Red coloration and the evolution of aposematism in arboreal sciurids
Cite this dataset
Sheets, Alec D.; Chavez, Andreas S. (2022). Red coloration and the evolution of aposematism in arboreal sciurids [Dataset]. Dryad. https://doi.org/10.5061/dryad.2z34tmppn
Abstract
An animal's coloration is associated with a variety of processes and is therefore subjected to multiple selective pressures. Mammals, especially, are typically inconspicuously colored, or cryptic, to avoid detection by predators. Alternatively, an animal may use conspicuous coloration to advertise the presence of an anti-predator defense. The association between signal and defense is called aposematism. Conspicuous black and white coloration has recently been associated with a range of defenses in mammals, including body size (Howell et al. 2021), however red coloration as a potentially aposematic signal has yet to be investigated in mammals. Squirrels, like most mammals, are unable to perceive red for use as a social signal. Here we use a comparative framework to test whether redness could be a means of background matching, serve a thermoregulatory function or be an honest warning of anti-predator defenses across a global distribution of tree squirrels which vary in size from 16g to 2.2kg in this study. We measured redness of the dorsum, the venter and of red accents of study skin specimens of 57 tree squirrel species (N=257) representing 25 genera. We then associated these phenotypic variables with environmental variables using phylogenetic generalized least squares regression. We find that increasing dorsal redness is associated with more humid environments and closed canopies, consistent with prior work that coloration on this body region under selection for crypsis (Sheets and Chavez 2020). However, we find that ventral redness and maximum redness is associated with large body sizes. Our findings suggest that crypsis and aposematism are not mutually exclusive, and that aposematism may be more widespread in mammals than is currently appreciated.
Methods
Phenotypic Data. We measured redness color from 318 study skin specimens at the Smithsonian National Museum of Natural History (NMNH, Washington, D.C.) representing 73 of the 148 recognized tree squirrel speciesand across 33 generawithin Sciuridae . Up to five individuals were selected per species(µ=4.1, ∂=0.14), if available, which maximized representation of available color morphs and subspecies for each species. We photographed the dorsal and ventral side of each specimen using a digital camera (Sony Alpha a5000) beneath a diffused light from two ESDDI 85 W 5500K Day Light Florescent bulbs. The specimens were collected from a range of years, between 1880 and 2018.
We extracted the mean A* channel pixel value from the digital images to acquire redness color from each study skin specimen. We did this by converting each image to CIE LAB colorspace using scikit-image in Python. Dorsal redness was measured as the mean A* channel pixel value across the entire visible dorsal surface of the study skin specimen. Ventral redness (mean A* channel pixel value) was focused solely on the single, contiguous monochromatic patch of color found in the ventral region of squirrels, and not from other visible areas of the body outside of this ventral patch. Both dorsal and ventral redness were averaged across all individuals within a species.
We also measured the maximum redness of the dorsum and ventrum from each study skin to capture potentially relevant chroma-based signal that is not constrained by averaging redness across either the dorsum or ventral patch. We measured a maximum value of chromaticity because the salience of a visual signal is dependent on both it’s chromaticity and apparent color patch size (would be good to find a citation CITE). . To do this, we set a minimum patch size of 10% of the visible surface (DORSAL OR VENTRAL OR BOTH??) of the study skin that still allowed a large enough patch to be perceived by a receiving animal. We convolved the A* channel values with a box linear filter, which is done by replacing the A* channel value of each pixel with the mean A* channel value of its neighbors plus itself. The size of the filter was set to be the smallest square with an area of at least 10% of the surface of the specimen visible in the photograph. Pixels that did not belong to the specimen and were part of the photo background were masked to a value of zero prior to the convolution. The highest A* channel value from the resulting convolution was recorded for each specimen’s dorsum and ventrum, not just the ventral patch as mentioned for the average A* channel values. We then chosethe maximum redness of each species as the maximum convolved A* channel value of any individual study skin from each species.
Body mass. Body mass data for 58 species was sourced from Thorington et al (2012). For another six species, body mass was sourced from Pantheria (Jones et al 2009). Finally, body mass for the remaining nine species was imputed using Rphylopars (Goolsby et al 2021) from body length measurements (Thorington et al 2012), first molar area (Smith et al 2018) and jaw size (Zelditch et al 2017) assuming a Brownian motion model of evolution and the phylogenetic tree produced by Menendez et al (2020). We then manually assessed imputations for reliability.
Environmental Data. We generated climatic data for the species used in this study using geographic information science approaches. Polygons representing extant ranges for these species were sourced from the IUCN Red List database. Raster files with high resolution data on spatial humidity and mean annual temperature were sourced from the NASA MODIS website (https://neo.sci.gsfc.nasa.gov). Raster files representing canopy closure were sourced from Hansen et al (2013). Climatic variables for each species range were calculated using the ‘clip’ function in the R package raster (Higmans 2020).
Funding