Deep learning image segmentation reveals patterns of UV reflectance evolution in passerine birds

Abstract

Ultraviolet colouration is thought to be an important form of signalling in many bird species, yet broad insights regarding the prevalence of ultraviolet plumage colouration and the factors promoting its evolution are currently lacking. In this paper, we develop a image segmentation pipeline based on deep learning that considerably outperforms classical (i.e. non deep learning) segmentation methods, and use this to extract accurate information on whole-body plumage colouration from photographs of >24,000 museum specimens covering >4500 species of passerine birds. Our results demonstrate that ultraviolet reflectance, particularly as a component of other colours, is widespread across the passerine radiation but is strongly phylogenetically conserved. We also find clear evidence in support of the role of light environment in promoting the evolution of ultraviolet plumage colouration, and a weak trend towards higher ultraviolet plumage reflectance among bird species with ultraviolet rather than violet-sensitive visual systems. Overall, our study provides important broad-scale insight into an enigmatic component of avian colouration, as well as demonstrating that deep learning has considerable promise for allowing new data to be brought to bear on long-standing questions in ecology and evolution.

Evolutionary predictors of the specific colors of birds

“…Such unaccounted variation may preclude the detection of certain effects, but should not lead to spurious results. However, we note that a recent study analyzing the evolution of UV reflectance in passerine birds using models of avian color vision reported similarly low marginal r 2 values (3 to 7%) and high conditional r 2 values (>85%) ( 63 ). This suggests that low explanatory power is unlikely the result of methodological limitations, and fits with the observation that low r 2 values are the norm in evolutionary studies ( 81 ).…”

“…This allows to assess whether specific selection factors have favored the evolution of, for example, red rather than blue colors. Splitting continuous color variation into useful categories has previously been done at smaller scales or focusing on specific color types ( 63 , 44 , 74 , 87 , 88 ).…”

“…One such candidate would be variation in retinal cone sensitivity functions, which in birds is largely restricted to the ultraviolet ( 14 ) and thus not applicable to our data. Although a recent analysis in passerines revealed an association between highly ultraviolet (UV)-sensitive vision and UV reflectance of the plumage, the effect was not statistically significant ( 63 ). Another candidate, applicable to the entire gamut of colors beyond UV, would be the proportion of the different photoreceptors present in the retina (cone proportions), which varies substantially between species ( 72 ) and affects how colors are perceived ( 73 ).…”

“…4 ) where the light environment (which is relatively rich in shorter wavelengths) renders them conspicuous visual signals ( 31 – 33 , 62 ). Indeed, a recent study showed that plumage colors rich in even shorter wavelengths (near-UV) are most common in the forest canopy ( 63 ). More broadly, our results show that the prediction based on comparative analyses on African Starlings ( 37 ), that darker forested environments harbor less chromatic colors because light levels are too low, does not apply across all birds.…”

Challenges and opportunities in applying AI to evolutionary morphology

“…Furthermore, the preparation and analysis of data workflows can be greatly improved with the use of AI and some of the most significant progress in this area has been conducted on museum bird specimens. DL methods have been applied to segment and extract plumage from images which greatly enhances the speed at which images can be processed and colour information extracted (He et al, 2022;Case Study 6). This approach has been taken to even higher levels of specificity by applying pose estimation methods to identify specific points of bird anatomy regions to improve the efficiency of the pipeline for extracting colour information from specimens (He et al, 2023).…”