There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.
Abstract
Coloration of animals is important for camouflage, for social behavior, or for physiological
fitness. This study investigates the color variation in adults of
Aglais urticae obtained on subjecting some pre‐imaginal stages to different temperature conditions
and their thermobiological consequences. To investigate the evolutionary–ecological
interactions of temperature and pigmentation in butterflies, caterpillars, and pupae
of the small tortoiseshell,
Aglais urticae (Lepidoptera, Nymphalidae), larvae from Central Europe and Scandinavia were reared
at temperatures between 7 and 34°C in the laboratory or in the field. After emergence,
the intensity of pigmentation of the imagines and their increase in body temperature
under defined full‐spectrum light irradiation were quantified by image analysis and
thermal imaging. At constant conditions, ambient rearing temperature and pigmentation
intensity of imagines were negatively and linearly correlated in Central European
butterflies, regardless of whether the pupal stage alone or, additionally, the last
period of the larval stage was exposed to these conditions: low temperatures induced
darker coloration and high temperatures led to lighter individuals. A thermal pulse
of a few days alone at the beginning of pupal dormancy led to a similar, albeit weakened,
effect. Caterpillars of the Scandinavian subspecies
A.
urticae polaris, whose pupal dormancy took place under Central European field conditions, developed
into strongly pigmented imagines. The thermobiological relevance of more intense pigmentation
was shown by significantly higher absorption of light, and thus stronger increased
body temperature after 5 min of defined illumination, but this difference ceased after
15 min. Our results show that phenotypic plasticity in wing coloration is adaptive
since temperature‐induced developmental changes provide thermobiological benefit in
adult butterflies. We propose that, in subpolar latitudes, darker coloration likely
has a selection advantage favoring individuals with reaction norms gradually shifted
to stronger pigmented phenotypes, possibly leading to the establishment of a pigmentation
cline.
Abstract
Coloration of animals is important for camouflage, for social behavior or for physiological
fitness.
Aglais urticae butterflies reared at temperatures between 7 and 34°C show at constant conditions
a negative correlation between ambient temperature and pigmentation intensity. Darker
imagines show a stronger increased body temperature after 5 min of defined illumination
which provides an ecological advantage.
During the last decades, considerable interest has arisen in predicting the distribution of species, assemblages of species and characteristics of assemblages under various scenarios of climate change1 2 3 4. Most of these predictions, however, are based on phenomenological models5 and provide little mechanistic understanding of the underlying processes6. A first step towards understanding the impact of climate change is to consider traits of species that underpin the relationship between species and climate7. This approach has been very successful for plants. For instance, leaf size as well as leaf shape have been used to reconstruct temperature and precipitation of paleoclimates8. However, the relationships between traits of animals and climate are more difficult to identify. One trait—the body size of endotherms—may be related to climate. Bergmann’s rule predicts a decrease in their body size within species and across closely related species with increasing temperature9 10. In line with this ecogeographic rule, a recent study found that the body size of some mammals decreases with increasing temperature11. Although many physiological and local-scale studies on insects have demonstrated associations between traits and climate, to what extent such mechanistic links may impact distribution patterns of insect species at a biogeographical scale remains largely unknown (but see ref. 12). In ectotherms, the appearance of the body surface and in particular its colour value are involved in thermoregulation13. Dark-coloured ectotherms are able to increase their body temperatures above ambient air temperature more effectively than light-coloured ectotherms, and therefore have an advantage in cool climates (thermal melanism hypothesis14 15). Most insects—by far the most species-rich lineage of ectotherms16—need to reach body temperatures above ambient temperature for flying, foraging or mating17. However, being dark is only advantageous in cool climates. In areas with high temperatures and insolation, insects need to protect themselves against overheating18. At high temperatures, ectothermic species with light colouration can be active for a longer period than species with dark colouration, and may be able to use a broader thermal range of habitats. Overall, heat avoidance and heat gain constraints lead to the prediction that insect assemblages should be dominated by dark-coloured species in cool climates and by light-coloured species in warm climates. The first hint that this holds even at a biogeographical scale was provided by Rapoport19 almost 50 years ago, when he used a rough estimate of springtail colour value to show a positive correlation of the percentage of dark-coloured species in species assemblages with latitude and altitude. In this study, we combine recent digital image analysis and phylogenetic statistics to demonstrate that colour lightness of insects is consistently correlated to the thermal environment across Europe. We furthermore show that assemblages of dragonflies became on average lighter-coloured during the last century, which we attribute to global warming. Results Mechanistic adaptation of species to climate Using a data set of nearly all European butterfly and dragonfly species, we show that insect species with dark and light colouration have advantages in cool and warm climates, and that this mechanistic adaptation of species to climate shapes the biogeographical patterns of species distributions. The results were obtained by measuring the colour value of the body and the dorsal and ventral basal wing areas of 366 butterfly species occurring in Europe using computer-assisted digital image analysis and the colour value (further on called colour lightness) of the body of 107 dragonfly species of Europe. The dorsal surface of butterflies was on average darker than the ventral surface, and the colour lightness strongly differed between families in both groups (Supplementary Figs 1 and 2; and Supplementary Tables 1 and 2). All statistical approaches consistently showed that the mean colour lightness of assemblages increased with a synthetic variable that characterized the thermal environment within grids, that is, colour lightness increased with increasing temperature (Fig. 1, Table 1). These results support a mechanistic link between climate and functional traits of species, and indicate that an ecophysiological phenomenon can cause noticeable biogeographical patterns of insects (Fig. 2). Phylogenetic and specific components of colour lightness The colour lightness of each species has two components: a phylogenetic component and a species-specific component. The phylogenetic component is in part determined by the response of the ancestors to paleoclimates, whereas the species-specific response measures the deviation of each species from the ancestors. The latter component is therefore the more recent response of species to environmental factors. When we considered these two components, we found a clear decrease in the mean colour lightness of insect assemblages for both the mean phylogenetic components and the species-specific components across Europe from the Mediterranean to Northern areas. This result was consistent for the ventral and dorsal surfaces of butterflies and for the dorsal surface of dragonflies (Figs 1 and 21 2, and Table 1). Impact of climate change on insects If colour lightness of assemblages across Europe is a reaction to climate, we would thus expect that climate change would lead to changes in the colour lightness of insect assemblages. To test this hypothesis, we compared recent distributional data of dragonflies with data before 1988 and plotted the shift of colour lightness within grids (Fig. 3). After correcting for phylogeny, we found a general shift towards lighter-coloured assemblages across Europe (Fig. 3c). In the phylogenetic component (Fig. 3b), we found shifts towards darker assemblages along the western margin of Europe, the Alps and the Balkans. However, the variation in the phylogenetic component still contains a portion of ‘phylogenetically structured environmental variation’20. In addition, the change in colour lightness was positively correlated to the change in annual mean temperature (AMT) between the time periods corresponding to the distribution data (Fig. 4). However, we would like to stress that the strength of the shift depends on the models used. Alterations in the distributions of species result in a complex spatial pattern in the colour lightness of assemblages, which is a clear indication that the response of assemblages of organisms to climate change does not lead to simple one-dimensional shifts (see for example, the Alps and the Scandinavian Mountains). Discussion The interpretation of our results is conditional upon the assumption that thermoregulation is the predominant selective pressure for the evolution of wing or body colouration. However, colour lightness and colouration in general may be influenced by numerous other selective forces, for example, cryptic colouration21 or disease and parasite resistance22. Our statistical models of butterflies showed that the positive relationship between the thermal environment and colour lightness is stronger for the ventral surface than for the dorsal surface (Table 1), which suggests that selective pressures other than thermoregulation are more important for the colouration of the dorsal surface than of the ventral surface. A stronger relationship between colour lightness and ventral wing surfaces is not entirely surprising, since several species exclusively use the ventral surfaces for thermoregulation (for example, Colias and Gonepteryx). Furthermore, ventral wing surfaces are additionally used for heat avoidance by reflecting insolation when wings are closed23. Despite these other potential pressures, the strong and consistent relationships between colour lightness and thermal environment across two insect groups underline the overall importance of climate for insects. After the last glaciation, butterflies and dragonflies recolonized wide areas of Europe, but most of the butterfly and dragonfly clades evolved before the Pleistocene16. The latitudinal gradient of the phylogenetic component suggests that exaptations24 have been of considerable importance during the assembly of northern insect faunas after the last glaciation, that is, colour lightness might have evolved for or as a by-product of other functions in a different spatio-temporal selective regime. But the trends in the species-specific component indicate that the comparatively dark-coloured species of all clades colonized northern areas. Our finding that the assembly of insect faunas in response to the thermal environment depends on the colour lightness of certain body parts clearly demonstrates the importance of thermal energy in structuring insect assemblages—even across larger spatial scales. This has implications in forecasting the effect of climate change1 2 3 4 5 6 25 26 27. With global warming, we would thus expect that dark-coloured insects will shift their distribution and possibly retreat from certain areas28 and/or on a smaller scale will shift their habitat preference to more shady conditions29. The latter scenario has consequences for conservation strategies : conservation efforts directed exclusively toward current habitat preferences without taking into consideration the effect of ecophysiological adaptation may be futile for the future as the climate changes. Methods Distribution data We used distribution maps of 434 European butterfly species30. After matching these with the species illustrated in ref. 31, we were able to consider 366 species across 1,825 grid cells of 50 km × 50 km. We created a database of 107 dragonfly species across 1,845 grid cells32 (Supplementary Tables 1–3, and Supplementary Methods). The butterfly distribution maps were presence/absence maps; the dragonfly distribution maps were outline maps. Computer-assisted digital image analysis Computer programs for the analysis of digital images provide possibilities to measure the colour value from scanned published illustrations33 34. We scanned ventral and dorsal butterfly wings and dragonfly bodies and converted the scanned RGB (red, green, blue) colour illustrations into 8-bit grey values. We assigned each pixel a value between 0 (complete black) and 255 (complete white), and averaged these values across all pixels of the considered area to obtain a single colour value for each species. Usually, software such as PhotoShop weights the three RGB channels according to the perception of human eyes when converting to grey values (for example, 0.21 × R+0.72 × G+0.07 × B). This is because some colours are subjectively perceived darker than others. For example, blue looks darker than green for humans at the same value of colour lightness. We decided to calculate the unweighted mean across the three RGB channels. Note that dark-coloured species have a low value and light-coloured species have a high value. In particular, we scanned the ventral and dorsal part of the wings of female butterflies illustrated in ref. 31 with an EPSON Perfection 4490 Photo Scanner with 1,200 dpi and 24 bit in the RGB colour spectrum. All further steps to estimate the colour value were done with Adobe Photoshop CS2, Adobe Photoshop Elements 2, Epson Scan Ver. 2.75G and ImageMagick 6.5.5-2. In our study, we predominantly considered females because their parental investment is higher and colouration of males may be biased by sexual selection35. Therefore, females can be expected to be closer to a ‘thermal optimum’. Only if an illustration of a female was not available, we used an illustration of the male (butterflies: 28 male and 338 female). Note that dorsal and ventral colour lightness was closely and significantly correlated between the sexes (t-test, P 0.80; Supplementary Fig. 6). This indicates that related species have more similar colour lightness than expected by chance, and hence the phylogenetic variance–covariance matrix needs to be scaled by lambda39. When compared with the phylogenetic signal in body size—which we assume to be a key niche trait—there was even no significant difference to the ventral colour lightness of butterflies. However, potential micro-evolutionary causes of this pattern, for example, gene flow, pleiotropy or lack of genetic variability cannot be distinguished within the framework of this study. Evolutionary model uncertainty The resulting correlations of the trait values of species in the phylogeny can be seen as a function of an evolutionary model-specific transformation of the branch lengths of the original phylogeny. Following ref. 44, a fundamental assumption of phylogenetic analyses is that the model of character evolution effectively recapitulates their history, and this implies to compare several evolutionary models. Hence, to support our main conclusions that (i) colour lightness is consistently correlated to the thermal environment, and that (ii) dragonfly assemblages became on average lighter-coloured during the last century, we simultaneously accounted for evolutionary model and phylogenetic uncertainty. However, we would like to highlight that it is not our intention to make statements about the most adequate model for the evolution of colour lightness in butterflies and dragonflies, so we arbitrarily chose four evolutionary models to underline the robustness of our conclusions: (i) the model ‘lambda’51 52 fits the extent to which the phylogeny predicts covariance among trait values for species by multiplying the off-diagonal elements in the correlation structure by the value of lambda. Interpreted as a tree transformation, values of lambda near 0 cause the phylogeny to become more star-like, and a lambda value of 1 recovers the Brownian motion model. Bounds for the estimation of lambda were set to 0 and 1. This model was used for the results presented in the main text. (ii) In the model ‘kappa’51, character divergence is related to the number of speciation events between two species. Interpreted as a tree transformation, the model raises all branch lengths to the power kappa. Bounds for the estimation of kappa were set to 0 and 1. (iii) The Early-burst57 or also called accelerating-decelerating (ACDC)58 model is a model where the rate of evolution increases or decreases exponentially through time. Bounds for the estimation of the scaling parameter a were set to −10 and 10. (iv) The Ornstein–Uhlenbeck model59 fits a random walk with a central tendency and an attraction strength proportional to the parameter alpha. Bounds for the estimation of alpha were set to 0 and 105. Model parameters were estimated for 1,000 randomly selected phylogenetic trees in each model with the function fitContinuous in the R package geiger 56 version 1.99–3. All trees were transformed with their estimated parameters (function transform.phylo) and used to partition colour lightness into a phylogenetic and a specific component with Lynch’s comparative method38. For each tree, we calculated for each grid cell in Europe the mean values of the phylogenetic and specific components of the subset of the occurring species and regressed these values against our thermal component 1. Regarding our main conclusion (i) that colour lightness is consistently correlated to the thermal environment, we found that all 24,000 regressions were highly significant (t-test, P 1 from the categories temperature and insolation, which characterize the thermal environment of the respective grid cell (Thermal 1 and 2), as well as two components for precipitation (Prec 1 and Prec 2) (Supplementary Table 5). The PCA extracted one eigenvector (Topo) for the topography (Supplementary Table 5). For temperature and precipitation, the first axis (called Thermal 1 and Prec 1) characterized the overall mean, and the second axis (Thermal 2 and Prec 2) characterized the variability of temperature or precipitation across the year. Since the directions of ecological associations between response and predictors in models with PCs are difficult to interpret because the sign of axis scores is arbitrary, we plotted selected variables of each category for illustration (Supplementary Fig. 9). We calculated multiple regression models using the phylogenetic and specific components of the mean colour value within grids and the principal components of the environmental variables. Assumptions of an ordinary least-squares regression are that the standard error of the error term is constant over all values of the response, and that explanatory variables and all estimates provide equally precise information. Because the standard error of the calculated mean colour value of assemblages within grids differed in relation to the number of species recorded within grids, using the number of species per grid cell as weights to the least-squares regression improves parameter estimation (see also ref. 65). To further improve the information quality of the data, we included only grid cells in the analysis of butterflies with more than five recorded species. To account for spatial autocorrelation in the model residuals after fitting the environmental components to the data, we build spatial simultaneous autoregressive models, where the error term is predefined from a spatial neighbourhood matrix, and autocorrelation in the dependent variable is modelled within a generalized least-squares framework using maximum likelihood66. Calculations were performed with the function errorsarlm in the R package spdep version 0.5–65 for neighbours within 1,000 km distance. We applied hierarchical partitioning to assess whether multicollinearity between the PCs of the different categories might affect the results. Hierarchical partitioning decomposes the variation contained in a response variable into independent parts, which reveal the absolute importance of each predictor by considering all possible variable combinations in a hierarchical multivariate regression setting67. Calculations were performed in R with the package hier.part version 1.0–3 using ‘r 2’ as the goodness-of-fit measure. We also included phylogenetic and specific parts of body size (digitally measured fore-wing length of butterflies and body length of dragonflies) in the analysis. We found qualitatively similar results compared with the findings of Table 1, indicating that multicollinearity and body size does not change our main results (Supplementary Fig. 10). All statistical analyses were performed with the software R ( www.r-project.org). Intensive calculations were conducted using the MaRC2 High Performance Computing cluster in Marburg. Change in dragonfly colour lightness To analyse changes in the colour value of dragonfly assemblages, we compared the data used in our statistical analysis (distribution of 2006, ref. 32) to maps of the distributions before 1988 (ref. 37), and plotted the distribution of the shift in colour lightness within grids and the average shift across all grids for multiple phylogenetic trees (Fig. 3). In addition, we extracted climatic data corresponding to the two distribution data sets of dragonflies in Europe from public resources (CRU TS3.2, ref. 68). We calculated the AMT for the period 1900–1988 and for 1988–2006; determined the change in AMT between these two periods; and correlated this change to the change in colour lightness of dragonfly assemblages (Fig. 4). We found positive correlations (t-test, P<0.001) for the raw data, the phylogenetic component and the specific component of colour lightness. Correlations using simultaneous autoregressive models to incorporate spatial autocorrelation were also all significant (t-test, P<0.05) and positive. This indicates that dragonfly assemblages became lighter coloured in regions where temperature increased during the last century, giving further support to our conclusion that climate warming favours light-coloured insects in Europe. Author contributions D.Z. compiled and analysed the data, prepared the figures, and assembled the dragonfly phylogeny; S.B. contributed the principal ideas and assembled the butterfly phylogeny; R.B. designed the study; and all authors discussed the results and contributed to the writing of the manuscript. Additional information How to cite this article: Zeuss, D. et al. Global warming favours light-coloured insects in Europe. Nat. Commun. 5:3874 doi: 10.1038/ncomms4874 (2014). Supplementary Material Supplementary Information Supplementary Figures 1-10, Supplementary Tables 1-5, Supplementary Methods and Supplementary References
The wings of Lepidoptera contain a matrix of living cells whose function requires appropriate temperatures. However, given their small thermal capacity, wings can overheat rapidly in the sun. Here we analyze butterfly wings across a wide range of simulated environmental conditions, and find that regions containing living cells are maintained at cooler temperatures. Diverse scale nanostructures and non-uniform cuticle thicknesses create a heterogeneous distribution of radiative cooling that selectively reduces the temperature of structures such as wing veins and androconial organs. These tissues are supplied by circulatory, neural and tracheal systems throughout the adult lifetime, indicating that the insect wing is a dynamic, living structure. Behavioral assays show that butterflies use wings to sense visible and infrared radiation, responding with specialized behaviors to prevent overheating of their wings. Our work highlights the physiological importance of wing temperature and how it is exquisitely regulated by structural and behavioral adaptations.
This is an open access article under the terms of the
http://creativecommons.org/licenses/by/4.0/ License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
scite shows how a scientific paper has been cited by providing the context of the citation, a classification describing whether it supports, mentions, or contrasts the cited claim, and a label indicating in which section the citation was made.