High-Arctic family planning: earlier spring onset advances age at first reproduction in barnacle geese
Abstract
Quantifying how key life-history traits respond to climatic change is fundamental in understanding and predicting long-term population prospects. Age at first reproduction (AFR), which affects fitness and population dynamics, may be influenced by environmental stochasticity but has rarely been directly linked to climate change. Here, we use a case study from the highly seasonal and stochastic environment in High-Arctic Svalbard, with strong temporal trends in breeding conditions, to test whether rapid climate warming may induce changes in AFR in barnacle geese, Branta leucopsis. Using long-term mark–recapture and reproductive data (1991–2017), we developed a multi-event model to estimate individual AFR (i.e. when goslings are produced). The annual probability of reproducing for the first time was negatively affected by population density but only for 2 year olds, the earliest age of maturity. Furthermore, advanced spring onset (SO) positively influenced the probability of reproducing and even more strongly the probability of reproducing for the first time. Thus, because climate warming has advanced SO by two weeks, this likely led to an earlier AFR by more than doubling the probability of reproducing at 2 years of age. This may, in turn, impact important life-history trade-offs and long-term population trajectories.
1. Introduction
Global warming may have dramatic eco-evolutionary consequences [1,2] by changing long-term population dynamics [3] and the evolution of life-history traits [4,5]. The fastest warming occurs in the Arctic [6], where, as a consequence, the timing of snow melt and vegetation growth onset in spring is advancing rapidly [7,8]. Since the snow-free season is extremely short at high latitudes, prolonged snow cover often has detrimental effects on reproduction in ground-nesting birds [9]. Accordingly, advancing springs due to recent climate warming have proven beneficial [3,10]. Changes in age-specific breeding success can trigger changes in key life-history traits like the age at which individuals mature [11] or reproduce [12] for the first time. Age at first reproduction (AFR) is linked to the fast–slow life-history continuum, where longer-lived species generally exhibit delayed, and larger individual variation in, AFR [13,14]. An individual's AFR will affect its fitness, owing to costs and benefits associated with different life-history strategies [14,15]. Earlier AFR can be beneficial, by increasing the total number of reproductive events, but can come at a cost if resources are used that would otherwise be allocated to growth, survival or future reproduction. Environmental stochasticity and density dependence can also induce variability in AFR [16,17], as high resource competition or poor breeding conditions can lead to individuals delaying maturation [18] or reproduction [11]. While weather conditions are known to influence annual AFR in some species (e.g. common tern, Sterna hirundo [19], red deer, Cervus elaphus [20]), the link between long-term climate change and trends in AFR remains largely unexplored (but see [21,22]).
Geese migrating to Arctic breeding grounds experience highly variable spring conditions. Consequently, their reproductive success exhibits large inter-annual fluctuations, while adult survival is generally high and buffered against variability [23,24], a common pattern in long-lived species. In Arctic geese, there is substantial age-related variation in reproduction [25], as well as temporal variation associated with timing of nesting, density dependence and food availability [26–28]. Although temporal variation in their AFR has been documented [29,30], potential environmental causes of this variation have received little attention. Accurately estimating AFR can be challenging owing to detection issues and because an individual's breeding state is not always ascertainable. Multi-event models are widely used to quantify state uncertainty, such as mortality [31] or breeding status [32], by evaluating them as a hidden Markov process [33]. Here, using a multi-event framework, we studied causes of variation in AFR, defined as the first production of goslings, in the female portion of a population of Svalbard barnacle geese, Branta leucopsis. We hypothesize that an early spring, which has proven beneficial for reproduction overall in this population [34], reduces individual AFR. Since spring onset (SO) is advancing rapidly, this predicts, in turn, a temporal decline in AFR.
2. Material and methods
(a) Study species and data collection
Our study population of breeding barnacle geese is located around Ny-Ålesund (Kongsfjorden), Svalbard (78.9° N, 11.9° E). The Svalbard flyway population overwinters at Solway Firth, UK (55° N, 3.30° W), then travels north in spring with a stopover along mainland Norway before arriving at the Svalbard breeding grounds. Barnacle geese are long-lived (up to 28 years-old) and become sexually mature at 2 years of age [25,35]. They are partial capital breeders, using reserves acquired at wintering and stopover sites earlier in the annual cycle to initiate reproduction [36,37]. Over a 26 year period (1991–2017), 480 female goslings were caught at Ny-Ålesund and ringed with unique colour and metal bands during moulting (July/August). Geese nest on islands during May–June. After hatching, families return to Ny-Ålesund to forage, where ringed adults and associated goslings are recorded, resulting in 3006 individual observations used to model AFR (electronic supplementary material, appendix S1a). Males were excluded from the dataset owing to lower recapture rates [35]. Date of SO and adult population density (POP) were included as time-varying covariates. Accumulated winter snowfall [38] was included initially, but showed no evidence of an effect. SO is the (ordinal) day when the 10 day smoothed daily temperature crosses 0°C and remains above for at least 10 days [39] and has been shown to affect egg production [34]. POP is an annual estimate of adult numbers in the study population, which negatively affects gosling production and fledgling recruitment [34,40].
(b) Statistical analysis
Mark–recapture data were used to estimate AFR, where reproduction is defined as a female producing goslings (recorded at the foraging grounds, see electronic supplementary material, appendix S1a). Data consisted of individual capture histories of female barnacle geese, recorded as observed with at least one gosling, observed without goslings, or not observed, in a given year. A multi-event model, run in program E-SURGE (Multi-Event SURvival Generalized Estimation; v. 2.1.4 [41]), was used to separate states, representing the ‘true' reproductive status of an individual in a given year, and events, i.e. the observed state of an individual. We modelled four states, pre-breeder (PB), non-breeder (NB), breeder (B) and dead (†). PB was any individual not breeding at year t that had never bred previously. NB included individuals not breeding at year t but that had bred in a previous year. B was any female that produced at least one gosling at year t and † includes dead and permanently emigrated individuals. Three events were considered: ‘not seen', ‘seen as breeder' and ‘seen as non-breeder'. Only individuals in the B state could give rise to a ‘seen as breeder' event, whereas both PB and NB states contributed to ‘seen but not breeding' events, and individuals in all three states could be recorded in a ‘not seen' event (figure 1). See table 1 for definitions.
Figure 1. Multi-state model of barnacle geese. Circles represent ‘true', unobservable states, with black arrows indicating transitions between states from time t−1 to t. Squares are observable events and grey-dotted arrows show which event(s) would be observed given an individual's state.
terminology | meaning | definition |
---|---|---|
AFR | age at first reproduction | the age at which a female first produces goslings that survive to the foraging area (around Ny-Ålesund) |
state | true annual state | PB, B, NB and †; not always observable; an individual without goslings may be PB or NB, depending on its reproduction history |
transition | shift between states from year t−1 to year t | transition probability from any (living) state at t-1 (i.e. B, PB, NB) to state B at year t represents the breeding probability at year t |
event | annual observed reproductive situation | events include seen as a breeder (i.e. with goslings), non-breeder and not observed |
PB | pre-breeder | state of females that have yet to produce goslings for the first time (Note: reproduction probability of PB refers to individuals in PB at t−1 that transitioned into B at t.) |
B | breeder | state of birds producing one or more goslings in a given year |
NB | non-breeder | state of birds not producing goslings during breeding season but having bred previously |
† | dead | state dead includes dead and permanently emigrated individuals |
SO | spring onset date | (ordinal) day when 10 day smoothed daily temperature crosses 0°C and remains above for at least 10 days |
POP | population density | annual estimated number of adults in the study population at Ny-Ålesund |
Goodness-of-fit (GOF) tests on a simplified, multi-state dataset (n = 687, four states: PB, B, NB, not observed) in program U-CARE (v. 2.3.4 [41]) indicated transience, which was accounted for by modelling age-dependent apparent survival, and trap-history-dependent recapture, which was not considered problematic for this analysis (see electronic supplementary material, appendix S1b for details). Details on model implementation are to be found in electronic supplementary material, appendix S1c.
Following [40] and the GOF tests, annual survival probabilities were modelled for goslings, yearlings and adults, including year effects, and recapture probabilities were modelled as year-specific. Transition probabilities (ψ) to the breeding state were assumed to be the same from NB and B states (ψNB/B → B). We compared models with covariates (SO, POP) on transition probabilities from PB to B (ψPB → B) and from NB and B to B (ψNB/B → B). An age effect was included on ψPB → B, where females of 4 years or older were pooled because of reduced sample sizes thereafter. Model selection was based on Akaike's information criterion corrected for small sample sizes (AICc). A model was considered a better fit when ΔAICc was reduced by at least 2 [42]. Confidence intervals for parameter estimates were calculated using the delta method [43].
Using the Viterbi algorithm in E-SURGE, we reconstituted the 30 most-probable life histories for each individual, and their probabilities, based on the highest-ranked model. From the output, we estimated the AFR distribution in the population and the annual proportion of breeding 2 year olds (electronic supplementary material, appendix S2).
3. Results
The best-fitting model (table 2) explaining the pre-breeder to breeder transition (ψPB → B) included an effect of SO and an interaction effect between age class and POP. The non-breeder/breeder to breeder transition (ψNB/B → B) also included a SO effect (ψNB/B → B (logit scale) β = −0.29; 95% CI = −0.40, −0.17), which was weaker than on ψPB → B (−0.44; −0.63, −0.25), as the mean estimate of ψPB → B was outside the confidence interval of ψNB/B → B. In other words, the probability of producing goslings decreased with delayed SO and more so for first-time breeders (figure 2a). POP had a negative effect on the probability of reproducing for the first time for females of age 2 years (−0.60; −0.93, −0.28) but no effect on ages 3 years and older (0.12; −0.09, 0.34) (figure 2b).
Figure 2. (a) Effect of spring onset date, SO, on reproduction probability of first-time (pre-breeders, PB) and experienced (non-breeders or breeders, NB/B) mothers. (b) Population density, POP, effects on age classes 2 and 3+ in PB. Annual (c) estimated proportion of 2 year olds reproducing, (d) SO and (e) POP. Dashed lines indicate (c) trend towards an increasing proportion of 2 year old individuals reproducing for the first time, estimated with E-SURGE (see Methods), and (d) advancing spring phenology.
rank | model ψPB → B | model ψB/NB → B | k | AICc | ΔAICc |
---|---|---|---|---|---|
1 | age2–3+ × POP + SO | SO | 6 | 9760.9 | 0 |
2 | age2–3+ × POP + SO | SO + POP | 7 | 9762.1 | 1.2 |
3 | age2–4+ × POP + SO | SO | 7 | 9762.4 | 1.5 |
4 | age2–3+ × POP age2–4+ × SO | SO | 9 | 9763.4 | 2.5 |
5 | age2–4+ × POP + SO | SO + POP | 8 | 9763.5 | 2.6 |
6 | age2–3+ × POP age2–3+ × SO | SO | 8 | 9764.9 | 4.0 |
7 | SO | SO | 4 | 9773.3 | 12.4 |
8 | SO + POP | SO | 5 | 9773.5 | 12.6 |
9 | age2–4+ × SO + POP | SO | 7 | 9774.3 | 13.4 |
10 | SO | SO + POP | 5 | 9774.6 | 13.7 |
Based on estimated individual AFR, 35% of individuals reproduced for the first time as 2 year olds, while 88 and 97% had reproduced by 5 and 10 years of age, respectively (electronic supplementary material, appendix S2). The top-ranked model suggested that a substantial number of individuals that were not observed as 2 year olds were breeding (appendix S2). Furthermore, the estimated proportion of 2 year olds reproducing each year more than doubled over the study (figure 2c) and the date of spring onset, SO, advanced by approximately two weeks (β = −0.55, s.e. ± 0.19, p-value < 0.01, figure 2d). This provides support for our prediction of declining AFR over time with advancing spring phenology. Population densities, POP, showed no significant temporal trend (β = −0.01, s.e. ± 4.1, p-value = 0.99, figure 2e).
4. Discussion
This long-term study of Svalbard barnacle geese documents empirically the link between global warming and AFR, a key life-history trait. Although some (poor) individuals produce goslings for the first time later in life, AFR appears strongly linked to annual fluctuations in nest-site and resource availability. Earlier SO increased the probability of producing goslings, especially for females reproducing for the first time, suggesting that inexperienced breeders are more affected by environmental variation. Advancing SO, associated with ongoing climate warming, led to an increasing proportion of reproducing 2 year olds (i.e. age of sexual maturity) over the study. Density dependence, also operating through resource availability, only affected the probability of producing goslings for 2 year olds. This suggests that barnacle geese generally start breeding as 2 year olds and only poor conditions–i.e. cold springs or high densities—force them to delay. In such cases, AFR is likely to change over time with long-term trends in breeding conditions.
Spring phenology can affect AFR since it impacts both clutch success/size and hatching success, through effects on the timing of nesting and food availability during incubation [26,27,34]. Colder springs delay snowmelt, and therefore nest-site availability, but also the timing of food availability by delaying plant growth onset [34]. Similarly, under delayed snowmelt, female geese initially use retained reserves for self-maintenance rather than egg production [44] and take more frequent, and longer, breaks from incubation to forage, increasing egg predation risk [28,45].
Density dependence affects reproduction and thereby potentially the age at which females produce goslings. Here, 2 year olds were less likely to produce goslings in years with higher densities (i.e. higher intraspecific competition), supported by similar findings from a Baltic population of barnacle geese [30]. Reproductive success was also found to be age-dependent in the Baltic population [25], explained by increasing experience/social status with age. This may explain the impact of increased competition on young geese that are forced to settle at sub-optimal nesting sites as densities increase [46]. Better nest-sites have more forage available, limiting time spent off the nest for incubating females, limiting egg predation risk. The same mechanism may also have contributed to stronger effects of SO on pre-breeders (typically younger individuals), since late springs increase snow cover and thereby nest-site availability.
Global warming is having profound effects on reproduction in Arctic geese and other Arctic herbivores [34,47]. Our results, from one of the most rapidly warming places on Earth [6], indicate that climate change is affecting key life-history traits like AFR. Climate change is advancing spring, providing an explanation for the increasing proportion of 2 year olds reproducing and thereby earlier AFR. Reproduction is the main driver of population dynamics in geese, and any changes have substantial population-level effects [40]. However, increased production of goslings will, to some extent, shift the age structure towards a larger proportion of young individuals that are more sensitive to density-dependent processes, potentially counteracting benefits of earlier AFR somewhat. Additionally, here, AFR refers to production of goslings, but survival to fledging is highly variable and susceptible to predation [34,48]. Earlier AFR may also incur a cost through reduced future reproduction or survival, which was not possible to test here, but care should be taken when inferring population-dynamic implications. For long-distance migrants like Arctic geese, following the food-peak across migratory sites is an important evolutionary strategy [49,50]. However, they may, eventually, be unable to keep up with fast-changing spring conditions [51], leading to phenological mismatch in food-web interactions [52,53], with potentially negative reproductive consequences [54]. Nevertheless, this population shows no current indication of mismatch effects [34]. On the contrary, Arctic climate change appears to allow higher gosling production and earlier AFR, which may have positive consequences for population persistence.
Ethics
Permissions for the fieldwork were given by the Bird Ringing Centre, Stavanger, the Animal Experimentation Board of Norway (FOTS) and the UK Wildfowl and Wetlands Trust (WWT) and the Governor of Svalbard.
Data accessibility
Data are available from Dryad (https://datadryad.org/stash/share/jVK4KBYR_SWuyvK83KgFY1bWtYvjPRxvRYubSqkbspw) [55].
Authors' contributions
K.L.-M., A.M.L., B.B.H. and V.G. designed the study. M.J.J.E.L. conducted the fieldwork and provided the data. M.A.F. analysed the data, with input from K.L.-M., A.M.L., B.B.H. and V.G. M.A.F. and K.L.-M. led the writing of the manuscript, which was commented on and revised by all co-authors. M.A.F. and K.L.-M. contributed equally (i.e. co-first authorship). All co-authors gave final approval for publication and agreed to be accountable for all aspects of the study.
Competing interests
We declare we have no competing interests.
Funding
NWO, Ministry of Foreign Affairs, BIRDHEALTH (851.40.071), Geese on Arctic Tundra (866.12.407), EU FP7-project FRAGILE and University of Groningen provided funding for data collection. The Research Council of Norway (FRIMEDBIO 276080, KLIMAFORSK 244647 and 273451, Centre of Excellence 223257 and AFG 269961) supported this work.
Acknowledgements
We thank scientists, students and volunteers for data collection and processing.