Winter color polymorphisms identify global hot spots for evolutionary rescue from climate change

See allHide authors and affiliations

Science  02 Mar 2018:
Vol. 359, Issue 6379, pp. 1033-1036
DOI: 10.1126/science.aan8097

Changing coats with the season

Many species of mammals and birds molt from summer brown to winter white coats to facilitate camouflage. Mills et al. mapped global patterns of seasonal coat color change across eight species including hares, weasels, and foxes. They found regions where individuals molt to white, brown, and both white and brown winter coats. Greater proportions of the populations molted to white in higher latitudes. Regions where seasonal coat changes are the most variable (molting to both brown and white) may provide resilience against the warming climate.

Science, this issue p. 1033


Maintenance of biodiversity in a rapidly changing climate will depend on the efficacy of evolutionary rescue, whereby population declines due to abrupt environmental change are reversed by shifts in genetically driven adaptive traits. However, a lack of traits known to be under direct selection by anthropogenic climate change has limited the incorporation of evolutionary processes into global conservation efforts. In 21 vertebrate species, some individuals undergo a seasonal color molt from summer brown to winter white as camouflage against snow, whereas other individuals remain brown. Seasonal snow duration is decreasing globally, and fitness is lower for winter white animals on snowless backgrounds. Based on 2713 georeferenced samples of known winter coat color—from eight species across trophic levels—we identify environmentally driven clinal gradients in winter coat color, including polymorphic zones where winter brown and white morphs co-occur. These polymorphic zones, underrepresented by existing global protected area networks, indicate hot spots for evolutionary rescue in a changing climate.

The importance of evolution in fostering the persistence of species facing rapid environmental change is a fundamental tenet of biology that underlies the modern field of conservation biology (13). Despite the central role of evolution for maintaining biodiversity, criteria to facilitate adaptation by wild species remain largely absent from conservation planning (4, 5). This is a particularly acute omission in a rapidly changing climate (6, 7) where evolutionary rescue may reverse population declines via adaptive evolutionary change in phenotypes (2, 8, 9).

As a first step to demonstrate how evolutionary rescue might enter conservation planning for climate change, we describe a fitness-relevant trait that exhibits clines of locally adapted morphs shaped directly by climate. At least 21 bird and mammal species undergo photoperiod-induced seasonal coat color molts from brown to white in some portions of their range to maintain crypsis against seasonal snow presence or absence (Table 1). This seasonal phenological trait is confronting decreased seasonal snow cover duration, one of the most persistent and widespread signals of climate change (10, 11). Field studies show that winter white animals mismatched against snowless ground suffer high fitness costs due to increased predator-caused mortality, which in the absence of evolutionary shifts would result in substantial population declines (12). In fact, coat color mismatch against decreased snow duration may have already contributed to range contractions for several species (1316).

Table 1 The 21 vertebrate species known to exhibit seasonal coat color molt.

The first eight species are those with sufficient sample sizes of georeferenced winter color phenotype to model range-wide distribution of color morphs. The other 13 species are those known to undergo seasonal coat color change in at least some populations. Species taxonomy follows the IUCN red list.

View this table:

Although the seasonal brown-white-brown color trait is a classic polyphenism—whereby multiple morphs are produced by a single individual (17)—individuals in some populations molt to brown winter coats, thereby not undergoing the circannual color change. This intraspecific variation results in monomorphic winter white and brown populations but also in polymorphic populations that include sympatric winter white and brown color morphs. Importantly, this phenotypic variation is genetically determined: Latitudinal transplants, common garden, and breeding experiments with several seasonal color molting species have consistently showed minimal plasticity in the expression of winter phenotype and instead suggested a simple genetic basis involving one or a few major loci [e.g., (1822)].

The enhanced standing phenotypic variation fostered by genetically based polymorphisms have long been linked to individual fitness and to potential for evolution to rescue populations from abrupt environmental change (23, 24). Specifically, color polymorphisms have served as powerful models demonstrating evolution in nature, including iconic examples of evolutionary response to anthropogenic stressors (2528). For the seasonal coat color trait, selection is expected to act on all winter color morphs based on local snow duration, but evolutionary rescue to changing climate should be enhanced by polymorphic regions where both brown and white winter morphs co-occur.

Here, we use a hierarchical approach across organismal scales (individual, population, and species) to spatially map geographic clines in winter coat color against local climate variables (29). We collated georeferenced descriptions of winter coat color from 2713 specimens spanning 60 countries across species ranges, with data sources including published accounts and specimens at 26 museums globally (table S1). From these georeferenced winter color morph samples, we built predictive models of winter color phenotypes across geographic ranges for eight mammal species that span trophic levels: four hare species and four carnivore species (three weasels and Arctic fox).

The response variable for our global generalized mixed model was the probability of an individual having a winter white coat, with species as a random effect and fixed effects including climate and landscape-level covariates (table S2). As expected for a trait under selection for crypsis against snow or bare ground, the most important covariates emerging from the global model were snow-cover duration and two climate variables affecting snow seasonality and transience. The probability of being white in winter (as opposed to brown) increased positively with snow duration and with seasonality (ranges of mean monthly temperatures), and negatively with isothermality (an index of snow transience).

Using the three environmental covariates identified in the best-fitting model, we created for each of the eight species a predictive range-wide map that assigned to each pixel a probability of an individual being white in winter (Fig. 1 and figs. S1 to S8). Based on fivefold cross validation, models fit georeferenced winter color morph data well (29). Across species, clinal gradients in winter color follow expected environmental gradients based on snow duration and ephemerality: Winter white morphs were more likely in regions with more persistent snowpack that tended to be more northern, higher elevation, and less maritime (Fig. 1). These results suggest that strong natural selection for camouflage against varying snow duration underlies phenotypic variation in winter color morphs across environmental gradients.

Fig. 1 Clinal variation in winter color phenotypes for six mammal species.

Colder colors (e.g., blue) indicate higher probability of winter white morphs (denoted by photo of a winter white snowshoe hare), warmer colors (e.g., orange) indicate higher probability of winter brown morphs (denoted by brown snowshoe hare), and greenish/yellow colors indicate polymorphic populations (see figs. S1 to S8 for larger versions of these maps and for maps of Arctic fox and mountain hare). [Photo credits: L. S. Mills research archives]

To identify hot spots that foster evolutionary rescue, we converted the continuous probabilities of individuals being winter white (versus brown) into polymorphic zones, using both a narrow (40% < Probability [winter white] < 60%) and broad (20% < Probability [winter white] < 80%) criteria. Depending on the species and criteria, polymorphic zones comprised 1 to 57% of a species’ range (table S4). The species with the most widespread polymorphic zones (for narrow/broad criteria) are arctic fox (10%/57%), white-tailed jackrabbit (13%/43%), and long-tailed weasel (9%/33%). Mountain hares have the most restricted polymorphic zone (1%/2%).

Given that the clinal gradient of winter color represents fine-tuned adaptation to local snow conditions, how must winter phenotypes shift to adaptively track projected reductions in snow duration? Based on our model, we plotted the current probabilities of being white in winter against snow duration for four species (table S6) to characterize “optimal” winter coat color as shaped by past selection (Fig. 2). Depending on the species and snow duration, a plausible reduction of 30 to 50 days of seasonal snow cover during this century (30) would require many winter white populations to become polymorphic and polymorphic populations to become winter brown to maintain optimal winter coat colors.

Fig. 2 Change in probability of being winter white as snow duration changes for four molting species.

Species are Japanese hare, dark blue; white-tailed jackrabbit, light blue; least weasel, yellow; and long-tailed weasel, red. The central colored area with both winter white and brown animals represents our broadly defined polymorphic zone (i.e., 20% < P[winter white] < 80%).

Next, we combined the polymorphic zones of the eight species to identify regions with multispecies polymorphic zones (Fig. 3). Although under the broad criteria, two or more species shared putative polymorphic zones across much of the Northern Hemisphere (Fig. 3, A and B), narrow criteria multispecies polymorphic zones were limited to a few regions in North America (Fig. 3C) and Great Britain (Fig. 3D).

Fig. 3 Regions with polymorphisms in winter coat color for multiple species.

(A to D) Where polymorphic zones overlap for two (red) or three (brown) species, derived from predictive maps for eight species (see Fig. 1 and figs. S1 to S8). Polymorphic zones defined broadly as 20% < P[winter white] < 80% in (A) North America and (B) Eurasia. (C) Polymorphic zones defined more narrowly as 40% < P[winter white] < 60%; found only in (A) North America and (D) Great Britain. (E) Example of camouflage mismatch (least weasel). In polymorphic zones, as snow duration decreases, mismatched winter white morphs would be selected against in favor of the sympatric winter brown morphs. [Photo credit: Karol Zub]

Polymorphic zones within and across these eight species ranges identify regions that currently hold disproportionately high potential to initiate evolutionary rescue from camouflage mismatch in this fitness-relevant trait affected by climate change. In addition to being hot spots for in situ evolutionary rescue, these areas may also facilitate gene flow of adaptive alleles to monomorphic populations (31, 32).

Although protected areas cover 13% of the world’s terrestrial area (33), multispecies polymorphic zones are poorly represented by existing protected areas (table S5). Even under our broad criteria, only 5% of multispecies polymorphic zones occur in the most strict protected areas described by the International Union for Conservation of Nature (IUCN) [categories I and II (34)]; all six IUCN categories of protected areas combined embrace only 10% of multispecies polymorphic zones (for the narrowly defined polymorphic zones; 4% fall in “strict” and 7% in “all”) (table S5).

The broad geographic ranges of color molting species, and their roles as flagships and strongly interacting predators and prey, amplify the value of understanding how climate-mediated evolution may foster their persistence in the face of climate change. Failed adaptation by these species could have indirect impacts that reverberate through their ecosystems. Further, because the codistributed species that make up the multispecies polymorphic zones represent both predators (e.g., weasels and Arctic fox) and prey (e.g., hares), differential molt responses in different species could exacerbate fitness costs and create cascading coevolutionary outcomes.

Mismatch in seasonal coat color provides a visual metaphor for how climate change may affect biodiversity, and regions of sympatric winter color polymorphisms identify multispecies hot spots for evolutionary rescue in the face of reduced snow duration. Our framework to identify zones of enhanced potential to initiate evolutionary rescue from climate change could be applied to polymorphisms in other morphological, physiological, or behavioral traits affected by climate change. Identification of hot spots for evolutionary rescue provides novel opportunities to integrate evolutionary processes to conservation planning in a changing climate.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

Tables S1 to S6

References (3546)

References and Notes

  1. Materials and methods are available as supplementary materials.).
  2. D. K. Hall, G. A. Riggs, MODIS/Terra Snow Cover Daily L3 Global 500m Grid, Version 6 (Sept. 1st, 2000 - Aug. 31st, 2015). Boulder, Colorado USA. NASA National Snow and Ice Data Center Distributed Active Archive Center. Date Accessed: 24 May 24; doi:
Acknowledgments: We thank S. T. Giery, D. Emlen, and four anonymous reviewers for insightful comments on the manuscript. We also thank museum curators (table S1) and substantial sample contributions from K. Zub, K. Garrison, T. Berry, J. Dines, E. M. Fouts, K. Nicholson, D. Oehlschlager, C. Opitz, A. M. Riedel, B. Wommack, S. Lado, M. Jones, J. Ivan, J. Whittington, and A. Angerbjörn. Funding: L.S.M.: National Science Foundation Division of Environmental Biology Grant 0841884; A.V.K.: National Science Foundation Graduate Research Fellowship under grant no. DGE-1252376; P.C.A.: FLAD, Luso-American Development Foundation; M.Z.: The Department of the Interior Southeast Climate Science Center Global Change Fellowship (Cooperative Agreement No. G10AC00624); J.M.-F.: Fundação para a Ciência e a Tecnologia Grants “CHANGE” PTDC/BIA-EVF/1624/2014 (Portuguese national funds) and IF/00033/2014/CP1256/CT0005 Investigador Fundação para a Ciência e a Tecnologia (Programa Operacional Potencial Humano-Quadro de Referência Estratégica Nacional funds from European Social Fund and Ministério da Ciência, Tecnologia e Ensino Superior), and SYNTHESYS grant SE-TAF-4695 (European Union Seventh Framework Programme; agreement 226506) to access the Swedish Museum of Natural History; J.M.G.: National Science Foundation EPSCoR grant 1736249; N.L.: Russian Foundation for Basic Research (17-04-00269 A). Author contributions: L.S.M. conceived the idea, helped with analysis, and led the writing. E.B., along with A.V.K., led the data analysis, contributed conceptually, and helped with writing. M.Z. helped with data analysis, figures, writing, and contributed conceptually. D.J.R.L., J.F., B.M.D., K.H., P.C.A., J.M.G., and J.M.-F. contributed conceptually and with writing. A.D., A.V.A., N.L., and K.F. helped provide key input data. All authors read and approved the submitted manuscript. Competing interests: None declared. Data and materials availability: The data reported in the paper are available in the supplementary materials and archived in the Dryad Digital Repository at Global SnowPack data are available as Web Map Service at DLR (German Aerospace Center) Geoservice,

Stay Connected to Science


Navigate This Article