Larger size and older age confer competitive advantage: dominance hierarchy within European vulture guild

Competition for limiting natural resources generates complex networks of relationships between individuals, both at the intra- and interspecific levels, establishing hierarchical scenarios among different population groups. Within obligate scavengers, and especially in vultures, the coevolutionary mechanisms operating during carrion exploitation are highly specialized and determined in part by agonistic behavior resulting in intra-guild hierarchies. This paper revisits the behavioral and hierarchical organization within the guild of European vultures, on the basis of their agonistic activities during carrion exploitation. We used a dataset distilled from high-quality videorecordings of competitive interactions among the four European vulture species during feeding events. We found a despotic dominance gradient from the larger species to smaller ones, and from the adults to subadults and juveniles, following an age and body size-based linear pattern. The four studied species, and to some extent age classes, show despotic dominance and organization of their guild exerting differential selection to different parts of the carrion. The abundance of these parts could ultimately condition the level of agonistic interactions. We discuss the behavioral organization and the relationship of hierarchies according to the feeding behavior and prey selection, by comparing with other scavenger guilds.


Methods
Study area and species. The study was carried out at four supplementary feeding sites (hereafter SFS) in southwestern Europe, two in the Pyrenean Mountain range (northeastern Spain, 42°25′41′′N 1°2′54′′E) and two in the Mediterranean Basin (central-western Spain, 39°10′55′′N 5°3′19′′W). The two regions were chosen to include representative populations of the four vulture species present in Europe 36 : the bearded vulture Gypaetus barbatus (BV) which is widely distributed across the Pyrenees; the cinereous vulture Aegypius monachus (CV) which is only common in central-western Spain; and the griffon Gyps fulvus (GV) and the Egyptian vulture Neophron percnopterus (EV) which are widespread throughout the two selected study areas 42 .
These species, which form the guild of European vultures, are all obligate carrion feeders and are physically adaptated to this food resource. They show different foraging strategies and carrion selection. The GV is the most gregarious species, attending in great numbers any feeding opportunity; this species prefers large amounts of carrion and entire carcasses of ruminants, mostly selecting soft tissues. The CV is the next most gregarious species and shows high numbers with small/medium-sized carcasses (rabbits or sheeps) or pieces of carrion, preferring tougher parts like muscles and tendons. The BV is a specialized bone-eater, mainly selecting extremities of different medium-sized ungulates. Juvenile and subadult individuals tend to concentrate around predictable sources of food, like supplementary feeding sites. Finally, adult EVs are non-gregarious visitors of both predictable and sporadic sources of carrion, mainly feeding on small peripheral soft tissues (skin, viscera, etc.) 18,34,36 . These species have differing population sizes, ecomorphological traits and territorial and migratory behaviors, which partly determine their conservation status and patterns of ocurrence while scavenging 43 . The GV is the most abundant species in Spain with more than 30,000 breeding pairs 44 , and overall consumes the largest amount of available biomass at feeding sites compared to other species 34,36 . In contrast, the CV and the EV are considered vulnerable, whilst the BV is endangered 45 . Study design and variables. We provided carrion on a monthly basis at each SFS on 100 occasions during two complete annual cycles from May 2009 to April 2011. Carrion was deposited during the day (randomly from 7:00 to 19:15 h GMT + 1) alternating different prey species (sheep Ovis aries, goat Capra hircus, pork Sus scrofa, cow Bos taurus, red deer Cervus elaphus and wild boar Sus scrofa), biomass (in kg) and distribution pattern (clumped and evenly spread) among the different SFSs [see details in 18,36]. Birds visiting the SFSs were monitored from the time of food provision until three days later using a high-resolution video camera (Arecont Vision Megavideo AV5100) connected to a computer, hard disk drive and rechargeable lithium batteries. The camera was located 20 m from the point of supply, allowing the observation of a wide area around the carrion (the 120° display angle of the camera also allowed the subsequent zooming of the recorded pictures with a sufficient sharpness to record species and behavior -see further details in 37 ). The recordings of each feeding event were later visualized using the AV program v.5.1.4.239 Application Manager (Arecont Vision, Glendale, California, USA) to register the activity of individuals of each vulture species. Video analysis was performed by the same person to avoid observer biases in the data interpretation.
We selected one individual corresponding to each species and age class attending the SFS for each feeding event so the total number of individuals studied per feeding input varied (mean 4.15 ± 2.70 birds studied/feeding event). We aimed to homogenize individual observations and ensure that the circumstances that led to competition for the resource were even throughout our analyses, so we established as a condition that at least 10 individuals of two different vulture species were present at the time of selecting the vulture species under study. Similarly, all observations of behaviors overlapped with sufficient presence of carrion, estimated at least 25% of the amount biomass initially provided. For the less abundant BV and EV we randomly selected the birds appearing at the SFS. For the most gregarious species -GV and CV-, as their numbers were usually rather great at feeding events, we chose birds exhibiting distinctive plumage or non-feathered parts as well as those marked with plastic rings or wing tags for easy recognition and to avoid losing sight of them in the crowd during their entire period of stay. This resulted in behavioral records of 415 individuals (99 BV, 93 CV, 214 GV and nine EV, see Table 1).
Data collection during the visualization of the video images began with the random selection of an individual. Its activity was then recorded from the time of its appearance until it disappeared from the visible recording area (this constitutes an observation in the data presented here). The duration of the continuous presence of the selected vulture in the recording area around the carrion averaged 22 minutes and 3 seconds ± 41 minutes and 43 seconds (minimum 4 seconds; maximum 4 hours, 26 minutes and 25 seconds). We first noted if the targeted individual interacted with another individual, to evaluate the interaction rates for each observation. For the purposes of our study, we considered an interaction as the direct action of a vulture to another vulture in which aggressive activity resulted from competition for food or another associated activity, such as the occupation of a certain place in the vicinity of the carrion (see Table 2). For each individual observed, we recorded its interaction with another individual, referred to here as the individual interacted, identifying their species and age classes and assigning the interaction to the categories of response variables indicated below ( Table 2). The interaction was considered complete once both individuals physically separated and the aggressive activity ceased. Each individual observed might perform several interactions during the same observation, 39 being the maximum recorded. We also recorded the interactions exhibited by other vulture/s directed towards the individual observed, to reveal their role as perpetrators or victims of the interaction.
Several other variables were sampled during the observational monitoring of each vulture on the basis of their linkage to agonistic behavior to, and dominance over, other individuals as well as to the competitive interactions observed 30,39 . The following response variables were included in the statistical analyses: Type of interaction, Result of the interaction and Role in the interaction (Table 2), following the categories and information used in previous studies 29,46,47 . We analyzed five explanatory covariates: species observed for the randomly selected individual, the age class of the individual observed, the species interacted (the individuals interacted by or to the observed vulture), www.nature.com/scientificreports www.nature.com/scientificreports/ the age of the individual interacted (Table 2), and the biomass in kg supplied as carrion during each feeding event.
We included this latter variable as a proxy of the size and availability of food, which in fact could modulate the type and results of interactions 18,23,35 .
The corresponding physical traits of the birds for denoting their age class and the assignment of the different behaviors were confirmed by zooming into the high-resolution images (at 5 Mpixels resolution). The individuals of the target species could not be weighed nor sexed due to the lack of observable sexual dimorphism in these species 48 . We consequently considered the body mass for each species following 49 and taking the lower value of the range for further analyses (7.0 kg for CV, 6.0 kg for GV, 4.5 kg for BV and 1.6 kg for EV, Table 3).

Statistical analyses.
First, we compiled a table with the totals and percentages obtained for each of the variables analysed ( Table 1).The database with all the observed interactions was analysed to determine the differences in agonistic behaviors between the various species and age groups. Aiming at recognizing what type of explanatory variables, and their interactions, influenced to a greater extent aspects of hierarchical dominance, we created three Generalized Linear Models (GLZ), one for each of the three categorical response variables considered (Table 2). We performed a Type 1 Likelihood Ratio (LR) test to find the log-likelihood value, the Chi-square statistic (χ 2 ) and the significance value (p) for each of the explanatory variables evaluated as well as for their interactions. All of the models developed included a logit function; the Type of interaction and Result of the interaction response variables showed a multinomial distribution of errors and the variable Role in the interaction had a binomial distribution. Because our models were based on categorical data and the limitations derived from our design were too large to compute reliable estimates, only the following three categorical explanatory covariates could be included in the models: species observed, age class of the individual observed and species interacted, together with the continuous covariate biomass. The models reported which of the explanatory variables, as well as the interactions between them, showed a greater likelihood that the observed data is most probable (log-likelihood) as well as which had a more robust level of significance (p). However, because the variable age of the individual interacted can influence the outcomes of paired interactions, individualized χ 2 analyses were performed for each of the three response variables related to age of the individual interacted.
We finally performed a General Regression Model (GRM) analysis to check the influence of body size and age on percentage success in the interactions recorded with the purpose of obtaining a global pattern between dominance rates and the size and the age of the individuals involved. For that, we considered the results of percentages of successful interactions of each group of individuals (Table 1) as a continuous response variable and the body mass and age class of the considered group as explanatory variables. The statistical analyses were conducted using Statistica 6.1 (StatSoft, Tulsa), and a standard p value of 0.05 applied.

Results
We recorded agonistic interactions with other individuals in 64.1% of observations of individual vultures (n = 415) taken from >7,500 hours of visualized video recordings ( Table 1). The total number of agonistic interactions registered were 1,049. The mean number of interactions in each observation was 2.5 ± 1.5; CV being the species with the highest number of interactions towards other species (4.8 interactions/observation), followed by GV, EV and BV ( Table 3). The majority of the interactions (65.0%) were intraspecific and the rest were between individuals of different species.

Variable Categories Description
Age class*  www.nature.com/scientificreports www.nature.com/scientificreports/ Regarding the type of agonistic interaction observed (see Table 2), the most frequent (58.8%) was fights/ attacks against other birds. In contrast, food stealing was rarely observed (5.0%), and was mainly performed by GVs (Table 1). The type of interaction observed was best correlated with the species observed (Log-likelihood = −797.30; χ 2 6 = 29.38; p < 0.001), followed by the age class of the individual observed (Log-likelihood = −784.91; χ 2 4 = 24.77; p < 0.001) and the interaction between the species observed*species interacted (Log-likelihood = −767.07; χ 2 8 = 15.85; p = 0.044). The age of the individual interacted (χ 2 2 = 54.44; p < 0.001) as well as the biomass provided at each feeding event (Log-likelihood = −811.99; χ 2 2 = 19.63; p < 0.001) also modulated the type of interaction. The BV instigated approximately the same number of fights/ attacks as displacements, whereas CVs and GVs performed a higher proportion of fights/attacks. Displacement was the most common type of interaction provoked by EVs (73.3%, Fig. 1). With respect to the different age-classes, in all of the species both adults and subadults showed a similar ratio between the types of interactions instigated (around 55% of fights/attacks and 40% of displacements). In contrast, juveniles were proportionally more aggressive towards other species, their percentage of fights/attacks being 68.2%. When acting as victims, non-adult individuals received a higher percentage of fights/attacks (84.8%) and displacements (64.6%) than the adults, and a similar ratio of stealings (51.1%). It is noteworthy that stealings always occurred at the intra-specific  Table 3. Ranks of the different vulture species and age classes based on the percentage of successful interactions, and body mass (in kg 49 ,) and age class 48 .  Table 3). The next highest were shown by adult GVs and juvenile CVs, while the lowest were shown by adult and subadult BVs and EVs, which therefore showed the least dominance ( Table 3). The success/failure ratio also varied depending on the interacting species. The percentage of success in conspecific confrontations was similar for BVs, CVs and GVs (around 60%, Fig. 3); that is, when an individual of these  www.nature.com/scientificreports www.nature.com/scientificreports/ species instigated an interaction. At the heterospecific level, CVs launching an interaction obtained high levels of success against BVs and GVs, while GVs and BVs only won more than 60% of the observations when competing with each other. The EVs observed were generally unsuccessful against all other species (Fig. 3). In relation to age class, adults obtained the highest rates of success, especially when facing juveniles (83.3%) and subadults (80.5%). In contrast, subadults and juveniles showed lower success rates, especially when competing with adults (38.4% and 45.8% respectively). The general success percentage decreased to 20.7% when the age of the species interacted was adult.
Only the species observed had a significant influence in relation to its role as an aggressor or victim in an aggressive interaction (Log-likelihood = −508.40, χ 2 2 = 72.39, p < 0.001). CVs were the most likely to initiate aggressive encounters (71.9% of their interactions were as aggressors) compared with the next highest GVs (52.3%), while EVs and BVs were most likely to be the subject of aggressive behaviors (91.6% of and 62.5% of their interactions, respectively, Fig. 4). In contrast, there were no significant differences between the different age classes in their role as aggressor or victim (Log-likelihood = −508.66, χ 2 2 = 1.63, p = 0.441). The summatory model explaining the success rates of competitive interactions (Table 1) was positively related to increasing body mass and older age classes of vultures, as well as to the interaction of both variables (R 2 = 0.98; p = 0.006; Table 4).

Discussion
Behavioral organization within the guild. In our study, the agonistic behaviors around carrion were grouped into three distinct types according to the level of aggression. The most frequent and aggressive were fights/attacks in which the vultures used their beaks, claws or opened wings to attack or try to attack, either to displace a competing vulture or to maintain their own position. In this type of interaction, physical contact was observed and serious, possibly even lethal, wounds might be inflicted 32 . Juvenile vultures of the studied species engaged in such fights more frequently both to con-and heterospecifics, possibly either because of their lack of experience and, consequent lower appreciation of the risk of fighting, or because their age class is more susceptible to starvation/dehydration 50 and is prepared to take greater risks. Displacements, which comprised more than a third of the interactions observed, were less aggressive in nature and were more commonly performed by the www.nature.com/scientificreports www.nature.com/scientificreports/ less belligerent species. Finally, food stealing attempts were considered as an independent category of interaction since they concern kleptoparasitic behavior, common among obligate and facultative scavengers [51][52][53] . The rarity of food stealing from individuals of other species might indicate pronounced resource partitioning and differential carrion selection behavior between the four species 18,36 .
Our results showed that among the four European vulture species, the CV is dominant: they exhibited the highest agonistic behavior success rates, had a greater number of interactions per observation, were the most frequent aggressor and most often showed fights/attacks behaviors. CVs, especially the adults and subadults, have adaptations which make them especially effective against other species and age groups. Their greater body size and more powerful beaks and claws enable better competitive performance against the other vulture species. Likewise, they have a number of displays and postures not present in other species, e.g. bristling of the plumage and behaviors such as the "threatening march" 24 .
The GV is next on the hierarchical scale with the second greatest number of successful confrontations, especially in the case of adults 27 . The GV´s second rank in the interspecific hierarchy is probably also related to its body size and common behavior of scavenging at gatherings, so that individuals can gradually gain experience in the numerous struggles against other individuals to obtain limited food. The most commonly observed GV behaviors were fights to maintain a preferential position close to the carrion -where most biomass is available 23,27and to defend individual pieces of food from stealing by other GVs and CVs.
Regarding the BV, juvenile individuals were more competitive than adults and subadults, and even more successful than juvenile and subadult GVs, showing the bold and social nature of that age class 49 . Juvenile BVs are more common at supplementary feeding sites (SFS), with their predictable food supply, compared with the territorial adult population 36,54 , which could explain their greater number of interactions per observation and their effectiveness in clashes with other vultures, especially other subadult and juvenile BVs and juvenile GVs. Finally, the EV was confirmed as the most passive of the guild since most of its interactions resulted from the aggression of the other species, especially GVs. Its smaller relative size, lower number at SFS and its means of food acquisition (see below) may explain its low success rate in confrontations.

Relationships between hierarchies and feeding behavior.
A species´ feeding behavior and the degree to which it is used to dealing with other scavengers at crowded carrion sites could in part explain the frequency (mean number of interactions/observations), types and results of interactions found in the vultures in our study. First, we have verified how carrion availability determines both the types of interactions and their result. When there was less food, more aggressive interactions (thefts and fights) happened due to increased competition for limiting resources (Fig. 2), which was also influenced by the number of birds present [34][35][36] . In this sense, as our findings show, the species most commonly observed in this competition and that monopolize most of the resources (i.e. the GV 34,36 and also the CV), would have a greater probability of success when resources are limited.
The CV frequently attends feeding events both at SFSs and naturally occurring carcasses, but is less frequently found in the scrums and clamours formed by numerous GVs around a carcass. On the contrary, CVs are more competitive in exploiting separated parts of a carcass when the number of GVs is lower 36 . It is also common for them to seize a small or medium sized isolated piece of carrion which they eat and defend 55 . CVs usually win agonistic encounters launched on other, preferably single, vultures in order to snatch small carcasses or pieces of muscle, skin or tendons 36 . This more aggressive and successful behavior would enable CVs to select and obtain choice carrion parts, which could be in limited supply in natural large cadavers (e.g. large herbivores) or in the piles of carrion provided at SFSs.
The GV is the most abundant vulture at scavenging events in Europe and takes the majority of the available carrion biomass overall 34,36 . It prefers the most abundant parts of carcasses such as the viscera and muscles, so large groups gather at large herbivore carcasses, favored by social facilitation (i.e. the processes in which conspecifics provide information about the position of food; 8,9,35 ). Its high position in the hierarchy at feeding events allows it to occupy the most favourable positions around a carcass and so to have the best access to the soft parts of the animal 35,36 . It also competes with other species for the scattered pieces of muscle/tendon that are also sought by CVs, and CVs are at a disadvantage if the numbers of GVs around those parts are too numerous.
BVs are specialist bone feeders 18,56,57 and >3/4 of their interactions observed in our study were directed towards conspecifics. Juveniles and subadults spend twice as much time as adults scavenging on the ground 18 and showed greater agonistic activity with the purpose of obtaining the most appropriate size and shape bones such as ribs, vertebrae, legs and hooves 57 . In contrast, adults usually locate and select the most favourable bones while still www.nature.com/scientificreports www.nature.com/scientificreports/ in the air and then carry them to other places outside the crowd once they are taken 18 . This practice of swooping down to pick up a bone before flying off with it would certainly contribute to their low observed interaction rate and the heterospecific interactions that did occur were mainly due to attacks by GVs that had detected active BVs.
The EV is the smallest and least dominant of the European vultures and minimises its interactions with other species at feeding events by almost always taking up a position at the edge of the feeding crowd and consuming small pieces previously discarded by GVs and CVs 18,36,58 . EVs rarely provoked interactions but suffered aggression from heterospecifics trying to displace them to take pieces of scattered carrion. Despite these apparent competitive disadvantages at SFSs and other feeding points attended by many vultures, EVs positively select these events and locations 34,36 because the resource availability is higher and more predictable, thus potentially improving their chances of survival 59,60 . Comparisons with other scavenger guilds. The establishment of hierarchies during carrion exploitation has also been shown in many other scavenger species. Regarding facultative carrion-eating birds, hierarchies in favour of larger species have been observed in some scavenger communities 61 . At the intraspecific level, older, larger males tend to be dominant, such as in the bald eagles Haliaaetus leucocephalus 50 and carrion crows Corvus corone 62 , while in other cases such differences have not been observed (e.g. in the sea eagle Haliaeetus albicilla 63 ).
Obligate scavengers show clear dominance hierarches and the behavioral patterns observed in our study of European vultures resemble those found in the vulture guilds of the Americas and Africa. Body size as a determining factor in the success rates of competitive interactions shows a similar pattern in American vulture guilds as in European ones; larger species are progressively dominant over smaller ones (i. e. Andean condor Vultur gryphus > king vulture Sarcoramphus papa > black Coragyps atratus/turkey vulture Cathartes aura > lesser vulture Cathartes burrovianus 28,39 10,26,[64][65][66]. When considering body size within the same species the largest individuals also dominate smaller ones, as seen in the Andean condor and the turkey vulture 29,30 . We were unable to examine the effect of differing body size in conspecifics in our study due to the lack of visible differences in body size between the different age or sex classes of the species we studied 48 . The patterns of a higher hierarchical position for adults over preadult age classes obtained in our study resembled those of New World vultures 28,30 . Moreover, sex was also a key factor in Andean condors 30 . However, the situation of European vultures could differ since, unlike Andean condors, they have a moderately reversed sexual dimorphism (i. e. females are slightly larger than males 43 ), so females could increase their competitiveness against males following the global pattern of positive dominance related to body size 60 .
In general, we would expect the dominance ranks among the different vulture guilds to be similar. First, Old World species would form three Genus groups based on phylogenetic similarities: Aegypius/Torgos/Trigonoceps/ Sarcogyps > Gyps > Neophron/Necrosyrtes/Gypaetus 67 . Each of the three proposed groups have similar morphological and behavioral traits during carrion exploitation so they occupy a similar hierarchical position in all the areas in which feeding behavior has been studied 16,24 . Similarly, the different European vulture species have trophic niches similar to those of New World species, both in diet selection and in their behavior during feeding: CVs are similar to Andean Condor, and GVs are similar to black and turkey vultures in this respect 30,39,68 . Implications on community ecology. The existence of a hierarchically structured organization in animal/ plant guilds linked to a specific resource shows the complex framework of social relationships as well as competition and facilitation processes 1,69 . The study of vultures allows the direct observation of agonistic behaviors around a trophic niche, such as carrion, offering a great energy input and appearing scattered in both space and time naturally. But this is not the usual pattern of availability of food in the wild for other niches, which usually occurs at broader scales 70 , so that intra-and interspecific competition often have to be evaluated through indirect methods, rather than individual behaviors observed directly through video recordings.
Our results reinforce a common behavioral pattern also detected in other groups of animals competing for a limited food resource: more experienced individuals with better physical condition occupy areas of optimal availability and quality of resources while suboptimal patches are exploited by younger and less competitive individuals 62,71,72 . This triggers relevant effects on their fitness, especially in terms of survival and breeding output, favouring the older and larger individuals 73,74 . In passerines, this has been shown, for instance, at wintering locations where both migratory and sedentary birds of the same and different species are present so that local, more experienced and better body conditioned individuals are more abundant in areas with a greater abundance of fruits of temporary occurrence 75,76 . Moreover, predatory birds and mammals compete for the occupation of territories with better prey availability and quality 77,78 . Thus, they tend to exhibit aggressive behaviors to expel other potentially competing individuals through struggles and exhibitions that are more frequently won by experienced and larger individuals 79 . In general, this selection by territorial species towards the best feeding and, consequently, high quality breeding habitats generates better breeding performance. Although it can also provoke a reduction in territory size and a greater investment in vigilance and aggressive interactions which, sometimes, leads to density dependence phenomena affecting demographic parameters 80 .
In relation to the above, and according to our results, animal communities find a balance between agonistic behaviors and resource partitioning. Competition and aggression establish dominance structures at the intraand interspecific levels but their subordinate effects are attenuated by a differential selection towards different parts of the resource allowing the exploitation of separated ecological niches [81][82][83] . This has been previously studied in nested communities and assemblages, both in different habitats and for taxonomically related guilds [84][85][86] . Consequently, adaptations for exploiting the same resource generate a dynamic balance in animal communities.

conclusions
We provide the first detailed observational study of the relative agonistic behavior of the four European vulture species, which result in despotic dominance patterns during carrion exploitation. Our results describe their species specific behaviors based on quantified and standardized annotated criteria and the hierarchical classification between individuals of different species and age classes. We show a general dominance gradient of body size (larger species to the smaller ones CV > GV > BV > EV), and age (from the adult age class to subadults and juveniles) as the outcome of agonistic encounters between European vultures. Aspects of dominance according to sex and the number of individuals of each species present have not been evaluated. However, according to phenotipic correlates of dominance, in European vultures, as moderate reversed dimorphic species, females could be dominant over males. Complementarily, intraspecific hierarchy seems to be also determined by behavioral factors such as hunger 64,65 .
In summary, the four vulture species studied, and to some extent their age classes, showed differential feeding behaviors that determined the despotic dominance and hierarchies formed during carrion exploitation. The different species selected different types of carrion, depending on the abundance of the available parts. Accordingly, food preferences and availability could also determine the level of agonistic interactions observed.

Data availability
Data in which the results of this article are based on is provided within the Dryad Digital Repository (https:// datadryad.org/review?doi=doi:10.5061/dryad.52m3q03).