Seasonal and circadian patterns of herring gull ( Larus smithsoniansus ) movements reveal temporal shifts in industry and coastal island interaction

1. Anthropogenic food subsidies attract opportunistic generalists like gulls in high densities, which may lead to negative impacts on human communities and local ecosystems. Managing impacts requires understanding why gulls use particular natural or industrial sites at different times of day or phases of the breeding


| INTRODUC TI ON
Human activity has significantly modified the structure of habitats and the availability of food resources in and around urban areas.
Understanding the behavioural response of wildlife to humandominated landscapes is important for assessing how urbanisation influences population dynamics, the function of local ecosystems, and the occurrence of conflicts between wildlife and humans (Ditchkoff et al., 2006;Schell et al., 2021).While urban development typically eliminates most native vertebrate species (McKinney, 2008), for those species capable of adapting to novel, human-provided food sources, urbanisation can provide significant and predictable dietary subsidies (Ditchkoff et al., 2006;Oro et al., 2013).This can lead to increases in human-wildlife conflict, such as physical attacks, property damage, wildlife-vehicle collisions or zoonotic disease transmission, and the need for management strategies (Schell et al., 2021).
Gulls in particular are considered strong urban adapters, as many species have undergone rapid population increases near urban and industrial areas worldwide linked to their capacity to exploit the predictable availability of food subsidies (Meléndez-Arteaga et al., 2022;Oro et al., 2013;Spelt et al., 2019;Wilhelm et al., 2016).Reliance on subsidies and increases in urban gull densities can have negative consequences for residents and businesses, local ecosystems and for the gulls themselves.Concentration of gulls in urban areas often leads to increases in conflicts with people and businesses from noise, aggressive behaviour toward people or damage to urban structures (Belant et al., 1998;Huig et al., 2016).Gulls also serve as biovectors by transporting and depositing pathogens (Belant et al., 1998;Nelson et al., 2008;Navarro et al., 2019) or persistent contaminants and toxic elements in areas where they congregate (Geizer et al., 2021;McIntyre et al., 2022;Signa et al., 2013), including urban and industrial sites but also island breeding and roost sites.On islands, gull aggregations can result in nutrient deposition and subsequent changes in plant communities (Ellis et al., 2006;Hill et al., 2019;Vidal et al., 2000) and can cause displacement or increased predation of other nesting seabirds (Oro et al., 2013;Scopel et al., 2017;Whittam & Leonard, 1999).Further, attraction of gulls to urban areas and industrial sites often has direct negative impacts on the gulls themselves, including increased risk of vehicle collisions (Bishop & Brogan, 2013), entanglement in or accidental ingestion of dangerous non-food items like plastics (Stewart et al., 2020) or harmful contaminant and pathogen exposure (Sorais et al., 2020).
To mitigate the negative consequences of attracting and subsidizing local gull populations, management strategies may include reducing gull access and reliance on particular food subsidies.However, the consequences of those management actions must be carefully considered, especially if target gull species are of conservation concern, or if reducing subsidies could result in knock-on effects on sensitive local ecosystems.Significant declines in local gull populations have occurred in response to sudden reductions or losses of reliable human-based resources, such as the closure of landfills (Delgado et al., 2021) or fur farms (Meléndez-Arteaga et al., 2022), or the collapse of fisheries or discard reforms (Bicknell et al., 2013;Wilhelm et al., 2016).Because many gull species exhibit high plasticity and adaptability in foraging strategy, they are likely to respond to the reduction or loss of one food resource by switching to an alternative (Calado et al., 2020;Langley et al., 2021;Zorrozua et al., 2020), which may or may not compensate for a subsidy loss.Regardless, this can potentially lead to higher pressure on local ecosystems from increased gull presence or predation on other seabirds (Bicknell et al., 2013;Stenhouse & Montevecchi, 1999;Votier et al., 2004) or to increased gull reliance on other local industries (Zorrozua et al., 2020), both of which could be expected to vary in intensity through time.It follows that the management outcome of reductions or incubation and early chick rearing.Occurrence at fish plants gradually increased until after fledging when attendance was highest from Aug-Oct coincident with the peak of Atlantic herring (Clupea harengus) processing and was more likely during the weekdays, during working hours, and during low and flood tide.4. Gulls in southwest Nova Scotia, Canada, have the behavioural flexibility to adapt to both natural rhythms and human schedules when beneficial, enabling them to thrive in a region where industry and natural resources are abundant.These findings can provide information to guide when and where to test different subsidy management strategies locally, while also considering potential increased pressures on island ecosystems.We emphasise that management outcomes of reductions of food subsidies for opportunistic species depend on multiple factors, including availability of alternative food sources and timing of use.
removal of particular subsidies depend, in part, on the proportion of the local gull population using the subsidy, the available alternatives and the degree and timing of reliance by gulls.
One means of inferring degree and timing of reliance is to examine how closely gull behavioural patterns align in time and space with predictable daily and seasonal structures in industry activity and subsidy availability.Studies have demonstrated that gull use of different natural and human-influenced habitats can vary temporally, often altering their schedules and habitat use to match fluctuating resource availability in accordance with industry seasonality, the work week, or working hours (e.g.Ouled-Cheikh et al., 2020;Parra-Torres et al., 2020;Spelt et al., 2021;Yoda et al., 2012).At the same time, predictable natural temporal rhythms of tide cycles and daylight influence natural prey activity, distribution and accessibility (Cox et al., 2013;Naylor, 2001;Yoda et al., 2012), while gull resource requirements and limitations on foraging range from the nest (i.e.central place foraging constraints) also shift in relation to the breeding cycle (Orians & Pearson, 1979;Spelt et al., 2019).The objective of this study was to quantify seasonal and circadian patterns in American herring gull (Larus smithsoniansus) access to different anthropogenic food subsidy types and coastal islands during the breeding season in southwest Nova Scotia (SWNS), Canada.Breeding gulls in this region are heavily influenced by local industry activity, where diets can contain a high proportion of fisheries-derived food sources (Shlepr et al., 2021), and birds travel large distances and spend long durations at particular industry sites, while otherwise paying lengthy visits to nearby coastal islands (Gutowsky et al., 2021).Like many gull species, it is also possible that there is a high degree of dietary specialisation and foraging strategy among colonies and individuals (Bolnick et al., 2003;Ceia & Ramos, 2015;O'Hanlon et al., 2022), and thus general patterns in resource use may be difficult to assess.We tested for consistent temporal patterns of gull-industry interactions in relation to phases of the breeding cycle (from pre-laying through fledging), natural rhythms (daylight, tide cycles), and human schedules (time of day, day of week, fisheries seasons).We discuss how our results might be considered in the development of strategies for reducing anthropogenic food subsidies available to gulls in the region, where the ultimate goal is to reduce pressure on sensitive components of coastal island ecosystems in SWNS (Smith, 2019).More broadly, our findings provide insight into the behavioural flexibility of opportunistic species to adapt to human schedules, and how their adaptability enables them to thrive in regions where both industry and natural resources are abundant.

| Study sites and data collection
Our study was undertaken in SWNS, in the Mi'kmaq district of Kespukwitk, which is considered an important Canadian biodiversity hotspot hosting a significant number of designated species at risk (SAR), including plants (e.g.eastern mountain avens (Geum pecki)), reptiles (e.g.eastern ribbonsnake (Thamnophis sauritus)) and birds (e.g.piping plover (Charadrius melodius); Government of Canada, 2020).This region is a hub of economic activity from fisheries-based industries, notably seasonal fisheries, wharfs and shore-based fish processing plants targeting primarily American lobster (Homarus americanus), Atlantic herring (Clupea harengus) and numerous groundfish species, as well as a declining American mink (Neovison vison) farming industry.The predictable availability of waste from these activities (e.g.fisheries discards or mink feed) attracts and supports locally high American herring gull populations (Gutowsky et al., 2021;Shlepr et al., 2021), which can be a nuisance and a health and safety hazard to these industries and the surrounding communities.Increased gull densities can also lead to biophysical effects on the coastal ecosystem (e.g.nutrient and heavy metal deposition, depredation of fauna; Geizer et al., 2021;McIntyre et al., 2022;Smith, 2019).Of particular concern are impacts of large gull populations on coastal islands; SWNS has an abundance of islands, many undeveloped and undisturbed, with highly productive intertidal zones, rare plant communities (Nova Scotia Department of Lands and Forestry, 2021), and significant nesting and migratory stop-over habitat for assemblages of bird species, including numerous SAR, distinct from those found in adjacent mainland areas (e.g.roseate tern (Sterna dougallii), Leach's storm-petrel (Hydrobates leucorhous), harlequin duck (Histrionicus histrionicus), purple sandpiper (Calidris maritima); Smith, 2019).
We deployed GPS devices on 15 herring gulls at two breeding colonies (Bon Portage Island (43.469°N, 65.753° W) and Whitehead Island (43.664°N, 65.867° W)) in 2019 in Lobster Bay, SWNS (Figure 1; Table 1).The two colonies are 23 km apart and are each within 25 km of at least 30 active fish processing plants and over 200 coastal islands (Figure 1).Incubating gulls at both colonies were captured in May with drop traps over nests and devices attached with a Teflon ribbon leg-loop harness (Mallory & Gilbert, 2008).Solar-powered devices from Ecotone Telemetry (Harrier-M or Harrier-L, Kite-M, and URIA 240) uploaded data to a colony base station using UHF radiowaves when within 0.2-8.0km.Thus, equipped gulls were not monitored in-the-field following device deployment.Duty cycles were programmed to record at 15 min sampling intervals for periods coincident with the breeding season (late April to early October, see Section 2.2), and switched to 1-24 h sampling intervals during the non-breeding period.Longer intervals and complete data gaps occurred sporadically for some devices, likely due to temporary battery depletion, or device or base station error.Some devices ceased transmission within 1 month of deployment while others continued to function through three breeding seasons.Weekly sample sizes are thus reported for the number of individual birds and the number of 'bird-years', since some individual birds contributed data from multiple breeding seasons in a given week (Table 1).This study was undertaken as a large collaboration between multiple primary investigators and institutions; fieldwork was carried out principally under Canadian Wildlife Service (CWS) Banding Permits Acadia University (Permit 04-18) and University of New Brunswick (Permit 14027).While we were not able to assess potential effects of device-deployment on equipped gulls, a related study deploying similar devices with the same attachment method on herring gulls at another colony in SWNS found no significant long-term effects on gull movement or productivity (Shlepr et al., 2021).

TA B L E 1
Weekly sample sizes for GPS-tracked herring gulls equipped at two breeding colonies in Lobster Bay, southwest Nova Scotia, Canada between 2019 and 2021.

| Data preparation
Data manipulation, visualisation and analyses were conducted using R 4.0.2(R Core Team, 2020).We quality-controlled raw positional datasets for each bird using the SDLfilter package (Shimada et al., 2012).
We removed temporal duplicates and positions associated with unrealistic movement rates using a maximum of 100 km/h for herring gulls (Rock et al., 2016).Quality-controlled datasets for each bird were interpolated to 15-min intervals using the interpolateTime function in the move package (Kranstauber et al., 2019), without interpolating over data gaps in the time series >1 h.We limited the dataset to positions within a time window overlapping the entire breeding period, from 30 April through 6 October.Because nest fate of equipped gulls was not monitored, we discuss results relative to general dates of the breeding period most likely to correspond with distinct breeding phases (incubation, early chick rearing and late chick rearing).Nests are initiated in May with peak laying around 14 May, and incubation is shared by the breeding pair to keep eggs covered ~98% of the time for 30-32 days until hatch (Weseloh et al., 2020).Following hatching, adult attendance to chicks is highest in the first 30 days (Weseloh et al., 2020), referred to as 'early chick rearing' in the 5-week period following incubation.
Chicks typically remain on the breeding territory until they fledge at 42-48 days old in late July or early August (Weseloh et al., 2020), thus the remainder of the breeding period until mid-August was approximated as 'late chick rearing', and adults are assumed to be mostly relieved of parental duties following this phase.Finally, we removed sequences of positions more than 150 km from SWNS in the first and last weeks of this window to discount movements of late migratory returns and early migratory departures.1).Minima and maxima in the sea-level time series were calculated using the extrema function in the Tides package (Cox & Schepers, 2017).Low tide was defined as 1.5 h before and after the minima, and high tide as 1.5 h before and after the maxima.Flood tide was assigned to all positions after low tide but before high tide, and ebb tide was assigned to all positions after high tide but before low tide.We used the getSunlightTimes function in the suncalc package (Thieurmel & Elmarhraoui, 2019)

| Evaluating seasonal and circadian patterns of occurrence
We ran generalised linear mixed effects models (GLMMs) with a binomial response variable for occurrence at each destination in Table 2. GLMMs were based on binary categorisation of positions (i.e.no = 0 and yes = 1), and a categorical fixed effect for week as well as time of day, tide phase, weekend or weekday, colony and a random factor for individual identity, using the glmer function of the lme4 package (Bates et al., 2015).Individual was included as a random factor to take into account the individual effect in the variability of the response, as gulls contributed positions repeatedly and unevenly throughout the year and over multiple years and are known to exhibit individual tendencies in occurrence patterns (Gutowsky et al., 2021).
We first ran a GLMM with occurrence at the territory to confirm that gull movement patterns reflected expected behaviour relative to likely breeding phase (incubation, early chick rearing and late chick rearing).Weekly patterns of territory attendance corresponded closely to presumed breeding phases (Figure S2), thus we are confident when discussing patterns of occurrence by week at other destinations in relation to likely breeding phase status, and also in our ability to detect general temporal patterns in gull behaviour.We then ran GLMMs for occurrence at each of the other destinations in Table 2 and the same set of fixed and random effects.We tested overall significance of the multi-level categorical predictors by evaluating the statistical difference in slopes estimated by our mixed effect models using a Wald Chi-square test of Type III calculated with the Anova function in the car package (Fox & Weisberg, 2019).Where support for predictors was found, post hoc tests were performed to compare pairwise differences in the effects within the categorical predictor with the emmeans package (Lenth, 2021).We assessed model predictive performance with the area under the curve (AUC) statistic using the pROC package (Robin et al., 2011) by testing how well models built with a training dataset (80%) predicted the outcome of the response on a test dataset (20%).AUC > 0.5 indicates that the model classifies the binomial response better than chance; the closer the AUC is to 1, the better the model.Models yielded predicted probabilities of occurrence with 95% confidence intervals generated for each destination type (using the effects package; Fox, 2003, Fox & Weisberg, 2019): probabilities were plotted against week to examine and compare seasonal patterns in occurrence at the destinations.

| Examining behaviour at high-use islands and fish plants
We provide in Data S1 further context for environmental conditions at island and fish plants in the study area, and gull behaviour at these destination types.We summarised the ecological value of the most heavily visited islands (e.g.species of birds using the islands during breeding, migration or overwinter, any SAR, rare plants or lichens), interviewed fish plant owners and workers regarding plant operations (e.g. products processed, seasonality, operating schedule) and gull activity, and made direct observations of gulls at island and fish plant sites in the study area.For each colony, we identified the primary island and primary fish plant for which occurrence was highest based on number of bird-years and positions, to provide detailed examples of how gulls interact with island and fish plant sites.We examined the geographical distribution of positions at each primary site relative to time of day and tide phase in order to visualize changes in space use relative to features of the sites.Areas of highest use during each time period of interest were approximated by generating pooled KDEs from all positions (i.e.positions are pooled at the group-level rather than by individual) at the primary site during a particular time of day or tide phase category using the adehabitatHR package (Calenge, 2006, with an ad hoc smoothing parameter and a grid size of 300), taking the 50% KDE contour to represent core area used by all birds occurring at the site during that time.

| RE SULTS
Over the course of the study (2019-2021), 15 gulls tracked over 3 years from two colonies overlapped with 59 coastal islands and 39 fish plants in SWNS (for ≥10 h total; Figure 1; Table 1).Total May (Table 1).

| Seasonal patterns of occurrence
There was strong evidence for week as a categorical predictor of probability of occurrence at six of seven destination types (all but mink farms) over the 23-week period (all 2 23 tests p < 0.001), revealing clear temporal in gull behaviour across the breeding season.Tests of model predictive performance indicated reasonable predictive power based on AUC values ranging from 70% to 82% accuracy in classification of the binomial response (Table 2).
Mean predicted probability of occurrence of gulls on the breeding territory peaked from mid-May to mid-June, corresponding to the incubation phase (Figure S2).Occurrence on the breeding colony island but outside of the territory followed a similar pattern to territory, but with lower overall probabilities (Figure 2).Probability on territory did not differ between colonies but there was some evidence that gulls from Whitehead were more likely to be found at the colony than gulls from Bon Portage (territory: 2 2 = 1.09, p > 0.1, colony: 2 2 = 5.80, p = 0.02; Figure 3).Occurrence on other islands was low throughout incubation and chick rearing, reaching lows of 3% in the week of incubation when probability of occurrence on the territory or colony was highest relative to other destinations (Figure 2).Probability of occurrence on mainland coastal areas also remained relatively low until late chick rearing when it increased through to when young had likely fledged (Figure 2).Colony was not a predictor of occurrence on other islands or coastal areas (islands: 2 2 = 1.78, p = 0.18, coastal: 2 2 = 0.8 , p = 0.78; Figure 3).Probability of occurrence offshore was higher than all other destinations from before incubation through to late chick rearing, ranging from 43% to 58%, then dropped to 9% in the week of Aug-26 (i.e. after fledging; Figure 2).There was strong evidence that occurrence offshore differed among colonies ( 2 2 = 106.40,p < 0.001): probability offshore holding all other effects at their means (i.e. over the 23-week period or any time of day) for Whitehead was higher (56%) compared to Bon Portage (8%; z = 10.31,p < 0.001; Figure 3).
Occurrence at fish plants gradually increased from incubation to late chick rearing, becoming the most likely destination after fledging (Figure 2).There was no evidence for colony as a predictor of occurrence at fish plants ( 2 2 = 0.02, p = 0.89).

| Circadian patterns of occurrence
Occurrence of gulls on the territory and colony differed markedly within a 24-h period in relation to the time of day and tide phase (support for pairwise differences among all levels of the time of day and tide phase effects, p < 0.001).The probability of occurrence at both destinations was most likely during dusk and night and was most likely during ebb tide (Figure 3).Gulls were also 2% more likely to occur on the territory on weekdays than weekends (z = 10.69,p < 0.01; Figure 3).
Gulls were also more likely to occur on other islands during dusk and night compared to dawn through afternoon (Figure 3), with support for pairwise differences among all levels of the time of day effect (p < 0.001) except morning:mid-day (z = 0.94, p = 0.34), morning:afternoon (z = 0.16, p = 0.87) and mid-day:afternoon (z = 0.82, p = 0.41).Occurrence on other islands was 1% more likely at flood and high tide compared to ebb and low tide (Figure 3), with no support for a difference between ebb:low (z = 0.52, p = 0.60).Gulls were more likely to occur on other islands on weekdays than on weekends (z = 7.76, p < 0.001) but only by 0.7% (Figure 3).
Occurrence on the coastal mainland was most likely at midday and afternoon (Figure 3), with support for pairwise differences among time of day levels (p < 0.001) except dawn:morning (z = 1.08, p = 0.28) and mid-day:afternoon (z = 0.70, p = 0.48).Occurrence was 2%-4% more likely at flood tide than at other tide phases (Figure 3), with no significant difference between low:high (z = −2.32,p = 0.02).
Gulls were 1% more likely on the coastal mainland on weekends than weekdays (z = 15.00,p < 0.001).
From dusk through dawn, the most likely destination for a gull was offshore; the probability of occurrence offshore was greater at dusk, night and dawn compared to other destinations, holding all other effects at their means (i.e. over the 23-week period), being 33% more likely than the next most probable destination (territory; Figure 3).Gulls were least likely offshore from morning to afternoon (Figure 3), with support for pairwise differences among all levels of the time of day effect (p < 0.001).Gulls were more likely offshore during high and ebb tide than at low or flood tide, with support for pairwise differences among all tide phases (p < 0.001) and were 3% more likely offshore on a weekday than weekend (z = 13.13,p < 0.001).
During mid-day, the most likely destination for a gull was at a fish plant; mean predicted probability of occurrence at a fish plant at mid-day (32%) was higher than all other destinations (with all other effects including week held at their weighted means; Figure 3).Gulls were most likely at fish plants from morning to afternoon (Figure 3), with support for pairwise differences among all levels of time of day (p < 0.001).Gulls were also 3% more likely at a fish plant during low and flood tide than at high or ebb tide (Figure 3), with support for pairwise differences among all tide phases (p < 0.001).Occurrence at fish plants was 4% more likely to occur on weekdays than weekends (z = 22.08, p < 0.001).and 50 m across at its widest (Figure S3), located 8.3 km from the centre of the Bon Portage territories and 1.7 km from the primary fish plant used by gulls from this colony.The island was less heavily and less consistently used across bird-years, compared to the primary island for Whitehead gulls, with 71% of birds and 56% of bird-years occurring on the island at some point, but only from dusk to dawn (Figure S3).Just 1.2% of all positions occurred on Little Green Island, and total time spent varied from 0% to 6% of known

| Use of fish plant sites
Gulls from Whitehead Island attended 12 fish plant sites each for ≥10 h.The primary fish plant site used by gulls tracked from Whitehead was 8.5 km from the colony, and was a cluster of fish processing plants and a large fishing wharf, with four known active effluent pipe sites (Figures S5 and S6).This site was attended by Whitehead gulls for all bird-years (20), with 9% of all positions.Total time spent ranged from 1% to 23% of positions per bird-year.
Bon Portage Island birds attended 32 fish plant sites each for ≥10 h.The primary fish plant site used by gulls tracked from Bon Portage was 10 km southeast from the colony, and comprised two fish processing plants and one relatively small wharf, and one active effluent pipe (Figure S5).This site was attended by 10 birds (12 birdyears) from Bon Portage, with 5.5% of all positions, but only three bird-years (two birds) spent ≥10 h each.These three bird-years spent 9%-33% of their time at the site, while the remaining birds spent <1%.More detailed information about the primary fish plants is provided in Data S1.

| DISCUSS ION
Our research examined the behavioural flexibility of an opportunistic species to adapt to a human-dominated landscape by taking advantage of predictable temporal rhythms in subsidy availability from multiple sources.American herring gulls in SWNS timed their interactions with both local industry activity and coastal ecosystems to match human schedules and natural cycles at multiple temporal scales.We found that gull occurrence on other islands and interaction with local industry varied markedly across the breeding period.
Moreover, we detected circadian patterns in occurrence relative  et al., 2020;Parra-Torres et al., 2020), but the degree of reliance on particular anthropogenic food subsidies can vary through the breeding season (e.g.Huig et al., 2016;Spelt et al., 2019).Local fisheries in SWNS follow distinct seasonal rhythms which, in concert with shifting demands of the breeding cycle, likely explain in part the observed temporal patterns in occurrence of American herring gulls in the offshore environment.The probability of offshore occurrence was higher than all other destinations from pre-breeding through the middle of chick rearing, and gulls were 33% more likely to occur offshore at dusk and night than the next most likely destination, the territory.As gulls forage by sight, incident light from brightly-lit vessels could improve natural foraging opportunities at night, while also indicating potential availability of discards, as observed for herring gulls in the North Sea (Garthe et al., 2016).
Local SWNS fisheries vessels typically operate around the clock on all days of the week, which may explain why occurrence offshore was only 3% more likely on weekdays than weekend.The highly productive lobster fishery season in the region runs late November to the end of May (Fisheries and Oceans Canada, 2019), overlapping the early gull breeding period.Lobster fishers use various baits (often Atlantic herring or unwanted bycatch from other fisheries) that are discarded and replaced with fresh bait while at sea, thus breeding birds may follow vessels offshore for an 'easy meal' during the incubation period, as reported for herring gulls in Maine (Goodale, 2001).In addition to lobster fishing, groundfish fisheries in SWNS run year-round with the most vessels in the water during summer when there are fewer storms (J.Kerr, fish plant owner, pers. comm.), and seasonal trends in neighbouring groundfish fisheries indicate that at-sea discard rates are highest from April to June (Keyser et al., 2022).Thus, gulls foraging offshore at night may switch from following and scavenging behind both lobster and groundfish vessels in April and May during incubation to focusing on groundfish discards after hatching.
Around fledging, fish processing plants became the most likely destination from mid-August through mid-September, especially during plant working hours, which typically peak at mid-day (see Data S1 for further detail of operations at primary fish plants).No longer constrained by central place foraging to food resources that can be obtained on shorter, targeted foraging trips while protecting and feeding chicks (Orians & Pearson, 1979;Weseloh et al., 2020), gulls can exploit other perhaps more ephemeral resources that require a 'sit-and-wait' foraging strategy (Yoda et al., 2012).The peak for processing of Atlantic herring in SWNS typically occurs from August to October (P.Pothier, Canadian Food Inspection Agency [CFIA], pers.comm.), thus gulls may have been taking advantage of the peak availability of this food subsidy (although reductions in total allowable catch (TAC) are likely to alter the predictability of this food source in the future; Government of Canada, 2022a).It may also be that effluent (i.e.pieces of discarded fish) from fish plant pipes are not preferred food items for feeding pre-fledging young.
Instead, fish plants may be a reliable but ephemeral food source suitable to increase energetic stores before migratory departure.
Gulls occur in the hundreds and even thousands at local fish plants (Figure S6), and are more often loafing or resting while waiting for potential scavenging opportunities, rather than actively foraging (N. Knutson, pers.obs.).Plant buildings provide ideal loafing locations for socializing and awaiting these and other feeding opportunities.
Our tracked gulls were most likely to occur at fish plants on weekdays from morning through afternoon, generally aligning with plant processing schedules (see Data S1 for further detail), similar to findings around fish plants in Japan (Yoda et al., 2012) Wilhelm et al., 2016).However, implementing such management strategies would be expected to redirect pressures to other food resources, at least in the short term (Bicknell et al., 2013;Langley et al., 2021;Oro et al., 2013).
Reductions to the subsidies upon which gulls currently rely could result in increased dependence on agricultural fields or other human settlements (Gutowsky et al., 2021), with negative consequences such as increased disease transmission to livestock (Coulson et al., 1983).Subsidy reductions could also increase pressures on the flora and fauna of coastal islands from increased deposition of nutrients or contaminants that impact native plant SAR (Geizer et al., 2021;McIntyre et al., 2022;Shochat et al., 2010;Vidal et al., 1998), or from increased displacement or predation of seabird SAR (Bicknell et al., 2013).Alternative or complementary approaches to subsidy reduction or removal at the source may include forms of direct lethal and non-lethal gull management on nesting colonies, including culls or nest destruction.While such approaches are used extensively in some jurisdictions (Coulson, 2015;Donehower et al., 2007), they can be capacity intensive, costly, inefficient, and publicly unpalatable at scale, limiting feasibility and overall impact as long-term strategies (Oro & Martínez-Abraín, 2007;Vidal et al., 1998).No single management method is likely to be successful on its own, and the timing of availability of different subsidies and the timing of use by gulls should be carefully considered when developing subsidy reduction strategies and anticipating management outcomes.For SWNS, reducing the amount of waste discarded from fishing vessels may depress the amount of food available to breeding gulls and force switching to alternative prey during the chick rearing period.In contrast, reducing effluent from pipes at fish plants in SWNS during breeding may not have significant impacts, but reducing food subsidies at fish plants following breeding may decrease both the number of adults concentrating in small areas as well as first-year survival in these local populations.However, implementing such strategies would be expected to increase pressures on other habitats or industries.The mink farming industry also attracts gulls in SWNS; for some colonies, mink farms may contribute up to 20% of herring gull diet in SWNS (Shlepr et al., 2021), but gulls also exhibit broad inter-individual and intercolony variation in mink farm use (Gutowsky et al., 2021).In our to calculate local time of key sunlight events for each day of the study period at Yarmouth tide station.We assigned each gull position to one of six time of day categories: (1) dawn, defined as the period from nautical dawn to the end of the sunrise; (2) morning, until 1.5 h before solar noon; (3) mid-day, until 1.5 h after solar noon; (4) afternoon, until the start of the sunset; (5) dusk, until nautical dusk; and (6) night, until nautical dawn.
number of days tracked during the breeding period varied among birds from 21 to 356 days (mean 194 days ± 114 SD), with five birds tracked over two breeding seasons and three over three breeding seasons.Number of birds and bird-years tracked each week from late April to the end of September varied from six birds and 10 bird-years with >3000 positions at the end of September to 15 birds and 31 bird-years with >18,000 positions in the week of 27 Locations of Whitehead gulls overlapped with 34 other islands within the Lobster Bay area each for ≥10 total hours.The primary island (South Brother Island) was attended by 88% of birds and 85% of bird-years tracked from Whitehead; South Brother Island measures 0.37 ha and 125 m at its widest (FiguresS3 and S4) and is located 5 km from the centre of the Whitehead gull territories and 3 km from the primary fish plant used by gulls from Whitehead.Two percent of all positions occurred on the island, and total time spent varied among bird-years from 0% to 10% of positions.The remaining 33 islands visited by Whitehead gulls ranged in size from 0.02 to 22.8 ha, with <1% of known positions occurring at each.Locations of gulls from Bon Portage Island overlapped 31 islands mostly within the Lobster Bay area each for ≥10 h.The primary island used by Bon Portage gulls, Little Green Island, is 0.23 ha

F
Predicted probabilities of occurrence of herring gulls in southwest Nova Scotia by week with 95% confidence intervals generated from each destination type model, where responses are conditioned on other predictors being held at their means.Estimated breeding phases are shaded in grey.positions per bird-year.The remaining 30 islands overlapped <1% of positions each, and ranged in size from 0.02 to 363 ha.Greater detail on the ecology of visited islands is provided in Data S1.
to the time of day, work week, and tide phase.Gulls primarily visited fish plants during operational hours and low tide, altering their use of these sites throughout the season, and spent time offshore at night, seemingly tied with the daily and seasonal activity of the various local fisheries.Taken together, our results indicate that herring gulls in SWNS have the behavioural flexibility to adapt to both natural rhythms and human schedules while balancing the seasonal F I G U R E 3 Predicted probabilities of occurrence of herring gulls in southwest Nova Scotia by circadian factors (a) time of day, (b) tide phase, and (c) weekday or weekend, and by (d) colony with 95% confidence intervals generated from each destination type model, where responses are conditioned on other predictors being held at their means.demands of reproduction.This ability of gulls to align their behaviour with the routines of human activities has been documented in a number of species living in close proximity to urban areas or active fisheries (e.g.black-tailed gulls (Larus crassirostris), Yoda et al., 2012; lesser black-backed gulls (L.fuscus), Spelt et al., 2021; yellow-legged gull (L.michahellis), Parra-Torres et al., 2020; Audouin's gull (Ichthyaetus audouinii),Ouled-Cheikh et al., 2020).Beyond gulls, temporal matching to the predictable behaviour of humans and availability of food subsidies occurs in other wildlife, including avian and nonavian species (e.g.shearwaters(Bartumeus et al., 2010), storks(Gilbert et al., 2016), bears (Lewis et al., 2015), elephants(Branco et al., 2019)), further highlighting the importance of understanding these relationships for sound management of human-wildlife conflict.Flexible species with generalist diets and opportunistic foraging strategies are likely to optimise foraging decisions in response to shifting availability of natural and human-provided food and in response to shifting dietary needs through the annual cycle.Numerous species of scavenging gulls have been shown to schedule their foraging around the offshore routines of local fisheries that provide predictable feeding opportunities (e.g.Ouled-Cheikh Fisheries closures in Witless Bay, Newfoundland resulted in increased predation pressure on Leach's storm-petrels at one of the most significant colonies in AtlanticCanada, Great Island (Stenhouse & Montevecchi, 1999).The population of Leach's storm-petrels on Bon Portage Island has already declined by approximately 20% in the last 16 years(Pollet & Shutler, 2019), and gulls (herring and great black-backed) account for about 42% of the predation of petrels on the island during the gull breeding period(Hoeg et al., 2021).While our results show that tracked herring gulls visited islands most after gull fledging, and during dusk and night in distinct edge-areas suggesting roosting, reducing the availability of fisheries-provided subsidies which are presently used extensively by gulls during breeding could result in increased gull presence and depredation of storm-petrels or roseate terns, and potentially exacerbate the declines of these vulnerable species.Our findings indicate that tracked herring gulls frequently visited islands after gull fledging, predominantly during dusk and night in specific edge-areas, implying roosting behaviour.Reducing the availability of fisheries-provided subsidies may lead to heightened gull presence on islands and consequently increase probability of roosting gulls predating upon storm-petrels or roseate terns, potentially exacerbating the decline of these vulnerable species.Inhibiting access to fisheries-based resources could also impact other nesting seabird species, documented to nest on at least 20% (12/59) of islands visited by tracked gulls, as well as fall migrants and early arrivals of overwintering bird SAR (particularly harlequin duck and purple sandpiper; see Data S1 for further detail on attributes of visited islands).Further study is required to understand the nature and intensity of interactions between gulls and the flora and fauna of islands in SWNS, and how these may change through the year.

Week Whitehead Bon Portage Total Birds (bird-years) Pos. Birds (bird-years) Pos. Birds (bird-years) Pos.
Overview map depicting location of the study area in southwest Nova Scotia, Canada.(b) Locations of the herring gull study colonies (Whitehead (43.664°N, 65.867° W) and Bon Portage (43.469°N, 65.753° W) islands) and the spatial extent of all islands and fish plants visited by tagged gulls.The Yarmouth tide station is indicated by the white asterisk.
Note: Sample sizes indicate the number of individual birds, bird-years (most birds were tracked over more than 1 year and thus contribute data from multiple years to a given week), and the number of GPS positions (pos.) at 15-min intervals.FI G U R E 1 (a) Description of assignment criteria used to categorise GPS-tracked herring gull positions at 15-min intervals from April-August to seven destination types treated as a binary response variable for GLMMs of probability of occurrence at each destination based on week, time of day, tide phase, weekend or weekday, and colony as categorical fixed effects and individual identity as a random factor.
(Gutowsky et al., 2021)Shlepr et al., 2021) with seven distinct 'destination' types of interest identified in previous research(Gutowsky et al., 2021;Shlepr et al., 2021): breeding territory, breeding colony (but off territory), other islands, other coastal areas, offshore, fish processing plants or mink farms (Table2).A high resolution shapefile of coastal islands of Nova Scotia was obtained from the Nova Scotia Coastal Islands Prioritization Program(Nova Scotia   Coastal Islands Working Group, 2021).Positions were designated as being at an individual's breeding 'colony' if it was within 100 m of the island of device deployment, thus including positions on shore and in the intertidal zone.Positions on the colony were differentiated between those within and those outside of the assumed breeding territory: territory boundaries were approximated for each birdfeeding, or territory defence;Weseloh et al., 2020), and time spent on the colony but away from the territory.Positions were designated as at an 'other island' if it was within 100 m of an island except that of device deployment.Coastal and nearshore positions (herein 'coastal') were those between 500 m inland and 1000 m offshore from the high tide line but not at an island, fish plant or mink farm, where birds are most likely to be influenced by the natural coastal environment.In previous work, we identified the point locations of fish plants and mink farms visited by these tracked gulls(Gutowsky et al., 2021).We plotted these sites with a spatial data layer of point locations of all known fish processing plants in the study region generated for a related project (N.Knutson, unpubl.data)with the complete positional time series on satellite imagery in Google Earth (Google Earth Pro v7.3.2.5776).We manually created polygons to capture the geographic extent of the infrastructure of each cluster of fish plants (encompassing buildings, wharfs and harbours to the waterline, occasionally capturing more than a single fish plant operation if positions spanned neighbouringTA B L E 2Note: The area under the curve (AUC) statistic for each destination indicates the model predictive performance for classifying the binomial response (-indicates insufficient data).dance.Removing sites with limited overlap thus improved our ability to detect temporal patterns in attendance at sites most important for providing food subsidies to gulls.'Offshore'positions were designated as those >1000 m from the high tide line of any landmass.We categorised each position by whether it occurred on a weekday or weekend, the phase of the tide, and the time of day relative to the sun.Hourly mean sea-level data were retrieved from the publicly available Canadian Tides and Water Levels Data Archive (https://www.meds-sdmm.dfo-mpo.gc.ca/isdm-gdsi/twl-mne/index -eng.htm) for the nearest station with continuous monitoring for the period 2019-2021 (Yarmouth, Station Number 365, within 50 km of the study colonies; Figure (Statistics Canada, 2021)Bon Portage and Whitehead islands visited mink farms from 2019 to 2021, which may reflect a response of gulls to a >80% decline in the intensity of mink farming over the past decade(Statistics Canada, 2021).The low occurrence at mink farms during this study suggests that other industries are more important at present when considering strategies and potential management outcomes of reducing anthropogenic food subsidies to herring gulls in SWNS.Coastal SWNS is heavily influenced by the seafood industry, with multiple different fisheries and types of fish processing plants, each with unique seasonal and circadian rhythms of operation.Consequently, interpretation of observed patterns in gull behaviour is challenging.Regardless, our results in light of local knowledge suggest overall that gulls in SWNS adapt to human schedules to access available food resources, and that their flexible, generalist nature enables them to thrive in this region where both anthropogenic and natural resources are accessible, generally abundant, and predictable.Collectively, food subsidies on land and at sea are likely playing a key role in maintaining locally high herring gull numbers in SWNS, despite the high spatiotemporal variability in the availability of individual food sources.Our findings can help guide when and where to test strategies for reducing different anthropogenic food subsidies available to gulls.Given their behavioural flexibility, we stress that mitigation programs must consider the impact on herring gulls as well as subsequent short and long-term potential impacts on local ecosystems.Overall, our findings emphasise the complexity of developing management strategies for human-wildlife conflict 26888319, 2023, 3, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1002/2688-8319.12274,Wiley Online Library on [20/09/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License