Paul Tol’s Colour Schemes

Introduction

Tol (2021) offers carefully chosen schemes, ready for each type of data, with colours that are:

Distinct for all people, including colour-blind readers,
Distinct from black and white,
Distinct on screen and paper,
Matching well together.

All the scales presented in Paul Tol’s technical note (issue 3.2, 2021-08-18) are implemented here, for use with base R graphics or ggplot2.

Qualitative data

According to Paul Tol’s technical note, the bright, contrast, vibrant and muted colour schemes are colour-blind safe.

The light colour scheme is reasonably distinct for both normal or colour-blind vision and is intended to fill labelled cells.

The pale and dark schemes are not very distinct in either normal or colour-blind vision and should be used as a text background or to highlight a cell in a table.

The qualitative colour schemes must be used as given (no interpolation): colours are picked up to the maximum number of supported values. Refer to the original document for details about the recommended uses (see references).

Scheme	Max. colours
`bright`	7
`high contrast`	3
`vibrant`	7
`muted`	9
`medium contrast`	3
`pale`	6
`dark`	6
`light`	9

bright

bright <- colour("bright")
plot_scheme(bright(7), colours = TRUE, names = TRUE, size = 0.9)

high contrast

highcontrast <- colour("high contrast")
plot_scheme(highcontrast(3), colours = TRUE, names = TRUE, size = 0.9)

vibrant

vibrant <- colour("vibrant")
plot_scheme(vibrant(7), colours = TRUE, names = TRUE, size = 0.9)

muted

muted <- colour("muted")
plot_scheme(muted(9), colours = TRUE, names = TRUE, size = 0.9)

medium contrast

mediumcontrast <- colour("medium contrast")
plot_scheme(mediumcontrast(6), colours = TRUE, names = TRUE, size = 0.9)

pale and dark

pale <- colour("pale")
plot_scheme(pale(6), colours = TRUE, names = TRUE, size = 0.9)


dark <- colour("dark")
plot_scheme(dark(6), colours = TRUE, names = TRUE, size = 0.9)

light

light <- colour("light")
plot_scheme(light(9), colours = TRUE, names = TRUE, size = 0.9)

Diverging data

If more colours than defined are needed from a given scheme, the colour coordinates are linearly interpolated to provide a continuous version of the scheme.

Scheme	Num. of colours	Bad data
`sunset`	11	#FFFFFF
`BuRd`	9	#FFEE99
`PRGn`	9	#FFEE99

sunset

sunset <- colour("sunset")
plot_scheme(sunset(9), colours = TRUE, size = 0.9)

BuRd

BuRd <- colour("BuRd")
plot_scheme(BuRd(9), colours = TRUE, size = 0.9)

PRGn

PRGn <- colour("PRGn")
plot_scheme(PRGn(9), colours = TRUE, size = 0.9)

Sequential data

If more colours than defined are needed from a given scheme, the colour coordinates are linearly interpolated to provide a continuous version of the scheme, with the exception of the discrete rainbow scheme (see below).

Scheme	Num. of colours	Bad data
`YlOrBr`	9	#888888
`iridescent`	23	#999999
`discrete rainbow`	23	#777777
`smooth rainbow`	34	#666666

YlOrBr

YlOrBr <- colour("YlOrBr")
plot_scheme(YlOrBr(9), colours = TRUE, size = 0.9)

iridescent

iridescent <- colour("iridescent")
plot_scheme(iridescent(23), colours = TRUE, size = 0.5)

rainbow

As a general rule, ordered data should not be represented using a rainbow scheme. There are three main arguments against such use (Tol 2021):

The spectral order of visible light carries no inherent magnitude message.
Some bands of almost constant hue with sharp transitions between them, can be perceived as jumps in the data.
Colour-blind people have difficulty distinguishing some colours of the rainbow.

If such use cannot be avoided, Paul Tol’s technical note provides two colour schemes that are reasonably clear in colour-blind vision. To remain colour-blind safe, these two schemes must comply with the following conditions:

discrete rainbow: This scheme must not be interpolated.
smooth rainbow: This scheme does not have to be used over the full range (Tol 2021 suggests starting at purple).

discrete_rainbow <- colour("discrete rainbow")
plot_scheme(discrete_rainbow(14), colours = TRUE, size = 0.7)

When using the smooth rainbow scheme:

Start off-white instead of purple if the lowest data value occurs often;
End at red instead of brown the highest data value occurs often.

smooth_rainbow <- colour("smooth rainbow")

## Start at purple instead of off-white
plot(smooth_rainbow(256, range = c(0.25, 1)))

## End at red instead of brown
plot(smooth_rainbow(256, range = c(0, 0.9)))

Diagnostic maps

Qualitative colour schemes

Diagnostic maps for the bright, vibrant, muted and light (from top to bottom) qualitative colour schemes.

Diverging schemes

Diagnostic maps for the sunset, BuRd and PRGn (from top to bottom) diverging colour schemes.

Diagnostic maps for the sunset, BuRd and PRGn diverging colour schemes.

Diagnostic maps for the sunset, BuRd and PRGn diverging colour schemes.

Sequential schemes

Diagnostic maps for the YlOrBr, iridescent, discrete rainbow and smooth rainbow (from top to bottom) sequential colour schemes.

Diagnostic maps for the YlOrBr, iridescent and smooth rainbow sequential colour schemes.

Diagnostic maps for the YlOrBr, iridescent and smooth rainbow sequential colour schemes.

References

Tol, Paul. 2021. “Colour Schemes.” Technical note SRON/EPS/TN/09-002 3.2. SRON. https://personal.sron.nl/~pault/data/colourschemes.pdf.