Green anoles (Anolis carolinensis), also described as the American chameleon, can change between brown and green coloration at will in a process known as physiological color change. Deciphering the adaptive purpose of this ability has captured scientists for over a century, with three major hypotheses dominating research: camouflage, social signaling, and thermoregulation. Social signaling is the most well-supported explanation in recent literature, while camouflage has lacked evidence. However, thermoregulation has remained contentious, as older studies show strong support for the hypothesis while newer studies show weak or no support. Seeing this disconnect, my coauthors (Robert Guralnick, Coleman Sheehy III, and Jacob Idec) and I attempted to evaluate these three hypotheses through a novel method to provide fresh insights into what drives color change in Anolis carolinensis.
Diagram of the computer vision pipeline
In our recent paper, we harness over 10,000 images from iNaturalist and recent advances in computer vision technology to evaluate the support for each of these hypotheses at a large scale. To determine the color of the anole in each observation, we utilized Meta’s new SegmentAnything Model (SAM) to generate segments of the anole in the image, filtered out poor segments, and then used a simple equation to determine whether the anole was presenting green or brown. Then, by using the metadata attached to community science posts, we were able to retrieve the exact date-time and estimate the temperature at the moment of image capture. Using these data, we found a strong correlation between the proportion of anoles observed as brown and lower temperatures. Interestingly, during the summer breeding season, this correlation completely disappeared. Additionally, the difference in proportions of green and brown presentation throughout the year was strongly linked to latitude. These observations combined provide evidence for both the thermoregulatory hypothesis and the social signaling hypothesis, which suggests multiple adaptive drivers of color change in this species.
Although big-data observational studies such as this are insufficient to prove the ultimate cause of physiological color change in green anoles, we believe that this paper can serve as a guide for future research that takes time of year and location into account when testing these hypotheses. Furthermore, this research shows that community science has immense potential in big-data studies, especially when working in tandem with artificial intelligence systems such as computer vision. Therefore, we must thank all of the spectacular citizen scientists on iNaturalist to thank for this amazing project, and we hope that more scientists take advantage of the breadth of data available from our communities.
If you would like to read the entirety of this paper, it can be read for free at this link: https://rdcu.be/eMrgE
Leave a Reply