General principles of color use in visualization applications

One of the most powerful ways to stretch the capabilities of on-screen displays is by taking advantage of the extraordinary ability of humans to use color information to aid in visualizing data. Using color logically requires a basic understanding of the three "dimensions" of color -- hue, lightness, and saturation.

 Hue. Color can be described by its wavelength; red at one end of the visible spectrum, violet on the other. A change in wavelength of visible light is manifested by a change in hue. "Red," "green," and "violet" are all hues.

Hue differences.

Lightness. Color can also be described by its shade, or its degree of lightness. Pink-white, pink, rose, red, and maroon are all are composed of the same hue (wavelength) but have different lightness. This dimension of color can also be called "value," or "brightness."

Lightness differences.

Saturation. A third dimension of color describes how "pure" the color is. If a color is made up of only one hue, it is called "saturated." The color of monochromatic light, like that which comes from a laser beam or neon sign, is completely saturated. However, if you imagine mixing together every glob of acrylic paint on an artist's palette, the result is a desaturated color, one made of many hues. Saturation could be called the "grayness" of the color.

Saturation differences.

As mentioned, color on the whole is one of the most powerful of the so-called "visual variables" in visualization applications. Differences in color can be used in limitless ways to differentiate objects or areas on a display. The following are some general guidelines for the logical use of color in an on-screen display.

Use of hue in visualization. Hue differences are best used to differentiate between two items that have no inherent order; that is, objects that differ in kind (nominal) and not amount (ordinal) are best represented by varying hue. Geologic maps use hue to a great extent to differentiate between soil or rock types. Land-use types, political party affiliations (as in the example below), and area code zones all represent nominal differences and should also be differentiated by a nominal visual variable like hue. Theoretically, hue differences imply no inherent order ("green" does not in itself imply "less" or "more" than blue). However, it has been argued -- successfully -- that a "spectral" color scheme, where blue and indigo represent small values (like cold temperatures) and red represents high values (warm temperatures), is an intuitive ordered use of hue.

Use of lightness in visualization. Unlike hue, differences in lightness do imply order. It has been shown that darker colors indicate higher values of a variable. Lightness variation, thus, is one of the most intuitive and powerful ways to represent ordered data, like rainfall totals, voter turnout, or traffic density.

It is generally accepted that the human eye can successfully discern as many as seven different hues and seven different lightnesses on a single display. If your display requires more classes than seven, you should exploit other visual variables, like texture, density, or shape differences.

Tip. In ArcView, select two colors for the outermost classes and then use Ramp Color to create intermediate colors in the intermediate classes.

These two examples come from the well-designed map of Czech voting results made by ARCDATA PRAGUE.

Use of saturation in visualization. Saturation is the least understood of the three dimensions of color. Saturation differences seem to be the least discernable by the eye, and thus this dimension should be used sparingly. Saturation has been best used to represent binary data -- data that can have one of two values, like public/private ownership of land, or attributes that either have or havenít been field-checked.