Sexes in sync: phenotypic plasticity, sexual selection and phenological synchrony between the sexes in a wild hibernator

Abstract

Desynchrony of phenological responses to climate change is a major concern for ecological communities. Potential uncoupling between one of the most fundamental divisions within populations, males and females, has not been well studied. To address this gap, we examined sex-specific plasticity in hibernation phenology in two populations of Columbian ground squirrels (Urocitellus columbianus). We find that both sexes display similar phenological plasticity to spring snowmelt dates in their timing of torpor termination and behavioural emergence from hibernation. As a result of this plasticity, the degree of protandry (i.e. males' emergences from hibernation preceding those of females) did not change significantly over the 27-year study. Earlier male behavioural emergence, relative to females, improved the likelihood of securing a breeding territory and increased annual reproductive success. Sexual selection favouring earlier male emergence from hibernation may maintain protandry in this population, but did not contribute to further advances in male phenology. Together, our results provide evidence that the sexes should remain synchronized, at least in response to the weather variation investigated here, and further support the role of sexual selection in the evolution of protandry in sexually reproducing organisms.

1. Introduction

To survive and reproduce in seasonal environments, organisms must organize schedules of energy expenditure in accordance with prevailing environmental conditions. Proper timing of the transitions between life-history stages (i.e. phenologies) is thus likely to affect fitness. Currently, climate change is altering the temporal energetic profiles of environments due to its effects on abiotic (e.g. weather) and biotic conditions (e.g. seasonal availabilities of food sources), and especially the interaction between them [1–5]. Phenological shifts are thus, unsurprisingly, the most commonly reported ecological responses to climate change [6].

Components of an ecosystem may respond to climate change differently, leading to asynchronous shifts among interactants. Asynchrony among trophic levels, for example, may lead to disruption of ecological communities [2,7–10]. Consequences could also be expected if asynchrony arises between two of the most common interacting components of communities, males and females of the same species. Multiple studies have reported the environmental causes and fitness consequences of shifted female reproductive phenologies [11,12], but logistical challenges have hindered similar assessments in males. As a result, examinations into potential desynchronizations between the sexes have been few (but see [13–15]).

In many animals, males enter the breeding population before females, and multiple hypotheses have been advanced to explain this phenomenon, termed ‘protandry’ [16]. Most support among vertebrate species is for hypotheses that rely on the reproductive benefits afforded to males arriving earlier than females [17]. Predicting how the sexes will respond to climate change, and whether they maintain synchronization, thus requires an appreciation both of sex-specific phenological responses to weather variation, and their fitness consequences. Many wildlife populations show phenological advances over time (e.g. migratory birds [13,14]), but some others show delays (e.g. hibernating ground squirrels [18,19]). Therefore, it is important to consider how climate variation could impact sex-specific phenology under both scenarios (figure 1). According to one hypothesis, that we term ‘selection balance’ (figure 1a), there is a balance between natural and sexual selection favouring, respectively, relatively later versus earlier arrival. If climate change ameliorates the natural selection costs of early arrival (e.g. by making resources available earlier in the season), we predict that the balance will tip towards earlier arrival by males, as they are typically subject to stronger sexual selection than females [20–22]. We may then expect the degree of protandry to increase under warmer springs if males plastically advance their phenology to a greater degree than females (figure 1.1a; as seen in migrating barn swallows, Hirundo rustica [13], and willow warblers, Phylloscopus trochilus [14]). However, climate warming is projected to increase atmospheric moisture and, over most of North America, increase winter precipitation and heavy precipitation events over time [19,23]. This could result in delays in phenology in years with, for example, delayed snowmelts. Under this scenario (figure 1.2a), if males are more plastic than females to later snowmelt, we would expect a reduction in protandry over time.

Figure 1.

Three hypotheses for sex-specific plastic responses to weather variation and their predictions under two scenarios of advanced and delayed phenology in response to an environmental variable. The example in (1) shows an advance in female arrival in the breeding population in response to warmer spring temperatures and (2) shows a delay in female emergence from hibernation in response to later snowmelt dates. According to (A) the selection balance hypothesis, male phenology is more plastic than females (steeper male slope) where males advance their arrival into the breeding population more than females in (1a) warmer springs or (2a) springs with earlier snowmelt. According to (B) the common cue hypothesis, males and females have similar plastic responses to the environment (similar male and female slopes) and protandry is unaffected among years and conditions in both scenarios. According to (C) the sex-specific constraint hypothesis, plasticity in males is reduced by a constraint (shallow male slope) where males do not show the same advances and delays in phenology in response to an environmental variable as females.

Not all costs of early arrival in the breeding population, however, will be ameliorated by warmer weather or earlier snowmelts. For example, hibernation may afford safety from predation while the hibernator is sequestered within the hibernaculum [24,25]. Therefore, males may benefit reproductively by having an advanced arrival relative to females, but further advances will be limited due to costs, such as an earlier exposure to predators as they emerge from the relative safety of the hibernaculum [25]. When this is the case, we would expect males to temporally adjust their phenologies, but to remain synchronized with females. In species in which males are not directly exposed to females during hibernation, this would be achieved through both sexes responding to similar environmental cues. Under this hypothesis (which we term ‘common cue’; figure 1b), we predict that male plasticity to weather variation is similar to that of females (whether advancing or delaying phenology over time; figure 1.1b and 1.2b), and the degree of protandry will remain unaffected by climate change.

Finally, there may be a constraint on phenology to which only one of the sexes is subject. For example, female arctic ground squirrels (Urocitellus paryii) plastically delayed termination of torpor (i.e. the final return to warm body temperatures during a hibernation bout) during a year of exceptionally delayed snowmelt, but reproductive males did not exhibit plastic delays [18]. Female ground squirrels typically emerge above ground on the day of torpor termination, and may re-enter torpor if conditions are not suitable for emergence. By contrast, reproductive males terminate torpor and are active in their hibernacula for one to three weeks prior to emergence above ground as they undergo physiological processes to prepare for the breeding season (i.e. elevations in circulating testosterone, testicular recrudescence and spermatogenesis), thus preventing re-entry into torpor [26–28]. Under this hypothesis (sex-specific constraint), male plasticity is limited, which would ultimately lead to an increase in the degree of protandry in years with later snowmelts if females plastically delay their phenology (figure 1.2c) or a decrease in protandry in years with warmer springs if females advance their phenology (figure 1.1c).

The goal of our study was to test these three hypotheses to explain how emergence phenology and protandry are affected by annual weather variation in two wild populations of hibernating Columbian ground squirrels (Urocitellus columbianus). Emergence from hibernation in this species is protandrous with males emerging above ground about a week prior to females [29,30]. Females typically emerge from the hibernaculum on the day that they terminate torpor [31] and become sexually receptive (i.e. enter oestrus) 3 to 5 days following emergence [29,32,33]. Males typically emerge from torpor much earlier than females and spend 7–10 days sequestered in the hibernaculum following termination of torpor before they emerge above ground and enter the breeding population ahead of females. We have previously shown that the likelihood of late season snow storms has increased in our study areas over time, which has resulted in increasingly later snowmelt dates over the study period [19]. We have also previously shown that females are phenotypically plastic to date of snowmelt by delaying their emergence from hibernation in response to delayed annual snowmelts over time, and that annual mean female fitness declines in years of delayed emergences [19]. Females are also phenotypically plastic to spring temperature and delay emergence from hibernation in colder springs, but no significant changes in spring temperature were reported in our study area over time [19]. In contrast to females, the phenological plasticity of males to annual weather variation is unknown. Here we examine the termination of torpor and behavioural emergence from the hibernaculum for males in relation to females. The fitness consequences of variation in male phenologies are central to all of the three hypotheses under consideration [16], but are currently unknown for any male hibernator or indeed, most mammals [34]. Here we evaluate these consequences with respect to male emergence phenology.

Each of the three described hypotheses provide the basis for an evaluation of the causes of variation in protandry among years. Female emergence from hibernation shows delays (as opposed to advances) in our study populations in response to delayed spring snowmelts over time [19], and so we base our predictions on the scenario presented in figure 1.2. We evaluate here sex-specific shifts in hibernation phenology in response to snowmelt dates where years of early spring snowmelt are assumed to advance forage conditions in the spring [18,19,35]. According to the selection-balance hypothesis (figure 1.2a), we predicted that males would plastically advance their hibernation phenology more than females in years of early spring snowmelt, resulting in an increase in protandry in these years. According to the common cue hypothesis (figure 1.2b), we predicted that males and females would exhibit similar plasticity in their hibernation phenology and, thus, the extent of protandry will be unaffected by weather variation among years. Finally, according to the sex-specific constraint hypothesis (figure 1.2c), we predicted that an inability to reenter torpor limits male phenological plasticity compared to females, leading to increased protandry in years with later snowmelt.

2. Methods

(a) . Study populations and hibernation phenology

We studied the hibernation phenology of two spatially proximate (within 4 km) populations of ground squirrels in Sheep River Provincial Park, Alberta, Canada. We captured ground squirrels in both populations using live traps (13 × 13 × 40 cm, Tomahawk Live Trap Co., Hazelhurst, WI, USA) and identified each individual with alphanumeric designations stamped onto paired metal ear tags (Monel #1; National Band & Tag Co., Newport, KY, USA) [35,36]. Individuals were either caught as young of the year at the time of weaning, thus their age was known with certainty (n = 54 of 120 at Kite Field and n = 258 of 308 at Meadow B), or their age could be estimated to within a year using criteria based on body mass and timing of behavioural emergence from hibernation in the population each spring.

(b) . Torpor termination and behavioural emergence (Kite Field)

We studied the first population on a 12.5 ha study site, named Kite Field, within an open meadow from 2010 to 2020 (lat: 50.65, long: −114.64; electronic supplementary material, figure S1). The study grid was bordered by the Sheep River, a secondary highway, and continuous meadow on two sides, allowing immigration into and emigration from the site. To assess the timing of torpor termination, we fitted a subset of individuals within this population (males = 61, females = 59) with collar-mounted temperature-sensitive data loggers (Thermochron iButton DS1921; Maxim Integrated, San Jose, CA; approx. 8g) that monitored thermoregulatory patterns during hibernation by recording skin temperature (T_sk). We outfitted individuals with collars during the late summer just prior to hibernation and aimed to recapture individuals every three days before hibernation to ensure proper fitting of the collar up until hibernation entry. We programmed iButtons to record T_sk every 240 min, and recovered them the following year after squirrels emerged from hibernation. We defined termination from torpor as the date of the final rewarming bout from hibernation (i.e. a sustained T_sk > 30°C, typically followed by behavioural emergence in the following hours in females or days in males). Each spring, we observed behavioural emergence from hibernation (i.e. first above ground appearance) through systematic surveys of the study grid conducted multiple times daily (sensu [35]), recorded the date, captured and identified individuals, removed their collars, weighed each individual with a spring balance (Pesola Co., Baar, Switzerland), and examined them externally for reproductive condition. Males were assessed as reproductively capable (i.e. testes descended into the scrotum) or not, and reproductive readiness of females was assessed based on observed behaviours and swelling or opening of the vulva monitored during recaptures [35].

(c) . Behavioural emergence and male fitness (Meadow B)

We monitored the second population within a 1.8 ha study site, named Meadow B, bordered on three sides by forest and on the fourth by the gorge of the Sheep River (lat: 50.64, long: −114.67; electronic supplementary material, figure S1). We monitored female and male behavioural emergences from hibernation at this site from 3 m observation towers between 1992 and 2018 with daily monitoring during annual spring surveys (methods outlined in [33,36–38]), and captured and processed ground squirrels in the same manner as in the Kite Field population. In spring each year, we typically captured ground squirrels on the day they emerged from hibernacula or rarely up to 3 days later. Territorial males successfully defended a central core of their activity range early in the mating season, where non-territorial males lost interspecific fights and chases across their activity ranges [36,39].

We estimated male reproductive success in the Meadow B population from 2005 to 2018 as the number of weaned offspring sired each year, with mothers usually censused with dependent juveniles on the day or subsequent day to their first emergence above ground from the nest burrow (mothers and litters were rarely censused two days following emergence [32,33,39]). We determined paternity following previously established methods [32,33] using 13 amplified microsatellite loci from DNA extracted from ear tissue samples taken from each individual upon their first capture or from a tissue biopsy from the hind feet (outlined in [40]).

(d) . Snowmelt data

Following previous work [18,19,41], we related emergence phenology of individuals to dates of snowmelt (Julian dates, number of days since Jan 1). While female ground squirrels are phenotypically plastic to both spring temperatures and dates of snowmelt, the two weather variables are colinear (fig. 1c in [19]). Therefore, we restrict our current analyses to dates of snowmelt. We determined dates of snowmelt for each year from weather station data collected in Okotoks, Alberta [19] (Environment Canada, http://climate.weatheroffice.ec.gc.ca; 50°N, 114°W; approximately 50 km downwind (eastward) of the study sites).

(e) . Statistical analysis

We conducted all analyses in R v. 4.0.3 [42] using mixed-effects models (lme4 package) [43]. We determined significance of fixed effects with the car package [44] using type II (non-interacting terms) and III (interacting terms) ANOVAs. We used F-tests with a KenwardRoger approximation for degrees of freedom for Gaussian models and Wald chi-square tests for non-Gaussian models [45]. We generated confidence intervals around estimates by bootstrapping, using the confint function with 10 000 simulations. We determined significance of random effects with likelihood ratio tests by comparing models fitted with maximum likelihood with and without the random term of interest [46]. We verified model assumptions and fit by visually inspecting histograms and QQPlots of residuals, as well as the relationship between the residuals and the fitted values.

(f) . Sex-specific plasticity in hibernation phenology

(i) . Torpor termination (Kite Field)

We applied a linear mixed-effects model to annual dates of torpor termination (Julian date range: 83–150) of individual ground squirrels from the Kite Field population as the response variable. For snowmelt dates (Julian date range: 29–145), a ‘between-individual’ effect (β_B) captured the average environment experienced by an individual (i.e. the average date of snowmelt an individual experienced over their lifetime) and a ‘within-individual’ effect (β_w) captured annual environmental deviations experienced by individuals from their lifetime average (i.e. annual date of snowmelt minus the average snowmelt experienced by an individual over their lifetime [47]). As we were interested in evaluating sex-specific differences in plasticity in torpor termination date as it responded to annual snowmelt, we fitted an interaction between sex and both the β_B and β_w effects. To control for potential variation in torpor termination with age and years, we included interaction terms between sex and both of these terms (age range: 1–11; year range: 2010–2020). We also included the quadratic effect of age, and its interaction with sex, to explore whether middle-aged individuals may emerge earlier than young and old individuals, and whether this trend differed by sex. As we had data for some individuals in multiple years and multiple observations in each year, we included individual ID and year as random intercept effects (mean number of observations per individual: 1.37, range: 1–4; mean number of observations per year: 16.5, range: 5–33). This level of replication, however, was insufficient to fit random slopes. Following [47], we evaluated whether the β_B and β_w effects differed by fitting an additional model where we substituted the β_w effect with the original snow melt date variable so the β_B effect could be interpreted as the difference between these effects (β_B − β_w). We scaled continuous covariates between 0 and 1, and grand mean centered on zero prior to fitting models [45,46].

(ii) . Behavioural emergence (Meadow B)

We examined sex-specific plasticity in behavioural emergence from hibernation in the Meadow B population following a similar approach as for the torpor termination data. We fitted a linear mixed-effects model with behavioural emergence date (Julian date range = 93–142) as the response variable. We again generated β_B (i.e. average snowmelt date experienced) and β_w snowmelt effects (i.e. annual deviations from average snowmelt date experienced by each individual), and included interactions between both of these effects and sex. We included interactions between sex and both years (range = 1992–2018) and age (range = 1–14). We also included the quadratic effect of age and its interaction with sex. As we had data for individuals in multiple years and multiple observations per year, we included individual ID and year as random intercept effects (mean number of observations per individual: 3.1, range: 1–13; mean number of observations per year: 35.89, range: 11–71). We fitted a random slope to this model in an additional step for individuals across the β_w effect to determine whether among-individual differences existed in within-individual plastic responses to weather conditions [47]. We included only individuals that were reproductively mature in the analysis, so our results were not biased by reproductively immature individuals in the population that tend to emerge later than adults (e.g. yearlings and young males; [30,48]). We included females greater than one year of age (n = 208 individuals) or if they had reproduced in their first year (n = 12 individuals) and males that showed signs of reproductive maturity (i.e. with scrotal testes; n = 88 individuals). Following the approach used for torpor data, we also evaluated whether the β_B and β_w effects differed for behavioural emergence by fitting an additional model replacing the β_w effect with the original snow melt variable [47]. We scaled continuous covariates between 0 and 1, and grand mean centered on zero prior to fitting models.

(g) . Male fitness (Meadow B)

We first explored whether being territorial (1) or not (0; as defined in [36]) was predicted by a male's emergence relative to females in a given year (n = 46 individuals). We fitted a generalized linear mixed-effects model with territorial status as the binomial response variable. We generated annual relative emergence dates for males by subtracting the average female emergence date in each year from each male behavioural emergence date within that year. We included the fixed effects of relative emergence date and year, and individual ID as a random effect. We did not control for male density as this variable was correlated with year in our dataset (r = 0.58, p < 0.001) and so would be at least partially accounted for by the inclusion of year in the model. Relative emergence dates and age are correlated since males tend to emerge earlier relative to females as they get older (rho = −0.28, p = 0.002). We therefore also evaluated relative emergence date when including age in the model as collinearity between the variables did not strongly influence results (variance inflation factors < 3 [49]).

We next examined potential fitness consequences of variation in male behavioural emergences. We lacked power to examine among-year variation in fitness as there were n = 44 males with fitness data (mean = 2.56 observations per individual, range = 1–6, n = 114 observations). Therefore, we relativized annual reproductive success by dividing the number of offspring weaned in each year by the mean number of weaned offspring sired by males in that year; we termed this within-year reproductive success. Relativizing fitness within each year controlled for sources of variation among years (e.g. density or sex ratio) which may affect annual fitness. Distributions of male reproductive success are shown in electronic supplementary material, figure S2.

We fitted a linear mixed-effects model with within-year reproductive success as the response variable. We included linear and quadratic relative emergence dates as fixed effects, and individual ID as a random effect. In a subsequent model, we included the linear and quadratic effects of age as fixed terms to evaluate relative emergence date and age effects together. Relative emergence dates and age are also correlated in our fitness dataset (rho = −0.37, p < 0.001). Collinearity did not strongly affect model estimates as all variance inflation factors were < 3 [49], so we present selection gradient estimates for relative emergence date in the absence and presence of age effects. To allow comparisons of selection gradients, relative emergence dates and age were standardized to a mean of zero and unit variance prior to fitting the models. Within-year reproductive success did not follow a Gaussian distribution (electronic supplementary material, figure S2), but model residuals were approximately normally distributed with some right skew [50].

3. Results

Annual snowmelt dates for our populations have become delayed by 17 days over the full study period (n = 27 years from 1992 to 2018, linear model estimate ± standard error: 0.65 ± 0.73 days/year, p = 0.38; figure 2a) with a high amount of among-year variation. Over the shorter 10-year study period in the Kite Field population, mean annual dates of torpor termination and behavioural emergence dates did not show statistically significant trends over time in either sex (e.g. torpor termination date in males: rho = −0.21, p = 0.56; females: rho = −0.08, p = 0.84; electronic supplementary material, figure S3). Torpor termination and behavioural emergence were highly correlated within individuals in the Kite Field population (figure 3), with this correlation being slightly higher in females (n = 81 obs, rho = 0.98, p < 0.001) than males (n = 84 obs, rho = 0.91, p < 0.001). On average, males in the Kite Field population tended to delay behavioural emergence following torpor termination more than females (average 7.9 days under ground in males and 1.6 days in females). Males at Kite Field terminated torpor on average 17 days before females, and emerged above ground (i.e. behaviourally emerged) on average 11 days before females over the study period. When examining annual mean trends over the larger 27-year study period in the Meadow B population, male and female average behavioural emergence from hibernation has shown a delay over the study period (figure 2b) in line with the delayed dates of snowmelt (figure 2a). Males in Meadow B behaviourally emerged above ground on average 8 days before females over this longer study period.

Figure 2.

Trends over the 27-year study period for the Meadow B population (range = 1992–2018) for (a) annual snowmelt dates and (b) mean annual behavioural emergence dates for each sex (light blue = female, dark blue = male). Both snowmelt and emergence dates are shown in Julian days since 1 January. Both plots show the predicted line of best fit and 95% confidence intervals from linear regressions between the variables, and means in panel (b) are presented ± 1 standard error.

Figure 3.

Within-individual relationship between date of torpor termination and behavioural emergence (both in Julian dates from 1 January between the years 2010 to 2020) in the Kite Field population shown separately by sex (light blue = 59 females, dark blue = 61 males). Points have been made semi-transparent to show overlap and lines of best fit are shown. Plot includes multiple observations for the same individual (total observations = 165, mean = 1.3 observations per individual, range = 1–4).

(a) . Sex-specific plasticity in hibernation phenology

(i) . Torpor termination (Kite Field)

The within individual (β_w) response to snowmelt date did not significantly affect the date of torpor termination (non-significant β_w effect in table 1a) suggesting weak plasticity in this trait to spring snowmelt conditions. Within- and between-individual effects did not significantly differ between the sexes (non-significant β × sex interactions; table 1a; electronic supplementary material, figure S4), suggesting that both males and females respond similarly to snowmelt conditions (figure 4). The β_w effect and β_w × sex interaction were significant (p = 0.04 and p = 0.05, respectively) when including only individuals with more than one observation in this analysis (n = 31; electronic supplementary material, table S1) providing support for stronger plasticity, especially in females in this reduced dataset. The β_B and β_w effects of snowmelt, and their interactions with sex, were not significantly different from each other (electronic supplementary material, table S2). Male and female dates of torpor termination did not significantly change between 2010 and 2020 (non-significant year effect and sex × year interaction, table 1a; electronic supplementary material, figure S5A). Torpor termination was significantly earlier with age only for males (significant sex × age interaction, table 1a) and tended to become slightly later again as males aged (marginally non-significant sex × age² interaction; electronic supplementary material, figure S5B). When examining behavioural emergences in the subset of individuals from Kite Field for which we had torpor data, most results were qualitatively similar to what we report for torpor termination. However, we found significant sex differences in the within-individual effect (β_w) suggesting that female behavioural emergence was more plastic than males to snowmelt conditions at Kite Field over the shorter study period (electronic supplementary material, table S3).

Table 1.

Evaluation of fixed and random effects on (a) torpor termination (from the Kite Field population between the years 2010–2020) and (b) behavioural emergence (from the Meadow B population between the years 1992–2018; both in Julian dates from 1 January) in male and female Columbian ground squirrels. Estimates are from linear mixed-effects models using a within-subject centering approach to separate between- and within-individual responses to annual snowmelt dates (also in Julian dates since 1 January). Continuous variables were scaled and grand-mean centered (i.e. scaled between 0 and 1 and then mean-centered on zero) and the model intercept (β₀) includes females as the reference category. Quadratic effects are fitted and interpreted as raw polynomials where the model intercept (β₀) represents hibernation phenology at the mean of other covariates (here mean = 0).

	(a) Torpor termination (Kite Field)			(b) behavioural emergence (Meadow B)
	n = 165 observations and 120 individuals			n = 969 observations and 308 individuals
fixed effects	estimates	CI	p	estimates	CI	p
intercept (β0)	123.55			117.13
between-individual effect (βB)	9.26	−10.99–29.52	0.346	11.52	5.05–17.99	0.001
within-individual effect (βw)	19.52	−3.20–42.23	0.087	9.31	2.66–15.97	0.008
sex [male]	−18.51	−22.53 – −14.49	<0.001	−8.01	−9.14 – −6.89	<0.001
age	−1.96	−14.24–10.32	0.753	−11.23	−14.44 – −8.02	<0.001
age2	−6.16	−47.03–34.71	0.766	26.53	15.81–37.26	<0.001
year	−1.38	−19.79–17.04	0.870	6.17	1.02–11.32	0.021
βB × sex [male]	1.99	−15.56–19.54	0.822	−7.26	−12.71 – −1.81	0.009
βw × sex [male]	−13.32	−29.42–2.78	0.103	1.28	−2.46–5.01	0.502
age × sex [male]	−29.76	−46.63 – −12.89	0.001	−13.35	−20.96 – −5.73	0.001
age2 × sex [male]	49.81	−6.67–106.29	0.083	34.75	1.66–67.85	0.040
year × sex [male]	−7.98	−17.98–2.01	0.117	−2.60	−5.85–0.66	0.117
random effects
individual variance	38.83		0.003	6.07		<0.001
year variance	51.32		<0.001	12.90		<0.001
residual variance	41.64			22.53

Figure 4.

Sex-specific effects and associated 95% confidence intervals of annual individual deviations from the average date of snowmelt experienced (within-individual plastic effect or β_w) on torpor termination date (left, from Kite Field between the years 2010–2020, n = 165 observations and 120 individuals) and behavioural emergence date from hibernation (right, from Meadow B between the years 1992–2018, n = 969 observations and 308 individuals). Hibernation phenology traits are shown in Julian dates from 1 January. Vertical line of observations at zero on the x-axes (more apparent in left panel) indicates individuals with just one observation (i.e. a deviation of zero).

(ii) . Behavioural emergence (Meadow B)

The significant within-individual (β_w) response to dates of snowmelt reflected strong phenotypic plasticity in behavioural emergence from hibernation over the longer study period at Meadow B (table 1b). The β_w effect did not significantly differ between the sexes for behavioural emergence (non-significant β_w × sex interaction; table 1b), suggesting that both males and females similarly adjust emergence from hibernation to snowmelt variation (figure 4). Excluding individuals from this analysis with only one observation did not qualitatively change this result (electronic supplementary material, table S1). At the β_B level, individuals of both sexes that tended to experience later years of snowmelt on average, also tended to emerge from hibernation relatively later. The β_B × sex interaction was significant and showed a stronger β_B effect in females compared to males (electronic supplementary material, figure S4). There was no significant difference between the β_B and β_w effects at the population level (β_B − β_w effect; electronic supplementary material, table S2), but there was a stronger difference in males where the β_w effect explained most of males' responses to snowmelt variation (significant and negative β_B − β_w × sex interaction; electronic supplementary material, table S2).

When fitting individual random slopes along the β_w effect in an additional model, we found weak evidence for among-individual variation in plastic responses to snowmelt (i.e. non-significant variance in individual random slopes; electronic supplementary material, table S4). The delay in behavioural emergence date (6 days over the 27-year study period) was not significantly different between the sexes (non-significant sex × year interaction; table 1b; electronic supplementary material, figure S5C). As well, emergence dates were earlier as squirrels aged (significant age effect; table 1b) and then tended to become later again after about seven years of age in both sexes (significant age² effect; table 1b). This advance and subsequent delay in emergence date with increasing age was more pronounced in males than females (significant sex × age² interaction; table 1b; electronic supplementary material, figure S5D).

(iii) . Male fitness (Meadow B)

Males were more likely to be territorial if they emerged earlier relative to females (electronic supplementary material, table S5A; n = 119 observations and 46 individuals, intercept or β₀ = 2.29, log odds estimate = −10.72, p < 0.001). Males were predicted to have a 0.49 probability of being territorial if they emerged from hibernation 5 days before females and this increased to a 0.97 probability if they emerged 15 days earlier. Older males were significantly more likely to be territorial, and when evaluating relative emergence date and age together, relative emergence date was no longer significant (electronic supplementary material, table S5B). Males that emerged earlier relative to females also tended to have higher reproductive success than those that emerged later (electronic supplementary material, figure S6; n = 113 observations and 44 individuals). We estimated a negative selection gradient for relative emergence date, with this estimate being higher and statistically significant only when excluding age effects (slope = −0.17, p = 0.02 in table 2a versus slope = −0.09, p = 0.26 in table 2b). Both age and the quadratic effect of age significantly affected relative within-year fitness (table 2b). Relative within-year fitness increased with age and then tended to decline after 6 years of age.

Table 2.

Evaluation of fixed and random effects on the relative within-year reproductive success of male ground squirrels from the Meadow B population (n = 113 observations and 44 individuals) (a) when excluding age effects and (b) when including age effects. Estimates are from a linear mixed-effects model with individual random intercepts. Within-year reproductive success was calculated as the number of pups weaned, which was relativized within each year. Relative emergence date and age are standardized to a mean of zero and unit variance. Quadratic effects are fit and interpreted as raw polynomials where the model intercept (β₀) represents the within-year reproductive success at the mean emergence date and age (both means centered on zero).

fixed effects	(a) no age effects			(b) age effects
	estimates	CI	p	estimates	CI	p
intercept (β0)	0.99			1.17
relative emergence date	−0.17	−0.32–−0.029	0.021	−0.089	−0.24–0.062	0.26
relative emergence date2	−0.036	−0.12–0.042	0.35	−0.020	−0.094–0.054	0.60
age				0.29	0.082–0.49	0.009
age2				−0.19	−0.33– −0.068	0.003
random effects
individual variance	0.04		0.45	0.03		0.62
residual variance	0.52			0.50

4. Discussion

The uncoupling of community interactions as a result of asynchronous phenological responses to climate change has been raised as a conservation concern [2,51–53]. Lane et al. [19] found that female Columbian ground squirrels plastically delay emergence from hibernation in response to delayed snowmelt, but a lack of similar information on males has prevented an understanding of potential desynchrony between the sexes. Here we show that male behavioural emergences are at least as phenotypically plastic as females to variation in spring snowmelt. Consequently, despite a modest trend for progressively delayed behavioural emergences (approx. 6 days over 27 years) observed in this study, levels of protandry remained unaffected. We also found support for the primary assumption underlying most hypotheses for protandry in vertebrates—reproductive benefits gained by relatively earlier emerging males [16,17,54]. We found that earlier emergences benefited males by raising their likelihood of securing a breeding territory and increasing their reproductive success relative to later emerging competitors. Older males tend to emerge earlier in our populations, and so it is unclear whether early emergence, age or both drive these reproductive benefits. Sexual selection could, thus, be a primary evolutionary influence on protandry in this population, and similar degrees of plasticity in behavioural emergence between the sexes synchronizes phenological advances in males and females and maintains the degree of protandry in the populations. These results provide insight both into the potential conservation implications of climate change to this species and the evolution of protandry in sexually reproducing organisms.

The selection balance hypothesis was based on Møller's [13] observation of a stronger phenological response to warming spring conditions in male, as compared to female, barn swallows (Hirundo rustica), as also more recently reported in willow warblers (Phylloscopus trochilus [14]). Møller [13] posited that an amelioration of the natural selection costs of early arrival tip the balance in favour of stronger sexual selection on males, leading to ever earlier arrivals of males relative to females (i.e. increasing levels of protandry with advancing warmer spring conditions). Under the selection balance hypothesis, we predicted that males would emerge from hibernation much earlier than females in years with earlier snowmelt to improve their reproductive success. Although earlier relative emergences by male ground squirrels were associated with increased reproductive success, sexual selection did not appear to lead to increased levels of protandry in our populations.

Our results more strongly support the common cue hypothesis, which predicts synchronous phenological adjustments to weather variation between the sexes. According to the common cue hypothesis, males and females should respond similarly to environmental cues because, although sexual selection favours earlier male arrival relative to females, the benefits have a limit. If a male Columbian ground squirrel can become reproductively ready, and secure a breeding territory by terminating torpor and emerging a set amount of time before the first female, there is probably little benefit to ending torpor even earlier, especially if the natural selection costs of early arrival reduce overall fitness. There is evidence in other species, for example, of diminished survival (e.g. due to predation) of emerged individuals relative to those remaining in hibernation, presumably as they have left the relative safety of the hibernaculum [24,25]. Male phenological plasticity to weather variation and selection for earlier male emergence dates, taken together, suggest that males are selected to emerge earlier than females but the benefits of a further male phenological advance are limited.

We found less support for the sex-specific constraint hypothesis because behavioural emergence over the longer study period at Meadow B was plastic to weather variation in both sexes. This hypothesis assumes that circulating testosterone in males during spermatogenesis prevents reentry into torpor and thereby constrains male phenological plasticity to snowmelt date [27]. Our results from Kite Field did not fully support weaker male plasticity. Despite a strong within-individual correlation between torpor termination and behavioural emergence dates, males had (i) limited, but similar, plasticity to females in torpor termination (this result was sensitive to repeat observations) and (ii) weaker plasticity in behavioural emergence compared to females. The inconsistent results between these correlated phenological traits may suggest we lack power to draw strong conclusions from the Kite Field data, notably as these results are drawn from a subset of collared individuals in the population for which we have torpor data. Since our previous study on female ground squirrels at Meadow B [19], we include seven years of additional data, of which four years had earlier spring snowmelt dates (figure 2a). This may explain why we see a weaker delay in snowmelt dates than previously reported in our study area, and why we find less evidence for phenological plasticity to snowmelt conditions over the shorter and more recent study period at Kite Field. Our finding that both sexes show strong and similar plasticity in hibernation phenology at Meadow B provides the strongest support against a male constraint, but further study will be required to determine whether this conclusion is robust with changing climate in the coming decades.

Our conclusion that male Columbian ground squirrels are plastic to variation in snowmelt over a 27-year period differs from findings in arctic ground squirrels [15,18]. The divergent responses may be due to different behaviours occurring during the period from torpor termination to behavioural emergence in the two species. While males of both species remain in their hibernaculum following torpor termination, this period of time is substantially longer in arctic (i.e. 15–25 days [18]) than Columbian ground squirrels (average of eight days in our study). During this time, arctic ground squirrel males are covered by a substantial snowpack that can remain for four to six weeks before green-up, whereas snowmelt in our Columbian dataset on average precedes torpor termination. As a result, they must subsist on a hoard of food cached the previous autumn [55]. The period of time between torpor termination and snowmelt is shorter for Columbian ground squirrels and males are not known to cache at our study location [31]. A more extended period of time underground while covered by a prolonged snowpack, along with access to a food cache, apparently partially disassociates the timings of the termination of torpor and emergence above ground for reproductive male arctic ground squirrels. In comparison, the two events occur closer in time in Columbian ground squirrel males and, as a consequence, weather conditions during torpor termination are more likely to resemble those during behavioural emergence. This would contribute to the strong correlation between these two phenological traits in both sexes, which refutes the sex-specific constraint hypothesis for Columbian ground squirrels.

Sexual selection theory states that the reproductive success of males, relative to females, is more limited by access to mates [21] and so achieving reproductive preparedness to maximize the duration of tenure in the mating season via earlier male phenology should be favoured by sexual section. Indeed, earlier male phenology has been shown to lead to increased reproductive benefits in barn swallows (Hirudo rustica [56]), great reed warblers (Acrocephalus arundinaceus [57]), and red deer (Cervus elaphus [58]). Such reproductive success benefits to males have previously been suggested as the reason for protandrous emergence from hibernation in Richardson's ground squirrels (Urocitellus richarsonii [54]) and, in a year of relatively early female emergences (due to a heat wave), many males of this species missed out on reproductive opportunities as they were physiologically incapable of breeding [59].

Similarly, we show that male Columbian ground squirrels that emerged earlier from hibernation relative to the mean female emergence in a given year tended to have higher reproductive success. This benefit was facilitated by a higher likelihood of securing and defending a breeding territory. Emergence dates of males were significantly earlier relative to females as they aged, and the estimated strength of selection and its statistical significance was conditional on the inclusion of age in our models (i.e. emergence date became a non-significant predictor of reproductive success when also considering age). Older males were also more likely to hold territories and it is, therefore, not clear whether selection acts indirectly on relative emergence dates through age. Parsing causation among these variables awaits further study. These patterns are similar to those reported for previously studied sexually selected traits in other species. For example, antler sizes of red deer correlate positively with lifetime breeding success, yielding a positive sexual selection gradient [60]. Antler size also increases as males enter prime age and then plateaus in the oldest age groups. Collectively, these patterns support that these are important sexually selected traits, and suggest that only mature males can support the energetic costs of emerging earlier from hibernation or growing larger antlers. More broadly, our study examining hibernation phenology in Columbian ground squirrels provides further evidence of the role of sexual selection in the evolution of protandry.

A linear sexual selection gradient on male emergence dates combined with consistent phenotypic plasticity between the sexes, and a lack of variation in protandry coincident with interannual variation in emergence dates is, in some ways, surprising. Specifically, if our assumption of an optimal and earlier emergence date in males relative to females were true, we would expect to see a nonlinear selection gradient. The nonlinear term was a weak effect in our analyses, and this may be due to the dependence of the selection gradient (linear or nonlinear) on the phenotypic variance that is expressed in the population [61]. Simply put, if individuals do not enter the breeding population early enough to show a fitness cost to this phenotype, it is impossible to quantify its selective consequences. Specifically in our study, if individuals terminated torpor too early and died underground, we would not have records for either torpor termination (which can only be determined by retrieving their temperature logger) or behavioural emergence.

In conclusion, we have shown that phenological synchrony between male and female Columbian ground squirrels is maintained through similar degrees of phenotypic plasticity. Despite an earlier phenology for males relative to females, and a sexual selection benefit of early emergence, protandry did not increase during years of earlier snowmelt as would be predicted by the selection-balance hypothesis, and as has been shown in some birds during warmer springs [13,14]. It also did not increase during springs with later snowmelt as would be predicted by the sex-specific constraint hypothesis, as has been shown in congeneric arctic ground squirrels [15,18]. Instead, our results most strongly support the common cue hypothesis, and suggest the occurrence of two sex-specific optimal emergence dates. More broadly, there is growing concern about how interacting components of ecosystems may differentially respond to climate change, and how this uncoupling may disrupt ecological communities [51,52]. Asynchrony between the sexes within populations could jeopardize the efficient functioning of mating systems and, if extreme, even compromise overall reproduction and population sustainability. Although such an uncoupling is not apparent in Columbian ground squirrels, the disparate results reported to date make clear that a general expectation across species may be inappropriate. Therefore, more studies of sex-specific responses to weather variation and climate change are needed before any generalities can emerge.

Ethics

Alberta Parks Research and Collection Permit: 21-326 & 22-202. Alberta Wildlife Research Permit and Collection License: 21-205. University of Saskatchewan Animal Use Protocol: 20140021. Auburn University IACUC protocol: 2021-3861

Data accessibility

Data and code to support the results can be accessed on Dryad at https://doi.org/10.5061/dryad.bg79cnpg7 [62].

Supplementary tables and figures are provided in the electronic supplementary material [63].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

M.J.T.: conceptualization, formal analysis, investigation, validation, visualization, writing—original draft; F.S.D.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, writing—original draft, writing—review and editing; D.W.C.: data curation, funding acquisition, investigation, project administration, resources, writing—review and editing; J.O.M.: conceptualization, funding acquisition, investigation, project administration, resources, writing—review and editing; S.R.: investigation, methodology, resources, writing—review and editing; J.E.L.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

M.J.T. was supported by a Canadian Graduate Scholarship from the Natural Sciences and Engineering Research Council (NSERC) of Canada, a doctoral scholarship from the Fonds de recherche du Québec Nature et technologies (FRQNT), and a PhD mobility grant from le Centre Méditerranéen de l'Environnement et de la Biodiversité (CeMEB). F.S.D. was supported by a National Science Foundation of the USA grant (DEB-0089473), by a fellowship grant from the Institute of Advanced Studies of the University of Strasbourg, and by a Gutenberg Excellence Chair from the Région Grand Est and the Eurométropole de Strasbourg. D.W.C., J.O.M. and J.E.L. were supported by Natural Sciences and Engineering Research Council of Canada Discovery Grants (D.W.C.: RGPIN-2018-04354, J.E.L.: RGPIN-2020-06781). S.R. was funded by a Swiss National Science Foundation grant to Peter Neuhaus (SNF 3100AO-109816). The authors have no conflict of interest to declare.