Halle Braun¹, Kushin Mukherjee¹, Seth R. Gorelik², Karen B. Schloss¹; ¹University of Wisconsin-Madison, ²Woodwell Climate Research Center

Colors have affective connotations, which influence people’s evaluations of data visualizations (Bartram et al., 2017; Anderson & Robinson, 2021). For example, observers prefer visualizations in which affective connotations of the dataset (positive/playful vs. negative/serious) and colors representing the data are aligned. Previous studies of individual colors and palettes of discrete colors (e.g., palettes in categorical data visualizations) suggested that lighter, saturated colors were associated with positive concepts, whereas darker, desaturated colors were associated with negative concepts (Bartram et al., 2018; Schloss et al., 2020). For visualizations of categorical data, it is possible to make visualizations overall dark or light to convey affective connotations while maintaining discriminability of data categories. However, for visualizations of continuous data where spatial structure is key, this approach could pose a problem because lightness variation is important for revealing spatial patterns in datasets (Ware, 1988; Rogowitz & Treinish, 2009). We investigated if color scales can be created to have strong affective connotations while preserving lightness variation for visualizations of continuous data. Starting with grayscale maps that evoked similar affective connotations (established in a pilot), we applied 16 color scales to each map: 4 hues (red/yellow/green/blue) x 2 lightness levels (light/dark) x 2 saturation levels (saturated/unsaturated). The maps were obtained from a global raster dataset of aboveground biomass stock (Harris et al., 2021) and were selected to have a roughly uniform distribution of pixel values (normalized to scale from 0 to 1). The color scales were generated using Color Crafter (Smart et al., 2020). Participants rated associations between each colormap and eight affective concepts. Positive concepts (positive/exciting/playful) were associated with lighter, saturated, and bluer colormaps, whereas negative terms (negative/disturbing/serious) were associated with darker, unsaturated, and yellower colormaps. The results indicate colormaps can have strong affective connotations while preserving lightness variation important for observing spatial patterns in data.

Acknowledgements: NSF grant BCS-1945303 to KBS