Heatmaps and 2D Fields
A heatmap maps a value at each cell of a grid to color, giving an immediate picture of a two-dimensional field or matrix.
Color as value
A heatmap assigns each grid cell or matrix entry a color drawn from a value-to-color map. It is the most direct way to show a 2D scalar field or a matrix of numbers, letting patterns, hot spots, and gradients appear at a glance. Its honesty depends entirely on the color map.
When it fits
- Dense 2D fields where per-cell precision is less important than overall pattern.
- Correlation or confusion matrices where structure is the point.
- Time-by-channel displays, with one axis time and the other a set of signals.
Color map matters most
Because a heatmap encodes everything in color, a non-uniform palette invents features and hides others. Use a perceptually uniform sequential map for one-sided data and a diverging map with a neutral center for signed data (see Color Maps). Show the scale, always.
Precision and gridding
Color is read imprecisely, so a heatmap conveys pattern better than exact values. When exact reading matters, annotate cells or pair with contours. Ordering rows and columns meaningfully (by cluster or by a variable) can turn apparent noise into visible structure.
Interpolation caution
Smoothly interpolating between cells implies data that may not exist; for sparse grids, show the cells as cells rather than a smooth surface, and mark missing data explicitly.
Kronos use
Simulated 2D field slices and diagnostic time-channel data are shown as heatmaps with perceptual maps and stated scales.