Behind the Headlines

MATLAB and Simulink behind today’s news and trends

This is machine translation

Translated by Microsoft
Mouseover text to see original. Click the button below to return to the English version of the page.

A dangerous rainbow: Why colormaps matter. 2

Posted by Lisa Harvey,

Rainbows typically signify a turn for the better. The storm has ended. The skies are clearing. But when the rainbow scale is used for weather and climate maps, the results can be misleading, and possibly downright dangerous.

A recent article in The Wall Street Journal, Weather Forecasts Should Get Over the Rainbow, detailed how the rainbow color scale, while adding visual interest, often leads to misunderstanding the data. Data scientists caution against using the rainbow colormap.

“Our vision is really good at drawing boundaries and clustering similar colors,” Rob Simmon, senior data visualization engineer at Planet, told The Wall Street Journal. “When it comes to the rainbow, this means that the transition from yellow to green jumps out at us, while shifts within, for example, the green part of the spectrum are much less apparent. This can create artificial boundaries in data that is actually smooth—a seemingly technical issue with serious consequences.”

Serious Consequences?

To explain the potential for serious consequences, The Wall Street Journal provided the recent example of the maps used during Hurricane Florence:

“For instance, the map consulted by Mr. Trump and his advisers during Hurricane Florence included a sharp boundary between yellow and green bands that appeared to represent a huge drop in the probability of tropical-force storm winds. This border stretched through Delaware, West Virginia, Virginia, North Carolina and South Carolina, affecting millions of people trying to determine the level of risk the storm posed. In fact, the difference was just 10%. Such visual confusion also can mislead policy makers tasked with issuing evacuation warnings.”

 

Image Credit: Just In Weather.

Perception is everything

For many years, scientists have considered the rainbow color map, also known as Jet, a poor choice for data visualizations. In the 2007 IEEE article, Rainbow Color Map (Still) Considered Harmful, authors David Borland and Russell M. Taylor II from the University of North Carolina at Chapel Hill stated, “The rainbow color map confuses viewers through its lack of perceptual ordering, obscures data through its uncontrolled luminance variation, and actively misleads interpretation through the introduction of non-data-dependent gradients.”

People perceive contrast in brightness and can easily order options in grayscale.

The authors then described how the rainbow colormap, based on the ordering of the wavelength of light, is not a perceptual ordering. Switching to the rainbow scale, people often have difficulty ordering the colors.

 

Default Colormap

Changing to a perceptually ordered colormap helps data understanding.

“Given these limitations, it is no surprise that data scientists have been calling for caution in the use of the rainbow scale for decades. Awareness of the issue seems to be increasing more recently,” Says The Wall Street Journal.

“MATLAB, a popular computer language used for image processing, long featured the rainbow as the default color scale. It recently replaced it with a new palette, called parula, that ranges from purple through yellow, with a rich range of blues and greens in the middle and no reds.”

Data scientists agree. Climate scientist Ed Hawkins launched his blog The Climate Lab Book as a tool to foster discussion among climate scientists. He explains why data visualization is critical:

“The blog has had helpful discussions about many aspects of climate change science, but also visualisation, such as our #endrainbow campaign to reduce the use of rainbow colour schemes, such as jet. Obviously, MATLAB has made a step forward by switching the default colour scheme to parula.”

2 CommentsOldest to Newest

Chad Greene replied on : 1 of 2
Nice post, Lisa! I always say that when color is used as a numerical axis, it should be just as linear and uniform as the tick marks on a normal x,y plot. Anything else is like arbitrarily stretching and squashing the number line. For more options of perceptually uniform colormaps in Matlab, try cmocean (https://www.mathworks.com/matlabcentral/fileexchange/57773) and crameri (https://www.mathworks.com/matlabcentral/fileexchange/68546).

Add A Comment

Your email address will not be published. Required fields are marked *

Preview: hide