Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The foraging ecology of Glausous Gulls preying on the eggs and chicks of Thick-billed Murres Gilchrist, H. Grant 1995

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1995-059642.pdf [ 2.8MB ]
JSON: 831-1.0088327.json
JSON-LD: 831-1.0088327-ld.json
RDF/XML (Pretty): 831-1.0088327-rdf.xml
RDF/JSON: 831-1.0088327-rdf.json
Turtle: 831-1.0088327-turtle.txt
N-Triples: 831-1.0088327-rdf-ntriples.txt
Original Record: 831-1.0088327-source.json
Full Text

Full Text

THE FORAGING ECOLOGY OF GLAUCOUS GULLS PREYING ONTHE EGGS AND CHICKS OF THICK-BILLED MURRESbyH. GRANT GILCHRISTHonors B.Sc., Trent University, 1990A THESIS SUBMHthD IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF ZOOLOGYWe accept this thesis as conformingto the required standardNIVERSITY OF BRITISH COLUMBIASeptember 1995© H. Grant Gilchrist, 1995In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of________________The University of British ColumbiaVancouver, CanadaDate °DE-6 (2/88)ABSTRACTThe glaucous gull (Larus hyperboreus) is a generalist predator with a circumpolar distribution. Itcommonly depredates the eggs and chicks of birds nesting in the Arctic, and often breeds inassociation with colonial nesting waterfowl and seabirds. The aim of this thesis was to examinehow environmental factors constrained the ability of glaucous gulls to depredate cliff-nestingthick-billed murres (Uria lomvia), and to determine whether these constraints vary through time,in space, or between individuals. I also wanted to determine how changes in gull foragingconstraints could affect the impact of gulls on murre reproductive success and populationdynamics. This study was conducted from 1990-1992 at a murre colony on Coats Island,Canada, and in 1993 at colonies in the Upernavik region of Greenland.Based upon results from egg placement experiments and field observations of predation,gull foraging success was constrained by high murre nesting densities, collective murre defence,and the accessibility of narrow cliff ledges. However, windy conditions enhanced the ability ofgulls to overcome these constraints. Wind improved the aerial maneuverability of gulls, andenabled gulls to reach weakly-defended narrow ledges and avoid contact with murres duringattack. Murre defence on narrow cliff ledges was less effective because murres had difficultyturning to face attacking gulls without dislodging their own eggs and chicks. Gull searchactivity, attack activity, and predation rates were strongly correlated with windy conditions.Consequently, the impact of gull predation on murre reproductive failure, which ranged from 0%to 21% depending on nest type, was determined largely by wind.Under calm wind conditions, adult gulls were often inactive and they rarely fed theirchicks. This inactivity may reflect the reluctance of gulls to forage on foot, because although thiswas the most successful attack mode, it also incurred the greatest contact with defending murres.Alternatively, gulls could have been responding to the varying energetic demands determined bychanging weather conditions, so that the danger of injury while foraging did not influence modeselection or periods of foraging activity. I explored these two alternatives using a dynamicIIsimulation model which integrated field data and energetic estimates of gull foraging behaviour.The model suggested that the first explanation, which was based upon energy considerationsalone, was not sufficient to explain gull foraging inactivity under calm wind conditions.However, the model supported the idea that gulls were sensitive to risk of injury, and that theyshould select low-danger foraging modes that provide low energetic gains. The model alsorevealed that gulls can afford to use foraging modes that yield low energetic gains relative tomore productive and dangerous ones (e.g. scavenging vs. foraging on foot under calmconditions), because even poor foraging modes are sufficient to meet their energetic demandsunder most circumstances.I predicted that a decline in the density of nesting murres should enhance the ability of gullsto overcome the constraints of calm wind conditions, cliff ledge accessibility, and prey defence.Thus, murre colony declines should increase the impact of gull predation on murre reproduction.To examine the impact of gull predation at declining thick-billed murre colonies, I compared gullforaging mode selection, predation rates, and murre nest site selection at Coats Island, N.W.T.,with that at declining murre colonies found near Upernavik Greenland. I found that gulls atdeclining colonies foraged on broad cliff ledges and were less constrained by calm windconditions, apparently because population declines increased the availability of low nestingdensity ledges where gulls could maneuver on foot and attack murres with little risk of injury.Perhaps because of this, predation rates at declining murre colonies were consistently higher thanat the stable Coats colony, particularly at low wind speeds.111TABLE OF CONTENTSAbstract iiTable of Contents ivList of Tables viiList of Figures viiiAcknowledgments xChapter 1: Introduction 1Predator foraging behaviour 1Study sites 3Ecology of the glaucous gull 4Glaucous gull ecology at Coats Island 5Aims of this thesis 9Chapter 2: Effects of murre nest density, cliff ledge width,and wind on predation by gulls 10Introduction 10Methods 11Study area and species 12Egg placement experiments 12Behavioural observations 13Statistical analysis 14Results 16Nesting density 16Ledge width 16Wind speed 20Vulnerability of eggs relative to timing of laying 22Discussion 26Constraints on gull foraging efficiency 26Consequences of predation for thick-billed murres 28Chapter 3: Wind and prey nest sites as foraging constraintson an avian predator, the glaucous gull 30Introduction 30Methods 31Study area and species 31Behavioural observations 32Gull gliding-flight dynamics 33Effects of gull predation on murre reproductive success 33Statistical analysis 34ivResults 36Murre colony structure and environment 36Gull search activity 36Gull attack rates 41Gull attack selectivity relative to murre nest site characteristics 41Gull predation rates 41Gull attack success 46Wind and gull maneuverability in flight 51Murre responses to gull attack 51Path diagram 53Colony-level effects of gull predation on seasonal murre reproduction 57Discussion 59Wind: a foraging constraint for glaucous gulls depredating murres 59The currency of glaucous gull foraging decisions 61Population level consequences of gull foraging behaviour for murres 63Conclusions 64Chapter 4: Foraging mode selection of glaucous gulls provisioningyoung: consequences of a danger-reward trade-offmediated by wind? 65Introduction 65Methods 67Study site and species interactions 67Field studies of gull foraging behaviour 67Dynamic optimization model 68Brood energy dynamics 69Model assumptions concerning provisioning 74Weather conditions 76Metabolic rate estimates 76Limits of chick fat stores 80Flight energetics 82Energy consumption 83Foraging mode energetic s and danger of contact with murres 84Modeling risk of fatal injury 89Results and Discussion 90Foraging mode selection with no risk of injury 90Foraging mode selection with risk of injury 90Comparisons with field data of gull foraging behaviour 94Further predictions 101Conclusions 102Chapter 5: Effects of glaucous gull predation at decliningthick-billed murre colonies: An inter-colony comparison 104Introduction 104Methods 105Study sites 105Gull foraging behaviour and predation rates 107Numerical response of glaucous gulls to murre colony declines 107VMurre nest site selection and breeding density 108Results 109Gull foraging mode selection, attack activity, and predation rates 109Numerical response of glaucous gulls to murre colony declines 112Murre nest site selection 115Discussion 117Gull foraging behaviour and predation rates 117Numerical response of glaucous gulls to murre colony declines 118Murre nest site distribution at declining colonies: a consequenceof gull predation? 119Population-level effects of gull foraging behaviour: multiple-stable states? 121Conclusions 123Appendix I 124Chapter 6: Conclusions 125Thesis synthesis 125Comparisons with other studies of gull foraging ecology 126From Larus to leo: identifying future research directions 132Literature cited 134viLIST OF TABLESpageTable 2.1 Description of attack modes of gulls and murre responses to attack. 15Table 2.2 Murre responses to gull attacks in relation to cliff ledge width. 19Table 2.3 Gull attack techniques in relation to cliff ledge width. 21Table 2.4 Murre response to gull attacks in relation to timing of murre egg laying. 25Table 3.1 General linear models of factors affecting rates of aerial search by gulls. 40Table 3.2 General linear models of factors affecting gull attack activity. 42Table 3.3 Analysis of covariance of factors affecting rates of gull attack inrelation to murre nest site characteristics. 43Table 3.4 General linear model of factors affecting gull predation rates. 47Table 3.5 Multiple logistic regression of factors affecting gull attack success. 49Table 3.6 a) General linear model of factors affecting murre response to gullattack, and b) multiple logistic regression of factors affecting theprobability that a gull was struck by murres during attack. 54Table 4.1 Probability of wind conditions in the future in relation to the windconditions in the current time interval 72Table 4.2 Estimates of Resting Metabolic Rates for a glaucous gull adult and chickin relation to weather conditions 81Table 4.3 Proportion of gull foraging time devoted to flight in relation to wind 85Table 4.4 Parameter estimates of glaucous gull foraging mode energetics and riskof injury in relation to wind conditions 87Table 6.1 A review of studies examining the foraging ecology of gulls 127viiLIST OF FIGURESpageFigure 2.1 Survival of exposed murre eggs in relation to thick-billed murrenesting densities. 17Figure 2.2 Survival of exposed murre eggs in relation to cliff ledge width. 18Figure 2.3 Survival of exposed murre eggs in relation to wind speed. 23Figure 2.4 Survival of exposed murre eggs in relation to timing of murre egg laying. 24Figure 3.1 Wind conditions in relation to date at Coats Island. 37Figure 3.2 Examples of glaucous gull aerial search activity in relation to windand time of day. 38Figure 3.3 Attack activity of glaucous gulls in relation to wind and murre nest sitecharacteristics. 44Figure 3.4 Glaucous gull predation rates in relation to date and year. 45Figure 3.5 Predation rates of murre eggs and chicks in relation to wind, murrenest site characteristics, and year. 48Figure 3.6 Glaucous gull attack success in relation to wind speed and year. 50Figure 3.7 a) Glaucous gull foraging patrol duration over murre breeding areas inrelation to wind, and b) gull attack-hover duration in relation to wind. 52Figure 3.8 Path analysis of the interactions between factors affecting gull predationrates and risk of contact with murres during attack. 55Figure 3.9 Proportion of murre eggs and chicks lost to gull predation in relation tomurre nest type 58Figure 4.1 Survival function of gull brood in relation to energy stores. 70Figure 4.2 Dynamic model predictions of gull foraging mode selection under calmconditions in relation to time, brood energy stores, and risk of injury. 91Figure 4.3 Dynamic model predictions of gull foraging mode selection under windyconditions in relation to time, brood energy stores, and risk of injury. 92Figure 4.4 Individual variation in attack mode selection in relation to wind conditions. 96Figure 4.5 Time budgets of aerial foraging specialists, and pedal foraging specialistsin relation to wind conditions. 99viiiFigure 5.1 Wind conditions at Coats Island and Kingittoq murre colonies. 106Figure 5.2 Attack mode selection of glaucous gulls in relation to wind conditionsand colony identity. 110Figure 5.3 Glaucous gull attack activity in relation to cliff ledge width, wind speedand colony identity. 111Figure 5.4 Gull predation rates in relation to wind speed and colony identity. 113Figure 5.5 The relationship between gull numbers and murre colony size 114Figure 5.6 Murre nesting characteristics in relation to the magnitude of declines atthree Arctic murre colonies. 116ixACKNOWLEDGMENTSThis study benefited greatly from the support and input of many people, and Iwould like to take this opportunity to thank them. Most notably, I thank my supervisorsJamie Smith and Tony Gaston who generously provided support and encouragementthroughout this project, and who contributed many ideas during the field work andwrite-up stages. Both Jamie and Tony were great supervisors, and I feel very fortunateto have studied with them. They struck a balance of involvement with this researchwhile at the same time allowing me to make key decisions relating to the work (for betteror worse). They also tempered my overly optimistic approach to field work withquestions like, “ what will you do if that doesn’t work?” at opportune times (i.e.prior to plane departure). I would also like to thank Erica Nol for introducing me toboth Jamie and Tony. I also greatly appreciate the input of my other research committeemembers Tony Sinclair, Ron Ydenberg, and David Jones, and also Carl Walters. Theyencouraged me to integrate a sound understanding of an organism’s Natural Historywith theoretical ideas. I particularly thank Ron Ydenberg for our many conversationsand his substantial contributions to chapter 4.At the University of British Columbia, I was also fortunate to be surrounded by adynamic and friendly group in the “Smith Lab” consisting of Lance Barret-Lenard,Fiona Schmiegelow, Linda Dupuis, Dave Westcott, Wesley Hochachka, David Ward,Arnon Lotem, Alice Cassidy, Victoria Campbell, and Christine Adkins. I benefitedespecially from my discussions with David Ward, Arnon, and Wesley. Thanks also toChristine Adkins for helping me translate gull behaviour into Quick Basic.While on Coats Island, I was fortunate to share my summers and bear encounterswith Leah deForest, Garry Donaldson, David Andrews, David Noble, Don Crol, MarkHiphner, Marco Passeri, Paul Prior, and Thomas Alogut. The experiences we sharedon Coats have been the basis for many long-lasting friendships. I especially thank “thegull guys”, Thomas Alogut, Marco Passeri and Paul Prior for their assistance andtireless commitment to this work. While in the Canadian arctic, I also thank BobLongworth and Lyne Paplinski of the Iqaluit Research Center for their logistical supportand Carribean weather reports over the radio. For the research conducted in Greenland,I thank the Greenland Home Rule Government, Henning Thing, and the community andMayor of Upernavik for their assistance and permission to work in the Upernavikregion. I would also like to thank Kaj Kampp, Jamie Smith, Tony Gaston, DavidNettleship, Vernon Byrd, David Nysewander, William Sydeman, and Daniel Roby forhelping me reach Greenland. I also appreciate the assistance and companionship ofGabanguak Bidstrup, Thomas Alogut, and Tara Gilchrist while in the field.This project was funded by the Canada Life Assurance Co., World Wildlife FundCanada, the American National Fish and Wildlife Foundation, the John K. CooperMemorial Trust, the Erickson Memorial Scholarship, the Hull Group, Sierra DesignsCanada, Mountain Equipment Co-op, Sigma Xi Research Support Fund, the CanadianWildlife Service (Student Research Support Program to H.G.G. and operating grants toAnthony J. Gaston), Natural Sciences and Engineering Research Council of Canada(grant to James N. M. Smith and student support for H.G.G.), the University of BritishColumbia Graduate Fellowship, the Northern Studies Trust Program, and the ScienceInstitute of the Northwest Territories.Finally, I would like to acknowledge the support of my family, Tara Gilchrist,Nancy and Brian, Susan and Cohn, and my parents Hugh and Myirea Gilchrist whohelped me in so many ways. I am especially grateful to my parents for encouraging meto pursue my interest in nature and for not pushing the more practical “architecturecareer path” too hard. Finally, I thank my wife Tara for her tremendous support andpatience, particularly during our times apart.x1CHAPTER 1INTRODUCTIONWhen I was young, I used to lie in the grass on summer afternoons to watch the aerialacrobatics of a kingbird (Tyrannus tyrannus) catching insects. I was amazed at how easily itcould fly up from its perch to snatch insects out of the air. On one of these days, a dense fogrolled over the countryside and the kingbird rarely moved; its feathers fluffed out against thedampness. As I was about to leave, I noticed a Cooper’s hawk (Accipiter cooperii) suddenlyappear out of the fog in a grey blur as it dived towards the kingbird on the fence. The kingbirdclimbed vertically into the air to avoid the attack, and the hawk followed effortlessly. At thepeak of its climb the kingbird dived straight back to the ground and at the last second, pulled upto glide neatly through the wire fence with the hawk in close pursuit. The hawk rocketedthrough the fence in a puff of feathers, and tumbled lifelessly into the grass on the other side;one of its wings shorn off by the wire. The kingbird re-appeared out of the grass and returnedsilently to its perch in the fog.Early observations like this convinced me that foraging and anti-predator behaviour arecritical components of an animal’s life history, and also important influences on the interactionsbetween species in the wild. Even predators like the Cooper’s hawk face constraints and riskswhen foraging, and in the most extreme cases, foraging decisions can cause crippling injury oreven death (as above). As I continued to study biology, I often found it useful to examineforaging behaviour from an economic perspective (Stephens and Krebs 1986). Early foragingstudies used the premise that animals should select foraging strategies which maximize their netenergetic gain. However, the simplicity of these early models obscured critical factors thataffect foraging decisions in the wild (learning, changing physiological states, predation risk,risk of injury during attack; Real and Caraco 1986; Lima and Dill 1986; Mangel and Clark1986). More recent foraging studies have attempted to identify factors that constrain theabilities of animals to maximize net energetic gain. Rather than asking, “do animals behave2optimally?’, foraging theory now considers why animals often select foraging strategies that donot yield the highest energetic gain (Dill 1986; Stephens 1990; Ward 1990).For example, foragers often appear to minimize mortality risks while foraging, so that atrade-off exists between immediate energetic gain and risks of mortality (Lima and Dill 1986).For predators like the Cooper’s hawk, foraging can be dangerous if prey fight back(physically or behaviourally), or if larger predators prey on them while they hunt (Curio 1974).In the first case, the value of attacking a prey should decrease as the danger of injury duringattack increases, because an animal should rarely jeopardize its current and future reproductionfor a small immediate increment in fitness (Clark 1994). The relative costs and benefits ofalternative foraging decisions should also vary with: 1) the availability of prey, 2) the dangersassociated with subduing prey, 3) environmental conditions, and 4) the energetic state of thepredator. Consequently, the impacts of predators on prey should also be dynamic.Only a few field-studies of vertebrate predator-prey interactions have quantified howchanges in the foraging constraints of predators influence the behaviour and populationdynamics of predators and prey (Werner et al. 1981; Spear 1993; Young 1993; Goss-CustardCt al. 1995). Young (1993) identified several environmental factors which determined theimpact of south polar skuas (Catharacta niaccorinicki) on the reproductive success of Adeliepenguins (Pygoscelis adeliae). The ability of skuas to take penguin chicks increased in windyconditions because wind increased their ability to avoid the defensive attacks of adult penguins.Late in the breeding season, penguin chicks also grew too large for skuas to kill, and thisplaced skuas in an energetic bottleneck at a critical time in their own chicks development. Inaddition, heavy ice conditions at sea affected the availability of alternative food sources and thisaffected the value of foraging at the penguin colony for skuas, and the risks skuas were willingto take when attacking penguins. These examples illustrate how seemingly unimportantenvironmental factors can interact to influence predator foraging behaviour, and consequently,the impact of predators on prey populations.3Interactions between changing environmental conditions and predator foragingconstraints like those described above, may affect the ability of glaucous gulls (Larushyperboreus) to prey on the eggs and chicks of colonial nesting thick-billed murres (Urialomvia). These topics are examined further in the chapters that follow using egg placementexperiments, behavioural observations, and computer simulations.Study sitesI studied the predator-prey interaction between glaucous gulls and thick-billed murresbreeding on Coats Island, N.W.T. Canada (1990-1992), and in the Upernavik region of northwest Greenland (1993). The Coats Island murre colony has expanded since the 1970?s,whereas the Upernavik murre colonies have experienced severe population declines due to acommercial hunt conducted by the local community between 1970 and 1985 (Kampp et al.1994).The stability of the cliff at the Coats colony permitted safe access with climbingequipment and un-obstructed views of the birds. In addition, 3 factors permitted me to studyan avian predator-prey system in unusual detail: 1) the numbers of birds present at the colony(ca. 32000 breeding pairs of murres, 13-16 breeding pairs of glaucous gulls), 2) my closeproximity to the birds allowed me to study many predator-prey interactions at close range andin a degree of detail that is rarely possible in the wild, and 3) several of the physical andenvironmental factors that affected these interactions were easily quantified (e.g. cliff ledgestructure, wind speed).A further benefit of the field site was that other recent research on the reproductiveecology of the thick-billed murre at Coats Island (Noble 1990; deForest 1993), provided mewith insights into this predator-prey system that I could not have achieved alone.4Ecology of the glaucous gullDISTRIBUTION - The glaucous gull is the second largest of all gull species. It is ageneralist predator with a circumpolar distribution. In North America, it is widely distributedin arctic regions and breeds along coasts from northern Alaska to Baffin Island, and Labrador.In the Atlantic, it is replaced by the greater black-backed gull (Larus marinus) below Labrador.In the Pacific, it is replaced by the glaucous-winged gull (Larus glaucascens) in the AleutianIslands and in British Columbia. Glaucous gulls from the eastern Arctic winter in the Maritimeregions of Canada, southern Greenland, and rarely on the Great Lakes. Gulls that breed in thewestern Arctic and Alaska winter along the west coast of North America as far south asnorthern California.NESTING BEHAVIOUR - Glaucous gulls use a variety of nesting habitats. Like manyother Arctic bird species, glaucous gulls select nest sites that are inaccessible to Arctic foxes(Alopex lagopus). They may nest on coastal islands and cliffs (Manning et al. 1956; Gaston etal. 1985), on islands in freshwater lakes near coasts (Martin and Moitoret 1986), and on sandyislets at river mouths (Sage 1974). In the Canadian and Greenland Arctic, glaucous gulls alsonest on inaccessible cliffs in association with colonies of black-legged kittiwakes (Rissabrevirostris), thick-billed murres (Uria lomvia), and Iceland gulls (Larus glaucoides).The density of breeding glaucous gulls ranges from single nests spaced kilometers apartover flat tundra, to colonies of more than 100 breeding pairs on small islands or cliffs (Johnsonand Herter, 1989). The selection and distribution of glaucous gull nests likely reflects aninteraction between limited nest sites that are inaccessible to foxes, and of the availability offoraging opportunities needed for successful reproduction.For example, at Digges Island, Quebec, Canada, all glaucous gulls nest on a cliff that isinaccessible to foxes (Gaston et al. 1985). Some gull nests are scattered within a thick-billedmurre colony, and the owners of these nests forage primarily on murre eggs and chicks(Gaston et al., 1985). In contrast, the gulls nesting at a colony of approximately 40 pairs nearthe upper margin of the murre colony apparently feed at distant shorelines and on the open5ocean. These gulls do not maintain feeding areas around their nests, and consequently, theirnests are spaced only meters apart within the gull colony (Gaston et al., 1985). A similardichotomy (i.e. where seabird-feeding specialists maintain territories and generalists nestcolonially), apparently occurs among glaucous gulls at most other seabird colonies in theeastern Canadian and Greenland Arctic (Gaston and Nettleship, pers corn., and pers. obs.),and among some other gull species (Spear 1994; Watanuki 1993; Nettleship 1972).DIET DURING THE BREEDING SEASON - Glaucous gulls are omnivores, although animalmaterial predominates in the diet of most individuals. Major categories of food during thebreeding season include: 1) marine invertebrates from the intertidal zone, 2) bird eggs andchicks, especially of colonial-nesting waterfowl and seabird species, 3) invertebrates and fishfrom rivers, 4) carrion, especially items washed up on shore or present on the sea ice, 4) smallmammals, especially lemmings where they occur, 5) human refuse at community garbagedumps or fish plants, and 6) food items pirated from other foraging birds (Ingolfsson, 1976;Gaston and Nettleship 1981; Strang 1982; Johnson and Herter 1989; Barry and Barry 1990).One of these food categories may predominate in the diet of some individuals. For example,some glaucous gulls nest within seabird or waterfowl colonies and prey primarily on the eggsand chicks of these other species (Gaston and Nettleship 1981). However, it appears that mostglaucous gulls in the Arctic have a more varied diet, and that these specialized gulls form asmall portion of the entire breeding population (Barry and Barry 1990).Glaucous gull ecology at Coats IslandNESTING BEHAVIOUR AN]) REPRODUCTIVE TIIvIING - Glaucous gulls arrive at CoatsIsland during the first and second weeks of May, which is several weeks prior to the arrival ofthick-billed murres at the colony. At this time, land-fast ice typically extends for over 50 kmfrom shore and the tundra of the island is largely snow-covered. Snow squalls occurfrequently in May and well into early June. Gulls construct their nests out of plant materialwhich they collect from ridges that have been blown free of snow. At Coats Island, gull nests6are constructed on cliff ledges and are spaced widely apart (>35 meters between most nests).This nesting distribution reflects the fact that breeding pairs defend feeding territories within themurre colony, and prevent other gulls from establishing nests in these territories by directattacks. Banding of gulls has shown that breeding pairs at Coats Island are philopatric to boththeir mates and to specific nesting ledges between years. This nesting distribution is unlike thatobserved in most other gull species, which typically nest colonially and commute to distantfood sources (e.g. Siegel-Causey and Hunt 1981; Sibly and McCleery 1983; Pugesek andWood 1992; Spear 1993). The number of glaucous gulls that have bred at the Coats Islandmurre colony since 1986, when researchers began to visit the colony regularly, has rangedfrom 12 to 16 pairs. A further 3 to 5 non-breeding glaucous gulls are usually present.Thick-billed murres begin to arrive at the cliff sporadically and briefly in early June. Theearly attendance patterns of murres appears to be determined largely by weather and thedistances they must travel to open water. During bad weather, murres often abandon thecolony and return to sea until weather conditions improve. In rare years where open wateroccurs close to shore, murres will return to the colony one to two weeks earlier in the year.The peak of murre egg laying typically falls between the 24th and 26th of June, whichcoincides with the hatching of most glaucous gull chicks. One factor that influences glaucousgull reproductive success is the weather that coincides with hatching period of gull chicks.Hatching chicks may be susceptible to exposure during wet and windy weather, becausebrooding adults sometimes stand up off their eggs when hatching is taking place. Once chickshave hatched, however, they are brooded more consistently, and are less vulnerable to theeffects of wind and rain.During late June, July and August, glaucous gull chicks remain on cliff nesting ledges.If chicks accidentally fall to lower ledges during this time, they are either killed outright bybreeding murres, or are forced by murres to leap to the ocean where they eventually die.Gull chicks fledge during the third and fourth weeks of August, so that most gull chickshave left the cliff prior to the peak of murre chick departures. Gull chicks remain in the area7around the colony and are sometimes attended by their parents at this time. Thus, breedingglaucous gulls spend approximately four months of the year at the Coats Island colony.Although very little is known about their migration pathways, it is likely that the glaucous gullsof Coats Island spend the winter in southern Greenland or Newfoundland, Canada.FORAGING OPPORTUNII1ES DURING BREEDING - When gulls first arrive at Coats islandin early May, foraging opportunities may be limited. They have poor access to the open ocean,and murres are absent from the colony. Examination of samples of regurgitated pellets (n=32)during May and early June revealed that gulls scavenge seal species (F. phocidae) and walrus(Odobenus rosmarus) carcasses on the sea ice.Coats Island supports a resident caribou herd (Rangfer tarundus) which experiencesmass die-offs on a five to eight year cycle. The carcasses from these winter die-offs, however,are mostly consumed by Arctic foxes prior to the arrival of breeding glaucous gulls. Anexception to this occurred in 1991, when two caribou fell to their deaths at the colony and werecovered by drifting snow and glaucous gulls scavenged these carcasses for several weeks inMay and June.In June, Arctic foxes provide gulls with scavenging opportunities on the sea ice belowthe colony. Murres often fall to the ice during intra-specific fights and when departing from thecolony in calm winds. Murres are rarely killed by the fall, but remain on the ice and are oftenunable to take flight. During calm wind conditions, it was common to see 10 to 20 murresstranded on the ice below the colony. Arctic foxes often killed these murres in quicksuccession, so that several carcasses remained to be scavenged by glaucous gulls. Followingthe break-up of the sea ice in late June, murres are no longer accessible to foxes and this foodsource no longer exists for gulls.Once murres begin to lay in late June, glaucous gulls concentrate much of their foragingeffort on stealing murre eggs. Observations of predation events, of food items provided tochicks, and the analysis of pellets regurgitated on nesting ledges (GG, unpublished data),8indicated that murre eggs and chicks make up >85% of gull diet in July and August. This isthe period of the breeding season that is the focus of this thesis.Murre eggs and chicks are a highly profitable food for glaucous gulls for the followingreasons. First, murre colonies provide gulls with a dependable and predictable energy sourcefor 2 and one half months during the breeding season. Second, adult gulls can almost forageand guard their chicks simultaneously, because they can hunt within feeding territories neartheir nest. As a result, glaucous gull chicks are sometimes left alone at the nest at a very earlyage while both parents forage in the vicinity. This contrasts with the situation in most colonial-nesting gull species, where one member of a pair must be present at the nest to defend theirchicks against cannibalistic conspecifics (Pierotti and Annett 1991 a). A further benefit to gullsbreeding in association with murres is that glaucous gulls can kill and easily ingest murrechicks throughout the breeding season. In some other avian predator-prey systems, prey cangrow too large for the predator to kill them (e.g. skuas and adelie penguin chicks, Young 1994;Glaucous gulls preying on snow goose goslings, Keith Hobson, C.W.S. personalcommunication.).In summary, the diet of breeding glaucous gulls at Coats island is apparently less diversethan the diets of glaucous gulls nesting elsewhere in the Arctic, and also less diverse than thediets of most other large gull species nesting in temperature regions. Glaucous gulls in thewestern Arctic have a diverse diet which can include rodents, intertidal organisms, scavengedcarcasses, mollusks, fish, and waterfowl eggs (see references above). Gulls in more temperateregions have a similarly diverse diet and may supplement natural prey with human refuseduring the breeding season (Siegel-Causey and Hunt 1981; Sibly and McCleery 1985; Pierottiand Annett 1986; Spear 1993). Coats Island supports no rodent species or Inuit communities.The shores near the murre colony at Coats Island also support few intertidal organisms,apparently due to scouring by sea ice. As a result, glaucous gulls that breed at the Coats Islandmurre colony feed primarily on murre eggs and chicks, fish at sea, and rarely on scavengedcarcasses. Thus, the foraging opportunities for breeding gulls at Coats Island are largely9determined by weather and ice conditions around the colony, and the reproductive timing ofmurres.Aims of this thesisMy objectives were to examine the foraging behaviour of glaucous gulls preying on theeggs and chicks of thick-billed murres, and to determine their impacts on murre reproductivesuccess. Specifically, I studied how environmental factors constrain the ability of glaucousgulls to attack murres, and if these constraints varied through time, in space, or betweenindividuals. I also considered the impact of gulls on the reproductive success and populationdynamics of murres.In chapter 2, I describe an experiment in which I placed murre eggs in the colony andmonitored their fate in relation to cliff ledge characteristics, murre nesting density, gull attackmode, and wind conditions. In chapter 3, I examine the hypothesis that gull foraging activityand predation rates are positively correlated with windy conditions. In chapter 4, I present adetailed energy budget of gull foraging activities and integrate this with field data of gullforaging behaviour using a dynamic model. I use this model to test the hypothesis that dangerof injury during attack is the main factor influencing the selection of foraging mode by gulls.Chapters 3 and 4 generated the prediction that decreased murre nesting densities decrease therisk of injury for gulls, thereby enhancing their foraging efficiency. Consequently, the impactof glaucous gulls on the reproductive success of thick-billed murres should increase followingany perturbation that decreases murre nesting densities. In chapter 5, I test a prediction of thishypothesis, that glaucous gull predation should be greater at a heavily-harvested and decliningmurre colony (Upernavik in northwest Greenland), than at Coats Island, N.W.T., Canada. Inchapter 6, I briefly summarize the thesis and identify the need for further research on mortalityrisks taken by top predators.10CHAPTER 2EFFECTS OF MURRE NEST DENSITY, CLIFF LEDGE WIDTH, AND WIND ONPREDATION BY GULLSPredation strongly influences the reproductive behaviour of colonial-breeding birds(Wittenberger and Hunt 1985; Burger and Gochfeld, 1994). Nesting in groups may providebenefits through increased vigilance, predator swamping, and group defence (Burger andGochfeld, 1994). Among seabirds, the effectiveness of predator mobbing may increase withgroup size and nesting density as neighbours defend nest sites collectively (Birkhead 1977;Siegel-Causey and Hunt 1981; Spear and Andersson 1989; Spear 1993). It may also vary overthe course of the breeding season, being most effective when most birds in a group have young(Andersson et al. 1980). Colony structure and topography may also influence predatorforaging efficiency because nest site characteristics, such as burrow depth, cliff ledge width,and ledge slope vary within seabird colonies and can affect the accessibility of nest sites toavian predators (Nettleship 1972), and the ability of prey to defend themselves (Siegel-Causeyand Hunt 1981; Young 1994). Therefore, the impact of avian predation is a function of howpredator foraging constraints vary in space and time, and how predators overcome theseconstraints.Glaucous gulls, Larus hyperboreus, are generalist predators that often breed inassociation with waterfowl or seabird colonies (Portenko 1989; Johnson and Herter 1989;Barry and Barry 1989). They commonly prey on thick-billed murre, Uria lomvia, eggs andchicks at colonies in the Arctic (Gaston and Nettleship 1981). Thick-billed murres breed indense colonies on exposed cliff ledges, and the principle habitat features affecting theirreproduction include the slope and width of cliff ledges, the numbers of adjacent rock walls,and the number of breeding neighbours (Gaston and Nettleship 1981). Murres that nest on theinterior of dense groups on broad cliff ledges experience the highest reproductive success11(Gaston and Nettleship 1981; Birkhead et al. 1985; Birkhead and Nettleship 1986; deForest1993), perhaps because they are most successful in avoiding gull predation.In this chapter, I identify some of the foraging constraints of glaucous gulls and examinewhether they vary spatially or temporally. I also explore how the dynamics of gull foragingconstraints affect the vulnerability of murre nest types. I placed eggs experimentally in themurre colony and monitored their fate in relation to cliff ledge width, murre nesting density,timing of murre egg-laying, murre defence, gull foraging mode, and wind conditions. Basedupon previous observations of gull foraging, I predicted that: 1) exposed eggs placed in highdensity nesting areas would survive longer than those in low density areas due to collectivedefence by murres, 2) that eggs placed on narrow ledges would survive significantly longerthan those on broad ledges because the large body size of glaucous gulls, 1.8-2.1 kg, makes itdifficult for them to forage on narrow ledges, and 3) that eggs placed early in the breedingseason would survive for a shorter period than eggs placed following the peak in murre laying.METHODSStudy Area and SpeciesThe study was conducted at a thick-billed murre colony on Coats Island, NorthwestTerritories, Canada (62°30’N, 83°OOW) in 1990, 1991, and 1992. Thick-billed murres bredon a vertical cliff up to 65 meters above the sea. Since 1984, 1500-2500 murre chicks havebeen banded each year with metal and colour bands, establishing a sample of birds of knownage. Glaucous gulls nested on ledges within the murre colony or occasionally on turfimmediately above the murres. Glaucous gulls were the primary predators at the Coats Islandcolony, and murre eggs and chicks made up >85% of gull diet for both adults and chicks.Prior to the onset of laying by murres, gulls fed on the carcasses of murre, ringed seal, Phocahispida, and caribou, Rangifer tarundus, present on the sea ice below the colony cliffs.12Egg placement experimentsIn 1990, 1991, and 1992 I placed large chicken eggs painted to mimic murre eggs in themurre colony in several sites and monitored their survival in relation to time of day, gull attacktechnique, murre defence, murre reproductive synchrony, and weather conditions. Inpreliminary studies, experimental eggs were taken readily by gulls and occasionally incubatedby murres. In 1991, I also used genuine murre eggs taken as part of a study examining thick-billed murre reproductive success (deForest 1993).In all studies, eggs were placed by climbing down into the colony from above using fixedropes. Murres often left nesting ledges in response to the disturbance caused during eggplacement, but usually returned within 5 minutes. Climbing equipment allowed me to moveslowly and methodically within the colony, and I was able to avoid significant eggdislodgment. Further, I remained above study plots and deterred gulls from taking exposedeggs until murres had returned to their nest sites. I began to record the survival time ofexperimental eggs after I left the cliff face and entered an observation blind. The survival ofexperimental eggs was monitored continuously for the first four hours after placement, andchecked hourly thereafter. Eggs that survived beyond twenty-four hours, were then checkedevery three hours. I terminated the experiment after 72 hours, and assumed that surviving eggswere either invisible or inaccessible to gulls.In 1990 and 1991, I tested the effects of nesting density and ledge width on the predationof experimental murre eggs. Eggs were placed in four situations: 1) broad ledges with highmurre nesting density; 2) broad ledges with low nesting density; 3) narrow ledges with highdensity; and 4) narrow ledges with low density. High density ledges had >80% of their areaoccupied by breeding murres, whereas low density ledges had <40% of their area occupied.The nesting density of ledges was assessed by eye using a 15-45 power spotting scope andwith the aid of photographs of nesting ledges. Broad ledges supported more than one row ofbreeding sites, whereas narrow ledges supported only one row.13In 1991, I tested the effects of wind speed on the accessibility of narrow ledges to gulls.I conducted the experiment following the peak of murre laying at which time >80% of murreshad eggs. Two eggs were placed on each of fourteen narrow, low-density murre nest sites.One egg was placed under windy conditions (>15 km/h), while the other was placed on thesame site under calm conditions (<5 km/h). I then compared the survival times of the eggsunder the two regimes. Wind speed categories were based upon a 15 km/h wind thresholdabove which glaucous gulls could glide without flapping their wings. This threshold wasdetermined from behavioural observations in the field and from data on flight dynamics of largegulls (Pennycuick 1987). Wind speed was measured to ± 1 km/h using a Weather-pro Digitaranemometer mounted 1.5 m from the cliff face. If the wind conditions changed during the firstfour hours of the experiment (e.g., from calm to windy), the trial was abandoned. I predictedthat eggs placed under calm conditions would survive longer than those placed under windyconditions due to the increased maneuverability of gulls in strong winds.In 1992, I used thick-billed murre eggs to test the effect of laying synchrony on eggpredation. For these trials, eggs were placed only when wind speeds were >15 km/hr. Iplaced eggs on broad (N=6) and narrow (N=6) ledges about a week before the peak of egglaying (< 17% of pairs with eggs). These placements were repeated on the same sites 12 dayslater at which time >82% of the pairs had eggs. Survival times were then compared. Layingsynchrony was monitored as part of a project examining the reproductive biology of thick-billed murres (deForest 1993). I predicted that eggs placed at the beginning of the layingperiod would survive for a shorter time, because birds not yet incubating their own eggsshould not contribute to group defence during gull attack.Behavioural ObservationsAfter egg placement, I observed gull foraging activity from blinds and recorded the attackmethods used, classifying them as shown in Table 2.1. In addition, I recorded whether eggswere removed successfully or dropped during the attempt. I also recorded the response of the14murres during attacks by gulls (Table 2.1). When more than one murre responded to an attack,I often recorded several types of behaviour simultaneously.Statistical AnalysisThe survival of eggs relative to ledge width and murre nesting density was comparedusing the Kolmogorov-Smirnov goodness of fit test. The survival of eggs placed on the samenest sites under two different conditions of both wind speed and numbers of neighbors witheggs were compared using the Wilcoxon matched-pairs test.The defensive responses of individual murres nesting on the same ledge could not beconsidered independent. In fact, it is likely that the response of an individual murre to a gullattack depends on the behaviour of its neighbours (Birkhead 1977). For each gull attack, Irandomly selected the response of one individual next to the egg for analysis of the effects ofledge width on murre defence behaviour. For my analysis I used the following categories ofmurre defensive response (see Table 2.1-a): 1) no defence = flush or no response, 2) moderatedefence = orient, and 3) strong defence = lunge. Finally, I used log-likelihood ratio (G) tests toexamine differences in weak, moderate, and strong murre defence behaviour relative to ledgewidth and laying date. I also compared the attack techniques of gulls using a G-test, with theWilliam’s correction for continuity applied because of small sample sizes (Sokal and Rohlf1981).15Table 2.1. a) Attack modes of gulls, and b) murre response to gull attack.Behaviour Descriptiona) Stand lunge Gull stands on murre nesting ledge, head lowered; lunges intonesting group of murres to take eggAerial lunge Gull flying along colony cliff drops into nesting murres to steal eggHover harass Gull flying along colony cliff stops to hover next to murre nest sites;gull then maneuvers into nest site to steal egg without landingb) No response Murres on nesting ledge do not alter behaviour during gull attackOrient Murres elongate their necks and direct their beaks towards attackinggullFlush Murres suddenly fly from nesting ledge avoiding contact with gullLunge Murre runs towards attacking gull standing on ledge; in response toaerial attack, murre drops from ledge and attempts to strike hovering gull16RESULTSNesting DensityOf the 77 experimental eggs that disappeared (n=92), 48 (66%) were taken within thefirst four hours and only 15 of a total of 92 eggs (16.3%) survived for 72 hours. I saw gullstake 58 (75.3%) of the 77 disappearing eggs. As predicted, eggs placed on ledges with highnesting densities remained for significantly longer than eggs placed under low-densityconditions regardless of ledge width (Fig. 2.1). Group defence by murres was important inpreventing gulls from reaching exposed eggs because gulls tried to avoid strikes from nearbymurres. On ledges with high nesting densities, murres collectively defended exposed eggs bylunging at or striking gulls as gulls tried repeatedly, and often unsuccessfully, to reach exposedeggs. On ledges with low nesting densities, gulls could walk among incubating birds whiletaking eggs. As predicted, gulls abandoned foraging attempts more frequently under highnesting density conditions (11 of 25 of high density attempts, 9 of 37 low density attempts;Gadj4.37, dfl, P<O.05).Experimental eggs may also have been hidden under incubating birds in high nestingdensity conditions. I considered that eggs were concealed when I could not see them fromseveral different vantage points on the cliff, and gulls did not investigate them when foraging intheir vicinity. Of the 15 eggs that survived to the end of the observation period (3 days), 5(33%) were concealed by neighbouring murres under high density nesting conditions.Ledge WidthAlthough nearly twice the number of eggs disappeared from broad ledges within the firsteight hours (Fig. 2.2), I found no statistically significant differences in the survival time ofeggs placed on narrow vs. broad ledges in 1990 (Fig. 2.2), contradicting my prediction.Murre responses to gull attack, however, did vary significantly with ledge width. Murres onbroad ledges responded more often to gull attacks with orient and lunge behaviours (Table 2.1;17Figure 2.1 Survival of exposed murre eggs in relation to thick-billed murre nesting densities(low murre nesting density - black; high murre nesting density - white). KolmorgorovSmirnov two-sample test, P<0.001). Wind conditions >10 km/h.10080CCi) 60-(15C,)c,)) 40w200500 1000 1500 2000 2500 >3000Survival time (mm)18Figure 2.2 Survival of exposed murre eggs in relation to cliff ledge width (broad ledge -black; narrow ledge -white). Kolmorgorov-Smirnov two sample test, P=0.483. Windconditions> 10 km/h.10080Cci)-60CeCl)40200II I H1000-H1500 2000Time (mm)2500 > 300019Table 2.2 Murre responses to gull attacks in relation to cliff ledge width. Each observationrelates to a randomly selected murre situated next to the experimental egg.Ledge widthMurre response Narrow BroadNo response 11 3Flush 6 5Orient 14 17Lunge 2 920G=8.41 dJ2, P<0.025), and were more successful in repelling gulls (15 of 34 cases on broadvs. 9 of 43 cases on narrow ledges; Gadj=4.653,dfl, P<0.05). Typically, murres on narrowledges only contributed to the defence of immediately adjacent nest sites even at high densities.Murres on narrow ledges also had difficulty turning on their sites to defend against attackinggulls without dislodging their own eggs. On broad ledges, birds moved tightly together andturned to face gulls that were attacking on foot. Murres in the interior of groups on broadledges often extended their heads over the birds situated on the edge and attempted to strikeadvancing gulls. In this way, several tiers of murres contributed to the defence of neighboursand of exposed eggs on broad ledges. Some individual murres on broad ledges also walkedbetween the gull and the group, and then lunged at the gull, forcing it from the ledge.Observations of banded murres revealed that these murres were often the mates of incubatingbirds (5 of 12 cases involving banded birds). However non-breeders ( 3 of 12) alsocontributed to defence.Gull attack behaviour also varied with ledge width (G=23.22, dfr2, P<0.001). Gullstypically made aerial attacks on narrow ledges, wheras they frequently landed on broad ledgesto make their attacks (Table 2.3). Aerial attacks were so rapid that gulls often failed to gripeggs firmly before flying off. Consequently, gulls dropped eggs taken from narrow ledges 3times as often (13 of 39) as eggs from broad ledges (2 of 19; Gadj3.7l,df1, P<0.07).While gulls were attacking on broad ledges, eggs sometimes rolled out onto unoccupied areasof the ledge beyond the reach of defending murres. This allowed gulls to pick up eggs moreslowly, to make several attempts if necessary, and even in some cases, to ingest eggs wholebefore leaving the ledge.Wind SpeedIn 1990, I did not find a significant difference in the survival of eggs placed on narrowand broad ledges. I therefore performed egg placements under windy conditions (>15km/hour) and found that gulls could reach narrow ledges using gliding flight. In 1991, I found21Table 2.3 Gull attack techniques on eggs placed experimentally in relation to cliff ledgewidth.Ledge widthAttack technique Narrow BroadStand Lunge 4 18Hover Harass 27 6Aerial lunge 12 1022that eggs placed under calm conditions on narrow, low nesting-density ledges survivedsignificantly longer than those placed on the same sites under windy conditions (Wilcoxon’ssigned-ranks test, paired comparisons, N=l4, Z2.88, P<0.00l). When it was windy, nearly90% of the eggs were taken within the first two hours, almost twice the percentage taken undercalm conditions during that period (Fig. 2.3).Vulnerability relative to timing oflayingEggs placed during the peak of murre egg laying survived significantly longer than thoseplaced on the same sites at the beginning of the laying period (Fig. 2.4; Wilcoxon’s signedranks test, paired comparisons, N=12, Z=2.81, P<0.025). All of the eggs were placed undermoderate wind conditions. Early in the laying period non-incubating birds often flushed whenattacked leaving experimental eggs exposed and incubating birds vulnerable to attack (Table2.4), although these differences in murre behaviour were not statistically significant in relationto timing of laying (Gadj=2.55,dJ2, P< 0.5, one-tailed). However, there was insufficientpower to reject the null hypothesis convincingly (1 -/3 = 0.381). The idea that murres weremore aggressive in their defence against gulls once most individuals were on eggs is supportedby the fact that murres repelled 3 times as many attacks following the peak in murre laying (9of 23, or 39% of gull attacks post peak; 2 of 23, or 13% of attacks prior to the peak;Gadj3.76,df=1, P<0.03, one-tailed).23Figure 2.3 Survival of exposed murre eggs in relation to wind (windy,> 15km/h, black;calm, <5km/h, white). Wilcoxon’s signed ranks test, paired comparisons, n=14, Z=2.88,P<o.o1.Cci)cts•1-C,)C)C)w100806040201jEl300 400Survival time (mm)500 > 60024Figure 2.4 Survival of exposed murre eggs in relation to the timing of murre laying (earlyblack; late - white). Wilcoxons signed ranks test, paired comparisons, n=12, Z=2.81.P<0.01. Wind conditions> 10 km/h.10080C 60-(‘5.4-CO3)C)uJ200100 200 300 400 500 > 600Time (mm)25Table 2.4 Murre responses during gull attacks in relation to timing of murre egg laying.Each observation is based upon one randomly selected murre situated next to the experimentalegg.Timing relative to laying peakMurre response Prior PostNone 2 3Flush 9 5Orient 7 17Lunge 2 426DISCUSSIONConstraints on gullforaging efficiencyMany colonial seabirds exhibit greater reproductive success at high nesting densities(Burger and Gochfeld 1994). Predator swamping, increased vigilance, and increasedcommunal defence may explain this phenomenon (Wittenberger and Hunt 1985; Burger andGochfeld 1994). Birkhead (1977) and Spear (1993) found that common murres which nestedon the periphery of groups were more vulnerable to gull attack and egg loss. I found that thegreater survival of thick-billed murre eggs at high nesting-densities was due largely to theability of murres in dense groups to strike gulls during attack. I found that gulls typicallystood on the edge of nesting groups and harassed the outermost birds even when attempting tosteal eggs from the center of large groups. Gulls often abandoned these attempts after beingstruck on the face and neck by the beaks of incubating birds. In contrast, gulls that foraged onledges with low nesting densities walked between brooding murres and often took murre eggswithout being struck by defending murres.Siegel-Causey and Hunt (1981) found that glaucous-winged gulls, Larus glaucescens,preferred to attack nests on level ground when foraging in a colony of pelagic, Phalacrocoraxpelagicus, and double-crested cormorants, P. auritus. Gulls were more likely to get struck bydefending cormorants when they foraged on steep terrain. In this study, colony topographyand cliff ledge width also influenced predation on exposed eggs. When brooding, murrestypically faced into the cliff and away from most gull attacks. Murre defence against gulls wasless effective on narrow ledges, apparently because murres had difficulty turning to faceattacking gulls without dislodging their own eggs. Consequently, murres on narrow ledgescould defend only their immediate neighbours, and most murres on narrow ledges defendedexperimental eggs weakly. In contrast, most murres on broad ledges turned to face gullsattacking on foot. Birds nesting in the interior of large groups typically extended their headsover the birds situated on the edge and contributed to group defence. Finally, more non-27breeding murres were present on broad than on narrow ledges (Noble 1990), and theysometimes defended against attacking gulls.Spear and Anderson (1989) concluded that wind affected the success of common ravens,Corvus corax, attacking colonial-nesting yellow-footed gulls, Larus livens. They noted thatravens attacked only during 25-40 km/hr winds, and suggested that increased maneuverabilityof ravens under these conditions enhanced their ability to avoid defensive mobbing by gulls(Spear and Anderson 1989). Similarly, south polar skuas (Catharacta maccormicki) weremore successful under moderate winds when stealing the chicks of Adelie penguins(Pygoscelis adeliae), apparently because wind increased their ability to out-maneuver defendingpenguins (Young 1993). In my study, wind influenced the ability of glaucous gulls to reacheggs on narrow ledges and avoid contact with murre during attack. Gulls reached weakly-defended eggs on narrow ledges when wind conditions enabled them to maneuver over to nestsites. Gulls apparently faced lower risks of getting struck by defending murres whenattacking nest sites on narrow ledges under windy conditions, and consequently, they wereable to remove eggs more readily from these sites.Surprisingly, wind effects have rarely been discussed in previous studies of avianpredation within seabird colonies (Spear and Andersson 1989; Young 1994; Thiel andSommer, 1994). In several studies, gulls and skuas, Cartharacta sp., foraged primarily onfoot in low nesting-density areas on broad cliff ledges or on level ground (Birkhead 1977;Furness 1981; Watanuki 1982; Watanuki 1983; Spear 1993). In contrast, the thick-billedmurre colony that I studied is currently expanding so most broad ledges were densely occupiedby murres (Gaston et al. 1993). Thus, the aerial foraging of gulls recorded in this studyreflected the lack of broad ledges with low nesting densities where gulls could forage on foot,and that wind enabled gulls to reach narrow ledges where murre defence was less effective.28Consequences ofgull predationfor thick-billed murresThe ability of glaucous gulls to take experimental eggs was greatly reduced by groupdefence of murres nesting at high densities. Eggs on the edge of a group were taken muchmore quickly than those in the interior. Despite this generality, several other notable factorsinfluenced the effectiveness of murre group defence.At Coats Island, communal defence by murres also depended on laying synchrony. Eggsplaced before the peak of murre egg laying were quickly taken even in high-density areas.Non-incubating murres often bolted from ledges when attacked, leaving experimental eggsvulnerable to gulls. Following the peak of laying, gulls were more likely to be repelled bymurres when attempting to take eggs. These results support the contention that individualswithout young should abandon the group and avoid contact with the predator (Anderson et al.1980). I conclude that the first murre eggs laid on a nesting ledge experience greater risk ofpredation because group defence is not firmly established early in the breeding season.Gulls used windy conditions to reach narrow ledges and avoid contact with murresduring attack. However, for birds nesting in the center of high-density groups, wind had noeffect because these sites were well defended regardless of wind conditions. Thus, thevulnerability of murre nest sites to gulls is partly determined by group defence, and partly bythe position of nest sites in relation to prevailing winds.Based upon these findings, I rank the vulnerability of thick-billed murre nest sites atCoats Island from high to low as follows: 1) low density edge sites on broad ledges; 2) lowdensity sites on narrow ledges; 3) high density sites on narrow ledges; and 4) high densitysites on broad ledges. I conclude that low density sites on broad ledges are most vulnerablebecause gulls could reach these sites on foot regardless of wind conditions. Nest sites onnarrow ledges are typically inaccessible during calm wind conditions (<15 km/hour).However, wind speeds often exceed 15 km/h for several hours on most days at the CoatsIsland colony (deForest 1993). Finally, gulls have difficulty reaching sites on high densitybroad ledges due to group defence, regardless of wind conditions. These findings provide a29mechanism to explain the results of Birkhead and Nettleship (1986) and of deForest (1993)who found that thick-billed murre reproductive success was lowest on broad, low densityledges, intermediate on narrow ledges, and greatest on broad, high-density ledges.30CHAPTER 3WiND AND PREY NEST SITES AS FORAGING CONSTRAINTS ON AN AVIANPREDATOR, THE GLAUCOUS GULLPredation is a major force influencing population dynamics, community structure, and preybehaviour (Zaret 1980; Taylor 1984; Sih 1989). The influence of predators on their prey isoften a function of changes in the environment (e.g. diet switching in response to changes inprey availability; reviewed in Sinclair 1989; Newton 1993). Consequently, it is of interest toquantify how predators respond to changes in their foraging constraints, and how thesedynamics influence prey populations (Werner et al. 1981; Erlinger, 1989; Cooper 1990; Millsand Shenk 1992; Spear 1993; Goss-Custard et al. 1995).It has been useful to examine predator foraging behaviour from an economic perspective(Stephens and Krebs 1986). For example, the most appropriate foraging strategy for apredator is to maximize the trade-off between energetic reward and risk of injury whileforaging, rather than to maximize net energy gain alone (Stein 1977; Pettifor 1990). Thus, thevalue of attacking prey should decrease as the danger of injury during attack increases for thepredator. The trade-off between energetic gain and risk of injury may be particularly importantamong avian predators which forage within seabird colonies, and who face the group anti-predator defences of their prey. The risk of injury while foraging, and the ability of avianpredators to overcome these dangers, may be influenced by colony topography, changes inprey nesting densities, and weather conditions (ch. 2 and references therein).Glaucous gulls are common predators of thick-billed murre eggs and chicks in the Arctic.Thick-billed murres breed in dense colonies on exposed cliff ledges, and they defendthemselves collectively against gull attack (Gaston and Nettleship 1981). In chapter 2, Ishowed that glaucous gulls are constrained by narrow cliff ledge width, murre nesting-density,and murre defence when they attack exposed murre eggs. Windy conditions enhance theability of gulls to overcome these foraging constraints (ch. 2). However, these conclusions are31limited to the conditions simulated by the egg placement experiments, because murres likelydefend their own eggs differently than experimental eggs. In this chapter, I examine theconstraints of gulls foraging on naturally occurring murre eggs and chicks using multivariatestatistical models. I also examine the affects of wind in greater detail to determine if windinfluences the choice of tactics by gulls and their foraging success. I expected that foragingactivity of gulls and predation rates would be positively correlated with windy conditions.Finally, I explore how wind speed and gull foraging behaviour interact to affect thevulnerability of murre nest types using a model which integrates data on predation rates andseasonal wind conditions.METHODSStudy area and speciesThis study was conducted at a thick-billed murre colony on the north eastern tip of CoatsIsland, Northwest Territories, Canada (62°30’N, 83°OOW) during 1990-92. For furtherdetails, see chapters 1 and 2.There was considerable variation in the types nesting ledges and in the density ofbreeding murres within the colony. I classified 5 nest site characteristics: 1) broad ledge siteshigh nesting-densities, 2) broad ledge sites at low densities, 3) narrow ledge sites with highdensities, 4) narrow ledge sites in low densities, and 5) crevice sites. I defined these sitecharacteristics following Gaston and Nettleship (1981): 1) broad ledges could support two ormore rows of breeding birds, whereas narrow ledges supported only one row, 2) high densitysites had 3 neighbours on broad ledges and 2 neighbours on narrow ledges. Crevice siteswere located under rock overhangs, in crevices, or within rock piles. The proportion of thetypes was determined by photographing study plots from the cliff at a distance ofapproximately 20-30 meters. Murres were counted from 15x25 cm black and white prints.The number of breeding pairs was estimated by multiplying the total number of birds presenton the cliff by a correction factor of 0.75 (see Nettleship 1976, for further details).32Behavioural observationsGull foraging behaviour was monitored from blinds on the cliff face. Study plotsextended Ca. 40 meters from the base of the blinds, and from the sea to the cliff top. Gullsappeared unaffected by my presence once I was inside a blind. Aerial search activity of gullswas monitored by recording the number of ffight patrols made by gulls over the study plotsduring half-hour time intervals. Gulls flying close to the cliff face with their head directedtowards brooding/incubating murres were considered to be patrolling. This distinct headposture made foraging patrols easily distinguishable from other flight behaviours (e.g. travel toand from the nest).The number of attacks made during each 30 mm. observation period was monitored. Iconsidered that an attack had occurred if a gull made an aggressive advance towards broodingmurres either on foot or from the air. For each gull attack, I recorded the attack behaviour(aerial or pedal), nest site characteristics of the murres attacked (cliff ledge width and nestingdensity), murre defensive response (do nothing, flush from ledge avoiding gull, oriented beaktowards gull, lunge at gull), and outcome (food item taken or not).Attack and patrol rates were monitored in relation to date, time of day, wind conditions,visibility, and the number of murres present on the study plots. Murre numbers on study plotsincluded both breeders and non-breeders. Wind was monitored with a Digitar anemometermounted 1.5 from the cliff. Visibility was recorded as clear (no fog), light fog (40m <visibility < 80 m), and dense fog (visibility <40m). Because visibility, wind conditions, andmurre attendance varied on a fine time scale, I chose 30 minutes as the duration of behaviouralobservation intervals.I examined the vulnerability of gulls to injury in two ways: first, I videotaped gull attacks;second, I used a gull model made of Styrofoam and plastered paper. The gull model wasintroduced into brooding murres from the edge of nesting groups using an extendible 4 cmdiameter pole to mimic natural attacks. With both methods, the location where murres struck33the heads of gulls, and the source of the defensive strikes (i.e. the individual being attacked orits immediate neighbours) was recorded.The influence of wind on aerial maneuverability was examined by timing the duration ofgull attack hovers and foraging patrols in relation to wind conditions on the cliff face. A hoverwas defined as a gull trying to maintain its position next to a murre nest site during aerialattack. Patrols were timed with a digital stopwatch as birds passed over the study plotsthrough a defined field of view (i.e. a 50 m x 40 m area of cliff). Patrols in which gulls madeattacks were not included in the analysis of patrol durations.Gull gliding-flight dynamicsI caught nine gulls, either on nesting ledges or at bait stations on the sea ice, by usingpadded leg-hold traps. I measured the body mass, wing length, and wing area of each bird,following the procedures outlined by Pennycuick (1989). I used these morphologicalmeasurements in Pennycuick’s gliding-flight model (Pennycuick 1989) to estimate theminimum air velocities at which glaucous gulls could glide.Effects ofgull predation on murre reproductive successPrevious experimental work suggested that gull predation was influenced by the width ofmurre nesting ledges, murre nesting density, and wind conditions (ch. 2 and referencestherein). Thus, I estimated the proportion of murre eggs and chicks lost to gulls annually asfollows:Ni=YZP(Tj*T) (1)100where N1 is the proportion of eggs that were originally laid which were taken from type i nestsites (e.g. broad ledge, high murre nesting-density); T is the total number of daylight foraginghours available to gulls in a breeding season at Coats Island (i.e. from the onset of murre egg-34laying to murre chick departure); T is the proportion of T hours at wind speedj; and P isthe predation rate (eggs taken! 1000 nest sites! hour) from nests of type i at wind speedj.Statistical analysisI used General Linear Modeling (SYSTAT, 1989) to explore factors affecting aerial searchactivity, gull attack rates, and predation rates. For each model, condition indices and tolerancediagnostics were examined to identify co-linearity among the variables (Systat 1989;Kleinbaum et al. 1988). If severe co-linearity existed between two variables, models were rerun and the variables were included in separate models. In cases where several variables wereexamined together, I did not conduct full factorial analyses. Instead, I included only a subsetof possible interactions which were selected after screening models and examining interactionplots (JMP software, SAS Institute), and from my previous knowledge of the system (ch. 2).Rates of gull attack and predation in relation to murre nest site characteristics were expressed asattacks! half hour! 1000 nest sites. I applied arcsine square-root transformations on rate dataof search activity, attack frequency, and predation prior to analyses (Sokal and Rohlf 1981).I used logistic regression analysis to examine factors influencing gull attack success, andthe factors affecting the probability that a gull was struck by a murre during an attack, becauseboth were binary variables. To test for the significance of the independent variable (orvariables) I used the likelihood ratio statistic (Kleinbaum et al. 1988). If year had asignificant influence on pooled data, models were derived for each year separately.Some of the behavioural data was collected during consecutive half hours of observation.To test if these data were independent, I conducted a time series analysis (Systat 1989) andfound that aerial search activity was indeed strongly auto-correlated for up to 2 and one halfhours. I therefore analyzed a subset of the original data in which one 30 minute period wasextracted from each 21!2 observation period. For the analysis which examined factorsaffecting attack success, each attack was considered to be an independent event. Although Iwas sampling gulls repeatedly, I was also studying all members of the gull population at Coats35Island roughly equally. Thus, the variation in the ANOVA is actually the population-widevariation over the course of the summer, rather than the variation of a limited sample.Following the above analyses, I constructed a path diagram to illustrate howenvironmental factors interacted with gull predation rates. The strengths of these influenceswere estimated by the standardized coefficients generated from the general linear models. Topermit comparisons between factors analyzed with general linear models, ANCOVA, andlogistic models, I re-ran the ANCOVAS and logistic analyses using multiple linear regressionspecifically for the variables included in the path diagram.36RESULTSColony structure and environmentThere was considerable variation in the structure of cliff ledges and the density of nestingmurres within the Coats Island colony. Of the murres nesting on the study plots, about halfnested on broad ledges in high-density groups (Mean = 47% ± 2.3 SE; n = 5 murre countplots). Approximately 25.0% (± 1.1) and 8.1% (± 0.78) of breeding pairs nested on narrowledges under high and low nesting densities respectively. A further 17.7% (±1.5) nested onbroad ledges under low nesting-density conditions. These included birds nesting alone or onthe edge of large groups. A further 5% (±1.2) of murres nested in crevices or caves. The lastproportion may be too low because these sites were difficult to count.Winds were frequently >40 1cm/h in 1990, and rarely exceeded 30 km/h in 1991 and 1992.Significantly stronger winds occurred in 1990 than in 1990-9 1 (90-91,t= 4.986, df=44,P<0.00l; 1990-92: t=5.34, df= 44, P= 0.001), however wind speeds in 1991 and 1992 weresimilar (t =0.973, df =97, P=0.33). As a result, most of my behavioural observations wereconducted under calm (<10 km/h) or moderate (10-40 kmlh)wind conditions in 1991 and 1992(Fig 3.1). Extended calm periods (>3 days at <10 km/h) occurred each summer, and weretypically accompanied by warm temperatures (15-25°C).Gull search activityAerial search activity was reduced under calm wind conditions in all years (Fig. 3.2, Table3.1, a-c). Date had a significant influence on search activity in all years, and there was astatistically significant interaction between date and wind conditions in 1990 (Table 3.1, a-c).This reflected daily fluctuations in wind speed.10080604020010 20 30 40 50 60 70604020010 20 30 40 50 60 70 80 9060199240 -20-.i\ /0 20LFigure 3.1. Wind conditions in relation to date at Coats Island.3719900-c2V0)0)CoV80 901991I — I — I I60 70 80 90Date (June 1 Day 1)38Figure 3.2. Examples of glaucous gull aerial search activity in relation to wind conditionsand time of day. Each figure represents three days of gull foraging activity. Solidhorizontal bars within each figure represent wind conditions above 10 km/h; openhorizontal bars represent wind conditions equal to, or below 10 km/h. Vertical barsrepresent periods of night when gulls could not forage.396O_60-______________________________2 - 6OTime (half h)40Table 3.1. General linear models of factors affecting rates of aerial search by gulls(patrols/half hour) as affected by five variates in (a) 1990, (b) 1991, and (c) 1992. Murreattendance was not monitored in 1990. Terms with more than one component, separated bya dot, indicate interactions between the component terms. t-values for single terms werecalculated after interaction terms were dropped from the model. See methods fortransformations conducted prior to the analysis and further rationale.Term Coefficient SE t-value P-valueModel (a) F5,165 = 44.7, 63.7% of variance explainedDate 0.033 0.011 3.028 0.003Time 0.000 0.000 0.814 0.417Wind 0.643 0.206 3.117 0.002Wind change 0.017 0.026 0.643 0.521Wind* date -0.001 0.001 -2.533 0.020Model (b) F5,395 28.0, 27.6% of variance explainedDate -0.031 0.009 -3.571 <0.001Time -0.001 0.001 -2.774 0.006Wind 1.009 0.089 11.338 <0.001Wind change -0.041 0.020 -2.007 0.045Murre attendance -0.002 0.002 -1.022 0.308Model (c) F5,638 = 30.1, 19.7% of variance explainedDate -0.010 0.010 -1.019 <0.001Time -0.000 0.000 -2.490 0.013Wind 0.458 0.051 9.017 <0.001Wind change 0.016 0.017 0.940 0.348Murre attendance -0.004 0.00 1 -3.030 0.00341Gull attack activityIn all years, attack rate was positively correlated with both aerial search activity and windconditions (Table 3.2 a-b). Wind conditions and aerial search activity contributed significantlyto the models when included independently or together. Changes in wind conditions betweenobservation periods had a significant influence in the calm years of 1991 and 1992 but not inthe generally windy year of 1990, indicating that gulls responded more strongly to increases inwind during calm years. Year, date, time of day, and visibility did not affect attack rate.Gull attack selectivity and murre nest site characteristicsBecause nesting ledge structure and murre nesting density varied within the Coats Islandmurre colony (see above), I examined how gulls selected for nest sites and in relation to wind(Table 3.3). Wind conditions had the strongest influence on rates of attack (ANcovA,F176674.54, P <0.001). As wind increased, gulls attacked a higher proportion of nest sites onnarrow ledges, apparently because wind enhanced their ability to reach these sites (Fig. 3.3,see below). Consequently, there was a strong interaction between wind conditions and thewidth of cliff ledges attacked (F1,6676 = 176.48, P > 0.001). Cliff ledge width alone had only aweak influence on attack rates (F16676 = 4.02, P = 0.045). Murre nesting-density had a strongand negative effect on attack rates (F1,6676 = 207.83, P <0.001). Gulls thus attacked areas oflow murre nesting-densities selectively (Fig. 3.3, see below).In summary, attack rates and the types of nest sites attacked by gulls were detenninedlargely by moderate to high winds, which enabled gulls to reach low nesting-density sites onnarrow ledges from the air.Gull predation ratesYear (199 1-1992), date, time of day, and murre attendance had no detectable influence onpredation rates. There were also no apparent seasonal trends in predation rates (Fig. 3.4).Instead, predation rates varied sharply over a few days, largely in response to changing wind42Table 3.2. General linear models of factors affecting rates of gull attack activity(attacks/half hour) in (a) 1990, and (b) 199 1-1992 combined. Murre attendance was notmonitored in 1990. Year (b), date, and time were not significant and are not reported.Search and attack rates were arcsine transformed prior to analysis.Term Coefficient SE t-value P-value(a) Model F4,165 100.7, 75.0% of variance explainedWind 0.178 0.034 5.303 <0.001Wind change 0.001 0.010 0.030 0.976Aerial search rate 0.363 0.037 9.791 <0.001(b) Model F8, 1033= 80.7, 38.7% of variance explainedWind 0.103 0.050 2.076 0.038Wind change 0.014 0.004 3.392 0.001Murre attendance -0.00 1 0.000 -2.695 0.007Aerialsearchrate 0.196 0.019 10.203 <0.00143Table 3.3. Analysis of covariance of factors influencing rates of gull attack in relation tomurre nest site characteristics. Variables separated by a (*) refer to interaction terms. Attackrate data were arcsine transformed prior to analysis.Source df SS MS F PWind 1 493.12 493.12 674.54 <0.001Cliff ledge width 1 2.94 2.94 4.02 0.045Murrenestingdensity 1 151.93 151.93 207.83 <0.001Ledge width* wind 2 129.01 129.01 176.48 <0.001Ledge width * nesting density 2 0.11 0.11 0.15 0.695Error 6676 4880.47 0.7344Figure 3.3 Attack activity of glaucous gulls in relation to wind and murre nest types.4-Broadlow 1990 J0---- Broad high3 - Narrow low TNarrow high2-1T//00 10 20 302T 19910//.11/4_%_,::;:::i.I3019923’OWind speed (km/h)45Figure 3.4. Glaucous gull predation rates in relation to date and year.121990108 I6 ICl,h..I IIHoo .q fo ••wI ‘iiHI0 I I10 20 30 40 50 60 70 80 90C81991.1Cl) 60oC;) rii0) I I0)CD I’ 7!3i ri2 t1:.10 20•‘,1992064 11* :.fI .1!2 . /.4 \ I’ I’ IIV\ /IT \?.J0 ... I I • I0 10 20 30 40 50 60 70 80 90Date (June 1 = Day 1)46conditions. Wind alone had a positive influence in 1990 but not in the generally calm years of1992 and 1991 (Table 3.4 a,b). Predation within the colony was determined by murre nestsite characteristics and by wind speed. Nest sites on narrow ledges experienced higher rates ofpredation when wind speeds increased above 10 km/h (Fig. 3.5). This was predicted basedupon the previous result that narrow ledges experienced higher rates of attack under windyconditions. Consequently, there was a significant interaction between narrow cliff ledge widthand wind conditions in all years (Table 3.4 a,b). Therefore, predation rates closely reflectedthe influence that wind conditions had on the selection of murre nest sites that were attacked.Gull attack successIn approximately 21% of 2407 gull attacks, an egg or chick was taken. Attack successdepended strongly on attack technique (Table 3.5). Attacks made on foot succeeded moreoften (308 of 671, 46%) than aerial attacks (191 of 1736, 11%). Attack success wasnegatively related to attack rate and wind speed (Table 3.5), because high attack rates occurredduring windy conditions when the less successful aerial attacks predominated. Attack successwas positively correlated with low murre nesting-densities and with weak defensive responsesby murres (Table 3.5). Cliff ledge width did not influence attack success, and contrary toprediction, there was no detectable interaction between ledge width and murre defensiveresponse. Fog and the attendance of non-breeding murres had no influence on attack success.In summary, high predation rates observed under windy conditions resulted from highattack rates and not from an increase in attack success.47Table 3.4. General linear model for predation rates (eggs-chicks taken! half hour) in (a)1990, and (b) 1991-1992 combined. Murre attendance was not monitored in 1990. Termswith more than one component, separated by a (*), indicate interactions between thecomponent terms. The effects of year (b), time of day, visibility, and previous conditionswere not significant and are not reported. t-values for single terms were calculated afterinteraction terms were dropped from the model. See Methods for transformations conductedprior to the analysis and further rationale.Term Coefficient SE t-value P-value(a) Model F6,210 = 12.96, 27.7% of variance explained, P < 0.001Wind 0.018 0.006 3.007 0.003Aerial search rate -0.027 0.039 -0.699 0.485Attackratebroadledges 0.167 0.095 1.761 0.080Attackratenarrowledges 0.176 0.070 2.527 0.012Wind * broad ledges -0.002 0.004 -0.436 0.664Wind * narrow ledges 0.006 0.003 2.4 10 0.0 17(b) Model F7,1120 = 47.1, 22.9% of variance explained, P < 0.001Wind -0.003 0.006 -0.433 0.665Murre attendance <0.00 1 <0.00 1 0.050 0.960Aerialsearchrate 0.020 0.004 5.575 <0.001Attackratebroadledges 0.051 0.011 4.759 <0.001Attack rate narrow ledges 0.004 0.0 12 0.328 0.743Wind * broad ledges -0.004 0.003 -1.229 0.2 19Wind * narrow ledges 0.009 0.003 2.583 0.0 1048Coci.1Co4-0ciC000a-c‘4--cCci)4-c,)c,)a)C,)ci)4-.C0t3a)ci0 10 20 30Figure 3.5. Predation rates in relation to wind, murre nest types, and year0. 10 20 3019920 10 20Wind speed (km/h)3049Table 3.5. Multiple logistic regression model of factors affecting gull attack success (eggschicks taken! attack). The effects of year, time of day, and interactions were not significantand are not reported.Term Estimate SE X2 value P -valueModel x2 [13, 12551 = 103.3, P < 0.001Wind speed -0.034 0.014 6.21 0.013Attackrate -0.104 0.031 11.55 <0.001Murre attendance -0.003 0.002 0.02 0.879Pedal attack techniques 0.701 0.105 44.50 <0.001Broadcliffledgewidth -0.003 0.138 0.00 0.971Low murre nesting density 0.506 0.265 3.63 0.051Low murre defensive response 1.368 0.4 18 10.76 0.00150Figure 3.6 Glaucous gull attack success in relation to wind speed and year. Error barsrepresent 1 Standard Error.U.b19900.30 000 10 20 300.6-CU)-19914—c,)ci)Cl) 0.3-0Ct 0.0 I • I •0 10 20 300.6 -19920.3 -0 0 ii.0 10 20 30Wind speed (km/ h)51Wind and gull maneuverability in flightI examined the duration of search patrols and attack hovers under various wind conditions.Aerial searching by gulls consisted of two distinct components: the patrol where gulls flewinto the wind along the cliff searching for prey; and the return, where gulls circled back overthe ocean flying with the wind. Search patrol duration increased significantly with wind speed(r 2 = 0.61, P <0.001; log transformed) because stronger head winds reduced gliding speeds(Fig. 3.7a). This likely increased their ability to locate exposed or poorly guarded murre eggsand chicks.Hover duration during attacks also increased with wind speed (r2 = 0.14, P < 0.002; linearregression with data weighted by their variance, Fig.3.7b). Wind conditions above 10 km/henabled gulls to maintain their position next to nesting ledges without flapping their wings(static-glide). Interestingly, the threshold wind speed above which gulls could static-glide waslower than that predicted by Pennycuick’s gliding model (Fig. 3.7b). I suggest that thisdifference occurred because gulls used updrafts that were not be measured by the anemometer.Murre responses to gull attackDuring attack, murres typically attempted to strike gulls with their beaks. Gulls attackedbrooding murres head-first, and this exposed their head and eyes to contact with defendingmurres. Video analysis of gull attacks and experiments with gull models indicated that thegreatest danger to gulls came from neighbours of intended victims. This occurred because theattacked murres typically struck at the beak of the gull, whereas its neighbours to either sidecould strike the head and eyes of the attacker. It is likely because of these risks of being struckthat gulls avoided large groups of murres and preferentially attacked murres nesting alone or onthe periphery of nesting groups.I used general linear modeling to explore how the width of nesting ledges, murre nesting-52C,)0ci)CoC0V04-00)C0)0LLFigure 3.7. a) Glaucous gull foraging patrol duration over murre breeding areas in relationto wind speed, and b) glaucous gull hover duration in relation to wind speed for gullsattacking murres from the air. The predicted line in figure b represents the wind speedat which glaucous gulls are predicted to glide generated by a ifight dynamics model(Pennycuick 1989), whereas the observed line is the wind speed threshold at whichgulls were observed to glide while foraging at Coats Island..a)I..•.• .: : 11.•$•. •I• •I •••1j I ••$ I.•1.I•.504030201000403020 -10010 20 30U)C.)ci)U)C0Dci>0-c-04-4-b) observed predictedI • .I. ILL .. IH I •.•••• . ...•.•••..• • • ••I • •• .• •• • •Z •.I_..I. • ! •0 10 20Wind speed (km/h)3053density, wind speed, and gull attack technique affected murre defence. The strongest murreresponse was found on broad cliff ledges (Table 3.6), because murres on broad ledges weremore likely to strike gulls than murres on narrow ledges. Gulls attacking on foot were struckmore often than gulls attacking from the air (Table 3.6), and consequently, wind conditionshad a significantly negative influence on murre response (Table 3.6). In summary, gulls thatattacked narrow ledges from the air under windy conditions were least likely to get struck bydefending murres.Path diagramI summarized the interactions of the various factors affecting gull predation rates using apath diagram which integrated the above analyses (Figure 3.8). Wind had a strong positiveinfluence on aerial attack technique, and both search and attack activity. Both search and attackactivity had a positive influence on predation rates of murre eggs and chicks. Aerial attack hada weak negative influence on attack success, which reflects that attacks in flight were lesssuccessful than attacks made on foot. However, a benefit of aerial attacks for gulls was thenegative affect they had on the effectiveness of murre defence, and presumably, the danger ofinjury for gulls. Therefore, wind facilitated the use of aerial attacks which had a lowprobability of success when compared with attacks made on foot, but which also wereassociated with low probabilities of contact with murres during attack.54Table 3.6. (a) General linear model for factors influencing the level of murre response togull attack, and (b) multiple logistic regression model for probability that gull was struck bymurres during attack. Murre response data for all three years were combined for analysisbecause they did not differ by year. All possible interactions were initially included andnone contributed a significant effect. The effects of year, and time were not significant andare not reported.Term Estimate SE t - value P -value(a) Model F [6, 1292] = 79.17, 17.5% of variance explained, P <0.001Wind speed -0.024 0.006 10.09 0.001Pedal attack techniques 0.086 0.066 1.67 0.196Broad cliff ledge width 0.339 0.055 37.12 <0.001Low murre nesting density -0.049 0.104 0.22 0.637Successful attack 0.153 0.076 4.04 0.046(b) Model x2 [6, 1294] = 84.12, 17.8% of variance explained, P <0.001Wind speed -0.026 0.022 1.37 0.242Pedal attack techniques 0.856 0.159 28.86 <0.001Broadcliffledgewidth 0.655 0.200 10.71 0.011Low murre nesting density -0.3 13 0.161 1.04 0.308Murre attendance 0.001 0.002 0.23 0.63 1Successful gull attack 0.307 0.160 3.66 0.05555Figure 3.8. Path diagram illustrating the relationships between factors affecting gullpredation rates. Solid arrows represent the direction of influences tested with multiplelinear regression. Shaded arrows indicate an absence of data, although an interactionlikely occurred. Values above arrows refer to the standardized regression coefficientsbetween. The sizes of coefficients indicates the magnitudes of influences, and thesecan be compared throughout the path diagram. The signs of the coefficients indicatewhether factors had negative or positive influences on each other. Arrows without acoefficient indicate an absence of significant influence, however negative or positivesigns of the non-significant coefficients are included for reference.+0.186SEARCHACTIVITY—PREDATIONRATE+0.761NESTINGDENSITYMURRENuMBERS--0.162+veLEVELOFMURREDEFENCENARROWLEDGESATTACKED+veAERIALATTTACK-0211TECHNIQUE+0.621-yeRISKOFINJURYFORGULLSATTACKACTIVITYye-ATTACKSUCCESS+veL157Colony-level effects ofgull predation on niurre reproductionMurre nest sites varied in their vulnerability to gull predation. Although newst sites onnarrow ledges experienced frequent predation at moderate wind speeds (Fig 3.5), calm windconditions restricted the ability of gulls to reach these nest sites. Therefore, nest sites onnarrow ledges suffered intermediate levels of predation over the total breeding season (Fig.3.9). Murres that nested at high densities on broad ledges could defend themselves effectivelyagainst attack regardless of wind conditions. In contrast, murres in low-density areas on broadcliff ledges suffered the highest levels of egg and chick loss because these sites were accessibleto gulls during all wind conditions. Thus, broad ledges supported both the safest (interior) andmost vulnerable (edge) nest sites within the colony (Fig. 3.9). These results are robustbecause there were no strong seasonal differences in attack rate, attack success, or predationrate over the course of the breeding season.58C,)Dc,)0.4-.4-U)0C,)c,)c,)ci)15C0002cLFigure 3.9. Estimates of the proportion of murre eggs that were originally laid which weretaken by glaucous gulls during an entire breeding season at Coats Island.Reproductive failure due to predation is presented in relation to murre nest sitecharacteristics (e.g, BH= broad cliff ledge width, high murre nesting-density;BL=broad cliff ledge, low nesting density; NH=narrow ledge, high density;NL=narrow ledge, low density; CAVE=crevice nest site). For further definitions ofmurre nest site characteristics, see Methods.20100BROAD HIGH BROAD LOW NARROW HIGH NARROW LOW CAVENest site characteristics59DISCUSSIONWind: a foraging constraintfor glaucous gulls preying on thick-billed murre eggs and chicksGlaucous gull foraging activity and predation rates were higher under windy conditions.Typically, gulls patrolled the murre colony when wind conditions exceeded 10 km/h. Whatadvantages could account for the strong interaction between wind and aerial foraging by gulls?First, gulls flew into the wind when searching, and higher winds slowed their rate ofsearch. Foraging patrols over nesting murres took twice as long under windy conditions thanunder calm conditions (Fig 3.7a). This likely increased the ability of gulls to locate exposed orpoorly guarded murre eggs because slower search speeds typically increase the likelihood that aforager detects concealed food items (Norberg 1977; Gendron and Staddon 1983; Bell 1991).Second, wind also increased the ability of gulls to attack vulnerable eggs and chicks oncethey had been detected. Under low winds (<10km/h), gulls had difficulty hovering next tomurre nest sites (Fig. 3.8b), and they typically had to circle over the ocean to maintain theirflight speed before returning to make an attack. Under these circumstances, gulls often couldnot relocate the vulnerable sites after circling. In contrast, gulls could maintain their positionnext to murre nest sites during attack under windy conditions (Fig. 2.8b). Indeed, Ioccasionally observed gulls gliding backwards under windy conditions to take eggs that theyhad initially passed over. Thus, windy conditions enabled gulls to make rapid and effectiveattacks almost immediately after prey were detected. This is important because, in most cases,eggs and chicks were only exposed to gulls briefly while an adult murre preened or anincubation change-over took place between brooding murres.Third, and perhaps most importantly, windy conditions allowed gulls to reach poorlydefended, and difficult to land on, nest sites. Murre defence was weak on narrow ledgesbecause most murres on narrow ledges faced into the cliff, and could not turn readily to facegulls without dislodging their eggs and chicks. Consequently, murres on narrow ledges could60only defend their immediate neighbours. In addition, most gull attacks at narrow ledges wereswift and from the air. In contrast, murres on broad ledges could turn to face gulls that wereattacking on foot. Murres on broad ledges could also act collectively to defend theirneighbours (see also Birkhead 1977; ch. 2). I therefore conclude that windy conditionsenabled gulls to overcome constraints imposed by both colony topography and prey defenceand thus reach narrow ledges where murre defence was ineffective.Other avian predators use wind to enhance their foraging success (Rudolph 1981,American kestrels; Spear and Andersson 1989, common ravens; Haney and Lee 1993, gulls atsea; Theil and Sommer 1994, herring and greater black-backed gulls; Young 1994, south polarskuas). However, not all aerial foragers are efficient in high winds because wind interfereswith prey detection and pursuit in other systems. In these cases, the benefits gained byincreased maneuverability and lower energetic search costs during flight in winds are notsufficient to increase the net foraging efficiency of the predator. For example, the foragingability of ospreys, Pandion haleaetus, and some terns, Sterna sp., decreases during windyconditions despite increased maneuverability, longer hovering bouts, and more time spentgliding. The surface turbulence on water created by wind interferes with their ability to locateprey (Dunn 1973; Grubb 1977; Taylor 1983; Machmer and Ydenberg 1990). Wind may alsoincrease the maneuverability of avian prey and their ability to escape predation orkleptoparasitism (Amat and Aguilera 1990). Wind and poor weather may also decrease theactivity of prey, and in turn, the probability that predators detect them (Mearns and Newton1988). Based upon these considerations and my findings, I predict that wind should enhancethe foraging efficiency of avian predators when: 1) aerial maneuverability increases theaccessibility to prey and/or the likelihood of successful attack, 2) the energetic costs of searchand attack during flight dramatically influence the net profitability of prey, and 3) wind doesnot greatly enhance the ability of prey to escape or avoid detection.61The currency ofglaucous guilforaging decisionsA basic theoretical premise of behavioural ecology is that animals select behaviouralstrategies that maximize fitness. In practice, indirect ‘currencies’ for estimating fitness arederived from the natural histories of foraging animals and the constraints facing them (Siblyand McCleery 1985; Mangel and Clark 1986; McNamara and Houston 1990). For example,classical foraging theory predicts that animals should select foraging strategies that maximizetheir net energetic gain while foraging (Stephens and Krebs 1986). This could occur eitherthrough maximizing the rate of energy gain or the efficiency with which it is obtained, andthere is considerable evidence that both strategies occur in the wild (reviewed in Ydenberg et al.1994). In my system however, glaucous gulls appeared to do neither of these things.Most gulls were inactive for extended periods of time during calm wind conditions at CoatsIsland (<10km/h; Figs. 3.2, 3.3). Once windy conditions returned, however, gulls which hadbeen loitering at the colony began to forage immediately which suggests that they had beenwaiting for wind conditions to increase. This is surprising since adult gulls were surroundedby their prey, and attacks made on foot under calm conditions were highly successful.Although there were several benefits from aerial foraging under windy conditions (see above),these benefits cannot explain both the foraging inactivity of gulls under calm conditions, andtheir reluctance to forage on foot. -The foraging inactivity of gulls could simply have been a response to changes in theirenergetic requirements under different weather conditions. For example, during calmconditions the thermal maintenance requirements for glaucous gulls would be low (Gabrielsenand Mehlum 1984). Under windy conditions, gulls might need to increase their forging activityto meet their higher energetic requirements due to higher rates of heat loss. However, threefactors suggest that this does not explain foraging inactivity of gulls under calm windconditions. First, gulls occasionally left the colony to forage at sea under calm conditions.This strategy yields less energy than could be obtained from a single murre egg obtained byforaging on foot (Spear 1993; ch. 4). Second, under windy conditions gulls foraged primarily62on the wing although this probably incurred higher energetic costs (Wiens 1984), and wasaccompanied by lower attack success (11%) compared with gulls foraging on foot (43%).Finally, gull chicks actively begged and harassed adults during calm conditions and this activityappeared to intensify as calm conditions persisted (pers. obs.). Collectively, these resultssuggest that gulls were not satiated during calm conditions, and that they selected foragingmodes which did not yield the highest net energetic gains even when foraging conditionsimproved. I therefore suggest that energetic considerations alone are insufficient to explainglaucous gull foraging behaviour at Coats Island.Foraging theory has recognized that environmental constraints often force foragers awayfrom the strategy that maximizes net energetic gain (McNamara and Houston 1986; Mangel andClark 1986; Lima and Dill 1990; Dill 1986). For example, prey may minimize their exposureto mortality risks while foraging (e.g. predation), and this may lower their foraging efficiency(Lima and Dill 1990). For predators, injury while foraging often results from prey fightingback (Curio 1974). Although these injuries are rarely fatal, the potential loss of fitness for thepredator is great. Among glaucous gulls, the risks of eye injury associated with foraging onfoot could reduce the fitness value of foraging on foot under calm conditions. I propose that atradeoff between energetic gain and risk of injury while foraging explains the reluctance ofgulls to forage on foot, and consequently, the reduced foraging activity observed under calmconditions. Sutherland and Moss (1985) predict such a foraging strategy in systems where, 1)poor foraging conditions are short-lived relative to the energy reserves of foragers and theirdependent young (e.g. the duration of calm wind conditions), 2) the forager is highlysuccessful once conditions improve, 3) the energetic value of a food item is large (here, 714 kJ/murre egg; Spear 1993), and 4) the mortality risks associated with improved conditions are low(e.g. the ability to reach poorly defended narrow ledges on the wing). Therefore, waiting forimproved foraging conditions could be an appropriate foraging strategy for glaucous gulls atCoats Island, particularly considering their long lifespan and the limited contribution of a63In summary, I suggest that glaucous gull foraging behaviour at Coats Island reflects atrade-off between the dangers of injury while foraging at the colony (which is generated bymurre defence) and the energetic gain of captunng a murre egg or chick. This tradeoff appearsto be mediated by wind conditions, which alter the reward/danger ratio of alternative foragingdecisions. Tests of these conclusions requires a closer examination of foraging energetics andrisk of contact with murres. I explore these issues further in chapter 4.Population-level consequences ofguilforaging behaviourfor murresGlaucous gulls preyed on murre eggs and chicks throughout the breeding season, and theirforaging activity and predation rates were positively correlated with windy conditions. Windalso influenced the characteristics of the murre nest sites that were attacked by gulls.Consequently, the level of gull predation was both a function of wind conditions over thecourse of the breeding season, and of predation rates in relation to nest site characteristics. Byintegrating data on wind conditions at Coats Island and gull predation rates relative to windconditions, I calculated the proportion of eggs and chicks lost to gulls in relation to murre nestsite characteristics using equation 1. From this, I ranked the vulnerability of murre nest sites togull predation from high to low as follows: 1) low-density edge sites on broad ledges; 2) low-density sites on narrow ledges; 3) high density sites on narrow ledges; 4) high-density sites onbroad ledges; and 5) crevice nest sites. Under both actual (this study) and simulated predation(ch. 2), these rankings are the same. My findings also partly explain the results of Birkheadand Nettleship (1986) and of deForest (1993) who found that thick-billed murre reproductivesuccess was lowest on broad, low nesting-density ledges, intermediate on narrow ledges, andgreatest on broad, high-density ledges.However, reproductive failure due to gull predation at Coats cannot account for all of theegg and chick loss experienced by murres. By integrating data on murre reproductive successcollected during the years of this study at Coats Island (deForest 1993), with the resultspresented in Figure 3.9, I estimate that gulls accounted for 0% (crevice), 13% (broad-high),6421% (broad-low), 13% (narrow-high), and 18% (narrow-low) of murre reproductive failure ateach of these nest sites. Murre nest site characteristics could also affect the likelihood that eggsand chicks are dislodged during incubation change-overs between members of a breeding pairor during fights between neighbours (Birkhead 1977; deForest 1993).Despite these low annual predation rates, my results illustrate how avian predation selectsfor murres to nest at high densities. The predation rate at low-density nest sites on broad cliffledges was 6 times greater than at similar sites at high nesting-densities. From this, I predictthat colony-wide predation rates should increase at murre colonies if the proportion of birdsnesting at low densities in open, level habitat increases. Avian predators would be lessconstrained by both calm wind conditions and by group defence if they could forage on footunder low risk of contact regardless of wind conditions. If this occurred, avian predationcould slow the expansion or colonization of new seabird nesting areas in open habitat (e.g.Johnson 1938; Spear 1993), or steepen declines in populations that are decreasing. I explorethese issues further below in chapter 5.ConclusionsIn this study, windy conditions allowed gulls to overcome constraints imposed by colonytopography and prey defence. Both attack activity and predation rates were positivelycorrelated with windy conditions. Consequently, the impact that predation had on murrereproductive success, which ranged from 0% to 21% of murre reproductive failure dependingon murre nest types, was determined largely by wind conditions and the accessibility of murrenest sites to gulls. I suggest that a decline in the density of nesting murres could enhance theability of gulls to overcome the constraints of calm wind conditions, cliff ledge accessibility,and prey defences.65CHAPTER 4FORAGING MODE SELECTION OF GLAUCOUS GULLS PROVISIONINGYOUNG: A DANGER-REWARD TRADE-OFF MEDIATED BY WIND?Life history theory is based on the assumption that trade-offs exist between various activitiesin an organism’s life. For example, the risks and energy allocation associated with presentreproduction may reduce an organism’s ability to reproduce in the future (Williams 1966).Trade-offs between the advantages and disadvantages of particular levels of reproductiveexpenditure may affect many life-history characteristics, including the age of sexual maturity,survival rate, brood size, reproductive frequency, and short-term investment in young.Among organisms with parental care, provisioning can add to the cost of reproductionor reduce the probability of adult survival (Drent and Daan 1980; Reznick 1985; Nur 1988;Hochachka 1990). Provisioning could increase energetic demands on the parent, thusincreasing risks of adult mortality or reducing energy available for future reproduction (e.g.Dijkstra et al. 1990; Beauchamp et al. 1991). It could also increase the exposure of the parentto predation (e.g. Harfenist and Ydenberg 1995). Thus, parents may be in conflict with theiroffspring about the optimal level of provisioning to be maintained (Ydenberg 1994).The provisioning choices of top predators could also be influenced by the risk of injurywhile foraging as a direct result of prey fighting back (Curio 1974). Risk of injury couldincrease if predators were forced to make more kills, or to switch to a more dangerous prey tomeet the energetic demands of their young. Under risk of injury, the provisioning strategywhich maximizes the energetic contribution to young might not be the one that maximizeslifetime reproductive success for the adult. Further, the relative costs and benefits of theseforaging decisions should vary with 1) changes in the availability of prey, 2) the risksassociated with subduing prey, and 3) the energetic state of the offspring. Applying a life-history framework could provide insight into whether variation in foraging behaviour within a66predator population is caused by evolutionary trade-offs between present and futurereproduction (e.g. Sibly and McCleery 1985; Dijkstra et al. 1990; Pierotti and Annet 1991).The glaucous gull, Larus hyperboreus, is well suited for a study of these topics. AtCoats Island, Northwest Territories, Canada, glaucous gulls prey on the eggs and chicks ofcolonial cliff-nesting thick-billed murres, Uria lomvia, and they use several foraging modeswhen doing so. In general, the most successful foraging modes incur the greatest contactwith defending murres so that a trade-off may exist between energetic gain and risk of injury(ch. 3). Weather conditions at the murre colony may also influence this trade-off. Windimproves the aerial maneuverability of gulls which in turn increases their ability to reach murrenesting ledges and avoid contact with murres during attack (ch. 2, 3).At Coats Island, breeding gulls are often inactive for extended periods of time undercalm wind conditions (ch. 3). During such periods of inactivity, their chicks are rarelyprovisioned. In chapter 3, I suggested that this foraging inactivity is the response of adultgulls to the dangers of foraging under calm wind conditions which is generated by murredefence. Thus, foraging inactivity under calm conditions may reflect a conflict betweenparent and offspring in which the adult ensures its survival and future reproduction at theexpense of its current brood. An alternative explanation is that gulls are simply responding tothe varying energetic demands determined by changing weather conditions, and that dangerof injury while foraging is unrelated to foraging inactivity.In this chapter, I present a dynamic optimization model of gull foraging modeselection which explores these two alternatives in the context of life history theory. In thismodel, the optimal decision for the adult gull is to choose the foraging mode that maximizesthe trade-off between current brood survival (via provisioning), and the risk of fatal injurywhile foraging. The model integrates field data on gull foraging mode selection withenergetic parameters obtained from the literature, to explore the above alternatives. I comparethe model’s predictions with field data on gull time budgets, and discuss how thesepredictions explain patterns of individual variation observed among wild gulls.67METHODSStudy site and species interactionsThe field component of this study was conducted at a thick-billed murre colony locatedon Coats Island. See chapters 2 and 3 for further details.Though the diet of breeding gulls and their chicks at Coats Island was made up mostlyof murre eggs and chicks (90%), gulls occasionally fed at sea. Trips to sea typicallyexceeded three hours duration. Food resources inland from the colony were limited becauseCoats Island has no rodents, and carrion on the tundra was quickly consumed by Arcticfoxes (Alopex lagopus). No waterfowl colonies existed near the murre colony.Gulls depredated murre eggs and chicks throughout the breeding season. Early in theseason, members of gull breeding pairs took turns foraging away from the nest so that oneadult was always present with the chicks. After 2 weeks, gull chicks were often leftunattended while both parents foraged. Members of a pair did not forage together nor didthey attack murres cooperatively. Gulls attacked murres on foot or from the air, and ingeneral, attacks made on foot were most successful (ch. 3). Attacks on foot also incurred thegreatest risk of contact with murres during attack because the presence of a gull standing on anesting ledge afforded murres the time to establish a collective defence (ch. 2, 3).Field studies ofguilforaging behaviourGull attack modes were studied among the general gull population. From blinds, Iobserved predation attempts and monitored the frequency of each attack mode, the probabilityof their success, and the frequency that the gulls were struck by defending murres duringeach attack (for further details, see ch. 3).In 1992, 8 glaucous gulls were captured using padded leg-hold traps and each wasindividually marked with acrylic paint The attack mode repertoire, attack rate, and attacksuccess of each bird was monitored. The length of time each gull devoted to the following68activities was also studied: 1) at the nest ledge; 2) loitering near but not on the nest ledge; 3)absent from colony; 4) foraging on foot; 5) aerial foraging; and 6) scavenging at the colony.Foraging behaviour and time allocation were studied in relation to two wind speed categories:calm < 5 km/h and windy> 10 km/h. Above 10 km/h gulls could search the colony usinggliding flight and were able to reach narrow murre breeding ledges (ch. 2, 3). Preliminaryanalysis of attack mode selection indicated that 2 of 8 birds (terrnedfoot specialists), attackedmurres predominantly on foot regardless of wind conditions. Thus, I partitioned time budgetdata so that aerial and foot specialists were analyzed separately.Dynamic optimization modelThe following assumptions describe interactions of gulls and murres at Coats Islandwhich were incorporated into the dynamic optimization model. Several of these assumptionswere identified through experimentation and behavioural observation in the field (ch. 2, 3).Aspects of gull energetics not studied at Coats Island were taken from the literature.Assumptions concerning model parameters are introduced here, and further details ofparameter values are provided below.The dynamic model simulates the foraging mode selection of an adult gull responding tochanges in both the energy stores of its chicks and to wind conditions. The brood obtainsenergy through provisioning from the adult. Adult gulls choose between the followingforaging modes: 1) do nothing and loiter at the colony, 2) attack murres by foraging on foot,3) attack murres using flapping ifight attack modes, 4) attack murres using gliding flightmodes, 5) scavenge at the colony, and 6) leave the colony to forage at sea. The gull cannotuse the gliding flight foraging mode during calm wind conditions (w=0, wind speed <5km/h) due to the limitations of flight dynamics (Pennycuick 1987, 1989; ch. 2, 3).However, when wind speed exceeds 10 km/h (w=1), all foraging modes are available to thegull. In the model, adult gulls select a foraging mode at the start of each hour interval (t= 1,2,3 ...7). Each foraging mode (i) carries with it a wind-dependent probability of69successfully obtaining a food item in each period t (2), and an energetic cost for the adult(q,) measured in U/h. Each foraging mode also carries with it a survival probability c,, sothe risk of fatal injury to the adult during the interval is, 1-c,.Brood energy dynamicsThe energy reserves of the brood in period t is x(t). The gross energetic gain that isavailable to the brood (/3j is a function of the energetic value of the food item () multipliedby the metabolizable assimilation efficiency of the food type, * = J3j. Assuming adelivery of /3 the energy status in the next period is,x(t÷l)=x(t)+/31 -ow-p1w (1)and if no delivery,x(t+1)=x(t)-8 -p (2)where 8 is the brood metabolic costs at rest, and q, is the energy requirement of the adultusing foraging mode i during that time period. /3 , 8, and q, are all dependent on windstatus w, and /3 and p are also dependent on i. Energy status evolves in this way until thefinal time period T, at which time brood success can be assessed.Survival of nestlings to breeding S (x(1)) depends on energy reserves x(t) (Fig. 4.1.The adult terminal fitness (F) depends on the success of the brood and also on adult survival.If the adult survives to the end of the terminal period T, fitness is assessed as,F (x, w, I) = S (x (7)) [w=O or 1] (3)Fitness is assessed at the end of each foraging interval. Wind conditions are set at thebeginning of each interval, following which, the gull is able to make its foraging modexCl,ci,ci;0>,2:Cl)Vci)4-,C.)C)0.w70Figure 4.1 Brood success S (x) as a function of energy reserves x at time tSmax is the maximum number of chicks that can survive to breed (3). TheXmin is the minimum reserves a 3 chick brood can carry (986 kJ), and xmaxis the maximum reserves (14985 kJ).S(x(T))=Smax *Smaxx(7) - Xminx(7) - Xmin+ 2000300 5000 10000 15000Brood energy reserves x (7)71decision. Wind conditions change from interval to interval according to the Markov processdefined in Table 4.1.At the end of the penultimate period T- 1, the adult still has one period to forage, duringwhich it may deliver J3 to the brood, or during which it may die with probability 1-Wind conditions during T- 1 are w. If it is calm (w=0) during period T, the adult’s expectedfitness at the start of the final period is,[ *]+[CL1 *,2 *F(X+/31w_3w 0,1)]+[cQ *(1)1)*f(x_& 0,1)]If it is windy (w= 1) during period T, the adult’s expected fitness at the start of the finalperiod is,[(1CL)*oJ+[CL1 *,2, *F(x+f3_& -p1,, 1,7)]+[cL1 * (1_2L,)*F(x_3 Piw’ 1,1)]In either case, the adult chooses foraging mode i (at the start of the final period T, after itknows w) to maximize the appropriate expression above. The expected fitness at the end ofthe penultimate period T -1 depends on the probability of calm or windy conditions in thenext period provided in Table 4.1. We can use these probabilities to derive the expectedfitness at the end of each period as,72Table 4.1 The probability of wind conditions in the next time period in relation to windconditions in the current time period.CurrentPeriodtNext period (t + 1)CalmWindyCalm Windyq=O.8 l-q‘-p p=o.873F(x, 1, t) = max { [(1- cq) * 0] (4a)+[ (p)*cj*.a,l*F(X+5_3(1)q,(1),1,t+1)]+[(1-p) * cj * * F(x-f-—8(o)- q(O),O, t+ 1)]+[(p) * yj * (1 -)) * F (x —5(1)-q1(1 , 1, t+ 1)]+[(1p)*j (l)*F(x_o(o)(o),0,t+1)]}A similar fitness algorithm is derived when wind conditions in the current time step are calm(w = 0), and given the probabilities that the wind would increase (1 -q) or stay calm (q) in thenext time period.F (x, 0, t) = max { [(1- cq) * 0] (4b)+[(1q)*j *F(x+f3_8(1)(1), l,t+ 1)]+[(q) * * * F(x +J3—ö(o)-q1(o), 0, t+ 1)]+[(1q)*oj *(1 l,t+ 1)]+[(q) * * (1 * F (x — (0) - (0), 0, t + 1)] }The model was implemented as described in Mangel and Clark (1986) by beginning with theterminal time period T (equation 3), and iterating backwards using equations 4a and 4b.74Model assumptions concerning provisioningThe gull being modeled was assumed to provide half the energetic requirements to itsbrood of three chicks. Behavioural observations at Coats Island indicated that members of apair contributed roughly equally to the provisioning of chicks. In some other gull species,males typically contribute more to provisioning than females (Watanuki 1989; Pyle et al.1991). The model also assumed that the adult could accurately assess the energetic state ofthe brood during each hourly time step, and that the adult delivered food items during thehour in which it was obtained. Observations in the field indicated that this latter assumptionwas generally the case.The model assumed that following a successful prey capture, the adult provisioned thebrood with the energy gained minus its own energetic requirements for that hour (q).Thus, the adult was assumed to maintain its own energy balance while provisioning, so thatthe model did not follow the energetic state of the adult. This simplification allowed me todevelop a tractable model which considered two states: chick energy stores x(t), and windconditions w(t). However, this approach assumed that provisioning strategies wereindependent of the energetic state of the adult. This simplification has important implicationsfor studies of provisioning strategies (Ydenberg 1994), and it therefore warrants furtherelaboration.The perceived value of a food item and the costs a forager is prepared to pay to obtain it,may differ if the food is for the foragers own consumption or for delivery to its young(Ydenberg 1994). In most cases, the interplay between the energetic states of both adult andyoung in studies of provisioning should be considered. This is particularly true if the forageris constrained by time or energy limitations, because as a forager approaches a time or energyboundary it may alter its provisioning strategies, foraging strategies, or both (Ydenberg et al.1994). However, I suggest that adult glaucous gulls have few time or energy constraints inrelation to the time scales considered by this model.75Gulls at Coats Island have up to 20 hours of daylight in which to forage during the periodof the breeding season that this model considers. Further, gulls raise their young surroundedby their primary food, so that the time between prey capture and delivery to the brood isnegligible. Indeed, gulls often take eggs and chicks from murres who share their ownnesting ledges. Considering the number of attacks that gulls can make in a day, the highprobability of their success, and the negligible travel time while searching for prey anddelivering prey to chicks, it seems unlikely that gulls are constrained by foraging time.I also assume that adult gulls at Coats Island are not constrained energetically. Thedistances traveled while searching for prey are small, and the energy obtained from a singleprey item is high relative to the energetic requirements of both foraging and maintenance (seebelow). For example, I estimate that a gull using the flapping flight mode for over 6 hours,could still achieve a positive energy balance for itself if it obtained a single murre egg duringthat time (i.e. a prey capture rate actually lower than that observed in the wild, see below).Further, large gulls typically carry fat stores during the breeding season which can easilymaintain them through short term periods of poor foraging conditions similar to thoseconsidered by this model (Spaans 1971; Coulson et al 1983; Sibly and McCleery 1985).Therefore, it is reasonable to assume that adult glaucous gulls generally maintain a constantenergetic state by assimilating portions of the prey that they deliver to their young.If the energetic state of the parent is constant, short-term provisioning strategies shouldbe more greatly influenced by the energetic state of the young, or by other factors influencingadult fitness (e.g. risk of injury while foraging). The model therefore concentrates on twoparameters: 1) the energetic state of the chicks, which, due to their dependence on theirparents, their smaller energy stores, and their higher thermal conductance, are more likely tobe affected by poor foraging conditions; and 2) the risk of fatal injury for the adult whileforaging.76Weather conditionsWind conditions at the Coats Island colony ranged between 0 and 70 km/h throughoutthe breeding season. However, wind conditions were usually moderate and ranged between0 to 25 km/h (ch. 3). This model considered the period of the breeding season when gullspreyed on murre eggs and when gull chicks were approximately 900g (half their final massof 1800g). During this period (early July), snow storms occurred occasionally andtemperatures regularly fell below 2°C at night, i.e. below the thermal neutral zone of adultglaucous gulls (2°C, Gabrielsen and Mehlum 1984).Metabolic rate estimatesThere are two basic approaches to studying avian energetics in the field. The first is tocalculate total gross energy intake and compare it with estimates of Field Metabolic Rate(1MR), which is the total energy cost that a wild animal pays during the course of a day(Nagy 1987). Estimates of the FMR are typically derived using the doubly-labeled watertechnique (Nagy 1987; Birt-Friesen et al. 1989). With this method, the relative energeticcosts of each activity are typically not partitioned (e.g. flight, brooding young, walking).The second approach is to compare total daily gross energy intake with estimatedexpenditures of each of the activities performed during the day (Tarboton 1978). In thiscase, most activities are given costs as multiples of the Basal Metabolic Rate (BMR) while theenergetics of flight are usually derived from flight dynamics models (Gabrielsen et al. 1987).BMR can be defined as the rate of energy used by animals at rest and in a fasting condition,and at an ambient temperature at which little or no extra heat is required for the maintenanceof body temperature (Kendeigh 1970). BMR may be estimated from either allometricequations based on mass (Tarboton 1978), from studies using metabolic chambers or doublylabeled water techniques (Gabrielsen and Mehlum 1984), or by combinations of thesemethods (Weathers et al. 1984; Birt-Friesen et al. 1989). I applied the BMR approach fortwo reasons. First, data existed for the BMR for glaucous gulls and other high latitude77seabirds existed (Birt-Friesen et. a!. 1989; Ellis 1984; Gabrielsen and Mehium 1984).Secondly, I could simulate foraging mode decisions and the energetic state of adults andchicks on a fine time scale, which permitted me to compare output of the model with fielddata on gull foraging behaviour.I estimated the basal metabolic rate (BMR, kJ day-’) of adult glaucous gulls (meanweight = 171 5g) using the allometric equation of Ellis (1981) for seabirds:BMR = 381.8 M°721 (5)where M is the mass of the bird measured in kilograms (kg) and BMR is measured in kJ/day.This equation predicts a BMR of approximately 563 kJ day-’ or 23.5 kJIh for adult glaucousgulls. I adjusted this estimate by +20% to account for discrepancies observed betweenallometric predictions of BMR for seabirds and the results of metabolic chamber studies ofglaucous gulls (Gabrielsen and Mehlum 1981). These high levels of BMR have beenattributed to a latitudinal gradient in which high-latitude seabirds are thought to maintainhigher metabolic rates to mediate the effects of their cold environment (Ellis 1984; Gabrielsenand Mehium 1984; Gabrielsen et al. 1987; Klaassen 1988). Following this adjustment, Iestimated a BMR of 675.6 kJ day-’ or 28.2 kJIh-1 for adult glaucous gulls. I assumed thatthe Resting Metabolic Rate at the nest was 1.9 xBMR or 53.6 kJ/h for adults. This multipleof BMR was the energetic expenditure of adult black-legged kittiwakes (Rissa tridactyla)nesting on exposed cliffs in the arctic (Gabrielsen et al. 1987) and closely approximates thevalue 2.13 xBMR for northern seabirds at rest (Birt-Friesen et al. 1989) and 1.9 xBMR forherring gulls at rest (Baudinette and Schmidt-Nielsen 1974).I estimated the BMR of a 0.9kg glaucous gull chick as 16.9 kJ/h using the adjusted Ellis(1981) allometric equation for seabirds (see above). I also assumed that glaucous gull chickshad higher RMRs than the value estimated for adults (RMR = 1 .9x BMR or 32.3 kJ/h) due theadditional requirements of growth. I calculated the energetic cost for growth (Etis; kJ/g)using the equations,E,1 = m ) +1 — m )t (6)78where etis is the energy density of body tissue of the chick (kJ/g), t is age of the chick indays, and m is mass of the chick in grams (following Klaassen 1988). The costs for thesynthesis of body tissue (E5)were calculated from the production costs (Et; kJIg) assuminga synthesis efficiency of 75% (Ricklefs 1974):E3= 1.25 (7)Energy density of tissue was estimated from the linear function of Drent et al. (1992) whichwas derived by plotting the energy density of body tissue from gull and tern chicks at a givenweight against the final asymptotic mass of each species. Thus, for a 900g glaucous gullchick at half asymptotic mass, the energy density of tissue was estimated to be 8.3 kJ/g.From the logistic growth curve of glaucous gull chicks at Coats Island (Gilchrist, chicks gained approximately 19g per day at this time. These values, when entered intothe equations above, provide an energetic estimate for growth of 9.1 U/h. When added topreviously calculated values of RMR for glaucous gull chicks (RMR l.9x BMR or 32.3 kJ/h,see above), the energetic requirement estimated for growth was 23% of the total energyexpenditure. This is close to the energetic requirements for growth of other non-passerinespecies at half asymptotic weight (as a proportion of total energy expenditure: 20-36% forherring gull chicks, Dunn 1980; 15-22% for Arctic terns, Klaassen 1988; 27-33% for long-eared owls, Wijnandts 1984). When maintenance metabolism and growth were combined, Iestimated that a 900g glaucous gull chick had a RMR of approximately 41.4 kJ/hour underneutral thermal conditions.Reviews of avian energy budget studies have emphasized the importance of evaluatingthe influence of environmental conditions on metabolic rates (Weathers et al 1984; Websterand Weathers 1988). During the night on Coats Island, the energy used by adults and chicksconsisted of resting metabolism plus additional thermoregulatory requirements. Theadditional metabolic heat production at night (Enignt kJIg.h.°C) was estimated using theequations:79Enigh: = Cn * (Tb - Tamb) - BMR (8)where Cn is the thermal conductance of the bird (kJIg.h.°C), Tb is the body temperature,and Tarn is the ambient temperature (Lasiewski et al. 1967; Weathers et al. 1984). Due to thedifferences expected in body temperature and thermal conductance between adults andchicks, I calculated thermal costs for each separately. Under conditions where ambientnightime temperature was -2°C, adult body temperature was assumed to be 39.6°C(Gabrielsen and Mehium 1984), adult mass was 1800g, and the thermal conductance was0.0004984 kJ/g.h.°C (0.0248 ml 02 /g.h.°C following the conversion 0.020 1 kJ/ml 02;Gabrielsen and Mehlum 1984), I calculate that adults required an additional 7.4 kJfh forthermal maintenance during the night (14% above daytime RMRs at thermal-neutraltemperatures). At -8°C , I calculated that adults would require 12.5kJIh above their RIvIR atthermal neutral temperatures. I assumed that chick thermal conductance was 0.0006977kJIg.h.°C (40% less efficient than adults; Klaassen 1994), and that chick body temperaturewas 41.3°C (4% higher than adults, Wijnandts 1984). At -2°C, I calculated that chicks(900g) required an additional 10.2 kJ/h for thermal maintenance during the night (24.6%above daytime RMRs) assuming that they were not brooded by their parents. This value roseto 14.1 kJ/h for chicks at Tamb= -8°C.Most gull nest ledges on Coats Island were exposed and supported no vegetation.Thus, nesting ledges provided little protection from the wind for either adults or chicks.Relatively few investigations have examined the influence of wind on convective heat lossfrom birds, and those that have, have generally concentrated on passerines in a laboratorysetting (Webster and Weathers 1988). Energetic studies under laboratory conditions haveshown that wind increases the convective heat loss from birds at air temperatures below andwithin the thermal-neutral zone, and that wind also has greater effects at lower airtemperatures (Gessaman 1972; Goldstein 1983; Webster and Weathers 1988). I applied theequations of Goldstein (1983) which predict the metabolic rate of a bird at any wind speed80below the lower critical temperature (T10 crit). These calculations occur in two steps. First,the energetic expenditure in relation to temperature is calculated,b = .0092M66 (Tamb -Ti0 Crit) .32 (9a)where b is in watts (W), M is mass in grams, and Ta and Tiowcrjt are in oC. The secondstep involves the equation,RMR = a + (b) (“Ju) (9b)where RMR is known for the temperature of interest, u is the wind velocity in mis, and a isthe Y-intercept. When the values for RMR, b , and u, are known for a given temperature, acan be calculated (Goldstein 1983). For an adult glaucous gull of mass 1715g at -2°C inwind conditions of 0.O6mls where RMR =53.6kJ?h, b =1.954, and “Ju= ‘J0.06, a will equal14.4. These values can then be substituted back into equation 8b, to calculate values forother wind speeds at -2°C (for details see, Goldstein 1983). Thus, for an adult gull at -2°Cin wind conditions of 25km/h, the RMR was estimated to be 61.2 kJ/h. Thethermoregulation requirements for chicks were calculated in the same way but I assumed thatchicks had a higher lower critical temperature than adults (Tk = 6°C rather than 2°C). Theseequations can only estimate the change in thermoregulatory costs below T,0 Theinfluence of wind within the thermal-neutral zone was estimated as 1.1 x and 1.1 5x RMR foradults and chicks respectively (Wijnandts 1984).Metabolic rate estimates are summarized in Table 4.1.Limits of chickfat storesEnergy stores are the nutrients accumulated by an animal which help it to surviveperiods of energy shortage. Energy can be stored as fat, carbohydrate, or protein (vanderMeer and Piersma 1994). However, protein is viewed as an energy store used only inemergencies because it typically constitutes the tissue necessary for normal functioning(references in vander Meer and Piersma 1994). I therefore assumed that fat was the mostcritical energy store. The maximum and minimum fat stores carried by glaucous gull chicks81Table 4.2 Estimates of Resting Metabolic Rates (RMR; kJ/h) for a glaucous gull adult(171 5g) and chick (900g) in relation to weather conditions. Weather categories refer to theconditions simulated in the model. See Methods further details.Adult ChickWeather* Day Night Day NightCalm 53.6 61.2 41.4 51.6Windy 56.0 77.9 47.6 65.2Storm 77.9 86.6 65.2 72.1* Calm < 5 km/h; Windy = 25 km/h; Storm-day = -2C, 25 km/h; Storm-night= -8C, 25km/h.82were estimated based upon the fat stores found in other gull species (i.e. as a proportion oftotal body mass). The fat stores of adult lesser black-backed gulls (Larusfuscus) rangedfrom 1-12% of total body mass (Sibly et al. 1987). Similarly, the range in weights ofherring gulls over a year was approximately 9% of upper body mass (Coulson et al. 1983).Herring gulls in northern England maintained fat stores which ranged between 62-lO6g or 6-11% of body mass (Sibly and McCleery 1985). Based upon these findings, I estimated thatthe upper limit of fat stores that a gull chick could carry (x,) was 15% of total body mass(135g). This proportion is higher than for adult gulls because chicks are not constrained bythe low weight requirements of flight. In the model, a three chick brood with an averagemass of 900g could thus carry a maximum of 405g of fat or 14985 Id of energy given anassimilation efficiency of 37.7-39.7 kJg of lipid reserve (Johnston 1970). The lowerminimum fat store (Xmin) was estimated as 3% of body mass, or 342 U per chick. I madethe simplifying assumption that fat stores were depleted linearly during fasting; this is nearlythe case among fasting herring gulls (Spaans 1971). Thus, a three chick brood that entered afasting period with maximum stores (xmax) could survive for approximately 4.5 days undercalm thermal-neutral conditions given my estimates of RMR (Table 4.1).Flight energeticsThe cost of level flapping flight was estimated using the equation provided by Pennycuick(1989):Cf= 1.1 [l.2(M *g)/2(spv)+o.5pv SbCdb+l.2Pamj+E+Pjb} (10)where M is body mass in kilograms (=1.75 1 ± 0.23 kg , n =9 birds), g is acceleration dueto gravity (9.8 mis2), Sd = wind disc area in square meters, p = air density at sea level, v =airspeed (mis), Sb body frontal area (in square meters), Cdb = drag coefficient of the body,Pam is the absolute minimum power (JIs), E = metabolic efficiency, 0.23; Pennycuick 1989),and jb = basal metabolic rate (3.73M°723). The cost of flight in mechanical units (W) wasthen converted to its metabolic equivalent (kJIh). A 1715g glaucous gull flying at its83maximum range speed (the velocity at which maximum distance is traveled per unit power,Pennycuick 1989) using flapping flight predicted an energy expenditure of 29.6 W or 106.6kJ/h under calm wind conditions (Pennycuick 1989). This estimate closely approximates the108 kJ/h predicted from the equation,logy =l.80+0.679logx (11)which was derived from activity-specific doubly labeled water studies of six northern seabirdspecies foraging at sea (Birt-Friesen et al. 1989). The minimum power velocity (Vmp=thevelocity at which flapping flight is most energy efficient) was calculated as 23.5 W or 84.6kJIh given the above morphology.I estimated that gliding required approximately 2.1 xBMR, which is roughly equivalentto estimates of power required to support the body mass of birds on outstretched wings(Pennycuick 1989). This multiple of BMR is similar to 2.4 xBMR measured for herring gullsgliding in a wind tunnel (Baudinette and Schmidt-Nielsen 1974) and for free-ranging herringgulls whose heart rates when gliding were 1 .3x their heart rates at rest (i.e. RMR; Kanwisheret al. 1978). This provides and estimate of 64.4 kJ/h when gliding for a 1715g glaucousgull. The energetic cost of hovering was estimated as lix BMR for a gull hovering undercalm wind conditions.Energy consumptionI estimated the rate of gross energetic gain ingested (GET) for each foraging mode bymonitoring the rate at which individually marked birds obtained food items (e.g. eggs, fish).Thick-billed murre eggs at Coats Island average iOOg wet weight (deForest 1993). Most eggcontents are ingested by glaucous gulls once the shell had been cracked with the bill. Iestimated, however, that gulls spilled approximately 1 5g from most eggs. The energycontent of fresh common murre eggs (Uria aalge) is 9.08 kJg-1 wet weight (Spear 1992)which provides 771.8 kJ. I applied this same value to thick-billed murre eggs and thenassumed a digestive assimilation efficiency of 0.85 for fresh eggs (Dunn 1976; Spear 1993;84Young 1994). Gulls also scavenged discarded fish (mean mass 11 g wet weight) andabandoned murre eggs at the colony, and I estimate that scavenged fish yielded a grossenergetic value of approximately 4.4 kJg (Dunn 1976; Fumess 1981; Spear 1993; Young1994). I assume a lower digestive assimilation efficiency of 0.70 for scavenged material(Young 1994).Foraging mode energetics and risk ofcontact with murresThe hourly energetics of each foraging mode i, reflected the number and characteristicsof the attacks that occurred, their probability of success (A), the metabolizable energy gained(/3), and the energetic costs of the search and attack activities for the adult during that time(q). As discussed below, these values often varied with wind conditions.Adult gulls often loitered on or near their nesting ledges and were assumed to loseenergy at their daytime RMR of 53.6 kJh and 56 kJ/h during calm and windy conditionsrespectively (Table 4.1). When adults were inactive, chicks were not provisioned and werealso assumed to lose energy at their daytime RMR (Table 4.1).The energetic costs of the stand foraging mode included both the costs of standingactivity during search and attack, and the short exploratory flights between murre nestingledges (Table 4.2). Kanwisher et al. (1978) found that the heart rates of herring gullsincreased by approximately 20% during aggressive interactions between conspecifics. Asimilar physiological response likely occurs among glaucous gulls attacking thick-billedmurres. Thus, I assumed that persistent attacks made on foot required 1.2 xRMR. Duringeach hour of stand foraging, glaucous gulls often visited several murre ledges. The flightsbetween ledges amounted to approximately 10 mm of flight each hour under windyconditions (Table 4.2). I combined the estimated costs of both standing (53.6 kJfh) andflight (84.6 kJ/h calm, flapping; 64.9 kJIh windy, gliding) and integrated these with the timeallocated to each of these activities per hour. This provided an overall energetic estimate of55.6 kJIh and 58.1 kJ/h under calm and windy conditions respectively. While stand85Table 4.3 Proportion of foraging time devoted to flight in relation to foraging mode andwind conditions (a) calm <5 km/h, b)> 10 km/h).% Total foraging timeForaging mode Flapping Gliding Hoveringx (SE) x (SE) x (SE)a) Standforaging 6 (2.1) 2 (0.2) 0 (0.0)Scavengeatcolony 14 (3.6) 4 (0.6) 2 (0.3)Flapping flight mode 85 (8.5) 8 (1.7) 6 (1.3)Gliding flight mode# -- -- -- -- --b) Stand foraging 4 (1.3) 10 (2.1) 0 (0.0)Scavenge atcolony 3 (0.7) 6 (1.1) 3 (1.9)Flapping flight mode* -- -- -- -- -- --Gliding flight mode 5 (0.2) 91 (5.6) 4 (2.0)# Gulls could not patrol the colony using gliding flight under calm conditions* Gulls rarely used flapping flight during windy conditions at the colony86foraging, marked individuals had a mean attack rate of approximately 4 attacks/h (± 2.6;8.2h). For gulls in the general population, each stand attack had a 46% chance of success(ch. 3), so that for each hour of stand foraging activity, gulls had approximately a 1.8 chanceof successfully obtaining a single egg (i.e. 0.9 probability of obtaining two items). Gullshad a 0.32 probability of being severely struck by defending murres during each attack sothat gulls were typically struck on average 1.3 times each hour while foraging (Table 4.4).Wind conditions did not influence these probabilities.The energetic costs of scavenging on foot included foraging on foot and short flightsbetween cliff ledges, and I estimate an energetic cost of 67.7 kJ/h and 62.8 kJ/h during calmand windy conditions respectively. This is similar to the energy requirements of the standforaging mode minus the costs of attack. However, scavenging gulls took more short flightswithin the colony (Table 4.2). At Coats Island, scavenging held a 0.6 probability of a gullfinding a single food item during one hour of foraging activity (n=4.8 h). The ratio of fooditems obtained while scavenging by the general gull population was approximately 8 fish foreach murre egg ingested. This yielded a mean energetic gain of 61 U/item when scavenging.However, scavenging held little risk from murre contact because most searching activityoccurred at the base of cliffs or on large ledges free of nesting murres where colony debriscollected. Wind did not alter these probabilities for scavenging gulls.The energetic costs of the flapping-flight foraging mode included the flight costs ofaerial search patrols and of hovering aerial attack. I assumed that gulls were flying at theirminimum power velocity rather than their maximum range velocity when foraging at thecolony. A slow and energy-efficient flight speed is expected for birds searching forconcealed prey in a localized area (Gendron and Staddon 1983), and it was obvious thatpatrolling glaucous gulls flew more slowly than gulls flying to their nest or out to sea. Theffight energetics model of Pennycuick (1989) estimated an energetic expenditure of 84.6kJfhour for a 1715g glaucous gull flying at its minimum power velocity under calm windconditions. Aerial attacks under calm conditions required that the gull hovered next to theTable4.4Parameterestimatesofglaucousgullforagingmodeenergeticsanddangerinrelationtowindconditions(a)calm;b)windy;c)sourceofestimate.Murrecontactreferstotheprobabilityofagullbeingseriouslystruckbymurresduringonehourofforagingactivity.Foragingsuccess(A)Grosenergeticgain(j3)Energeticcost(q,)MurrecontactSurvival()Foragingmode(i)(Prob.oftwoitems/h)(kJ/fooditem*DAE5)(kJ/h)(Prob.ofcontlh)(prob./hour)a)Standforaging0.80652e55.61.30.9990Scavengeatcolony0.4061s62.70.20.9999Flappingflight0.55652e89.80.90.9995Glidingflight--------Triptosea0.45171f108.0#0.01.0000b)Standforaging0.80652e58.11.30.9990Scavengeatcolony0.4061s62.80.20.9999Flappingflight----Glidingflight0.40652e67.20.50.9997Triptosea0.45171f151.2#0.01.0000c)Source’FDLIFD,LIFDVAR$DAE=DigestiveAssimilationEfficiency(e=murreegg;s=scavengedmaterial;f=arcticcod)#Estimatedbaseduponallometricequationsofnorthernseabirdsforagingatsea(Birt-Friesenetal.1990)@FD=Fielddata;U=Literaure;VAR=Variableinthemodel88murre nest site. These hovering attacks were rarely maintained for more than six secondsunder calm conditions (ch. 3). Individually marked gulls had a mean attack rate of 6attacks/h (± 3.4; n=2.67h) using flapping-ifight under calm conditions. In the general gullpopulation, flapping ifight attacks had a 0.19 probability of success and a 0.16 probabilitythat the gull was struck by defending murres (w=572). Thus, for each hour of flapping flightforaging activity, gulls had a 0.55 probability of successfully obtaining two food items and a0.9 probability of serious contact with defending murres (Table 4.3). Under windyconditions, gulls rarely used the flapping flight mode and I therefore had insufficient data toestimate these values for windy conditions (Table 4.3).The energetic costs of the gliding-flight foraging mode under windy conditionsincluded the flight costs of search patrols and of aerial attack. However, under windyconditions, gulls glided almost continuously, even when attacking murres. I estimated thatgliding foraging required 64.4 kJIh for a 1715g gull (2.1 xBMR). Individual gulls had amean attack rate of 11 attacks/h (± 5.6; n=6.7 foraging hours). Each attack had a 0.08probability of success and a 0.049 probability of being struck by defending murres(n=1551). Thus, for each hour of gliding flight foraging activity, gulls had a 0.85probability of success and a .054 probability of contact with defending murres (Table 4.3).Gulls could not glide extensively when patrolling under calm wind conditions due to theconstraints of flight dynamics (Pennycuick 1989; ch. 3), so gliding was excluded from themodel as a foraging option for gulls during calm conditions (Table 4.3).The energetic gains and costs of glaucous gulls foraging at sea are uncertain. Little isknown about the distribution of fish at sea around Coats Island or how the availability of fishfor gulls varies with season, ice, or weather conditions. Based upon observations ofregurgitations to chicks and of crop contents, gulls foraged primarily on Arctic cod(Boreogadus saida ) when at sea. Foraging trips averaged three hours away from thecolony(± 1.3; 42 trips; n=8 birds). I used equation 10, to obtain an estimate of 108 kJ/hunder calm wind conditions for a 171 5g gull. During windy conditions, the cost of foraging89at sea was 40% higher for black-legged kittiwakes in the Arctic (Gabrielsen et al. 1987) and Iuse this value for glaucous gulls. Young (1994) found that the foraging success of skuas atsea in Antarctica was not influenced by strong gales. Similarly, I found that glaucous gullsreturned consistently with large fish loads regardless of weather conditions at sea. I thereforeassume that the probability of a gull finding fish during a trip to sea was 0.9 regardless ofwind. The prey load obtained from a trip to sea was estimated as 137 grams of Arctic cod.This load is approximately 8% of gull body mass which is consistent with fish loads of otherlarge gull and skua species (Furness and Hislop 1981; Spear 1993; Young 1994). Thisyields a gross energetic gain of 171.1 id/h for a three hour foraging trip assuming that thegull was successful (Table 4.3).Modeling risk offatal injuryIn this model, the probability of contact with murres while foraging (Table 4.3) wasused as an index by which foraging modes could be ranked in terms of danger. Thus, thestand foraging mode held the highest degree of risk of fatal injury; scavenging and foragingat sea the least (Table 4.3). Tn the model, the parameter c was varied so that adult foragingmode selection could be estimated in the presence or absence of risk of fatal injury. Thus,when the model was run with no risk (i.e. x=1), gulls were assumed to select foragingmodes that simply maximized the rate of net energetic gain. As risk increased (CL<l), themodel selected the foraging mode which maximized expected fitness by maximizing thetrade-off between the survival of the current brood (via provisioning) and the adult.90RESULTS AND DISCUSSIONForaging mode selection with no risk of injuryTo model foraging mode selection under no risk of injury, o was set to one. Thus, theoptimal foraging mode was selected based upon energy considerations alone. Under calmwind conditions, the model selected the stand foraging mode which provided the mostreliable and sizable energetic gains per hour. Alternative foraging modes such as flappingflight, foraging at sea, and scavenging were not selected (Fig. 4.2b).Under windy conditions with no risk of injury, the gull also selected the stand foragingmode only (Fig. 4.3a). Other potential foraging modes which included gliding, scavengingand trips to sea were not selected. The model predicted that foraging activity under windyconditions should occur at chick energetic states higher than predicted for calm conditions.In other words, gulls would be expected to forage more and maintain their broods at higherenergetic states under windy conditions, simply because the costs of thermoregulation arehigher under windy conditions (Table 4.2).To summarize, if all foraging modes held little risk of injury so that foraging decisionswere based upon energy considerations alone (c=l), gulls should select the stand foragingmode regardless of wind conditions (Fig. 4.2b, 4.3b). Thus, foraging mode should not varywith changing wind conditions, even though energetic demands are higher during windyconditions (Table 4.2). However, most gulls did vary their foraging mode selection as windconditions changed (ch. 3). Thus, changes in gull foraging patterns observed under varyingwind conditions were not simply the response of gulls to changing energetic demandsimposed by weather.Foraging mode selection with risk of injuryThick-billed murres aggressively defend themselves against gull attack, and although Inever witnessed a fatality of an adult gull, it was not uncommon to observe gulls withbloodied faces and legs following aggressive interactions with murres. Further,91Figure 4.2 a, b - Gull foraging mode selection under calm wind conditions in relation totime, brood energy reserves, and risk of injury (a) no risk, b) risk). Hyphenated strategiesrefer to mixed foraging strategies. DO NOTHING = Inactivity; SCAV=Scavenge; SEA=Forageat sea; STAND = Forage on foot; FL = Foraging using flapping flight.nightC’)ci)ci)Cl)ci)________________>c,)ci)C)-cC-)DO NOTHING /SCAV - DO NOTHINGRISKNO RISKx maxx mmx maxx mmSTANDa)DO NOTHINGSCAV - DO NOTHINGSEAFLSTANDb)DO NOTHINGSTANDJ.TTime (hours)92IxminFigure 4.3 a, b - Gull foraging mode selection under windy conditions in relation to time,brood energy reserves, and risk of injury (a) no risk, b) risk). Hyphenated strategies refer tomixed foraging strategies. DO NOTHING = Inactivity; SCAV=Scavenge; SEA=Forage at sea;STAND = Forage on foot at colony; FL = Foraging using flapping flight at colony; GLIDE =Glide foraging at colony.xmax RISKNO RISKx maxx mmTTime (hours)93several fully-grown gull chicks were blinded and killed when they fell into densely occupiedmurre nesting ledges (Donaldson, pers. comm.), so the potential of fatal injury exists. Thismodel was built to explore whether this potential was high enough among adult gulls toinfluence their foraging activity and provisioning strategies.To simulate foraging mode selection under risk of injury, the probability that gullswould survive an attack (x) was lowered to reflect the relative risks of contact associatedwith each foraging mode (see Methods). Therefore, the optimal foraging mode was selectedbased upon both the energetic gain for the brood, and potential risk of injury for the adultwhile foraging. This generated predictions that were in sharp contrast to the foragingdecisions made under energy considerations alone.Under calm wind conditions with danger, the model predicted that gulls should select avariety of foraging modes depending on the energetic state of their brood (Fig. 4.2a). Forbroods with high energy stores during calm conditions, gulls should do nothing or scavengeat the colony (Table 4.4). At brood energy stores below 50% of maximum reserves (Xmax),the model predicted that gulls should leave the murre colony to forage at sea. However,scavenging and trips to sea both yield lower net energetic gains compared to flapping-flightor stand foraging at the colony (Table 4.4).Under windy conditions, the model predicted that gulls should switch to the glidingflight foraging mode for brood states 8% xmax <x(t) <65% Xmax (Fig. 4.2a). Trips to seaand scavenging are not selected. Under low brood energy stores (x(t) <8% xmax), the standforaging mode is selected. However, this should be rare under windy conditions, becausegliding-flight yields energetic gains sufficient to meet the energetic demands of the brood(compare Tables 4.2 and 4.4).In summary, the model predicts that gulls foraging under risk of injury should notselect modes which maximize rates of net energetic gain unless brood energy stores fall tonear-critical levels. This prediction holds for both wind conditions and is in sharp contrast tothe provisioning strategies predicted for gulls foraging under low risk of injury (see above).94Comparisons with field data ofguilforaging behaviourThe model generated two distinct foraging strategies based upon the degree of dangerassociated with foraging mode decisions. Further, these two strategies were marked bystrong differences in the effects that wind had on them. Under low risk of injury, the modelpredicted that gulls should select the stand foraging mode regardless of wind conditions. Incontrast, gulls foraging under risk of injury are predicted to spend a greater amount of time atsea or fasting under calm conditions, and much of their time glide-foraging during windyconditions. How do these predictions compare with observations of glaucous gulls foragingin the wild?To address this question, I studied the affects that sudden wind change had on foragingmode selection of gulls in the wild. In the context of the dynamic model, gulls move fromone decision array to another following a wind change (Figs. 4.2 a,b and 4.3 a,b). Forexample, if wind increased, the model predicted that gulls foraging under risk should switchsynchronously from loitering, scavenging, or foraging at sea to the gliding flight mode iftheir broods fell below 60% of Xmax (Fig. 4.2a to 4.3a). Similarly, if gulls were foragingusing gliding flight and the wind dropped below 10 km/h, a synchronous switch toinactivity, scavenging, or absenteeism from the colony is expected. In contrast, the modelpredicts that gulls foraging under low danger should not switch foraging modes with a windchange, because stand foraging is always preferred (Fig. 4.2b and 4.3b).On 6 occasions during the years of this study, my assistants and I monitored gullforaging activity continuously for several days in relation to wind conditions. As predicted,both aerial search and attack activity increased immediately when wind conditions increasedfollowing extended periods of calm (ch. 3). During calm conditions, attacks made on foot orfrom flight were initially infrequent. As calm conditions persisted beyond 14 hours, attackfrequency using flapping flight or stand foraging modes gradually increased. The modelsuggested that this gradual increase in foraging activity would be expected if parents were95responding to the worsening energetic states of their chicks. To test this prediction morerigorously requires that adult foraging decisions, chick energetic state, and wind conditionsare monitored simultaneously. This is a challenging undertaking because it requires that theadults that are captured and marked also nest on accessible cliff ledges. More importantly, itrequires that the energetic states of chicks are monitored on a fine time scale in a way thatminimizes disturbance to both gull chicks and the murres nesting around them. This workwas beyond the scope of this thesis.Although the foraging behaviour of the general gull population was similar to thatpredicted for gulls foraging under danger, marked individual variation existed among gulls(Figs. 4.4, B l-B6). The majority of marked birds (6 of 8) exhibited a diverse foraging-mode repertoire which included inactivity, scavenging, trips to sea, and both stand and aerialforaging. Their attack mode selection also varied with wind conditions, so that aerial attackspredominated under windy conditions. In contrast, 2 of 8 birds showed a consistentpreference for stand attack modes regardless of wind conditions (Fig. 4.4 B 1, B2). Iobserved that these two gulls attacked murres by pulling them off their nest sites by the wingor tail, and they usually did this without being struck by murres. Thus, these gulls used anattack technique which apparently decreased the risks associated with foraging on foot.As predicted by the model, the time allocated to foraging among such foot specialistsdid not vary significantly with wind conditions (Fig. 4.5), while aerial specialists spentsignificantly more time away from the colony under calm conditions, and significantly moretime foraging on the wing at the colony under windy conditions (Fig. 4.5). These patterns offoraging behaviour are in general qualitative agreement to the two foraging mode patternspredicted by the model.Thus, individual variation observed within this gull population could be partlyexplained by the differences in how individuals respond to the risk of injury while attackingmurres. Among avian predators breeding at seabird colonies, some individuals commonlytravel further for small energetic rewards compared to the resources available to them at96Figure 4.4 Attack mode selection of 6 glaucous gulls (Bird 1 - Bird 6), in relation towind conditions (calm <5 km/h; windy> 10 km/h). Attack modes include: 1- standforaging; 2 - aerial lunge; 3 - pull murre off site while standing; 4 - drop into murres fromabove; 5 - ascend aerially from the ocean and lunge into murres; 6 - hover harass from flight;7 - descend aerially from the ocean and lunge into murres; 8 - pull murre off site from flight;9 - circle and hover harass from flight; 10- descend from above and harass. Note that birds,B 1 and B2 attacked murres using stand foraging modes regardless of wind conditions.97100’ — 100’Bi CALM B1 WIND75’ 75.50 50’25 25’0’ —0’ —________________________1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10100’ 100’B2 CALM B2 WINDC.)( 75. 75.50’ 50’025’ 25’0.00’ —. 0• — .1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10100’ 100’83 CALM 83 WIND75. 75.50’ 50’25’ 25’ 70’ - .. r,fl 0’1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10Foraging mode98100 100’-B4CALM B4WIND75 75.50 50’25-.fl on... -___1 2 345 6 78910 1 2 3 4 5 6 7 8 910100• 100-86 CALM B6WIND75. 75.50- 50’C0‘€ 25- 25-00. ri El_____________ __________0-— . 0-1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10100- 10085 CALM B5 WIND75 7550 50’25- 250- — . — r—iF1r,1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10Foraging mode99Figure 4.5 Gull time budgets in relation to wind conditions and foraging modespecialization. (*) refers to a significant difference in the time devoted to an activity underwindy and calm conditions (t-test, paired comparisons, P<O.05).AERIALTrSTAND80 60 40 2080a) E 4- 0) -o ‘6 C 0 0 0. 2 a-T*-rDCALMWIND0’60 40 20 0Tflra..z=1ABSENTLOllERSTANDFORAERIAL.FORBehavioural categoriesJ]jABSENTLOllERSTANDFORAERIALFORBehavioural categoriesC C101colonies (e.g. Southern and Southern 1984; Watanuki 1989, 1991; Spear 1992; Young1994). This again suggests that the economics of alternative foraging behaviours differbetween individuals. For some individuals, high risk of injury while foraging at the colonycould increase the relative benefits of foraging elsewhere. A danger-reward trade-off mayalso explain why large males are typically the individuals who specialize on seabird prey(Parsons 1971; Watanuki 1989; Watanuki 1991; Spear 1993). Being large may dilute therisk of injury during attack. However, there is also likely a correlation between low risk ofinjury, killing ability, and the ability to ingest large seabird prey; all of which may increasewith bill size (Harris and Jones 1969; Watanuki 1989). Although risk of injury is just onefactor influencing foraging strategies of avian predators, my findings suggest that itcontributes to the behavioural variation observed among individuals foraging in the wild.Further predictionsWhy do most gulls at Coats Island forage under the constraints of murre defence andcalm wind conditions while a minority appear to be able to obtain murre eggs and chicks atwill? A similar dichotomy exists in other avian predator populations (Watanuki 1991; Spear1992; Young 1994). Perhaps the reward-danger ratio of foraging modes change asindividuals improve their foraging performance with experience. For example, learning thesubtleties of attack could help glaucous gulls overcome murre defenses more efficiently, andthis would enhance the profitability of the more productive but risky modes (e.g. foraging onfoot). The time and experience probably required to learn how to overcome prey defencesprobably ensures that individuals within a long-lived population differ in their foragingabilities at any given time (Fig. 4.4). For example, it is likely that young glaucous gullsforage under greater risk than adults when preying on murres. If this is true, the modelpredicts that young birds should use a variety of foraging modes under both calm and windyconditions (Fig. 4. la, 4.2a). Thus, young birds should use modes which carry low risk ofinjury and low energetic gains (e.g. scavenging, foraging at sea, gliding flight). If learning102occurs, foraging mode repertoire of individuals should shrink with age as they become moreadept at attacking murres. In fact, the oldest and most experienced gulls should foragealmost exclusively on foot (Fig. 2a, 2b), and consequently, escape the constraint of calmwind conditions (ch. 2, 3). Therefore, learning to improve foraging efficiency could enableglaucous gulls to increase their reproductive success without necessarily increasing the costsof reproduction. An improvement in foraging efficiency has been invoked to explain howolder gulls can achieve higher reproductive success without increasing their energeticexpenditure (Reid 1988; Pugesek and Wood, 1992; Pugesek 1995).My model also generated predictions on how gulls might differ in their foragingsuccess and mode selection at colonies with lower inherent risks. At Coats Island during theyears of this study, the colony was increasing in size so that murres were nesting at very highdensities. Indeed, most cliff ledges were completely occupied by brooding murres and therisk of contact with murres when foraging on foot was high (ch. 2, 3). If a murre colonywere to decline suddenly following a perturbation (e.g. oil spill, gill-netting by-catch, or anera of over-hunting), murre nesting densities on ledges and/or the proportion of ledge areaoccupied by brooding murres would decline (Birkhead 1977). Under both circumstances,gulls should overcome murre defences more easily (Spear 1992; Siegel-Causey and Hunt1981; ch. 2, 3). If this occurred, the model predicts that a higher proportion of gulls wouldforage on foot and thereby overcome the constraints of calm wind conditions, colonytopography, and murre defence (ch. 2, 3). Consequently, the proportion of murre eggs andyoung lost to avian predation should increase during and following colony declines. Iexplore these issues further in chapter 5.ConclusionsMy aim in this chapter was to integrate field data and energetic estimates of gullforaging behaviour to explore the influence that mortality risks have on the foraging andprovisioning strategies of glaucous gulls. Previous field studies indicated that gull foraging103activity, foraging mode selection, and provisioning are strongly affected by wind (ch. 2, 3).There are at least two possible explanations for this. First, an increase in foraging activityduring windy conditions may occur because gulls require more energy to maintainthemselves and their young due to higher thermoregulatory costs. Alternatively, a trade-offmay exist between the risk of injury and energetic gain while foraging, and perhaps windyconditions improve the economics of this trade-off in favor of gulls.The model suggests that the first explanation, which is based upon energyconsiderations alone, is not sufficient to explain the observed patterns of foraging activity.However, the model supports the idea that glaucous gulls are very sensitive to risk of injurywhile foraging. Gulls at Coats Island face higher risks of contact with murres when foragingunder calm wind conditions (ch. 2, 3), and this dynamic model predicts that under calmconditions, gulls should select low-danger modes that provide low to moderate energeticgains (e.g. do nothing, scavenge, forage at sea). A general feature of the results is that gullscan afford to use foraging modes that yield low energetic gains relative to more productiveand dangerous ones, because even poor foraging modes are sufficient to meet their energeticdemands under most circumstances. In addition, poor foraging conditions in the wild (i.e.calm wind conditions) are usually short-term in relation to the energy stores that are typicallymaintained by large gulls (Spaans 1971; Coulson et al. 1883; Sibly and McCleery 1985).104CHAPTER 5EFFECTS OF GLAUCOUS GULL PREDATION AT DECLINING THICK-BILLEDMLJRRE COLONIES: AN INTER-COLONY COMPARISONMany prey species have evolved adaptations which decrease their profitability as prey byincreasing the risks of injury for the predators attacking them (e.g. toxicity, defensiveweapons, aggressive group defense; Edmunds 1974; Endler 1986; Sih 1989). Thus, the valueof a prey should increase for a predator as the risk of injury for the predator declines.Consequently, if the effectiveness of a prey’s defence depends on large group size, (e.g.dolphins, Norris and Dohl 1980; muskox, Gray 1974), predators may have an increasingimpact on prey numbers as prey populations decline.Such predator-prey dynamics may apply to colonial-nesting seabirds that breed inassociation with avian predators. The effectiveness of group defense often increases withnesting density (Hatch 1970; Birkhead 1977; Gotmark and Andersson 1984), and avianpredators respond to the dangers generated by these defensive behaviours (Spear and Anderson1989; Spear 1993; Young 1994; Thiel and Sommer 1994). For example, gulls (Larus sp.) arehighly selective when attacking the nest sites of cliff-nesting murres (Uria sp.). They avoiddense nesting aggregations and preferentially attack birds nesting alone or on the edge ofgroups, apparently because they can forage without being struck by defending neighbours(Birkhead 1977; Spear 1993; ch. 2, 3). Because most avian predators do not attack seabirdsnesting at high densities, most of a breeding colony is typically safe from predation (Spear1993; Young 1994). Consequently, avian predation has only a minor impact on seabirdreproduction at most colonies (Siegel-Causey and Hunt 1981; Spear 1993; Young 1994; ch. 2,3). If nesting densities of seabirds declined however, predation should have a greater impacton seabird reproduction because avian predators could overcome the constraint of group preydefence (ch. 2, 3). This prediction has conservation implications for the recovery of heavilyharvested. seabird colonies, or those reduced in size by natural catastrophes.105To examine the relative impacts of avian predation at increasing and declining seabirdcolonies, I compared glaucous-gull (Larus hyperboreus) predation at thick-billed murre (Urialomvia) colonies in the Canadian and Greenland Arctic. Glaucous gulls are the primary avianpredator of thick-billed murre eggs and young in the Arctic. I compared gull predation rates,foraging behaviour, and thick-billed murre nest site selection at an expanding colony (CoatsIsland, Canada) with declining colonies in the Upernavik region of Greenland. Thesecomparisons tested: 1) if gulls were more successful when stealing eggs and young at decliningcolonies, 2) if murre nesting densities and group structure on breeding ledges were altered ascolonies declined, and 3) how/if the number of breeding gulls changed at declining colonies.METHODSStudy sitesI worked at two field sites. The first was at the north eastern tip of Coats Island,Northwest Territories, Canada (1990-1992). See chapter 1 and 2 for further details. Thesecond site was in the southern Upernavik region of Greenland at the Kingittoq (5625 breedingpairs) and Timmiak (175 breeding pairs) colonies. I carried out behavioural observations ofgulls and murres at Kingittoq, and visited Timmiak once to census the colony. A commercialharvest of murres by the community of Upernavik in the 1980’s substantially reduced themurre populations at these colonies (Evans 1984; Kampp et al. 1994). Although thiscommercial hunt ceased in 1988, these colonies have not recovered and hunting for personaluse continues in the early spring (Falk and Durinck 1992; pers. obs.).At both study sites, weather conditions were generally moderate during my studies andwind conditions were often below 10 km/h (Fig. 5.1), and periods of calm conditions (>2days), commonly occurred.Figure 5.1 Wind conditions at Coats Island (1990-1992) and Kingittoq (1993)colonies.106King ittoqa)E4-04-.ct5G)Co-00J5040302010•0-5040302010•0•I0-4 5-9 10-14 15-19 20-24 >25Coats0-4 5-9 10-14 15-19 20-24 >25Wind speed (km/h)107Gullforaging behaviour and predation ratesBehavioural observations of gull-murre interactions were carried out daily fromconcealment from blinds at Coats Island (1990-92) and from behind rocks at Kingittoq (1993).For each gull attack, I recorded the following: 1) gull foraging mode (aerial flapping flight,aerial gliding ifight, stand foraging); 2) murre nest site characteristics (narrow ledge highnesting density, narrow ledge low density, broad ledge high density, broad ledge low density,and cave/crevice), and 3) outcome of the attack (egg/chick taken or not). See chapters 2 and 3for definitions of nest sites and foraging modes, and other details relating to methodology.Wind conditions on the cliff were monitored with a computerized Digitar anemometer at Coats,and a hand-held anemometer at Kingittoq.I used analysis of covariance to explore factors affecting foraging mode selection by gullsat Coats and Kingittoq colonies. Colony location (Coats vs. Kingittoq), wind conditions(continuous variable), and the interaction colony x wind were included in the model. Ianalyzed the stand foraging mode only because stand and aerial proportional data were notindependent (i.e. they both combined to make 100%).I used analysis of covariance to examine factors influencing gull attack activity. Colonyidentity, ledge width, wind conditions, colony x ledge width, and colony x wind were includedin the model. All proportional and rate data were arcsine transformed prior to analysis.The population-level response ofglaucous gulls to murre population declinesPredator foraging efficiency and reproductive success may fall as a predator’s primaryprey decline (Sinclair 1989). Consequently, the impact of glaucous gull predation shouldremain constant regardless of colony state assuming that the numbers of gulls present at murrecolonies declined in parallel. However, it is more likely that gull numbers would decline moreslowly than those of murres, particularly if colony declines increased the vulnerability of murrenest sites to predation. If this were true, a higher ratio of gulls to murres would be predicted atdeclining colonies over the short term.108I studied the numerical relationship between glaucous gull numbers and murre colonysize using photographs of the number of gulls present at Coats and the Upernavik colonies,and using information of gull and murre numbers in the literature (App. I). I used analysis ofcovariance to examine the numerical relationship between gull and murre numbers in relation tocolony state (stable at Coats vs. declining at Upernavik). Colony state, murre numbers, andthe interaction between colony state and murre numbers were included in the model. Numbersof gulls and murres were log-transformed prior to analysis.Murre nest site selection and breeding densityI photographed each colony from the sea during mid July and then counted the murreson the cliff face from black and white prints. The numbers of birds present on the cliff weremultiplied by a correction factor of .75 to estimate the number of breeding pairs (followingNettleship 1976). From these photographs, I also quantified the proportion of murres breedingon the various nest types. For definitions and characteristics of murre nest site characteristics,see ch. 2.I examined how the magnitude of colony decline, nest site characteristics, and theinteraction between colony state influenced the use of each type of nest site using ANCOVA (i.e.the proportion of each colony nesting on each nest type). Because only 4 of the 5 proportionscould be considered independent, I only included 3 of the 5 characteristics in the model tomaintain independence (cave, broad high density, and narrow high density).109RESULTSGullforaging mode selection, attack activity, andpredation ratesGulls at Coats and Kingittoq used different foraging modes (Fig. 5.2 a, b; ANCOVA, F6 = 11.6, P 0.014). Gulls at Kingittoq frequently attacked murres on broad ledges on footeven under windy conditions (Fig. 5.2b). However, the proportion of aerial attacks by gullsat Kingittoq increased with wind speed (5.2a), which implies that gulls still benefited fromforaging on the wing. In contrast, gulls at Coats Island mostly attacked from the air evenduring calm conditions (Fig. 5.2). Therefore, the greatest differences in foraging modesbetween gulls at the two colonies occurred at low wind speeds (<10 km/h). The influence ofwind, however, was not quite significant (F 1,6 5.5, P = 0.057).Gull attacks were more frequent at Kingittoq colony, particularly under calm conditions(Fig. 5.3a). Gulls at Kingittoq attacked murres on broader ledges than gulls at Coats Island atall wind speeds. At Coats Island, broad ledges were densely occupied by murres and therewere few opportunities for gulls to forage on foot (ch. 3). Instead, gulls at Coats relied heavilyon aerial attacks (Fig. 5.2). Consequently, gull attack activity and foraging success increasedwith wind speed at Coats Island (Fig. 5.3) when windy conditions increased their aerialmaneuverability and enabled them to attack weakly defended narrow ledges (ch. 3). Thus,attack activity was significantly influenced by wind (ANCOVA,F155 = 16.15, P < 0.001),ledge width (F 1,3455 16.28, P <0.001), and colony identity (F 13455 = 13.59, P =0.001). There was also a significant interaction between colony identity and ledge (F =6.88, P =009), which supports the idea that gulls attacked different ledge types at the twocolonies. The idea that risk of contact with murres varied between the two colonies is alsosupported by the observation that gulls at Kingittoq ingested eggs on the nesting ledges, whilegulls at Coats typically carried murre eggs to un-occupied ledges above the colony. Tosummarize, attack activity at Kingittoq was consistently higher than at Coats, particularly undercalm wind conditions when gulls at Kingittoq could forage on foot on broad ledges.110Figure 5.2 Attack mode selection of glaucous gulls in relation to wind conditions and colony(a) Aerial foraging, b) Stand foraging).100a) - .- 0 .- - --80 ..-- 0- 0- -60.$...40• /1•20 -----0” Coats Island* KingittoqC)0- I • I0 10 20 30D100b)C8060-•fr40-—..0.20- -...-...--..- 0’ - • -.•00- I • I0 10 20 30Wind speed (km/h)U)E000-c-cCl)-C-)-I-.4-a5>>C)-C)4—4-c111Figure 5.3 Glaucous gull attack activity in relation to cliff ledge width (a) broad, b) narrow),wind speed, and colony (i.e. Coats Island, Canada, and Kingittoq, Greenland).Values are off-set for clarity. 10 20 300 10 20 30Wind speed (km/h)112Predation rates at Kingittoq were consistently higher than at Coats Island although theywere more similar under higher wind speeds (Fig. 5.4). This is expected because gull foragingbehaviour and attack rates were most similar at higher wind conditions (see above). Thus, thegreatest differences in predation rates between the two colonies occurred at low wind speeds,and predation rates were significantly influenced by wind (ANCOVA, F1 1436 = 17.82, P =0.001). Although mean predation rates were more than twice as high at Kingittoq than at Coats(Fig. 5.4), no significant colony effect was detected (F 11436 = 2.45, P =0.13).In summary, gulls at Kingittoq had consistently higher attack rates and were lessconstrained by calm wind conditions, apparently because they could forage on foot on broadledges regardless of wind conditions. Consequently, gull predation rates at Kingittoq werehigher than at Coats Island during calm wind conditions..Numerical response ofglaucous gull populations to murre colony declinesA review of the literature suggests that there is a consistent numerical relationshipbetween the number of breeding predatory gulls and murre colony size at stable colonies (opensymbols, Fig. 5.5, Appendix I). As predicted, the number of glaucous gulls present at thedeclining Upernavik colonies exceeded the observed relationship (Fig. 5.5). Consequently,both murre numbers (F 1,16 = 15.97, P =0.001) and colony location (F 1,16 = 17.01, P=0.001) had a significant influence on the numbers of gulls present at murre colonies. Theseresults imply that a time lag exists in the population response of glaucous gulls to murrepopulation declines.Although the 8 pairs of breeding glaucous gulls at Kingittoq foraged almost exclusivelyon murre eggs and chicks, it is unlikely that the other declining murre colonies entirelysupported the numbers of gulls that were present. This is particularly true at the Kingittoq,Upernavik, and Timmiak colonies where high proportions of murres bred on nest sites113Figure 5.4 Gull predation rates in relation to wind speed and colony (i.e. Coats IslandCanada and Kingittoq, Greenland). Sample sizes for each value refer to the numberof half hours of observation at each wind speed. Values are off-set for clarity.0.12Coa)ID 0.102000 0.08-cU)a)C 0.04-a)0.02-0.000 10 30----0-- Coats IslandKingittoq211531275I I I15 20 25Wind speed (km/h)114Figure 5.5 The relationship between numbers of predatory gulls, murre colony size, andmurre colony state (i.e.Upernavk region vs. other stable or increasing murrecolonies).O Other murre colonies• Upernavik region murre colonies1000•I o• O0 10-.00I01- I I I100 1000 10000 100000 1000000Murres (breeding prs)115that were inaccessible to gulls (see below). It is likely that some of the gulls at Upemavik weregeneralists who foraged at sea, on fish offal from the Upernavik fishing fleet, or at a garbagedump (10-25 km away). The extent to which these gulls exploited the other Upernavikcolonies requires further study of their foraging behaviour and diet (e.g. analysis of pelletcontents at nest sites).Murre nest site selectionMurres selected different nest sites at Coats, Kingittoq, and Timrniak. At Coats Island,most cliff ledges were densely occupied so that most murres nested at high densities on broadcliff ledges (Fig. 5.6). At the Kingittoq colony, murres were more evenly distributed over thecliff face and fewer of them nested at high densities on broad ledges as I predicted (Fig. 5.6).Guano stains on the cliff clearly indicated that empty ledges had once been occupied by murres,and this was confirmed through interviews with local Inuit. At Kingittoq, and contrary toexpectation, few birds nested on broad ledges under low nesting-densities. Instead, mostmurres nested on narrow cliff ledges. At the Tinmiiak colony, which had experienced thegreatest declines, murres nested almost exclusively on narrow ledges at high densities or increvices (Fig. 5.6). Collectively, these results suggest that as colonies declined, murres movedfrom broad ledges to narrow ledges or caves. This is supported by a significant interactionbetween colony state and nest site characteristics (ANCOVA, F 1,6 = 11.6, P = 0.014).Figure 5.6 The % of murres nesting under 5 nest types (BH=broad ledge, high murrenesting density; NL=narrow ledge, low density; NH-narrow ledge, high density;BH=broad ledge, high density; CAVE=cave or crevice nest site), at Coats Island,Kingittoq, and Tinimiak colonies.116>C00C)a)S080706050403020100• COATSQ KINGI1TOQTIMMIAKBL NL NH BH CAVENest site characteristics117DISCUSSIONGuliforaging behaviour andpre&ztion ratesMany seabirds must nest at high densities as a defence against avian nest predators(Burger and Gochfeld 1994). Among murres, high nesting densities are particularly importanton broad cliff ledges where gulls can forage on foot regardless of wind conditions (Birkhead1977; Spear 1993; chapters 2, 3). My inter-colony comparison showed that gulls at thedeclining Kingittoq colony foraged more often on broad ledges and were less constrained bycalm wind conditions, perhaps because population declines increased the availability of low-nesting density ledges where gulls could maneuver on foot and attack murres with little risk ofinjury. This is supported by the observation that gulls at Kingittoq ingested eggs on thenesting ledges, while gulls at Coats typically carried murre eggs to un-occupied ledges abovethe colony.The ability of gulls to forage routinely on foot during calm conditions likely increasesannual predation rates at colonies, because calm conditions typically restrict the ability of gullsto reach murre nest sites at stable colonies (ch. 2, 3). Indeed, at Coats Island calm periodscoincided with, 1) greater absenteeism of gulls from the murre colony (ch. 3, 2) a higherproportion of fish provisioned to gull chicks (a food resource of lower caloric value thanseabird chicks or eggs; Pierotti and Annet 1991; Watanuki 1992), and 3) gull chick weight lossat most nests (Gilchrist, unpubi.). Thus, the ability of gulls at Kingittoq to forage effectivelyon broad cliff ledges regardless of wind should enhance their chick’s growth, and hence, theirreproductive success. This could help maintain high gull numbers at colonies despite murrecolony declines (discussed further below).To summarize, changes in the nesting structure of colonies may influence the economicsof gull foraging behaviour, and this suggests that colony declines resulting from humaninduced mortality of murres can alter natural predator-prey relationships at colonies.118Response ofgulls populations to murre colony declinesAs predicted, the number of gulls per 1000 murre pairs present at colonies in theUpernavik region was higher than at stable murre colonies found elsewhere in Arctic andtemperate regions. This result may reflect the ability of gulls to escape both the constraints ofcalm wind conditions and the reduced effectiveness of murre group defence. Both of thesefactors could enhance adult survival and reproductive success of gulls. However, the highnumber of gulls present at even the smallest colonies (where few murres were accessible, seebelow), implies that gulls in the Upernavik region also utilized food sources at sea, or in thecommunity of Upernavik.The murre colonies closest to the community of Upernavik have suffered the greatestdeclines (Kampp et al. 1994), and are also nearest to alternative food sources for gulls (e.g.garbage dump, fish plant). These alternative food sources may maintain high gull numberseven as murre colonies decline. Consequently, an interaction may exist between the proximityof a murre colony to a human community, hunting intensity, and the numbers of predatorygulls present. These interactions will likely ensure that gulls will be maintained at highnumbers at Upernavik and will continue to depress murre reproductive success, affect nest siteselection, and perhaps prevent the re-colonization of abandoned nesting ledges.The effects of gull predation at declining murre colonies may also be complicated byfeeding territoriality within colonies. Territoriality of seabird predators occurs among skuas,Catharacta sp. (Andersson 1976; Trilimich 1978; Furness 1981; Young 1994), western gulls,L. occidentalis, (Spear 1993), and slaty-backed gulls, L. schistisagus (Watanuki 1992). Atthe Coats Island murre colony, I also found that glaucous gulls defended feeding territories(Gilchrist, in prep.). Further, the distribution of glaucous gull nests at other Arctic colonies(Gaston, pers. corn.), as well as the consistent numerical relationship observed between gullsand murres (Fig 5.5), suggests that feeding territoriality is a widespread phenomenon in thisspecies.119If I assume that the dynamics of glaucous gull territoriality are governed by constraintssimilar to those of raptors (Village 1982; Janes 1984; Temeles 1987; Finck 1990), I predictthe following: if thick-billed murres move from broad, low nesting-density ledges toinaccessible nest sites during population declines (see Results and discussion below), glaucousgull territories should expand during colony declines. In general, declining colonies shouldeventually be occupied by fewer gulls that specialize on murres once murres have made thetransition to inaccessible nest sites.Murre nest site distribution at declining colonies: a consequence ofgull predation?Murres nested on different types of sites at the three colonies, after the availability of thenest types had been controlled for. At Kingittoq, a smaller proportion of murres bred onavailable broad ledges compared to Coats. At a colony near extinction, a greater proportion ofmurres bred on narrow ledges or in caves inaccessible to gulls. These results suggest thatduring the population declines at these colonies, murres had moved from broad ledges tonarrow ledge or cave nest sites. Gull predation may explain the levels of reproductive failureon broad ledges required to establish the odd pattern of nest distribution observed at theUpernavik colonies.Because the Upernavik region colonies have experienced heavy hunting mortality inrecent decades (Evans 1984; Evans and Nettleship 1985), it seems reasonable to assume thatthe sizes of breeding groups on nesting ledges decreased within the colony as murres wereremoved from the breeding population. Similar nesting patterns have been observed at othermurre colonies following rapid population declines (Johnson 1938; Birkhead 1977; Ainley andBoekelhede 1990). If this occurred at Upernavik, it is likely that a greater proportion of birdswould have nested on the periphery of breeding groups (i.e. a greater circumference to arearatio) where they would have been more vulnerable to gull attack (Birkhead 1977; Spear 1993;ch. 2, 3). More importantly, larger unoccupied areas on broad cliff ledges would haveincreased the accessibility of murre nest sites to gulls foraging on foot (Spear 1993; ch. 2). If120these processes occurred during the Upernavik population declines, the predator refugeprovided by high nesting densities may have broken down and murres nesting on broad ledgeswould have experienced higher predation rates. Under these conditions, birds on broad ledgeswould be predicted to eventually abandon their nest sites and prospect for new ones.Assuming that these dynamics are correct, why did displaced murres not converge to fillup a small number of broad ledges after abandoning high-predation areas? If they had, theKingittoq colony would have consisted of a mosaic of empty and high nesting-density broadcliff ledges. In this situation, even a small number of murres could have established a refugeagainst predation by maintaining some high nesting-density areas within the colony where gullscould not land to forage on foot.It is likely that prospecting murres at the declining colonies could not improve their nestsite by recruiting to other broad ledges, because their new nest sites would likely be positionedon the edge of small groups as well. Thus, their new sites would also be vulnerable topredation. Further, the potential for others to join the group would be low because otherprospectors would be thinly distributed over the cliff (due to population declines), and wouldlikely be philopatric to their original nesting areas (Gaston and Nettleship 1981; Noble 1991).Thus, a synchronous colonization of a broad ledge would not likely occur unless there weremany prospectors, or if a structural refuge provided some initial protection from gull attack(e.g. rock overhang, loose rock). Consequently, prospectors would be expected toconsistently abandon broad ledges until they were attracted to the remaining nest sites withinthe colony that were relatively free from predation (i.e. narrow ledges or caves). If thisprocess continued, the majority of birds would eventually nest on sites inaccessible to gulls(Timmiak, Fig. 5.6). Consequently, the per capita predation rate would be predicted to belowest at the colonies that had experienced the greatest declines.I propose that this pattern of prospecting and nest site selection, driven largely by murresattempting to avoid gull predation and harassment, would result in the nesting distributionobserved at the Upernavik area colonies and elsewhere (Johnson, 1938; Birkhead 1977; Ainley121and Boekeihede 1990). The nest distribution at the Upernavik colonies also suggests that gullsalone cannot force a murre colony to extinction, because the structural refuges on the cliffensure that not all nest sites are accessible to gulls.Population level effects ofguilforaging behaviour: multiple-stable states?Once seabird population declines have occurred, can predation of eggs and young holdcolonies at low numbers once sources of adult mortality have been removed? In other words,are there multiple stable states among seabird populations in which avian predation has only aslight influence on the prey population at high densities, whereas at low densities predators canhold the population permanently below the level at which resources become limiting (reviewSinclair 1989; Newton 1993)? For predators to maintain stable prey densities, theory suggeststhat the following criteria must be met: 1) that the predators involved are generalists which canswitch rapidly between alternative food sources as prey numbers change; 2) an ample supply ofalternative prey exists to maintain a high predator population; 3) that prey occur in a patchyhabitat with predator refuges so that prey cannot be forced to extinction by predation; 4)variation in the vulnerability of individual prey and the hunting skills of individual predators;and 5) that predation varies proportionally with prey density which may be determined by intraspecific interactions among the predators themselves, e.g. interference or territoriality (Sinclair1989; Newton 1993). All of these requirements appear to be met at the Upemavik colonies(see above), which suggests that gulls should be able to maintain murre colonies at lownumbers following a population decline. Despite this, data from a limited number of sites(Birkhead 1977; Hatchwell 1992; Ainley and Boekethede 1990), and computer simulationsbased upon thick-billed murre population parameters obtained in the Canadian Arctic (Gilchrist,in prep) suggest that in general, gulls cannot hold murre populations at low levels indefinitelyonce sources of adult mortality have been removed. There are at least two possibleexplanations for this.122First, the pattern of nest site selection during colony recovery may enhance the rate atwhich murres can overcome high levels of egg and chick predation by gulls. During recovery,new recruits to the colony are predicted to join high-density nesting areas to breed, and indoing so, re-establish high-density areas more quickly than if nest sites were selectedrandomly. This pattern is in contrast to the distribution of nest sites predicted during thecolony declines when birds were removed randomly from the breeding population (e.g. shot oroiled at sea). Also during the declines, high philopatry by established breeders to their poorsites would likely prolong the time over which a large portion of the colony nested on sitesvulnerable to gulls. Therefore, I predict that per capita predation would be lower during colonyrecovery than during decline for a given population size. This prediction has implications forseabird colony monitoring because it suggests that the status of a colony at a given populationsize can only be assessed when current census data are integrated with historical information.A second and related explanation regarding colony recovery in the presence of avianpredators considers the relationship between structural habitat refuges, recruitment of young tothe colony, and the re-establishment of high-nesting density areas. In general, predation on theeggs and chicks of birds is much less likely to affect subsequent breeding numbers than ispredation on the adults themselves, particularly among long-lived species such as the thick-billed murre (Newton 1993). Therefore, it is important to consider that glaucous gulls lowerthe effective breeding population by limiting the habitat where murres can reproducesuccessfully, rather than through the direct mortality of breeding adults. As well, at mostdeclining colonies a breeding population is maintained in structural refuges (Birkhead 1977;Ainley and Boekelhede 1990; this study). It appears that members of these remnant breedingpopulations can often produce enough offspring in their lifetime to eventually re-establish highnesting-densities in open habitat (Birkhead 1977, Hatchwell 1991; Ainley and Boekelhede1990). This assumes that there is high philopatry to the natal colony and that offspring returnto breed even after initial reproductive failures, and both of these assumptions appear to be metwith murres (Gaston and Nettleship 1981; Noble 1991; deForest 1992). Therefore, I suggest123that a gradual increase in a breeding population could modify the breeding habitat of murresfavorably and eventually enable them to escape high levels of predation. This could explainthe recovery of murre colonies in the presence of gull predation following initial populationdeclines (Birkhead 1977; Hatchwell and Birkhead 1991; Ainley and Boekelhede 1990).These findings also suggest that a colony could be held at a low population equilibriumby predators if other sources of adult mortality continued to erode the breeding populationand/or the recruits returning to breed at their natal colony. This appears to be the situation atthe southern Upernavik colonies where these colonies continue to persist at low numbers withno indication of recovery, despite the cessation of the summer commercial hunt (Kampp et al.1994; pers. obs.).ConclusionsMy findings illustrate how seemingly unimportant details of predator-prey interactionscan have practical consequences. In this system the interactions between wind speed, cliffledge accessibility, and risk of injury for the predator while forging, varied with preypopulation size. Consequently, human-induced changes in murre numbers appeared to alterthe natural predator-prey dynamics at murre breeding colonies, and the perhaps increased theimpact of predation on murre reproductive success. These findings could be relevant to otherecological systems where the economics of predator foraging decisions are partly determinedby the risk of injury established through group prey defence.124Appendix I. Numbers of predatory gulls in relation to murre colony size (expressed asbreeding pairs). Gulls present in the vicinity of colonies which did not feed primarily onmurre eggs and chicks were excluded.LOCATION GULLS* MURRES@ CENSUS# SOURCEIcefjord, Spitsbergen 2 a 750 a ii Pennycuick 1956Bear Island, Norway 19 a 52 500 a ii Williams 1975Akpatok Island - North, Can., 1985 138 a 450 000 a ii Chapdelaine, pers corn., 1994Akpatok Island - South, Can., 1985 61 a 200 000 a ii Chapdelaine, pers corn., 1994Coats Island, Canada, 1990 22 a 30 000 a i Gilchrist and Gaston, unpubDiggs Island, Canada, 1985 85 a 180 000 a i Gaston et a!. 1985; pers corn. 1994Hantch Island, Canada 21 a 50 000 a ii Gaston, pers corn., 1994Minarettes, Canada, 1985 70 a 133 000 a ii Gaston and Smith, 1985Prince Leopold east, Canada, 1981 40 a 70 000 a i Gaston & Nettleship 1981Halduyt Island, Greenland 50 a 28 000 a iii Kampp 1990; pers. corn. 1994Carey Islands, Greenland 5 a 5 000 a iii Kampp 1990; pers. corn. 1994Saunders Island, Greenland 50 a 107 000 a iii Kampp 1990; pers. corn. 1994Parker Snow Bay, Greenland 50 a 38 000 a iii Kampp 1990; pers. corn. 1994Agpat Agpai, Greenland 50 a 36 000 a iii Kampp 1990; pers. corn. 1994Upernavik Agparssuit, Greenland, 19936 a 350 a ii this studyKingigtoq - general, Greenland, 1993 20 a 7500 a ii this studyTimmiakulussuiot, Greenland, 1993 16 a 200 a ii this studyAppatsiaat, Greenland, 1993 30 a 320 a ii this studyBezymyannaya Bay, Russia, 1956 60 a 180 000 a iii Uperski, 1956SE Farallon Island, U.S.A. 8 b 4 500 b i Spear 1993* a) glaucous gull, Larus hyperboreus, b) western gull, Larus@ a) thick-billed murre,Uria lornvia, b) common murre, Uria aalge.# i) detailed census with gull behaviour! diet information, ii) detailed census of both species, iii) estimate ofat least one species125CHAPTER 6CONCLUSIONSThesis synthesisIn chapter 2, I presented one of the first experimental studies to examine the foragingconstraints of an avian predator in the wild. I found that the foraging ability of glaucous gullswas constrained by narrow cliff ledge width, high murre nesting densities, and group defence.My results confirm descriptive behavioural studies of other avian predators foraging at seabirdcolonies (Watanuki 1989; Spear 1993; Young 1994). However, I found that glaucous gullscould overcome these foraging constraints using wind conditions which enabled them to reachweakly-defended nest sites on narrow ledges. Thus, the foraging constraints affectingglaucous gulls were not constant, but varied with wind conditions.In chapter 3,1 conducted observations of gulls preying on the eggs and chicks of naturallyoccurring murre eggs and chicks and found that gull foraging activity and predation rates werepositively correlated with wind speed. Using a model which integrated field data on gullpredation rates and wind conditions, I found that the vulnerability of murre nest sites was afunction of both colony topography and seasonal wind conditions.In chapters 3 and 4, I concluded that a trade-off existed between energetic gain and risk ofinjury to the foraging gulls, and that this trade-off was mediated by wind. Although I neverwitnessed an adult gull being killed by murres, I did observe adult gulls being injured by thedefensive strikes of murres. I suggest that the potential for injury during attack altered therelative costs and benefits of foraging decisions. For example, although foraging on footunder calm wind conditions yielded the greatest rewards for gulls, it also incurred the greatestrisk of contact with defending murres. Consequently, gulls preferred to forage aerially duringwindy conditions, although aerial attacks were much less successful than attacks made on foot.In the most extreme response to this trade-off, gulls were inactive during calm conditions (Fig.3.2), perhaps because they were waiting for the wind to increase and for foraging conditions to126improve. To my knowledge, this is the first field study to support the theoretical prediction ofSutherland and Moss (1985), that foraging inactivity among top predators is influenced by thedangers and rewards of current foraging conditions, rather than simply being a result ofpredator satiation.In chapter 5, I tested the prediction that gull predation should increase following anyperturbation that decreases murre nesting densities. Although previous seabird studies havediscussed this possibility (Johnson 1938; Birkhead 1977; Ainley and Boekeihede 1990), mystudy is among the first to quantify this influence and to examine why it occurs. I found thatgulls at declining colonies in Greenland foraged on broad ledges and were less constrained bycalm wind conditions, apparently because population declines increased the availability of lownesting-density ledges where gulls could attack murres on foot regardless of wind conditions.Consequently, predation rates at the declining Upernavik murre colonies were higher duringcalm wind conditions compared to the Coats Island colony. I suggest that the effects ofpredators on their prey may be particularly among prey species who rely on group defence toavoid predation.Comparisons with other gull foraging studiesI now compare and contrast my results with other studies that have examined theforaging ecology of large gulls, and to identify topics that warrant further research (Table 6.1).Few studies of gulls have examined how risk of injury while foraging influences gull foragingbehaviour, although several studies have identified prey defence as an important constraint ongull foraging efficiency. Most studies to date have emphasized the effects of gull predation onthe reproductive success of prey, or have identified factors that affect diet selection (e.g. tradeoffs between foragingtime allocation and nest defence). Of those studies that have focusedspecifically on attack behaviour, several have found, as I did, that windy conditions enhancegull foraging efficiency. There are too few studies, however, to determine whether the effect127Table 6.1. A review of studies examining the foraging ecology of 7 species of large gullduring the breeding season. %POP refers to the proportion of the population thatspecialized on a specific food item. FT refers to gulls that maintained feeding territories.SUBSTR refers to the slope and other habitat characteristics where gulls were foraging.WIND refers to studies that detected significant wind effects on gull foraging efficiency.Question marks refer to topics that were not examined in the research. In studies wheretwo foraging behaviours occurred within a single gull population, each strategy isdescribed separately and they share the same literature source.SPECIESREGIONNESTDIET%POPFORAGEFACTORSINDIVDSUBSTR.WINDSTUDYSOURCEBEHAV.MODEAFF.FOR.VAR.EFFECTFOCUSL.hyperboreuslowArcticterritorialseabird>90%predationpreydefense,yes:for.cliffyesforagingmode,thisstudyGlaucouscoastaldispersedeggs,fish.aerial,ontopography,,rodents,???yes:shoreline,?dietselection,BarryandBarry,coastalorsolitaryfish,diettundra,sea,seasonalvar.1990wat.eggs.indiet.LhyperboreusArcticterritorialwat.eggs,>70%???tundra,?gullforagingStrang1976coastaldispersedfish.coasts.andbreedingFTecology.L.hyperboreuslowArcticterritorialseabird?predationpreydefence,?cliffyespred.effectsGastonetal.,1985coastaldispersedeggs.aerial,ontopography.onprey.Fr?foot.lowArcticcolonialfish,??shoreline?Gastonetal.,1985coastalintertidal.L.hyperboreuslowArcticterritorialwaterfowl100%predationpreydefenceyes:for.island,nopred.effectsSchamel1977dispersedeggs,fish.onfootpreyconcealedbehav.flat.onprey.FTL.hyperboreuslowArctic?intertidal,?searching?yes:shoreline?dietselection.Ingolfsson1976coastalshellfish.onfootdietinter-specificcompetition.L.marinuslowArctic?garbage,fish,?scavenge?yes:shoreline?dietselectionIngolfsson1976Greaterblack-backedseabirdeggs.dietofgulls.L.rnarinustemperateterritorialseabirdeggs?predationpreydefence,yes:for.flatisland?pred.effectsJohnson1938coastaldispersedaerial,onhumandist.behav.onprey.FT?foot.Lglaucascenstemperatecolonialseabirdeggs,predationpreydefence,?island,flatnopred.effectsSiegel-CauseyandGlaucous-wingedcoastalfish.onfoottopography,andsloped,onprey,pred.Hunt1985.eagledist.for.constraints.colonialgarbage,fish?atsea????Siegel-CauseyandHunt1985.L.Californicustemperatecolonial“inlandlakes”0%??gullreprod.Pugesek1992Californiagullcoastalstrategies.L.schistisagustemperatecolonialseabirdeggs.?predation,preydefenceyes:diet,island,gullfor.behav.Watanuki1983,Slaty-backedpillaerial,topographyfor.behav.sloped.individ.diff.1989,1992indiet.Loccidentalistemperateterritorialseabirdeggs,1%predationpreydefenceyes:for.island,yespred.for.behav.Spear1993WesternIlcoastalFTfish.aerial,ontopographybehav.sloped,andeffectsonfoot.intra-speccomp.prey.colonialgarbage,99%atseaatsea,Spear1993fish.shorelines.Largentatustemperatecolonialseabirdeggs,30%predation,preydefence,7island,flatyespred.effectsTheil andSommer,HerringpIllcoastalfish.aerial,wind,onprey,wind1994.effects.colonialfish.70%77?sea,shorelines?TheilandSommer,‘-1994.Largentatustemperatecolonialseabirdeggs,4%predationpreydefenceyesisland,flat.?pred.for.SouthernandFr?fish.aerial,onbehav.andSouthern,1984.foot.effectsonprey.temperatecolonialgarbage,96%???distantgarb.?SouthernandfishdumpSouthern,1984L.argentatustemperatecolonialgarbage,?scavenge,temporalyes:distantgarb.?trade-offs:SiblyandMcCleeryearthworms,onfoot.variationindietagriculturedietselection1991.intertidal,preyavail,shorelinesandnestguarding.L.argentatustemperatecoloniaL,intertidal45%L.onfoot.timeawayyes:distantshores,trade-offs:PierottiandAnnett,variousgarbage..23%4’onfootfromnest.dietdumpsdietselection1991.habitats.seabirdadults.9%ijaerialtopography.atcolonyandnest guarding.L.argenratuslowarctic?waterfowl?predation,preydefence,?shorelinesyesfactorsaffectingMendenhall andchicks.aerial,wind,openwaterpreysurvivalMime,1985.C131of wind is related to the topography of the habitat as I have suggested (e.g. gulls that forage onfoot in level habitat may be less reliant on wind than gulls that forage in cliff habitat; Ch. 2, 3).I now consider topics that warrant further study. Gulls are well-suited for comparativestudies because they have been extensively studied in the wild, and they have extremely diversediets, and nesting and foraging behaviours. By comparing the reproductive and foragingecology of gull species, it should be possible to explore how foraging behaviour andenvironmental constraints interact to influence the evolution and maintenance of coloniality.For example, my literature review showed that two distinct foraging strategies often occuramong gull populations that breed near avian prey (e.g. waterfowl or seabird colonies). Thesepopulations can be divided into ‘specialist predators’ which feed primarily on eggs and chicks,and ‘generalists’ who have a varied diet. Nest site selection by gulls within a populationappears to be related to these foraging strategies (Table 6.1). Gulls that commute to distantforaging sites typically nest colonially. In contrast, the nests of predator-specialists aretypically dispersed because these gulls often maintain feeding territories around their nests.It would be interesting to study the costs and benefits of these two foraging strategies, andto explore why both are maintained within a population. I suggest that predation on seabird orwaterfowl eggs by specialists is typically the most profitable foraging strategy for 3 reasons.First, gulls that nest colonially and commute to foraging sites must balance foragingopportunities with the risk that their chicks may be killed by conspecifics in their absence(Pierotti and Annett 1991; Sibly and McCleery 1991). This risk appears to be reduced forpredator-specialists because they forage near their nest, and often exclude other gulls from thearea (Spear 1993; this study). Second, travel to distant food sources likely decreases the netprofitability of prey, which in turn, may decrease chick growth and reproductive success (Siblyand McCleery 1991). Third, birds that commute to sea or to intertidal areas typically forage onunpredictable (e.g. fish) and less profitable prey (e.g. intertidal organisms) when comparedwith seabird prey (Watanuki 1992; Theil and Sommer 1994; Southern and Southern 1984; thisstudy; but see Pierotti and Annett 1991).132Despite the benefits for gulls of specializing on seabird or waterfowl prey, predator-specialists often make up a small proportion of gull populations (Table 6.1). Perhapscompetitive exclusion by territoriality restricts most individuals within a population fromaccessing the most profitable prey. However, little is known about how individual gullswithin a population become specialized. From this, I identify a need for researchers to studyhow foraging specialists come to occur in a gull population, whether the ratio of generalists tospecialists remains stable during environmental change, and whether competitive exclusionwithin a population prevents more individuals from adopting a specialized diet.From Lw-us to Leo: the need to examine mortality risks among top predatorsIt is difficult to quantify the mortality risks of decisions made by animals in the wild.Given the consequences, it is not surprising that fatal outcomes are rarely observed.Nevertheless, mortality risks influence the costs and benefits of behavioural decisions in thewild. For example, prey appear to be sensitive to both the dangers associated with behaviouralalternatives, and to environmental factors that may alter the degree of danger (Milinski andHeller 1978; Lima 1986; Lima and Dill 1990; Sih 1989; Ydenberg 1994). Although many toppredators face mortality risks while foraging, little attention has been devoted to how thesedangers influence their behaviour (Stein 1977; Pettifor 1990; Merav et al. 1991). It is likelythat the trade-off between energetic gain and risk of injury that I found among glaucous gullsdepredating murres, is common among top predators. For example, the largest and mostenergetically valuable prey are often the most dangerous to attack. Indeed, it is common forsome predators to “test” the defensive abilities of their large prey before attempting to make akill (lions, Schaller 1972; wolves, Mech 1970; hyaenas, 1990).The trade-off between energetic gain and risk of injury while foraging could affectseveral aspects of predator behaviour, and in many cases this has been over-looked. Forexample, several authors have concluded that the average group size of African lions (Pantherleo) is larger than required to capture and subdue even large prey (reviewed in Packer 1986;133Mangel and Clark 1988). Several explanations have been invoked to explain “larger thanoptimal’ pride size. Large prides may have a lower variance in daily food intake (Caraco1981), increased mating success (Caraco and Wolf 1975), or be able to maintain largerfeeding territories. However, to my knowledge, no one has considered that individuals thatforage as part of a larger pride face less risk of injury when subduing prey. Given thefrequency with which lions hunt (every few days), their potential longevity (10-15 years),and the size and defensive abilities of their large prey, it seems reasonable that reducedmortality risks while hunting could out-weigh the energetic costs of sharing prey amongmore individuals. I do not suggest that this single danger-reward trade-off explains all thecomplexities of lion sociality, but I find it surprising that this topic has not been examinedquantitatively.In summary, mortality risks can influence foraging and provisioning behaviour amongtop predators. Specifically, my study illustrates how seemingly unimportant changes inenvironmental conditions can alter the relative costs and benefits of foraging decisions madeunder risk of injury (e.g. the interaction between wind speed, ledge accessibility, and preydefence). These dynamics can have practical implications because long-term changes in thedanger-reward trade-offs of top predators can influence their foraging behaviour andconsequently, their impacts on prey populations.134LITERATURE CITEDAinley, D. and Boekeihede, 1990. Seabirds of the Farallon Islands: Ecology, dynamics andstructure of an upwelling system community. Stanford Univeristy Press: Stanford.450 pp.Amat, J. A., and E. Aguilera. 1990. Tactics of black-headed gulls robbing egrets and waders.Animal Behavior, 39, 70-77.Andersson, M. 1976. Predation and kieptoparasitism by skuas in a Shetland seabird colony.Ibis, 118, 208-217.Andersson, M., C. G. Wiklund, and Rundgren, H. 1980. Parental defence of offspring: amodel and an example. Animal Behaviour, 28, 536-542.Barry, S. J. and T. W. Barry. 1990. Food habits of glaucous gulls in the Beaufort Sea.Arctic, 43, 43-49.Baudinette, R. V. and K. Schmidt-Nielsen. 1978. Energy cost of gliding flight in herringgulls. Nature, 248, 83-84.Beauchamp, G., and A. Kacelnik. 1990. On the fitness functions relating parental care toreproductive value. Journal of Theoretical Biology, 146, 513-522.Beauchamp, G., B. J. Ens, and A. Kacelnik. 1990. A dynamic model of food allocation tostarling, Sturnus vulgaris, nestlings. Behavioural Ecology, 2, 2 1-37.Bell, W. J. 1991. Searching Behaviour: The Behavioral Ecology of Finding Resources.Chapman and Hall. London.Birkhead, T. R. 1977. The effect of habitat and density on breeding success in the commonguillemot, Uria aalge. Journal of Animal Ecology, 46, 751-764.Birkhead, T. R., Greene, E., Biggins, J.d. & Nettleship, D. N. 1985. Breeding sitecharacteristics and reproductive success in thick-billed murres. Canadian Journal ofZoology, 63, 1880-1884.Birkhead, T.R. and D. N. Nettleship. 1986. Ecological relationships between commonmurres, Uria aalge, and thick-billed murres, Uria lomvia, at the Gannet Islands,Labrador. II: breeding success and site characteristics. Canadian Journal of Zoology,65, 1630-1637.Birt-Friesen, V. L., W. A. Montevecchi, E. K. Cairns, and S. A. Macko. 1989. Activity-specific metabolic rates of free-living northern gannets and other seabirds. Ecology, 70,357-367.Buckley, F. G. and P. A. Buckley. 1980. Habitat selection and marine birds. In J. Burger,B. L. Olla and H. E. Winn (eds.), Behavior in Marine Animals. Vol. 4: Marine Birds.pp. 69-112. Plenum Press, New York.Burger, J. and M. Gochfeld. 1994. Predation and effects of humans on island-nestingseabirds. In, Seabirds on Islands: Threats. Case Studies, and Action Plans. ed. by, D.N. Nettleship, J. Burger, and M. Gochfeld. Birdlife International: Cambridge.135Calder, W. A. and J. R. King. 1974. Thermal and caloric relations of birds. In: D. S. Farner& J. R. King (eds.). Avian Biology. vol. 4: 259-413. Academic Press. New York.Caraco, T. 1981. Risk sensitivity and foraging groups. Ecology, 62, 527-53 1.Caraco, T. and L. L. Wolf. 1975. Ecological determinants of group size in foraging lions.American Naturalist, 109, 343-352.Carpenter, F. L. 1987a. Territoriality: conceptual advances in field and theoretical studies.American Zoologist, 27, 223-228.Carpenter, F. L. 1987b. Food abundance and territoriality: to defend or not to defend?American Zoologist, 27, 387-399.Castellini, M. A. and L. D. Rea. 1992. The biochemistry of natural fasting at its limits.Experientia, 48, 575-582.Clark, C. W. 1991. Modelling behavioral adaptations. Behavioral and Brain Sciences, 14,85-117.Clark, C. W. 1994. Antipredator behavior and the asset-protection principle. BehavioralEcology, 5, 159-170.Cooper, S. M. 1990. The hunting behaviour of spotted hyaenas, Crocuta crocuta, in a regioncontaining both sedentary and migratory populations of herbivores. African Journal ofZoology, 28, 131-141.Coulson, J. C., P. Monaghan, J. Butterfield, N. Duncan, C. Thomas, & C. Shedden. 1983.Seasonal changes in the herring gull in Britain: Weight, moult and mortality. Ardea, 71,235-244.Curio, E. 1974. The Ethology of Predation. Springer-Verlag. New York.Daan, S. & Tinbergen, J. 1979. Young guillemots, Uria lomvia, leaving their arctic breedingcliffs: a daily rhythm in numbers and risk. Ardea, 67, 97-100.deForest, L. N. 1993. The effects of age, timing of breeding, and site characteristics on thereproductive success of the thick-billed murre, Uria lomvia.. M.Sc. thesis. Universityof Ottawa.Dill, L. M. 1986. Animal decision making and its ecological consequences: the future ofaquatic ecology and behaviour. Canadian Journal of Zoology, 65, 803-811.Dijkstra, C., A. Bult, S. Bijlsma, S. Daan, T. Meijer, and M. Zijistra. 1990. Brood sizemanipulations in the kestrel, Falco tinnunculus: effects on offspring and parent survival.Journal of Animal Ecology, 59, 269-285.Drent, R. H. and S. Daan. 1980. The prudent parent: energetic adjustments in avian breeding.Ardea, 68, 225-252.Drent, R. H., M. Klaassen, and B. Zwaan. 1992. Predictive growth budgets in terns andgulls. Ardea, 80, 5-17.136Dunn, E. H. 1976. The development of endothermy and existence energy expenditure inherring gull chicks. The Condor, 78, 493-498.Dunn, E. H. 1980. On the variability in energy allocation of nestling birds. Auk, 97, 19-27.Dunn, E. K. 1973. Changes in the fishing ability of terns associated with windspeed and seasurface conditions. Nature, 244, 520-521.Edmunds, M. 1974. Defense in animals: a survey of anti-predator defenses. London:Longmon.Ellis, H. I. 1984. Energetics of free-ranging seabirds. Pages 203-234 in G. C. Whittow & H.Rahn (ed.). Seabird Energetics. Plenum Press, New York.Endler, J. A. 1986. Defense against predators. In: Predator-prey relationships: perspectivesand approaches from the study of lower vertebrates. (Ed. by M.E. Feder and G.V.Lauder), pp 109-134. Chicago: University of Chicago Press.Erlinger, T. J. 1989. Learning and individual variation in bluegill foraging: habitat-specifictechniques. Animal Behavior, 38, 643-658.Evans, P.G.H. 1984. The seabirds of Greenland: Their status and conservation. In:Croxall, I.P., Evans, P.G.H., and Schreiber, R.W., eds. Status and conservation of theworld’s seabirds. ICBP Technical Publication No 2: 49-84.Evans, P.G.H. and Nettleship. 1985 Conservation of the Atlantic Alcidae. In: The AtlanticAlcidae. Orlando: Academic Press. 53-154.Falk, K. and J. Durinck. 1992. Thick-billed murre hunting in West Greenland, 1988-89.Arctic, 45, 167-178Finck, P. 1990. Seasonal variation of territory size with the little owl, Athene noctua.Oecologia, 83, 68-75.Fitzgibbon, C. D. 1990. Mixed-species grouping in Thomson’s and Grant’s gazelles: theantipredator benefits. Animal Behaviour, 39, 1116-1126.Flint, E. N. and K. N. Nagy. 1984. Flight energetics of free-living sooty tems. Auk, 101,288-294.Furness, R.W. 1981. The impact of predation by great skuas, Catharacta skua, on otherseabird populations at a Shetland colony. This, 123, 534-539.Furness, R. W. and J. R. G. Hislop. 1981. Diets and feeding ecology of great skuas,Catharacta skua, during the breeding season in Shetland. Journal of Zoology, 195, 1-23.Gabrielsen, G. W. and F. Mehium. 1984. Thermoregulation and energetics of arctic, G. C. Whittow & H. Rahn (eds.). Seabird Energetics. Plenum Press, New York.Gabrielsen, G. W., F. Mehium, and K. A. Nagy. 1987. Daily energy expenditure and energyutilization of free-ranging black-legged kittiwakes. Condor, 89, 126-132.137Gaston, A.J. & Nettleship, D.N. 1981. The thick-billed murres of Prince Leopold Island: Astudy of the breeding ecology of a colonial, high arctic seabird. Canadian WildlifeService Monograph Series No. 6, Ottawa.Gaston, A. J., Cairns, D. K., R. E. Elliot, and D. N. Noble. 1985. A Natural History ofDigges Sound. Can. Wild. Serv. Report Ser. No. 46.Gaston, A. J. and R. D. Elliot. 1991. Studies of high-latitude seabirds: Conservation biologyof thick-billed murres in the northwest Atlantic. Can. Wild. Serv. Occasional Paper. 69.Gaston, A. J., L. deForest, H. G. Gilchrist, and D. N. Nettleship. 1993. Monitoring thick-billed murre populations at colonies in northern Hudson Bay, 1972-1992. CanadianWildlife Service, Occasional Paper, No. 80.Gendron, R. P. and Staddon, J. R. 1983. Searching for cryptic prey: the effect of searchrate. American Naturalist, 121, 172-186.Gessaman, J. A. Bioenergetics of the snowy owl. Arctic and Alpine Research, 4, 223-238.Gochfeld, M. 1980. Mechanism and adaptive value of reproductive synchrony in colonialseabirds. In: Behaviour of Marine Animals. Vol. 4 (ed. by J. Burger, B.L. Olla &H.E. Winn), pp 207-270. New York: Plenum Press.Goldstein, A. 1983. Effect of wind on avian metabolic rate with particular reference togambel’s quail. Physiological Zoology, 56, 485-492.Goss-Custard, J. D., R. W. G. Caldow, R. T. Clarke, and A. D. West. Deriving populationparameters from individual variations in forging behaviour. U. Model tests andpopulation parameters. Journal of Animal Ecology, 64, 277-289.Gotmark, F. and M. Andersson. 1984. Colonial breeding reduces nest predation in thecommon gull, Larus canus. Animal Behaviour, 32, 485-492.Gray, D. R. 1974. The defence formation of the Musk-ox. National Museum of NaturalSciences. The Musk-ox, 14, 25-29.Grub, T. C. 1977. Weather-dependent foraging in Ospreys. Auk, 94, 146-149.Hamilton, W. D. 1971. Geometry for the selfish herd. Journal of Theoretical Biology, 31,295-3 11.Haney, J. C. and D. S. Lee. 1993. Air-sea heat flux, ocean wind fields, and offshoredispersal by gulls. Auk, 111, 427-440.Harfenist, A. and R. C. Ydenberg. 1995. Parental provisioning and predation risk inrhinoceros auklets, Cerorhinca monocerata: effects on nestling growth and fledging.Behavioral Ecology, 6, 82-86.Harris, M. P. and H. Jones. 1969. Sexual differences in measurements of herring and lesserblack-backed gulls. British Birds, 62, 129-133.Hatch, J. 3. 1970. Predation and piracy by gulls at a ternery in Maine. Auk, 87, 244-254.138Hatchwell, B .J. & Birkhead, T.R. 1991. Population dynamics of common guillemots, Uriaaalge , on Skomer Island, Wales. Ornis Scandinavica, 22, 55-59.Hatchwell, B .J. 1991. An experimental study of the effects of timing of breeding on thereproductive success of common guillemots, Uria aalge. Journal of Animal Ecology,60, 721-736.Hixon, M. A. 1980. Food production and competitor density as the determinants of feedingterritory size. American Naturalist 115, 510-530.Hoochachka, W. 1992. How much should reproduction cost? Behavioral Ecology, 3,42-52.Hunt, G. L. Eppley, Z.A. & Schneider, D.C. 1986. Reproductive performance of seabirds:The importance of population and colony size. Auk, 103, 306-317.Ingolfsson, A. 1976. The feeding habits of great black-backed gulls and glaucous gulls inIceland. Acta Naturalia Islandica, 24, 1-19.Janes, S. W. 1984. Influences of territory composition and interspecific competition on red-tailed hawk reproductive success. Ecology 65, 862-870.Johnson, S. R., and D. R. Herter. 1989. Shorebirds, gulls, terns and alcids. In: The birds ofthe Beaufort sea. pp. 142-22 1. Alaska: Bridgetown Printing.Johnson, R.A. 1938. Predation of gulls in murre colonies. Wilson Bulletin., 50, 161-170.Kampp, K., D. N. Nettleship, and P. H. Evans. 1994. Thick-billed murres of Greenland:status and prospects. In, Seabirds on Islands: Threats. Case Studies, and Action Plans.ed. by, D. N. Nettleship, J. Burger, and M. Gochfeld. Birdlife International:Cambridge.Kanwisher, J. W., T. C. Williams, J. M. Teal, and K. 0. Lawson. 1978. Radiotelemetry ofheart rates from free-ranging gulls. Auk, 95, 288-293.Kendeigh, S. C. 1970. Energy requirements for existence in relation to size in birds. Condor,72: 60-65.Klaassen, M. 1994. Growth and energetics of tern chicks from temperate and polarenvironments. The Auk, 111, 525-544.Kleinbaum, L. L., L. Kupper, and K. E. Muller. 1988. Applied Regression Analysis andOther Multivariate Methods.Lack, D. 1968. Ecological adaptations for breeding in birds. Methuen, London.Lima, S. L. 1986. Predation risk and unpredictable feeding conditions: determinants of bodymass in birds. Ecology, 67, 377-385.Lima, S. L. and L. M. Dill. 1990. Behavioral decisions made under the risk of predation: areview and prospectus. Canadian Journal of Zoology, 68, 6 19-640.Machmer, M. M. and R. C. Ydenberg. 1989. Weather and osprey foraging energetics.Canadian Journal of Zoology, 68, 40-43.139Mangel, M. and C. W. Clark. 1986. Towards a unified foraging theory. Ecology, 67,1127-1138.Mangel, M. and C. W. Clark. 1988. Dynamic modeling in behavioral ecology. PrincetonUniversity Press: New Jersey.Manning, T. E., E. 0. Hohn, and A. H. MacPherson. 1956. The birds of Bands Island. Nat.Mus. Canada. Bull. No. 143. Ottawa, Canada.Martin, M. and C.S. Moiteret. 1981. Bird populations and habitat use, Canning River delta,Alaska. U.S. Fish and Wildlife Service. Fairbanks, Alaska.McNamara, J. M. and A. I. Houston. 1990. The value of fat reserves and the trade-offbetween starvation and predation. Acta Biotheretica, 38, 37-61.Mech, L. D. 1970. The Wolf: The Ecology and Behaviour of an Endangered Species. NaturalHistory Press: New York.Mearns, R. and Newton. 1988. Factors affecting breeding success of peregrines in southScotland. Journal of Animal Ecology, 57, 903-9 16.Mendenhall, V. M., and H. Milne. 1985. Factors affecting duckling survival of Eiders innortheast Scotland. This, 127, 148-158.Merav, B., Pellis, S.M. and V. C. Pellis. 1991. Feeding habits and predatory behaviour in themarbled polecat, Vormela Peregusna syriaca: Killing methods in relation to prey size andprey behaviour. Behaviour, 118, 127-143.Milinski, M. and R. Heller. 1978. Influence of a predator on the optimal foraging behaviourof sticidebacks, Gasterosteus aculeatus. Nature, 275, 127-143.Mills, M. G. L. and T. M. Shenk. 1992. Predator-prey relationships: the impact of lionpredation on wildebeest and zebra populations. Journal of Animal Ecology, 61,693-702.Mills, M.G.L. 1990. Kalahari Hyaenas. Unwin Hyman: London.Mock, P. J. 1991. Daily allocation of time and energy of western bluebirds feeding nestlings.Condor, 93, 5948-611.Nagy, K. 1987. Field metabolic rate and food requirement scaling in mammals and birds.Ecological Monographs, 57, 111-128.Nelson, D. A. 1989. Gull predation of Cassin’s aukiet varies with the lunar cycle. Auk,106, 495-497.Nettleship, D.N. 1972. Breeding success of the common puffin, Fratercula arctica L., ondifferent habitats at Great Island Newfoundland. Ecological Monographs, 42, 239-268.Nettleship, D. N. 1976. Census techniques for seabirds of arctic and eastern Canada.Canadian Wildlife Service Occasional Paper No. 25, 1-3 3.140Newton, I. 1993. Predation and limitation of bird numbers. Current Ornithology, Volume 11,143- 197.Nisbet, I. C. T. and M. I. Welton. 1984. Seasonal variations in breeding success ofcommon terns: consequences of predation. Condor, 86, 53-60.Noble, D. 1990. Factors affecting recruitment of thick-billed murres, Uria lomvia, on CoatsIsland, N.W.T. M.Sc. thesis. Queen’s University.Norberg, R. A. 1977. An ecological theory on foraging time and energetics and choice ofoptimal food-searching method. Journal of Animal Ecology, 46, 511-529.Norris, K. S. and T. P. Dohl. 1980. The structure and function of cetacean schools. CetaceanBehavior: Mechanisms and Processes. ed. L. M. Herman. Wiley and Sons: NewYork.Nur, N. 1988. The cost of reproduction in birds: an examination of the evidence. Ardea,76, 155-168.Ollason, J. C. and G. M. Dunnet. 1988. Variation in breeding success in fulmars. In,Reproductive Success: Studies of Individual Variation in Contrasting Breeding Systems.edt. T. H. Clutton-Brock. Chicago Press: Chicago.Packer, C. 1986. The ecology of sociality in felids. In Ecological Aspects of socialEvolution. D. I. Rubenstein and R. W. Wrangham, eds. Princeton Univeristy Press:Princeton. pp 429-451.Parsons, J. 1971. Cannibalism in herring gulls. British Birds, 64, 528-537.Pennycuick, C. J. 1987. Flight of Auks (Alcidae) and other northern seabirds comparedwith southern Procellariiformes: Ornithodolite observations. Journal of ExperimentalBiology, 128, 335-347.Pennycuick, C. J. 1989. Bird Flight Performance: A Practical Calculation Manual. OxfordScience Publications: Oxford. 153 pp.Pennycuick, C. 1. 1956. Observations on a colony of brunnich’s guillemot, Uria lomvia , inSpitsbergen. Ibis, 98, 80-99.Pettifor, R. A. 1990. The effects of avian mobbing on a potential predator, the Europeankestrel, Falco tinnunculus. Animal Behaviour, 39, 821-827.Pierotti, R. and C. A. Annett. 1991a. Diet choice in the herring gull: Constraints imposed byreproductive and ecological factors. Ecology, 72, 319-328.Pierotti, R. and C. A. Annett. 199 lb. Diet and reproductive output in seabirds. Bioscience,40, 568-575.Portenko, L. A. 1989. Larus hyperboreus: Glaucous gull. In: Birds of the ChukchiPenninsula and Wrangel Island. pp. 36-52. Washington: Smithsonian InstitutionLibraries.Pugesek, B. H. 1995. Offspring growth in the California gull: reproductive effort and parentalexperience hypotheses. Animal Behaviour, 49, 64 1-647.141Pugesek, B. H. and P. Wood. 1992. Alternate reproductive strategies in the California gull.Evolutionary Ecology, 6, 279-295.Pyle, P., L. B. Spear, W. J. Sydeman, and D. G. Ainley. 1991. The effects of experience andage on the breeding performance of western gulls. Auk, 108, 25-33.Real, L. and T. Caraco. 1986. Risk and foraging in stochastic environments. Annual Reviewof Ecology and Systematics, 17, 371-390.Reid, W. V. 1988. Age-specific patterns of reproduction in the glaucous-winged gull:Increased effort with age. Ecology, 69, 1454-1465.Reznick, D. 1985. Costs of reproduction: an evaluation of the empirical evidence. Oikos,44, 257-267.Richards, L. J. 1983. Hunger and the optimal diet. The American Naturalist, 122, 326-334.Rosenzweig, M. E. and R. H. Macarthur. 1963. Graphical representation and stabilityconditions of predator-prey interactions. American Naturalist, 97, 209-223.Rudolph, S. G. 1981. Foraging strategies of American kestrels during breeding. Ecology,63, 1268-1276.Sage, B. L. 1974. Ecological distribution of birds in the Atigun and Sagavanirktok Rivervalleys, arctic Alaska. Canadian Field Naturalist, 88, 281-291.Schaller, G. B. 1972. The Serengeti Lion. University of Chicago Press: Chicago.Schamel, D. 1977. Breeding of the common eider on the Beaufort Sea coast of Alaska.Condor, 79, 478-485.Schnell, G. D. and J. J. Hellack. 1979. Bird flight speeds in nature: optimized or acompromise? The American Naturalist, 113, 53-66.Siegel-Causey, D. and G. L. Hunt, Jr. 1981. Colonial defence behavior in double-crestedand pelagic cormorants. Auk, 98, 522-531.Sibly, R. and R. McCleery. 1985. Optimal decision rules for herring gulls. AnimalBehaviour, 33, 449-465.Sibly, R. M. and R. H. McCleery. 1983. Increase in weight of herring gulls while feeding.Journal of Animal Ecology, 52, 35-50.Sih, A. 1989. Predator and prey lifestyles: an evolutionary and ecological overview, inPredation: Direct and Indirect Impacts on Aquatic Communities. eds. W. C. Kerfootand A. Sih. University Press of New England: London.Sinclair, A. R. E. 1989. Population regulation in animals. (edt.) J. M. EcologicalConcepts: the contribution of ecology to an understanding of the natural world.Blackwell: Boston.Sokal, R. R. and F. J. Rohif. 1981. Biometrics: The Principles and Practice of Statistics inBiological Research. 2nd Edition. Freeman: San Francisco. 859 pp.142Southern, W. E. and M. D. Southern. 1984. Herring gulls specialize as ring-billed gullpredators. Colonial Waterbirds, 7, 105-110.Spaans, A. L. 1971. On the feeding ecology of the herring gull, Liirus argentatus, in thenorthern part of the Netherlands. Ardea, 59, 73-186.Spear, L.B. 1993. Dynamics and effect of western gulls feeding in a colony of guillemots andBrandt’s cormorants. Journal of Animal Ecology, 62, 399-4 14.Spear, L. B. and D. W. Andersson. 1989. Nest site selection of yellow-footed gulls.Condor, 91, 91-99.Stein, R. A. 1977. Selective predation, optimal foraging, and the predator-prey interactionbetween between fish and crayfish. Ecology, 58, 1237-1253.Stephens, D. W. 1987. On economically tracking a variable environment. TheoreticalPopulation Biology, 32, 15-25.Stephens, D. W. 1991. Change, regularity, and value in the evolution of animal learning.Behavioural Ecology, 2, 77-89.Stephens, D. W. and J. R. Krebs. 1986. Foraging Theory. Princeton University Press: NewJersey.Strang, C. A. 1976. Feeding behavior and ecology of glaucous gulls in western Alaska.Ph.d. thesis, Purdue University, Layfayette, Indiana.Sutherland, W. J. and D. Moss. 1985. The inactivity of animals: influence of stochasticityand prey size. Behavior, 92, 1-8.Stephens, D. W. 1990. Foraging theory: up, down, and sideways. Studies in Avian Biology,13, 444-454.Suhonen, J., K. Norrdahl, and E. Korpimaki. 1994. Avian predation risk modifies breedingbird community on a farmland area. Ecology, 75: 1626-1634.Tarborton, W. R. 1978. Hunting and the energy budget of the black-shouldered kite. Condor,80, 88-91.Taylor, K. 1985. Great black-backed gull, Larus marinus, predation of seabird chicks onthree Scottish Islands. Seabird, 8, 45-52.Taylor, I. R. 1983. Effect of wind on the foraging behaviour of common and sandwich terns.Ornis Scandinavica, 14, 90-96.Taylor, R. J. 1984. Predation. New York: Chapman and Hall.Temeles, E. J. 1987. The relative importance of prey availability and intruder pressure infeeding territory size regulation by harriers, Circus cyaneus. Oecologia, 74, 286-297.Thiel, M. and T. Sommer 1994. Wind-dependent impact of gulls (Larus sp.) on the breedingsuccess of common terns, Sterna hirundo. Ophelia Supplement, 6, 239-251.143Tiebout, H. M. 1991. Daytime energy management by tropical hummingbirds: Responses toforaging constraints. Ecology, 73, 839-851.Tilman, D. 1989. Population dynamics and species interactions. In, Perspectives in EcologicalTheory. ed., J. Roughgarden, and R. M. May. Princeton University Press: Princeton.Trilimich, F. 1978. Feeding territories and breeding success of south polar skuas. The Auk,95, 23-33.van der Meer, J. and T. Piersma. 1994. Physiological inspired regression models forestimating and predicting nutrient stores and the composition in birds. PhysiologicalZoology, 67, 305-329.Village, A. 1982. The home range and density of kestrels in relation to vole abundance.Journal of Animal Ecology, 51, 4 13-428.Wanless, S. & Harris, M.P. 1988. The importance of relative laying date on breeding successof the guillemot, Uria aalge. Ornis Scandinavica, 19, 205-211.Ward, D. 1990. Putting ecology back into optimal foraging theory. Comments TheoreticalBiology, 1, 387-410.Watanuki, Y. 1982. Size selective hunting by slaty-backed gulls, Larus schistisagus, andinfluence on fledging success of black-tailed gulls, L. crassirostris. Journal YamashinaInstitute of Ornithology, 14, 25-34.Watanuki, Y. 1983. Predation and anti-predation behaviour in seabirds on Teuri Island,Hokkaido. Journal Yamashina Institute of Ornithology, 15, 167-174.Watanuki, Y. 1989. Sex and individual variations in the diet of slaty-backed gulls breeding onTeuri Island, Hokkaido. Japanese Journal of Ornithology, 38: 1-13.Watanuki, Y. 1992. Individual diet difference, parental care and reproductive success in slatybacked gulls. Condor, 94, 159-171.Weathers, W. W., W. A. Buttemer, A. M. Hayworth, and K. A. Nagy. 1984. Anevaluations of time-budget estimates of daily energy expenditure in birds. Auk, 101,459-472.Webster, M. D. and W. Weathers. 1988. Effect of wind and air temperature on metabolicrate in verdins, Auriparusfiaviceps. Physiological Zoology, 61, 543-554.Werner, E. E., G. G. Mittelback, and D. J. Hall. 1981. The role of foraging profitability andexperience in habitat use by the bluegill sunfish. Ecology, 62, 116-125.Wiens, J. 1984. Modelling the energy requirements of seabird populations. In, SeabirdEnergetics, Whittow, G. C. and Rahn, H (eds.), Plenum Press, New York.Wijnandts, H. 1984. Ecological energetics of the long-eared owl (Asio otus). Ardea, 72,1-92.Williams, G. C. 1966. Adaptation and natural selection: a critique of some currentevolutionary thought. Princeton University Press, Princeton, New Jersey.144Wittenberger, J.F. & Hunt, G.L. 1985. The adaptive significance of coloniality in birds. In:Avian Biology, Vol.8 (Ed. by D.S. Farner, J.R. King & K.C. Parkes). pp 1-78.Orlando: Academic Press.Ydenberg, R. C. 1994. The behavioral ecology of provisioning in birds. Ecoscience, 1,1-14.Ydenberg, R. C., C. V. J. Weiham, R. Schmid-Hempel, P. Schmid-Hempel, and G.Beauchamp. 1994. Time and energy constraints and the relationships betweencurrencies in foraging theory. Behavioural Ecology, 5, 28-34.Young, E. 1993. Skua and penguin: predator and prey. Cambridge University Press:Cambridge.Zach, R. and J. N. M. Smith. 1981. Optimal foraging in wild birds? In: A. C. Kamil andT. S. Sargent (eds.) Foraging Behaviour: Ecological. EthologicaL andPsychological Approaches. Garland: New York. and Smith 1981Zar, J. H. 1984. Biostatistical Analysis. Toronto: Prentice-Hall.Zaret, T. M. 1980. Predation and freshwater communities. New Haven, Conn: YaleUniveristy Press.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items