RESEARCH ARTICLE | ECOLOGY

Behavioral plasticity shapes population aging patterns

in a long-­lived avian scavenger
Marta Acácioa,1 , Kaija Gahmb, Nili Anglistera , Gideon Vaadiaa, Ohad Hatzofec, Roi Hareld , Ron Efrate , Ran Nathanf , Noa Pinter-­Wollmanb ,
and Orr Spiegela,1

Affiliations are included on p. 8.

Edited by Nils Stenseth, Universitetet i Oslo, Oslo, Norway; received April 11, 2024; accepted July 13, 2024

Studying the mechanisms shaping age-­related changes in behavior (“behavioral aging”) is

important for understanding population dynamics in our changing world. Yet, studies that Significance
capture within-­individual behavioral changes in wild populations of long-­lived animals are
still scarce. Here, we used a 15-­y GPS-­tracking dataset of a social obligate scavenger, the Aging is a universal
griffon vulture (Gyps fulvus), to investigate age-­related changes in movement and social phenomenon, yet how behavior
behaviors, and disentangle the role of behavioral plasticity and selective disappearance in changes with age (“behavioral
shaping such patterns. We tracked 142 individuals for up to 12 y and found a nonlinear aging”) and the mechanisms that
increase in site fidelity with age: a sharp increase in site fidelity before sexual maturity shape behavioral aging (plasticity
(<5 y old), stabilization during adulthood (6 to 15 y), and a further increase at old age and/or selective disappearance)
(>15 y). This pattern resulted from individuals changing behavior throughout their life are still poorly understood. We
(behavioral plasticity) and not from selective disappearance. Mature vultures increased
individually tracked griffon
the predictability of their movement routines and spent more nights at the most popular
vultures for up to 12 y and
roosting sites compared to younger individuals. Thus, adults likely have a competitive
advantage over younger conspecifics. These changes in site fidelity and movement routines tracked up to 60% of the
were mirrored in changes to social behavior. Older individuals interacted less with their population simultaneously. We
associates (decreasing average strength with age), particularly during the breeding season. found that older vultures display
Our results reveal a variety of behavioral aging patterns in long-­lived species and under- higher site fidelity, fixed
score the importance of behavioral plasticity in shaping such patterns. Comprehensive movement routines, and weaker
longitudinal studies are imperative for understanding how plasticity and selection shape social relationships compared to
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

the persistence of wild animal populations facing human-­induced environmental changes. younger conspecifics. Some of
these patterns were driven by
behavioral aging | behavioral plasticity | selective disappearance | movement ecology | sociality
individuals changing their
behavior throughout their lives
Aging is ubiquitous across living organisms, leading to behavioral changes throughout life
rather than by the selective
(1–4) (“behavioral aging”). Younger, sexually immature individuals tend to differ from
older, mature individuals in a number of behavioral traits, including how they move [e.g., disappearance of particular
flight performance (5)] and how they interact with conspecifics [e.g., strength of social phenotypes. Our results highlight
interactions (6, 7)]. Such behavioral changes may influence space use (3, 8), the spread the role of individual plasticity in
of infectious diseases (1), and even the lifespan of individuals (9). Behavioral changes with shaping behavioral aging, which
age may also play an important role in how populations adjust to environmental change: may have implications for our
Young individuals may adopt novel behaviors and be the agents of change (10, 11), while understanding of how population
old individuals, with their accumulated knowledge and experience, may adjust to the dynamics are impacted by a
environment by shifting behavioral strategies over their lifetimes (12–14). Despite the changing world.
importance of understanding behavioral aging in nature, most ecological studies focus on
binary comparisons between young and old animals, failing to track individuals through-
out their lives (15). Specifically, longitudinal studies that follow long-­lived animals
Preprint server: A previous version of this manuscript
throughout most of their lives are rare, mostly due to methodological constraints (15). has been deposited in EcoEvoRxiv, under the DOI https://
This gap hinders the identification of gradual and nonmonotonic behavioral changes in doi.org/10.32942/X22321

the wild or the mechanisms that underlie population-­level aging patterns. The authors declare no competing interest.

Research on behavioral aging reveals a spectrum of patterns at the population level This article is a PNAS Direct Submission.

(Fig. 1): Some behaviors remain fixed throughout life (16), while others change, either Copyright © 2024 the Author(s). Published by PNAS.
This article is distributed under Creative Commons
gradually (2), or drastically at specific ages [e.g., early (17, 18) or late in life (19, 20); the Attribution-­NonCommercial-­NoDerivatives License 4.0
latter usually associated with senescence and loss of physiological or physical capacities (CC BY-­NC-­ND).
(3, 21, 22), Fig. 1A]. Population-­level behavioral changes with age can arise from two, Although PNAS asks authors to adhere to United Nations
naming conventions for maps (https://www.un.org/
nonmutually exclusive mechanisms. First, individuals may change their behavior through- geospatial/mapsgeo), our policy is to publish maps as
out their lifetimes [behavioral plasticity (6, 23, 24)]. Second, individuals with particular provided by the authors.
behavioral phenotypes may have lower survival than others (25, 26). Across generations, 1
To whom correspondence may be addressed. Email:
[email protected] or [email protected].
the selective disappearance of these phenotypes that confer lower fitness can result in
This article contains supporting information online at
changes to the behavioral composition of the population with increasing age, without https://www.pnas.org/lookup/suppl/doi:10.1073/pnas.​
within-­individual behavioral plasticity (27, 28) (Fig. 1B). Ultimately, examining the pat- 2407298121/-­/DCSupplemental.
terns and mechanisms of behavioral aging provides a foundation for understanding how Published August 20, 2024.

PNAS 2024 Vol. 121 No. 35 e2407298121 https://doi.org/10.1073/pnas.2407298121 1 of 9

 A

Behaviour

Age

B Individual plasticity Individual plasticity Selective disappearance

(no individual variability) (with individual variability) (no individual plasticity)
Behaviour

Age

Fig. 1.   Theoretical relationships and potential mechanisms of behavioral aging. (A) Five theoretical relationships between age and behavior at the population
level: no relationship; a linear relationship (monotonic change in behavior with age); an exponential, convex, relationship (sharp change in behavior in old
individuals); a quadratic, saturating, relationship (drastic change before maturation and then a stabilization in behavior); and a third-­degree polynomial relationship
(different behavior at each life stage, for example, juveniles behaving differently from reproducing individuals, and old individuals behaving differently from the
first two age groups). Note that we depict an increasing relationship between age and behavior, but, depending on the behavior, a decreasing relationship is
possible too. The vertical dotted line indicates the age of maturation, and the different gray hues represent distinct life stages (e.g., juvenile, reproducing, old). (B)
Mechanisms that may explain population-­level relationships between age and behavior: individual plasticity without individual variability (all individuals behave
and change their behavior in the same way); individual plasticity with individual variability (individuals differ in their average behavior but change it similarly
with age); variation in behavior among individuals with no behavioral plasticity, instead, selective disappearance of individuals with a particular behavioral traits
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

leads to population-­level behavioral aging. These three mechanisms can apply to any of the patterns in A. Thick black lines are the population-­level relationship
between behavior and age and the thinner gray lines are individual-­level relationships between age and behavior.

populations might adjust to the environment, if through plasticity species can have potential implications for their effective conser-
or selection, and how changes in population age-­structure can vation. For instance, uncovering different spatial requirements by
influence ecological processes and the ability of a species to age may help focus conservation efforts on areas that preserve
respond to environmental changes (19). individuals of all age cohorts.
In this study, our main goal is to examine patterns of behavioral We hypothesized that site fidelity, movement routines, and social
aging in a wild animal population. We also aim to disentangle the behaviors will change with age either gradually or more sharply early
relative importance of individual plasticity and selective disap- or late in life (Fig. 1). We tested these hypotheses using a unique
pearance of individuals from the population in shaping observed GPS-­tracking dataset, of 319 griffons followed between 2008 and
aging patterns in movement and social behaviors. To do so, we 2022. To account for the differences in the GPS transmitters’ sam-
use a dataset of 15-­y dataset of GPS-­tracking of griffon vultures pling rate over the 15-­year sampling period (38), and to maximize
(Gyps fulvus hereafter griffons). The griffon is a long-­lived obligate the available data, we analyzed different measures of roosting behavior
scavenger: In captivity, griffons live beyond the age of 40 while in as a proxy for movement and social behavior [e.g., higher roost fidelity
the wild the survival rates decrease (actuarial senescence) after the representing higher site fidelity and higher co-­roosting strength rep-
age of 28 y (29). Griffons have a slow life cycle, reaching sexual resenting stronger social bonds (33)]. We built seasonal co-­roosting
maturity around the age of 5 and laying a single egg each year (if proximity-­based social networks during the breeding, summer, and
they breed at all). Griffons may perform seasonal long-­range forays transient seasons [when most long-­range forays occur (30)], using the
and movements (30, 31), particularly early in life (32). Griffons full dataset of roost locations, and for the years during which at least
use social cues while flying and foraging (33) and they sleep and 25% of the griffon population was tracked (39) (2016 to 2022).
nest in communal roosts (34) that act as information centers, Using a subset of 142 griffons, aged 0 to 24 y (Fig. 2B), that remained
where individuals gather information about the location of in the main study area and that were tracked for a minimum of 30 d
resources in the landscape (35). Their large body size and longevity and maximum of over 12 y (Fig. 2C), we examined how age affected
(which facilitate long-­term GPS-­tracking), as well as their high the likelihood of using the same roost-­site on consecutive nights
mobility and sociality, make the griffon vulture an ideal candidate (“roost fidelity”), and the predictability of roost switching sequences
for studying behavioral aging in the wild. Furthermore, most vul- [i.e., if vultures switched roosts in an ordered sequence (40)–“rou-
ture species are globally endangered or critically endangered (36). tine”]. We also evaluated how age influenced roost-­site selection (i.e.,
Likewise, Israeli griffons are regionally critically endangered and selecting popular roosts–“roost popularity”), as well as griffons’ social
are the target of an intensive conservation management program, relationships, measuring the number of co-­roosting partners and the
aimed at rehabilitating the population (37). Thus, understanding average number of nights co-­roosting with their social partners.
the patterns and mechanisms underlying behavioral aging in this Finally, we tested whether these population-­level behavioral changes

2 of 9 https://doi.org/10.1073/pnas.2407298121 pnas.org
 Fig. 2.   Study area and GPS-­tracking dataset. (A) Map showing GPS tracks of two griffon vultures in a single day: one individual remained at the same roost-­site
on consecutive nights (blue), and the other switched between roost sites on consecutive nights (red). The inset shows the location of the study area, as well as
the movements of griffon vultures traveling to East and Central Africa and to the Arabian Peninsula on long-­range forays. (B) Number of unique griffon vultures
tracked at each age. Individuals tracked over multiple years appear in multiple age cohorts. (C) Tracking duration (in years) of the 20 griffon vultures with
longest tracking duration since the time of their first GPS deployment. Gaps indicate periods when the individual was not tracked due to the loss or failure of
the GPS transmitter. For a similar figure of all individuals in the study, see SI Appendix, Fig. S7. In B and C, colors indicate the age of each individual, with younger
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

individuals represented in lighter colors.

arise from individuals changing their movement and social behaviors (e.g., A -­> B -­> C -­> A -­> B -­> C -­> A -­> …, Fig. 3B and SI Appendix,
as they age or through the selective disappearance of individuals with Tables S4 and S5). Furthermore, the strength of these roost-­switching
particular traits. We did so by disentangling the effects of within-­ routines [index of routine (40)] increased exponentially -­changing
individual changes and of individual longevity (age at death) on most drastically after the age of 15 y (Fig. 3C and SI Appendix,
behavioral aging. Tables S6 and S7). Due to the smaller sample size in this analysis,
we could not evaluate whether this pattern emerged from individual
Results and Discussion plasticity or selective disappearance. Thus, we could not determine
whether behavioral predictability provides a selective advantage over
Griffons Increase Site Fidelity and Movement Predictability the course of an individuals’ lifetime (individuals with stronger
with Age, Shaped by Individual Plasticity and Not by Selective movement routines live longer), or if as individuals mature, they
Disappearance. We found that griffons’ roost fidelity increased with learn which sites they prefer and subsequently tend to frequent those
age. Interestingly, this population-­level relationship took the form selected locations.
of a third-­degree polynomial. There was a rapid increase in roost Past research, from a single population in France, found that
fidelity until the age of 5 (griffons’ age of maturation), no change actuarial senescence (an increase in mortality rates) in griffon vul-
between the ages of 6 and 15 y, and then another substantial increase tures starts at 28 y (29). However, in the population studied here,
in roost fidelity after the age of 15, indicating that old vultures tend adult griffons have lower survival than in other populations [0.86
to return to the same roost each night (Fig. 3A and SI Appendix, apparent survival (41) vs 0.94 in the French population (29)],
Tables S1 and S2). This result was not determined by the long-­range most mortality is caused by human activities (42). Consequently,
forays performed predominantly by younger griffons, during which griffons in this population rarely live past the age of 24. While it
they might switch roosts more often than they would in our main is unlikely that the observed increase in site fidelity and predict-
study area (Fig. 2 and SI Appendix, Fig. S1). This population trend ability starting at 15 y old results from age-­related changes in
was mostly shaped by individual plasticity (individuals increasing cognitive or physical abilities, it is worth noting that in long-­lived
roost fidelity throughout their lives, SI Appendix, Fig. S2) and not by bird species phenotypic senescence may emerge well before actuarial
selective disappearance, despite the slight, nonsignificant indication senescence. For example, wild wandering albatrosses (Diomedea
that individuals with higher roost fidelity may live shorter lives exulans) may live to the age of 50 but show a decrease in survival
(Fig. 4 and SI Appendix, Table S3). rates after the age of 35 (43) and signs of reproductive senescence
In addition to higher roost fidelity, older griffons were also more after the age of 30 (3, 43).
predictable in their movement routines: When switching between It is plausible that the differences in movement behavior of
roosts, older griffons were more likely to follow an ordered sequence older griffons emerge from changes in their breeding duties.

PNAS 2024 Vol. 121 No. 35 e2407298121 https://doi.org/10.1073/pnas.2407298121 3 of 9

 1.00 A 1.00
B C

Probability of having a routine

Probability of roost fidelity

0.75 0.75

0.50 0.50

Season
0.25 0.25 Breeding
Summer
Transient
All year
0.00 0.00

1.00 D E F
F
Probability of occupying a popular roost

0.75
0.75
Normalized degree

0.50
0.50

0.25

0.25

0.00

0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
Age (years) Age (years) Age (years)

Fig. 3.   Age-­related changes in movement and social behavior of griffon vultures. Each panel A–F represents the model predictions, backtransformed to the
original data scale, and the 95% CI for the effect of age on movement (A–C) and social behaviors (D–F): (A) probability of remaining at the same roost-­site on
consecutive nights (roost fidelity); (B) probability of predictably switching between roosts in the same order (routine); (C) strength of roost-­switching routines
(index of routine); (D) probability of occupying a popular roost (defined as the top 20% most used roosts in the study area); (E) number of unique individuals
a vulture interacted with over a season, normalized to the size of the GPS-­tracked population (normalized degree); (F) average number of social interactions
over a season (average strength: strength divided by degree). Colors and line styles indicate the different seasons: breeding (blue, dotted), summer (green,
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

dashed), and transient period (orange, long dashed). Red solid lines indicate analyses that did not include a seasonal effect. The inset in each panel shows the
corresponding theoretical prediction described in Fig. 1.

However, to the best of our knowledge, there are no data on how frequenting popular roosting sites may expose griffons to patho-
griffons’ reproductive success changes after reaching old age in gens (1). Furthermore, individuals that frequent popular roost
wild populations (either increasing due to experience or decreasing sites may be more informed about the location of resources (35).
due to reproductive senescence). Future work that examines breed- While this may be a benefit most of the time, the most common
ing status might help determine the mechanisms that underlie cause of mortality in this population is poisoning, thus having
behavioral aging. The observed increase in roost fidelity may fur- more information about foraging resources may be maladaptive
ther result from a competitive advantage of older individuals over (46) and may expose highly informed individuals to poisoned
younger ones in occupying prime locations (44, 45). Indeed, as carcasses, decreasing their overall longevity. Future work may
individuals aged, they roosted more frequently at the 20% most examine lifelong fitness metrics (such as breeding success) to eval-
frequently used roost sites within our study area (“popular roosts,” uate the trade-­off between the benefits of proximity to feeding
Fig. 3D and SI Appendix, Tables S8 and S9), which tend to be stations and the costs of crowding at popular roost sites. Future
closer to supplementary feeding stations (SI Appendix, Fig. S3). analysis on individuals tracked from birth to adulthood (38) could
This pattern was so pronounced that griffons older than 5 (when also help to determine whether older vultures tend to converge
sexual maturity is reached) spent on average 80% of their nights their usage on roosts they frequented when they were young, or
at only 20% of roost sites. Furthermore, griffons over the age of if they become faithful to new sites as they age.
20 only used the most popular roosts (Fig. 3D). From a conser-
vation standpoint, protecting effectively the 20% most used roost Griffons Decrease the Strength of Social Relationships with Age.
sites within our study area would be sufficient to safeguard the In addition to the changes in movement behavior, there was also
roosting sites of adult griffons (>5 y old). Future work uncovering an effect of age on the social behavior of griffons. Contrary to
why these roosts are so popular may be important for guiding studies in other species (24), the proportion of the population
future conservation management actions. that an individual coroosted with (normalized degree) remained
Similar to roost fidelity, changes in roost selection with age were constant at all ages (Fig. 3E and SI Appendix, Table S10). This
mostly determined by individual plasticity, with individuals mov- stability probably reflects a balance between juveniles having
ing into more popular roosts as they age (SI Appendix, Fig. S2), lower roost fidelity (with frequent roost switches likely increasing
despite the slightly, nonsignificant, lower longevity of the individ- their degree), and adults’ tendency to use more crowded roosts
uals that occupy the most popular roosts (SI Appendix, Fig. S4 (exposing them to more vultures and possibly compensating for
and Table S3). These results suggest that while the proximity of their high roost fidelity). The long temporal scales of our social
popular roosts to feeding stations may provide short-­term benefits, networks, the small spatial scale of the study area, and the small
high fidelity to popular sites may also bear costs. For example, population size also likely contributed to this result; degree is a

4 of 9 https://doi.org/10.1073/pnas.2407298121 pnas.org
 early or late in life. This work adds to an increasing body of lon-
gitudinal studies analyzing behavioral aging in a multitude of
vertebrate species (for example, nonhuman primates (2, 24), red
deer Cervus elaphus (6), black kites Milvus migrans (25), among
others [reviewed in (48)]). We show that, with age, animals
become more faithful to their known sites and routines, and
potentially become more selective in their social relationships –
interestingly, behaviors that are commonly attributed to aging
humans (49, 50). While we found that within-­individual behav-
ioral plasticity is a major driving force of behavioral aging in a
long-­lived bird, selection may still act on the rate of behavioral
Compared models
change, for example, by favoring individuals that change more
Age
Age + ID
drastically earlier in life (17, 25). At the population level, the rate
Longevity + ID of behavioral change, and its timing in life, seem to differ among
Age + Longevity + ID
behaviors: Griffons displayed the highest site fidelity, stronger
movement routines, and weakest social relationships at different
stages in life. Indeed, it might be more important to be exploratory
Fig. 4.   Individual plasticity, rather than selective disappearance, shapes the and more social earlier in life, when individuals are learning about
increase in roost fidelity with age. To isolate the effect of individual plasticity the environment and searching for potential mates (8), and may
(i.e., within-­individual behavioral changes) from the effect of selective benefit from obtaining social knowledge for a longer part of their
disappearance (i.e., between-­individual behavioral differences and selection
acting on different behaviors), we compared four alternative statistical models: lives (51). In contrast, remaining in the most popular locations
age only (light yellow); age and individual ID as a random intercept (dark (here defined as the most commonly used sites) may be more
yellow); longevity (i.e., age-­at-­death) and individual ID as a random intercept beneficial later in life, when older individuals can better compete
(gray); and age, longevity, and individual ID as a random intercept (black). All
models included season as a fixed effect, and year as a random intercept.
for resources (44) or benefit from following others to discovered
The plot shows the predicted effects and 95% CI of age (modeled as a third-­ resources (35). Furthermore, long-­term information gathering in
degree polynomial, y = ax3 + bx2 + cx + d) and longevity on roost fidelity, for old age may be less needed, making older individuals more selec-
all four models. Both age and longevity are scaled. Adding longevity did not tive in their social relationships (2, 51). From a conservation per-
change the effect of age within the model and resulted in a nonsignificant
negative effect size for longevity, suggesting that behavioral plasticity, and spective, understanding what drives juvenile movements may help
not selective disappearance, shapes the relationship between age and roost improve the connectivity between endangered vulture populations.
fidelity in the population. Similar plots for other response variables are Furthermore, when attempting to establish new populations, trans-
included in SI Appendix, Fig. S4.
locating breeding-­aged individuals, whose site fidelity is high, may
improve the success of these conservation efforts.
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

measure that is sensitive to brief interactions, and over the course Similar to other species, we identified a strong effect of season-
of an entire season, most griffons of all age groups were likely to ality on behavior (52, 53), with periods when site fidelity and
have coroosted together for at least one night, resulting in high social selectivity may have a greater adaptive value, like the breed-
normalized degree values. ing season, showing more pronounced relationships between
In contrast, at the population level, the average strength of behavior and age. This seasonal effect is not likely driven by fluc-
roosting relationships was lower after vultures reached the age of tuations in food availability, given that vultures are provisioned at
10 y, during the breeding and summer seasons, but not during several feeding stations across the study area, as a part of ongoing
the transient season, suggesting that the breeding behavior of adult conservation management (54). In fact, this high predictability
individuals limits their interaction rates (Fig. 3F and SI Appendix, in food availability and distribution (55, 56) enabled us to decouple
Tables S11 and S12). We have observed anecdotally that when within-­individual behavioral changes from changes in the envi-
one vulture tends to the eggs or chicks in the nest the other partner ronment. Nevertheless, it is plausible that the observed sharp
tends to roost at a different site. This behavior would result in low increase in behavioral predictability in old age may be more mod-
average strength of social relationships between breeding individ- erate in areas with unpredictable foraging conditions (51, 57). In
uals. Future studies analyzing the movement and social behaviors such challenging environments, older individuals of long-­lived
of breeding couples would help test this hypothesis. Interestingly, species may play a particularly important role in buffering the
older griffons have weaker co-­roosting relationships despite grif- effects of poor environmental conditions, as they may act as repos-
fons’ increasing site fidelity to the most popular areas as they age. itories of ecological knowledge gathered throughout their lives
This may suggest that, similarly to primates (2, 24), older griffons (12, 14). As anthropogenic activities are threatening to change
may be more selective in their social relationships than their the age structure of populations of endangered species (20, 58),
younger counterparts. Because social and spatial behaviors are the loss of knowledge and behaviors exhibited by older individuals
intertwined (47), the changes we observe in social behavior may can hamper the ability of populations to adjust to changing envi-
also emerge from changes in spatial behavior with age (6). The ronmental conditions (12–14, 59).
mechanisms that shape age-­related patterns of social interactions
remain elusive. A strong correlation between age and individual
longevity rendered it impossible to test whether the population-­level Methods
change in average strength resulted from individuals reducing their Study System and Data Collection. The study was carried out in Israel, where
reliance on social information with age (6) or whether the pattern griffon vultures are critically endangered (37). A historical population of thou-
is due to the selective disappearance of individuals with higher sands of individuals (60) declined to about 400 griffons three decades ago and
average strength, for example, due to higher exposure to pathogens to less than 200 individuals today. To prevent the local extinction of this spe-
(1) (SI Appendix, Figs. S2 and S4 and Table S3). cies, there has been an intense conservation and management effort, including
Our results support our hypothesis that behaviors change dif- provisioning of food at supplementary feeding stations for vultures, release of
ferently with age, some gradually, while others change sharply captive-­bred and translocated individuals (61), minimizing mortality caused by

PNAS 2024 Vol. 121 No. 35 e2407298121 https://doi.org/10.1073/pnas.2407298121 5 of 9

 energy infrastructures, seasonal censuses, and monitoring through individual nighttime locations, if available). If there was no GPS position during that time
marking and GPS transmitters (38). frame, we used the first GPS position of the following day that was obtained
Between 2008 and 2022, we deployed GPS-­Accelerometer tags (160 g E-­Obs within 60 min after sunrise or earlier. If none of the GPS positions met one of
GmbH) and GPS-­GSM Accelerometer transmitters (50 g Ornitrack from Ornitela) these two criteria, we calculated the geodesic distance between the last GPS
on 319 griffons in Israel, using Teflon harnesses in backpack (2008 to 2015) and position of the day (if this position was obtained during the afternoon) and the
leg-­loop (2016 onward) configurations. Vultures were captured in cage traps (62) first GPS position of the next day (if this position was obtained in the morning).
and a small fraction were released from rehabilitation centers, captive-­breeding, If the distance between these two locations was 2 km or less, we considered the
and translocation programs (61). During tagging, all griffons were ringed with last position of the day to be the roosting location.
metal and color rings and marked with patagial tags for field identification. During To define roost-­sites, we grouped the identified roost locations (N = 145,270)
capture, they were aged based on their morphological characteristics, including using the DBSCAN clustering algorithm (Density-­Based Spatial Clustering of
feather shape and molting patterns (flight feathers, great coverts, and neck ruff Applications with Noise). We used 1 km as the epsilon distance (the maximum
feathers), and the bill and eye colors (63, 64). Every year, about 100 griffons are distance between two points for one to be considered close to the other), and
trapped or retrapped; therefore, almost all individuals are aged when they are still a minimum of 50 locations to identify the core points. We identified a total of
younger than 4 y old. We designated individuals as age 0 in their first year of life, 10,720 different roost-­sites. The most popular roost-­site included 25,834 roost
changing to age of 1 on the 15th of December of the same year (the approximate locations. Roost locations that were not clustered into a roost-­site (for example,
start of the breeding season). Nine griffons were initially captured when they were for griffons that were performing long-­range forays and roosting in unusual loca-
already adults (>5 y old) and therefore it was not possible to reliably age them in tions) were identified as their own unique roost-­site (N = 10,666). To visualize
the field. To estimate their age, we compared their roost fidelity with individuals the roost-­sites, we built convex hulls based on the roost locations of each site
of known age, starting when they were at least 15 y old. On average, their roost (as shown in Fig. 2A). This analysis was performed using R package “fpc”(69).
fidelity was not different from the cohort of 18 y old, and therefore, we assumed
Roost Fidelity and Routine Behaviors. We computed a daily metric of “roost
this was their average age (SI Appendix, Fig. S5). To ensure that the addition of
these individuals did not shape our results, we performed all statistical analysis fidelity”: If a vulture remained in the same roost-­site on two consecutive nights,
with and without the individuals of unknown age and the results remained the we considered the vulture to have used the same roost. To ensure our results
same (SI Appendix, Fig. S6). were not sensitive to the clustering method, we repeated all analyses using a
Considering our high recapture rate, we were able to replace the GPS devices threshold of 1 km and 20 km for considering two consecutive nights as using the
that had either stopped working or that had fallen off the griffons. The average same roost-­site (SI Appendix, Fig. S1). In addition, to guarantee that our results
duration of a single deployment was 457 d (min = 1, max = 2962, including were not influenced by the younger griffons’ long-­range forays (during which
only terminated deployments, SI Appendix, Fig. S7), and the median number of they might switch roosts more often than they would in our main study area),
deployments per bird was 1 (mean = 1.5). Yet, one individual was redeployed we repeated the models with these long-­distance travels removed, resulting in
with new GPS devices 6 times during its life. Overall, the gap between deploy- the same patterns (SI Appendix, Fig. S1).
ments was 543 d (min = 0 d, max = 3,622 d). We then investigated movement routine for griffons that switched roosts
The GPS transmitters typically recorded location and ground speed every within our main study area (Israel and surrounding areas: longitude 34.2 to 36.5,
10 min during the day (average 13 min, SD = 68 min), and provided one or latitude 29.3 to 33.2), as most of the griffons remained in this area throughout
the study period. We quantified movement routine as the degree of predictability
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

two additional locations during the night, when vultures are inactive. Yet, the
sampling rates vary according to the battery charge, specific research questions in a sequence of roost sites when griffons switched roosts (40, 56). To do so, for
[e.g., periods of high-­resolution tracking at 1 Hz (5)], and the GPS transmitter’s each griffon, we considered the order of the roost sites they used within a year as
longevity; at the end of the battery’s lives, transmitters often fail to charge and a roost sequence (removing consecutive nights at the same roost-­site), and for
provide only 1 point per day, usually at night, when vultures are in the roosts [see each sequence, we computed a routine index using an algorithm developed by
Acácio et al (38) for an example]. We initially filtered the data to remove the first Riotte-­Lambert et al. (40). This routine index ranges from 0 (complete unpredicta-
date of GPS deployment, as well as any dates when the individuals were captured bility of roost switching sequence) to 1 (complete predictability of roost switching
or, on rare occasions, at the wildlife hospital. We also filtered outlier positions; sequence). Because longer sequences are more reliable for routine identification,
an observation was considered an outlier if the speed between two consecutive we performed this analysis at the yearly level rather than breaking sequences by
locations was over 50 m/s, or if two consecutive nighttime locations were over season, thus maximizing the length of the roost switching sequences (average
10 km apart. We then manually identified vultures that spent most of their time sequence length = 46 roost-­switches, range = 10 to 243).
in the southern region of Israel during the study period (n = 210), excluding Identifying Popular Roosts. We hypothesized that age may affect the inten-
griffons that spent most of their time in the northern region of Israel or in other sity of use of popular roosts. We started by identifying the “popular roosts” by
countries (Türkiye, Greece, Bulgaria), as their movement patterns might differ examining the number of roost locations at each roost-­site. We then considered
from the ones in the south of Israel. All distance metrics were calculated as the “popular roosts” to be the top 20% of roost sites with the highest number of
shortest distance between the two points on an ellipsoid (WGS84), using the R locations (out of all roost sites with at least two roost locations) (SI Appendix,
package “geosphere”(65). Fig. S8). In total, we identified six popular roost sites out of 31 available roost
We then divided each year into three distinct seasons: breeding, summer, and sites within our main study area.
the transient period. We considered the “breeding” season to last from December We controlled for three possible biases that could influence our results: the
15th of the previous year (when griffons start to select their nest site and partner) to threshold for defining roost popularity, the tendency of younger individuals to
May 14th of the current year (when griffons are either still caring for an old chick or, leave the main study area, and the high roost fidelity of older individuals. To
considering the low breeding success in our studied population, have already failed examine whether the results were influenced by the popularity threshold, we
their breeding attempt). We defined the “summer” season between May 15th and repeated our analyses using 25% and 30% thresholds for roost popularity. We
September 14th (postbreeding for some individuals, and late breeding for others), also investigated whether our results could be explained by younger individuals
and the “transient” period [when long-­range forays typically occur (31)] between being more likely to leave the main study area (to peripheral regions where there
September 15th and December 14th. All data processing, analyses, and mapping are no “popular roosts”). To do so, we repeated the analyses without the long-­
were conducted in R (66). All code and data are available at https://github.com/ range forays. We found that the results were not affected by the chosen popularity
msa2015/Ageing_Vultures (67) and the data is available in Zenodo (68). threshold or by the elimination of the long-­range forays (SI Appendix, Fig. S9).
Finally, we also note that our popularity index was not driven by the roost usage
Roost Location and Roost-­Site Identification. Vultures roost in deep canyons of old individuals: First, older individuals are uncommon in the dataset (Fig. 2B)
with poor GPS coverage, resulting in frequent GPS failures within the roost. Thus, and thus do not drive roost-­site popularity estimate. Second, when calculating
to determine where vultures roosted, we identified, for each individual, the last the top 20% of roost sites using a random and balanced design (in terms of the
GPS position of the day that was recorded 60 min before sunset or later (including ages of the individuals), our results remained the same (SI Appendix, Fig. S10).

6 of 9 https://doi.org/10.1073/pnas.2407298121 pnas.org
 Social Behavior. To quantify the social behavior of griffon vultures, we exam- Gaussian distributions and identity link functions, continuous variables >0). To
ined co-­occurrence at roosting sites. For this analysis, we only considered years ensure that our indices were not biased by an incomplete representation of the
when at least 25% of the total Israeli griffon population was continuously GPS population’s social network in particular seasons, for these analyses, we only
tracked (39) (2016 to 2022, SI Appendix, Fig. S11). A conservative estimate of the considered individuals that were tracked for at least 30 nights within a season,
minimum percentage of the tracked population ranged between 26% and over and seasons that had at least 30 tracked individuals. We were left with 16 unique
60%. We excluded any nights that the griffons roosted outside the main study networks and 108 griffons (total of 510 data points).
area, because we could not be sure if they were roosting alone or with non-­GPS Models for each of the social behavior measures included age, season, and
tracked griffons from other populations. We then constructed a series of seasonal their interaction as explanatory variables. For normalized degree, the interaction
social networks (N = 20 y-­season combinations), in which vultures that roosted was not statistically significant and was dropped from the model. All models had
within 1 km of each other were considered to be co-­roosting. The weight of the year and individual as random intercepts. The model with strength as a response
interactions between two individuals was calculated as a simple ratio index, based variable did not fit our data correctly (high heteroscedasticity), and therefore this
on their seasonal interaction frequency, divided by the total number of nights variable was removed from further analysis. For roost popularity, normalized
within a season when both individuals were tracked and could have roosted degree, and average strength, we tested the four functional relationships of age
together (70–72). From these networks, we derived three individual-­level net- (linear, quadratic, exponential, third-­degree polynomial), as detailed above, and
work centrality measures: normalized degree, strength, and average strength. chose the best model according to the AIC.
Normalized degree is the number of unique individuals a griffon coroosted with
Mechanisms Shaping Age-­Dependent Movement and Social Behaviors. To
within a season, divided by the number of individuals in the seasonal network
minus 1 (average network size = 87, range = 30 to 133). Strength is a measure assess the relative importance of selective disappearance and/or individual plas-
of how many interactions an individual had at roosts and is calculated as the sum ticity for shaping behavioral aging in movement and social behavior of griffons,
of the weights of all interactions (edges) of an individual. Average strength is the we considered the subset of individuals with known death year (and consequently
ratio between strength and degree (i.e., strength divided by degree), represent- their age at death, i.e. longevity). Because the Israeli griffon population is heav-
ing how strongly connected a griffon is on average to each of the individuals it ily monitored, if a GPS tag shows no movement a ranger is sent to the field to
shared a roost with. investigate a possible mortality event (38). Furthermore, almost all individuals
are observed in the field during their lifetime, either during the seasonal census
Changes in Movement Behavior with Age. To evaluate the influence of age or at their nest or at feeding stations (average number of yearly observations per
on the movement behavior of griffon vultures, we fitted a series of generalized GPS-­tracked individual = 3, range = 0 to 67). As a result, we considered a vulture
linear mixed models (GLMMs), using the R package “glmmTMB” (73). We con- to be dead if it was found dead in the field, or if it was not seen in the field for
sidered three response variables to describe the movement behavior of griffons: at least 5 times its usual visual observation rate. This rate was calculated from all
“roost fidelity” (binomial distribution with a logit link function, 1 = remained at the observations for each individual during their lifetime, with or without a GPS
the same roost; 0 = switched between roosts), “probability of having a routine” transmitter (e.g., if individual A was seen every 2 mo, then it would be considered
(binomial distribution with a logit link function, 1 = index of routine is larger dead if it was not seen for 10 mo; if individual B was seen every 36 mo, then it
than 0, 0 = index of routine is 0), and “routine index” (beta distribution with a would be considered dead if it was not seen for 180 mo). The age of death was
logit link function, values ranging between >0 and 1, with higher values indi- the age the individual was last seen alive plus its usual observation rate (as an
cating higher predictability in roost switching sequence). For these analyses, we individual could have died at any point during that period). Individuals that were
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

considered only individuals with at least 30 identified roosts in a season (roost never observed in the field (e.g. if they were mainly active outside Israel due to
fidelity), or at least 30 roosts in a sequence within a year (routine). On average, we natal or breeding dispersal) were not considered dead unless their body was
identified 607 roosts for each individual (min = 36, max = 3,120). To ensure that found (n = 3 individuals). To ensure our results were robust to the used 5-­fold
tracking duration did not impact our results, we performed a sensitivity analysis threshold, we repeated all analyses with a 10 -­fold rate threshold, as well as with
and repeated our data analysis twice: using individuals that were tracked for at an alternative approach of a uniform 2 y-­gap since the last observation (for all
least 182 d (6 mo) and again, only with individuals that were tracked for 365 d individuals). All three age-­at-­death estimates provided similar results for the
(1 y) or more (SI Appendix, Fig. S12). The results of both analyses did not differ relative importance of longevity and behavioral plasticity (SI Appendix, Figs. S15
from results of the full dataset (SI Appendix, Figs. S13 and S14). Overall, using and S16). For the analysis of the mechanism of behavioral aging, we had 59
the 30 d’ threshold, we identified the roost fidelity for 142 griffons (86,192 roost individuals for the “roost fidelity” and “probability of occupying a popular roost”
locations); we modeled the probability of routine for 135 griffons (627 roost analyses, 55 individuals for the “probability of routine” analysis, 41 individuals
switching sequences); and we modeled the influence of age on the routine index for the “routine index” analysis, and 29 individuals for the “average strength”
for 115 griffons (312 roost switching sequences, with routine index >0). analysis. With this reduced dataset for the effect of age on the probability or
The structure of each model was as follows: roost fidelity as a function of age, index of routine (due to the absence of older individuals in this subset of data),
season, and their interaction; probability of routine as a function of age and we could not evaluate the role of selective disappearance or individual plasticity
length of the roost sequence (log transformed); routine index as a function of for these variables.
age (because length of the roost sequence was not statistically significant and We then performed a test for selective disappearance as described by Van de Pol
was therefore dropped from the model). All models had year and individual as & Verhulst (6, 27). We fitted the best model of each of our response variables (“roost
random intercepts. For each response variable, we fitted four models, each with fidelity,” “probability of routine,” “routine index,” “probability of occupying a pop-
a different functional relationship between behavior and age: linear (y = ax + b), ular roost,” and “average strength”) using four different models: age only, without
quadratic (y = ax2 + bx + c), exponential (y = ax), and third-­degree polynomial any other variables (“Age” model); age and individual ID as a random intercept, to
(y = ax3 + bx2 + cx + d) (Fig. 1). Depending on the behavior, these models may account for individual variation (“Age + ID” model); longevity (i.e., age at death) and
represent an increasing or decreasing relationship between behavior and age. individual ID as a random intercept (“Longevity + ID” model); and age, longevity
We also fitted a null model with only random effects. We analyzed the model as fixed effects and Individual ID as a random intercept (“Age + Longevity + ID”
residuals using R package “DHARMa” (74). For each response variable, we ranked model). All models had season as a fixed effect and year as a random intercept.
models according to their AIC and selected the model with the lowest AIC. If the The model residuals were analyzed, as well as the collinearity between variables
top models had delta AIC < 2, we selected the simplest model. (variance inflation factor, VIF). These four models allowed us to isolate the effect of
individual plasticity (i.e., within-­individual changes) from the effect of selective
Changes in Sociality with Age. To analyze how age affects griffon vultures’ disappearance (i.e., between-­individual changes). If adding longevity changes the
social behavior, we fitted a series of GLMMs considering four response variables: effect of age within the model, it implies that selective disappearance (and not
“probability of occupying a popular roost” (binomial distribution with a logit link individual plasticity) shapes the age-­dependent pattern found at the population
function, 1 = spent the night in a popular roost, 0 = spent the night in a non- level [more details in Van de Pol and Verhulst (6, 27)]. After building the four models
popular roost), “normalized degree” (beta distribution with a logit link function, (“Age,” “Age + ID,” “Longevity + ID,” and “Age + Longevity + ID”) for each response
values ranging between 0 and 1), “strength” and “average strength” (both with variable, we selected the model with highest predictive power based on AIC. For the

PNAS 2024 Vol. 121 No. 35 e2407298121 https://doi.org/10.1073/pnas.2407298121 7 of 9

 “average strength,” the VIF between age and longevity in the “Age + Longevity + Miller, Asher Perez, Avishai Bar-­On, Yaniv Levy-­Paz, Dvora Shilo, Lior lev, Sappir
ID” model was over 10 implying these two variables were collinear; thus, we could Simchi, Elad Zisso, Elya Maatuf (Elika), Zehava Sigal, and Arye Rosenberg. We
not assess whether plasticity or selection drives the changes in average strength. also acknowledge the members of the Movement Ecology Laboratory at the
To ensure that our results were not driven by the complex relationship between Hebrew University of Jerusalem. Finally, we thank all members of the Movement
behavior and age, or by a nonlinear relationship between behavior and longevity, Ecology and Individual Behavior Laboratory at Tel Aviv University, in particular
we fitted all models with age as a linear effect, second-­degree and third-­degree Assaf Uzan for all support, and thank Elvira D’Bastiani for helpful discussions on
polynomials. We also fitted longevity as a quadratic term. Doing so did not affect our earlier versions of this work. Funding for this work was provided by the NSF-­BSF
conclusions regarding the mechanisms of behavioral aging in griffons (SI Appendix, grant: NSF IOS division 2015662/BSF 2019822 to N.P.W. and O.S., as well as BSF
Table S13, Figs. S17, S18). In the main text, present the models with the lowest AIC grant 255/2008 to R.N. M.A. was supported by the George S. Wise Postdoctoral
(third-­degree polynomials and longevity with a linear effect) while the remaining Fellowship (Tel Aviv University).
model outputs can be found in Table S13.

Data, Materials, and Software Availability. Analysis code is available at Author affiliations: aSchool of Zoology, Faculty of Life Sciences, Tel Aviv University, Tel
Aviv, Israel; bDepartment of Ecology and Evolutionary Biology, University of California
https://github.com/msa2015/Ageing_Vultures (67). The data are available in Los Angeles, Los Angeles, CA; cScience Division, Israel Nature and Parks Authority,
Zenodo, under the DOI: 10.5281/zenodo.10651582 (68). The GPS coordinates Jerusalem, Israel; dDepartment for the Ecology of Animal Societies, Max Planck Institute
were shifted a few kilometers from the original location to ensure species safety, of Animal Behavior, Konstanz, Germany; eMitrani Department of Desert Ecology, Jacob
Blaustein Institutes for Desert Research, Ben-­Gurion University of the Negev, Midreshet
while maintaining all geometric attributes needed for reconstructing the analyses. Ben-­Gurion, Israel; and fDepartment of Ecology, Evolution and Behavior, The Hebrew
University of Jerusalem, Jerusalem, Israel
ACKNOWLEDGMENTS. We want to sincerely acknowledge all individuals who Author contributions: M.A., N.P.-­W., and O.S. designed research; M.A., K.G., and O.S.
have contributed to the conservation of griffon vultures in Israel, and without performed research; M.A. and K.G. analyzed data; M.A. and K.G. critically revised the
paper; N.A., G.V., R.H., and R.E. collected data; critically revised the paper; O.H. collected
whom this work would not have been possible. In particular, we thank the work- data; critically revised the paper; secured funding; R.N., N.P.-­W., and O.S. critically revised
ers and rangers of the INPA for their hard work and dedication, specially Ygal the paper; Secured funding; and M.A. wrote the paper.

1. G. F. Albery, A. R. Sweeny, Q. Webber, How behavioural ageing affects infectious disease. Neurosci. 27. M. van de Pol, S. Verhulst, Age-­dependent traits: A new statistical model to separate within-­and
Biobehav. Rev. 155, 105426 (2023). between-­individual effects. Am. Nat. 167, 766–773 (2006).
2. A. G. Rosati et al., Social selectivity in aging wild chimpanzees. Science 370, 473–476 (2020). 28. D. H. Nussey, T. Coulson, M. Festa-­Bianchet, J.-­M. Gaillard, Measuring senescence in wild animal
3. V. J. Lecomte et al., Patterns of aging in the long-­lived wandering albatross. Proc. Natl. Acad. Sci. populations: Towards a longitudinal approach. Funct. Ecol. 22, 393–406 (2008).
107, 6370–6375 (2010). 29. S. Chantepie et al., Age-­related variation and temporal patterns in the survival of a long-­lived
4. R. Efrat, O. Hatzofe, T. Mueller, N. Sapir, O. Berger-­Tal, Early and accumulated experience shape scavenger. Oikos 125, 167–178 (2016).
migration and flight in Egyptian vultures. Curr. Biol. 33, 5526–5532.e4 (2023), 10.1016/j. 30. R. Nathan et al., Using tri-­axial acceleration data to identify behavioral modes of free-­ranging
cub.2023.11.012. animals: General concepts and tools illustrated for griffon vultures. J. Exp. Biol. 215, 986–996
5. R. Harel, N. Horvitz, R. Nathan, Adult vultures outperform juveniles in challenging thermal soaring (2012).
conditions. Sci. Rep. 6, 27865 (2016). 31. O. Spiegel et al., Moving beyond curve fitting: Using complementary data to assess alternative
6. G. F. Albery et al., Ageing red deer alter their spatial behaviour and become less social. Nat. Ecol. explanations for long movements of three vulture species. Am. Nat. 185, E44–E54 (2015).
Evol. 6, 1231–1238 (2022), 10.1038/s41559-022-01817-9. 32. K. L. Bildstein, M. J. Bechard, C. Farmer, L. Newcomb, Narrow sea crossings present major obstacles
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

7. P. A. Cook, R. A. Costello, V. A. Formica, E. D. Brodie, Individual and population age impact social to migrating Griffon Vultures Gyps fulvus. Ibis 151, 382–391 (2009).
behavior and network structure in a long-­lived insect. Am. Nat. 202, 667–680 (2023). 33. N. Sharma, N. Anglister, O. Spiegel, N. Pinter-­Wollman, Social situations differ in their contribution
8. H. Weimerskirch et al., Lifetime foraging patterns of the wandering albatross: Life on the move!. J. to population-­level social structure in griffon vultures. Ecol. Evol. 13, e10139 (2023).
Exp. Mar. Biol. Ecol. 450, 68–78 (2014). 34. P. Mateo-­Tomás, P. P. Olea, The importance of social information in breeding site selection increases
9. S. Verhulst, M. Geerdink, H. M. Salomons, J. J. Boonekamp, Social life histories: Jackdaw dominance with population size in the Eurasian Griffon Vulture Gyps fulvus. Ibis 153, 832–845 (2011).
increases with age, terminally declines and shortens lifespan. Proc. R. Soc. B Biol. Sci. 281, 35. R. Harel, O. Spiegel, W. M. Getz, R. Nathan, Social foraging and individual consistency in following
20141045 (2014). behaviour: Testing the information centre hypothesis in free-­ranging vultures. Proc. R. Soc. B Biol.
10. M. A. Verhoeven et al., Generational shift in spring staging site use by a long-­distance migratory Sci. 284, 20162654 (2017).
bird. Biol. Lett. 14, 20170663 (2018). 36. D. L. Ogada, F. Keesing, M. Z. Virani, Dropping dead: Causes and consequences of vulture population
11. J. A. Gill, J. A. Alves, T. G. Gunnarsson, Mechanisms driving phenological and range change in declines worldwide. Ann. N. Y. Acad. Sci. 1249, 57–71 (2012).
migratory species. Philos. Trans. R. Soc. B Biol. Sci. 374, 20180047 (2019). 37. A. Mayrose, Z. Labinger, O. Steinitz, N. Leader, The Red Book of Birds in Israel (INPA, SPNI, ed. 2017,
12. L. J. N. Brent et al., Ecological knowledge, leadership, and the evolution of menopause in killer 2017).
whales. Curr. Biol. 25, 746–750 (2015). 38. M. Acácio et al., A lifetime track of a griffon vulture: The moving story of Rehovot (Y64). Ecology 104,
13. C. S. Teitelbaum et al., Experience drives innovation of new migration patterns of whooping cranes e3985 (2023).
in response to global change. Nat. Commun. 7, 12793 (2016). 39. M. J. Silk, A. L. Jackson, D. P. Croft, K. Colhoun, S. Bearhop, The consequences of unidentifiable
14. K. McComb, C. Moss, S. M. Durant, L. Baker, S. Sayialel, Matriarchs as repositories of social individuals for the analysis of an animal social network. Anim. Behav. 104, 1–11 (2015).
knowledge in African elephants. Science 292, 491–494 (2001). 40. L. Riotte-­Lambert, S. Benhamou, S. Chamaillé-­Jammes, From randomness to traplining: A
15. T. Clutton-­Brock, B. C. Sheldon, Individuals and populations: The role of long-­term, individual-­ framework for the study of routine movement behavior. Behav. Ecol. 28, 280–287 (2017).
based studies of animals in ecology and evolutionary biology. Trends Ecol. Evol. 25, 562–573 41. E. Schohat, O. Ovadia, O. Hatzofe, Estimating Griffon Vulture (Gyps Fulvus) Survival in Israel Based
(2010). on Mark-­Resight Data (Israel Nature and Parks Authority, 2012).
16. J. T. Cunningham et al., Reduced activity in middle-­aged thick-­billed murres: Evidence for age 42. N. Anglister et al., Plasma cholinesterase activity: A benchmark for rapid detection of pesticide
related trends in fine-­scale foraging behaviour. Anim. Behav. 126, 271–280 (2017). poisoning in an avian scavenger. Sci. Total Environ. 877, 162903 (2023).
17. S. Rotics et al., The challenges of the first migration: Movement and behaviour of juvenile vs. adult 43. D. Pardo, C. Barbraud, H. Weimerskirch, Females better face senescence in the wandering albatross.
white storks with insights regarding juvenile mortality. J. Anim. Ecol. 85, 938–947 (2016). Oecologia 173, 1283–1294 (2013).
18. A. G. Hertel et al., Time constraints may pace the ontogeny of movement behaviour. Proc. R. Soc. B 44. M. Bosè, F. Sarrazin, Competitive behaviour and feeding rate in a reintroduced population of Griffon
Biol. Sci. 290, 20222429 (2023). Vultures Gyps fulvus. Ibis 149, 490–501 (2007).
19. D. Pardo, C. Barbraud, M. Authier, H. Weimerskirch, Evidence for an age-­dependent influence of 45. J. A. Donázar, J. E. Feijóo, Social structure of andean condor roosts: influence of sex, age, and
environmental variations on a long-­lived seabird’s life-­history traits. Ecology 94, 208–220 (2013). season. The Condor 104, 832–837 (2002).
20. H. Weimerskirch, A. Corbeau, A. Pajot, S. C. Patrick, J. Collet, Albatrosses develop attraction to fishing 46. M. Sigaud et al., Collective decision-­making promotes fitness loss in a fusion-­fission society. Ecol.
vessels during immaturity but avoid them at old age. Proc. R. Soc. B Biol. Sci. 290, 20222252 Lett. 20, 33–40 (2017).
(2023). 47. Q. M. R. Webber et al., Behavioural ecology at the spatial–social interface. Biol. Rev. 98, 868–886
21. F. Angelier, S. A. Shaffer, H. Weimerskirch, O. Chastel, Effect of age, breeding experience and (2023).
senescence on corticosterone and prolactin levels in a long-­lived seabird: The wandering albatross. 48. J. P. Woodman et al., The ecology of ageing in wild societies: linking age structure and social
Gen. Comp. Endocrinol. 149, 1–9 (2006). behaviour. [Preprint] (2024). Available at: https://ecoevorxiv.org/repository/view/6715/. Accessed 29
22. P. Catry, R. A. Phillips, B. Phalan, J. P. Croxall, Senescence effects in an extremely long-­lived bird: The May 2024.
grey-­headed albatross Thalassarche chrysostoma. Proc. R. Soc. B Biol. Sci. 273, 1625–1630 (2006). 49. C. Wrzus, M. Hänel, J. Wagner, F. J. Neyer, Social network changes and life events across the life
23. A. R. Martinig, K. J. Mathot, J. E. Lane, B. Dantzer, S. Boutin, Selective disappearance does not span: A meta-­analysis. Psychol. Bull. 139, 53–80 (2013).
underlie age-­related changes in trait repeatability in red squirrels. Behav. Ecol. 32, 306–315 (2021). 50. N. Thompson González, Z. Machanda, M. Emery Thompson, Age-­related social selectivity: An
24. E. R. Siracusa et al., Within-­individual changes reveal increasing social selectivity with age in rhesus adaptive lens on a later life social phenotype. Neurosci. Biobehav. Rev. 152, 105294 (2023).
macaques. Proc. Natl. Acad. Sci. U.S.A. 119, e2209180119 (2022). 51. O. Spiegel, M. C. Crofoot, The feedback between where we go and what we know—information
25. F. Sergio et al., Individual improvements and selective mortality shape lifelong migratory shapes movement, but movement also impacts information acquisition. Curr. Opin. Behav. Sci. 12,
performance. Nature 515, 410–413 (2014). 90–96 (2016).
26. F. Sergio et al., Compensation for wind drift during raptor migration improves with age through 52. L. J. N. Brent, A. MacLarnon, M. L. Platt, S. Semple, Seasonal changes in the structure of rhesus
mortality selection. Nat. Ecol. Evol. 6, 989–997 (2022). macaque social networks. Behav. Ecol. Sociobiol. 67, 349–359 (2013).

8 of 9 https://doi.org/10.1073/pnas.2407298121 pnas.org
 53. S. Nandini, P. Keerthipriya, T. N. C. Vidya, Seasonal variation in female Asian elephant social structure 64. I. Zuberogoitia et al., The flight feather molt of griffon vultures (Gyps fulvus) and associated
in Nagarahole-­Bandipur, southern India. Anim. Behav. 134, 135–145 (2017). biological consequences. J. Raptor Res. 47, 292–303 (2013).
54. O. Spiegel, R. Harel, W. M. Getz, R. Nathan, Mixed strategies of griffon vultures’ (Gyps fulvus) 65. R. Hijmans, Geosphere: Spherical Trigonometry. (2022). Deposited 2022.
response to food deprivation lead to a hump-­shaped movement pattern. Mov. Ecol. 1, 5 (2013). 66. R Core Team, R: A language and environment for statistical computing. Deposited 2023 (2023).
55. A. Cortés-­Avizanda et al., Supplementary feeding and endangered avian scavengers: Benefits, 67. M. Acácio et al., “Ageing Vultures”. GitHub. https://github.com/msa2015/Ageing_Vultures.
caveats, and controversies. Front. Ecol. Environ. 14, 191–199 (2016). Deposited 6 February 2024.
56. J. Fluhr, S. Benhamou, L. Riotte-­Lambert, O. Duriez, Assessing the risk for an obligate scavenger to 68. M. Acácio et al., Dataset for “Behavioural plasticity shapes population ageing patterns” [Dataset].
be dependent on predictable feeding sources. Biol. Conserv. 215, 92–98 (2017). Zenodo. https://doi.org/10.5281/zenodo.10651583. Deposited 12 February 2024.
57. L. Riotte-­Lambert, J. Matthiopoulos, Environmental predictability as a cause and consequence of 69. C. Hennig, fpc, Flexible procedures for clustering. Deposited 2023 (2023).
animal movement. Trends Ecol. Evol. 35, 163–174 (2020). 70. S. J. Cairns, S. J. Schwager, A comparison of association indices. Anim. Behav. 35, 1454–1469
58. J. M. Milner, E. B. Nilsen, H. P. Andreassen, Demographic side effects of selective hunting in (1987).
ungulates and carnivores. Conserv. Biol. 21, 36–47 (2007). 71. P. He et al., A guide to sampling design for GPS-­based studies of animal societies. Methods Ecol.
59. C. Foley, N. Pettorelli, L. Foley, Severe drought and calf survival in elephants. Biol. Lett. 4, 541–544 (2008). Evol. 14, 1887–1905 (2023).
60. H. B. Tristram, The Land of Israel: A Journal of Travels in Palestine (Clay, Son and Taylor, 1865). 72. J. R. Ginsberg, T. P. Young, Measuring association between individuals or groups in behavioural
61. R. Efrat, O. Hatzofe, Y. Miller, O. Berger-­Tal, Determinants of survival in captive-­bred Griffon Vultures studies. Anim. Behav. 44, 377–379 (1992).
Gyps fulvus after their release to the wild. Conserv. Sci. Pract. 2, e308 (2020). 73. M. Brooks et al., glmmTMB balances speed and flexibility among packages for zero-­inflated
62. S. Izekiel, B. Woodley, O. Hatzofe, Cage trap for Gyps fulvus. Vulture News 49, 14–16 (2003). generalized linear mixed modeling. R J. 9, 378–400 (2017).
63. O. Duriez, B. Eliotout, F. Sarrazin, Age identification of Eurasian Griffon Vultures Gyps fulvus in the 74. F. Hartig, DHARMa: residual diagnostics for hierarchical (multi-­level/mixed) regression models.
field. Ringing Migr. 26, 24–30 (2011). Deposited 2022 (2022).
Downloaded from https://www.pnas.org by 58.213.155.250 on April 8, 2025 from IP address 58.213.155.250.

PNAS 2024 Vol. 121 No. 35 e2407298121 https://doi.org/10.1073/pnas.2407298121 9 of 9