Figure legends
Figure 1. PCA loadings on climatic factors (left) and the relationship between climatic PC1 and butterfly reflectance mapped on the butterfly phylogeny (right). The dimensionality of the six reflectance variables (mean reflectance of dorsal/ventral thorax, basal wings, and entire wings across 320 – 1050 nm range) was reduced using PCA. Reflectance PC1 explained 68% of the variation and correlated strongly with all six reflectance variables (r > 0.71). Colours in the phylogenetic tree indicate climatic PC1, and colours in the heatmap show butterfly reflectance PC1.
Figure 2. Average reflectance of European butterfly species for each body region. The colour of each grid (50 × 50 km) represents the average reflectance (over 320 – 1050 nm) of all butterfly species assembly found in each grid. The colour code of each map was assigned using the Jenks natural break classification method to maximise the variance between each colour class. Red indicates that the butterfly assemblage in the area has higher reflectance while blue indicates lower reflectance (N = 343 species).
Figure 3. The relationship between climatic PC1 and the mean reflectance of each body region of butterflies. The trend lines represent the prediction from the multivariate phylogenetic regression models after accounting for the phylogenetic relationships.
Figure 4. The relationship between climatic PC1 and residuals of the model where near-infrared reflectance (670–1050 nm) was linearly fitted by log form of visible reflectance (320–680 nm). Only ventral thorax and ventral basal wing regions showed significant trends. The trend lines represent the prediction from the PGLS models.
Figure 5. The relationship between climatic PC1 and ventral-dorsal reflectance differences in butterflies. The difference was calculated by subtracting dorsal reflectance from ventral reflectance. The trend lines represent the prediction from the multivariate phylogenetic regression models.