Professional Documents
Culture Documents
Data graphics shape the way science is data being represented. One common data appear to have sharp edges and broad
communicated, and the color schemes colormap that is perceived in false rela- plateaus when plotted with jet. A com-
we employ can either faithfully repre- tion to the data it represents is the mon argument in favor of jet is that its
sent or tacitly obscure the data a fig- rainbow-colored jet colormap, whose sharp gradients allow proximal colors to
ure is intended to convey (Tufte, 1983). popularity stems largely from its decades- be distinguished by a meticulous viewer,
Tasteful use of color can make data long residence as the default setting in and indeed, Figure 1c shows that in some
graphics visually appealing and can draw MATLAB and similar programs (Eddins, subranges within the color scale, jet offers
viewers in, engaging the audience and 2014). The jet colormap visually distorts superior point-to-point discrimination.
encouraging further inspection of a fig- information through two primary mech- However, the high-performing regions of
ure. But wherever color is used to rep- anisms. First, jet places emphasis at arbi- jet are offset by regions elsewhere in the
resent numerical values, its role tran- trary locations along the color scale cor- spectrum where changes of the same size
sitions from a mere aesthetic nicety to responding to local maxima and minima are nearly imperceptible.
carrying the responsibility of conveying in colormap lightness (Rappaport, 2002;
data honestly and accurately. Yet, biases Stauffer et al., 2015). This effect leads to HOW TO SELECT AN HONEST,
introduced by some common colormaps higher rates of error in identifying regions EFFECTIVE COLORMAP
have gone widely unrecognized within of maximum and minimum data val- To avoid the pitfalls of jet and other per-
the oceanographic community. Here, we ues, even among experts who are familiar ceptually nonuniform colormaps, we
describe the pitfalls of some commonly with the data type under evaluation and offer the following guidelines for select-
used colormaps, provide guidelines on who have experience with jet colormap ing data-appropriate colormaps that
effective, accurate colormap selection, (Spence et al., 1999; Borkin et al., 2011; show numerical values effectively, intui-
and present a suite of perceptually uni- Bryant et al., 2014). tively, and in proper proportion.
form cmocean colormaps that have been The second mechanism by which jet
designed for oceanographic data display. deceives viewers is through false gradients Reflect the Nature of the Data
The cmocean package is available across introduced by a non-monotonic light- The first step in selecting an appropri-
multiple software programs, including ness profile, which accelerates at a differ- ate colormap relates to the nature of
MATLAB, Python, R, Generic Mapping ent rate than the data it represents (Light the data being displayed. Numerical
Tools, and Ocean Data Viewer. and Bartlein, 2004; Borland and Taylor II, data represented by pseudocolor can
2007). For oceanographers, the danger of typically be categorized as sequential,
RAINBOW DECEPTION jet’s false gradient profile is its ability to diverging, or cyclic.
When color is employed to show numer- covertly exaggerate fronts in some regions
ical data, viewers expect the overall dis- of the color scale while minimizing the 1. Sequential. There is a regular interval
tribution of color they perceive to match presence of fronts elsewhere (Ware, 1988; relationship in sequential data, such as a
the distribution of the underlying data. Mersey, 1990). Figure 1e shows jet’s mis- range of salinity values. Sequential data
However, perception of color is often behavior, where a cone with linear slop- should be represented with a monotoni-
overlooked in oceanography as we con- ing sides viewed from the top appears cally increasing range of lightness values
tinue to rely on colormaps that have com- smoothly varying when plotted with a (Stevens and Marks, 1965; Rheingans,
plex transfer functions between what is simple, unbiased, perceptually uniform, 2000), such as in a grayscale colormap,
visually perceived and the underlying grayscale colormap, yet the same smooth but can also have variations in hue. The
FIGURE 3. Examples of colormap use. The free surface in the northwestern Gulf of Mexico is shown (subplots a and b, where solid line/red are
positive values, dash-dot/blue are negative values, and gray solid lines are bathymetric contours) with a ship track indicated. Data collected
from an undulating towed vehicle along the ship track are shown are shown in the following rows. Depth vs. distance plots along the track are
temperature (c–d), salinity (e–f), and oxygen (g–h). Each property is shown at left in jet and at right in a colormap from cmocean. Data are recent
work of authors DiMarco and Zimmerle for an atlas of oceanographic observations of the mechanisms controlling hypoxia
2
http://matplotlib.org/cmocean
3
http://www.mathworks.com/matlabcentral/fileexchange/57773
4
http://dankelley.github.io/oce
5
https://github.com/kthyng/cmocean-odv
6
http://soliton.vm.bytemark.co.uk/pub/cpt-city/cmocean/index.html