Regular readers will be aware of the #endrainbow campaign to reduce the use of rainbow colour palettes in scientific figures. At the recent EGU conference, I gave a talk on ‘making better figures’, which included an example of a published conclusion which was incorrect due to the use of a rainbow colour scheme.
In a paper in 2006, a series of ‘fronts’ were identified in spatial ozone fields (Fig. 1). In the original figure, there is a clear apparent boundary between the yellow and green regions, which was identified as a ‘sub-tropical’ front, and highlighted with a blue line.
How robust is this published finding of a sub-tropical front and associated changes in its behaviour?
Sean Davis (NOAA) reanalysed the same data of total column ozone (TCO) with a similar rainbow colour scale (Fig. 2, left), and a sequential colour scale (Fig. 2, middle). The same sub-tropical ‘front’ is apparent in rainbow, but far less clear in the sequential scheme.
So, which is right?
The right hand panel shows the latitudinal gradient in the same ozone field. Dark colours indicate large gradients, and hence possible ‘fronts’. The yellow ‘mid-latitude’ front is clear in the gradient field, whereas the blue front is far less clear – it is a mirage, an artefact of the choice of colour scale.
Why is the front so ‘visible’ in rainbow?
It is because rainbow scales are not ‘perceptually uniform’ – they create sharp artificial boundaries between colours (particularly involving yellow) that are not necessarily representative of the underlying data (see another example here).
Relying on a change of colour to define a front is clearly not right in this case. Using a perceptually uniform scale, such as the sequential scheme shown here or others, would have avoided this erroneous conclusion.
Summary: use rainbow colour schemes with caution, or avoid them if possible!
Thanks to Sean Davis for letting me use his graphics.