Share |

Dear NASA: No more rainbows, please

VISUAL | April 25, 2012

A color-shaded relief map of the far side of the moon. Image courtesy of NASA's Goddard Space Flight Center/DLR/ASU.

Drew Skau has a message for NASA:

"The visualization community has noticed your insistence on using rainbow color scales for representing continuous data. This is a plea to you (and anyone else doing the same thing) to stop."

Skau, a PhD computer science visualization student at the University of North Carolina - Charlotte, gives several reasons why rainbow color scales are a bad idea:

Color-blind people cannot use them, whereas they can use two hue diverging color scales.
The divisions between hues create edges in visualizations that are not actually significant.
The spectral order of hues has no inherent and intuitive meaning.
Because the color yellow activates both red and green cones in our eyes, the color appears more intensely bright, making it stand out. Most rainbow scales do nothing to compensate for this type of artifact.
Detail is actually harder to see in a rainbow.

Want the details? Check out Skau's original blog post at visual.ly, a visualization startup for which Skau occasionally works.

As for Skau's request, for the sake of accessibility, we hope that his pleas will be heard. But we sure will miss the pretty rainbows!

Via VizWorld.

Comments

Add New Comment

by hyutars | April 14, 2013 - 8:04pm

Thanks so much ^^ nice page.

Thanks so much ^^ nice page. Happy to read this watch Photokano anime

by Anonymous | March 5, 2013 - 3:46pm

I think the research with

I think the research with this is so good. You are able to learn so much from the work here. So much more has to be done with this so much here. Keep up the good work. http://www.realwinners.ca

by Anonymous | February 19, 2013 - 4:49am

The visualization community

The visualization community has noticed your insistence on using rainbow color scales for representing continuous data. This is a plea to you (and anyone else doing the same thing) to stop."

Skau, a PhD computer science visualization student at the University of North Carolina - Charlotte, gives several reasons why rainbow color scales are a bad idea:
Usługi transportowe Wrocław
AGD i RTV
dom
edukacja
internet i komputery

by Anonymous | April 25, 2012 - 4:06pm

Don't understand these rationale

Re: Skau's comment 3. How is it possible that "the spectral order of hues has no inherent or intuitive meaning"? The scale bar/decoder ring on the moon map example above looks just like the natural progression of hues in an actual rainbow. ROYGBIV? Hello??

If you mean that the assignment of red to "high" and blue to "low" end of the range of the parameter depicted is arbitrary, I somewhat agree. Hence the need to depict the scale bar/decoder ring.

Re: Skau's comment 2. If you think the use of a graduated spectrum of values gives "false edges", how is moving to a "diverging color scale" with discrete values of color going to solve that? The example cited using such a scale was a map, where borders are supposed to be sharp, and parameters aren't necessarily graduated or continuous (as is the elevation on the lunar map).

by Anonymous | May 11, 2012 - 3:56pm

ROYGBIV

In response to Anonymous' questions:
ROYGBIV has no inherent meaning other than being the order of the colors the spectrum is in, but semantically, this order means nothing. In addition, if you were never taught ROYGBIV, you would have no idea that the colors went in that order.

The false edges appear in rainbow gradients because changes in luminance and hue are not uniform across the gradient. It is technically possible to make sure these perceptual edges line up with a significant threshold in the data, but the overwhelming majority of the time this is not done. Borders are not supposed to be sharp in all maps. In fact, in many scientific maps, political borders are completely irrelevant information.

I hope this answers a few of your questions,
Drew Skau