Robert Simmon is the lead visualizer for the Nasa Earth Observatory and has an extensive background in using colour as a representation tool. After watching his talk from the 2014 Open Vis Conference, I gained great insight into how colour can be used of data visualization.
Simmon says, “The purpose of data visualization is to illuminate the data. To show patterns and relationships that are otherwise hidden in an impenetrable mass of numbers...colour for spatial or multi-dimensional data, is one of the most effective and common ways to conveying this information" (3:27 - 4:07). I find this quote powerful as it breaks down the goal of data visualization and how colour can be used as a tool.
Diving into basic colour theory, Simmon's addresses how we perceive colour vs. how colour is computed. He suggests that we move into thinking about colour in terms of lightness, hue and saturation, particularly using the LCH colour space (not the HSB scale). He then discusses how colour palettes using LCH, can impact sequential data, divergent data and qualitative data. In qualitative data, the goal is to make your colours choices as distinct as possible to separate the classes. However, he shows a 12-class palette. He notes that this is not ideal, alluding to Bertin’s theory of the ‘rule of 7’. Beyond this he states that you can use intuitive or semantically associated colours.
I found this talk very informative, particularly this distaste for ‘the rainbow palette’. Do you think this point against the rainbow palette is valid? How can we incorporate some of his principles into our own work?