Sequential and divergingPalettes for maps and heatmaps
Seven palettes for data with an order, and two well-known ones to avoid.Each is measured here with this site's color engine: whether its lightness climbs evenly, whether it keeps its order in grayscale, and whether its steps and its two sides can still be told apart with color vision deficiency.
The nine palettes, low to high. Select one to read why.
Hours of daylight by latitude and month, in nine classes.Choose a palette and a type of color vision; every cell keeps its number, so color is never the only cue.
The controls need JavaScript. Without it the map shows viridis, as typical vision sees it.
Sequential
Diverging
Vision
Palette
viridisSequential
Closest neighboring classes
ΔE 10.70Distinct
Verdict
Use it
Hours of daylight
5.25–6.75
6.75–8.25
8.25–9.75
9.75–11.25
11.25–12.75
12.75–14.25
14.25–15.75
15.75–17.25
17.25–18.75
Shorter daysAbout 12 hours, day as long as nightLonger days
Hours of daylight on the 15th of each month, by latitude
Latitude
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
60°N
6.3
8.8
11.3
14.2
16.8
18.4
17.7
15.4
12.5
9.7
7.1
5.6
45°N
8.9
10.2
11.6
13.3
14.7
15.4
15.1
13.9
12.3
10.7
9.3
8.6
30°N
10.3
11.0
11.8
12.7
13.5
13.9
13.8
13.1
12.2
11.3
10.5
10.1
15°N
11.2
11.5
11.9
12.3
12.7
12.9
12.8
12.5
12.1
11.7
11.3
11.1
0°
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
12.0
15°S
12.8
12.5
12.1
11.7
11.3
11.1
11.2
11.5
11.9
12.3
12.7
12.9
30°S
13.7
13.0
12.2
11.3
10.5
10.1
10.2
10.9
11.8
12.7
13.5
13.9
45°S
15.1
13.8
12.4
10.7
9.3
8.6
8.9
10.1
11.7
13.3
14.7
15.4
60°S
17.7
15.2
12.7
9.8
7.2
5.6
6.3
8.6
11.5
14.3
16.9
18.4
Day length works with both kinds of palette.Read as a quantity, from short days to long ones, it is sequential.Read against 12 hours, when day and night are equal, it is diverging: the middle class holds the equator all year, and every latitude in March and September.
The hours are computed, not measured: sunrise to sunset at the center of the sun, with Cooper's 1969 formula for the sun's declination and no atmospheric refraction, so real days run a few minutes longer.The classes are 1.5 hours wide and centered on 12 hours, so every palette colors the same cells the same class.
Measured
Each palette, measured
Lightness says whether a palette shows order; CIEDE2000 under simulation says whether its steps, and its two sides, can still be told apart.The verdict follows from the checks: a sequential palette must rise in lightness and keep that order in grayscale, a diverging one must keep its sides at least noticeably apart under every color vision.
Lightnessis OKLCH L, from 0 for black to 1 for white.A sequential palette should rise at every step; a diverging one should be lightest in the middle and darken toward both ends.
Even stepsis the smallest step in lightness divided by the largest. At 1.00, equal steps in the data look equal.
Closest neighborsis the smallest CIEDE2000 difference between two neighboring stops, under any of the five simulations, grayscale included. Under 5 they read as a close match.
Balanceis the largest difference in lightness between a stop and its mirror across the middle. Near 0, equal distances above and below the middle carry equal weight.
Two sides apartis the smallest CIEDE2000 difference between a stop and its mirror, under typical vision and the three dichromacies.Grayscale is left out: a diverging palette's two sides are equally light by design, so no diverging palette can show its side in gray.
Color vision
As each color vision sees them
Each palette under typical vision, the three dichromacies at full severity, and in grayscale.A sequential palette should still run from one end to the other; a diverging one should still show two sides.
viridisSequential
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
cividisSequential
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
BluesSequential
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
YlGnBuSequential
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
jetSequential
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
RdBuDiverging
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
PuOrDiverging
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
BrBGDiverging
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
RdYlGnDiverging
Typical vision
Protanopia
Deuteranopia
Tritanopia
Achromatopsia
The simulation is the model of Machado, Oliveira and Fernandes (2009) at full severity, the one this site's simulator uses; achromatopsia is shown as luminance only, which is also how the palette prints in black and white.A simulation approximates how colors collapse; it does not reproduce anyone's vision.
Use it well
Which one, and why
First decide what kind of data you have; then pick from that kind.Every reason below is a measurement from this page.
Sequentialwhen the data runs from low to high and low is not special: counts, rates, hours of daylight.Lightness carries the order, so it must rise at every step.
Divergingwhen the data has a meaningful middle and runs away from it both ways: change from last year, anomalies, correlation, days longer or shorter than nights.Hue carries the side; lightness carries the distance from the middle.
Neitherwhen the data has no order at all, such as categories: use thechart palettesinstead.
viridis
SequentialUse it
Use it forThe default for a quantity on screen: heat maps, density, any value from low to high.Lightness climbs in near-even steps (0.82), so equal differences in the data look about equal, and it keeps its order in grayscale.
Watch forIts stops 8 and 9, green-yellow to yellow, come within ΔE 4.55 under protanopia.Label the top of the legend.
cividis
SequentialUse it
Use it forWhen readers with color vision deficiency are a main audience: it was designed for them.Every neighboring step stays at ΔE 6.83 or more under all five simulations, the only palette here that does.
Watch forIt runs from blue to yellow only, so it is quieter than viridis.Its lightness steps are the most even here (0.84).
Blues
SequentialUse it
Use it forA map of one quantity in a few classes: rates, shares, density by region.A single hue keeps the message simple: darker means more.
Watch forIts lightest stop is 1.03:1 against white, so a white page swallows it: outline every region.In grayscale its stops 1 and 2 come within ΔE 3.53.
YlGnBu
SequentialUse it
Use it forMore classes than one hue can separate: hue changes with lightness, so neighboring classes are at least ΔE 8.19 apart in typical vision, against 5.09 for Blues.
Watch forIts pale end bunches: stops 1 and 2 are ΔE 2.44 apart in grayscale and 3.18 under tritanopia.Label the low end of the legend.
jet
SequentialAvoid
Why to avoid itLightness rises to stop 6 and falls again, so the brightest band, cyan to yellow, looks like a feature of the data wherever it lands.In grayscale, stops 4 and 5 are the same gray (ΔE 0.22) and the order is lost.
Use insteadviridis, which spans the same range of lightness without turning back.
RdBu
DivergingUse it
Use it forData with a meaningful middle: change from a baseline, anomalies, correlation from −1 to 1.Its two sides are balanced in lightness (within 0.013) and stay at least ΔE 16.47 apart under every color vision.
Watch forIn grayscale the two sides meet, never more than ΔE 3.65 apart: print the values, or a legend with signs.
PuOr
DivergingUse it
Use it forA diverging palette with no red and no green, and none of red's sense of alarm.Its two sides stay at least ΔE 20.12 apart under every color vision.
Watch forIts purple side is darker than its orange side, by 0.172 in lightness at the ends, so high values look weightier than low ones.If both directions matter equally, use RdBu.
BrBG
DivergingUse it
Use it forTwo directions with an earthy feel: wetter or drier, vegetation gained or lost.Balanced within 0.034.
Watch forUnder protanopia, stops 5 and 6 come within ΔE 3.23.Label the classes nearest the middle.
RdYlGn
DivergingAvoid
Why to avoid itUnder deuteranopia, the most common color vision deficiency, its two sides meet: the stops beside the middle are ΔE 0.52 apart, and even its two ends only 9.57.A reader cannot tell above the middle from below it.
Use insteadRdBu or PuOr, whose sides stay apart under every color vision.
In code
Already in your charting library
All seven ship with matplotlib and with d3.The calls below give exactly the nine stops measured here; for the three diverging palettes in matplotlib, pass the stops yourself.
Each recommended palette in matplotlib and d3
Palette
matplotlib
d3
Note
viridis
colormaps["viridis"].resampled(9)
d3.quantize(d3.interpolateViridis, 9)
Both give exactly these nine.
cividis
colormaps["cividis"].resampled(9)
d3.quantize(d3.interpolateCividis, 9)
d3's cividis is a polynomial fit, up to 13 of 255 away in one channel; matplotlib's is exact.
Blues
colormaps["Blues"].resampled(9)
d3.schemeBlues[9]
Both give exactly these nine.
YlGnBu
colormaps["YlGnBu"].resampled(9)
d3.schemeYlGnBu[9]
Both give exactly these nine.
RdBu
ListedColormap(stops)
d3.schemeRdBu[9]
matplotlib's own RdBu runs through ColorBrewer's 11 classes: these nine and a darker stop at each end.
PuOr
ListedColormap(stops)
d3.schemePuOr[9].slice().reverse()
d3 lists PuOr from the purple end, so reverse it to match ColorBrewer and this page.
BrBG
ListedColormap(stops)
d3.schemeBrBG[9]
matplotlib's own BrBG runs through the 11 classes, as RdBu does.
In matplotlib, colormaps and ListedColormap come from matplotlib and matplotlib.colors; stops is the palette's list from the JSON export below.Every name was run against matplotlib 3.10.8 and d3-scale-chromatic 3.1.0 before publishing.
Code
Export the palettes
The seven recommended palettes as tokens: scale-viridis-1 to 9, and so on, low to high.
Copy a format and rename the tokens to fit your system.
viridis, cividis and jet are matplotlib's 256-color tables at entries 0, 32, 64 … 224 and 255; the ColorBrewer schemes are their 9-class versions, the same in the colorbrewer package and in d3.Lightness is OKLCH L; differences are CIEDE2000; simulations apply Machado's matrices in linear light at full severity.Node tests recompute every number on this page from the site's color engine.Checked with axe and the keyboard in Chromium; not yet tested with a screen reader.Last reviewed on 14 October 2026.
This product includes color specifications and designs developed by Cynthia Brewer (http://colorbrewer.org/).