Evidence for senescence in survival but not in reproduction in a short‐lived passerine

Abstract Senescence has been studied since a long time by theoreticians in ecology and evolution, but empirical support in natural population has only recently been accumulating. One of the current challenges is the investigation of senescence of multiple fitness components and the study of differences between sexes. Until now, studies have been more frequently conducted on females than on males and rather in long‐lived than in short‐lived species. To reach a more fundamental understanding of the evolution of senescence, it is critical to investigate age‐specific survival and reproduction performance in both sexes and in a large range of species with contrasting life histories. In this study, we present results on patterns of age‐specific and sex‐specific variation in survival and reproduction in the whinchat Saxicola rubetra, a short‐lived passerine. We compiled individual‐based long‐term datasets from seven populations that were jointly analyzed within a Bayesian modeling framework. We found evidence for senescence in survival with a continuous decline after the age of 1 year, but no evidence of reproductive senescence. Furthermore, we found no clear evidence for sex effects on these patterns. We discuss these results in light of previous studies documenting senescence in short‐lived birds. We note that most of them have been conducted in populations breeding in nest boxes, and we question the potential effect of the nest boxes on the shape of age‐reproductive trajectories.

long-term datasets rather than lack of senescence itself (Botkin & Miller, 1974). Over the last decades, accumulated empirical findings have shown that senescence is ubiquitous in natural animal populations affecting several fitness components (Nussey, Froy, Lemaitre, Gaillard, & Austad, 2013).
More recent studies have investigated how the rate of senescence varies between sexes and among life-history traits (Bouwhuis & Vedder, 2017;Hayward et al., 2015). Senescence may evolve differently in the two sexes, because selection pressure could be strongly asymmetric between males and females (Bonduriansky, Maklakov, Zajitschek, & Brooks, 2008). One major theoretical prediction is that the sex experiencing higher mortality should also exhibit higher rates of senescence (Williams, 1957). When mortality due to extrinsic causes is high, there is strong selection favoring early reproduction and weak selection for survival and reproductive success later in life. Such patterns have been reported in polygynous, dimorphic ungulate species where males suffer high mortality and exhibit stronger senescence than females (Clutton-Brock & Isvaran, 2007, but see Tidiere et al., 2015). In birds, and especially in short-lived passerines, it is often observed that females suffer from higher mortality than males because they bear the cost associated with egg production and egg-offspring care and defense (Bennett & Owens, 2002;Donald, 2007;Promislow, Montgomerie, & Martin, 1992). Beyond the physiological cost, females are exposed to higher predation risk due to nest attendance (e.g., Low, Arlt, Eggers, & Pärt, 2010). It is therefore predicted that females show more rapid senescence than males in short-lived passerines (Williams, 1957). Recent studies have provided some support for this prediction. For instance, house sparrow Passer domesticus and willow tit Parus montanus females have a shorter life span and higher reproductive and survival senescence rates, respectively, than males (Orell & Belda, 2002;Schroeder, Burke, Mannarelli, Dawson, & Nakagawa, 2012). In contrast, in a population of Seychelles warblers Acrocephalus sechellensis free from predator, survival did not differ between sexes and both sexes experience the same actuarial senescence rates (Hammers et al., 2019;Hammers, Richardson, Burke, & Komdeur, 2013).
Although comparisons of senescence patterns between sexes became more frequent in literature, they remain challenging to establish. First, the monitoring of natural animal populations is frequently biased toward females for practical reasons (Bouwhuis, Choquet, Sheldon, & Verhulst, 2012;Nussey et al., 2013). Second, obtaining descent sample sizes remains challenging especially for short-lived species like passerines. Indeed, the only reliable method to determine the age of a bird in the wild is to mark it at a known age, which is typically only possible at the juvenile stage in many birds. A marked bird needs then to be repeatedly observed as an adult to obtain individual longitudinal data, which are mandatory to investigate senescence rates (Nussey, Coulson, Festa-Bianchet, & Gaillard, 2008).
Obtaining such data is particularly difficult for passerines due to the extensive juvenile dispersal and high mortality (Cox, Thompson, Cox, & Faaborg, 2014;Weatherhead & Forbes, 1994). Thousands of chicks need to be ringed to obtain sufficient sample size to explore age pattern in adulthood. Therefore, the vast majority of studies on senescence are conducted for a limited number of species that are common and breed in high density in nest boxes, such as tits, flycatchers, or sparrows. Yet, to fully understand the evolution of senescence, it is important to test predictions for species with contrasting ecology.
In this study, we investigate for the first time age-specific survival and reproduction in male and female whinchat (Saxicola rubetra), a short-lived ground nesting migratory passerine. To overcome the difficulty of obtaining descent sample size from known age individuals, we jointly analyzed seven individual-based long-term whinchat datasets. According to the evolutionary theory of senescence and previous empirical studies, we expected that (a) senescence would occur in both, survival and reproduction, and that (b) the senescence rate would be higher in females than in males due to higher mortality of females (Donald, 2007).

| Model species
The whinchat is a small (15 g), insectivorous Afro-Palearctic migrant that inhabits open grasslands in both the breeding and nonbreeding areas (Cramp, 1988). Whinchats show high annual mortality, and only few individuals live more than 5 years (this study). The maximal life span ever recorded is 7 years (Fransson, Jansson, Kolehmainen, Kroon, & Wenninger, 2017). Whinchats are monogamous, nest on the ground, and usually raise one brood per year (Cramp, 1988).
Individuals are sexually mature at the age of 1 year. Females incubate the clutch (typically 5-6 eggs) and brood the hatchlings, but both sexes contribute to feeding the nestlings (Cramp, 1988). Similar to most farmland birds, whinchats have undergone massive population declines over the last decades (86% between 1980 and 2016 at the European scale, IUCN, 2019) due to agricultural intensification.

| Data collection
This study is based on the long-term monitoring of seven whinchat breeding populations from five European countries (United Kingdom 1, Slovenia 1, Russia 1, Germany 3, Switzerland 1). The duration of data sampling varied from 5 to 16 years depending on the population (Table 1). These populations are too far away from each other to allow substantial exchange among them and are thus demographically independent. In each population, both adults (1 year old or older) and nestlings were ringed with aluminum and color plastic rings. Based on plumage characteristics, some adults that were captured for the first time could be aged as 1 year old.
Although the sex of nestlings was unknown, sex was systematically recorded for birds ringed or reobserved as adults based on plumage dimorphism. In total, 5,553 individuals have been ringed.
A substantial proportion of them, however, have been ringed as fledgling and were never observed thereafter. Only 6% of all ringed nestlings have been recaptured as adults, which is a standard return rate in passerines (Weatherhead & Forbes, 1994). In the end, the life histories of 1,461 adults have been recorded, and the exact age was known for 493 of them. Additionally, whinchat nests were searched and the presence of fledglings was recorded to determine the breeding success. A marked individual was considered successful if it had produced at least one fledgling during the breeding season. The number of nest visits was kept at a minimum in order to reduce the potentially negative impact of the monitoring. The available information about reproduction was therefore limited to whether or not a brood was successful (produced at least one fledgling).

| Data analysis
All capture-recapture datasets were analyzed jointly with a Cormack-Jolly-Seber (CJS) model using the state-space likelihood fitted in the Bayesian framework (Kéry & Schaub, 2012). We fitted two models to the data: the first model allowed for sex-specific senescence rates while the second model assumes a common senescence rate and a sex-specific intercept. Parameters directly estimated by the model were φ i,j,t , the apparent survival probability of individual i in population j in year t, and p i,j,t , the recapture probability (Lebreton, Burnham, Clobert, & Anderson, 1992). We applied the following linear model for the survival probabilities of the known- Here, γ is the mean logit survival of females, β p is the difference of survival between sexes in population p, and pop j and ind i are the same random effects (variability among populations and variability among individuals) as for the known-aged individuals (Equation 1). The population-specific difference in survival between sexes was modeled with a further random effect as where d is the mean difference and 2 d the variability of the differences among populations. The recapture probabilities for both the individuals of known and of unknown age were modeled as log it p i,j,t = j,sex i + p t,j , where p,s is the mean logit recapture probability of individuals in population p and of sex s, and p t,j ∼ N 0, 2 p . 2 p is the population-specific temporal variability in recapture probability.
The second model that was fitted to the same data assumed that the age effect on survival is the same in both sexes; thus, there is no interaction anymore between age and sex. Consequently, Equation 1 is replaced by: where ′ a is the annual logit survival of individuals of age a. The other parameters are the same as explained above, and also, the model parts for the individuals of unknown age and for the recapture probabilities are the same as explained above. The tow analyses Monitoring period 2010-2014 2002-2014 2001-2016 1983-1993 1979-1984 1990-1994 1989-1993 Breeding success (number of broods) were restricted to individuals that survived to adulthood (age 1) treating the first breeding observation as the year of initial marking (Pradel, Hines, Lebreton, & Nichols, 1997). By doing this, we excluded the survival estimates from fledging to age 1, which include high rates of natal dispersal and preclude a direct comparison with survival estimated at older age (Weatherhead & Forbes, 1994).
Breeding success of individuals of known age was analyzed using the following generalized linear mixed model.  (Plummer, 2003). We used vague normal priors for the regression coefficients and uniform priors for the standard deviations of the random effects. Posterior summaries from three MCMC chains were based on 100,000 iterations after a burn-in of 20,000 and a thinning of 10. We confirmed parameter convergence using the Gelman-Rubin statistic. All R-hat values were below 1.1 indicating convergence of the Markov chains.
We report posterior means, their associated 95% credible intervals and the probabilities that the estimates were higher or lower than zero. The goodness of fit of the capture-recapture models was assessed with the program U-CARE (Choquet, Lebreton, Gimenez, Reboulet, & Pradel, 2009). Results are presented in Appendix S1.

| RE SULTS
The average recapture probabilities were lower in females (.49)

TA B L E 2
Posterior means and 95% credible intervals of the slopes of the regression of breeding success and survival against age in the whinchat. Given are also the probabilities that the estimates are positive. Sample sizes are given on Figures 1-3

Mean (slope) 95% CRI p(slope)<0
Breeding success .07], p < 0 = .81). The estimated decline in survival from age 1 to 5 was roughly 10% per year (Figure 1). When estimating sex-specific actuarial senescence, the survival decline seemed clearer in males than in females (p(slope)<0 = .86 and .68 respectively, Figure 3), but this may be due to the lower sample size in females. Available data provided weak evidence for sex-specific actuarial senescence.  Figure 1) nor when the sex was accounting for (p < 0 = .42 and 62, Figure 2). These results suggest that, in contrast to survival, breeding success remained constant across the life span in both sexes. Survival decreased monotonously from age 1 to 5 without evidence of a difference between sexes in the whinchat. Such a continuous decline of survival matches the average-age survival trajectory described in song sparrows (Melospiza media) for both sexes (Keller, Reid, & Arcese, 2008). Characteristically, survival in short-lived passerines decreases continuously after a maximum value that is reached early in life (Balbontín & Møller, 2015;Bouwhuis et al., 2012;Orell & Belda, 2002;Sendecka, 2007;Sternberg, 1989). The rapid decline of survival in passerines is in line with our evolutionary understanding of senescence in the framework of life-history theory.

| D ISCUSS I ON
In accordance with predictions, senescence has been shown to occur proportionally earlier and stronger in species with increasing speed of life (Jones et al., 2008;Péron, Gimenez, Charmantier, Gaillard, & Crochet, 2010).
Whereas higher natural mortality rates in females are widely expected in passerines (Donald, 2007), there is only limited evidence for sex differences in actuarial senescence rates from the wild. The few studies investigating survival senescence in both sexes of passerines provide generally no evidence of differences between sexes (Balbontín & Møller, 2015;Hammers et al., 2013;Sendecka, 2007;Sternberg, 1989). However, when a sex effect was present, senescence was stronger in females (Orell & Belda, 2002). The difference between theory and empirical studies could reflect the overestimation of the expected female mortality due to the confounding effect of dispersal (Dale, 2001). Alternatively it could also reflect the difficulty to detect small sex effects in senescence due to low statistical power (Nussey et al., 2008). The absence of evidence is not evidence of absence (Altman & Bland, 1995). In our case the modest sample sizes especially for females did not allow for a definitive conclusion.
Surprisingly, we found no support for reproductive senescence.
This is unexpected because reproductive decline at old age seems to be ubiquitous in animal populations (Lemaître & Gaillard, 2017;Nussey et al., 2013). Although the onset of reproductive senescence may be slightly delayed by the progressive improvement of competence in early life, studies examining survival and reproductive senescence in passerine observed generally a decrease in both fitness components (Bouwhuis et al., 2012;Keller et al., 2008;Sendecka, 2007 (Martin & Li, 1992). The whinchat as a ground-breeding species in open landscapes is exposed to a high predation risk and other environmental perturbations making breeding success fairly random. This may be accentuated in agricultural landscape by mowing activities, which lead to nest destruction. This high stochasticity may impede the detection of an age-related decline.
Interestingly, most previous studies documenting reproductive senescence in passerines have been carried out in populations breeding in nest boxes (tits, flycatchers, sparrows, swallows). Nest boxes are known to positively affect the breeding success because the exposition to predators and weather is reduced (Fay, Michler, Laesser, & Schaub, 2019;Møller, 1989 As in most demographic studies, survival estimated from capture-recapture data includes permanent emigration. Although dispersal is much more important at the juvenile stage, adults, and especially adult females, may still disperse substantially (Clarke, Saether, Røskaft, Saether, & Roskaft, 1997). Thus, the higher survival of whinchat males may partly be due to sex-specific differences in dispersal (Bastian, 1992). This raises the question whether the detected senescence effect on apparent survival is in fact caused by increasing dispersal. To the best of our knowledge, increasing dispersal with increasing age has never been documented in passerines. It seems more reasonable to assume that breeding dispersal is age-independent, and the decline in apparent survival is a consequence of senescence in true survival. Consistently, the strongly philopatric whinchat males show survival estimates that clearly decrease with age, similarly with the average pattern reported in Figure 1.
To conclude, this study contributes to our understanding of senescence in natural populations showing for the first time evidence of actuarial senescence in the whinchat. Surprisingly we found no evidence of reproductive senescence and speculate whether stochastic factor such as predation may have contributed to this result.
We hope that this study will stimulate more empirical studies in a wider range of species.

ACK N OWLED G M ENTS
We are grateful to all people that have participated to the monitoring of these whinchat populations. We thank Damijan Denac and Urša Koce for their contribution during the fieldwork in Slovenia.

CO N FLI C T O F I NTE R E S T
None declared.