Choose Colors for Scientific Data Visualization

Some general suggestion

• Choosing color palettes for scientific figures
• Color in Scientific Figures
• Design Tips for Scientists
• 2014, Plos Computational Biology, Ten Simple Rules for Better Figures

Background in color space

1. RGB color space
   • Mixture of red, green and blue color
2. HSB color space
   • Hue (色相)
   • Saturation (饱和度)

   • Brightness (亮度)， some tines also called value or lightness

   • See https://www.codeproject.com/Articles/1202772/Color-Topics-for-Programmers
3. HSLuv
   • A alternative to HSL, claimed to be more human friendly

python for color space mapping

• For detailed implementation of such conversion see:
  • https://stackoverflow.com/questions/3018313/algorithm-to-convert-rgb-to-hsv-and-hsv-to-rgb-in-range-0-255-for-both

import colorsys
# convert rgb to hsv 
colorsys.rgb_to_hsv(0.2, 0.4, 0.4)
# (0.5, 0.5, 0.4)
colorsys.hsv_to_rgb(0.5, 0.5, 0.4)
# (0.2, 0.4, 0.4)

Some useful tricks

• Given rgb colors:
  1. Change the lightness / saturation
     • The easist way may be convert rgb values to hsv values, change s/v, than map back to rgb value
  2. Take complementary color
     • Convert to hsv, then set h=(h+0.5)%1
     • Also see https://stackoverflow.com/questions/40233986/python-is-there-a-function-or-formula-to-find-the-complementary-colour-of-a-rgb
  3. Interpolate from one color to another
     • We shall still convert rgb to hsv, then h …
     • See discussion in https://stackoverflow.com/questions/13488957/interpolate-from-one-color-to-another

Color map for visualization

Color Map Requirements

• According to Diverging Color Maps for Scientific Visualization – The map yields images that are aesthetically pleasing. – The map has a maximal perceptual resolution. – Interference with the shading of 3D surfaces is minimal. – The map is not sensitive to vision deficiencies. – The order of the colors should be intuitively the same for all people. – The perceptual interpolation matches the underlying scalars of the map

  Claasification of color maps

• qualitative / nominal color maps: discrete, unordered classes
• sequential / ordinal / saturation color maps: hue is nearly fixed, (nearly monochromatic), difference in saturation and lightness indicate numericial difference
• diverging / ratio / bipolar / doubleended color maps: two major color components
• cyclic: change in lightness of two different colors that meet in the middle and beginning/end at an unsaturated color

Collection of perceptually accurate colormaps

• https://colorcet.holoviz.org/
• colorcet, https://github.com/holoviz/colorcet
• https://www.kennethmoreland.com/color-maps/
• https://github.com/1313e/CMasher

Caveats

• https://arxiv.org/abs/1509.03700
• Diverging color map and Mach bands
• Don’t use color map like jet or hot
  • 2020, Nature Communication, The misuse of colour in science communication

Examples

Choose which type of color scale to use

• qualitative: bar chart, violin plot, etc. To illustrate contrast, use complementary hues, to illustrate similarity, use analog hues.
• sequential: indicates difference toward one direction, like expression level, frequency, etc
• diverging: indicates difference toward two direction, like Z-score of differentially expressed genes

Customization of color maps

• See https://matplotlib.org/2.0.2/examples/pylab_examples/custom_cmap.html

Qualitative

• Choose of discrete colors is rather subjective, which you think is beautiful, you boss may tnink it’s ugly …
• May pick the one you love from ggsci, Wes Anderson Palettes, GeoDataViz-Toolkit, https://pattern-library.economist.com/color.html, or other resources, or anything you want …

Sequential

• Perceptually uniform cmap recommanded by seaborn here
  • rocket,mako,flare,crest,magma,viridis
• Perceptually uniform cmap by matplotlib
  • viridis, plasma, inferno, magma, cividis

Diverging

• seaborn
  • vlag,icefire
• matplotlib
  • “BrBG and RdBu are good options. coolwarm is a good option, but it doesn’t span a wide range of L values”

python examples

• See this notebook

R example

• https://www.nceas.ucsb.edu/sites/default/files/2020-04/colorPaletteCheatsheet.pdf
• pals package
  • The author also recommand coolwarm for diverging palette