (B) also has a green set at the middle. Ramp 2 runs from yellow to green and from green to blue, rather than directly from yellow to blue. The map shows Census 2010 data for Salt Lake City census tracts: family households as a percent of total households. Data source: US Census Bureau. Maps by P. Limpisathian, Penn State Geography.
If you would like more saturated colors in the ramp, you need to choose a route that arcs outward rather than going straight through the middle of the color space. You can approximate this by specifying interim colors as well as endpoints. A straight yellow-to-blue ramp includes some muddy midrange colors, but specifying a midrange green, as shown in figure 8.39B, improves the scheme. Likewise, including an orange midway between yellow and red will produce a ramp with colors that are easier to tell apart and are more pleasing in appearance (figure 8.40).
Figure 8.40 A yellow-red ramp (A) is improved more subtly with this technique. A saturated orange midpoint was set in the second ramp (B). The map shows Census 2010 data for Salt Lake City census tracts: households with one person as a percent of all nonfamily households. Data source: US Census Bureau. Maps by P. Limpisathian, Penn State Geography.
The strategy of fixing midrange colors for ramps also allows you to design flexible diverging schemes. Custom ramping gives you a flexible number of classes to make a logical lightness and hue structure. In figure 8.41, white symbolizes the 0.96 to 1.09 class, straddling 1.00 (equal number of males and females sixty-five and over). Figure 8.43 shows a custom ramp with a hue change at the 1.1 class break. Preset ramps do not provide enough control to symbolize classes that are asymmetrically arranged above and below a critical midrange data class.
Figure 8.41 Dark red, white, and dark blue are used to make this asymmetrical diverging scheme. The white class represents nearly equal numbers of elderly in two household situations. The map shows Census 2010 data for Salt Lake City census tracts: the ratio of gender of people sixty-five and older in all households.
Data source: US Census Bureau. Maps by P. Limpisathian, Penn State Geography.
Ramping is useful for reducing the tedium of explicitly specifying colors for numerous classes. If you have so many classes, however, that typing color specifications is time-consuming, you probably have too many color classes to be differentiated by the map reader. Consider your reader and your design goals carefully before you concoct a map with more than about ten classes.
Figure 8.42. A diverging ramp without a middle light class. Four colors were set (light and dark for each hue) before creating ramps that diverge from the class break at 1.1. The map shows Census 2010 data for Salt Lake City census tracts: the ratio of married family households with their own children under eighteen at home to married families without their own children at home.
Figure 8.43 shows 40 classes; 33 between white and dark red. This approach produces an essentially continuous gradation, called an n-class or unclassed choropleth map. Unclassed choropleth mapping is useful in some contexts, but not when your readers need to identify data ranges for individual areas on the map. Simultaneous contrast among the mixed colors on the map affects adjacent colors, making comparison between enumeration units inaccurate.
Figure 8.43 An n-class choropleth map for the same data shown in figure 8.41. Be cautious about creating so many classes that readers cannot identify data ranges. Data source: US Census Bureau. Map by P. Limpisathian, Penn State Geography.
Another more suitable use for many classes is as a continuous elevation ramp (figure 8.44). Unlike a choropleth map, elevation classes follow one after the other over the land surfaces, as with other continuously changing surfaces. Splitting the continuous ramp into many classes allows precise control of where the land-water break, or another critical value, falls in the ramp.
Figure 8.44 A classified raster map of hypsometry and bathymetry for southwest Washington State. Data source: National Oceanic and Atmospheric Administration (NOAA). Map by P. Limpisathian, Penn State Geography.
