Why do zoos attract crows? A comparative study from Europe and Asia

Abstract Crows have successfully colonized many cities, and urban zoos have been important in this process. To evaluate why zoos attract crows, we quantified crow numbers and behavior in three zoos in Europe (Debrecen, Edinburgh, Vienna) and one in Asia (Sapporo). Data were collected in 445 surveys over 297 days in summer 2014 and winter 2014–2015. We found that crow numbers were highest in Vienna, intermediate in Debrecen and Edinburgh and lowest in Sapporo, increased significantly from summer to winter (Debrecen, Edinburgh, Vienna), and from mornings to afternoons (Debrecen, Sapporo, Vienna), and were higher in sunny weather than in cloudy weather with precipitation and when visitor numbers were low (Debrecen, Vienna). The crows' use of natural food was highest in Vienna, intermediate in Edinburgh and Sapporo, and low in Debrecen. The use of anthropogenic food was high in Debrecen and Sapporo, where the availability of open grassy areas typically used by crows for natural foraging was low. In Sapporo, food availability was more limited than in other zoos, resulting in strong territoriality and few crows in summer, which decreased further in winter. Our study indicates that crows are primarily attracted to zoos by food availability and secondarily by breeding opportunities and that the relative importance of natural versus anthropogenic food sources may vary with zoo habitat structure. Our study draws attention to a previously overlooked role of zoos in urban biodiversity conservation. It may also provide useful information for the management of crow populations, if necessary, and for the planning of urban areas.

The benefits that urban environments provide to bird species include the year-round availability of food, milder local climate, decreased or no predator pressure, and diverse nesting opportunities (Eötvös, Magura, & Lövei, 2018;Marzluff et al., 2001;Vuorisalo et al., 2003). An understanding of the relative importance of these factors, however, is difficult due to the variation in urban environments in terms of human population size, the built-in and natural components of urban landscapes, and type and availability of food sources.
A comparison of corvid use of urban landscapes that are similar in human population size, urban landscape composition, and food availability can thus provide insights into the relative importance of the factors suspected to explain corvid colonization of urban environments (Preininger et al., 2019).
Zoos offer a great possibility for such comparisons. Zoos are found in many cities of the world and are typically viewed as cultural landscapes (Axelsson & May, 2008;Hallman & Benbow, 2006) that often show similarities in area (up to few tens of hectares), landscape composition (parks with trees and open spaces), and food availability to free-living birds (food given to zoo animals, leftover/garbage from humans). Zoos that allow access to free-living birds and that provide permanent sources of food may be the first to be colonized by corvids in a city and are usually characterized by high nesting density (Kövér et al., 2015). Zoos thus provide an excellent opportunity to study the factors that influence the establishment and colonization of corvids in urban environments (Uhl et al., 2018). However, to our knowledge, there is no study specifically investigating the role that zoos play in the colonization of cities by free-living birds. We thus know little on why zoos attract free-living birds such as crows, that is, how crows use the zoos and what are the factors that may make zoos attractive to crows.
The aim of this study was to evaluate why crows are attracted to zoos, that is, the role that zoos play in the establishment and colonization of free-living corvids in urban environments. We focused on food availability and breeding opportunities as factors that may explain the attractiveness of zoos to crows. We hypothesized that (a) crow abundance may vary between zoos with different geographic settings, habitat structures, and/or food management practices, (b) crow abundance may change between seasons, as we expected more crows in the winter (when food is generally limited and crows are gregarious) than in the summer (when food is usually not limited and crows are territorial), and (c) crow abundance may be higher in the afternoons, as zoo animals are typically fed during the day, for example, at scheduled feeding times for visitors, and by the afternoon, the leftovers will become available to crows. To study these factors, we quantified the number of corvids, including, as far as possible breeding (i.e., resident) and nonbreeding corvids (i.e., nonresident), and recorded their behavior over one summer and one winter in four urban zoos that differed in geographic settings, habitat structures, and/or food management practices. We studied the effects of food availability, season, time of day, weather, temperature, and snow cover on corvid numbers. We also quantified crow behavior to evaluate the importance of natural versus anthropogenic food sources and breeding opportunities in order to detect differences between zoos that can be used to infer the importance of the above factors in corvid establishment and colonization of urban areas.

| Study locations and species
We studied three crow species (Figure 1)

| Data collection
We collected data by recording the number and behavior of crows in the zoos in the summer of 2014 and winter of 2014/2015 (Table 2).
We surveyed crows while walking along a predefined route that covered the entire zoo area. One or more surveys were conducted on every study day (mean 1.6 ± S.D. 0.37, range 1.2-2.7 surveys per day in the summer and 1.5 ± 0.16, range 1.2-1.9 surveys per day in the winter). The total number of study days was 85, 72, 82,and 58 in Debrecen, Edinburgh, Sapporo, and Vienna, respectively, and the number of surveys was 100, 88, 114, and 143, respectively (Table 2).
On average, surveys were conducted once every 2.4 ± 0.17 days (range 2.1-2.9) in the summer and once every 2.5 ± 0.33 days (range 1.9-3.3) in the winter (all zoos combined).
In each survey, we recorded the location of sighting for each crow and the behavior of the bird. We defined three categories of crow behavior: (a) "foraging" was searching for or consuming naturally occurring food on the ground, in the grass, or in the trees, (b) "feeding" was searching for or consuming food from anthropogenic sources either as food provided to the zoo animals, given directly to crows by humans or indirectly as leftover near restaurants, kiosks, or trash bins, and (c) "breeding" included all behaviors suggesting that an animal was breeding (collecting nest material, incubating eggs, guarding a nest, guarding or feeding young etc., only in summer). Surveys were con- The number of visitors was then classified into three categories as few (between 1 and 5 visitors), intermediate (between 6 and 10 visitors), and many (>10 visitors). Finally, we recorded season (summer, winter), time of day (morning, afternoon, with threshold of 13:00 in summer and 12:00 in winter), weather conditions (sunny, cloudy, cloudy with precipitation such as rain or snow), air temperature (measured by handheld thermometers), and presence of snow cover (only in winter). Air temperature measurements were classified into three categories (cold: <10°C, moderate: between 10°C and 20°C, warm: >20°C). No surveys were conducted in heavy rain or snow or when air temperature was below −10°C or above 30°C.

| Data analyses
Our first response variable was the number of observed crows per survey. To correct for the different area of zoos, we calculated this as the number of crows per 10 ha. We evaluated the effect of predictor variables on the response variable by constructing a generalized linear mixed model (GLMM) with a log-linear link function and a Poisson error distribution. For the GLMM, we pooled summer and winter data, and "day" was entered into the model as random factor to account for repeated measures, that is, more than one survey on the same day and to minimize the effect of temporal autocorrelation in the observations. Fixed main effects included zoo (Debrecen, Edinburgh, Sapporo, Vienna), season, time of day, weather, temperature, snow cover, and number of visitors. We did not differentiate between observations on weekdays and weekends/holidays because the number of visitors was considered a more direct measure of human presence in the zoos.
Because our primary interest was to find between-zoo differences (i.e., interactions) and similarities (i.e., lack of interactions) in the effect of predictor variables, we entered all first-order interactions with "zoo" to allow for different relationships with fixed effects in the four zoos (full model). Nonsignificant (p < .05) interactions and main effects were removed sequentially to obtain a final reduced model. When any interaction with "zoo" was significant, we applied pairwise comparisons of zoos using t tests, with a significance level corrected for multiple comparisons (Bonferroni α = .005). To confirm our final model, we also carried out model selection in an information theory-based approach based on AICc using the "dredge" function of the R package "MuMin." This analysis resulted in one best model (for the second best model, ΔAICc = 2.218 or >2), which was identical to our final model; thus, we concluded that an alternative approach would lead to qualitatively identical results.
Our second response variable was crow behavior. To compare crow behavior within and between zoos, we calculated percentages of the three behavioral variables separately for each zoo as the percentage of occasions when birds were observed performing any of the three behaviors in each survey. As we only recorded foraging, feeding, and breeding and crows could engage in other behaviors, these percentages did not necessarily add up to 100% in each survey. We compared behavioral data (percentages) between and within zoos separately for seasons. As the behavioral data were not normally distributed, we used Kruskal-Wallis and Mann-Whitney U tests for between-zoo comparisons. In pairwise Mann-Whitney U tests, we applied Bonferroni adjustment for multiple comparisons by setting α to .008 (between-zoo comparisons) or to .016 (within-zoo comparisons). Significance values given are two-tailed, and all analyses were performed with IBM SPSS Statistics 20.

| Factors influencing the number of crows
We observed significantly more crows in Vienna than in the other three zoos (Figure 3 Table 3).

Significant interactions between zoos and four factors
showed that the effects of season, time of day, weather, and number of visitors on the number of crows differed between the zoos ( Table 3). The interaction between zoo and season was because there were more crows in the winter than in the summer in Debrecen (F 1, 415 = 68.181, p < .001), Edinburgh (F 1, 415 = 17.068, p < .001), and Vienna (F 1, 415 = 17.47, p < .001), whereas we found the opposite in Sapporo (F 1, 415 = 6.589, p = .011) ( Figure 3A, Table 3).
The interaction between zoo and time of day indicated that there were more crows in the afternoon than in the morning in Debrecen (F 1, 415 = 5.439, p = .02), Sapporo (F 1, 415 = 9.743, p = .002), and Vienna (F 1, 415 = 128.526, p < .001), whereas a similar difference in Edinburgh was not significant (F 1, 415 = 0.938, p = .333) ( Figure 3B, Table 3). The interaction between zoo and weather indicated that there were more crows in sunny weather than in cloudy weather with precipitation in Debrecen (F 2, 415 = 9.531, p < .001) and Vienna (F 2, 415 = 18.829, p < .001), whereas there were no such differences in Edinburgh (F 2, 415 = 1.008, p = .389) and Sapporo (F 2, 415 = 2.49, p = .084) ( Figure 3C, Table 3). Finally, the interaction between zoo and the number of visitors was because there were more crows when there were a few or some visitors than when there were many visitors in Debrecen (few vs. many: t = 3.881, df = 415, p < .001; some vs. many: t = 3.995, df = 415, p < .001), whereas there were more crows when there were some visitors compared to few or many visitors in Vienna (few vs. some: t = 4.709, df = 415, p < .001; some vs. many: t = 2.657, df = 415, p = .008). The number of crows observed was not related to the number of visitors in Edinburgh (F 2, 415 = 0.1, p = .905) and Sapporo (F 2, 415 = 2.722, p = .067).

| Crow behavior differences between and within zoos
Between-zoo comparisons revealed significant variation in each of the three crow behaviors recorded between zoos (Figure 4). In summer, the frequency of foraging (use of natural food sources) was highest in Vienna, followed by Edinburgh, and was low in Debrecen and Sapporo ( Figure 4A, Table 4). The frequency of feeding (use of anthropogenic food sources) was high (but also highly variable) in Debrecen, relatively high in Edinburgh, followed by Vienna, and was lowest in Sapporo ( Figure 4C, Table 4). The frequency of breeding behaviors was highest in Sapporo, followed by Edinburgh, then by Vienna and Debrecen, with no significant difference between the latter two ( Figure 4E, Table 4). In the winter, the frequency of foraging behavior was highest in Vienna and significantly lower (15%-20%) in the other three zoos, indicating a decrease from summer levels in Edinburgh and increases in Debrecen and Sapporo ( Figure 4B, Table 4). The frequency of feeding behavior increased from summer to winter in Debrecen and Edinburgh and was significantly higher in the winter than in Sapporo and Vienna ( Figure 4D, Table 4).
Within-zoo comparisons showed that feeding was the most frequently observed behavior of crows in Debrecen in the summer, and its frequency was higher than that of foraging ( Figure

| D ISCUSS I ON
Our study provides two main results. First, we found that the varia- Afternoons are probably important feeding times in the winter for crows because they need to collect energy to survive the cold nights at this time (Baltensperger et al., 2013). In Edinburgh, zoo birds and primates are fed twice a day (morning and afternoon), resulting in food potentially being available to crows all day, which may be related to the absence of a time-of-day effect here.

Weather was important for crow numbers in Debrecen and
Vienna, with more crows observed in sunny weather than in cloudy weather with precipitation. It is likely that crows move around less when there is precipitation, for example, we observed that crows tended to stay longer in roosting sites in adverse weather and extreme cold in Sapporo, probably to save energy. In rainy weather, most crows stop feeding, become less active, and seek shelter from rain (Hume, 1986). Moreover, air uplift, which is necessary for crows for medium to long-distance flight, does not occur on rainy days (Elkins, 1983). Although it is likely that crows taking refuge in trees were less detectable to observers than active crows in sunny weather, the differences in crow numbers were far greater than just a few individuals overlooked (Figure 4).

The number of visitors influenced crow numbers in Debrecen
and Vienna similarly, with more crows when the number of visitors was few/intermediate, that is, 10 or less than when the number of visitors was many (over 10) in Debrecen, and with more crows with intermediate (6-10 visitors) than with few (5 or less) or many visitors in Vienna. This was expected because although crows can easily get used to the presence of humans in urban environments, they usually keep some distance from humans (Clucas & Marzluff, 2012;Matsyura, Jankowski, & Zimaroyeva, 2015) and thus are expected to avoid areas with too many visitors. , and crows were hardly if ever seen foraging in wooded areas or forests. In contrast, feeding on anthropogenic food was typically observed near sources of such food, for example, outdoor eateries, kiosks, trash bins, and animal enclosures, and on trees where crows carried food items for further handling and consumption.
One likely reason for the relative importance of natural versus anthropogenic food sources may be related to the structure of habitats within the zoos (Figure 2, Table 1 (Table 1), and there are strict rules for the handling of zoo animal food. For example, large carnivores are given their food in the indoor parts of their pens. As a result, food availability is low and is mostly restricted to naturally occurring food. As a possible consequence of the scarcity of food, in the summer, the Sapporo zoo area is divided up into territories that are fiercely defended by territorial pairs of breeding crows. The crows observed in the zoo in the summer are, almost exclusively, resident breeding birds, which drive nonresidents out of the zoo area, which in turn explains the small and constant number of crows. During the surveys, for example, we observed territorial pairs defending their territories against flocks, sometimes of up to 250 and 300 crows that attempted to trespass or perch in their territories. In the winter, cold weather and snow cover reduce naturally occurring food so even the resident birds have to leave the zoo to find food elsewhere from time to time.
Our results on the importance of food sources do not refute the general observation that zoos provide ample opportunities for the F I G U R E 3 Mean ± SE number of crows observed in the four zoos in summer and winter surveys (A), in morning and afternoon surveys (B), and in surveys in sunny, cloudy weather, and cloudy weather with precipitation (C). Values shown are means adjusted for fixed and random effects by the final GLMM. Lowercase letters above (A) indicate significant differences (p < .01) in the number of crows in the between-zoo comparison, with data pooled across season, time of day and weather. Asterisks indicate results of within-zoo comparisons, *p < .05, **p < .01, ***p < .001, ns: not significant breeding of the crows (Table 1). For example, in Vienna, as many as 45 active nests were found in 2012 within the zoo (C. Schwab, pers. obs.), corresponding to a high nesting density of 2.6 nests per hectare. This value is one magnitude higher than that reported anywhere else previously, for example, one such published maximum was 0.255 nests per hectare (Vuorisalo et al., 2003). High nesting densities may have important consequences on the social structure of crows through the emergence of colonial nesting, which is not expected in crows that traditionally nest solitarily in rural areas (Cramp & Perrins, 1994;McGowan, 2001). In Vienna, the social structure of crows shows an environmentally influenced fission-fusion dynamics centered around the zoo throughout the year (Uhl et al., 2018). In Debrecen, the center of establishment and colonization of crows in the city was the zoo (Kövér et al., 2015). Between 2006 and 2012, the city nesting population increased continuously, and the rate of increase was highest in the zoo area, which has many tall trees available for nesting. Here, nesting density increased from 2 to 8 nests/ km 2 in only seven years (Kövér et al., 2015). These previous results and our current results suggest that high availability of food and potential nesting sites are likely candidates to explain why crows are attracted to zoos.
We did not measure food availability directly, for example, by quantifying ground-dwelling invertebrates or the amount of anthropogenic food accessible to crows. Rather, we used an indirect measure of food availability based on the behavior of the birds observed. In some cases, however, a higher frequency of foraging may not necessarily reflect higher availability of quality foraging areas with abundant natural food. For example, it is possible that crows are forced to forage for natural food rather than feed on anthropogenic food if the latter is not available, for example, by strict management of food for zoo animals, bird-proof enclosures and garbage collection/storage, etc. Thus, foraging and feeding can be difficult to separate and more detailed information on natural and anthropogenic food sources, including their spatial and temporal variability, are necessary to disentangle the effects of natural versus anthropogenic food sources and the importance of the quantity versus quality of food to crows.
Our findings may have implications for the planning of urban areas and green infrastructure, as well as potential management implications for zoos regarding free-living corvids, which may be especially relevant if crows become high in number. In many urban areas of the world, free-living corvids come into conflict with humans (e.g., Soh, Sodhi, Seoh, & Brook, 2002), as large concentrations or high nesting densities of corvids may impact the soil and vegetation, decrease urban bird diversity, and present sanitary risks from the crows' use of waste dumps and leftovers for feeding (Matsyura, Zimaroyeva, & Jankowski, 2016;Zeller & Schuffenecker, 2004

F I G U R E 4
Percentage of crows observed engaged in foraging (natural food, A, B), feeding (anthropogenic food, C, D), and breeding (E) behavior in the summer (left-hand column) and winter (right-hand column) in the four zoos. Lowercase letters above the graphs indicate significant differences between zoos (pairwise Mann-Whitney tests, Bonferroni-adjusted two-tailed probabilities, p < .008) We conclude that crows are primarily attracted to zoos by food availability and secondarily by the availability of nesting sites. Both natural food and anthropogenic food contribute to increased food availability, and the relative importance of each appears to vary with habitat structure within and around the zoos. Our study thus draws attention to a previously overlooked role of zoos in the conservation of biodiversity (Conde, Flesness, Colchero, Jones, & Sceheurlein, 2011). It also provides useful information for the management of crow populations if they become high in number and for the planning of urban areas and green infrastructure.

ACK N OWLED G M ENTS
We thank the study zoos for permits to conduct fieldwork and two anonymous reviewers for constructive comments on an earlier draft. AK, FU, RM and CS were supported by the Vienna Science and Technology Fund (WWTF) through project CS11-008 to CS. SL was supported by two grants from the National Research, Development and Innovation Office of Hungary (OTKA K106133, GINOP 2.3.3-15.2016.00019).

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data used in the analyses will available from Dryad upon acceptance. https ://doi.org/10.5061/dryad.d2547 d7zm TA B L E 4 Between-zoo pairwise comparisons of foraging (use of natural food sources), feeding (use of anthropogenic food sources), and breeding behaviors in summer (data in Figure 2) and foraging and feeding in winter Note: Mann-Whitney tests, Bonferroni-adjusted two-tailed probabilities, significant (p < .008) differences are highlighted in bold.