OrionAI chooses colours and classes that fit your data

Changes since 2.11.1.

#New

• Smooth gradients. Ask for a smooth or continuous map and OrionAI colours each area along one gradient between two end colours, with no classes. The legend shows it as a gradient bar. When a few values are far above the rest (a capital city in a population map), the gradient is spread by rank so the other areas do not all look the same.
• More colour scales. 18 ColorBrewer scales join the list: two-colour gradients such as Yellow-Orange-Red, Yellow-Green-Blue and Purple-Red, diverging scales such as Brown-Blue-Green for drier or wetter than usual, and palettes for categories (Set 2, Dark 2, Paired, Set 3). They are in the Style panel too.

#Improved

• Colours chosen by meaning. OrionAI picks the scale by what the values are: warm scales for population, density and heat, blues for rainfall and water, greens for vegetation, two-sided scales for change, and distinct colours for types. It no longer reuses one scale for every map.
• Natural breaks by default. Classed maps now use natural breaks, which follow the gaps in the data, so skewed values no longer crowd into one or two classes. OrionAI can also use quantile and equal interval, checks how well the classes fit the data, and says on the result card which method it used and why.
• Categories look like categories. Land cover, types and other unordered categories get distinct colours instead of a light-to-dark scale that made them look ranked. The Style panel does the same for the new category palettes.
• No silent substitutes. If OrionAI asks for a colour scale that does not exist, it is told the closest real ones and picks again, instead of quietly getting Viridis.

#Fixed

• The map matches its colour key. Coloured areas are now drawn almost opaque, and other filled boundary layers under them become outlines. Before, a see-through fill could mix with a boundary layer underneath, so a map keyed in oranges could show as blue and mauve.