Every time I pick colors for a figure, I end up looking up palette codes again. So I gathered the palettes I use most as swatches on one page. Click a color to copy it.
c(...)), Python (a list), CSS (custom properties), or JSON.The examples below use only base R graphics, with no extra packages.
People perceive color differently. Palettes that rely on hue alone, such as red versus green or blue versus purple, can be hard to tell apart for some readers. The palettes below were designed with that in mind. For simulating how they look, see Simulating Color Vision Deficiency in R.
An eight-color palette proposed by Masataka Okabe and Kei Ito (Okabe and Ito 2008). It was also introduced in a Nature Methods column (Wong 2011) and is common in journal figures. In R, palette.colors(palette = "Okabe-Ito") returns it (R 4.0 or later) (R Core Team, n.d.).1 Note that R’s built-in version has nine colors: a gray (#999999) is added at the end.2 The leading black is handy for elements that should not depend on hue, such as lines and text.
As an example, here is a scatter plot of weight against fuel economy in mtcars, colored by the number of carburetors (carb). It takes the seven colors of R’s built-in version that remain after dropping black and gray.
# 9 colors: black, 7 chromatic colors, gray. Keep the 7 chromatic ones.
okabe_ito <- unname(palette.colors(palette = "Okabe-Ito"))[2:8]
carb <- factor(mtcars$carb)
par(mar = c(4, 4, 1, 1))
plot(mtcars$wt, mtcars$mpg, col = okabe_ito[carb], pch = 19, cex = 1.4,
xlab = "wt", ylab = "mpg")
legend("topright", legend = levels(carb), title = "carb", pch = 19,
col = okabe_ito[seq_len(nlevels(carb))], bty = "n")
One of the schemes compiled by Paul Tol, an instrument scientist at SRON, the Netherlands Institute for Space Research. Bright, Vibrant, and Muted all appear in his color-scheme document, which SRON published as a technical note rather than a journal paper (Tol 2021). It has seven colors with a calmer look than Okabe-Ito.
As an example, here is a box plot of insect counts by insecticide type (spray) from InsectSprays.
More saturated than Bright, and it includes orange. It stands out on both light and dark backgrounds.
As an example, here is a line chart of mean chick weight by diet.
tol_vibrant <- c("#EE7733", "#0077BB", "#33BBEE", "#EE3377",
"#CC3311", "#009988", "#BBBBBB")
# mean weight at each Time (rows) for each Diet (columns)
m <- with(ChickWeight, tapply(weight, list(Time, Diet), mean))
cols <- tol_vibrant[seq_len(ncol(m))]
par(mar = c(4, 4, 1, 1))
matplot(as.numeric(rownames(m)), m, type = "l", lty = 1, lwd = 2.5,
col = cols, xlab = "Time", ylab = "weight")
legend("topleft", legend = colnames(m), title = "Diet",
col = cols, lty = 1, lwd = 2.5, bty = "n")
For when you need up to ten colors. The lower saturation keeps a crowded plot from looking noisy, at the cost of smaller differences between neighbors. The pale grey at the end suits missing values or an “other” category.
As an example, here is a stacked bar chart of each region’s share of telephones worldwide (WorldPhones) by year. There are seven regions, so it uses seven of Muted’s ten colors.
tol_muted <- c("#CC6677", "#332288", "#DDCC77", "#117733", "#88CCEE",
"#882255", "#44AA99", "#999933", "#AA4499", "#DDDDDD")
# rows: regions, columns: years, each column sums to 1
share <- t(prop.table(WorldPhones, margin = 1))
par(mar = c(4, 4, 1, 6), xpd = TRUE) # xpd lets the legend sit in the margin
barplot(share, col = tol_muted[seq_len(nrow(share))],
ylab = "share of telephones", legend.text = rownames(share),
args.legend = list(x = "right", inset = c(-0.2, 0), bty = "n"))
Stéfan van der Walt and Nathaniel Smith designed it to replace jet, Matplotlib’s former default colormap (Smith and van der Walt 2015). It was presented in a SciPy 2015 talk rather than a paper. It became Matplotlib’s default in 2.0, and R users can get it from the viridis package (Garnier et al. 2024). It was later discussed in the literature too. A PLOS ONE paper by Nuñez and colleagues describes viridis as the current “gold standard” colormap (Nuñez et al. 2018). The paper proposes cividis, a colormap derived from viridis and further optimized for color vision deficiency. It also notes that viridis is optimal for viewers with normal color vision, rather than being optimized specifically for color vision deficiency.
Lightness changes monotonically from low to high values, which suits heatmaps and other figures that encode magnitude as color. Eight steps are shown here. The ribbon above them blends the eight colors smoothly.
As an example, here is a heat map of a volcano’s elevation data (volcano). The eight colors are interpolated to 256. Base R’s hcl.colors(256, "viridis") gives a similar look, though its color values differ slightly.
viridis_8 <- c("#440154", "#46337E", "#365C8D", "#277F8E",
"#1FA187", "#4AC16D", "#9FDA3A", "#FDE725")
# interpolate the 8 colors to 256 steps
pal <- colorRampPalette(viridis_8, space = "Lab")(256)
par(mar = c(1, 1, 1, 1))
image(volcano, col = pal, axes = FALSE)
These were not made for color vision diversity. If you use them, combine color with line types, markers, or labels.
Tableau’s default categorical palette, redesigned for Tableau 10 in 2016 (Stone 2016). It is familiar from dashboards and has ten colors. R also ships it as palette.colors(palette = "Tableau 10") (R Core Team, n.d.). It is not R’s default palette, though. Since R 4.0, the default is a different palette called "R4"; “Tableau 10” is one of the built-in palettes you can choose.
As an example, here is a bar chart of the ten largest of the world’s major landmasses (islands). Each bar takes one of the ten Tableau 10 colors built into R.
This site’s background and other colors are based on Flexoki, a color system by Steph Ango. Its accent colors come in 600 for light backgrounds and 400 for dark ones.
As an example, here is the same scatter plot as in the Okabe-Ito example, using the 600 accents on Flexoki’s paper color (#FFFCF0).
flexoki_600 <- c("#AF3029", "#BC5215", "#AD8301", "#66800B",
"#24837B", "#205EA6", "#5E409D", "#A02F6F")
carb <- factor(mtcars$carb)
# background (bg), axis lines and text (fg) colors
op <- par(mar = c(4, 4, 1, 1), bg = "#FFFCF0", fg = "#100F0F",
col.axis = "#100F0F", col.lab = "#100F0F")
plot(mtcars$wt, mtcars$mpg, col = flexoki_600[carb], pch = 19, cex = 1.4,
xlab = "wt", ylab = "mpg")
legend("topright", legend = levels(carb), title = "carb", pch = 19,
col = flexoki_600[seq_len(nlevels(carb))], bty = "n", text.col = "#100F0F")
par(op) # restore the previous settingsAs an example, here is the same scatter plot with the 400 accents on a dark background (#1C1B1A).
flexoki_400 <- c("#D14D41", "#DA702C", "#D0A215", "#879A39",
"#3AA99F", "#4385BE", "#8B7EC8", "#CE5D97")
carb <- factor(mtcars$carb)
op <- par(mar = c(4, 4, 1, 1), bg = "#1C1B1A", fg = "#FFFCF0",
col.axis = "#FFFCF0", col.lab = "#FFFCF0")
plot(mtcars$wt, mtcars$mpg, col = flexoki_400[carb], pch = 19, cex = 1.4,
xlab = "wt", ylab = "mpg")
legend("topright", legend = levels(carb), title = "carb", pch = 19,
col = flexoki_400[seq_len(nlevels(carb))], bty = "n", text.col = "#FFFCF0")
par(op)The vector from “Copy all” → R can be set with palette(), after which it works for plots that specify colors by number or factor.
The list from Python can be set as Matplotlib’s color cycle.
import matplotlib.pyplot as plt
okabe_ito = ["#E69F00", "#56B4E9", "#009E73", "#F0E442",
"#0072B2", "#D55E00", "#CC79A7"]
plt.rcParams["axes.prop_cycle"] = plt.cycler(color=okabe_ito)Paste the lines from CSS into :root to use them as custom properties.
:root {
--okabe-ito-orange: #E69F00;
--okabe-ito-sky-blue: #56B4E9;
}The swatches on this page are drawn in the browser by JavaScript, not generated as images (the R figures below them are images produced by R at render time). No libraries are involved. Three files do all the work:
| File | Role |
|---|---|
palettes-data.js |
An array holding only the color data (names and HEX) |
palette.js |
Reads the data and builds the swatches and copy features |
palette.css |
Appearance. Light and dark follow the site theme |
The text only needs an empty block like the one below. palette.js looks up the data by the data-palette value and fills the block in.
::: {data-palette="okabe-ito"}
:::The front matter of the QMD loads the CSS and JavaScript.
css: palette.css
include-after-body:
text: |
<script src="palettes-data.js"></script>
<script src="palette.js"></script>To add a palette, add an entry to palettes-data.js and place an empty block in the text. Copying uses navigator.clipboard, and falls back to execCommand("copy") where it is unavailable.
The actual files are included below in collapsible sections. Each starts with an overview comment, and sections inside have their own comments.
palette.colors() was added in R 4.0.0. It is listed under New Features in the R NEWS (Changes in R 4.0.0).↩︎
In the development source of R, colorstuff.R, .palette_colors_hex defines "Okabe-Ito" with nine colors, ending in gray = "#999999".↩︎