Best Colour Palettes for Scientific Figures

Colour is one of the most powerful and most misused tools in scientific figures. The right palette guides the reader's eye and represents data honestly; the wrong one invents patterns that aren't there and shuts out a meaningful fraction of your audience. This guide covers the palettes to use, the ones to avoid, and the accessibility rules that make your figures clear for everyone.

Why colour choice is a scientific decision

Around one in twelve men and roughly one in two hundred women have some form of colour vision deficiency, most commonly difficulty distinguishing red from green. If your figure relies on those colours to carry meaning, a real portion of your readers — and possibly a reviewer or editor — simply cannot read it. Beyond accessibility, a poorly chosen colour scale can distort the data itself, which makes palette choice a matter of scientific accuracy, not just aesthetics.

Avoid the rainbow (jet) colormap

The rainbow or "jet" colormap remains common but is widely discouraged for good reason. It is not perceptually uniform: equal steps in your data do not produce equal-looking steps in colour. That creates sharp false boundaries in smooth regions (often around the yellow and cyan bands) while flattening genuine variation elsewhere. It also collapses into ambiguity for colourblind viewers and turns to mush in greyscale. If you have been reaching for rainbow by habit, switching is the single biggest improvement you can make.

Palettes for continuous data

Viridis and its relatives (Magma, Inferno, Plasma) are perceptually uniform, colourblind-friendly, and legible in greyscale — they have become the default for heatmaps and continuous scales in modern plotting libraries. Cividis is specifically optimised for colour vision deficiency. For data that diverges around a meaningful midpoint (such as positive and negative values), use a diverging palette like blue-to-white-to-red, which keeps the centre neutral and reads symmetrically.

Palettes for categorical data

For distinct categories rather than a continuous scale, choose a qualitative palette designed for accessibility. The Okabe-Ito palette is an eight-colour set built explicitly to be distinguishable for all common types of colour blindness, and ColorBrewer offers curated colourblind-safe qualitative sets. Whichever you pick, never rely on colour alone to separate categories — add a second cue such as shape, line style, or a direct label so the figure still works in greyscale or for a colourblind reader.

Practical rules that always apply

A few habits keep figures clear regardless of palette. Use as few colours as the information requires — two or three plus neutrals is usually plenty. Maintain strong contrast between elements and the background. Reserve your most saturated colour for the most important element, and let everything else recede. Test your figure by converting it to greyscale and by running it through a colourblindness simulator; if it survives both, it will read for almost everyone. Finally, keep your palette consistent across all the figures in a paper so colours mean the same thing throughout.

A quick workflow

Start by deciding whether your data is continuous, diverging, or categorical, then pick the matching palette type. Choose a colourblind-safe option from the start rather than fixing it later. Build the figure, then check it in greyscale and a simulator before you finalise. This takes only a couple of minutes and prevents the most common — and most avoidable — figure problems reviewers flag.

Related reading: Publication-Quality Figures: DPI, TIFF & CMYK and 3D Scientific Illustration: A Beginner's Guide.