Understanding sociality and behavior change associated with a nesting event in a captive flock of great white pelicans

Zoo ‐ housed pelicans are commonplace, but their breeding record is poor and little research is published on the activity patterns, as potential predictors of nesting, of captive flocks. Existing literature shows that comparative research can provide useful information for husbandry and conservation planning for pelican populations. The opportunity arose to investigate the time ‐ activity budget and social network of a breeding flock of captive great white pelicans. Three chicks were hatched in June and July 2016 and one in March 2017. Data on state behaviors, space use, and association preferences were collected around these nesting events, from October 2016 to February 2017 and July to October 2017. Results suggest that pre ‐ nesting periods were associated with heightened flock ‐ wide vigilance, suggesting that vig-ilance may be a precursor for courtship or nesting activity. Social network analysis revealed nonrandom associations between birds and a social structure across the flock, in which subadults seemed to associate more with each other than with adult birds. A limited visitor effect was noted; whilst no overall behavior change was apparent with different numbers of visitors, pelicans did widen their enclosure usage with increased visitor presence. These data are relevant to those attempting to breed this pelican, who wish to know more about the daily behavior patterns of this species across the season and physiological state, and who wish to understand pelican social structure, which is useful to the planning and implementation of bird moves or changes to the social environment of the flock. Further extending such research to include uninterrupted observation over a successful breeding event is recommended.


| INTRODUCTION
Behavioral studies are excellent ways of advancing animal husbandry for ex situ populations by providing evidence of individual and population responses to captive care (Hosey, 1997;Melfi, 2005). Focused research enables the development of best practice guidelines to uphold positive welfare states for specific species (Fidgett & Gardner, 2014;Tonkins et al., 2015;Troxell-Smith & Miller, 2016).
For many species of frequently housed zoo animal, an up-to-date research focus may be lacking. Pelicans (Pelicanidae) are excellent examples of common zoo birds that appear to receive little research attention (Danel et al., 2020) compared with other popular zoo birds such as penguins (Sphenisiformes). Current (March 2020) global pelican holdings for institutions that provide data to the Zoological Information Management System database (ZIMS) show that, across all pelican species, 3545 birds are listed (ZIMS, 2020). Despite their popularity with visitors and representation in many zoos globally, ex situ populations of pelicans may not be fully sustainable, with some flocks achieving only sporadic breeding success. Even in wellestablished flocks, poor breeding results caused by infertility, trampling of eggs and limited display of breeding activity (Ober & Verkade, 1998) can be noted.
The great white pelican (Pelecanus onocrotalus), hereafter referred to as "GWP," is a particularly good example of the need for more study in population demographics and potential husbandry issues. Data in the Zoological Information Management System (ZIMS) database shows a population of 1693 animals globally as of March 2020 and a review of these data identifies many aging birds and few regularly breeding flocks (ZIMS, 2020), which may suggest a decline in captive numbers in the future. Poor reproductive performance seems a feature of pelican colonies and Dathe (1962) notes that breeding attempts in the past may have been hampered by unreliable literature that was used to guide efforts. Many decades later, captive pelican research is still limited in number, and evaluation of bird activity patterns in zoological collections is hard to find. Older reports also state that empirical data on the behavior of wild GWP be rarely collected and published (Brown & Urban, 1969) and this may be being rectified as behavioral ecology papers from this century are noted (de Ponte Machado, 2007;Izhaki et al., 2002); a need for increased research activity to inform conservation and management (Megaze & Bekele, 2013) is noted.
In Europe, the largest nesting colony of GWPs is in the Danube Delta in Romania, where breeding takes place over May and June (Marinov et al., 2016), which is similar to the timing of breeding in other temperate regions (Hatzilacou, 1996). In Africa, where the largest population of this species is found (BirdLife International, 2018), GWPs can breed all year round provided suitable nesting and feeding grounds are available (Brown & Urban, 1969).
Wild GWPs can be erratic in their breeding attempts and breeding colonies can be susceptible to disturbance and nest failure (Bowker & Downs, 2008). The social nature of GWPs influences their key state behaviors, with foraging, roosting, migration, and maintenance (as well as nesting activities) occurring communally (Crivelli et al., 1997;Elliot, 1992;Hatzilacou, 1996;Saino et al., 1995). As a behavioral study of pelicans is noted as being useful to population management planning for both wild and captive flocks (Gokula, 2011), the aim of this study was to use a multiple methods approach to investigate the behavior and enclosure use of a breeding flock of GWP under human care. We collected data on time-activity budgets, enclosure utilization, and the social network of these GWPs T A B L E 1 Great white pelican population information to provide information on the bird's welfare state and to aid identification of potential behavioral triggers of reproduction.

| MATERIALS AND METHODS
Data were collected at Blackpool Zoo during two research periods running from October 2016 to October 2017 on 13 GWPs ("Obs 1") and 14 GWPs ("Obs 2"). Observations ran in two blocks: 24th October 2016 to 28th February 2017 (Obs 1), and then from 14th July 2017 to 27th October 2017 (Obs 2). Observation periods were based around the availability of the two students who collected data, and it was attempted to gather data as close to breeding events as possible within the constraints of the academic year. GWPs hatched in summer 2016 were fledged juveniles when Obs 1 data collection commenced, and the chick hatched in spring 2017 was a fledged juvenile for the start Obs 2 data collection. In total, 18 days of data collection were included in Obs 1 and 19 days for Obs 2.
Most birds wore plastic leg rings for individual identification.
Un-rung birds were identified from distinguishing features (e.g., bill or leg color). Details of the sample population are provided in

| Bird husbandry
GWPs had ad lib access to all the exhibit during the observation period, including indoor housing. Birds were fed twice daily, during the morning and afternoon, with whole fish in Zone 2 of the F I G U R E 1 Schematic representation of the great white pelican enclosure including size and description of zones used to determine overall enclosure usage by the flock enclosure ( Figure 1). A mixture of handfeeding and scatter feeding was used to ensure all birds received adequate daily food intake.
During the nesting period, keepers avoided approaching the nesting area except to conduct health checks.
For each day of study (at the start of each hourly observation, association data were recorded. Associations were defined by individual bird proximities based on one beak length from a neighboring bird. A chain rule method (Croft et al., 2008) was used to determine group membership; a chain rule denotes that two individuals are associating within a specific group based on their direct connections as well as the connections each has with others around them. As all individuals could be identified at every observation, the Simple Ratio Index (Cairns & Schwager, 1987) was used to calculate association rates for each bird. A total of 2663 identifications were made across the entire period of study.

| Enclosure use
The total area (m 2 ) of the GWP enclosure was calculated using Google Earth Pro v. 7.3.2.5776 (Google, 2019). The enclosure was further divided into nine zones ( Figure 1) based on accessible resources and their size (m 2 ) was also calculated via Google Earth Pro. For each day, at the start of each hour observation period, the zone occupancy for all pelicans was recorded. To assess the evenness of the flock's enclosure zone occupancy, a modified Spread of Participation Index (SPI) was calculated (Plowman, 2003). The SPI formula compares an expected frequency of zone occupancy with an observed frequency and provides a value between 0 (equal usage of all zones) and 1 (biased usage of one zone).

| Data analysis
To compare the daily SPI value for enclosure zone occupancy in the second observation period, a one-sample t-test was run on these data using the overall mean SPI value for the first observation period. To determine any influence of visitor number, temperature, and humidity on pelican behavior, a repeated measures model was run in RStudio using the "lmertest" package (Kuznetsova et al., 2016). Date was blocked as a random factor and the ANOVA (model name) function was used to evaluate each of the predictors inputted into the model (visitor category, temperature, and humidity) on performance of state behaviors for each observation period (alert and maintenance). The same modeling approach was used to assess the influence of these predictors on pelican enclosure usage (based on SPI value). The fit of each model was checked using the plot(model name) function of the residuals in RStudio. Where appropriate post hoc testing was conducted in RStudio using the "pbkrtest" and "lsmeans" packages (Halekoh & Højsgaard, 2014;Lenth, 2016).
To further understand any relationship between visitor number and pelican space usage, a negative binomial regression was run using the "Mass" package (Venables & Ripley, 2002) and

Incubation
The pelican sits on a nest. The pelican may be sitting on top of an egg or hatchling.

Foraging
The pelican searches for and consumes food. The individual may attempt to find food by dipping its bill in water.

Maintenance
The pelican engages in preening, wing flapping, sunbathing, or dust bathing activities.

Moving on land
The pelican uses its legs to walk across a solid substrate.

Rest
Sitting or standing motionless, with head facing forward or resting between wings in roosting position.

Social
Individual engages in behavior that is directed toward another pelican. Behaviors include bill clasping, chasing, biting and calling.
compared to the stocking density of each zone (population divided by the area of the zone).

| Social network analysis
Social network analysis was conducted in Socprog v. 2.9 (Whitehead, 2019). Values for social differentiation, the coefficient of variation of the true association index (AI) (Whitehead, 2008) were calculated using the likelihood method to determine the degree of homogeneity of the network. Whitehead (2019) (Whitehead, 2008).
Permutation tests to determine the number of preferred and avoided dyads in the network were run in Socprog using the "test for preferred/avoided associations" function (Whitehead, 2019). The number of trials was set at 1000 and the number of permutations increased from 1000 to 10,000 until the coefficient of variation p value (critical level set at 5%) stabilized, as per Whitehead (2019). Calculated p values for the list of preferred and avoided dyadic associations provide confidence in the accuracy of this identified pairing.
To identify predictors of pelican association, the "multiple measures analysis" function of Socprog was used and Multiple Regression Quadratic Alignment Procedure (MRQAP) testing was conducted (Whitehead, 2019). The flock's association matrix was uploaded as a Matlab file into Socprog and Socprog also converted specific attributes of the birds (sex, age, source of the bird, relatives in the flock, and years in the zoo) into a matrix of association measures using the "association measure from supplemental data" function. MRQAP tests were run for 10,000 permutations.
To determine a similarity between the associations present in There is no influence of enclosure zone on occurrences of pelicans being social and this is not influenced by the maximum stocking density of each enclosure zone (F 1,5 = 0.73; r 2 = 12.7; p = .433). Table 3 shows the most "popular" zones that groups of associating GWPs were observed within.

| Network structure and social assortment
Within this GWP network, the mean associations per dyad, i.e. the number of sampling periods in which a dyad was associated, averaged over all dyads (Whitehead, 2009) The estimate of the correlation between true and estimated association indices using likelihood was 0.928 (±0.224), which provides excellent confidence in the calculation of social differentiation.
Consequently, the network of these GWPs was not homogeneous (calculated social differentiation just greater than 0.3) but overall, birds were relatively loosely associated.
Permutation testing calculated an expected number of significant dyads (if this network was random) to be 4.55 but an actual F I G U R E 5 Daily cumulative number of birds seen in the zone closest to visitor areas (white bar) and in the breeding area (gray bar) compared against total daily visitor number (black line and marker). Associations remained stable over the two study periods, with a highly significant correlation noted between association matrices from before and after the 2017 nesting period (Mantel Z-test, r = .744; p < .001).

| DISCUSSION
Overall, our results showed that these pelicans used their enclosures unevenly, but SPI alone cannot be used reliably as a predictor of breeding activity for this species. The flock showed several key differences in time-activity budget between the first and second observation period, suggesting behavior change to be a good predictor of breeding activity. Social network analysis revealed the presence of very strong dyadic associations ( assortment between pelicans in a flock. This study shows the usefulness of a range of methods for understanding animal behavior and exhibit usage to gain an insight into the breeding activities of captive wild animals.

| Influences on behavior and exhibit use
Across the study period, high SPI values (varying between 0.6 and 0.9) were regularly calculated, suggesting that pelicans favored specific enclosure zones over others (Figure 4). Such a finding is similar to the uneven enclosure usage published in research on another colonial waterbird, the flamingos, Phoenicopteriformes (Rose et al., 2018). This unequal zone usage maybe occurring due to the gregarious nature of pelicans when performing key daily activities; breeding, maintenance, and feeding behaviors are all conducted as a social group in wild GWPs (Dathe, 1962;Megaze & Bekele, 2013;Saino et al., 1995). In the wild, GWPs will also flock together outside of the breeding season, often roosting communally (Elliot, 1992;Hatzilacou, 1996), therefore enhancing the chances of social transmission of information (perhaps useful for increasing the foraging or nesting success of individual birds) between GWPs in their flock (Danel et al., 2020). The strong bonds present in this GWP network as well as the high degree of observations of the birds together in specific enclosure zones throughout all seasons demonstrates a similar pattern of social choice as noted in the wild.
Whilst temperature and humidity were not significant predictors of pelican enclosure use, visitor number was, with wider enclosure usage (lower SPI values noted when more visitors were present). However, Figure 5 shows a complicated interaction between specific zone occupancy and visitor number. Pelicans may use their island more when visitor number increases, and this is likely influenced by other factors (such as the draw to use the island for socializing and reproductive behaviors) but this is worthy of further investigation. No relationship between visitor number and the occupancy of the zone closest to the public is noted. Zone 2 was the feeding area, clearly a draw to this area of the exhibit for the birds. It may be that pelicans are likely to enter this zone in readiness for feeding but may move away once feeding has been completed, irrespective of the potential size of an audience at this exhibit areas. Large groups of people can be intimidating to pelicans (Brown & Urban, 1969) so monitoring of the flock's activity is recommended at busy periods. An animal's overall behavior pattern may not be affected by increased visitor presence in exhibits where individuals have the opportunity to move away from the main areas of visitor congregation (Learmonth et al., 2018). This GWP flock may be an example of this, in that these birds avoid public proximity during periods of greater visitor intensity so therefore no overall significant changes in behavior occurred. It is likely that the large exhibit size and stable flock structure (based on the identification of key pelican dyads and the Mantel tests between years) facilitated this, thereby acting as a buffer against any negative visitor effects. This finding is useful for pelican keepers to consider, as the propensity of wild GWPs to abandon nesting colonies or suffer reproductive failure due to nesting disturbance is high (Bowker & Downs, 2008).
GWP husbandry may also influence space usage as the pelicans were normally fed in Zone 2 (the land area nearest the visitors).
Often, individual birds were fed by hand, to ensure all animals received food containing a fish-eater pellet (L. Forster, pers. comm, 20/08/2016). Birds that were satiated postfeeding may then move to other, quieter areas of the enclosure to rest and digest their food, as is noted in the activity patterns of wild American white pelicans (P. erythrorhynchos) that used roosting areas distant to human presence (Bunnell et al., 1981). Likewise, visitors may be attracted to the pelican enclosure to see the feeding and therefore any relationship between zone occupancy and visitation may be spurious; the lack of behavioral change associated with increases in visitor number suggests no apparent welfare compromise experienced by these GWPs of being on display.
Zones 7, 8, and 9, were rarely used by GWPs, with birds observed in these locations only sporadically. The resources provided in these zones may have resulted in their avoidance; the woodland section does not replicate the wild habitat of GWPs (Birdlife International, 2018) and the water and land zones (8 and 9) were elevated, which may have been inaccessible to these birds who were flight restrained. The function of all exhibit zones should be considered from an ecological perspective (Rose & Robert, 2013); this study shows that woodland is rarely used by these GWP and that steep slopes may be challenging for these birds to negotiate and this information on enclosure zone occupancy is helpful to evidence replication of ecologically relevant environments in future exhibit developments.
Considerable changes to the activity budget of these GWPs occurred between the two observation periods, with birds spending considerably more of their time being alert in Obs 1 period and comparatively more time resting and swimming during Obs 2. The higher proportion of alert behavior during the first observation period may be related, in part, to breeding as alertness can precede group courtship events amongst pelican species (Dathe, 1962). These GWPs may be more alert before courtship commences to ensure that the environment is suitable and safe for the rearing of young, and this has been detected in these behavioral observations. Extending this project, to cover an uninterrupted period of time before, during and after a successful nesting event would add more evidence to the indicators of breeding as noted from our study.

| Network structure and social assortment
The  (Crivelli et al., 1997) suggests that nonrandom associations are likely to exist in the wild too. The zoo-housed GWPs of our study were more often in close proximity to one another compared to being seen alone; this may be a factor of enclosure design, or how the enclosure has been zoned for this study, or (as expected) the colonial nature of this bird (Elliot, 1992;Hatzilacou, 1996). The lack of relationship between each enclosure zone's stocking density and the occurrence of socializing pelicans within that zone suggests that individual bird social preferences are more important than the influence of environmental variables on encouraging sociality between pelicans. Some zones were more likely to see more occurrences of solitary pelicans ( the type of social interaction that occurs in specific zones would add more information to differences in group sizes within each zone. The close association between known dyads results from the bird's natural tendencies to be social (Brown & Urban, 1969) and this social attraction and tolerance are shown by Danel et al. (2020) to facilitate social learning that influences foraging efficacy. Investment in specific social bonds may therefore convey an advantage to pelicans by enabling them to access resources more easily. Some of the highest association indices include male-female pairings, even outside of the breeding season. In the wild, the GWP has been described as monogamous but extra-pair copulations have been observed (Brown & Urban, 1969). The extent to which captive GWPs are monogamous is unknown, although these results suggest that each individual bird picks a single partner, at least per breeding season.
Other strong relationships were identified, including between newly fledged male birds and between adult females. These strong associations between birds may act as a social support network (Rault, 2012); as has been noted in other species, including humans, individuals with familiar characteristics can band together to provide a useful and reliable support network for all individuals involved (McPherson et al., 2001). These bonds can also reduce intraspecific aggression within the flock (Gokula, 2011). Nonbreeding birds, for example, may associate with other individuals to develop their social communication. In wild flocks during the breeding season, sub-adult birds were not seen (Brown & Urban, 1969 (Danel et al., 2020) especially during synchronous feeding events (Saino et al., 1995). Given the propensity for social learning in this species, that is, a long lifespan, a colonial nature, slow maturity, and a big brain (Danel et al., 2020), it is likely that individual birds become more competent at breeding by watching the attempts of other birds within their network. Large flocks, particularly of mixed age groups, may have value in supporting successful captive breeding efforts for GWPs; and as ZIMS data (species360, 2021) shows this study flock to have successfully bred several times since these original data were collected, this exchange of behavioral information may be occurring and having a positive impact on nesting success.

| CONCLUSION
Our research on these GWPs provides insight into the sociality and behavior of a successfully breeding flock, allowing keepers of this species to better understand individual and flock behavioral needs and choices. The application of enclosure use measures has enabled the prediction of a potential visitor effect but not of breeding behavior. We identified preferential enclosure usage, and areas of limited occupancy and that could be altered to increase their value for pelicans. Social network measures reveal an underlying structure within this flock, with very strong associations between specific dyads that were not predicted by age, sex, or origin of the birds involved. Our results suggest to pelican keepers that flock behavior change (particularly time spent to alert) could be an indicator of interest in breeding and these methods could be applied to other captive GWP flocks to further understand potential impediments to breeding.

ACKNOWLEDGMENTS
With thanks to the Bird Department at Blackpool Zoo for providing access to these birds and for answering questions on their husbandry and history at the Zoo. Thanks to Jamie Lunn for assisting with the data collection aspects of this project. There is no funding associated with this project.

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

ETHICS STATEMENT
The methods for this project were reviewed by the Bird Department at Blackpool Zoo in summer 2016, before the project commences, and then by the ethics committee for student projects at University Centre Reaseheath in Autumn 2016. Any correspondence relating to ethical review can be provided by JEB.

DATA AVAILABILITY STATEMENT
Data will be made available upon reasonable request to the corresponding author. Raw data from this project are available upon reasonable request from the corresponding author.