Introduction

Some species of animals are able to adapt to urban habitats (Luniak 2004; Baker and Harris 2007; McKinney 2008; Francis and Chadwick 2012). Mammals and birds are two groups of animals that include urban adapters such as red squirrel Sciurus vulgaris (Luniak 2004; Babińska-Werka and Żółw 2008; Beliniak et al. 2022) and Corvidae birds (e.g. Żmihorski et al. 2010; Benmazouz et al. 2021). Urban populations of corvids and squirrels may differ from those inhabiting natural habitats, i.e. they will reach much higher densities, especially in urban parks (Luniak 1981; Jokimäki and Suhonen 1998; Babińska-Werka and Żółw 2008; Jadczyk et al. 2013; Kövér et al. 2015; Beliniak et al. 2022). Urban animals are often bolder (i.e. their FIDs are shorter) than their non-urban counterparts (Uchida et al. 2016; Samia et al. 2017), which may help them to explore new feeding opportunities. People provide supplementary food (offered directly or left in bird feeders) that is used by both squirrels (Rézouki et al. 2014; Reher et al. 2016; Jokimäki et al. 2017; Kostrzewa and Krauze-Gryz 2020; Krauze-Gryz et al. 2021a) and corvids (Marzluff and Neatherlin 2006; Tryjanowski et al. 2015; Plaza and Lambertucci 2017). These animals can also utilise leftovers searched for in bins (Plaza and Lambertucci 2017; Wist et al. 2022; García-Arroyo et al. 2023). The supplementary feeding may be crucial for maintaining high population density (Magris and Gurnell 2002; Verbeylen et al. 2009; Jokimäki et al. 2017; Benmazouz et al. 2021) and deciding on survival in cities, especially during winter (Luniak 2004).

The co-existence of urban squirrels and corvids may lead to interactions that, when food sources are involved, are mostly neutral or competitive (Bosch and Lurz 2012; Kopij 2014), i.e. if they forage at the same sites and on the same food type (Jayne et al. 2015). A specific form of competition, which involves stealing food that has been obtained by another individual of the same or another species, is kleptoparasitism (Brockmann and Barnard 1979). Most kleptoparasitic studies focus on birds (reviewed in Morand-Ferron et al. 2007), however it is common among other taxa (reviewed in Iyengar 2008). Kleptoparasitism allows certain individuals to obtain energetic benefits by robbing food from another animal, who previously invested time and effort to possess it (Thompson 1986; Ens et al. 1990).

Hoarding food items for later consumption is widespread among birds and mammals (Vander Wall 1990), including squirrels (Rice-Oxley 1993) and corvids (de Kort and Clayton 2006). Thanks to this, animals control feeding resources availability in time and space, and this gives them an advantage above other animals that do not cache food (Vander Wall and Jenkins 2003). Red squirrels are scatter hoarders, i.e. they cache food in numerous small caches as opposed to making big larders (Wauters and Casale 1996).

Stealing squirrels’ food caches by corvids was reported in the literature, e.g. Clark’s nutcracker (Nucifraga columbiana) exploited food items stored by American red squirrel (Tamasciurus hudsonicus) (Brockmann and Barnard 1979), carrion crow (Corvus corone) and Eurasian magpie (Pica pica) that of grey squirrel (Sciurus carolinensis) (Leaver et al. 2007) and eastern scrub jay (Aphelocoma californica) caches of fox squirrel (Sciurus niger) (Anderson 2012). In the case of European red squirrels, such interactions were reported for Eurasian jays (Garrulus glandarius), Siberian jays (Perisoreus infaustus) and Eurasian nutcrackers (Nucifraga caryocatactes) (Bosch and Lurz 2012). Nevertheless, it was assumed that the extent of food stealing was not high enough to seriously deplete the food supplies of the host species (Leaver et al. 2007).

Some birds and mammals change their caching behaviour with the aim to minimise the stealing risk (Dally et al. 2006). Nevertheless, the response of grey squirrels did not seem to be consistent. According to one study, grey squirrels altered their foraging and vigilance behaviour in response to corvids’ auditory playbacks (Jayne et al. 2015). According to the other study, they did not appear to be sensitive to the presence of corvids while caching their food (Schmidt and Ostfeld 2008). Interestingly, grey squirrels turned back and spaced caches further apart when conspecifics (but not when corvids) were present (Leaver et al. 2007). According to our knowledge, such a response of red squirrels has not been studied so far. Investigating if corvids are responsible for stealing squirrels’ supplies is necessary to check if they pose an important stealing risk to squirrels (Leaver et al. 2007).

We focused on interactions between red squirrels and corvids (hooded crows Corvus cornix and rooks Corvus frugilegus as representatives) in an urban park with abundant populations of both taxa and in relation to supplementary food use. The park and the possibility to feed its red squirrels are tourist attractions, and squirrels make use of food offered by park visitors (Krauze-Gryz and Gryz 2015; Kostrzewa and Krauze-Gryz 2020; Krauze-Gryz et al. 2021a; Krauze-Gryz et al. 2021b). They were also found to alter their diurnal activity and behaviour (by spending more time on the ground, approaching people and begging for food), assumingly to increase their chances to obtain supplementary food (Beliniak et al. 2021; Krauze-Gryz et al. 2021a; Krauze-Gryz et al. 2021b). During our previous studies, we often observed abundant corvids interact with red squirrels (Krauze-Gryz et al. 2021a; Krauze-Gryz et al. 2021b). We assumed that corvids, with their high cognitive skills (Clayton and Emery 2005), have also learnt to make use of the supplementary food, i.e. by following squirrels being offered nuts from visitors and trying to steal the obtained food. We hypothesised that if squirrels treat corvids (and other squirrels) as food competitors, they will respond to their presence by altering their behaviour while food handling, i.e. making more deceptive caches (Steele et al. 2008), as squirrels may alter their foraging and vigilance behaviour in the presence of corvids (Jayne et al. 2015). We supposed that this response will change seasonally, i.e. the reaction will be most vivid in winter/spring time when supplementary food is most important for squirrels (Shuttleworth 2000; Magris and Gurnell 2002) and corvids at this time mostly search for rood in urban parks (Luniak 2004).

Materials and methods

Study area

The study was conducted in Warsaw, the capital city of Poland, in Łazienki Park (52° 12′ 51.06″ N, 21° 1′ 58.27″ E), placed approximately 2 km from the city centre, bordered by busy roads. The city has more than 1.8 mln inhabitants, and the population density is 3600 ind/km2 (https://stat.gov.pl/). The region is affected by the relatively mild and wet oceanic climate of Western Europe, as well as the harsh and dry continental climate of Eastern Europe and Asia. The duration of the growing season is approximately 210 days; the total precipitation measures 600 mm per year; and the mean ambient temperature ranges from 4 °C in January to 18 °C in July. During typical winter in Poland, the mean number of days with snow cover varies from 29.6 to 65.8 days per winter (Szwed et al. 2017); however, during the last years, a decrease in snow cover depth and an increase in ambient temperature are noted (Tomczyk et al. 2021) (e.g. winter 2019/2020 was extremely warm and snowless). In the study year (2016), the mean year ambient temperature was 9 °C, and it was an extremely warm year (https://dane.imgw.pl/).

Łazienki Park covers 76 ha, was founded in the 17th century, and is among the biggest and oldest parks in Warsaw. The park is very popular among tourists and local inhabitants, and it is visited by over 2 million people per year (Kruczek 2015). According to the Warsaw Tourist Organization, in 2019, this park with a museum complex was the second most common tourist attraction in Warsaw (https://wot.waw.pl/wiedza/). The park has 92 species of trees and shrubs; tree cover is about 70% (Babińska-Werka and Żółw 2008). Tree species are mostly deciduous, e.g. hornbeam (Carpinus betulus), common oak (Quercus robur), beech (Fagus sylvatica), as well as hazel (Corylus avellana), walnut (Juglans regia) and North American walnut (Juglans nigra) (Babińska-Werka and Żółw 2008). There are also some conifer species, e.g. common yew (Taxus baccata), European spruce (Picea abies) and Duglas fir (Pseudotsuga taxifolia). Approximately 25% of trees in the park are older than 130 years, while some reach as many as 250 years (Babińska-Werka and Żółw 2008).

The park has free entrance, but it is closed during the night. Avian predators included the tawny owl Strix aluco (Gryz et al. 2008). As for mammals, red fox Vulpes vulpes (Jackowiak et al. 2021), stone marten Martes foina and free-ranging domestic cats (Felis catus) are present (Beliniak, pers. obs.).

Park is commonly known for an abundant population of red squirrels (Babińska-Werka and Żółw 2008), reaching 1.05 to 1.89 ind./ha (Beliniak et al. 2022). During this study, we also counted the number of red squirrels and corvids recorded along a 2500 m transect route to assess their relative abundance. The transect route went around the whole park, and the counts were done once a month during the whole year. During the count, red squirrels, hooded crows and rooks were recorded if they stayed at a distance of up to 20 m from the transect route. Number of seen squirrels ranged from 9 (in October) to 45 (in November) and corvids from 2 (in October) to 101 (in January) (see Table S1 in Appendix). Red squirrels and hooded crows were present in the study site during all the study; rooks were absent from April to July. Hooded crows are present in Warsaw all year round; they breed and stay during all months. Rooks, in turn, are less numerous and are present mostly in winter when they come from the north and east of Europe (Luniak et al. 2001; Jadczyk et al. 2013).

Data collection

The study was done from January to December 2016. The year was divided into four seasons: spring (1 March–31 May), summer (1 June–31 August), autumn (1 September–30 November) and winter (1 December–28 February). Observations were done before midday, usually between 8 and 11 a.m., and in good weather conditions, when red squirrels were most active (Wauters and Dhondt 1987; Wauters et al. 1992; Babińska-Werka and Żółw 2008; Beliniak et al. 2021). They were conducted during weekdays, avoiding Saturdays and Sundays when very high numbers of visitors (especially in spring and summer) could hinder observations (animals obtained plenty of food so they lost interest in it and spent less time on the ground as they were often chased by children, Beliniak, pers. obs.). We did not estimate the number of visitors during our study. Yet, according to data collected in the past, the number of visitors observed by one person walking a random path across the park and during 1 h was on weekdays 243 but on Saturday and Sunday reached 309–456 (Rykaczewska 2013).

The research consisted of two parts: (a) feeding trials and (b) observations of groups of animals (details are given below). Squirrels and corvids were not marked. To avoid testing twice the same animal on the same day, the next test was conducted in another park area. It is possible that the same individuals could have participated in the study more than once but on different days. However, sessions were taken in various park areas to make sure that as many individual squirrels as possible were involved. All observations were done by one person. Time was counted using a stopwatch and was reported in seconds. The distance was counted in metres, using a rangefinder.

Feeding trials

We planned our tests (i.e. feeding trials) to simulate situations in which red squirrels are offered nuts by park visitors. Each trial lasted 10 min. During that time, whole (unshelled) hazelnuts were offered to squirrels. The first nut was always put on the ground. Sometimes, when a squirrel was really bold and fast, the nut was given on a hand, in a way visitors usually give food (Krauze-Gryz et al. 2021a). The test started when at least one squirrel was present (up to 10 m from an observer, not necessarily with other squirrels or corvid(s) nearby). Time count started with the first nut being taken by a squirrel. Another nut was offered (i.e. placed on the ground or shown on a hand) directly after the squirrel took the first one. Three scenarios were possible: (a) corvid(s) were present from the beginning of a trial, (b) corvid(s) joined during the trial, c) corvid(s) were absent. A corvid was considered present when it was up to 10 m from a squirrel or apparently was interested in the food. Squirrels were offered nuts in random spots in the whole park.

During the trial, the number of squirrels taking food or approaching an observer was recorded every minute (more squirrels could join as the trial proceeded). Also, the number of corvids was recorded every minute. The way a squirrel handled the food was recorded, i.e. (a) cached, buried or hid between branches or in the grass; (b) ate, instantly consumed the food; and (c) unspecified, when it was not possible to note behaviour due to lost sight of the animal. To investigate if red squirrels modified their caching behaviour when other animals were present, the number of deceptive caches before the final cache was recorded. Deceptive caches were defined as covering one or more excavated or unexcavated sites without putting food inside (fake, empty sites) (Hopewell et al. 2008; Steele et al. 2008). After a final cache (where the nut was hidden), distance from the closest (but staying no further than 30 m) animal (squirrel or corvid) was noted and put into one of the three categories: 0–5 m, 6–10 m and 11–30 m. If there were several squirrels and corvids near the final cache, distance to the closest animal was only noted. When no animals were present up to 30 m from the final cache, the ‘no audience’ category was attributed. Attempts to steal nuts obtained by the squirrel from an observer by corvids or other squirrels (and the success rate) were recorded. Also, the way the stolen nut was handled: cached, eaten or unspecified (see above) was recorded. The trials were done six times per month (regularly every 3 to 5 days) during the whole year (72 days in total). During each visit to the park, two to nine feeding trials were done (17 to 28 monthly). Altogether, 280 trials were done (over 46 h of observations).

Observations of groups of animals

We started the observations when at least one squirrel and one corvid distant up to 10 m from each other were noted in any random spot in the park. In this case, an observer did not offer nuts to squirrels. We observed groups of animals during the same days as the feeding trials, but in other parts of the park to test different individuals whose behaviour was not affected by a feeding trial. Nevertheless, they could have obtained food from park visitors. Each observation lasted 5 min. The behaviour of one individual at the same time was noted. Observation was done if a squirrel (or squirrels) were on the ground, corvids could be on the ground or in the trees/on shrubs. If more individuals (either squirrels or corvids) joined the group, the observation continued. If a group fell apart before 5 min passed, the observation was stopped and was not taken into analysis. Different types of behaviour and for how long they were observed (in seconds) were recorded: (a) foraging and caching—searching for food, eating, caching; (b) inter- and intraspecific interactions—any kind of visible interactions between animals and/or humans; (c) other—drinking, travelling, grooming, no movement (without signs of stress).

Interactions between animals (both inter- and intraspecific) were noted in detail as follows: (a) attempting and stealing food (both successful and unsuccessful attempts to steal previously cached food); (b) chasing and following (when an animal tried to chase or frighten another animal to obtain nut); (c) chasing away (when an animal tried to drive away another animal to prevent stealing after or before caching food); (d) flight (from other animal). Also, interactions with people were recorded: (a) taking food from people (approaching, begging for food and getting food); (b) following people (approaching and following people but not obtaining any food); (c) flight (from people). The interactions were counted per each 5-min observation.

The observations were done six times per month throughout the whole year. Altogether, 915 min of observations were collected (over 15 h). Red squirrels were observed 89, hooded crows 67 and rooks 27 times.

Statistical analysis

A chi-square test was used to test differences in the proportions of cached (as opposed to all obtained) nuts, feeding trials with corvids trying to steal nuts obtained by squirrels, nuts that were attempted to be stolen by corvids, and time dedicated to different behaviours observed during 5 min observations. A Kruskal–Wallis test was used to compare the numbers of animals (i.e. squirrels and corvids) per each minute of a feeding trial, seasonal differences in the number of deceptive caches, differences in the number of deceptive caches in the presence of other animals (and depending on how close they stayed) and a frequency of interactions recorded per 5-min observations. The normality of data distribution was tested with the Shapiro–Wilk test (P < 0.05). Dunn’s test was done for post hoc analysis (Bonferroni-corrected P values). The analyses were done in Past 4.05 (Hammer et al. 2001) software.

Results

Feeding trials

Squirrels mostly cached obtained nuts (Table 1), and this happened most frequently in autumn (chi-square test, χ2 = 92.79, DF = 3, P < 0.0001). The number of feeding trials, when corvids were present (either from the start or joined the trial) and tried to steal nuts cached by red squirrels, changed seasonally and was highest in autumn and winter (chi-square test, χ2 = 55.05 DF = 3, P < 0.0001, Table 2). Nevertheless, the share of nuts that were attempted to be stolen from squirrels was similar in all seasons (chi-square test, χ2 = 2.15 DF=3, P > 0.05). Corvids were constantly present during 21% of all feeding trials (N = 58), and in those trials, they tried to pilfer almost every third nut obtained from the observer and cached by squirrels (29%, N = 825, Table S2).

Table 1 Number of feeding trials, given nuts and way of handling by squirrels during seasons
Full size table
Table 2 Number of all feeding trials (corvid(s) were present or absent), number of feeding trials with corvid(s) presence (from the start or joining the trial), number of cached nuts by squirrels (during trials with corvids present) and the percent of nuts stolen by corvids (in all trials with corvids present) seasonally
Full size table

Squirrels stole about 1% of cached nuts from other squirrels (N = 26), and all attempts were successful. They also mostly cached stolen nuts (88%, N = 24). Corvids more often ate (58%) than cached (14%) stolen nuts (N = 100). In the remaining cases, the way of handling stolen nuts was not assessed as corvids flew away.

Regardless of the presence or absence of corvids, the number of squirrels observed during the feeding trial increased (Fig. 1). The number of corvids was stable during the trial (Fig. 1a) and increased only if birds were not present from the very beginning (Fig. 1b).

Fig. 1
figure 1

Mean (± SE) number of animals (red squirrels and corvids) present from the beginning or joining the 10 min feeding trials (when an observer offered nuts) conducted in the urban park in Warsaw (Łazienki Park). The results of the Kruskal–Wallis test for the three scenarios were as follows: a corvid(s) were present from the beginning of a trial, squirrels: H = 20.17, DF = 9, P < 0.005; corvids: P > 0.05, b corvid(s) joined during the trial, squirrels: H = 63.7, DF = 9, P < 0.0001; corvids: H = 1919.6, DF = 9, P < 0.001, c corvids were absent during the whole trial, squirrels: H = 75.1, DF = 9, P < 0.0001. In order to start a trial at least one squirrel had to be present

Full size image

The number of deceptive caches done by squirrels was assessed for 765 nuts obtained from an observer. On average, squirrels performed 2.86 deceptive caches (SE = 0.12, min 0, max 32) for each nut. Almost every fourth nut was cached without any deceptive caches (24%, N = 183). The mean number of deceptive caches did not differ between seasons (Kruskal–Wallis test: H = 5.062, DF = 3, P > 0.05). Also, the presence or absence of other animals did not affect the number of deceptive caches done by a squirrel (Kruskal–Wallis test: H = 3.478, DF = 2, P > 0.05).

Corvids stayed closer (X ± SE= 6.84 ± 0.47) to squirrels caching nuts than other squirrels (X ± SE = 9.7 ± 0.39, Kruskal–Wallis test, H = 33.43, DF = 1, P < 0.001). Nevertheless, distance of a corvid from a squirrel (i.e. 1–5 m, 6–10 m and 11–30 m) did not affect the number of deceptive caches. On the contrary, when another squirrel was the closest audience, the number of caches depended on how far it stayed. The number of deceptive caches was lower when the other squirrel was 1–5 m as compared to 11–30 m apart (Fig. 2).

Fig. 2
figure 2

Mean (± SE) number of deceptive caches done by a squirrel during caching a nut given by the observer in the urban park in the presence of an audience: another squirrel, Kruskal–Wallis test, H = 9.876, DF = 2, P < 0.01, the difference between 1–5 m and 11–30 m according to Dunn’ post hoc test; corvid, H = 2.489, DF = 2, P > 0.05) in three distance categories

Full size image

Observations of groups of animals

Both squirrels and corvids spent most time foraging or caching, but the proportion of time dedicated to different activities differed (chi-square test, χ2 = 7186.1, DF = 2, P < 0.001) (Table 3). When squirrels interacted with other animals (both squirrels and corvids), they most often flew away. The other most frequent behaviour was chasing other animals away. On the contrary, in the case of corvids, flight was observed the least frequently. Corvids usually followed and chased other individuals or tried to steal their food (Table 4). Squirrels often interacted with people, i.e. most frequently they took food (Table 4). Crows followed or flew away from people, but such observations were very infrequent (i.e. on average 0.02 interactions per 5 min observations, min 0, max 2, while no interactions between rooks and people were observed. Squirrels interacted mostly with corvids, and less often with people or other squirrels. Corvids focused mostly on squirrels—they interacted with squirrels 2–3 times per each 5-min observation (Table 5).

Table 3 Behaviour of red squirrels and corvids registered during direct 5 min observations of groups (at least two individuals) of animals in the urban park
Full size table
Table 4 Frequency of inter- and intraspecific interactions between red squirrels and corvids in the urban park (n of recorded interactions per one 5 min observation (± SE)
Full size table
Table 5 Frequency of inter- and intraspecific interactions between red squirrels and corvids in the urban park (n of recorded interactions per one 5 min observation (± SE)
Full size table

Discussion

In this study, we focused on inter- and intraspecific interactions between red squirrels and corvids in the urban park and in relation to supplementary food offered by park visitors. Squirrels often interacted with people and took offered food, while corvids rather focused on squirrels, trying to steal obtained food. Despite this, the squirrel did not try to prevent corvids from stealing the food by performing more deceptive caches. They only reacted this way to their conspecifics.

During feeding trials, squirrels mostly cached obtained food, which was similar to other studies (Tamura et al. 1999; Kostrzewa and Krauze-Gryz 2020) and which suggests that they treated this as supplementary supplies (Shuttleworth 2000). Corvids, in turn, usually ate stolen food, which may point to the importance of this limited food source (review in Morand-Ferron et al. 2007, but see: Jokimäki et al. 2022). In general, corvids joined squirrels that were offered food. Yet, when other corvids were present at the start of the trial, new birds seemed to refrain from joining in. This may point to the intraspecific competition in an abundant urban population. They also seemed to be vividly interested in this food source, attempting to steal or stealing every third nut. On the contrary, red squirrels generally joined other squirrels during the trials, but they did not seem to be much interested in food obtained by other squirrels; they stole nuts only occasionally. Squirrels are likely to obtain nuts directly from numerous park visitors (Kostrzewa and Krauze-Gryz 2020; Krauze-Gryz et al. 2021b), while for corvids (being often perceived as overabundant pests), this is rather unlikely. Instead of stealing nuts, squirrels seemed to observe their conspecifics to take an opportunity if supplementary food was being offered. It suggests that this (especially to some individuals, Krauze-Gryz et al. 2021b) is a valuable source of food, and squirrels may rely on it. It is assumed that animals that cache food can lose up to 30% of their storage per day to pilferers (see review Vander Wall and Jenkins 2003). In the light of this, attempts of stealing caches of red squirrels by corvids can be assumed high. This may be due to the fact that corvids are very numerous in the park, but it also suggests that corvids learned to utilise food caches of red squirrels and points to their high cognitive abilities (Morand-Ferron et al. 2007). It was beyond the focus of this experiment, but corvids may have also remembered areas where nuts were cached with the aim to come back later using specific observational spatial memory (Watanabe and Clayton 2007; Scheid et al. 2008; Grodzinski et al. 2012). This may explain why in some feeding trials, corvids (despite being present) did not try to steal nuts.

We assumed that stealing food by corvids would be more frequent during winter and spring when corvids mostly search for food in urban parks (Luniak et al. 2001). However, in this study, the proportion of nuts stolen by corvids remained the same throughout the year. That suggests that anthropogenic food can play an important role in corvids’ diet regardless of season.

Squirrels can change their caching behaviour when other animals are present in aim to lower the risk of stealing food (Leaver et al. 2007; Hopewell et al. 2008; Steele et al. 2008). Interestingly, in our study, we did not notice changes in the caching behaviour (i.e. the number of deceptive caches) with the presence of either conspecifics or corvids. During all seasons, the mean number of deceptive caches was similar; despite in autumn and winter, the proportion of feeding trials with corvids present was much higher than in other seasons. In another study, grey squirrels were shown to modify their caching behaviour (by turning back) when other squirrels were present. At the same time, they did not respond to the presence of corvids as this strategy would not be effective with corvids’ ability to fly and observe from the height (Leaver et al. 2007). We observed that corvids kept closer to squirrels caching nuts than other squirrels. Yet, the distance of a corvid from a squirrel did not affect the number of deceptive caches. According to Stapanian and Smith (1984), scatter hoarding minimalizes food stealing by birds, as birds have a weak sense of smell. On the contrary, when another squirrel was the closest audience, its distance affected the number of deceptive caches. In a similar study on grey squirrels, they made more deceptive caches, when conspecifics were up to 20 m (Steele et al. 2008). Yet, in our case, the number of deceptive caches differed mostly when other squirrels were very close and at a farther distance. This may suggest that with another squirrel being very close (like up to 5 m in our case), the chances of protecting the cache were slight, so the host did not invest much time in handling the nut. Also, this might have been unprofitable; as in our test, new nuts were offered just after the previous one was taken. Moreover, all reported trials of reburying food cached by conspecifics were successful which means that red squirrels were not deceived by deceptive caches. This stays in contrast to the findings of another study, where grey squirrels looked for food in deceptive caches and were deceived by conspecifics (Steele et al. 2008).

All of the observed animals spent most time foraging. Indeed, in the case of squirrels, this is the dominant activity on which they spent from about half (Shuttleworth 2000) to even 80% of their daily activity time (Wauters et al. 1992). Interspecific and intraspecific interactions accounted for only 3% of the time in the case of squirrels but as much as 16% in the case of corvids. Corvids most often interacted with squirrels (i.e. they chased squirrels or followed them while they foraged) and very seldom with people. During our study, we never saw people giving nuts to corvids, as it was seen elsewhere in Poland (Jadczyk et al. 2013). On the contrary, park visitors mostly tried to whisk corvids away while feeding squirrels. This shows that corvids adjusted to this situation of limited human tolerance and focused on thieving food that squirrels obtained from people, rather than trying to get it directly.

Squirrels rarely interacted with other squirrels, more often with corvids or people. With corvids mostly agonistic interactions were recorded, i.e. flight or chasing away corvids during hoarding food. Squirrels, as scatter hoarders, are thought not to invest much energy in cache protection (Brockmann and Barnard 1979; Steele et al. 2008), but in some cases, it is more beneficial to protect the cache than collect more food items in time (Hopewell et al. 2008). On the contrary, when squirrels interacted with people, they mostly attempted to obtain food. Indeed, supplementary food can play important role in squirrels’ diet (Shuttleworth 2000; Magris and Gurnell 2002; Babińska-Werka and Żółw 2008; Bosch and Lurz 2012; Krauze-Gryz and Gryz 2015; Reher et al. 2016; Jokimäki et al. 2017; Krauze-Gryz et al. 2021a) and can have an impact on their demographics (Magris and Gurnell 2002).

The results presented here need to be treated with some caution—the study was conducted in one park and with non-marked animals some of them could have participated in the study more than once. However, the study site was big (almost 80 ha), with a very abundant population of squirrels and corvids, and the study was conducted in a way which minimised the possibility of testing the same animal twice in a short time. A similar study conducted in the future should involve more localities (with varying abundance of the two species and with varying natural food base and importance of supplementary food source) and encompass a longer period (as winter conditions and food base availability are not stable) to better understand how animals cope with the availability of new human-provided food sources in urban habitats. Another weakness of our study is that it refers only to a single year. Food availability and abundance of red squirrels and corvids may be different during the years; thus, inter- and intraspecific interactions may also be variable.

Overall, we showed that corvids can be food competitors and kleptoparasites for red squirrels. Red squirrels, with whom people often have affinity relationships, benefited from direct supplementary feeding and often interacted with people. Corvids, in turn, learnt how to steal human-delivered nuts from squirrels instead of trying to get it directly. This may be an evidence of how animals adapt to urban habitats and how the presence of anthropogenic food influences their feeding strategies.