66
MaP Design
nature of the Data
The nature of the data is a major factor. Data may be qualitative—that is, non-
ordered, or quantitative, ordered and numerical—and within these categories there
are subcategories. Table 4.1 is based on Cynthia Brewer’s work with color schemes
and shows four different data conceptualizations and their associated color schemes.
ColorBrewer (
www.ColorBrewer.org) is a useful site for learning how to understand
and choose color (Plate 4.6).
Qualitative schemes
Qualitative schemes are used for data that have no magnitude or size difference
between classes; they are
unordered and show only qualitative information (differ-
ences in kind), such as land use or vegetation types. Qualitative schemes show differ-
ences between categories with different hues; lightness should be similar, although
small lightness differences are necessary for the colors to be distinguished from one
another and allow areas to be identified more easily. There should not be large differ-
ences in lightness or saturation because these imply differences in importance.
Since the advent of GIS, which makes use of color easier, maps are frequently
created that have as many as forty qualitative categories, often made up of varying
lightness steps of hues. These maps are unreadable; the user tends to group the reds
together, the greens, the yellows, and so forth, making perceptual categories. If there
is no logic to the hue/lightness choices, a strange impression, indeed, is the result. An
example of lack of color logic is shown in Plate 4.7 Red is used for coniferous trees,
pink for grasslands, green for deciduous trees, light green for bare land, brown for
mixed forest, and tan for shrubland. While this is an extreme example, many pub-
lished maps make similar errors.
A qualitative map with many categories presents a major challenge and requires
thought and experimentation to achieve a clear map. If possible, the number of cat-
egories should be reduced, but if this is not possible, then lightness and saturation
differences can be introduced, but by using some logical system. For example, urban
and suburban areas imply an ordered relationship, so that lightness steps of one hue,
red, perhaps, can be used to distinguish the categories. For the vegetation example
above, using green for all trees makes sense, with the variations of deciduous and
coniferous shown with variations in saturation and lightness.
Brewer
considers binary schemes to be a special case of qualitative schemes.
Binary variables are those which have only two categories, such as rural/urban, pri-
vate/public, and populated/unpopulated. For these categories she suggests one hue or
two hues or a neutral with the difference between them a lightness step (Plate 4.8)
Quantitative schemes
Quantitative maps are those that represent some numerical aspect of a spatial distri-
bution, such as temperature, rainfall, elevation, and the like.
If data show a progression from low to high, a simple
sequential color scheme is
normally used. The scheme may be made up of neutrals, that is, shades of gray, one
hue, such as shades of blue, or a transition of two hues, such as yellow to blue (Plate
Color in Cartographic Design 67
4.9). In these cases the differences are of lightness. There is a limit to the number of
lightness steps that can be perceived, so the number of categories is usually restricted,
and the number possible depends upon the hue or hues selected. Most cartographers
use between five and seven steps for quantitative maps (see Chapter 8). Yellow has
fewer distinguishable lightness steps than red. A single hue permits the fewest distin-
guishable steps and a two-hue sequence permits the greatest number. Thus, if more
than
seven steps are needed, a two-hue sequence is a good choice.
Some data do not progress from high to low, but are seen as two-ended or diverg-
ing from a midpoint. Positive and negative change with zero as a midpoint is an obvi-
ous example of this. Other examples include data above or below a median, or aver-
age temperatures above or below freezing. For these data a
diverging scheme works
well. Diverging schemes basically utilize two sequential schemes joined at the critical
or midrange figure. Some examples are shown on Plate 4.9.