Study species and data collection:
Yellow-bellied marmots are large (adult females weighing on average 2.5 kg and adult males weighing on average 3 kg; Armitage, 2014) hibernating rodents living up to 15 years. They have a four-month active season, from late-April/early-May to late-September, during which they need to reproduce and accumulate fat reserves to survive the eight months of hibernation (Armitage, 2014). A population of yellow-bellied marmots has been studied at the Rocky Mountain Biological Laboratory (RMBL) in Gothic, CO, USA since 1962.
Marmots were live trapped in Tomahawk traps regularly during the active season. When they were caught, data on their weight, sex, and reproductive status were collected. Upon first trapping, marmots were assigned a unique identifier and given a permanent ear tag for identification. For observations at a distance, Nyanzol-D, a semi-permanent dye, was applied in a unique pattern on each marmot. Starting in 2000, parentage has been determined using genetic assignment (for detailed methodology on genetic assignment see Blumstein et al., 2010). Prior to this, maternal identity could be reliably determined via behavioural observations while paternal identity remained unknown since males do not contribute to parental care. Daily climate data were collected by an on-site weather station since 1975. Data collected included daily minimum and maximum temperatures, daily precipitation, and depth of the snowpack. Mass on June 1st and August 15th were estimated for each individual every year using best linear unbiased predictors from age- and sex-specific linear mixed models (for detailed methods see Kroeger et al., 2018). For pups, mass was estimated for emergence date and not June 1st. Age was calculated using birth year and the year of capture. Since 83% of females are captured for the first time when they are juveniles, they are of known age. The study is divided into an up-valley and a down-valley that differ in elevation by 165 m (Ozgul et al., 2010) resulting in a delay in the phenology of the up-valley by about two weeks compared to the down-valley (Monclúset al. , 2014). For our analysis, we restricted our dataset to include only main colonies which are observed with a higher frequency (near daily) compared to satellite colonies and we could thus have a more accurate estimation of emergence dates.
To estimate the timing of reproduction, we used the date pups first emerge from their burrow after being weaned as a proxy. Since the length of gestation and lactation are considered fixed in the marmots, with 30 days spent gestating and 25 days spent lactating (Armitage, 2014), and pups emerge immediately following weaning (Monclús et al. , 2014) this is an excellent proxy for the timing of reproduction. Given that adult emergence in the spring is related to average spring temperature and average spring snowpack, we focused on these two environmental variables for our analysis of pup emergence date. Since seasonal averages of environmental variables can be estimated between any two arbitrary timepoints (Bailey & van de Pol, 2016), we used a statistical approach to determine which phenological window of these two variables had the greatest association with pup emergence date. This was done using the statistical approach built in the R package climwin (Bailey & van de Pol, 2016; van de Pol et al., 2016). This package allows the fitting of multiple models with different phenological windows used to estimate environmental averages and determine, using AIC based model comparison, which window has the strongest relation with the biological variable of interest.