Scientific Color Palettes: Okabe-Ito, Paul Tol, ColorBrewer, Viridis, Cividis, RdBu, and Crameri vik

Which Scientific Color Palette Should You Use?

If you are making a research figure, the fastest safe rule is this: use Okabe-Ito or Paul Tol Bright for unordered categories, Viridis or Cividis for one-way continuous values, and ColorBrewer RdBu or Crameri vik when the data has a meaningful midpoint.

That one decision prevents most color mistakes in posters, journal figures, heatmaps, correlation matrices, classroom diagrams, and slide decks. The palette is not decoration. It tells readers which values belong together, which values are higher or lower, and whether the difference they are seeing is real.

Scientific Color Palette Generator

Create colorblind-friendly palettes for research figures and export copy-ready hex codes.

Generate a palette →

A quick comparison of the palette families covered in this guide. The cover image is a 1920x1080 WebP and stays under 100 KB.

Quick Decision Table

Use the table before opening any color picker. If the data type is wrong, no amount of aesthetic tuning will rescue the figure.

Okabe & Ito Palette: The Categorical Default

The Okabe-Ito palette comes from Masataka Okabe and Kei Ito's Color Universal Design work. It was designed so that colors remain distinguishable for readers with common forms of color vision deficiency, while still looking vivid enough for presentations and printed figures.

It is the best first choice when your groups have names but no ranking: treatment A vs treatment B, species, cell types, experimental conditions, or survey groups.

Okabe-Ito works best for unordered categories. Do not use it for a heatmap or any smooth numeric scale.

Copy it into Python:

okabe_ito = [
    "#E69F00", "#56B4E9", "#009E73", "#F0E442",
    "#0072B2", "#D55E00", "#CC79A7", "#000000",
]

The main limitation is category count. Eight colors is already a lot for a legend. If you need more groups, merge minor categories, split the figure into panels, or add a second encoding such as marker shape, line style, or hatch pattern.

Paul Tol Bright: A Softer Line-Chart Alternative

Paul Tol's color schemes are built around a practical scientist's problem: figures need to be clear on screen, on paper, and for color-blind readers. His Bright qualitative scheme is especially useful for multi-line plots because the colors have enough separation without feeling as harsh as many default plotting cycles.

Paul Tol Bright is a good fit for line charts, labeled series, and figures where seven categories are enough.

Use Paul Tol Bright when you want a category palette that feels publication-ready but not overly saturated. It is also a good companion to direct labels: place the label close to the line or point it describes so readers do not have to match distant legend swatches.

Viridis and Cividis: Sequential Palettes for Continuous Data

Use a sequential palette when the data moves in one direction: low to high, weak to strong, small to large. This is where Viridis and Cividis are stronger than categorical palettes.

Viridis became popular because it is perceptually ordered: equal steps in the data look more like equal steps in color. It also holds up far better than rainbow or jet maps when printed in grayscale. Cividis was designed with color-vision deficiency robustness as an even stronger constraint, trading some visual drama for dependable reading.

Viridis and Cividis are for continuous magnitude. They are not good legend colors for unrelated groups.

Use Viridis for:

• heatmaps of expression, density, or signal intensity
• microscopy intensity maps
• elevation, concentration, or uncertainty surfaces
• any single variable that rises from low to high

Use Cividis when:

• the figure must be especially robust for red-green color vision differences
• the chart will be projected in poor lighting
• a more restrained blue-yellow ramp fits the publication style

Python example:

import matplotlib.pyplot as plt

plt.imshow(matrix, cmap="viridis")  # or cmap="cividis"
plt.colorbar(label="Signal intensity")

ColorBrewer RdBu: The Familiar Diverging Scale

ColorBrewer began as a cartography project, but its palettes are now widely used in scientific visualization. RdBu is one of its most recognizable diverging schemes: red on one side, blue on the other, and a light neutral center.

Use it when a value can go in two meaningful directions around a midpoint: positive vs negative residuals, above vs below baseline, hot vs cold anomalies, or correlation matrices.

ColorBrewer RdBu is a diverging palette. Use it only when the center point has meaning.

The 11-color RdBu set:

#67001F, #B2182B, #D6604D, #F4A582, #FDDBC7, #F7F7F7,
#D1E5F0, #92C5DE, #4393C3, #2166AC, #053061

Two cautions matter. First, red and blue can carry domain-specific meaning. In some fields red means danger, hot, loss, or significance; in others it is just one side of a comparison. Second, do not use RdBu for ordinary low-to-high values. If there is no meaningful midpoint, choose Viridis or Cividis instead.

Crameri vik: A Scientific Diverging Map

Fabio Crameri's Scientific Colour Maps are designed around scientific accuracy: perceptual uniformity, color-vision-deficiency friendliness, and grayscale behavior. vik is a diverging map for fields where the midpoint should read as neutral and the two sides should feel balanced.

Crameri vik is a scientific diverging map. Use the official LUT files for production figures; the image above is a compact visual preview.

Good uses for vik include:

• anomaly maps around zero
• positive and negative model residuals
• geoscience and climate fields with a neutral center
• any symmetric scientific quantity where the midpoint is not just another value

For production work, download the official Scientific Colour Maps package and use the LUT file for your software. A short blog swatch can preview the character of the palette, but a scientific colormap should be applied from its full lookup table.

How to Pick the Right Palette in Figviz Workflows

If you are using Figviz to generate a figure, include the palette family in the prompt rather than describing vague style preferences.

Better prompts:

• "Create a grouped bar chart with the Okabe-Ito palette and direct labels for each treatment."
• "Create a heatmap using Viridis with a labeled color bar from low signal to high signal."
• "Create a correlation matrix using a diverging RdBu scale centered at zero."
• "Create an anomaly map using Crameri vik, with the midpoint shown as neutral white."

For science diagrams rather than charts, use color more sparingly. In a scientific diagram maker, the palette should separate structures without turning the whole image into a legend puzzle. In an AI chart generator, the palette carries data meaning directly, so the family decision matters more.

For research submissions, pair palette choice with typography and export checks. Our guide to scientific poster and figure fonts covers label size and font choices, and the scientific diagram guide covers figure preparation workflows.

Common Mistakes

Using Rainbow or Jet for Continuous Data

Rainbow-style ramps create false visual bands because the perceived brightness does not change smoothly. A reader may see a boundary that is not present in the data. Replace rainbow with Viridis, Cividis, or a scientific colormap from Crameri.

Using Categorical Colors for Ordered Values

Okabe-Ito and Paul Tol Bright are excellent for categories, but they imply separation, not magnitude. A heatmap with categorical colors is hard to interpret because readers cannot reliably judge which cells are higher or lower.

Using a Diverging Palette Without a Midpoint

RdBu and vik need a meaningful center. If the values run from 0 to 100 with no natural middle, a diverging palette makes the midpoint look more important than it is.

Relying on Color Alone

Okabe and Ito's own Color Universal Design guidance emphasizes redundant coding: use shapes, labels, positions, line styles, and patterns as well as color. A figure should still communicate when printed in grayscale or projected in a dim room.

Source Notes

The palettes in this guide are drawn from primary or near-primary resources:

• Okabe & Ito's Color Universal Design guide
• Paul Tol's colour scheme notes
• ColorBrewer for RdBu
• Matplotlib's colormap documentation for Viridis and Cividis context
• Fabio Crameri's Scientific Colour Maps for vik
• Google Search Central's helpful content guidance for how this post is structured around reader tasks rather than keyword stuffing

Summary

Choose the palette by data type:

1. Unordered groups: Okabe-Ito first, Paul Tol Bright when you prefer a softer seven-color set.
2. One-way magnitude: Viridis as the general default, Cividis when accessibility robustness is the priority.
3. Meaningful midpoint: ColorBrewer RdBu for familiar red-blue divergence, Crameri vik for scientific fields that need a balanced neutral center.

Then run three checks before publishing: colorblind simulation, grayscale preview, and a label/legend review. If a reader can still understand the figure without perfect color perception, the palette is doing its job.

Scientific Color Palette Generator

Build accessible palettes with copy-ready hex codes for charts, diagrams, posters, and papers.

Try free →

Related Resources

• Okabe-Ito Palette Hex Codes Reference
• Scientific Poster & Figure Fonts
• How to Make Scientific Diagrams for Research Papers
• AI Chart Generator
• Scientific Diagram Maker

FAQ

Q: What is the best scientific color palette overall? A: There is no single best palette. Okabe-Ito is the safest categorical default, Viridis is the safest general sequential default, and RdBu or Crameri vik are better when your data diverges around a meaningful center.

Q: Are Okabe-Ito and Wong palette the same thing? A: In most research-figure discussions, yes. Bang Wong popularized the same eight-color set in a Nature Methods column, while Okabe and Ito's Color Universal Design work is the original source most people cite for the palette.

Q: What is Paul Tol Bright best for? A: Paul Tol Bright is best for multi-line plots and small categorical sets where seven colors are enough. It feels less harsh than many default plotting palettes while staying clear for color-blind readers.

Q: When should I use Cividis instead of Viridis? A: Use Cividis when color-vision accessibility is more important than maximum visual contrast. Use Viridis when you want a stronger and more familiar continuous ramp that still behaves well in grayscale.

Q: Is ColorBrewer RdBu colorblind-safe? A: RdBu can work well as a diverging palette, but you should still test the final figure. The issue is not only the palette; line thickness, labels, adjacent colors, and whether red-blue is meaningful in your field all matter.

Q: Why use Crameri vik instead of RdBu? A: vik is designed as a scientific colour map with balanced perception and color-vision-deficiency behavior. It is a strong option for anomaly fields and residual maps where you want the two sides of the scale to feel visually balanced.

Q: How many categorical colors can I safely use? A: Keep categorical colors to about six whenever possible. Okabe-Ito gives eight and Paul Tol Bright gives seven, but large legends are hard to read. Add shapes, direct labels, or separate panels when the category count grows.

Q: Can I use these palettes in Excel, PowerPoint, Python, or R? A: Yes. Copy the hex codes into Office custom colors, use them in Python matplotlib lists, or pass them to R scale functions. For full continuous colormaps such as Crameri vik, use the official lookup-table files.