1 Introduction

The phenologies of wild populations represent their timing of life history events and are assumed to have evolved so as to synchronize these occurrences with temporal abundances of resources (Both et al. 2009). These resource distributions are rapidly shifting worldwide, however, due to ongoing climate change (IPCC 2007) and phenological adjustments will thus be necessary in adapting populations to these new conditions. Not surprisingly, phenological adjustments are the most often cited ecological response to climate change (Parmesan 2006) and those species which have either not adjusted (Møller et al. 2008), or have exhibited an insufficient response to remain synchronized with their primary resources (Both et al. 2006) are more likely to be in decline.

Hibernation is widely hypothesized to be a seasonal adaptation to long-term energy and/or water shortages (Wang 1989; Humphries et al. 2003a, b; but see Lovegrove 2000; Liow et al. 2009). Yet, despite the important role that hibernation plays in the life cycle of many mammalian species, compared to phenological traits in other taxa (e.g., Przybylo et al. 2000; Charmantier et al. 2008) it has received substantially less attention in this context. We consequently have comparatively little understanding of how evolution has shaped natural hibernation expression and are limited in our ability to predict how populations of wild hibernators will be affected by climate change.

In this chapter, I will outline the sources and consequences of individual variation in hibernation phenology. Phenotypic plasticity represents one avenue by which individuals can adjust their phenology in response to variable climatic conditions (sensu Réale et al. 2003). In response to long-term directional change, however, it is thought that, because phenotypic plasticity has a theoretical limit, microevolution of phenological traits will be necessary (Visser 2008). Evolution by natural selection requires a combination of three fundamental ingredients: phenotypic variation, selection acting on this variation, and a heritable genetic basis underlying it (Endler 1986). While these features are often assumed in investigations of hibernation phenology, I will use concepts and theory from quantitative genetics and animal breeding to show how we can accurately quantify them. Although this data is currently limited, the framework presented in this chapter should allow for its collection from a diversity of species.

2 Intraspecific Variation and Phenotypic Plasticity

Whereas the focus in hibernation research, to date, has been on species-level patterns, if we are to understand how selection has shaped current hibernation phenologies and how these phenologies may respond to climate change, a tightened focus on individual variation is necessary. Individual variation in hibernation phenological traits is commonplace (e.g., Table 5.1) and its most striking examples are obtained by studying individuals across a range of environments. Black-tailed prairie dogs (Cynomis ludovicianus) are facultative hibernators resident throughout the great plains of North America (Lehmer et al. 2001). Although populations from a more benign environment (Colorado, USA) hibernate rarely (Lehmer et al. 2006), all focal individuals from a more northern population (in Saskatchewan, Canada) were shown to hibernate (Gummer 2005). Similarly, echidnas (Tachyglossus aculeatus) range throughout Australia and New Guinea and hibernation in the comparatively more benign climate of Kangaroo Island is more shallow, later and of shorter duration (Rismiller and McKelvey 1996) than in other locations (e.g., Tasmania; Nicol and Andersen 2002). There is also good reason to expect that the ecological consequences of this variation may be profound. For example, Landry-Cuerrier et al. (2008) showed that the energy expenditure across individual eastern chipmunks (Tamias striatus) expressing varying degrees of torpor varied by over threefold.

Table 5.1 Individual variation in five phenological traits from a wild mammal population (Coumbian ground squirrels, Urocitellus columbianus)
Full size table

Davis (1976) hypothesized that physiological processes underlying hibernation are integrated through annual changes in the environment. Specifically, metabolism exhibits a free-running circannual cycle of approximately 11 months (Pengelley and Fisher 1963) and torpor occurs when the low phase of the metabolism cycle coincides with environmental energy shortages. This hypothesis was originally formulated to explain species-level variation among the Marmotini. However, comparison to other mammalian hibernators may shed light on the traits where we should expect to see individual variation as well as the levels of variation, and its environmental correlates. Whereas a pronounced circannual rhythm appears to limit hibernation to winter seasons in the Marmotini, other species are more flexible in their torpor patterns. Eastern chipmunks (Munro et al. 2008), edible dormice (Glis glis; Bieber and Ruf 2009) and pygmy possums (Cercartetus nanus; Geiser 2007) are all highly sensitive to environmental conditions and may hibernate for up to a year during resource shortages. I suggest, therefore, that Davis’ hypothesis can be expanded. Relatively large hibernators (e.g., larger members of the Marmotini) are able to store more energy as fat prior to hibernation. In addition, the seasonality of more predictable environments was likely a useful ‘guide’ to the availability of resources. Over evolutionary time, circannual rhythms that matched this seasonality were thus likely to have been favored by selection. Larger hibernators and those occupying more predictable environments (Lovegrove 2000) should be more likely to have hibernation patterns heavily influenced by an innate circannual rhythm. Smaller species and populations from more stochastic environments, by comparison, should be more plastic.

In the wild, phenotypic plasticity has most often been assessed as a change in the mean value of a trait within a population (e.g., Przybylo et al. 2000) or as interpopulation differences in mean values of traits (e.g., Sheriff et al. 2011). However, phenotypic plasticity is inherently an individual-based phenomenon. Specifically, it is defined as the differential phenotypic expression of a genotype across an environmental gradient (Pigliucci 2001). For traits that are expressed multiple times throughout the lifetime, it is thus most appropriately evaluated using the reaction norm approach in which variation within an individual (or genotype) is compared against variation in the environmental trait of interest (Nussey et al. 2007). With repeated measurements of phenological traits from hibernators, levels of phenotypic plasticity can thus be estimated by using a specific type of mixed-effects model, the random-regression model:

$$ y_{ij} = \mu + {\ss}E_{j} + p_{i} + p_{Ei} E_{j} + e_{ij} $$

In this model, the phenological trait, y (e.g., emergence date) is measured on individual i, at time j, and related to the environmental variable E. The fixed effects describe the population average response of y to variation in E: μ is the population mean in the average environment (i.e., the intercept if the environmental variable is mean-centered) and ß is the population response to the environmental variation (i.e., the slope of y on E). The random effects then model the individual-specific deviations from the population response: p i models consistent individual differences in the phenological trait (i.e., the variance in the intercepts of individuals; Fig. 5.1a) and p Ei models individual differences in phenotypic plasticity (i.e., the variance in the slopes of individuals; Fig. 5.1b). For a more detailed description of the application of random-regression models to studies of life history variation in natural populations see Nussey et al. (2007).

Fig. 5.1
figure 1

Hypothesized relationships of the sources (a, b), and consequences (c, d) of individual variation in hibernation phenology. a represents the influence of an environmental variable (ambient temperature in this case) on variation in a hibernation phenological trait (emergence date). In this scenario, all individuals emerge earlier in response to warmer temperatures (they are phenotypically plastic) and individual intercepts differ, reflecting individual variation in mean emergence date which may, or may not, have a heritable genetic component. In b the slopes of the individuals differ, reflecting variation in their plastic response to the environmental predictor. c depicts negative selection on emergence date as individuals with increasingly delayed emergences experience diminished fitness. d depicts stabilizing selection as individuals with the mean emergence date have the highest fitness. Positive and disruptive selection would be represented by the reverse of c and d, i.e., respectively, a linear increase and a concave function

Full size image

A number of examples of phenotypic plasticity have been reported from wild populations (e.g., Kobbe et al. 2011) and perhaps the most convincing evidence comes from experimental studies that have manipulated resource abundance to directly examine its influence on hibernation expression (e.g., Humphries et al. 2003b). However, to my knowledge, no studies to date have quantified reaction norms for hibernation traits. Although such estimates undoubtedly require large sample sizes (Martin et al. 2011), appropriate dataset are increasingly available (e.g., Ozgul et al. 2010; Lane et al. 2011) and should provide for the ability to explicitly test a number of key hypotheses. For example, by statistically comparing reaction norm slopes across species/populations, one can test the hypothesis that those resident in more stochastic environments should display more plasticity in hibernation traits. Similarly, through appropriate phylogenetic comparisons, the relationship between phenotypic plasticity and body mass could be evaluated.

3 Selection

Gender-related variation in hibernation phenology has been described for a number of species (e.g., Young 1990; Kawimichi 1996). In many north temperate sciurids, for example, the emergence from hibernation of reproductive males precedes that of reproductive females (e.g., Murie and Harris 1982). This pattern (termed ‘protandry’; Morbey and Ydenberg 2001) is hypothesized to have arisen due to differing selection pressures across the sexes (Michener 1983). Natural selection is presumed to favor female emergences that synchronize reproduction with the period of vegetative growth, while sexual selection favors earlier emergence of males so as to gain reproductive opportunity. However, an appropriate assessment of the adaptiveness of sex-specific emergence phenology, or indeed of much of hibernation phenology, has yet to be conducted because the strengths and directions of selection have never been calculated for any phenological traits.

Lande and Arnold (1983), using recognizable principles from multiple regression, developed the approach for measuring selection in wild populations that is most widely used today. Estimating selection in this way offers a number of advantages. First, standardized coefficients of selection are provided that can be compared across traits and taxa. Second, by analyzing multiple traits within a multiple-regression framework, direct estimates of selection are decomposed from the action of indirect selection on correlated traits. Third, by estimating the curvature in the trait-fitness relationship (by fitting quadratic functions), the action of directional, stabilizing, and disruptive selection can be evaluated (Fig. 5.1c, d). Finally, when coupled with estimates of the heritability of traits (see below), standardized selection estimates provide for an assessment of not only the direction, but also pace of microevolution (Falconer and MacKay 1996).

Numerous studies have since applied Lande and Arnold’s (1983) analytical framework to investigate selection in wild populations (reviewed in Endler 1986; Kingsolver et al. 2001). However, the vast majority of these have been obtained for morphological traits (of the more than 1,500 estimates reviewed by Kingsolver et al. (2001), over 80% were for morphological traits). By comparison, only 13–17% were for life history/phenological traits and 61% of these were obtained from plants (no estimates for hibernation phenological traits were represented). To what extent the results from other traits and taxa may hold for hibernating mammals is thus currently unknown and there is a pressing need for selection estimates for hibernation traits. However, there are practical considerations to consider. For example, the median sample size represented in Kingsolver et al. (2001) was 134 and the authors’ estimated that the probability with which one could reject a null hypothesis of no selection at the 95% level using this sample size was <50%. Nonetheless, I would suggest that the benefit gained from providing estimates of this sort should in many cases outweigh concerns over statistical significance.

4 Genetic Variation

In order for selection to lead to the evolution of a trait, shifts in the phenotypic distribution in one generation need to be passed reliably to the subsequent generations (Lynch and Walsh 1998). In other words, the trait must be heritable. Heritability refers to the proportion of phenotypic variance that has a genetic basis and is estimated by comparing measurements of phenotypic traits among individuals of known genetic relatedness (Lynch and Walsh 1998). Arguably the most intuitive way to do this is with a parent–offspring regression, in which the values of the phenotypic trait in the offspring are regressed against these values in one, or both, of their parents. The slope of the regression [or twice the slope if a mid-parent value (i.e., the mean of the maternal and paternal trait) is used] then represents the heritability of the trait. This technique has been used extensively but it suffers from a number of limitations. Specifically, by only taking advantage of a single genetic relationship (parent–offspring), it does not make efficient use of all available data (Kruuk 2004). In addition, for many species maternity is easier to identify than paternity and mother–offspring regressions are thus more commonly used than father–offspring regressions. However, maternal effects (i.e., non-genetic influences of a dam on her offspring) can lead to phenotypic similarity among kin and, unless explicitly accounted for, can confound estimates of heritability (Kruuk and Hadfield 2007). For mammals, this is especially problematic due to their prolonged periods of maternal care. For these reasons, a specific type of restricted maximum likelihood model, the ‘animal model’, is now used more often (Kruuk 2004), i.e.,

$$ {\mathbf{y}} = \, {\user2{X}}{\mathbf{b}} + \, {\user2{Z}}_{1} {\mathbf{a}} + \, {\user2{Z}}_{2} {\mathbf{m}} + \, {\user2{Z}}_{3} {\mathbf{pe}} + \, e$$

This model uses pedigree-derived measures of genetic relatedness and estimates of phenotypic similarity among all members of a population to estimate the variance components underpinning observed phenotypic variation. Specifically, y represents a vector of observed phenotypic values and b represents a vector of fixed effects. Importantly, because this is a type of mixed-effects model, variables such as age and sex can be controlled for by fitting them as fixed effects. a, m, and pe then represent, respectively, the vectors of additive genetic, maternal, and permanent environment effects and X and Z 1–3 are the corresponding design matrices (Lynch and Walsh 1998; Kruuk 2004). The permanent environment effect arises due to individual differences other than those due to additive genetic or maternal effects (Kruuk and Hadfield 2007). Incorporating these additional random effects (m and pe), in the mixed-effects model framework, thus allows for a non-confounded estimate of the additive genetic variance. The heritability can then be estimated as the variance component associated with a (V a) divided by the total phenotypic variance (V p).

Despite its advantages, to my knowledge, our analysis of hibernation phenology in Columbian ground squirrels (Urocitellus columbianus) represents the only application of an animal model to a wild hibernator (Lane et al. 2011). In this population from Alberta, Canada the heritability of female emergence date was h 2 = 0.22 ± 0.05 and the genetic correlation between male and female emergence dates was r G = 0.76 ± 0.22. In addition, the genetic correlation between female emergence date and oestrous date as r G = 0.98 ± 0.01. These results suggest that emergence date can evolve in response to selection and selection-induced changes will have corresponding effects across the sexes and traits.

5 Discussion and Future Directions

In this chapter, I have taken a perspective from evolutionary ecology to address the sources and consequences of individual variation in hibernation phenology. It is my view that, although these sorts of data are currently limited, appropriate datasets have been built (or are being built) and the analytical methodology with which to analyze them are readily accessible. By applying these frameworks, we should be able to gain a greater depth of insight into how evolution has shaped the hibernation phenotypes we currently observe, and how these phenotypes may respond to environmental change. Here, I outline three potential questions, incorporating both the theoretical and applied relevance of hibernation phenology, that could be addressed in future study.

How does selectionact on variable torporexpression during hibernation? Torpid metabolic rates during hibernation average 5% of euthermic metabolic rate and can be <1% of the metabolic rate expressed during cold exposure (Geiser and Ruf 1995). These pronounced metabolic savings have led many authors to suggest that torpor expression should be maximized during hibernation (e.g., Harlow and Frank 2001). Others, however, have pointed to the ubiquitous arousals to euthermy during hibernation, as well as the ecological benefits of elevated energy expenditure, in general, to suggest that torpor entails non-trivial costs and an optimum level should thus balance energetic benefits against these costs (French 1988; Humphries et al. 2003a). Support for this hypothesis has been provided by the observation that animals fed supplemental food express less torpor (e.g., Humphries et al. 2003b) and the apparently polyphyletic loss of hibernation in southern populations of woodchucks (Marmota monax) and California ground squirrels (Otospermophilus beecheyi) (Davis 1976). The individual variation expressed in wild hibernator populations provides ideal raw material with which to test these hypotheses from an evolutionary standpoint. Specifically, by quantifying the strength and shape of selection on phenological traits (e.g., immergence and emergence dates, number and depth of torpor bouts) and coupling these measurements to estimates of the heritability of these traits, information can be gained as to the fitness consequences of individual variation and predictions can be made as to how they may evolve.

How will the phenologiesof wild populations of hibernatorsrespond to environmental change and will this change be sufficient to prevent population declines? In addition to providing for a more in depth appreciation of the evolutionary ecology of hibernation, an understanding of the phenotypic plasticity and microevolutionary potential of phenological traits is increasingly becoming of applied relevance. Their relevance to population dynamics has recently been shown in yellow-bellied marmots (Marmota flaviventris). Climate change has resulted in earlier emergence of the marmots, which has effectively lengthened the active season for the population and led to higher over-winter survival (Ozgul et al. 2010). However, this is the only investigation of its kind to date and predicting how populations of hibernators will respond to environmental change, more generally, clearly requires additional data. This is especially important as the environmental factors influencing phenological variation in hibernators (e.g., duration of snow cover) may be different for those species upon which most research has been conducted (i.e., ambient temperature for insectivorous birds), thus limiting our ability to generalize across taxa.

Are hypotheses to explain protandrousemergence from hibernationvalid and, if so, how will shifting selectiondue to climate changealter these patterns? Enhanced degrees of protandry have recently been observed in some migratory bird populations and investigators have suggested that this is due to climate change (Møller 2004). Specifically, warming climates lead to more benign conditions during migration and, hence, lowered survival costs to early arrival. This relaxation of natural selection pressures then allows sexual selection to promote earlier male arrival. Thus, climate change could potentially influence not only the viability of populations, but also the social relationships within them. As many hibernators exhibit naturally protandrous emergence, climate change could conceivably further enhance this gender difference. A strong between-sex genetic correlation in Columbian ground squirrels suggests that such independent evolution of the sexes will be limited. However, estimates of natural and sexual selection on emergence dates of the two sexes have not been made in this, or any other species, thus rendering most predictions premature at this point.

In describing the focus of hibernation research over 30 years ago Davis (1976) stated: “Research on hibernating species often concerns primarily the physiological responses to low temperatures. While this topic is highly meritorious for its clinical and biochemical applications, it may perhaps neglect some aspects of adaptations to natural conditions that may provide clues to physiological processes”. Although the intervening decades have yielded great advances in both hibernation physiology and evolutionary ecology, it is fair to argue that the current focus resembles that in 1976. I suggest, however, that a tightened focus on the genetic and environmental sources, as well as the fitness consequences, of variation in hibernation phenology will provide for a more thorough understanding of the ecological and evolutionary significance of natural hibernation expression.