Short answer
Use a diverging scale only when the data have a meaningful midpoint — zero, an average, a target — and both directions away from it matter. Two hues meet at a light, neutral centre and darken symmetrically towards each extreme. Choose the two hues so they stay distinct for readers with colour-vision deficiency: blue against orange survives; red against green usually does not.
A diverging scheme is two sequential scales joined at a shared light centre. ColorBrewer describes such schemes as giving equal emphasis to a critical mid-range value and to the extremes at both ends, and says they work best when the central break is meaningfully related to the data: a mean, a median, zero change. Temperature anomalies, election swings, profit and loss, and survey agreement all qualify. Population density does not — it has no meaningful middle — and putting it on a diverging scale invents a boundary at whatever value happens to land in the centre. Before choosing a diverging palette, name the midpoint; if you cannot, use a sequential one.
Real data are rarely symmetric about their midpoint. If changes run from −5% to +40%, a symmetric scale centred on zero spends half its colours on a range with little data. There are two defensible responses. One is to keep the scale symmetric (−40 to +40) so equal colours mean equal magnitudes in either direction, accepting that the negative arm is barely used. The other, which ColorBrewer notes for cases such as population change, is to move the critical break towards one end so more classes serve the side with more data — at the price that the darkest blue and darkest orange no longer represent the same distance from zero. Say which choice you made in the legend; readers otherwise assume symmetry.
The two arms must stay distinguishable when hue perception is reduced, because the sign of the value is carried by hue alone: a class at −2 and a class at +2 are designed to have the same lightness. The table tests three pairings at matched OKLCH lightness and chroma. Red against green, the default in many spreadsheet tools because of its traffic-light meaning, keeps the two arms far apart for typical vision but brings them close under simulated deuteranopia, so a reader with the most common colour-vision deficiency can tell 'how far from zero' but not 'which way'. Blue against orange stays well separated, which is why it and purple–green dominate designed diverging schemes. Moreland's cool–warm map, a blue-to-red design with a neutral centre, is a widely used scientific example.
The centre should be neutral and light — near-white or pale grey — so values near the midpoint recede and extremes stand out. A saturated centre (yellow is a frequent offender) makes the least interesting values the most eye-catching. The two ends should reach similar darkness; if one arm ends in dark navy and the other in mid orange, the navy side looks more extreme at equal values. On dark backgrounds the logic inverts: a dark neutral centre with arms that brighten outwards. Finally, for continuous diverging maps, check that each arm is itself perceptually uniform; a diverging map is only as good as its two sequential halves.
| Pairing | Arm colours (OKLCH L 0.55, C 0.13) | ΔE00 typical | ΔE00 deuteranopia | ΔE00 deuteranopia, pale classes (L 0.80) |
|---|---|---|---|---|
| Blue ↔ orange | #3772bb / #a65c07 | 46.9 | 59.8 | 40.1 |
| Purple ↔ green | #8959aa / #38853e | 46.5 | 45.4 | 29.8 |
| Red ↔ green | #b14e49 / #448335 | 54.3 | 2.5 | 1.3 |
Why: A diverging scale was used for data with no meaningful midpoint.
Fix: Use a sequential scale unless you can name the midpoint.
Why: The arms are red and green, which collapse for red–green colour-vision deficiency.
Fix: Use blue–orange or purple–green, and add + and − signs to labels.
Why: The centre colour is saturated (often yellow).
Fix: Make the centre a light neutral grey or off-white.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
ConventionStrong evidence
ColorBrewer describes diverging schemes as giving equal emphasis to mid-range critical values and extremes at both ends, best used when the central break is meaningful, such as a mean, median or zero.
Source: ColorBrewer 2.0 — colour advice for cartography; ColorBrewer.org: An Online Tool for Selecting Colour Schemes for Maps
ConventionModerate evidence
ColorBrewer notes that a diverging scheme may need its critical break moved towards one end of the sequence when the data are asymmetric, giving population change as an example.
Colourwise analysisModerate evidence
At matched lightness and chroma, the two arms of a red–green diverging scale come much closer together under simulated deuteranopia than blue–orange or purple–green arms do.
Based on: Computed by Colourwise with the Viénot–Brettel–Mollon simulation and CIEDE2000; see the table.
Caveat: Simulation is of full dichromacy; exact colours vary between published schemes.
Source: Digital video colourmaps for checking the legibility of displays by dichromats; CIE (International Commission on Illumination) publications
FactModerate evidence
Moreland proposed a cool–warm diverging colour map with a neutral centre for scientific visualisation, and described the rainbow map as known for obscuring data and introducing artefacts.
Reviewed 29 September 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.