Full size image available through contact |
The scientific establishment is drowning in data, but whether it is census data or the vast amounts of satellite and computer-generated information created every day, visual representation and the use of color can help scientists understand and extract important patterns from this deluge, according to a Penn State cartographer.
"The smarter we are at mapping data, the more likely it is that we will see relationships," says Dr. Cynthia Brewer, associate professor of geography. "We can look at complicated data using color intelligently and see these relationships and generate hypotheses," she told attendees today (Feb 14.) at the annual meeting of the American Association for the Advancement of Science in Denver.
Presenting more than one variable is sometimes difficult, but careful color and texture choices can make representations much clearer. One option is to use a color scheme for the actual data and a hatched texture to indicate that the data is uncertain.
"Sometimes, it is important to overlay something about uncertainty to prod the viewers into thinking about what the data means," says Brewer. "The color says here is the data, but the hatching says this data is suspect."
No more than three types of data can be shown using color. Confusion sets in at the fourth variable, but three variables can be successfully shown on a map. In topographic data for example, color hues could show the direction of slopes, while saturation of the colors could indicate the steepness of slope. Lightness represents the third variable – shape -- and is the illumination on the relief produced by an arbitrary sun above the map. With these three variables, the map takes on shape and conveys information. With only the first two, it is difficult to read.
Brewer warns that there are mistakes that can be made when choosing colors for maps. If ordered data is being represented, choosing a group of random color hues obliterates any patterns. Also, if vivid colors are used for values that are unimportant, the color pops off the page and causes the viewer to give more weight to that variable than it warrants.
"For diverging spectral color schemes, make sure the central value, often bright yellow, is actually matched with an important data range," Brewer says. "If mapping areas of drought versus flooding, use the central yellow as the normal precipitation level moving to orange brown in one direction and blues in the other."
Another error is designing a map without considering those who are color blind. Most color blind people are red-green color blind and schemes that contain versions of red, orange, yellow and green with the same lightness can be impossible to read. Often, simply removing the green from a spectral scheme will improve readability for the color blind. But avoiding unreadable schemes and finding ones that are pleasant and best show the data is not as difficult as it seems.
Brewer and Mark Harrower, assistant professor of geography, University of Wisconsin, Madison, developed an online web tool that can provide pre-designed color palettes. ColorBrewer (www.ColorBrewer.org) allows you to identify the number of classes you will have on the map, the type of data being represented – sequential, diverging or qualitative – and then a color scheme. Schemes are marked to indicate color blindness friendly choices, usefulness for projection, on a computer screen, photocopyability or printability. The program also indicates the color specifications for commercial printing, computer screens or web page.
"For those who need to do full four-color printing, a paper soon to be published in the January issue of Cartography and GIS will provide color swatches and color-separation ink codes for a large variety of color schemes," said Brewer.
With the advent of color monitors for computers, scientists in every discipline embraced the use of color to clarify and illuminate data, but back then, monitors could only display the eight basic colors. Consequently, many scientists adopted a simple rainbow color scheme for their work.
"Now, computers can show millions of colors and researchers are still using just eight," says Brewer. "There is a whole world of subtlety and shade available now."
Additional images available at http://www.psu.edu/ur/2003/MapColorphotos.html