Dominance, gender, and season influence food patch use in a group-living, solitary foraging canid

Behavioral Ecology The official journal of the ISBE International Society for Behavioral Ecology Behavioral Ecology (017), 8(5), 130 1313. doi:10.1093/beheco/arx09 Original Article Dominance, gender, and season influence food patch use in a group-living, solitary foraging canid Jo Dorning and Stephen Harris School of Biological Sciences, Life Sciences Building, University of Bristol, 4 Tyndall Avenue, Bristol BS8 1TH, UK Received 4 November 016; revised 16 May 017; editorial decision 6 June 017; accepted 16 June 017; Advance Access publication 5 July 017. In patchy environments, foragers adopt different strategies to acquire resources depending on their internal state and external physical and social environment: this has important fitness consequences. Linking individual variation in patch use to tangible characteristics is key to understand many higher-level ecological processes. We studied patch use by red foxes (Vulpes vulpes) in the city of Bristol, UK. We placed camera traps in gardens where householders provisioned foxes (patches) to investigate whether 1) foxes discriminated between patches based on food availability, quantified as provisioning frequency (predictability) and the energy value of provisioned food; and ) individual patch use varied with dominance, gender, and season. Increased frequency of provisioning encouraged more foxes to visit and to stay longer in patches. All foxes visited the most predictable patches first each day, but females were more selective and generally more efficient foragers than males. Females increased foraging effort during cub rearing, whereas males reduced patch use in the dispersal and mating season. Dominants and subordinates shared patches spatiotemporally, possibly facilitated by relatedness and familiarity between group members. However, dominants visited more food patches on their territory, spent more time in predictable patches and fed earlier than subordinates. Subordinates may compensate for competition by visiting patches of lower quality or outside their territory, which is inefficient and risky. Our results demonstrate gender differences in behavioral motivation, show how subordinates forego foraging efficiency to mitigate intra-group competition and reveal how human provisioning influences fox space use in urban areas. Key words: activity patterns, camera trapping, foraging, human wildlife interactions, provisioning, social status. INTRODUCTION Food resources are clustered into discrete patches of variable quality (Wiens 1976). To forage efficiently, animals generally select patches that offer maximum energy intake for minimal energy expenditure (Charnov 1976; Sims and Quayle 1998; Wei et al. 015; Mahenya et al. 016; Seidel and Boyce 016). However, individuals are likely to adopt different foraging strategies depending on their internal state and external physical and social environment (Pettorelli et al. 011). Relatively few empirical studies have examined how attributes such as social status, gender, and season might explain individual variation in foraging behavior within social groups (Marshall et al. 01). Understanding how animals respond to their environment is a central issue in behavioral ecology (Davies et al. 01) and modeling individual behavior is key to understanding population-level ecological processes (Evans 01). Address correspondence to J. Dorning. E-mail: jrdorning@hotmail.co.uk. Within social groups, individual foraging strategies are influenced by competition. Many mammalian societies exhibit dominance hierarchies (Christian 1970) that help reduce aggression among competitors (Alexander 1974; Drews 1993; Clutton-Brock and Huchard 013). Dominant individuals use their higher competitive ability to monopolize resources, so generally have priority access to food (Baker et al. 1981; Bosè et al. 01; Duriez et al. 01; Takanishi et al. 015) and higher feeding rates (Thouless 1990; Stahl and Kaumanns 003; Bijleveld et al. 01; Ceacero et al. 01; Wright et al. 014), especially where resources are patchy (Stahl and Kaumanns 003; Vahl et al. 005). This increases their foraging efficiency (Murray et al. 006; Gilbert-Norton et al. 013) and hence fitness (reviewed by Ellis 1995). Subordinates adopt various foraging strategies to compensate for contest competition, including being less cautious (Sih et al. 015), increasing exploration to discover new patches (Stahl et al. 001; Akbaripasand et al. 014), targeting lower quality patches (Duriez et al. 01), spending more time foraging (Stahl et al. 001; Murray et al. 006), and dietary diversification (Murray et al. 006) or specialization (Newsome et al. 015a). Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018 The Author 017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com

Dorning and Harris Patch use by red foxes 1303 Foraging is influenced by season and the associated energetic requirements, especially during reproduction when gender differences arise. Female mammals forage at higher rates when rearing young (Saunders et al. 1993; Cripps et al. 011) to offset the energetic costs of lactation (Gittleman and Thompson 1988). Conversely, male mammals typically invest in fitness and reduce foraging effort in the mating season in favor of fighting, displaying and/or mate guarding, often leading to loss of condition (Mitchell et al. 1976; Alberts et al. 1996; Forsyth et al. 005; Georgiev et al. 014; but see Girard-Buttoz et al. 014). To date, most carnivore foraging studies have been based on diet (e.g. Saunders et al. 1993; Newsome et al. 015a; Newsome et al. 016) or circadian activity, which tends to coincide with prey activity (e.g. Linkie and Ridout 011; Monterroso et al. 013; Díaz- Ruiz et al. 016). Little is known about how carnivores use patches, probably because habitats containing free-living prey are difficult to segment into discrete, spatiotemporally stable patches where food availability and exploitation can be measured. However, human provisioning as a conservation tool (Yarnell et al. 015), leisure activity (Baker et al. 004a) or inadvertently (Gilchrist and Otali 00; Bino et al. 010; Karlsson and Johansson 010; Paul et al. 015; Yirga et al. 015), creates stable, predictable food patches that are readily exploited by opportunistic carnivores (Newsome et al. 015b) and facilitate patch use studies. The red fox (Vulpes vulpes) is a contentious carnivore of global economic and ecological importance as a predator, competitor, and disease vector (Deplazes et al. 004; Soulsbury et al. 007; Saunders et al. 010; Nouvellet et al. 013), and one of the world s most invasive species (Lowe et al. 000). Foxes thrive in urban areas, where population densities are higher than rural areas (Harris and Rayner 1986; Šálek et al. 015), and food of human origin forms much of their diet (Harris 1981a; Saunders et al. 1993; Contesse et al. 004; Savory et al. 014; Šálek et al. 015). Although solitary foragers, foxes form mixed-sex social groups of increasing size as population density rises (Baker and Harris 004; Iossa et al. 008). These comprise a dominant pair and one or more subordinate adults/subadults that are often related (Baker et al. 1998; Iossa et al. 008). The accessibility of urban foxes, their exploitation of provisioned food and flexible social structure makes them ideal for answering questions about patch use both as individuals and competing members of social groups. Our aim was to quantify the effects of patch quality (food availability), individual attributes, and season/life history patterns on patch utilization in an urban fox population. Specifically, we investigated whether 1) foxes target high-quality patches, ) dominants have better access to food, 3) patch use is related to season, and 4) foraging tactics vary with gender. Because dominant foxes weigh more, live longer and have greater reproductive success than subordinates (Baker et al. 1998, 004b; Iossa et al. 008), we expected them to have greater access to resources. As fox life history patterns coincide with seasonal variation in ranging behavior (Saunders et al. 1993; White and Harris 1994; Baker et al. 001; Soulsbury et al. 011) and body condition (Harris 1981a; Saunders et al. 1993; White and Harris 1994; Iossa et al. 008; Shi 014), we also expected foraging effort and efficiency to vary seasonally, with clear gender differences during the dispersal/mating and cub rearing seasons. We used camera trapping to address our hypotheses. Although GPS and radio-tracking are widely used to study space use in freeliving animals (e.g., White et al. 1996; Lai et al. 015), they are invasive, and the sample size is limited by battery life, manpower and trappability (e.g., Baker et al. 001). Camera trapping is noninvasive and enables continuous data collection from multiple locations and individuals concurrently, without the need for trapping or direct observations (Kays et al. 011; Rowcliffe et al. 014), so is ideal for monitoring how all members of a social group utilize food patches. METHODS Study site and camera locations The study took place between July 013 and June 015 in northwest Bristol, UK, in a 1.5 km area of predominantly 1930s semidetached housing with one of the highest fox densities in the city (Harris 1981b). We estimated territory boundaries for 7 social groups using radio-tracking (N = 5 foxes), live trapping (N = 75 foxes), and camera trap photos and public sightings. We used data collected within 3 years of the study start date to assume stable territories (White et al. 1996; Baker et al. 000; Iossa et al. 009). As fox territory boundaries tend to follow roads and other topographical features (Saunders et al. 1993), we could produce accurate hand-drawn range estimates (Supplementary Figure S1). We studied gardens where foxes were provisioned deliberately and regularly (at least twice a week) by householders (food patches). This allowed us to study the key food patches on each territory, as although urban fox territories contain many potential food sources, provisioned food is the most important. To locate food patches, we delivered a questionnaire to 1706 households; of 738 responses, 80 households (10.8%) fed foxes and 143 (19.4%) other wildlife. Follow-up visits to nonrespondents yielded few new patches, so we assumed that nonrespondents did not provision foxes (Baker et al. 000). Of the 80 food patches we identified, 35 were within the target territories, provisioned foxes at least twice a week in a private residential garden and were willing to participate in the study. This provided 4 6 patches per territory (Supplementary Figure S1). Camera trap survey design We conducted 4 camera trap surveys per territory in 4 consecutive seasons, but not all were studied concurrently due to logistical constraints: see Supplementary Table S1. Seasons coincided with life history stages: they were spring (March May; birth, early cub-rearing), summer (June August; late cub rearing, the onset of juvenile independence), autumn (September November; onset of dispersal), and winter (December February; peak dispersal, mating). Season length varied from 90 to 9 days; camera trap surveys lasted 40 consecutive days (~44% of each season) and included a minimum of 4 patches per territory. Pilot studies showed that this would enable us to identify and record multiple patch visits from all foxes resident on a territory (Dorning 016). We used one camera trap (ScoutGuard SG565F-8M, Boly Media Communications, Inc., USA) per garden tied to a tree or stake, 40 70 cm above ground overlooking the provisioning area (Supplementary Figure S). Cameras were active continuously and recorded a burst of 1 3 photos per motion-triggered event. Batteries and memory cards were changed weekly. Quantifying food availability Householders were asked to record provisioning time and type and quantity of food provided for foxes each day. This was confirmed visually from camera trap photos and used to quantify provisioning frequency and energy value of provisioned food. Provisioning Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018

1304 Behavioral Ecology frequency was the mean number of days per week that foxes were provisioned (provisioning days) and energy value was the mean nutritional value of food supplied on each provisioning day in megajoules (MJ) (Saunders et al. 1993). Provisioning frequency and energy value were not strongly correlated [r s (486) = 0.45, P < 1] and variance inflation factors were < in models that included both terms, permitting their use in the same models (Graham 003; Field et al. 01). Identifying individuals We identified individual foxes from camera trap photos using a combination of fur markings and body shape, and spatiotemporal information such as location and photo timestamps (Dorning 016). We identified the foxes in 98.6% of photos and identified 175 individuals aged > 5 months old (101, 4, 3 unknown sex). Foxes <5 months old were excluded as they were not fully independent. Of the 175 foxes, 7 had been captured and eartagged previously (Baker et al. 001); 4 of the 7 tagged foxes were also radio-collared. Tags and radio-collars provided additional identifying features. Gender of unmarked animals was determined from genitalia and/or signs of lactation. Social status was established using body size, breeding status, and behavior during social interactions. Interactions were interpreted from photos containing multiple individuals and occasionally observed directly in situ. The dominant male was the largest dog fox on the territory that most often evoked submissive postures from other foxes; the dominant female was the breeding vixen that most often evoked submissive postures from other foxes (Dorning 016). All other foxes were assumed to be subordinates. Defining patch visits Photos were managed in Camera Base version 1.6. (http://www. atrium-biodiversity.org/tools/camerabase). We deducted 1 h from each photo timestamp so that nocturnal activity could be analyzed as whole nights: because foxes are primarily nocturnal in urban areas, days started and ended at noon for all analyses. Photos of an individual at a patch were grouped into independent patch visits using an inter-record time interval threshold of 15 min, where photos taken >15 min apart were assumed to represent separate visits. This threshold was chosen by plotting the number of seconds between pairs of consecutive photos of the same individual at the same patch, which revealed a clear break between consecutive photos from the same and different visits (Dorning 016). Timestamps of the first and last photo in each visit were used to calculate patch residence time (PRT) in seconds. Territory residency We used sighting frequency (number of days seen) and the number of associations ( foxes visiting the same food patch concurrently, inferred from overlapping visit times) to identify the foxes resident on each territory. During each survey, foxes were considered resident if they 1) visited a patch on the territory on at least 0 of the 40 survey days, a threshold based on distributions of sighting frequencies (Dorning 016), and ) shared at least associations with another territory resident, suggesting they were socially integrated into the resident group. Foxes not fulfilling both criteria were considered nonresidents. Foxes could only be resident in one territory; the 9 individuals fulfilling these criteria in multiple territories were considered resident in the territory with most associations and the highest sighting frequency, and nonresident elsewhere. Unless specified, analyses were conducted on data from territory residents only, as foxes were expected to behave differently outside their territory (Supplementary Table S). Statistical analyses We used several measures of foraging behavior to gain a more comprehensive understanding of patch use (Bastille-Rousseau et al. 010). We measured general trends in patch selection using the daily number of visitors to patches and the first patch visited by individual foxes each day. We measured individual foraging effort using the number of patches visited per day and daily PRT. We examined competition and foraging efficiency using all these measures, circadian activity patterns, and the identity of the first fox to arrive after provisioning. All analyses were conducted in R version 3.3.0 (R Core Team 016). Foraging activity patterns Visit start times were used to examine activity patterns in each season. We calculated mean visit start times and provisioning times in package circular (Agostinelli and Lund 013). We used kernel density estimation in package overlap (Meredith and Ridout 016) to calculate probability density functions of activity and the coefficient of overlap (delta-hat-4, which ranges from 0 to 1), between dominant and subordinate visit times, and between fox visits and provisioning times. Confidence intervals were estimated from 5000 bootstrap samples. Food availability and patch utilization All statistical models were linear or generalized linear mixed models (LMM or GLMM), fitted by maximum likelihood with package lme4 (version 1.1 11) (Bates et al. 015). We used random effects to account for repeated measures, as multiple visits were recorded for the same individuals, patches, territories, and seasons. Depending on the model, random effects were included for: territory, to account for variation in territory size, patch distribution and social structure; season, to account for seasonal life history patterns; patch, to account for variation in provisioning habits and unmeasured factors such as cover, size, and disturbance; individual, to account for variation in ranging behavior or other unmeasured factors; and an individual-by-patch interaction, to account for individual variation in the effect of patch, for example, due to site fidelity (Wakefield et al. 015) or perceived risk (Dammhahn and Almeling 01; Mella et al. 015). Model structures are summarized in Table 1. We checked residuals for homogeneity of variance and overdispersion using simulations in package DHARMa (Hartig 017). We used package influence.me (version 0.9 6) (Nieuwenhuis et al. 01) to confirm that no patch had an unusually high influence on the regression estimates. We identified the minimal significant model using stepwise model simplification, comparing model deviance with and without each fixed effect using likelihood ratio tests until only significant terms remained. Reported chi-squared values for fixed effects were obtained by comparing the deviance of each final minimal model with and without the term. We used package lsmeans (version.3) (Lenth 016) for post hoc tests and model predictions. We first tested whether food availability at patches varied between seasons (Table 1, models a and b) and whether patches with higher food availability attracted more foxes, both residents and nonresidents (Table 1, model c). We then investigated whether Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018

Dorning and Harris Patch use by red foxes 1305 Table 1 Summary of mixed models used in the analyses Measure Model Units Sample size Model type Fixed effects Random effects Provisioning a) Prov. freq. Mean days 14 Gaussian LMM Season Patch (N = 35) provisioned per week b) Energy value Mean MJ supplied 14 Lognormal (log 10 ) Season Patch (N = 35) per provisioning day LMM Patch utilization c) Visitation rate N visitors per day 4864 Poisson GLMM Prov. freq. Patch (N = 35) (log link) Energy value Season (N = 4) d) Proportion of patches visited 584 Poisson GLMM (log link) N patches visited per day, with N patches on territory as offset term e) First patch Binary score whether patch was the first visited on territory that day f) PRT Total PRT (seconds) per day g) First visitor Binary score whether first visitor was dominant each day foxes visited a greater proportion of patches on their territory if they were dominant or female and whether this was influenced by season (Table 1, model d). On each day a fox was observed, we used a binary score to indicate whether or not it visited a particular patch first, to test whether foxes were more likely to visit patches with higher food availability first, and whether this depended on gender and/or social status (Table 1, model e). To control for between-territory differences, we ranked patches by their energy value within territories, where zero was the highest and included in the intercept, one was the second highest and so on. We used these ranks as the fixed effect. We tested whether seasonal PRT reflected gender differences in seasonal energetic demands and whether PRT in patches with higher food availability varied with gender and social status (Table 1, model f). We also included a fixed effect for provisioning day (yes/no), as we expected PRT to be higher on provisioning days due to food handling time. On each provisioning day (N = 3640), we used binary scores to denote whether the first fox to arrive was dominant or subordinate and male or female. We excluded days when the first fox photo after provisioning contained more than one individual (N = 340) and when the first visitor was nonresident (N = 197). When householders put food out several times per day, we only used the first provisioning time after 1:00 GMT (the start of each day) to avoid oversampling. We used exact binomial tests to examine whether first visitors were more often dominant or female and a GLMM to test the effect of gender and food availability on the probability that the first visitor was dominant (Table 1, model g). 584 Binomial GLMM (probit link) 15 870 Lognormal (Log 10 ) LMM 3103 Binomial GLMM (logit link) Gender Individual (N = 53) Social status Season Gender Season Territory (N = 7) Status Season Prov. freq. Individual (N = 53) Energy value rank Gender prov. freq. Patch (N = 35) Status prov. freq. Gender energy value rank Individual patch Status energy value rank Gender Individual (N = 53) Social status Season Prov. freq. Energy value Patch (N = 35) Gender season Status prov. freq. Status energy value Gender Patch (N = 35) Prov. freq. Energy value Individual patch Terms shortened for clarity are: prov. freq. = provisioning frequency, that is, mean days provisioned per week; (G)LMM = (generalized) linear mixed model; PRT = patch residence time in seconds; energy value = mean energy value of food supplied per provisioning day, in megajoules; energy value rank = withinterritory rank of patch based on energy value. Ethical statement Animal capture and handling procedures followed Sikes et al. (016), were approved by the University of Bristol Ethics Committee (UB/14/015) and carried out under the Animals (Scientific Procedures) Act 1986. RESULTS We recorded 3850 patch visits by 175 foxes, of which 34 476 (89.5%) were by territory residents (N = 53 individuals: 15 dominants [8, 7 ], 38 subordinates [19, 19 ], Supplementary Table S). The 4044 visits by nonresidents (N = 157 individuals: 17 dominants [11, 6 ], 1 subordinates [70, 7, 5 unknown sex], 18 unknown statuses [10, 1, 7 unknown sex]) were excluded from all analyses except model c (Table ). Thirtyfive foxes were recorded both in their own territories as residents and in other territories as nonresidents. Foraging activity patterns Activity at foraging patches was unimodal and lowest around midday in all seasons; patch visitation peaked between 19:30 and 1:30 GMT in spring, 0:00 and :00 in summer, 16:30 and 18:30 in autumn, and 17:00 and 19:00 in winter (Figure 1). Average provisioning times were 19:33 GMT in spring, 19:47 in summer, 18:18 in autumn, and 18:34 in winter; visit times showed considerable overlap with householder provisioning in every season (coefficients of overlap: 0.57 0.67, Supplementary Figure S3). Patch visits occurred Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018

1306 Behavioral Ecology Table Summary of mixed model results Fixed effects Random effects Model In min. model? Parameter Coef. SE χ df P Parameter Var. SD % Total var. a) Prov. freq. Intercept 5.451 0.8 Patch.7 1.49 8.5 Season 4.849 3 0.183 Residual 0.47 0.687 17.5 b) Energy value Intercept 0.048 0.046 Patch 0.066 0.58 90.4 Season 8.637 3 0.035 Residual 7 0.086 9.6 Season (SU) 0.03 0.0 Season (AU) 0.061 0.0 Season (WI) 0.049 0.0 c) Visitation rate d) Proportion of patches visited Intercept 0.91 0.096 Patch 0.074 0.7 NA Prov. freq. 0.064 0.01 5.484 1 <1 Season 0.011 0.107 NA Energy value.344 1 0.16 Residual NA NA NA Intercept 0.4 0.105 Individual 0.059 0.4 NA Social status (sub) 0.161 0.080 4.04 1 0.045 Territory 0.036 0.189 NA Gender (M) 3 0.077 Residual NA NA NA Season Season (SU) 0.091 0.031 Season (AU) 0.03 Season (WI) 4 0.031 Gender season 10.188 3 0.017 Gender (M): season (SU) 0.039 0.049 Gender (M): season (AU) 0.063 0.047 Gender (M): season (WI) 0.086 0.047 Status season 4.704 3 0.195 e) First patch Intercept 1.431 0.168 Individual 0 0 NA Prov. freq. 0.094 0.01 Patch 0.44 0.651 NA Prov. freq. gender (M) 0.055 0.01 7.110 1 8 Individual 0.984 0.99 NA patch Energy value rank 0.089 0.019.489 1 <1 Residual NA NA NA Prov. freq. status (sub) 0.89 1 0.591 Energy value rank gender (M).03 1 0.138 Energy value rank status (sub) 0.606 1 0.436 f) PRT Intercept 5.481 0.05 Individual 0.013 0.116 0.4 Gender (M) 0.330 0.104 Patch 0.403 0.635 10.9 Season Individual patch 0.931 0.965 5.1 Season (SU) 0.307 0.045 Residual.358 1.536 63.6 Season (AU) 0.308 0.045 Season (WI) 0.335 0.045 Gender season 63.081 3 <1 Gender (M): season (SU) 0.308 0.070 Gender (M): season (AU) 0.453 0.068 Gender (M): season (WI) 0.05 0.068 Social status (sub) 0.468 0.1 Prov. freq. 0.046 0.09 Social status (sub) prov. freq. 0.087 0.035 6.171 1 0.013 Energy value 0.67 0.050 7.381 1 <1 Prov. day (no) 1.115 0.03 1043.800 1 <1 Social status (sub) energy value 3.640 1 0.056 g) First visitor Intercept 0.680 0.636 Patch 6.380.56 NA Gender (M) 0.550 0.134 16.641 1 <1 Residual NA NA NA Prov. freq. 0.30 0.086 7.077 1 8 Energy value 1.07 1 0.7 Coefficients are reported on the response scale for (a), the log 10 scale for (b) and (e), the log scale for (c), (d) and (f) and the probit scale for (g). Coefficients are presented for fixed effects included in the minimal significant model, indicated by tick marks. P-values from likelihood ratio tests are reported for each fixed effect, with significant values in bold; P-values for main effects are not reported for terms also included in significant interactions. All random effects included in the minimal model are shown. In Gaussian and lognormal models, residual variance was used to calculate the proportion of total variance explained by each random effect. Terms shortened for clarity are: prov. freq. = mean days provisioned per week; SU = summer, AU = autumn, WI = winter; energy value = mean energy value of food supplied per provisioning day, in megajoules; sub = subordinate; M = male; PRT = patch residence time in seconds; energy value rank = within-territory rank of patch based on energy value, where 0 is the highest and included in the intercept, 1 is the second highest etc. Reference categories were season = spring, gender = female, social status = dominant. Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018 over a longer time span in autumn and winter than spring and summer. There was substantial overlap between visit times of dominants and subordinates in all seasons (coefficients of overlap: 0.89 0.94, Figure 1). This reflected patterns within individual patches; overlap in 4 patches selected at random and tested individually was 0.70. Food availability Householders at each patch fed foxes a median of 5.950 days per week (Median Absolute Deviation [MAD] = 1.557; mean = 5.380, SD = 1.666) and provided a median of 0.931 MJ per provisioning

Dorning and Harris Patch use by red foxes 1307 Spring ^ 4 = 0.91 (0.89 0.9) Summer ^ 4 = 0.90 (0.88 0.91) 0.0 Dominant Subordinate 0.0 Dominant Subordinate 0.10 0.10 day (MAD = 0.88; mean = 1.301, SD = 1.090) (Supplementary Figure S4a). So an average patch in our study provided around half an adult fox s daily energy requirement of MJ (Saunders et al. 1993) on 5 6 days per week. There was no significant seasonal variation in provisioning frequency (LMM: χ 3 = 4.849, P = 0.183, Table, model a), despite the lower median in autumn (Supplementary Figure S4b). The energy value of provisioned food varied seasonally (log 10 LMM: χ 3 = 8.637, P = 0.035, Table, model b); householders provided the greatest energy value per provisioning day in spring (median = 1.148 MJ, MAD = 0.55) and the lowest in autumn (median = 0.818 MJ, MAD = 0.468) (Tukey contrast: SP AU t-ratio =.709, P = 0.039, Supplementary Figure S4c). Differences between patches explained >80% of the total variance in food availability in both models (Table, model a and b), suggesting that provisioning varied greatly between households (Supplementary Figure S4). Patch utilization Proportion of visits 0.0 0.10 1:00 18:00 0:00 6:00 1:00 Visitation rate During each 40-day survey, patches were visited by to 3 different foxes (mean = 8.0, SD = 3.9). Patches were visited by the largest number of foxes in winter (mean = 11.0, SD = 3.) and the fewest in summer (mean 5.5, SD =.3). Although all patches were visited by both dominant and subordinates on most days, patches (in Territory 1) were only visited by subordinates in spring. On average, each patch was visited by 3.7 foxes per day (SD = 1.3; mean range between patches 1.1 7.5). The daily number of visitors increased with provisioning frequency (GLMM: χ 1 = 5.484, P < 1, Table, model c, Supplementary Figure S5) but was not influenced by energy value (χ 1 =.344, P = 0.16, Table, model c). Autumn ^ 4 = 0.89 (0.88 0.90) Winter ^ 4 = 0.94 (0.91 0.94) Dominant Subordinate 1:00 18:00 0:00 6:00 1:00 0.0 0.10 Time of day (GMT) 1:00 18:00 0:00 6:00 1:00 Dominant Subordinate 1:00 18:00 0:00 6:00 1:00 Figure 1 Daily activity patterns of dominant and subordinate foxes, based on patch visit start times. The lines show kernel density estimates and shaded areas represent the coefficient of overlap (delta-hat-4) between dominants and subordinates. The coefficient of overlap is shown above each plot with 95% confidence intervals calculated from 5000 bootstrap samples. Number of patches visited The number of patches foxes visited on their territory varied between seasons depending on gender (GLMM: χ 3 = 10.188, P = 0.017, Table, model d, Figure ). Females visited most patches in summer (Tukey contrasts: SU SP z-ratio =.933, P = 0.043; SU AU z-ratio = 3.635, P = 4; SU WI z-ratio =.846, P = 0.056), with no significant differences between spring, autumn, and winter. Males visited fewest patches in winter (Tukey contrasts: WI SU z-ratio = 3.331, P = 0.01; WI AU z-ratio = 3.351, P = 0.011), with no significant differences between spring, summer, and autumn. There were no significant gender differences in any season. Dominant foxes visited more patches on their territory than subordinates year round (χ 1 = 4.04, P = 0.045, Table, model d, Figure ). Which patch do foxes visit first? Foxes visited patches with higher provisioning frequencies first: this effect was particularly strong for females (GLMM: χ 1 = 7.110, P = 8, Table, model e, Figure 3a) but did not vary with social status (χ 1 = 0.89, P = 0.591). Foxes also visited patches that supplied food of higher energy value first (χ 1 =.489, P < 1, Table, model e, Figure 3b), regardless of gender (χ 1 =.03, P = 0.138) or social status ( χ 1 = 0.606, P = 0.436). Patch residence time PRT for resident foxes was around 5 min per day (median = 8 s, MAD = 73). Foxes spent significantly longer in patches on provisioning than nonprovisioning days (log 10 LMM: χ 1 = 1043.800, P < 1, Table, model f, Figure 4a). The interaction between gender and season had a significant effect on daily PRT (χ 3 = 63.081, P < 1, Table, model f, Figure 4b). In spring, females had higher PRTs than males (Tukey contrasts: Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018

1308 Behavioral Ecology Dominant Subordinate 0.75 Male Female Proportion of patches visited 0.50 0.5 Spring z-ratio = 3.169, P = 0.04) and spent more time in patches in spring than any other season (Tukey contrasts: SP SU z-ratio = 6.800, P < 1; SP AU z-ratio = 6.766, P < 1; SP WI z-ratio = 7.508, P < 1). In winter, males had lower PRTs than females (Tukey contrasts: z-ratio =.945, P = 0.050), and spent significantly less time in patches compared with all other seasons (Tukey contrasts: WI SP z-ratio = 6.03, P < 1; WI SU z-ratio = 5.405, P < 1; WI AU z-ratio = 8.86, P < 1). Foxes spent more time in patches that supplied food of higher energy value ( χ 1 = 7.381, P = 1, Table, model f, Figure 4c) and this did not vary with social status (χ 1 = 3.640, P = 0.056). However, the effect of provisioning frequency on PRT varied with social status ( χ 1 = 6.171, P = 0.013, Table, model f, Figure 4d). PRT declined with provisioning frequency for subordinates Summer Autumn Winter Spring Summer Autumn Winter Figure The proportion of patches that male and female dominants and subordinates visited on their territories each day. Predictions are based on the GLMM in Table, model d. Error bars show 95% confidence intervals. (a) Probability 0.5 0.0 0.15 0.10 0.05 Male Female 1 3 4 5 6 7 Days provisioned per week (b) Probability 0.0 0.15 0.10 0.05 0 1 3 Rank of MJ per provisioning day Figure 3 The influence of (a) the interaction between gender and weekly provisioning frequency and (b) mean energy value of provisioned food in megajoules, MJ (patches were ranked within territories; patches scored 0 supplied the highest energy value, patches scored 1 suppled the second highest, etc.) on the probability that a fox will visit a patch first each day. Predicted means were generated from the GLMM in Table, model e. 95% confidence intervals are shown by shaded ribbons in (a) and error bars in (b). (ls trend = 0.041, 95% CI = 0.09 to 9) and increased for dominants (ls trend = 0.046, 95% CI = 0.011 to 0.103), though neither slope was significantly different from zero. Subordinates spent significantly more time than dominants in patches where provisioning frequency was low (Tukey contrast dominants vs. subordinates when fed 1 day per week: z-ratio =.09, P = 0.037), while there was a weak trend for dominants to spend more time in patches where provisioning frequency exceeded 4 days per week (Tukey contrast dominants vs subordinates when fed 7 days per week: z-ratio = 1., P = 0.). In the LMM, 5% of the variance in PRT was attributable to variation between individuals at different patches (Table, model f), suggesting that different individuals used patches with different intensities. However, most variation in PRT (64%) Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018

Dorning and Harris Patch use by red foxes 1309 (a) Patch residence time (s) (c) Patch residence time (s) 400 300 00 100 0 1000 750 500 50 Provisioning day was within individuals or within patches. This may be due to higher PRTs due to food handling soon after provisioning compared with other times of day, or higher PRTs for individuals that waited in gardens to be fed, or a variety of other possible factors. Who is the first visitor? The first visitor after provisioning time was significantly more likely to be dominant than subordinate (exact binomial test: probability the first visitor was dominant = 0.586, 95% CI = 0.569 0.604, P < 1, N(dominant:subordinate) = 1819:184 days, 14:8 individuals) and to be female than male (exact binomial test: probability the first visitor was female = 0.711, 95% CI = 0.695 0.77, P < 1, N(F:M) = 06:897 days, 3:19 individuals). Male first visitors were more likely to be dominant than female first visitors (GLMM: χ 1 = 16.641, P < 1, Table, model g, Figure 5) and the probability the first visitor was dominant increased with provisioning frequency (χ 1 = 7.077, P = 8) but not energy value ( χ 1 = 1.07, P = 0.7). DISCUSSION Our study is the most detailed examination of food patch use in a free-living carnivore and one of the few camera-trapping studies to investigate individual behavior (Zimmermann et al. 016). Social status influenced how foxes exploit local resources; dominants improved foraging efficiency, whereas subordinates reduced conflict and adopted risky compensatory strategies such as extraterritorial foraging. Female foxes were more efficient foragers than males, and seasonal variation in foraging effort reflected gender differences in resource prioritization. (b) Patch residence time (s) 0 Non-provisioning day Spring Summer Autumn Winter (d) Patch residence time (s) 0 1 3 4 5 6 1 3 4 5 6 7 MJ per provisioning day Days provisioned per week 300 00 100 300 50 00 150 Male Female Dominant Subordinate Figure 4 The influence of (a) provisioning day, (b) the interaction between gender and season, (c) the mean energy value (in megajoules, MJ) supplied per provisioning day and (d) the interaction between social status and weekly provisioning frequency on daily PRT. Predicted means and 95% confidence intervals, shown as error bars, were generated from the GLMM in Table, model f. Householder provisioning Foxes synchronize their activity with their main prey (Cavallini and Lovari 1994; Lovari et al. 1994), and because anthropogenic food forms the majority of the diet of urban foxes (Harris 1981a; Saunders et al. 1993; Contesse et al. 004), coinciding visit times with provisioning may improve foraging efficiency in urban areas. However, whether foxes visited patches to coincide with provisioning or householders provisioned to coincide with fox activity is unclear. As predicted, foxes increased foraging efficiency by visiting more predictable patches and/or those with food of higher energy value first. Targeting the best patch first during each foraging bout may reduce energy expenditure by minimizing the number of patches visited and help beat competitors (MacArthur and Pianka 1966; Charnov 1976; Devenport et al. 005; Deygout et al. 010). More visitors and longer PRTs at patches where householders provisioned more often may explain the perception that human provisioning increases fox numbers. However, provisioning was not related to group size: while groups were smallest in spring and summer and largest in autumn and winter (Dorning 016), there was no seasonal variation in provisioning frequency and provisioned energy value peaked in spring and was lowest in autumn. Foxes are flexible foragers exploiting a wide range of food types (Kidawa and Kowalczyk 011; Bakaloudis et al. 015), so are unlikely to be limited by householder provisioning. However, we recorded a high rate of patch use by non-residents (10.5% of all patch visits), which may explain why previous attempts to link group size in Bristol foxes to territorial resources were only partially successful (Baker et al. 000, 004a). Foxes and other carnivores share high-quality patches with neighbors (Gilchrist and Otali 00; Eide et al. 004), which may Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018

1310 Behavioral Ecology Male Female 1.00 0.75 Probability 0.50 0.5 1 3 invalidate theories relating group size to territorial resources (Hofer and East 1993). Effects of life history and gender Females were more efficient foragers than males. Despite visiting the same number of patches on their territory as males, females were more likely to target predictable patches, were more often the first visitor, and had higher PRTs than males in winter and spring. Females also foraged according to their energy requirements; PRTs were highest in spring, which was when householders supplied the greatest energy value, they visited the largest number of patches in summer, and travel further in spring and summer (Saunders et al. 1993). Increasing foraging effort during cub rearing may help females meet the energetic demands of breeding (Harris 1981a; Gittleman and Thompson 1988). We identified more than twice as many males than females (101 vs. 4 ), mainly due to nonresident males making extraterritorial movements in winter in search of mating and/or dispersal opportunities (Woollard and Harris 1990; Soulsbury et al. 011). Resident males also visited fewer patches on their territory in winter, had lower PRTs, and did not meet their energetic requirements: male foxes have lower fat reserves in late winter (Harris 1981a) and lose weight between winter and spring (Saunders et al. 1993). Competition at food patches Foxes did not have exclusive food patches (cf. Poulle et al. 1994). The highest number of visitors (3) was to a patch near the boundary between 3 territories, following the death of the resident dominant male; many visitors were nonresidents, highlighting the role of the dominant male in territorial defense (Gese and Ruff 1997; Arnold et al. 011). Although it has been argued that dominant and subordinate foxes forage in separate patches (Macdonald 1980), no patch was visited exclusively by dominants, and were only visited by subordinates, although dominants visited more patches on their territory 4 5 6 7 1 3 4 5 6 7 Days provisioned per week Figure 5 The influence of gender and provisioning frequency on the probability that the first patch visitor after provisioning time is dominant rather than subordinate. Predicted probabilities were generated from the GLMM in Table, model g. Error bars show 95% confidence intervals. than subordinates. The considerable overlap between dominant and subordinate visit times shows that foxes did not avoid competition by temporal separation, unlike other species (Bethge et al. 009; Heurich et al. 014). It is unlikely that resources were insufficiently clumped to defend (Gyimesi et al. 010). However, because ~10% of households in our study area provisioned foxes, food availability may have been high enough to permit patch sharing (Stears et al. 014); this may have been facilitated by the high level of relatedness within groups (West et al. 00; Iossa et al. 009) and familiarity from repeated encounters (Sillero-Zubiri and Gottelli 1995; Utne-Palm and Hart 000). However, subtle behavioral differences suggest competition. Dominants improved their foraging efficiency by spending more time in patches where food was more predictable, although the trend was weak. Because we often observed foxes waiting to be fed, higher PRTs could indicate that foxes waited longer to be fed in more reliable patches. Conversely, subordinates appeared to avoid more predictable patches, thereby avoiding dominants (Duriez et al. 01); less predictable patches are more difficult and/or less profitable for dominants to defend (Goldberg et al. 001; Revilla and Palomares 001). This suggests dominants and subordinates use information about the same resources differently (Carter et al. 016). Because we only studied patches where foxes were provisioned at least twice per week, we may have under-estimated the number of less predictable patches only visited by subordinates. Many nonresidents were subordinates, suggesting they compensate for intra-group competition by visiting patches in other territories: subordinates take more risks when foraging (Koivula et al. 1994; Shi 014) and extraterritorial movements are stressful as intruders risk agonistic encounters with residents (Young and Monfort 009). As we predicted, dominant foxes were more often the first to arrive after provisioning, particularly at more predictable patches (Deutsch and Lee 1991; Duriez et al. 01), whereas subordinates visited later (Gilbert-Norton et al. 013; Carter et al. 016). Conflict avoidance by subordinates has a greater influence on access to food than aggressive interactions (Thouless 1990), and Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018

Dorning and Harris Patch use by red foxes 1311 plays a greater role in maintaining stable dominance relationships (Kaufmann 1983). Male first visitors were more likely to be dominant than females, largely because subordinate males were rarely ( 9% of provisioning days) the first visitor. There was little difference between dominant and subordinate females in the probability of visiting first. In conclusion, our data confirm that human provisioning in urban areas attracts more foxes for longer, and demonstrate how dominance, gender, and season influence the behavior of urban foxes. This behavioral flexibility facilitates coexistence between foxes living at high population densities and provides some insight into the worldwide success of the species. SUPPLEMENTARY MATERIAL Supplementary data are available at Behavioral Ecology online. FUNDING This work was supported by the League Against Cruel Sports and the Royal Society for the Prevention of Cruelty to Animals. We thank the householders who allowed us access to their gardens for this study and the editor and two reviewers for their constructive comments that greatly improved our manuscript. Data accessibility: Analyses reported in this article can be reproduced using the data provided by Dorning and Harris (017). Handling editor: David Stephens REFERENCES Agostinelli C, Lund U. 013. R package circular : circular statistics, version 0.4 7. Available from: https://r-forge.r-project.org/projects/circular. Accessed June 017. Akbaripasand A, Ramezani J, Krkosek M, Lokman PM, Closs GP. 014. Does social status within a dominance hierarchy mediate individual growth, residency and relocation? Oecologia. 176:771 779. Alberts SC, Altmann J, Wilson ML. 1996. Mate guarding constrains foraging activity of male baboons. Anim Behav. 51:169 177. Alexander RD. 1974. The evolution of social behavior. Annu Rev Ecol Syst. 5:35 383. Arnold J, Soulsbury CD, Harris S. 011. Spatial and behavioral changes by red foxes (Vulpes vulpes) in response to artificial territory intrusion. Can J Zool. 89:808 815. Bakaloudis DE, Bontzorlos VA, Vlachos CG, Papakosta MA, Chatzinikos EN, Braziotis SG, Kontsiotis VJ. 015. Factors affecting the diet of the red fox (Vulpes vulpes) in a heterogeneous Mediterranean landscape. Turkish J Zool. 39:1151 1159. Baker MC, Belcher CS, Deutsch LC, Sherman GL, Thompson DB. 1981. Foraging success in junco flocks and the effects of social hierarchy. Anim Behav. 9:137 14. Baker PJ, Funk SM, Bruford MW, Harris S. 004b. Polygynandry in a red fox population: implications for the evolution of group living in canids? Behav Ecol. 15:766 778. Baker P, Funk S, Harris S, Newman T, Saunders G, White P. 004a. The impact of human attitudes on the social and spatial organization of urban foxes (Vulpes vulpes) before and after an outbreak of sarcoptic mange. In: Shaw WW, Harris LK, Vandruff L, editors. Proceedings of the 4th International Symposium on Urban Wildlife Conservation. Tucson: University of Arizona College of Agriculture and Life Sciences. p. 153 163. Baker PJ, Funk SM, Harris S, White PC. 000. Flexible spatial organization of urban foxes, Vulpes vulpes, before and during an outbreak of sarcoptic mange. Anim Behav. 59:17 146. Baker PJ, Harris S. 004. The behavioural ecology of red foxes in urban Bristol. In: Macdonald DW, Sillero-Zubiri C, editors. Biology and conservation of wild canids. Oxford (UK): Oxford University Press. p. 07 16. Baker PJ, Harris S, Robertson CPJ, Saunders G, White PCL. 001. Differences in the capture rate of cage-trapped foxes (Vulpes vulpes) and their implications for rabies contingency planning in Britain. J Appl Ecol. 38:83 835. Baker PJ, Robertson CPJ, Funk SM, Harris S. 1998. Potential fitness benefits of group living in the red fox, Vulpes vulpes. Anim Behav. 56: 1411 144. Bastille-Rousseau G, Fortin D, Dussault C. 010. Inference from habitat-selection analysis depends on foraging strategies. J Anim Ecol. 79:1157 1163. Bates D, Mächler M, Bolker B, Walker S. 015. Fitting linear mixed-effects models using lme4. J Stat Softw. 67:1 48. Bethge P, Munks S, Otley H, Nicol S. 009. Activity patterns and sharing of time and space of platypuses, Ornithorhynchus anatinus, in a subalpine Tasmanian lake. J Mammal. 90:1350 1356. Bijleveld AI, Folmer EO, Piersma T. 01. Experimental evidence for cryptic interference among socially foraging shorebirds. Behav Ecol. 3:806 814. Bino G, Dolev A, Yosha D, Guter A, King R, Saltz D, Kark S. 010. Abrupt spatial and numerical responses of overabundant foxes to a reduction in anthropogenic resources. J Appl Ecol. 47:16 171. Bosè M, Duriez O, Sarrazin F. 01. Intra-specific competition in foraging griffon vultures Gyps fulvus: 1. Dynamics of group feeding. Bird Study. 59:18 19. Carter AJ, Torrents Ticó M, Cowlishaw G. 016. Sequential phenotypic constraints on social information use in wild baboons. Elife. 5:e1315. Cavallini P, Lovari S. 1994. Home range, habitat selection and activity of the red fox in a Mediterranean coastal ecotone. Acta Theriol. 39:79 87. Ceacero F, García AJ, Landete-Castillejos T, Bartošová J, Bartoš L, Gallego L. 01. Benefits for dominant red deer hinds under a competitive feeding system: food access behavior, diet and nutrient selection. PLoS One. 7:e3780. Charnov EL. 1976. Optimal foraging, the marginal value theorem. Theor Popul Biol. 9:19 136. Christian JJ. 1970. Social subordination, population density, and mammalian evolution. Science. 168:84 90. Clutton-Brock TH, Huchard E. 013. Social competition and its consequences in female mammals. J Zool. 89:151 171. Contesse P, Hegglin D, Gloor S, Bontadina F, Deplazes P. 004. The diet of urban foxes (Vulpes vulpes) and the availability of anthropogenic food in the city of Zurich, Switzerland. Mammal Biol. 69:81 95. Cripps JK, Wilson ME, Elgar MA, Coulson G. 011. Experimental manipulation of fertility reveals potential lactation costs in a free-ranging marsupial. Biol Lett. 7:859 86. Dammhahn M, Almeling L. 01. Is risk taking during foraging a personality trait? A field test for cross-context consistency in boldness. Anim Behav. 84:1131 1139. Davies NB, Krebs JR, West SA. 01. An introduction to behavioral ecology. 4th ed. Chichester (UK): Wiley & Sons. Deplazes P, Hegglin D, Gloor S, Romig T. 004. Wilderness in the city: the urbanization of Echinococcus multilocularis. Trends Parasitol. 0:77 84. Deutsch JC, Lee PC. 1991. Dominance and feeding competition in captive rhesus monkeys. Int J Primatol. 1:615 68. Devenport JA, Patterson MR, Devenport LD. 005. Dynamic averaging and foraging decisions in horses (Equus callabus). J Comp Psychol. 119:35 358. Deygout C, Gault A, Duriez O, Sarrazin F, Bessa-Gomes C. 010. Impact of food predictability on social facilitation by foraging scavengers. Behav Ecol. 1:1131 1139. Díaz-Ruiz F, Caro J, Delibes-Mateos M, Arroyo B, Ferreras P. 016. Drivers of red fox (Vulpes vulpes) daily activity: prey availability, human disturbance or habitat structure? J Zool. 98:18 138. Dorning J. 016. Social structure and utilisation of food patches in the red fox, a solitary foraging canid [PhD thesis]. [Bristol (UK)]: University of Bristol. Dorning J, Harris S. 017. Data from: dominance, gender, and season influence food patch use in a group-living, solitary foraging canid. Behav Ecol. doi:10.5061/dryad.36910. Drews C. 1993. The concept and definition of dominance in animal behaviour. Behaviour. 15:83 313. Duriez O, Herman S, Sarrazin F. 01. Intra-specific competition in foraging griffon vultures Gyps fulvus:. The influence of supplementary feeding management. Bird Study. 59:193 06. Downloaded from https://academic.oup.com/beheco/article-abstract/8/5/130/394496 by guest on 8 December 018