Environmental and social correlates, and energetic consequences of fitness maximisation on different migratory behaviours in a long-lived scavenger

Partial migration is one of the most widespread migratory strategies among taxa. Investigating the trade-off between environmental/social factors — fitness and energetic consequences — is essential to understand the coexistence of migratory and resident behaviours. Here, we compiled field monitoring data of wintering population size and telemetry data of 25 migrant and 14 resident Egyptian Vultures Neophron percnopterus to analyse how environmental and social factors modulate overwintering immature population size, compare energetic consequences between migratory and resident individuals across wintering and non-wintering seasons and evaluate fitness components (i.e. survival and reproduction) between the two migratory forms. We observed that social attraction may influence the number of overwintering immature individuals, which increased linearly with adult birds surveyed. Residents spent more energy but exhibited higher survival probabilities and lower breeding activity. On the contrary, migratory birds showed lower energy expenditure during winter but also lower survival and more breeding attempts. These results suggest that social attraction may modulate population dynamics and promote residency in immature birds. Resident individuals benefit from enhancing their survival at the expense of higher energy expenditure during winter. Migrant birds, on the contrary, may compensate for the higher costs in terms of survival by a reduction in the energy cost, which may benefit more frequent breeding. Our results offer new insights to understand how species benefit from one strategy or another and that the coexistence of both migratory forms is context-dependent. Animal populations exhibiting partial migration are composed of migrant and resident individuals who share the same breeding areas but different overwintering quarters. Deciphering the causes and consequences that affect each migratory behaviour is essential to understand the balance and persistence of the two strategies. Here, we investigate the environmental and social factors affecting number of immatures during winter and evaluate both migratory strategies in terms of energy expenditure, reproduction and survival. We found that social attraction modulates wintering population size of immature individuals and that the residency is energetically more costly but beneficial in terms of survival but not for breeding. By contrast, migration lessens the energy costs, increases the breeding activity, but also reduces survival probability.


Introduction
The drivers of animal movement and its consequences pose some of the most challenging research questions in ecology. Among the various forms and realisations of animal movement, migration is the most conspicuous phenomenon and has attracted scientists' attention for centuries. Animal Communicated by W. Wiltschko migration is defined as a bidirectional and repeated movement between two different places, usually breeding and wintering areas, with a residency period in each location (Hansson and Akesson 2014). One of the most widespread strategies in migratory species is partial migration, in which part of the population remains in the breeding grounds while some individuals move to a distinct place to overwinter (Chapman et al. 2011a). The life-histories of individuals, in species simultaneously exhibiting these two well-marked behaviours, are subject to (and shaped by) different environmental, social and evolutionary forces (Lundberg 2013). Disentangling which factors determine the stability and persistence of these two strategies is crucial to understand the emergence of the migratory behaviour (Turbek et al. 2018).
Environmental (e.g. climate, resources) and social cues (e.g. location or presence of other individuals) are among the most influencing factors modulating migratory behaviour in partial migratory species (Shaw and Couzin 2013). Therefore, changes in these factors may favour migration or residency, and ultimately affect survival and reproduction (Buchan et al. 2020). Among the environmental cues, milder winters and higher food availability during winter could favour residency at higher latitudes, which increases individual survival and enables early reproduction and higher breeding success (Meller et al. 2016;Gilroy 2017). On the contrary, migration could be more advantageous by abandoning the resource-depleted regions during the non-breeding season which could increase both the chances of survival and reproduction (Winger et al. 2019;Winger and Pegan 2020). In terms of social cues, in long-lived species, the decision to migrate or not may come from learning or following more experienced individuals which could enhance the survival of less-skilled individuals (e.g. juveniles; Mellone et al. 2011;Teitelbaum et al. 2016). These findings show that despite the great progress that has been made in understanding the contribution of survival and reproduction to the maintenance of these two migratory behaviours, less attention has been paid to their energetic consequences. Energy expenditure is a key link between behaviour and overall fitness (Grémillet et al. 2018), therefore addressing energetic consequences of different migratory strategies is of paramount interest.
The recent advent of high-resolution Global Positioning Systems (GPS) in combination with tri-axial acceleration data has allowed researchers to study how much energy is spent by animals during crucial periods of their annual cycle (i.e. breeding season; Grémillet et al. 2018), estimate how expensive or efficient it is to move depending on their kinematic patterns (i.e. flapping flight; Williams et al. 2020) and even investigate which factors limit activity duration (Pokrovsky et al. 2021). Integrating this energetic perspective into the study of migratory strategies could help to understand (1) the indirect energetic consequences of changes in those environmental and social factors that modulate migration and residency and the trade-offs between energy expenditure and fitness (e.g. survival and reproduction) of both strategies. Overall, these two aspects could help explain the relative contribution of environmental conditions, social factors, fitness and energy expenditure, to the coexistence and persistence of these two strategies in partial migratory species (Gilroy 2017). This is of great importance to predict not only how current and future environmental changes could impact populations but also to design effective conservation measures that account for within-species behavioural diversity and preserve species functionality and their role in ecosystems (Gilroy et al. 2016).
Here, we untangle from a mechanistic perspective which factors modulate and shape both migratory and residency strategies in a long-distance soaring migrant, the Egyptian Vulture Neophron percnopterus, by combining field monitoring and GPS tracking data of a partially migratory population in Spain. This endangered species ranges across southern Europe, northern Africa, the Middle East and Central and South Asia (BirdLife International 2020). While migratory individuals regularly travel more than 4000 km between their northern breeding and southern wintering grounds by using several distinct migratory flyways (Phipps et al. 2019;Buechley et al. 2021), other individuals overwinter in southern and south-western Spain (García et al. 2000;Morant et al. 2020). The Egyptian Vulture is an obligate scavenger that frequently consumes carrion from both livestock and wild ungulates (Donázar 1993). The species exhibits complex social behaviour forming large individual congregations outside the breeding season at highly preferred feeding stations (e.g. farms) and nearby temporary roosting sites (García-Alfonso et al. 2020;van Overveld et al. 2020a). Altogether, these traits make it an ideal study species to assess how different migratory strategies coexist.
We firstly investigate the consequences of either residency or migration and aim to identify the factors that balance between the costs and benefits associated with both strategies. Secondly, we investigate the correlates underlying the partial migratory strategy observed in the Egyptian Vulture population overwintering in south-west Spain. We therefore hypothesize that (1) environmental (climate, food availability) and social (attraction of immature individuals to wintering sites by presence of adults) factors may influence the decision to migrate or not; (2) We expected that both migratory strategies yield different outcomes in terms of flight efficiency during winter; (3) that these energetic differences (if there are any) might explain differences in survival and reproduction; and (4) ultimately whether there is a fitness cost of migration and/ or residency.

Field monitoring data
We gathered data from censuses of the wintering Egyptian vulture population in Cáceres, Extremadura, Spain (see Supp. Mat. Appendix S1 for details). We accessed data from monitoring project, where the wintering population was surveyed twice monthly from December to February between 2014 and 2019 with adults and immatures classified according to plumage characteristics (more details in Morant et al. 2020). Surveys were carried out in one roosting site until 2014. From 2014 onwards, four additional wintering roosting sites were discovered and surveyed taking advantage of GPS tracking of some individuals that were tagged in 2014 (n = 5, see Morant et al. 2020 for details). It was not possible to record data blind because our study involved focal animals in the field.

Movement data
We used data from 39 Egyptian Vultures belonging to three different populations in the Iberian Peninsula, namely Extremadura, Duero/Douro and Castilla-La Mancha/ Valencia (see Supp. Mat. Fig. S1). Of the 39 birds, seven (n = 6 adults and n = 1 juvenile) were captured and tagged in Castellón and Guadalajara provinces (Spain) between 2007 and 2009 with a solar-powered GPS tag from Microwave Telemetry (Columbia, Maryland, USA), 19 (n = 11 adults, n = 5 subadults, n = 3 juveniles) were captured in Cáceres (Extremadura, Spain) between 2014 and 2020, and fitted with solar-powered GSM-GPS-ACC transmitters (n = 16) (E-obs GmbH, Munich, Germany) and Ornitela (n = 3) devices, while 13 were equipped in Duero/Douro (Bragança, Portugal; Zamora, Spain; Salamanca, Spain) with Ecotone-Skua (n = 3 adults, n = 1 subadults, n = 2 juveniles) (Ecotone skua), and Ornitela (n = 2 adults, and n = 5 juveniles) devices (see Supp Mat Table 1 for details). GPS fixes and associated data were acquired at temporal resolutions ranging from one location per 5 min to one location every 2 h, with dormancy periods during the night (from 22.00 p.m. to 4.00 a.m.) (see López-López et al. 2014;Phipps et al. 2019;Morant et al. 2020 for more details on the tagged individuals; see also Supp. Mat. Table S1). Individual ages at deployment were estimated in calendar years based on plumage traits of different age classes. We classified juveniles as individuals in the first calendar year, subadults as individuals in the second to fifth calendar year and adults as individuals in the sixth calendar year or older (Forsman 2016). The sex of individuals was determined by using molecular sexing techniques (Fridolfsson and Ellegren 1999).
All captured individuals were equipped with yellow and red alphanumeric plastic rings, metal rings and a GPS transmitter. All transmitters weighed 24-63 g, < 3% of body mass, which is below the recommended limits to avoid adverse effects (Bodey et al. 2018) and were attached using backpack or leg-loop harness systems. All the GPS and accelerometry data were automatically incorporated and downloaded from the online Movebank data repository (www. moveb ank. org; Wikelski and Kays 2019).
Birds were divided into "migrant" (n = 25) (i.e. birds that exhibited usual migratory behaviour overwintering in the African quarters; Supp. Mat. Fig. S2A; see for example García-Ripollés et al. 2010) and "resident" (n = 12) (i.e. birds that did not migrate and remained in the Iberian Peninsula during the study period; Supp. Mat. Fig. S3B; see Morant et al. 2020 for details). Two of the tagged birds, initially resident, exhibited migratory behaviour the following wintering season, when they left the breeding grounds to migrate to Africa. Therefore, we considered them resident during the period they remained in the study area and migrant after this period in our analyses (see also Supp. Mat. Table S2 for details of individuals used in each analysis).

Environmental and social correlates of migratory behaviour
Surplus food or other types of waste and carrion from human activities may promote congregations of scavengers (Plaza and Lambertucci 2017). However, vultures may just take advantage of the readily available food or, rather, gather at those sites for social purposes, which in the case of immature birds, may confer advantages in terms of shelter and crucial social information exchange from more experienced individuals that may ultimately affect survival (van Overveld et al. 2020a). In species such as Egyptian Vulture, immature individuals tend to abandon breeding quarters and remain in African quarters until reaching adult stage (5 years; Serrano et al. 2021), where individuals experience milder conditions besides higher food availability. Therefore, the decision to remain in the breeding quarters during winter might be influenced by a set of factors that ultimately could enhance individuals' survival (e.g. food abundance, presence of conspecifics and milder climatic conditions at the start of the migratory period). Here, we analysed the relationship between the number of immature birds during winter and different environmental and social factors to elucidate which factors most influence the number of immature birds and therefore may indirectly affect the decision to stay or leave. Among the environmental factors, we selected (1) temperature in the period of departure to African quarters (i.e. mean temperature between Sep-Oct), and (2) food abundance in the study area (e.g. livestock numbers). We assumed that higher livestock numbers are related to higher carrion availability (see for instance Arrondo et al. 2018), and that carrion is readily available to scavengers due to regulations approved a decade ago (EC 142/2011; see Margalida et al. 2012) which allow farmers to abandon livestock carcasses in 62% of Spanish territory, including our study area (see Morales-Reyes et al. 2017). As for the social factors, we selected the number of adults recorded in the surveys carried out between 2014 and 2019 wintering seasons (Nov-Feb). Temperature information was recorded for 2014-2019 (Agencia Estatal de Meteorología (2020)). Livestock numbers, including cows, pigs, goats and sheep, were obtained from the annual census conducted in the study area (data provided by the regional government, Junta de Extremadura).
We then investigated how the above-mentioned factors influenced the number of wintering vultures by computing Generalized Linear Mixed Models (GLMMs) with Poisson distribution error and log-link function by using "lme4" package (Bates et al. 2015). We entered the year and roosting site as a random factor in our model to account for the effects of a repeated census every year in each roosting site. We estimated marginal and conditional R 2 by using "piece-wiseSEM" R package to assess the models' overall explanatory power (i.e. for fixed and random factors). Prior to modelling, we computed Pearson correlations to avoid collinearity problems (Dormann et al. 2013) and dropped those variables showing > 0.5 correlation values. We therefore excluded sheep and pig abundance from our analyses due to their high correlation with cows and goats respectively.

Energetic consequences of migratory behaviour
We investigated the energetic consequences of adopting one migratory strategy or the other. In particular, we examined differences in percentage of time spent in costly activities such as flapping flight as a proxy of energy expenditure among migratory and resident birds.
We estimated the energy expenditure in two different seasons (non-winter: March-Oct, and winter: Nov-Feb) using the Overall Dynamic Body Acceleration (hereafter ODBA), calculated from the tri-axial accelerometry data (hereafter ACC) Gleiss et al. 2011). We used birds from which ACC data were recorded (migrant = 9, resident = 13). ODBA can be considered a proxy of energy expenditure since it is positively associated with oxygen consumption and carbon dioxide production (Wilson et al. 2006 and the mechanical work produced by muscles and internal organs Bishop et al. 2015). Furthermore, the integrative summary of energy expenditure from ODBA is even more effective when the parts with high-energy locomotion (e.g. flapping flight) are modelled separately from other behaviours (Duriez et al. 2014;Stothart et al. 2016).
ACC data were collected in bursts on three axes (Xsway, Y-surge, Z-heave) for a duration of 2-3 s every 5-10 min at 20 Hz from Ornitela and E-Obs devices, respectively (see Supp. Mat. Table S1). Firstly, we estimated the energy expenditure calculated as the average ODBA value for each burst of 2-3 s along the three axes (X, Y, Z). We transformed raw acceleration data into physical unit "g" (Laich et al. 2011) by using "moveACC" package (Scharf 2018). To this end, we assigned the calibration values of intercept and slope provided by manufacturers for each device type (Ornitela and E-Obs, respectively). We then estimated the average ODBA value (in gravitational units) for each burst. We finally estimated the mean daily ODBA by averaging the ODBA values per day, for each individual, year, month and day (see Wilson et al. 2020). Secondly, we estimated the flight type (flapping or non-flapping) for each burst, by extracting wingbeat frequency from the ACC data (O'Mara et al. 2019). We classified each wing beat frequency as flapping or non-flapping. We identified and removed outliers in wingbeats (being < 2 or > 6 beats per second), representing about 1% of all bursts classified as flapping (see also O'Mara et al. 2019 for a similar approach). Finally, we interpolated the spatial location of each ACC burst based on the closest GPS locations using the R package "move" (Kranstauber et al. 2020). We associated each ACC burst with the height above ground corresponding to the GPS location closest in time and we applied a height threshold of 100 m above ground to select our flapping locations (see Scacco et al. 2019 for similar approach). By doing so, we ensure that we exclude the flapping associated with taking off, assuming that above this height the birds were using flapping flight only in response to the absence of uplifts. We then calculated the percentage of time spent daily on flapping flight during non-winter and wintering seasons. We selected flapping flight since it is considered the costliest activity in soaring birds, given the disproportionate energy expenditure compared to all other behaviours (e.g. Williams et al. 2020). We estimated the percentage of time spent flapping by counting the daily flapping events for each individual from the former energy expenditure database.
To analyse whether there were differences in proportion of time invested in flapping flight among migratory and resident birds in wintering and non-wintering periods, we modelled the daily percentage of time spent flapping as a function of migratory behaviour (migrant and resident). Since we expect that the highest difference among both migratory behaviours would occur in winter, we also included season (wintering and non-wintering period) as a factor in our model. We used a beta regression model and considered the proportion of time spent flapping as our response variable. We ran a glmmTMB model implemented in the "glm-mTMB" package with beta family and logit link (Brooks et al. 2017). Individual identity and year were entered as random terms to account for the repeated measurements of the same individuals and within the same year. We also estimated marginal and conditional R 2 by using "performance" package (Lüdecke et al. 2020).

Breeding and survival consequences of migratory behaviour
We estimated the number of breeding years of each tagged individual (i.e. number of years that an individual had bred independently of the breeding output since the tagging date) and successful breeding years (i.e. number of years an individual had bred and raised at least one chick since the tagging date).
We only selected data from adult individuals (> 5 calendar years, n = 22) since subadult individuals do not usually breed (Serrano et al. 2021). We used data recorded during field monitoring during the breeding season for each individual and year. Individuals' nests were identified by using GPS locations and later confirmed in the field. Breeding status (i.e. breeding/non-breeding) of each individual was confirmed during April-June, when the tagged individual and its partner were observed copulating, arranging the nest and incubating. Breeding success of tagged individuals was confirmed when at least one chick was successfully raised at the end of the breeding period (August; see Morant et al. 2019 for details). In some cases, it was not possible to confirm neither the breeding status nor the breeding success of tagged birds in the field (n = 14 individuals/breeding events/ year). We then used the "nestR" package (Picardi et al. 2020) to identify (1) nest location when these data were not available for a certain individual/year; (2) breeding status; and (3) breeding success (see Supp. Mat. Appendix S2 for details). We first tested whether the package discriminated well using nests from individuals for which the nest location was known (n = 10) and we visually inspected whether the results matched the observed breeding nesting sites.
We used the breeding dataset to examine whether different migratory strategies exhibited differences in reproduction. In order to control for the effect of tracking years on the number of breeding years for each individual, we estimated the ratio between number of breeding years and total number of tracking years. Similarly, we estimated the ratio between the successful breeding years and cumulated number of breeding years to control for the effect of number of times that each individual bred. We then modelled the breeding/tracking years and the successful/breeding years ratios as a function of migratory type (migrant and resident) by running Linear Models (LMs). We ran separate models for each of the above mentioned parameters.
To construct our survival database, we gathered data on each GPS-tagged individuals' fate for each year since the individuals were tagged. We assigned a binary value as an event indicator, being 0 if an individual was alive at the end of each year until reaching the fixed study end date (28th February 2021) or the date of the last GPS location in case of transmitter failure. We assigned 1 if an individual had died within a given year. In such cases, we estimated the number of days that individual was alive until the last GPS location date when casualties occurred. Thereby, we were able to consider the change in the age-class transition between years (e.g. from juveniles to adults) and accurately estimate the annual survival for each individual and age class. In case there was no clear evidence of individual death (e.g. a picture of the dead individual or reliable information from collaborators or official entities), we could reliably separate deaths from cases of transmitter-failure based on three simple indicators extracted from the tags (see Supp. Mat. Appendix S3). According to the latter, we excluded six individuals from our analyses since we were not able to reliably ascertain their fate (i.e. whether they died or not), and therefore avoiding survival over/underestimation. Although we cannot rule out the possibility of some effect, we assume that tagging with the transmitters had a negligible impact on individuals' absolute survival (e.g. Sergio et al. 2015;Bodey et al. 2018;Buechley et al. 2021).
We evaluated the effect of migratory behaviour (migrant and resident) on annual survival by running Cox regression models with right-censored data (Pollock et al. 1989). We also entered age class (adult, subadult and juvenile) at the end of each year to account for the effects of age-specific variation on survival. Because for the same individuals we had different observations, we used the "cluster" function within our models to account for the non-independence of each observation. The model was fitted by using "coxph" function from "survival" package and results were plotted by using "survfit" function from the same package (Therneau 2018).
In the case of the breeding model (i.e. GLM), we explored the overdispersion of selected models using the "AER" package (Kleiber and Zeileis 2008) to ensure that our model did not violate the assumption of Poisson distribution (i.e. variance and the mean are the same). We also computed the overall explanatory power of the selected model (D 2 ) by using "modEva" package (Barbosa et al. 2015) to inspect the proportion of variation explained by our best models. In the case of the survival models, we checked the overall explanatory power by using the "Rsq.ad" function which measures the proportion of variance explained by the best models.
Spatial and statistical analyses were done in R version 4.0.0 (R Core Team 2020). In all analyses, we adopted a multimodel inference approach. We used the dredge function from the "MuMin" package (Barton 2020) to perform an automated model selection with subsets of the supplied 'global' model (i.e. model including all the terms). Model selection was performed by using "model.sel" function from the "MuMin" package by which all the models (including null model) were compared by using the Akaike Information Criterion (AIC; Burnham and Anderson 2002), corrected for small sample sizes (AICc). The best model was that with the lowest AICc value. We considered models within two AICc units of the best model (that is, the model with lowest AIC) as having similar support (Burnham and Anderson 2002). For the best model, homogeneity of variance and normality of residuals were inspected by using "ggresid" package to check the goodness-of-fit of our best models (Goode and Rey 2019). When there was no clearly supported model, we performed model averaging by using "model.avg" function from the "MUMin" package of those top-ranked models with Akaike weights > 0.001 (Burnham and Anderson 2002).All tests were two-tailed, statistical significance was set at p < 0.05, and all means were given together with standard error.

Environmental and social correlates of migratory behaviour
Our results showed that there was strong evidence of the relationship between the number of immatures and number of adults (Table 1; Supp. Mat. Table S3). The number of immature birds increased linearly together with the number of adults, explaining almost 38% of the total variance of the data (Fig. 1).

Energetic consequences of migratory behaviour
Overall, we found strong evidence of the effect of season and migratory type on percentage of time spent on flapping flight. We observed that resident birds spent more time on flapping flight compared to migrant birds (Table 1; Supp. Mat. Table S3; Fig. 2). Moreover, resident birds spent more time in flapping flight during the wintering season in the Iberian Peninsula compared to migrant birds (mean = 4.86%, SD = 2.77%, and mean = 2.95%, SD = 2.58%, respectively). Both resident and migrant birds exhibited similar investment in flapping flight during the non-wintering season (mean = 4.22%, SD = 2.79%, and mean = 3.40%, SD = 3.04%, respectively).

Breeding and survival consequences of migratory behaviour
Overall, our results showed weak evidence of the effect of migratory behaviour on breeding/tracking years ratio, with the null model as good as that including migratory behaviour (i.e. within the 2 AIC points). However, migrant birds bred slightly more often than resident birds (mean = 0.701 breeding/tracking years, SD = 0.365, and mean = 0.421 breeding/   Table S3). Similarly, no effect of migratory behaviour on breeding/successful breeding years ratio was found. Indeed, migratory and resident birds experienced similar breeding success (mean = 0.625 breeding/successful breeding years, SD = 0.370, and mean = 0.561 breeding/ successful breeding years, SD = 0.473, respectively) (Supp. Mat Table S3; Table 1; Fig. 3B).
We observed a total of 13 casualties of the tagged birds (33% of all tagged birds) at the end of the study period, of which 12 were migratory birds and one was a resident bird. The casualties occurred in Spain and Portugal (n = 2) and the rest of them in African quarters (n = 11). Our results showed moderate evidence of the effect of migratory behaviour and between age classes on annual survival probability (Table 1; Supp. Mat. Table S3). Resident birds exhibited higher annual survival rates than migrants for all age classes (mean = 0.9, SD = 0.09, and mean = 0.71, SD = 0.16, respectively; Fig. 4A, B). Moreover, annual survival of juvenile birds was lower (mean = 0.50, SE = 0.19) compared to subadults and adults (mean = 0.75, SE = 0.17, and mean = 0.79, SE = 0.1, respectively) (Fig. 4C).

Discussion
In this study, we disentangled the causes and consequences of migratory behaviour in a long-distance partial migratory bird, the Egyptian Vulture. We showed that social (attraction to conspecifics) but not environmental (food or climate) cues are the primary correlates of variation in immature wintering population size across time. We found that overwintering in the breeding grounds seems to be energetically more expensive than migrating. Overall, resident birds spent more energy during winter than migrants. Interestingly, we observed that migrants exhibited higher breeding activity (i.e. number of breeding years) than resident individuals, whereas residents showed higher survival probabilities than migrant birds. In summary, our findings showed that the optimal migratory strategy is context-dependent. Residency may be energetically costly but advantageous in terms of survival. On the contrary, migration is less costly and may allow individuals to invest more energy in reproduction which may explain the observed high breeding rates but lower survival probabilities. Therefore, our results suggest that the coexistence of different migratory behaviours may be balanced by a complex trade-off between optimal energy allocation and fitness shaped by mainly social factors.

The role of environment and sociality
Global change has become one of the most influencing factors on partial migratory species in recent decades (e.g.

Fig. 2
Differences between migrant and resident birds in percentage of time spent daily in flapping flight (as a proxy of energy expenditure) among migrant and resident birds during non-winter (in yellow) and wintering (in light blue) seasons. Shaded areas represent 95% confidence intervals Fig. 3 Predicted values of the significant variables included in the best models of breeding. A The ratio between breeding years and tracking years for each migratory type throughout the study period. B Differences between migrant and resident birds on successful/Breeding years ratio. The bars represent 95% confidence intervals Gill et al. 2019;Van Doren et al. 2021). Species rapidly respond to milder winters and increasingly predictable and available food pulses at higher latitudes by shortening migratory routes and even by remaining in the breeding areas during winter (Rotics et al. 2017;Arizaga et al. 2018;Haest et al. 2019;Nuijten et al. 2020; Riotte-Lambert and Matthiopoulos 2020 and references therein). However, our results showed that responses of birds are more correlated with proximate factors such as social cues rather than food or climate (e.g. temperature). Our findings suggest that adult Egyptian Vultures may attract immatures and less experienced individuals by remaining in a particular place where resources are abundant and predictable (Morant et al. 2020). This pattern has also been observed in White Storks (Rotics et al. 2016) and other sedentary populations of Egyptian Vultures where collective foraging in areas of high food predictability and availability (e.g. farms) could benefit individuals with lower social status, such as juveniles or subadults in finding food (García-Alfonso et al. 2020;van Overveld et al. 2020b). Additionally, immature individuals may also benefit from the size of the roosting site given the proportional relation that exists between the year (in days) since the first tracking day. Each step in the lines represents the death of an individual in each case. The shaded area represents 95% confidence interval roost size and individuals' fitness (Brouwer et al. 2020;Parker and Rubenstein 2020). Indeed, both factors may act ultimately enhancing survival on species with high immature mortality such as the Egyptian Vulture (Grande et al. 2009).
Despite we did not find any effect of food abundance on the number of immature individuals, it should be noted that these variations in abundance of carrion at small spatio-temporal scales (i.e. due to illegal releases of carrion carcasses by farmers) are difficult to infer from just variations in livestock numbers at coarser scales (e.g. García Alfonso et al. 2019). Therefore, it cannot be completely discarded that these unnoticed changes in food releases together with other specific geographic location, historical factors and habituation effects might also potentially play an important role (e.g. 'socially inherited routines', Andrew 2017) in attracting less experienced and immature individuals. Indeed, this pattern has been also observed in sedentary populations of Egyptian Vultures where almost all marked individuals within the population have been observed feeding at the same farms from an early age (van Overveld et al. 2020b). Finally, it should be noted that the small sample size and/or time window could have affected the observed results in our models and hence should be cautiously interpreted.

Impact of migratory behaviour on energy allocation
Interestingly, we observed that residency is more energetically costly than migration, particularly during the wintering season. These results are in contrast to previous studies showing that residency in breeding grounds in Europe could decrease energy expenditure and increase survival probability (see Flack et al. 2016;Rotics et al. 2017). We observed that resident individuals invested more effort flying on wintering grounds than migrants. In particular, they invested more time in costly flight types like flapping. These results suggest that resident birds have shorter days and higher thermoregulatory costs, so they need to ingest more energy in a shorter period of daylength, which might necessitate more flapping flight as it is faster to get from one food source to another (Pokrovsky et al. 2021). On the contrary, migratory birds experienced better flying conditions and can travel farther by using gliding and soaring flights which minimize travel costs (see for example Rotics et al. 2016). Therefore, our findings may indicate that the decision to stay offset the cost of migration with enduring harsher conditions and higher energy expenditure in winter (Rotics et al. 2018). However, residents may also compensate for the higher energy expenditure by reducing their wintering foraging areas and exploiting highly predictable food resources such as farms close to their roosting sites (Supp. Mat. Fig. S3. Table S4; Morant et al. 2020;Soriano-Redondo et al. 2021).

Fitness maximisation
Our findings, despite the weak evidence of migratory behaviour effect, showed a certain breeding advantage for migrant birds. However, it could be expected that resident birds could increase breeding performance due to their earlier access to best-breeding sites and earlier reproduction (Pulido and Berthold 2010). According to our results, it is also possible that in resident birds specific components of fitness are maximized (i.e. survival) at the expense of reproduction. Migrants, on the contrary, may allocate more resources towards reproduction while they are subjected to direct mortality costs of migration (Soriano-Redondo et al. 2020;Buechley et al. 2021). However, the observed similar breeding success between migrants and residents may indicate a clear advantage towards residents, since improved conditions during breeding season can result in better productivity for both migrants and residents, in addition to improved survival for residents (e.g. Griswold et al. 2011). In this context, maximisation of certain fitness components could occur if some individual traits such as physical condition are compromised due to unfavourable conditions experienced in winter (Chapman et al. 2011b).
Recent studies on fitness benefits of migration or residency strategies yield clear evidence that, in terms of survival, residency is more beneficial than migration (Buchan et al. 2020). Our results are in agreement with these findings and suggest that favourable climate conditions and year-round resource availability could contribute to the observed higher survival rates (e.g. Satterfield et al. 2018). In fact, we found that all age classes exhibited higher survival rates in residents as compared to migrants. Hence, residency could be particularly beneficial for subadult and juvenile birds that exhibited higher mortality rates associated with migration (Grande et al. 2009;Sánz-Aguilar et al. 2017). However, Buechley et al. (2021) found that migratory Egyptian Vultures experienced higher survival rates at African wintering grounds. These results may be explained by differences in how survival was estimated. Buechley et al. (2021) estimated the monthly survival of the same individuals, while our survival estimates were based on individuals that follow a particular behaviour (resident or migratory) in a given year.
The higher survival at non-breeding grounds, linked to the fact that subadult Egyptian Vultures usually spend the first years of life in African wintering grounds until reaching adulthood (Donázar 1993), showed that a certain parity might exist among survival rates among migratory and non-migratory birds that could contribute to the coexistence of both behavioural strategies (Gilroy 2017), or rather that higher energy allocation was in expense of survival and maintenance (e.g. thermoregulation) during winter in the Iberian peninsula.
Overall, our results showed a complex trade-off between survival and reproduction which could lead to a selection of an optimal strategy that maximises certain fitness components for migratory behaviour (Chapman et al. 2011b). These results may also indicate a negative frequency dependent selection of migration that could lead to a coexistence of individuals with different strategies (Dall et al. 2004). More importantly, they might indicate that migratory and resident species have different life-history strategies (e.g. migratory species live faster than resident ones) that promotes the coexistence of both forms (Soriano-Redondo et al. 2020). Nonetheless, it cannot be discarded that individuals may opt for one strategy or another according to the behaviour of others, their own condition and current ecological circumstances, which in turn are related to the space-time dependent fitness payoffs that could enable the maintenance of different behavioural types within populations (Gross 1996). The latter, might support the fact that, as we observed, individuals could have suddenly switched from being residents and start migrating in a given winter.

Conclusions
We showed that species-specific behavioural traits such as attraction to conspecifics and not environmental factor such as food or climate might influence the decision to stay during winter in immature individuals. We also observed that each migratory form may maximise a certain fitness component (i.e. survival or reproduction). Resident individuals may maximize their survival by exploiting predictable and easily available trophic resources (e.g. farms, López-López et al. 2014;Morant et al. 2020) that compensate for the higher energy cost of moving in unfavourable conditions during winter. Migrant individuals could benefit from more seasonal and unpredictable resources but better environmental conditions in African wintering quarters that improve their flight capacities (e.g. reducing flapping flight due to higher availability or thermals; Flack et al. 2016). In summary, our results reveal a complex trade-off between fitness components between migrant and resident behaviours which could offset the energetic consequences of selecting one strategy or the other. Taken together, these insights could help better understand coexistence of both migratory forms in partial migratory species.