From Noisy Data to Useful Color Palettes: One Step in Making Biodiversity Data FAIR

Abstract

Due to the differences in individual’s color perception and the variations in color naming and color rendering under different settings, color has historically been a challenging trait in describing species for taxonomic and systematic research. Reusing a noisy color dataset collected from high-quality images of Carex specimens, we developed a data mining method (e.g., clustering and classification) for constructing domain-specific color palettes. Color palettes associated with color values measured in a color space help systematists record color data in a way that the differences in colors can be more accurately compared and computed, making color data interoperable and reusable. The Carex color palette was evaluated by Carex experts and the evaluation data showed that experts overwhelmingly preferred using color palette over color strings.