Using large-scale community science data and computer vision to evaluate thermoregulation as an adaptive driver of physiological color change in Anolis carolinensis

Facultative, physiological color change has many potential adaptive functions, and the ability of the green anole (Anolis carolinensis) to shift between brown and green coloration is no exception. Three non-mutually exclusive hypotheses for such color changes include: 1) The camouflage hypothesis, which states that individual anoles use brown and green coloration to blend into their background; 2) The social signaling hypothesis, which states that coloration shifts convey intraspecific signals such as dominance, submission, and mating status during interactions; 3) The thermoregulation hypothesis, which states that shifting to darker brown coloration during colder temperatures allows for increased absorption of solar radiation as heat. We showcase the utility of a computer vision pipeline to derive individual-level color (green versus brown) from a large dataset of citizen science observations spanning the southeastern USA. We used this color information along with climate, seasonal timing information and background in images to test associations between color morph, temperature and time of year. Results show that brown-presenting A. carolinensis were observed more frequently at lower temperatures during winter. However, the observed correlation between presenting color and temperature was absent during the summer breeding season. We did not find strong evidence for background color matching. We found support for both the thermoregulatory hypothesis and social signaling hypothesis dependent on time of year, which suggests multiple independent drivers are influencing physiological color changes in A. carolinensis. Further, this work shows the power of leveraging large-scale digital field images and machine learning to derive insights about how species can regulate phenotype to maintain their thermal and biotic niche optima.