Why should engineers and scientists be worried about color? Because presenting complex data with the default rainbow colormap is misleading:
In the view on the left, you see large areas of yellow, with a dark blue region, rimmed with cyan and green moving in from the left, and some dark red regions particularly in the upper right. Based on this distribution of color, you may make some assumptions about the underlying structure in the data.
Now, consider the [second image], which you will recognize as a map of the southeastern United States, with the Florida peninsula clearly illustrated. In this representation, you can easily see the coastline and the surrounding continental shelf in the Gulf of Mexico and Atlantic Ocean. The boundary of the continental shelf and areas of deep ocean are shown in purple, which darkens with depth. The Appalachian Mountains are clearly distinguished in lighter colors from the piedmont and coastal plain regions, primarily in green.Which picture do you think best represents the underlying topography and bathymetry data? Would it surprise you to learn that the two pictures not only show the same data, but are represented by colormaps which are mathematically equivalent? In both cases, each value of the continuous variable, elevation in meters above and below sea level, has been mapped onto a unique value on a continuous pseudo-color scale.
In the [first] picture, elevation at each point has been mapped onto the most commonly-used colormap in visualization, the so-called “rainbow” colormap. In this hue-based colormap, show to the right of the visualization, the lowest value is mapped to blue, the highest value is mapped to red, and the intervening values are mapped continuously onto values interpolated between these two colors in red-green-blue space.
In the [second] picture, elevation has also been mapped onto a pseudo-color map. However, in this case, the colormap has been designed to take into account characteristics of the data and the human visual system. In particular, the colormap has been designed so that equal steps in the data variable will be perceived as equal steps in the representation. Since the data has a threshold value or boundary of interest to the user of the data (i.e., sea level), this characteristic of these interval data is also explicitly incorporated into the colormap.