Migration shapes senescence in a long-lived bird

Significance

Our study demonstrates that migration strategies have a significant impact on senescence in a bird roaming the Mediterranean Sea, the Greater Flamingo. Residents have higher survival and reproduction performances early in life than migrants. However, residents possibly bear the cost of their early-life advantages in advanced age, exhibiting both accelerated reproductive and actuarial senescence. We show that migration has critical impacts on life trajectories, shaping the mortality and reproduction outcomes in one of the longest-lived and most charismatic birds from temperate regions.

Abstract

Each year, billions of animals migrate across the globe on diverse spatial and temporal scales. Migration behavior thus plays a fundamental role in the life cycle and Darwinian fitness of many organisms. While the influence of migration on early-life survival and reproduction is well documented, its effects on senescence (aging) in advanced age remain largely unexplored. Using a unique 44-y ring-resighting dataset from a long-lived, partially migratory bird species, the Greater Flamingo (Phoenicopterus roseus), we demonstrate that migration plays a key role in shaping age-specific trajectories of mortality and reproduction. Resident flamingos exhibit higher early-life demographic performances, with lower baseline mortality than migrants, resulting in longer adult lifespan. Residents also have a higher probability of breeding than migrants, though their breeding success is similar. However, residents seem to pay for their early-life advantages in old age, experiencing accelerated actuarial and reproductive senescence compared to migrants. Overall, our study highlights the critical impact of migration on survival and reproduction throughout life, thereby illustrating the role played by behavioral decisions in the biology of aging in long-lived vertebrates.

Every year, billions of animals undertake migrations—cyclical movements between different habitats—across a wide range of spatial and temporal scales (1, 2). Migration allows animals to combine the use of different, geographically distant resources that are essential for their survival, growth, and reproduction (1, 3). In many organisms, migration is partial—namely facultative—and depends on individual decisions that have profound consequences for fitness (4, 5). Partial-migration systems offer a unique opportunity to investigate the evolutionary implications of migration as they allow for comparisons of mortality and reproduction outcomes among residents and migrants within the same population. To date, however, our understanding of the impacts of migration on the two main components of fitness—survival and reproduction—remains incomplete, particularly because previous studies either did not consider the effect of age (e.g., 6, 7) or lacked the statistical power to quantify it accurately (8). Yet, migration could have critical effects on the demography of age-structured populations by influencing a major life process: senescence.

Senescence is defined as the decline in survival and reproduction with increasing age—referred to as actuarial and reproductive senescence, respectively (9–11). According to evolutionary theories, senescence arises from a reduction in the strength of natural selection with age (11). Medawar (12) proposed that this weakening of selection leads to the accumulation of deleterious mutations with late-life effects, resulting in senescence. Alternatively, Williams (13) suggested that senescence is driven by alleles with antagonistic pleiotropic effects, which enhance early-life fitness at the cost of performance in later life (13, 14). Kirkwood later extended Williams’ theory by proposing a hypothesis based on the life history framework, suggesting that senescence results from a resource allocation trade-off between survival mechanisms and reproduction (15). According to this hypothesis, shifts in resource allocation toward reproduction occur at the expense of somatic maintenance, leading to the accumulation of unrepaired physiological damage over time and accelerating the process of senescence. Together, these theories provide a well-established conceptual framework for predicting how migration may influence senescence.

Migratory birds are excellent biological models for investigating the impact of migration on senescence, a topic that presents a major challenge in terms of demographic data. Population studies conducted on a few partially migratory species over nearly half a century (e.g., 16, 17) provide the necessary individual longitudinal data to address this issue. Furthermore, current knowledge about bird migration and its demographic impacts, combined with the principles of evolutionary theories of senescence, allows for the introduction of a working hypothesis, which we have termed “early benefit to residents.” In birds, migration is often considered as the least optimal strategy [“best of a bad job,” (18)] adopted by small-bodied or socially subordinate individuals to reduce the deleterious effects of competition and climate on winter survival (19, 20). Migrants usually have lower mean adult survival and breeding outcomes than residents (7, 21–23). Accordingly, based on the key concept of early-late life trade-offs from senescence theories, three predictions can be made regarding the “early benefit to residents” hypothesis: 1) Residents are expected to have higher fitness early in life, with greater adult survival and reproduction compared to migrants. 2) Higher adult survival in early adulthood should lead to a longer lifespan for residents than for migrants. 3) The increased reproduction of residents at the start of adulthood is expected to come at the cost of accelerated actuarial and reproductive senescence (e.g., reduced breeding probability and breeding success) later in life.

In this study, we investigate whether and how migration is associated with actuarial and reproductive senescence, as well as adult lifespan, in a long-lived colonial bird with partial migration, the Greater Flamingo (Phoenicopterus roseus) (Fig. 1A). We test the “early benefits to residents” hypothesis (Fig. 1B) and evaluate the associated predictions using a long-term dataset spanning 44 y, which includes 1,840 ringed flamingos resighted across the Mediterranean basin (Fig. 1 C and D). To analyze how individual migration strategy (migrant, resident, or mixed) influences age-specific survival and reproduction patterns, we employ two complementary approaches. First, we build different mortality models (Gompertz, Weibull, exponential, and logistic) to generate age-specific mortality curves and assess the relationship between migration strategies and three key mortality parameters: basal adult mortality (the mortality rate in young adults), actuarial senescence rate (the increase in mortality with age), and adult lifespan. Next, we use Bayesian ring-resighting multievent models to examine how migration strategies affect age-dependent reproduction and reproductive senescence, while accounting for imperfect detection in the field. We especially consider age-specific breeding (i.e., laying an egg) probability and success (i.e., producing a fledgling).

Fig. 1.

Migration strategies and monitoring of Greater flamingo. (A) Schematic migratory movements between wintering and nesting sites in the three migration strategies (resident, migrant, and mixed). (B) Hypothesis “early benefits to residents”: predictions about the effects of migration strategies on demographic parameters. (C) Map of nesting and wintering sites used by the Greater flamingo. (D) Pictures of Greater flamingo in Camargue © Hellio & Van Ingen. The two pictures on the right show the capture and ringing operations © Tour du Valat.

Results

Migration Strategies and Age-Dependent Mortality.

Migration strategies have a strong effect on the age-dependent mortality patterns of Greater flamingos (Fig. 2 A–C), which are best described by Gompertz models (SI Appendix, Tables S1–S5). The model fit tends to decline at the oldest ages (beyond age 45), particularly among migrants and individuals with a mixed strategy. The mean predictions of the mortality rate, µ(x), diverge from the observed values, m(x), which exhibit a tendency toward deceleration—although m(x) remain within the credibility interval of the predicted mortality curve (Fig. 2 A–C).

Fig. 2.

Age-dependent mortality estimates from Gompertz models built for three different migration strategies of Greater Flamingos. Individuals can be migrants, residents, or with a mixed migratory strategy. The mortality curves from Gompertz models associated with the three migratory strategies are presented on panel A–C—the death rates m(x) (gray dots) are also presented. The posterior distribution of Gompertz model parameters are provided (D–F): basal mortality rate ($b_0$ noted in the R package BaSTA; F), actuarial senescence rate ($b_1$; E), and actuarial senescence rate corrected for lifespan (Std senescence rate; D) which was calculated by multiplying $b_1$ by life expectancy. Three adult lifespan metrics (expressed in years) were calculated using the posterior distributions of Gompertz model parameters: life expectancy, i.e., the average age at death (G); lifespan 50% (H) and 80% (I), i.e., the age at which 50% and 80% of the individuals alive at adulthood are dead. Finally, we investigated a potential trade-off between early- and late-life survival across migration strategies by examining the negative correlation between basal mortality rates $b_0$ and senescence rates $b_1$ (J)—this negative relationship is known as the “Strehler–Mildvan correlation.”

Consistent with the “early benefits to residents” hypothesis, residents show lower basal mortality rate—i.e., at the beginning of adulthood—than migrants (Fig. 2F). Differences in mortality parameters between migration strategies are evaluated using calibrated Kullback–Leibler divergences [q(k), SI Appendix, Table S6], where a q(k) value greater than 0.85 indicates a significant difference between two posterior distributions, and a value of 1 means that there is no overlap between them (24–26). The posterior distributions of basal mortality rate do not show any overlap: q(k) values are always 1. Moreover, the basal mortality rate of residents and individuals with a mixed strategy is relatively similar, with a negligible overlap of the posterior distributions of the basal mortality rates [q1(k) = 0.843 and q2(k) = 1.00].

Mortality also increases with age in the three migration strategies (Fig. 2 A–C), revealing actuarial senescence. Consistent with the “early benefits to residents” hypothesis, the actuarial senescence rate ($b_1$) is higher in residents than in migrants [Fig. 2E; see q(k) values in SI Appendix, Table S6]. This pattern is even more pronounced when the senescence rate is corrected for lifespan (i.e., standardized senescence rate; Fig. 2D). The standardized senescence rate is 1.4 times lower in migrants than in residents, with a negligible overlap of the posterior distributions [q1(k) = 1.00; q2(k) = 0.98]. Furthermore, individuals with a mixed strategy have the highest standardized senescence rate (Fig. 2 D-E): the posterior distributions of their standardized senescence rate slightly overlap that of residents [q1(k) = 1.00; q2(k) = 0.90], but do not overlap that of migrants [q1(k) = 1.00; q2(k) = 1.00].

The differences in mortality parameters among migration strategies result in significant changes in adult lifespan (Fig. 2 G–I and SI Appendix, Table S6). Consistent with the “early benefits to residents” hypothesis, residents have the longest lifespan (both 50% and 80%; Fig. 2 H and I) and life expectancy (Fig. 2G) while migrants have the shortest ones—the posterior distribution of the different longevity metrics never overlap for these two groups of individuals [q(k) is almost always 1]. Migrant life expectancy is 6.7 y lower than that of residents. Moreover, individuals with a mixed strategy have intermediate lifespan and life expectancy, whose distributions partially overlap those of migrants and residents (SI Appendix, Table S6).

Using simulations based on estimates from the Gompertz models, we identify a negative relationship between the senescence rate and the baseline mortality rate across all three migration strategies (Fig. 2J). The correlation observed for migrants differs from that of residents and individuals with a mixed strategy (Fig. 2J), suggesting distinct mortality regimes among migration strategies. Overall, these results could reflect a trade-off between early-life and late-life survival performance, as predicted by the “early benefits to residents” hypothesis, but should be interpreted with caution given the mathematical properties of the Gompertz model (see Discussion section).

Migration Strategies and Age-Dependent Reproduction.

Migration strategies have strong influence on age-dependent reproduction trajectories of Greater flamingos (Fig. 3 A–F), according to q(k) values (SI Appendix, Table S7) and model WAIC (SI Appendix, Table S11). Breeding probability increases with age early in life and decreases afterward in the three migration strategies (Fig. 3 A–C), indicating reproductive senescence. Consistent with the “early benefits to residents” hypothesis, the breeding probability is higher in residents than in migrants and in individuals with a mixed strategy (Fig. 3 A–C), up until the onset of senescence—the age at which breeding probability is maximal and then decreases. At the reproductive senescence onset, the breeding probability of residents is 0.37 (0.37, Fig. 3D) while it is 0.28 and 0.30 in migrants and individuals with a mixed strategy, respectively—q(k) is almost always 1 (SI Appendix, Table S7). However, the onset of reproductive senescence is slightly earlier in residents (20.44 y, Fig. 3E) than in migrants (21.90 y) and individuals with a mixed strategy (21.65 y), with a low overlap of the posterior distributions [residents vs. migrants: q1(k) = 0.954; q2(k) = 1.00; residents vs. individuals with a mixed strategy: q1(k) = 0.955; q2(k) = 1.00]. The age at which 90% of the individuals ceased to breed (Fig. 3F) is relatively similar in residents (36.77 y) and migrants (36.82 y) and is a little later in individuals with a mixed strategy (38.04 y) although the overlap between posterior distributions is broad (SI Appendix, Table S7).

Fig. 3.

Age-dependent reproduction estimates (breeding probability, A–F; breeding success probability, G–L) from a multievent model for three different migration strategies of Greater flamingos. The curves of breeding and breeding success probabilities associated with the three migratory strategies are presented on panel A–C and G–I, respectively. The posterior distributions of the following parameters are provided: breeding probability at the onset of senescence (D), age at the onset of senescence (i.e., the age at which breeding probability was maximal and decreased afterward) (E), age at which 90% individuals stopped breeding (F), breeding success probability at first reproduction (J), breeding success probability at the onset of senescence (i.e., values provided on panel E) (K), and breeding success probability late in life (i.e., age at which 90% individuals stopped breeding; values provided on panel F) (L).

When reproduction is initiated, breeding success probability decreases with age in the three migration strategies [Fig. 3 G–I; see the q(k) values in SI Appendix, Table S7], once again indicating reproductive senescence. We find mixed support for the “early benefits to residents” hypothesis when examining breeding success. At first reproduction, residents (0.52) and migrants (0.56) have relatively similar breeding success (Fig. 3J), with a broad overlap of the posterior distributions [residents vs. migrants: q1(k) = 0.770; q2(k) = 1.00]. The individuals with a mixed strategy have lowest breeding success (0.42), with a low overlap of the posterior distribution of breeding success probabilities [migrants vs. individuals with a mixed strategy: q1(k) = 1.00; q2(k) = 0.987; residents vs. individuals with a mixed strategy: q1(k) = 0.976; q2(k) = 1.00]. Later in life, breeding success is relatively similar among individuals of the three migration strategies once they reach their respective maximum probability of breeding (Fig. 3K; SI Appendix, Table S7). At the end of their reproductive lifetime, residents show the lowest breeding success probability (Fig. 3L), consistent with the “early benefits to residents” hypothesis. In the last ten percent of individuals reproducing at an old age, breeding success is lower in residents (0.38 at an age of 36.77 y) than in migrants (0.48, 36.82 y) and individuals with a mixed strategy (0.45, 38.04 y)—although the credible intervals are broad in the latter (Fig. 3L and SI Appendix, Table S7).

Discussion

Using an exceptional 44-y dataset, our results provide evidence of the critical role migration strategies play in shaping patterns of actuarial and reproductive senescence. Overall, we find strong support for the “early benefits to residents” hypothesis: Residents exhibit higher early-life demographic performances, with lower basal mortality and higher breeding probability than migrants. Lower mortality among young adults also allows residents to achieve longer adult lifespan compared to migrants. However, residents possibly bear the cost of their early-life advantages in advanced age, exhibiting both accelerated reproductive and actuarial senescence relative to migrants. Our study provides one of the few pieces of evidence of the demographic impacts that migration can have on senescence in a long-lived vertebrate.

Early-Life Adult Mortality and Migration Strategies.

We find that residents and individuals with a mixed strategy have lower basal mortality rate than migrants at the beginning of adulthood. This result is congruent with the idea that migration is a “best of a bad job” strategy used by low-quality individuals to reduce the deleterious effects of competition or environment on winter survival (19, 20). It is also in accordance with previous studies that have reported higher mean adult survival of residents versus migrants or short-distance versus long-distance migrant in long-lived bird species (23, 27–29). Higher basal mortality rate in migrants may result from the higher energy expenditures caused by migration movements. Migratory flights of hundreds of kilometers require a large amount of energy to face the variety of environmental conditions encountered during the flight but also to simply be able to resume the whole migratory flight (21, 30–32). In addition, higher mortality of young migrants may result from the accumulation of risks encountered on the flyway (33, 34). Dry wetlands, harsh weather conditions, and destruction of historical stopover sites are adverse events that can be met all along the migratory flight and might cause higher mortality in young adults with limited experience (35).

Lower basal mortality rate in individuals with resident and mixed migratory strategies might also be facilitated by specific management practices in northern wintering sites. In Camargue, for almost 40 y, baiting for ducks and waterbirds within hunting estates and in a large public ornithological park has been a common practice during winter either for hunting or for ecotourism (36). Thus, flamingos are often seen feeding on these seeds or broken rice during winter. Previous work has suggested that these baiting practices coupled with suitable winter conditions increase ducks body condition (36, 37). Hence, it is possible that resident flamingos also benefit from this food supply, as well as from the increasingly benign ambient winter conditions in southern France due to climate warming, to better survive than migrants (38). However, rare cold spells may have a considerable negative impact on the survival of wintering flamingos (38), which may explain the occasional advantage of the migratory strategy and the maintenance of both strategies in the population in the long term (16). In sum, our study clearly shows that migration strategies have a marked influence on the survival of young adults: Residency and partial migration provide an early-life survival benefit compared to full migration, which leads to an increase of 26 and 21% in adult life expectancy in individuals with resident and mixed migratory strategies, respectively.

Early-Life Reproduction and Migration Strategies.

Our results show that residents have higher breeding probability early in life than migrants and individuals with a mixed strategy. Nevertheless, when reproduction is initiated, residents and migrants have fairly similar breeding success, at least until the onset of reproductive senescence (36 to 38 y). Breeding success of individuals with a mixed strategy is lowest at the first reproduction and increases afterward. At the onset of reproductive senescence, breeding success is similar among the three migration strategies. Overall, our results are congruent with the findings of previous studies highlighting higher breeding performances in resident birds (22), in accordance with the concept of migration as a “best of a bad job” strategy (18).

In our study system, the difference of breeding probability between residents and migrants may be caused by two processes. First, the lower breeding probability of migrants may result from constraints linked to late arrival dates on the breeding colony. While residents arrive earlier on the breeding colony (because wintering in the surrounding areas), migratory birds arrive on average 6 d later. The consequence of this later arrival is that migrants may occupy nesting sites of poorer quality on the breeding island where settlement has been found to follow a despotic distribution (39, 40). In colonial breeding birds, the nest place on the breeding colony largely influences breeding probability, central places often benefiting from less disturbance or less predation risk (41, 42). Second, the higher reproductive success of residents could be facilitated by the accumulation of breeding experience throughout life. Previous studies on flamingos have shown that past breeding experiences influence the shape of age-specific breeding probability trajectories (43). Residents, which reproduce more frequently early in life compared to migrants and individuals with a mixed strategy, may benefit from a cumulative positive effect of past breeding experiences. This process could contribute to higher breeding probabilities in residents up until the onset of senescence.

Senescence and Migration Strategies.

Our analyses reveal that flamingos exhibit an actuarial senescence rate (0.06 to 0.09) comparable to that of other long-lived bird species [e.g., Larus canus, 0.07; Sterna hirundo, 0.08; Gyps fulvus, 0.08; (44)]. After accounting for differences in lifespan, this rate is found to be 44% higher in residents than in migrants, consistent with the “early benefits to residents” hypothesis. Interestingly, the fit of the Gompertz model tends to deteriorate at older ages, particularly in migrants. The observed death rate m(x) initially follows the exponential trajectory predicted by the Gompertz model, before showing a deceleration later in life. Observed in humans and various animal species (45), mortality deceleration can be explained by two alternative hypotheses (46). According to the “heterogeneity hypothesis,” deceleration is a statistical artifact resulting from selective survival. Because more frail individuals tend to die earlier, the individuals who survive to older ages are generally in better health and physiological condition, leading to a decrease in the average mortality rate later in life (46, 47). In flamingos, our results show that migration leads to high mortality in young adults, which may filter out the most vulnerable individuals early in life, potentially causing a deceleration in mortality at older ages. The development of frailty models (48, 49), which account for among-individual heterogeneity, could help test this hypothesis in flamingos, provided a sufficiently large number of individuals are available in the oldest age classes. Alternatively, according to the “individual-risk hypothesis”, the deceleration in mortality could also result from physiological processes at the individual level. For example, a slowing of the “rate of living” with age—linked to the regulation of cellular and molecular processes—and age-specific changes in resource allocation patterns could lead to a late-life mortality deceleration (46).

We also find that the rate of actuarial senescence decreases with baseline mortality and that the form of this correlation differs between residents and individuals with a mixed strategy, on the one hand, and migrants, on the other. This negative relationship, known as the Strehler–Mildvan correlation, has long been viewed as the result of a trade-off between survival performance at early and late life stages—individuals with high vitality early in life experiencing faster senescence (50, 51). In addition, changes in the Strehler–Mildvan correlation were also interpreted as resulting from variations in the regimes of mortality selection (51). However, simulation studies have highlighted the potentially artifactual nature of this correlation, which arises at least partially from the mathematical properties of the Gompertz model (52, 53). As noted by Burger and Missov (52), interpreting the Strehler–Mildvan correlation requires the use of other measures of late-life condition or nondemographic indicators of senescence. Our analyses already suggest that residents exhibit higher vitality early in life and a steeper decline in late-life performance compared to migrants, not only in survival but also in reproduction. However, physiological measures of aging would be needed to provide more formal evidence for the existence of the early–late life trade-off predicted by the “early benefits to residents” hypothesis, beyond the potential mathematical artifact of the Strehler–Mildvan correlation.

Importantly, we detect a clear signal of reproductive senescence in flamingos, with residents exhibiting a stronger age-specific decline in breeding performance compared to migrants and individuals with a mixed strategy—consistent with the “early benefits to residents” hypothesis. By refining age-specific reproductive estimates from a previous model (43), our analysis highlights that flamingos exhibit a late onset of reproductive senescence similar to patterns reported in other long-lived bird species such as albatrosses (54). This decline in breeding probability begins after age 20 across all three migration strategies—starting at 20 y in residents and around 22 y in migrants and individuals with a mixed strategy. Interestingly, breeding probability shows a more pronounced senescence signal than breeding success, aligning with meta-analytic evidence that maternal age has only a moderate effect on offspring survival in birds (55). However, despite this weaker signal, we find that residents experienced a 20% decline in reproductive success at the end of their reproductive lifetime compared to migrants—although the credible intervals were broad.

The demographic patterns of actuarial and reproductive senescence highlighted in our study are likely underpinned by physiological aging processes that are still poorly documented in flamingos. One study (56) showed notably that, in captivity, resistance to oxidative stress increases between ages 0 and 15, reaches a plateau between 16 and 25, and then declines between 25 and 45 y. Interestingly, these age-specific changes in oxidative stress resistance are closely synchronized with the acceleration of mortality and the decline in reproductive performance, both of which begin between 20 and 25 y of age, depending on the migration strategy considered. Oxidative stress resistance, which is influenced by environmental factors (e.g., stress, diet) and decreases with reproductive effort in birds (57–59), may thus represent a proximate mechanism involved in the variation in senescence across migration strategies in flamingos. A more rapid accumulation of oxidative damages caused by high reproductive effort early in life could lead to accelerated senescence in residents. By contrast, in migrants, less frequent reproduction could slow down the process of degradation of the physiological machinery, thus decreasing the intensity of actuarial senescence and maintaining reproductive functions—and therefore reproductive success—for longer.

Conclusion.

Our findings add to the growing body of research documenting intraspecific variation in senescence and its causes across vertebrates (60–63). Using one of the most long-lived and charismatic birds of temperate regions as a biological model, we uncover the significant role migration can play in shaping survival and reproductive trajectories. We find strong support for the “early benefits to residents” hypothesis: Residents show higher fitness at the beginning of adulthood but likely bear the costs of these early-life advantages later, exhibiting accelerated senescence compared to migrants. Although the proximate mechanisms [“hallmarks of aging,” (64)] driving senescence differences between residents and migrants remain largely unknown, we highlight the crucial impact of individual decisions and behavioral movement on aging in the wild.

Materials and Methods

Species and Collection of Long-Term Individual Data.

The Greater flamingo is a long-lived colonial waterbird species with records of individuals older than 45 y in the wild (65). Flamingos live in coastal brackish wetlands or salt lakes all across the Mediterranean region and western Africa, where they feed on aquatic invertebrates or seeds (65, 66). Flamingos often breed in colonies of several thousands of individuals settled on isolated islands, preventing disturbance or terrestrial predation. Flamingos can breed from the age of three, although access to reproduction usually occurs later on between 4 and 10 y (43, 67). This species gives birth to only one chick per breeding. Flamingos have bred almost every year since 1977 at two nearby sites in the Camargue, southern France (Fig. 1C): the Fangassier lagoon (43°25’39.8”N 4°37’25.9”E, Salin-de-Giraud) and the Roi lagoon (43°31’18.8”N 4°12’29.8”E, Aigues-Mortes). The size of the colony is usually around 10 000 breeding pairs.

Since 1977, with some exceptions (2002, 2007, 2016, and 2019), 7 to 30% of the fledglings reared each year in Camargue have been marked individually with PVC rings (Fig. 1D) that can be read at a distance up to 300 m with a telescope (65). As flamingos were ringed as fledglings, their age was perfectly known. Our initial dataset consisted of 27,519 individuals for which ring resighting was performed both during the breeding season (March to September) and in winter (November to January) between 1977 and 2020. We then considered a subset of this dataset comprising individuals for which migratory strategies could be reliably assessed. We selected individuals resighted at least once at the Camargue breeding colony during the breeding period (reduction of sample size from 27,519 to 13,413 individuals), observed at least during three different winters (13,413 to 4,041) and reproducing in the Camargue more than 95% of the breeding occasions (4,041 to 1,840 individuals). Within this subset, breeding attempts ranged from 1 to 23 depending on the individuals.

The definition of migration strategies was based on the flamingo movement behavior ecology (Fig. 1A). Flamingos overwinter all over the Mediterranean basin and can be seen in wetlands from Türkiye to Southern Spain and North Africa [Fig. 1C; (16, 65)]. However, the most important sites for the wintering of French individuals are southern Spain, Italy, and North Africa (Fig. 1C). As flamingos exhibit a relatively high fidelity to their wintering sites after their first winter [>90%; (16, 68)], we categorized individuals into three migration strategies. Residents were defined as individuals for which > 90% of winter resighting events occurred in continental France (n = 1357), while migrants were those for which >90% of winter resighting events occurred outside France (n = 237; see Fig. 1C for the distribution of resighting locations of ringed flamingos). Individuals with a mixed migratory strategy are those for which none of the two previous strategies predominate (n = 246). Migration strategy was attributed to each individual and treated as a fixed categorical individual covariate.

Analyzing Age-Dependent Mortality Patterns.

The associations between migration strategies and age-specific mortality were analyzed using the R package “BaSTA” (69, 70). BaSTA allowed us to consider imperfect detection and right-censored (i.e., unknown death date) ring-resighting data in our analysis. We analyzed the three groups of individuals (residents, migrants, and mixed strategy) separately to identify the models best describing mortality patterns for each migratory strategy. From BaSTA models, we also estimated adult lifespan for the different migratory strategies using three metrics: “lifespan 50%” and “lifespan 80%,” defined as the age at which 50% and 80% of the individuals alive at the onset of adulthood (4 y) are dead, respectively; and “life expectancy” from the onset of adulthood, defined as the average adult age at death. We calculated calibrated Kullback–Leibler divergences (24, 25) to quantify how mortality parameters differed among migratory strategies (SI Appendix, Table S6)—q(k) ranges from 0.5 to 1, where a value of 0.5 means that posterior densities are identical and 1 that there is no overlap between them. The models and related analyses are described in detail in SI Appendix, section 1.1.

Analyzing Age-Dependent Reproduction Patterns.

A multievent ring-resighting model (71) was used to analyze age-dependent breeding and breeding success probabilities. This approach allowed us to model individual breeding state and success, which cannot be assess with certainty due to imperfect detection on the field. The model, implemented using a Bayesian state space formulation (72, 73), distinguishes initial state probability (i.e., at first capture), state-transition probabilities, and the event probabilities (i.e., field observations) conditional on the underlying states (71). Individual heterogeneity was also considered through the inclusion of individual random effects (49). Values of q(k) were calculated to compare the posterior distribution of reproductive parameters (SI Appendix, Table S7). The model structure and related analyses are described in detail in SI Appendix, section 1.2.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (CSV)

Dataset S02 (CSV)

Dataset S03 (CSV)

Code S01 (TXT)

Code S02 (TXT)

Code S03 (TXT)

Data, Materials, and Software Availability

All study data and code are available from Github: https://github.com/sebroques/Flamingo_ageing_code (74).