Figure 4. Relationships between pup emergence date and fitness proxies. Panel a presents the output of our selection model examining the relationship between pup emergence date and the weighted proportion of pups surviving their first winter. Panel b presents the output of our selection model examining the association of pup emergence date with the total number of pups surviving their first winter. The black line represents the prediction from the models. The grey shading is the associated confidence intervals. Data points have been jittered to enable clearer visualization (number of females = 176; number of litters = 417).

Discussion:

We found that pup emergence date was weakly linked to female emergence date, with late emerging females mating in their burrow and early emerging females delaying reproduction. We found a positive effect of spring snowpack on the timing of pup emergence, but no effect of spring temperature. We found directional, but not stabilizing selection for pup emergence date, with pups that emerged earlier better surviving their first winter. We additionally found among-individual variation at the female level in pup emergence date, but no additive genetic basis for that variation. While there was population-level plasticity in response to average spring snowpack, there was no inter-individual variation in plasticity for either average spring snowpack or temperature.
We showed a weaker relationship between female and pup emergence date than expected. Indeed, there was substantial variation in pup emergence date, with the earliest pup emerging about a month before and the latest emerging about a month later than expected based on their mother’s emergence date (Figure 1). Gestation and lactation length were assumed to have a fixed duration (30 and 25 days respectively), but there might be some among-individual variation in both of their lengths. Yellow-bellied marmots are considered capital breeders and mate when little to no food is available in the environment (Armitage, 2014). Therefore, the body condition of a female might shorten or lengthen gestation by a few days. During lactation, most females have emerged from their burrow and thus both a female’s body condition and micro-environmental variation in food availability could lead to inter-individual variation in lactation length before pups are weaned. Again, variation by a few days is expected. Overall, among-individual and environmental variation in gestation and lactation length, would only explain a variation by a few days in the relationship between pup emergence date and their mother’s emergence date. Given that pups could emerge up to a month before and up to a month after expected based on their mother’s emergence date, it is clear the females are in some cases delaying reproduction after emerging and in others, able to mate in their burrow before emerging. Females delaying reproduction after emerging might be due to environmental variation, poor body conditions, and/or the absence of a male to mate with. Indeed, our results showed that pup emergence date was related to spring snowpack, with pups emerging later in springs with heavier snow (Table 2). This possibility of delaying reproduction because of spring snowpack may also explain why pup and adult emergence date are not similarly associated with spring temperature. Given that the date of emergence from hibernation of adult marmots is strongly related to spring temperature, (Inoyue et al. , 2000; Edic et al., 2020), we would also expect a positive relationship between pup emergence and spring temperatures. Yet, we find no association of spring temperature with pup emergence date (Table 2) and therefore, female marmots may delay their reproduction until there is less snow regardless of spring temperatures.
However, Andersen et al. (1976) postulated that delaying reproduction decreased fitness as the growing season was shortened for pups and females. Indeed, we found directional selection for earlier pup emergence even though females are emerging earlier from their hibernacula (Edic et al., 2020) and growing season length has increased (Cordes et al., 2020). To elaborate, litters that emerged earlier had increased probability to survive and increased number of pups surviving to one year old than litters that emerged later, indicative of directional selection (Table 4). This may be a result of increased time during the growing season to forage and gain weight when pups are born earlier in the season (Monclús et al. , 2014). Further, earlier born pups tend to be heavier at weaning than later born pups and this weight is positively correlated with overwinter survival (Monclús et al. , 2014). We do not find a similar pattern with litter size (Table 4). Given the selective pressures for earlier births in the marmots, we would predict that females that reproduce later in the season would produce fewer, but heavier pups than those females that reproduce earlier (Stearns, 1992). Indeed, it is predicted that in unfavourable environments, it is advantageous to not reproduce to your full capacity (Monclús et al. , 2011). However, in our study, it seems that regardless of the fitness costs associated with giving birth later in the season, females will give birth to the same number of pups regardless of when they will emerge. This may stem from previous stabilizing selection on the number of pups that can be produced. Further, Monclús et al. , (2014) report that mothers will not attempt to provide the offspring born later with more resources compared to offspring born earlier to mitigate the effect of being born later. Therefore, even though pups born later in the season will have lower survival, mothers will not invest more resources in them to reduce the fitness costs (Monclús et al. , 2014).
Despite these existing directional selection pressures to reproduce early, pup emergence date will not show an evolutionary response because of its low additive genetic variation (Table 3). There are two plausible explanations for this low variation. First, female marmots can re-absorb fetuses if they are not viable. By using pup emergence date as a proxy for the timing of reproduction, we are effectively removing all those females that may have reproduced, but not given birth to any pups. This removes a potentially significant source of variation in the trait and may explain the lack of observable heritability. If females reproducing too early or late tend to reabsorb or abort their pregnancies, this may also decrease variation through stabilizing selection. Secondly, the timing of reproduction is a fitness trait and fitness traits are generally reported to be less heritable compared to other traits (Merilä & Sheldon, 2000). This phenomenon is generally attributed to Fisher’s fundamental theorem (Price & Schluter, 1991) which proposes that there should be strong selection on fitness traits that maximally increase fitness (Merilä & Sheldon, 2000). This reduces the heritability by reducing the amount of variation in the trait – only the variation that provides the greatest fitness benefit will be left in the population following this strong selection (Merilä & Sheldon, 2000). This may explain the pattern we observe here. There may have been strong selection on the timing of pup emergence date to increase fitness, causing a reduction in the amount of additive genetic variance present and as a result lowering the heritability of the trait. However, there have been challenges to this theorem with suggestions that the lower heritability of fitness traits is not due to decreased additive genetic variance, but rather increased residual (Merilä & Sheldon, 2000), or environmental variance (Price & Schluter, 1991). In our model we report both low additive genetic variance and high residual variance.
Past selective pressures may also explain why we found no inter-individual variation in the plasticity of pup emergence date with spring snowpack (Figure 3). Female marmots are responding in the same way to the same changes in average spring snowpack. Inter-individual differences in the intercept in our plasticity model indicated that in the average environment, individuals are reproducing at different times (Nussey et al. , 2007). No covariation between the intercept and the slope indicated that there is no relationship between the timing of reproduction in the average environment and how plastic an individual is (Brommer, 2013). The lack of inter-individual variation in the slope in our population indicated that individuals do not differ in their response to changes in the environment. This may be explained by canalization (Stearns, 1982). Marmots are heavily constrained by their climate and have a relatively short period of time to reproduce and gain mass again prior to hibernating. Since there is strong selection to reproduce within a short window of time where fitness is optimized, and strong selection is expected to decrease the magnitude of inter-individual differences (Westneat et al. , 2009), this may explain the lack of IxE in our study population. Predation might also drive the small variation. If females varied substantially in the timing of their reproduction in response to the same environmental conditions, pups would emerge at different times, exposing them to increased predation risk as there are fewer pups available at any given time as prey (Michel et al., 2020).
Significant sources of variation in our animal model were valley, permanent environment, and year (Table 2; Table 3). Pup emergence date is earlier in the down-valley compared to the up-valley. This is to be expected as these two sites differ in elevation by about 200 m, causing a two-week delay in the phenology of the up-valley compared to the down-valley (Monclús et al. , 2014). Inter-individual variation in pup emergence date may be due to microenvironmental differences experienced by females such as burrow quality, foraging ability, or differences in environmental conditions experienced (e.g., trees preventing snow melt; van Vuren & Armitage, 1991). Inter-annual variation in pup emergence date may be expected due to yearly variations in environmental conditions such as variation in the number of males present or amount of snow in the area. We find no association of colony area with the date of pup emergence, but this may be because permanent environment effect and colony are correlated as females do not generally leave once they are reproductively mature (Edic et al., 2020). Colony effects that may have been confounded with permanent environment may be the number of individuals present, as marmots can produce more pups when there are fewer individuals in the colony (Maldonado-Chapparroet al., 2015), the number of males present in the colony, or the degree of reproductive suppression present in the colony. These factors could all impact the timing of reproduction in a colony-specific way.
For the model examining the annual number of pups surviving their first winter, we found that more pups survive their first winter in the up-valley compared to the down-valley (Table 4). This is likely due to differing predation rates between the valleys, with higher predation in the down-valley compared to the up-valley. Predation and winter conditions are the main causes of death in marmots, and young marmots are very susceptible to predation (Armitage, 2014). For the models on litter size and annual number of pups surviving, we report a positive effect of a mother’s mass in June (Table 4). June body mass of a mother has been reported to have a positive effect on the mass of her offspring, and heavier offspring are expected to have higher chances of overwinter survival (Monclús et al. , 2014). Additionally, as marmots are capital breeders, higher body masses are often associated with more resources available for reproduction, potentially explaining larger litter sizes for larger females.
There are some limitations to our dataset that may have impacted our results. First, despite best efforts, we might have some error on emergence date for pups and mothers because we rely on visual observations to determine emergence. While our observation effort is high in this study with colonies observed on a near-daily basis with approximately 1000 hours of observations logged per year, exact emergence dates may still be missed. We additionally tried to control for this by limiting our analyses to only the main colonies as these are observed with a higher frequency than satellite colonies. Therefore, we are less likely to have missed emergence dates in the main colonies compared to the satellite colonies. Further, we are only able to use pup emergence date as our proxy for the timing of reproduction. Being able to see inside burrows and know exactly when pups are born would provide a better estimate of the timing of reproduction in addition to identifying cases when all pups died during lactation. Similarly, being able to know when a female mated would also provide more information about pregnancy interruption (reabsorption and abortions). Additionally, we unfortunately only have data on female emergence date between 2003-2017. It would have been interesting to analyze pup and female emergence dates for more years. Finally, since we only have one weather station on site, the climate variables used are the same between valleys. In the future it would be interesting to separate weather variables between the valleys since the up-valley environment is harsher and there is phenology delay of about two weeks between the valleys.

Conclusions:

Overall, we report that while marmots are plastically adjusting the timing of pup emergence date in response to changing spring snowpack, individuals do not differ in their plasticity level. Further, pup emergence date is not heritable but there is selection for pups to emerge earlier. This indicates that pup emergence date may not have an optimum time and it is just better to emerge earlier. Without having inter-individual variation in plasticity and without being able to evolve in response to natural selection, this population may be limited in its ability to track optimal environmental conditions for reproduction. If climate continues to change, this may prove problematic. For instance, the length of the active season may change, altering the timing of food availability. If pups do not emerge early enough, they may not be able to gain enough mass prior to hibernation. Similarly, if the mother reproduces too late in the season, she will also be limited in her ability to gain sufficient mass for hibernation. This potential mismatch in the length of the active season and when pups emerge may impact population fitness, causing a decrease in pup and dam survival. Additionally, future research should investigate the discrepancy we report between female and pup emergence to determine the ecology behind this pattern.