Colour figures in journal articles are common these days. Many of the people reading them, however, will print them in black and white. Consequently, when designing figures, we should ensure they remain meaningful if converted to grayscale. Here are a few tips to keep in mind.
Use different line styles (and weights)
Different colours are a good way to distinguish different lines. However, the time-tested method of using different line styles (dashed, dotted, etc) is more reliable. The best option is to distinguish by colour and style as the example shows.
Using different line weights (thicknesses) can augment the number of distinct lines available. However, be careful to ensure the reader doesn’t infer anything you didn’t intend, e.g., that thicker lines relate to more important quantities.
Use varying brightness
Certain colour schemes can be converted to grayscale and have their components remain identifiable by virtue of their brightnesses (also known as value or lightness). Typically, these schemes start with closely related colours. For example, red, orange, and yellow are related in that they are warm colours and are adjacent on a colour wheel. Careful tinting or shading (mixing with white or black) results in a colour scheme with evenly separated shades of gray when printed without colour.
For line graphs, these colour schemes are beneficial for situations where dashed and dotted lines would not work well. Note, however, that a practical upper limit to the number of lines that can be distinguished this way is five. Any more than that and the reader will be unable to accurately differentiate between lines.
For filled contour plots, these schemes are appropriate for sequential (as opposed to diverging) data. A reasonable limit on the number of different shades is eight.
Use dashed lines for negative contours
In an earlier post, I recommended the blue–red colourmap as one possibility for diverging data. Unfortunately, when printed in black and white, information about the sign is lost; only the magnitude is shown. For contour plots, this can be overcome by using the following established convention: solid lines for positive contours and dashed lines for negative contours. Here’s an example:
It might appear that the inline contour labels make the need for dashed contour lines redundant. However, the dashed lines help demarcate the positive and negative regions. This is useful for illustrating the overall structure of the data.