Factors associated with Herring Gulls Larus argentatus stealing food from humans in coastal towns

ABSTRACT Food stealing from humans by Herring Gulls Larus argentatus increased with human food availability but was not associated with any food preferences. Gulls obtained significantly more food when stealing than when scrounging, and the response of most people was neutral, suggesting that food stealing is a low-risk, high-reward strategy for urban Herring Gulls.

Increasing interactions between humans and wildlife have created a myriad of conflicts. As a result, human-wildlife conflict has become of major research interest in the last two decades (Nyhus 2016). While this large body of research has informed strategies to resolve human-wildlife conflict in many cases, there are taxa that remain under-represented (Nyhus 2016). One such species is the Herring Gull Larus argentatus. While their foraging behaviour of targeting fishery discards (Pons & Pierre 1995, Monaghan 1980, Shelpr et al. 2021) and landfills (Hudson & Furness 1989, Stenhouse & Montevecchi 1999, Wilhelm et al. 2016 has been studied extensively, their foraging behaviour in towns is not nearly as well understood and represents a significant gap in the research literature. This is problematic as towns are one anthropogenic landscape where arguably a substantial part of human-gull conflict lies (Rock 2005). Recent studies have addressed kleptoparasitic (i.e. food-stealing) interactions between gulls and humans, showing that gulls' approach to human food is deterred by human gaze (Goumas et al. 2019) and that urban gulls make more kleptoparasitic attempts than their coastal counterparts (Spencer et al. 2017). However, the conditions that may influence kleptoparasitic behaviour of urban gulls, such as human food availability, location and season, have not yet been extensively investigated in towns.
Kleptoparasitism can be an effective foraging strategy for opportunistic species, such as the Herring Gull, when certain ecological conditions are met (Brockmann & Barnard 1979), including large concentrations of hosts and food, reliable and predictable food availability, and the presence of high quality, easily detectable foods (Carr & Ryes-Galindo 2017), all of which can be found in many towns. Observational studies have been conducted on gull kleptoparasitic interactions with American Black Ducks Anas rubripes (Jones III et al. 2011), Atlantic Puffins Fratercula arctica (Busniuk et al. 2020) and other species (Källander 2006), providing speciesspecific information about gull-host interactions. Similarly, this study aimed to identify which factors are associated with gull kleptoparasitic behaviour in urban areas where humans are the hosts, including time of day, location, season, food availability, human group composition and food type. The identification of these factors may help to inform conflict mitigation strategies, and thus reduce harm to gulls as well as the negative economic consequences of their kleptoparasitic behaviour.
First, we predicted that time of day would be associated with the frequency of kleptoparasitic attempts by Herring Gulls. As anthropogenic food becomes available daily at highly predictable times (e.g. lunch time around noon), gulls would benefit from timing their food acquisition attempts accordingly, as observed in gulls in Bristol (Spelt et al. 2020). We also tested for a correlation between overall food availability at a given location and overall number of food-stealing attempts, and considered how these might vary with season and location. Human food availability is likely to increase in summer in popular coastal towns, and some townssuch as St Ives, Cornwallare renowned for their 'problem gulls' (Rock et al. 2016). We further tested whether there was an association between gull food-taking success and human group composition (Broom et al. 2004), as groups of only adults are likely to behave differently from groups consisting of both adults and children. Other factors we quantified were the type of food targeted and the quantity obtained (Capretta et al. 1973); we speculated that gulls may try to take certain foods more than others, as they may be easier to handle and/or provide more energy. Lastly, to quantify the trade-off between these rewards and the potential risks associated with kleptoparasitism, we measured the rate of food obtainment success by the gulls and the frequency of people's negative, neutral and positive responses.
We collected data in six towns in the county of Cornwall, in the south-west of the UK, in areas where gulls consistently make food acquisition attempts due to high footfall of humans and high food availability: the harbour areas of Porthleven and Padstow, the seafronts of St Ives and East Looe, the market square ('Lemon quay') of Truro, and the pier of Falmouth (for coordinates see online Table S1).
We made observations from 11:00 to 14:45 h, on 12 days, sampling each of the six towns twice, between 7 April and 22 July 2021. The first time we visited each town, before July, was considered to be before the tourist season, while the second visit to each town occurred in July, during the tourist season (Cornwall Government 2020). We visited locations, in random order, based on the weather forecast and, where relevant, the tide forecast, to avoid low food availability as rainy weather and high tides tend to result in low footfall (personal observation). High tides will also affect how gulls behave as their natural prey is less available during this time (Kubetzki & Garthe 2003).
We conducted a scan sample of food availability every minute throughout the observation period. The observer stood at a fixed point and the radius within which data were collected was determined by the observer's visual periphery (see online Table S1 for exact area). Town locations were chosen based on previous work on gulls in Cornwall (Goumas et al. 2019, Goumas et al. 2020. We used binoculars to identify food items and recorded food type, number of food items accessible to gulls, number of people within a 5 m radius, and time of day.
During the observation period, we used continuous sampling to record all the food acquisition attempts made by gulls; these attempts were categorized as either 'scrounging' or 'stealing'. We considered a gull to be scrounging when it attempted to take food possessed by a person, but not in their hand; only in close proximity to them, for example in a shopping bag or on a bench. We considered a gull to be stealing when it attempted to take food that a person was holding in their hand. A gull made an 'attempt' when it tried to contact an anthropogenic food item using their feet or beak.
For each food acquisition attempt, we also recorded the time of day, success at obtaining food (yes/no), age category of gull (adult/immature; determined by plumage), group composition (group with children or without), total number of feeding opportunities (each individual item is an opportunity) at the exact time of the attempt, types of food available and their quantities, type and quantity of food consumed by the gull, and human response to gull behaviour. We created five categories for human response to gull behaviour: 'Positive' if the human response to the gull approaching was to feed the gull, 'Neutral' if they simply covered their face or food, 'Negative' if they attempted to physically or verbally deter the gull, 'Mixed' if humans physically/verbally threatened the gull but still fed them and 'None' if the associated people ignored the gull. Gull food-taking attempts where gulls were purposely being fed by humans were excluded from the kleptoparasitism analyses.
We used the scan sample data to determine the total number of feeding opportunities and quantities of types of food available for each kleptoparasitic attempt. We compared the food type taken against its relative availability to determine whether gulls targeted the most abundant food types available. We estimated the percentage of each food item that the gull consumed upon its kleptoparasitic attempt as follows: <10%, 10%-25%, 25%-50% and ≥50%. For example, if a person was about to consume two pieces of fish, and the gull stole one piece, the quantity of food stolen would be 100%; if it stole half of one fish (due to the other half falling mid-flight), the quantity would be 50%. We used this metric to quantify and compare the relative rewards obtained from stealing versus scrounging food from people.
To make the time-of-day data suitable for analyses, we generated four time periods: 11:00-12:00 h, 12:00-13:00 h, 13:00-14:00 and 14:00-14:45 h. In addition to comparing food availability in different time periods, we also compared how food availability changed from one time period to another.
People most commonly consumed the following food items: pasty, ice cream, fish and chips, chips, crisps and sandwich. We used their availability and the relative frequency with which gulls attempted to take each food type to test if gulls appeared to have a food preference.
To test whether there was a significant difference between scrounging and stealing in (i) the probability of success and (ii) the proportion of the food item consumed, we used (i) a Pearson's chi-square test and (ii) a Fisher's exact test, respectively. For both analyses, we tested the observed distributions against equal distributions between categories. To test whether gulls showed any food preference in their food-taking attempts, we used the total number of different food items observed to be available to calculate the expected probability of attempts on each. We then used a chi-square test to compare the expected probability of food-taking attempts per food item to the actual number of attempts observed. We also used a chi-square test to test whether gulls were more successful in taking food from groups with children versus from adult-only groups.
We used one-way analysis of variance (ANOVA) to test whether there were significant differences in (i) the mean number of gull food acquisition attempts (i.e. stealing plus scrounging) between time periods and (ii) the mean number of gull food-stealing attempts between locations. We then used a linear mixed-effects model to test how the number of food items available and the season (pre-tourist/tourist) affected the number of stealing attempts observed, while controlling for location and time period by including these variables as random effects. All analyses were conducted in R version 4.0.2 (R Core Team 2021).
During 43.75 h of observation across the six Cornish towns we sampled, a total of 1962 food items became available. We observed 142 kleptoparasitic attempts by gulls; 90 of these were stealing attempts, which were successful 50% of the time, and 52 were scrounging attempts, successful 81% of the time. Gulls were significantly more successful at scrounging than stealing (χ 2 = 11.884, df = 1, P < 0.001), but obtained more food from stealing than from scrounging (Fisher's exact test: P < 0.001). Gulls secured ≥50% of the targeted food item in 68% of their successful stealing attempts as compared to scrounging, where they only did so in 7% of their successful attempts (online Table S2). Gulls showed no significant food preference (χ 2 = 2.868, df = 5, P = 0.720) and targeted the most common food item available (pasties) in 69% of all successful food acquisition attempts.
Controlling for time period and location, we found that higher food availability was associated with more food-stealing attempts (linear mixed-effects model: slope estimate ± se = 0.033 ± 0.013, t 23 = 2.600, P = 0.016; Figure 2). In addition, more stealing attempts were observed in the tourist season than in the pretourist season for the same food availability (slope estimate ± se = 1.148 ± 0.393, t 39 = 2.920, P = 0.006; Figure 2). We found that the majority of people's responses to stealing attempts were neutral (51 responses; 60%), followed by negative (27 responses; 32%). Other responses were rare but there were no positive responses. With regards to human group composition and its potential influence on gull stealing attempts, 85 human groups were targeted. Gulls attempted to steal from 49 groups without children where they succeeded 45% of the time, as compared to groups with children, where 36 groups were targeted and gulls succeeded 52% of the time. However, this difference in success rate was not significant (χ 2 = 0.249, df = 1, P = 0.618).
Through this study, we identified various factors associated with the frequency of gull kleptoparasitic behaviour targeting anthropogenic food in six Cornish towns. Time period was an influential factor; gulls made most food-taking attempts between 12:00 and 13:00 h. As many humans regularly consume food during this time, the high predictability of food availability explains increased kleptoparasitic behaviour (Carr & Ryes-Galindo 2017).
Location was significantly associated with stealing frequency, with Looe and St Ives hosting more stealing attempts by gulls. This may be due to the  towns' layout, where the seafront with numerous food outlets faces the sea which provides an easy escape route following food theft. While we observed differences in gull stealing frequency between towns, change in food availability from spring to the tourist season influenced different towns similarly. All towns, except for Truro, saw an increase in food availability and, subsequently, an increase in stealing attempts. This change in food availability and consequent change in food-stealing frequency suggest that gull kleptoparasitism frequency may follow an ideal free distribution, where animals distribute themselves proportional to resource availability (Emlen 1966, Morand-Ferron et al. 2009). Coastal towns attract tourists during the summer months, which could explain the trend observed in Truro as it was the only inland town sampled (Visit Britain 2016).
Finally, our findings suggest that stealing from humans tends to be highly rewarding for Herring Gulls and relatively low-risk in the towns we sampled. Gulls were successful in 50% of their attempts to steal food from people, which is at the higher end as compared to their success when attempting to steal food from other species (Källander 2006). They secured large portions of food, thus making each successful attempt that much more rewarding. In addition, the risk of serious injury is relatively low; we never observed a human contact or injure a gull, and human responses were generally neutral. Interestingly, gulls displayed no obvious food preferences; this could be due to the potential costs of being choosy (Emlen 1966). While we could identify gulls in some locations (e.g. Falmouth) due to distinguishing features or leg bands, this was usually not possible. Determining whether it is the number of gulls that leads to high rates of kleptoparasitism, or instead due to just a few individual gulls, is a fruitful area for future research.