Data from: Gloger's rule or historical conjecture? Tests in mammals
Data files
Aug 28, 2025 version files 873.65 KB
-
GvHC_Data_040725.csv
868.99 KB
-
README.md
4.66 KB
Abstract
Gloger’s rule states that homeotherms are darker at lower latitudes; however, a number of 19th-century naturalists also suggested that animals are more brightly coloured in the tropics than in temperate regions. Using phylogenetic comparative methods, we investigated and compared both ideas across a global sample of 2,726 species of mammals, examining their head, torso, legs, and tail regions. Coloration data were obtained from photographs and compared with a colour chart specifically devised for mammals; ecological data were extracted from pre-existing, open-source databases. All analyses were conducted using phylogenetic comparative generalised linear mixed models in a Bayesian framework. We found strong support for mammals being darker in the tropics and in areas of high precipitation and evapotranspiration, little support for them being darker in warmer areas, little support for them being redder in more arid regions (a more nuanced interpretation of Gloger’s rule), and virtually no support for 19th Century naturalists’ conjecture regarding coloration, contrast, or patterning being more conspicuous in the tropics. These results were replicated at both Class and Order levels. Our findings provide clear evidence for eumelanic coloration to be more prevalent in more humid climates (one facet of Gloger’s rule), operating at a Class level, but indicate that 19th Century observations about bright coloration in the tropics do not pertain to mammals. Our results confirm the importance of Gloger’s rule across mammals as a whole, and add to a growing tide that darker coloration is linked to humidity at a macroecological scale.
https://doi.org/10.5061/dryad.2bvq83c04
The Excel dataset GvHC_Data_040725.csv contains taxonomic information for 4,411 extant, wild, terrestrial, non-volant mammal species, followed by colour trait data and ecological trait data.
Colour traits are split into four distinct regions of the body: head, torso, legs, and tail.
- Primary and all secondary colour scores were generated through comparison of photographs of each species to a mammal pelage melanisation chart
- Numbers of colours is a simple count of the amount of distinct colour scores that were present on each body region of each species
- Colour contrast of a body region was determined through comparison of the first- and second-most prevalent colours upon each body region
- Body region pattern scores were split into discrete categories
Ecological trait data were collated from the existing database PanTHERIA, and consist of:
- Mean annual temperature values measured in degrees Celsius (°C)
- Mean monthly precipitation values measured in millimeters per month (mm/m)
- Mean actual evapotranspiration rate values (mm/m)
- Median latitude and longitude, both in degrees (°) - all of these traits were measured in accordance with each species' known geographical range
- In addition, 'PC1' (a principal component score derived from principal component analysis of precipitation and evapotranspiration, two inter-correlated variables) values for each species can be found in the final dataset column
Columns called Dark Head, Red Head, and Dark or Read Head (and so on for each of the four body regions) were used for mapping only, where data points for different colour values were given different colours and shapes upon the maps.
Description of the data and file structure
To see which pelage colours each of the colour scores describes, see Figure 2 in the related manuscript for a copy of the pelage melanisation chart used to generate the scores.
Colours were considered to be contrasting on each body region if primary and secondary colours differed markedly. To be conservative, if the primary colour of a body region was any score of 1 on the chart, and the secondary colour was any score of 5 (or vice versa; see Fig. 2 of the manuscript), then that body region was considered to be contrastingly-coloured.
Pattern categories are as follows:
Head: 0 = uniformly-coloured, 1 = adjacent blocks of colour, 2 = stripes, 3 = complex patterns (see Fig. 5 of Howell & Caro, 2024 for examples).
Torso: 0 = uniformly-coloured, 1 = flecked, 2 = vertical stripes, 3 = longitudinal stripes, 4 = countershaded (dark dorsum, light ventrum), 5 = reverse countershaded (light dorsum, dark ventrum), 6 = irregular blocks of colours, 7 = disordered stripes resembling contour lines, 8 = spots, 9 = stripes and spots (see Fig. 3 of manuscript for examples).
Legs: 0 = uniformly-coloured, 1 = horizontal bands, 2 = vertical bands, 3 = complex patterns (see Fig. 4 of manuscript for examples).
Tail: 0 = uniformly-coloured, 1 = longitudinal stripes, 2 = adjacent blocks, 3 = alternating bands, 4 = flecked, 5 = spotted, 6 = complex patterns (see Fig. 6 of Howell & Caro, 2024 for examples). The category of ‘complex patterns’ for the head, leg, and tail regions was a catch-all category comprised of any other pattern that could not be scored as one of the other available categories.
Cells were filled with 'NA' where species had no available information recorded in any of the chosen databases for specific variables. In the context of coloration data, colour values were marked as 'NA' where there were no available photographs meeting the selection criteria for colour data to be collected from for a species.
Sharing/Access information
Data were derived from the following sources:
- Ecological data derived from PanTHERIA: https://esapubs.org/archive/ecol/E090/184/metadata.htm
Code/Software
The R code script GvHC_SOMCode_040725.r contains all lines of code necessary to reproduce the analyses that the dataset was used for. Each line of code is annotated with regards to its function as well as any parts of the script that might need changing depending upon which variables are being analysed at any one time.
Statistical software R is needed to run this script (latest version recommended, but v4.2.1 of R onwards should be sufficient). Supporting software RStudio can also be used. Packages used are as follows: phangorn, MCMCglmm, beepr (optional for help with time-keeping), tidyverse, ggplot2, maps, ggdist, and tidyquant.