New approaches for capturing and estimating variation in complex animal color patterns from digital photographs: application to the Eastern Box Turtle (Terrapene carolina).

Abstract

Color pattern plays a crucial role in various aspects of an organism's biology, including camouflage, mating, and communication. Despite its significance, methods to quantify and study color pattern variation are often lacking, especially for complex patterns that defy simple categorization. In this study, we developed algorithms to capture and obtain data on 19 different pattern measurements from digital images of 55 individuals of the Eastern box turtle Terrapene carolina sampled in the field and in a museum. The Eastern box turtle is an ideal species to study variation of complex color patterns as this species is easily encountered in the field and in museum collections in Northeastern US, has a relatively easy to identify bright color pattern against a dark background, and has a rigid shell structure, which removes problems related to body distortion. The selected measurements capture the different aspects of the complexity of the color pattern, including the symmetry of the pattern on the turtles' scutes, a critical component in developmental and evolutionary studies. We estimated the variation of each of these 19 measurements across our samples. We determined how much of this variation was influenced by the sensitivity of the pattern capture algorithm due to non-standardized elements of the image acquisition, lighting conditions, and animal shape on pattern variation. To our knowledge, this is the first study to use a comprehensive set of pattern measurements to capture variation in a complex color pattern while also assessing the susceptibility of each of these measurements to noise introduced during data collection. Additionally, we carried out a citizen science approach to characterize the complexity of the color pattern based on human perception and determine which of the 19 pattern measurements best describe this complexity. The most variable measurements across individuals were blue and yellow contrast between the pattern and non-pattern coloration and the average size of objects. From our estimates of the measurement noise due to image acquisition and analysis, we found that the contrast differences reflected true pattern variations between individual turtles, whereas differences in the average size of objects were influenced by both individual turtle variation and measurement inconsistencies. We found that due to the complexity of the patterns, measurements had lower variability if they did not depend on the algorithm defining a set of discrete objects. For example, total area had much less measurement variability than average object area. Our study provides a comprehensive workflow and tools to study variation in complex color patterns in organisms sampled under non-standardized conditions while also estimating the influence of noise due to biological and non-biological factors.