Examples of how to use (and how not to use) Color

First and foremost when using color (or even using grayscale) that your data using the color to send the message that you went to convey. Your data should own the color, not the other way around. Here are some examples of maps with temperature for color overlaid. The first example is not a very good example.

It has nothing to do with the image being pixilated, it has everything to do with the choices in color. For us everyone is trained to think that red is hold and blue is cold, when looking at the key blue is cold; however, the brightest pink (there is no red) is not the hottest temperature. Also around the 61-67 temperature range because the colors jump drastically you think on the map that there is a drastic change in color as well, but there isn’t. Its only 6 degrees. But this map isn’t all bad, it does some things right it does not use a rainbow color map, tries to use colors that are more intuitive and from 39-63 degrees it has a beautiful gradient.

This next example does en excellent just of using colors that are intuitive, having a color map that works with the data.

The one critique I have of this map is that perhaps it should have used more ‘earth tones’ for the two colder temperature ranges since the other colors are more muted.

Ways to Say Nothing with Scientific Visualization

If you never using a color legend then people don’t know what values the color corresponds to then all you have is a pretty picture, and it says nothing about your data. If annotation is avoised then very subtle nuiances will go unnoticed by the reader. You need to make call outs so people will notice what is important. You shouldn't be afriad to smooth data, it makes the picture look pretty, and can obscures error. It makes it easier and sometimes quicker to publish. But don't smooth too much, other wise you obscure more than just error. Those are just some of the important take home points from Globus and Raible.