Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Factors limiting population growth of non-cyclic collared lemmings at low densities Reid, Donald Grant 1995

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

Item Metadata


831-ubc_1995-060519.pdf [ 4.47MB ]
JSON: 831-1.0088414.json
JSON-LD: 831-1.0088414-ld.json
RDF/XML (Pretty): 831-1.0088414-rdf.xml
RDF/JSON: 831-1.0088414-rdf.json
Turtle: 831-1.0088414-turtle.txt
N-Triples: 831-1.0088414-rdf-ntriples.txt
Original Record: 831-1.0088414-source.json
Full Text

Full Text

FACTORS LIMITING POPULATION GROWTH OF NON-CYCLICCOLLARED LEMMINGS AT LOW DENSITIESbyDONALD GRANT REIDB.Sc.(Hons), The University of Guelph, 1977M.Sc., The University of Calgary, 1984A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF ZOOLOGYWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIASeptember 1995© Donald Grant Reid, 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 LOThe University of British ColumbiaVancouver, CanadaDate 1 3 SM’3 199SDE-6 (2/88)11ABSTRACTI examined factors limiting population growth of collared lemmings(Dicrostonyx kilangmiutak) at low densities (<3 adults/ha), at Pearce Point, NorthwestTerritories, Canada. Populations were followed by mark-recapture, andradiotelemetry. They fluctuated annually, typically with summer declines, and winterincreases.I tested the hypothesis that predation mortality limits population growth insummer, by comparing a population in an 11 ha predator exclosure (PE) with threecontrol populations (18-25 ha). Predation, principally by red fox (Vulpes vulpes) andrough-legged hawks (Buteo lagopus), was the proximate cause of most adult and neonatal mortality. No other mortality factor compensated for decreased predationmortality in PE. Adult survival and recruitment increased significantly in PE.Controls declined in 1990 and 1991. PE declined less quickly. It did not grow, becauseweanlings dispersed long distances (53 mId), and frequently left PP. I conclude thatpredation mortality is sufficient and necessary to limit summer population growth.Three other factors potentially limiting population growth in PP wereinvestigated. Social interactions did not inhibit reproduction because neither theproportion of lemmings reproductively active, nor litter sizes, differed between PEand controls. Food availability did not limit growth because principal foods were notdepleted in PP. net primary production was similar to that in regions wherelemmings irrupt, and enhanced production with fertilization was not consumed bylemmings. Behavioural sensitivity to predation risk appeared limiting becausegrowth rates of neonates were higher in PE than on controls.111Lemmings bred in all winters (1987-92). Variance in rates of winter populationgrowth was significantly explained by a combination of mean daily temperature inautumn, when lemmings change morphology, and an index of cold stress in winter.Summer predation mortality was destabilizing. Specialist and semi-generalistpredators drove lemmings to densities too low for persistence of these predators.Generalist predators continued to limit lemmings in the absence of specialists andsemi-generalists. The summer predator community at Pearce Point, consisting mostlyof generalists or semi-generalists, contrasts with a predominantly specialist predatorcommunity at arctic sites where lemmings irrupt. A relatively diverse prey base,especially including arctic ground squirrels (Sperniophilus parryii), seemed critical forthe maintenance of predators that limit lemmings in summer.ivTABLE OF CONTENTSABSTRACT.iiTABLE OF CONTENTS ivLIST OF TABLES viiLIST OF FIGURESACKNOWLEDGEMENT xiiINTRODUCTION 1CHAPTER 1:LIMITATION OF COLLARED LEMMING POPULATION GROWTHAT LOW DENSITIES BY PREDATION MORTALITYINTRODUCTIONMETHODSStudy areaPopulation densityMortality and litter fatesPredator exciosurePredator populationsStatisticsRESULTSCharacteristics of study gridsPredator communityPredation mortality- adultsPredation mortality- subadultsPredator exclusion- densityEffect of radiocollarsPredator exclusion- survivalPredator exclosure - recruitment and populationJuvenile dispersalTundra volesDISCUSSIONPredation limitationDispersalTundra volesRelevance to population cyclesDO PREDATORS REGULATE LEMMINGS AT LOW DENSITIESIN SUMMER ?INTRODUCTIONgrvth6688811121314141414161922293232374043434647485050CHAPTER 2:METHODS.Study AreaPrey populationsPrey MortalityPrey Reproductive Status and ProductivityPrey GrowthFood LimitationRESULTSDensity Dependent Predation MortalitySocial Inhibition of ReproductionBehavioural Sensitivity to Predation RiskFood LimitationDISCUSSIONDensity Dependence in Predation MortalityInhibition of ReproductionFood LimitationBehavioural Sensitivity to Predation Risk.Other Limiting FactorsCommunity DynamicsPredator RegulationINTRODUCTIONCHAPTER 3:POSSIBLE FACTORS LIMITING WINTER POPULATIONIN COLLARED LEMMINGSMETHODSStudy AreaClimate DataPopulation EstimationWinter Snow DistributionMicrotine Winter NestsRESULTSExtent of winter breedingErmine predationNest DistributionSnow and Temperature RegimesDISCUSSIONErmine PredationSnow DistributionWinter Temperature RegimesOther Factors Affecting Winter Population GrowthGROWTH• . 101ioi105105105107• . 108108• . 109109111111114125125126• . 127V5252535355565659596570778888919395969798128viINTRODUCTIONMETHODSStudy AreaMicrotine PopulationsHabitatsPredator Numerical ResponsesPredator DietsPredator Functional ResponsesRESULTSMicrotine DemographyNumerical Responses of PredatorsPredicting Lemming Body WeightFox dietGrizzly Bear DietRough-legged Hawk Feeding ExperimentRough-legged Hawk DietDiets of Other RaptorsDISCUSSIONGeneralist Predator HypothesisProceduresImpact on the Lemming PopulationThe Specialist-Generalist ContinuumWhat Allows Generalist Predators to Persist?WHY DON’T ALL LEMMING POPULATIONS IRRUPT?INTRODUCTIONCHARACTERISTICS OF POPULATION IRRUPTIONSGEOGRAPHIC PATTERNS IN LEMMING IRRUPTIONSFACTORS LIMITING POPULATION GROWTH AT LOWDENSITIESHOW DO REGIONS OF IRRUPTIVE AND NON-IRRUPTIVEDYNAMICS 7Summer PredatorsSnow DepthFood AvailabilityHYPOTHESESCONCLUSION197197198200207212212224227227229CHAPTER 4:PATTERNS OF PREDATION ON NON-CYCLIC LEMMINGS:THE GENERALIST PREDATOR HYPOTHESIS. 131131133133135136136137144144144148152152173173173179185185190191193196CHAPTER 5:LITERATURE CITED 232viiLIST OF TABLESTable 1.1. Characteristics of the predator exciosure (PE) and three control(Cl, C2 & C3) study grids 10Table 1.2. Numbers and breeding success of the principal lemming predators . . . 15Table 1.3. Fates of resident adult lemmings on control grids (C) with predatoraccess, and in the predator exclosure (PE) 17Table 1.4. Fates of lemming litters initiated on control areas (C) and predatorexciosure (FE) 20Table 1.5. Tests of the null hypothesis that the Dicrostonyx density in the predatorexclosure was not larger than the mean of sample estimates from controls. 25Table 1.6. Instantaneous weekly rates of population change (r) for microtines . . . . 28Table 1.7. Jolly-Seber estimates of average probability of survival 30Table 1.8. Proportion of adult lemmings retrapped after one and two weeks 31Table 1.9. Total adult recruits divided by total first and second litterpregnancies 38Table 1.10. Intensity of recruitment (number of recruits per hectare) 39Table 2.1. Proportion of lemming-weeks in which individual adult-sizedlemmings were reproductively active 68Table 2.2. Litter sizes of primiparous (spring-born), multiparous(overwintering) and first summer litter females 69Table 2.3. Mean ± S.E. post-partum weights of adult female lemmings 78Table 2.4. Mean ± S.E. percent cover (± 0.5%) of principal lemming foods 79Table 2.5. Mean above-ground mass (g/0.125 m2) of vascular plants andlichens, and net above-ground primary production 80Table 2.6. Comparison of estimates of net above-ground primary productionin plant communities at Pearce Point, N.W.T., with estimates for similarcommunities at Canadian arctic sites where lemmings are known toundergo substantial fluctuations in density 84vu’Table 2.7. Mean changes in standing crop (% cover) from August 1990 toAugust 1991 for three collared lemming foods under four differentfertilization treatments 85Table 2.8. Two-way analysis of variance investigating the effects ofherbivore exclosure and fertilization treatments on variancein mean change in standing crop (% cover) of live Drvasintegrifolia from August 1990 to August 1991 86Table 2.9. Mean changes in percent cover of two collared lemming food groupson paired one hectare control and fertilized plotsin two heath communities 87Table 2.10. Absolute abundance of Dryas integrifolia (% cover) at three radiifrom maternal burrows of three females at parturition 89Table 3.1. Numbers of females on study grids in late summer (end of August)and in early summer (early June) 110Table 3.2. Instantaneous weekly rates of change of microtines, microtine nestdensities, and nest occupancy by ermine 112Table 3.3. The association of microtine winter nests with remnant winter snow. . . 115Table 3.4. Mean ( S.F.) instantaneous weekly rates of population change (r)from late August to early June for Dicrostonyx, combined withthree indices of thermal stress which might effect r 122Table 4.1. Numbers of confirmed predation mortalities of radio-collaredadult collared lemmings attributable to individual predatorspecies 147Table 4.2. Numbers of raptor pairs establishing breeding territories,numbers of nests successful and young fledged, and numbers ofadult and weaned juvenile mammalian predators, in relation tomean adult Dicrostonyx density (#/ha) and combined mean adultDicrostonyx and Microtus densities (#/ha) in spring (earlyJune) and summer (early July) on three study grids 149Table 4.3. The linear regressions of live body weight (g) of Dicrostonyx(dependent variable) on lengths (cm) of mandible, ulna, and upper andlower molar tooth rows, and from weight (g) of hair 153Table 4.4. Percent frequency of occurrence of all prey remains in red fox scats. . . 156Table 4.5. Mean ± S.E. biomass (g) per scat for red fox, and percent of totalbiomass of principal prey groups in summers 1990 and 1992 159ixTable 4.6. Mean (± S.E.) biomass (g) per scat for red fox and percent of totalbiomass of principal prey groups in summer 1991 161Table 4.7. Adult and juvenile red fox defecation rates (scats/d) and biomassconsumption rates (g/d) converted to per capita consumption rate ofadult and subadult Dicrostonyx, and total numbers of Dicrostonyx killedper summer (a) 1990 and 1992, and (b) 1991 165Table 4.8. Summary of conditions and results of the captive feeding experimentwith a rough-legged hawk, including proportion of lemmings recoveredin pellets, and pellet casting rate 174Table 4.9. Percent frequency of occurrence of remains in rough-legged hawkpellets 175Table 4.10. Conversion of Dicrostonyx remains in rough-legged hawk pellets toconsumption rates of individual lemmings and lemmingbiomass (g/d) by adult hawks 176Table 4.11. Percent frequency of occurrence of all items in peregrine falconpellets 182Table 4.12. Percent frequency of occurrence of all items in pellets of gyrfalcons,gulls and golden eaglesTable 4.13. Mean (S.D.) number of individual lemming remains per pellet, foravian predators other than rough-legged hawks 186Table 5.1. Summary of data collected by the Canadian Arctic Wild LifeEnquiry, concerning relative stability in lemming abundance, for 1935 to1949, in eleven regions of the Canadian arctic 201Table 5.2. Summary of factors which are sufficient to limit lemming populationgrowth at low densities, and those that are necessary to curtail irruptivepopulation growth 208Table 5.3. Presence (Y) or absence (N) of lemming predators as summerbreeders at sites where lemmings have been intensively studied 213xLIST OF FIGURESFig. 1.1. Jolly-Seber estimates of collared lemming densities in the predatorexclosure (PE) and three control grids (Cl, C2 and C3) 23Fig. 1.2. Jolly-Seber estimates of collared lemming density, and the 95%confidence intervals around these estimates, on the predator exclosure(PE) and three control grids (Cl, C2 and C3) 26Fig. 1.3. Kaplan-Meier survivorship functions for collared lemmings in thepredator exclosure (PB) and on two control grids (Cl and C2) duringsummer 1990 33Fig. 1.4. Kaplan-Meier survivorship functions for collared lemmings in thepredator exclosure (PE) and on two control grids (Cl and C2) duringsummer 1991 35Fig. 1.5. Jolly-Seber estimates of tundra vole densities 41Fig. 2.1. The relationship between percent of adult lemmings killed during thesummer and adult lemming density in spring for two populations: Cl(triangles) and C2 (stars) from 1988 to 1992 60Fig. 2.2. The relationship between percent of resident adult lemmings killedwithin a two-week period and lemming density 63Fig. 2.3. The relationship between litter size and female weight post partum asobserved in a predator exciosure (squares) and on control grids (stars). . . . 66Fig. 2.4. The relationship between total growth rate of litters and femaleweights post partum as observed in a predator exclosure (squares), apredator exclosure with fertilization (triangles) and control grids (stars). . . 71Fig. 2.5. The relationship between mean per capita neonatal growth rate on alitter by litter basis and female weight postpartum as observed in apredator exclosure (squares), a predator exdosure with fertilization(triangles) and control grids (stars) 73Fig. 2.6. The relationship between growth rate of juvenile lemmings and theirbody weights 75Fig. 3.1. Map of the distribution of winter nests (dark points) with respect todistribution of remnant snow in late May (hachured line) on PB grid,spring 1991 116xiFig. 3.2. Profiles of mean weekly snow depth (cm) over the nine month periodsfor winters 1987 through 1990. Data were collected at Clinton Point,approximately 65 km east of the study area 118Fig. 3.3. Profiles of mean weekly snow depth (cm) over the nine month periodfor winters 1990 through 1992. Data were collected at Clinton Point 120Fig. 3.4. Relationship between instantaneous weekly rate of population changeoverwinter and the average of mean daily temperatures (°C)during the autumn (September and October) when Dicrostonyx arechanging from summer to winter morphology 123Fig. 4.1. Mean densities (numbers/ha) of resident collared lemmings (solidline) and tundra voles (broken line) on the three study grids 145Fig. 4.2. Frequency distributions (percent) of maximum diameters of red foxscats collected at (a) the natal den from 29 May to 6 July 1991 (n=83), (b)the natal den from 7-20 July 1991 (n=47), (c) the natal den in August1991 (n=58), and (d) away from the natal den in June and early July1991 (n=23) 154Fig. 4.3. Functional response of adult foxes to adult lemming density 171Fig. 4.4. Functional response of adult rough-legged hawks to adult lemmingdensity 180Fig. 5.1. Map illustrating jurisdictions with arctic tundra (a, Alaska; Y, Yukon;N, Northwest Territories; Q, Quebec; L, Labrador), the eleven regions(numbers) summarized by the Arctic Wild Life Enquiry (Chitty 1950),and, in northern Alaska, the division (dotted line) between the coastalplain and the foothills (after Hartman and Johnson 1984) 202Fig. 5.2. Ranges of (a) red fox (Vulpes vulpes), and (b) arctic fox (Alopex lagopus),in northern North America (from Banfield (1974) and Hall (1981)) 217Fig. 5.3, Ranges in arctic North America of two lemming species complexes, (a)Lemmus , and (b) Dicrostonyx , and ranges of four rodent speciespotentially sympatric with lemmings, and acting as potential alternatefood for red fox: (c) arctic ground squirrel, (d) tundra red-backed vole,(e) tundra vole, and (I) singing vole 221Fig. 5.4. Map of isohyets (dotted lines) of mean total annual snowfall (cm) inarctic North America 225xiiACKNOWLEDGEMENTIt is very inaccurate to call this thesis mine alone. Without the intellectual,emotional and logistic support of many others, it would not exist. The lemming studyat Pearce Point was started by Charley Krebs, Rudy Boonstra and Alice Kenney. Iowe a great deal to their willingness to accept me, inexperienced in small mammalbiology, into the team. Without their encouragement and groundwork, in developingstudy techniques and relevant ideas, this research would not have been possible.Charley and Alice also gathered a great deal of the data in all three field seasons, andAlice analysed the remote sensing data. Able and dedicated field assistance was alsoprovided by Carita Bergman, Maria Leung, Xavier Lambin, Beth Scott, NicholasWilliams, Jason Ruben, Chris Ruben, Marcus Ruben and Joe Ruben. NicholasWilliams ably tackled population estimation of ground squirrels. Thanks to MicheleCherry, Shannon Beglaw, Sand Russell and Anna Reid for patient and exactingassistance in the laboratory.Members of my supervisory committee, Tony Sinclair, Jamie Smith, DolphSchiuter and Fred Bunnell, frequently challenged my thinking, and helped focus myefforts. Many other colleagues were equally stimulating and encouraging; specialthanks to Christoph Rohner, David Hik, Dennis Chitty, Erkki Korpimaki, Nic Larter,Fritz Mueller, Per-Olof Palm and Erik Lindstrom.The Hunters and Trappers Committee of Paulatuk gave me permission tostudy lemmings on Inuvialuit land. Thanks to the Committee members, particularlyPeter Green, Noel Green and Tony Green, for their interest in the research. Theresearch would not have been possible without the logistic support of: Irene Wingateand Dick Cannings of the Zoology Department at U.B.C.; Gary White and Les Kutnyof the Inuvik Research Laboratory, Science Institute of the Northwest Territories;Bruce McLean, Chris Shank, Derek Melton and Paul Fraser of the WildlifeManagement Division, Renewable Resources, Government of the NorthwestTerritories; and the Tuktoyaktuk staff of the Polar Continental Shelf Project, EnergyMines and Resources Canada. Thanks to Bev Day and coworkers at the OrphanedWildlife Rehabilitation facility (OWL) in Delta, B.C., for their enthusiastic cooperationin the rough-legged hawk feeding experiment. Thanks to Dr. Nic Larter, RenewableResources, Government of the N.W.T., Inuvik, for providing lemming carcasses usedin some analyses.Funding was provided by National Science and Engineering Research Council(NSERC) of Canada operating grants to C.J.Krebs, Canadian Wildlife ServiceUniversity Research Support Fund grants, and Department of Indian and NorthernAffairs Northern Scientific Training Grants. I was supported by University of BritishColumbia Graduate Fellowships, the Anne Valleé Memorial Scholarship U991), andthe Arctic Institute of North America Jennifer Robinson Scholarship (1992).My wife, Maria Leung, provided perpetual love and encouragementthroughout this undertaking. Likewise my parents, Ian and Barbara Reid, and mysister, Anna Reid, gave fully of their love and support. Thanks. I remain grateful tothe lemmings themselves, and the other creatures on the arctic tundra at PearcePoint, all of whom offered me wonderful insights into their lives.1INTRODUCTIONEcologists have expended a great deal of effort in attempts to document andexplain the population dynamics of microtine rodents. The central focus of this efforthas been the periodic irruptions or ttcycles’ in the abundance of many vole andlemming species (Elton 1942, Krebs and Myers 1974, Taitt and Krebs 1985, Stensethand Ims 1993). Cycles are multiannual patterns of change in abundance or densitycharacterized by four phases: a low density phase lasting at least one and often twoor three years, an increase phase of rapid population growth generally within oneyear, a peak phase with high densities at least ten times the low and generally lastingless than a year, and a decrease phase of variable duration leading back to a low(Krebs and Myers 1974, Taitt and Krebs 1985). Cycles are not ubiquitous. Theserodents exhibit a variety of dynamics within and between species, ranging frommultiannual patterns of varying period and amplitude, to relatively constant densitieswith annual fluctuations of small magnitude (Taitt and Krebs 1985, Hansson andHenttonen 1985).Two themes are evident in the history of this scientific investigation ofmicrotine population dynamics. First, there has been a persistent debate betweenthose searching for a single ecological factor both sufficient and necessary to explainthe entire cyclic dynamic, and those who hold that multiple factors must be invoked,perhaps individually or in combination, at various stages of the populationfluctuation (Taitt and Krebs 1985, Lidicker 1988). Second, there has been a recurringdifference in orientation between those who investigate extrinsic factors (e.g.,predation, food, weather), and those who investigate intrinsic factors (e.g., social2behaviour, physiological stress, age-structure) (Chitty 1960, Caughley and Krebs 1973,Stenseth and Ims 1993).Collared lemmings (Dicrostonyx spp.) are microtine rodents, considered to be aholarctic species complex. Various populations apparently survived the Pleistoceneglaciations in different refugia, and subsequently colonized most arctic tundra regions(Macpherson 1965, Jarrell and Fredga 1993). Collared and brown lemmings (Lemmusspp.) were subjects of one of the first scientific investigations of cyclic patterns inrodent dynamics (Elton 1942). Because of their dramatic density fluctuations theycaptured the attention of other research efforts (Krebs 1964, Fuller et a!. 1975a,b, Batzliet a!. 1980). However, even early investigations suggested that lemmings did notundergo multiannual density fluctuations in all regions (Chitty and Nicholson 1942,Chitty 1950). At Pearce Point, Northwest Territories, Canada, within one of theregions of more constant dynamics, collared lemming densities fluctuate annuallywithin a small range which is typical of low densities in cyclic situations (Krebs et a!.1995).This thesis is a more detailed, and partly experimental, investigation of factorslimiting collared lemmings to low densities at Pearce Point. I do not directly addressthe two debates mentioned above. However, I implicitly follow a multiple, extrinsicfactor approach, if not in experimental design, at least in the focus of investigation. Ibelieve this is desirable for a number of reasons. First, population regulation is bestdemonstrated by investigating all potential limiting factors to discover which one(s)are necessary to curtail population growth (Krebs 1995). Second, different patterns ofpopulation fluctuation exhibited by collared lemmings in different portions of their3range, suggest that different limiting factors operate either geographically, ortemporally within populations (cf. Hansson 1987). Third, lemmings live in a seasonalenvironment with major changes in predator community and food availabilitybetween seasons. Different limiting factors are therefore likely to operate seasonally.Fourth, theory (e.g., McNamara and Houston 1987), and field experiments (Taitt andKrebs 1983, Desy and Batzli 1989), indicate that two key factors, predation mortalityand food availability, can have interactive effects on population dynamics. Fifth, atrue cycle, though sufficiently described in theory by one driving variable with timelag (May 1976), does not adequately describe empirical patterns of population change.Instead, these patterns seem better described as having two relatively quick changesin dynamics: the initiation of increase from a low, and the initiation of decrease froma peak. I interpret this as an escape from one set of limiting factor(s) in the low, andthe exposure to a potentially different set of limiting factor(s) at the peak.This is a study of factors limiting collared lemming population growth at lowdensities. Peak populations irrupt from such low densities, and a preliminaryassessment of a non-cyclic situation indicates fairly constant low densities (Krebs et a!.1995). This implies that irruptive situations consistently differ from non-irruptivesituations in the changing action through time of one or a few ecological factorslimiting population growth at low densities.This is largely a study of the limiting effect of predation on population growth.The role of predation in both cyclic and non-cyclic situations deserves attention. Fieldstudies on cyclic lemmings demonstrate a strong limiting effect of predationmortality, at least at the beginning of a low phase (Pitelka et al. 1955, Maher 1970,4MacLean et al. 1974). Lemmings in the low phase of a cycle continue to breed, andtherefore have high potential rates of population increase, but this increase is notimmediately realized (Krebs 1964). This is the case at Pearce Point, where lemmingsalso suffer high rates of summer predation mortality (Krebs et a!. 1995). Therefore,predation is strongly implicated as the principal limiting factor at low densities.This thesis contains the following five chapters:1. Limitation of collared lemming population growth by predation mortality.This is an experimental test of the hypothesis that predation mortality is sufficientand necessary to limit collared lemmings at low density at Pearce Point.2. Do predators regulate lemmings at low densities in summer?This chapter addresses predation mortality within two paradigms of populationregulation, the density-dependent and mechanistic paradigms (Sinclair 1989, Krebs1995). It includes an experimental and comparative investigation of the potentialoperation of alternate limiting factors (food availability, reproductive inhibition andpredation risk) on lemming population growth at low density in summer.3. Possible factors limiting winter population growth in collared lemmings.This chapter uses correlation and regression analyses to evaluate the role of erminepredation, autumn temperatures, and winter and spring snow depths, as factorsexplaining variance in overwinter rates of population change.4. Patterns of predation on non-cyclic lemmings: the generalist predator hypothesis.This chapter uses the generalist predator model, which attempts to explain relativestability in prey densities, as a hypothesis to investigate patterns of summerpredation by the predator community at Pearce Point. It includes assessment of the5predators’ numerical and functional responses, and changes in their diets withchanging lemming density.5. Why don’t all lemming populations irrupt ?This chapter develops alternative hypotheses suggesting why some North Americanlemming populations irrupt and others do not. It includes summaries of geographicalpatterns in irruptions, our knowledge of factors necessary to curtail lemmingpopulation growth, and our knowledge of the distribution of predator and preyspecies in arctic tundra.6CHAPTER 1LIMITATION OF COLLARED LEMMING POPULATION GROWTHAT LOW DENSITIES BY PREDATION MORTALITYINTRODUCTIONIn many regions of arctic North America collared lemmings of the genusDicrostonyx exhibit substantial, and often cyclic, multi-annual population fluctuations.These have been reported for D. hudsonius in Ungava (Elton 1942), D. richardsoni innorthern Manitoba (Shelford 1943; Scott 1993) and Eskimo Point (Mallory et a!. 1981),and D. groenlandicus near Baker Lake (Krebs 1964), near Igloolik (Rodgers and Lewis1986), on Devon Island (Fuller et al. 1975b), and on the Alaskan north slope (Batzli eta!. 1980).Ecological factors causing cyclic population changes have not beenexperimentally investigated for Dicrostonyx, but hypotheses cover the full spectrum ofexplanations for microtine cycles (Taitt and Krebs 1985; Stenseth and Ims 1993). Highdensities of Dicrostonyx, often in synchrony with brown lemmings (Lemmustrimucronatus), are generally followed by strong numerical responses of arctic (Alopexlagopus) and red fox (Vulpes vulpes) (Elton 1942; Macpherson 1969), ermine (Mustelaerminea) (Krebs 1964: MacLean et a!. 1974) and a number of avian predators (Pitelka eta!. 1955; Watson 1957; Maher 1970). Predation by nomadic avian and mammalianpredators may synchronize lemming declines over broad geographical regions(Ydenberg 1987; Ims and Steen 1990). Enhanced breeding success by these predators,when lemmings are numerous, increases their numerical response (Watson 1957;Macpherson 1969), and is hypothesized to intensify the depth of the decline in7lemming densities and prolong the phase of low density (one or two years) throughstrong limitation on population growth (Pitelka et a!. 1955; Maher 1967, 1970;MacLean et a!. 1974). Similar explanations for the population dynamics of othermicrotine rodents during the decline and low phase of periodic irruptions have beenprovided by other studies in North America (Pearson, 1966, 1971; Fitzgerald 1977),and Europe (Hansson 1984a; Henttonen 1985; Henttonen et a!. 1987; Korpimaki et a!.1991). Limitation by predation mortality is assumed to wane, as predators declinewith increasingly scarce prey densities, and eventually be insufficient to limitexponential prey population growth.The first detailed study of D. kilangmiutak, at Pearce Point, on the mainland ofthe western Northwest Territories, Canada, revealed persistent low densities (lessthan three per hectare) over three consecutive years, and frequent declines in summerpopulations (Krebs et a!. 1995). Populations often recovered with winter breeding, aspreviously observed by Krebs (1964) and Fuller et al. (1975a). Predation was theproximate cause of at least 66% of summer mortalities in 1988, and 79% in 1989,leading Krebs et a!. (1995) to hypothesize that predation mortality was the most likelyfactor limiting summer population growth.This chapter reports an experimental investigation of the hypothesis thatpredation mortality is sufficient and necessary to limit collared lemmings (D.kilangmiutak) to low densities at Pearce Point. I tested four predictions of thepredation mortality hypothesis: (i) predation is the proximate cause of the greatmajority of mortalities, (ii) predator removal decreases mortality (predation is notreplaced by another source of mortality in a compensatory fashion) and therefore8increases survival, (iii) predator removal enhances rates of population change, (iv)predator removal enhances recruitment. I also report demographic data on the tundravole (Microtus oeconornus), the only other common microtine rodent in the area, tolook for possible patterns shared with lemmings.METHOD SStudy areaThe study area (40 km2) was the vicinity of Pearce Point (69°48’N, 122°40’W),on the south shore of Amundsen Gulf, western mainland Northwest Territories,Canada. The bedrock of dolomites and limestones, interrupted by dikes of basalt, isfrequently exposed in cuffed hills rising to 130 m a.s.l.. Surficial geology has beeninfluenced by Pleistocene glacial scouring and deposition, and by movement of windblown sands in the delta of a local river, draining the Melville Hills to the south.Population densityFour areas of tundra, each 18 to 25 ha and each including a range of availablemicrotine habitats, were chosen for study grids. Each was fairly discrete, beingbordered in part by rocky outcrops or water courses, but partly linked to other areasby continuous similar habitats. Vegetation communities were primarily upland heath(Dryas integrifolia / Carex rupestris / Salix arctica) and mesic hummock (Carexmembranacea / Dryas integrifolia / mosses), and also included some ribbon-like wetmeadow (Carex aquatilis) bordering water bodies (Table 1J). Collared lemmingsoccupied drier habitats, and tundra voles occupied the wetter communities on thisspectrum (Bergman and Krebs 1993).On each area a grid of reference stakes was surveyed, and all microtine9burrows were marked. I quantified the population densities of microtines by live-trapping every one or two weeks from early June (end of snow-melt) to late August(first snow-fall) of 1990, 1991 and 1992, using Longworth live traps, and mark-recapture with ear tags. I used the Jolly-Seber open model for population estimation.In early spring, open traps (c. 60/grid) and nest boxes (c. 12/grid) left over winterwere the best places to find microtines, as most burrows were flooded. For an initialtrapping session, and alternate ones thereafter, microscope slides covered in talcumpowder were placed in all burrow mouths, and examined 18-24 h later for presenceof microtine tracks indicating active burrows (Boonstra et at. 1992). The trappingsession then involved immediately setting traps at these active burrows in themorning, and checking them in the afternoon and again in the evening, when theywere locked open. At alternate sessions, I used trap locations chosen by tracks onslides the previous session.To assess how representative study grids were of regional population patterns,I counted active microtine burrows in one-hectare quadrats located randomly throughthe study area, both on study grids and elsewhere, in early August 1990.Juveniles were rarely caught more than once. Consequently, density estimatesare based on residents alone, defined as adults plus those few juveniles caught atleast twice on the same grid.For each capture, location, weight and reproductive condition (males - testesscrotal or abdominal; females - vagina perforate or non-perforate, lactating or nonlactating, pubic symphysis closed, partly open or open) were recorded.10Table 1.1. Characteristics of the predator exciosure (PB) and three control(Cl, C2 & C3) study grids.PE Cl C2 C3Surface cover (%)Dryas heath 55.2 65.6 67.1 83.7Carex-Dryas hummock 39.6 16.1 22.2 9.0Carex marsh 1.0 0 1.4 0.4Unvegetated 4.3 18.3 9.3 6.9Elevation range (m) 8 3 3 25Distance (m) toNatal fox den 4000 3200 2800 1400Rough-legged hawk nest 200 50 600 240011Mortality and litter fatesAdult Dicrostonyx were fitted with radiotransmitters (Biotrack Inc., model SS-1)mounted to cable ties as neck collars. Radio and harness weighed approximately 3 g.Individuals were relocated every two or three days on grids, or by extensivesearching throughout the study area. This allowed estimation of the date of death towithin two or three days. Radios frequently led to microtine carcasses, remains, ormerely the radios in dens or nests of predators. Cause of death was classified byautopsy on carcasses, or by assessment of lemming remains and disturbance atburrows.The Jolly-Seber model was used to estimate survival rates. I also used theKaplan-Meier procedure for survivorship estimation of radiocollared lemmings, withstaggered entry of individuals during the course of the summer, and progressivecensoring due to radio failure or survival to the end of the study period (Pollock et al.1989). To test the possibility that radio packages compromised survival, I followeddensities on one of the four grids in 1991 (Control #3) without using any radiocollars.The numbers of litters born, and their fates, were assessed indirectly, bymonitoring female reproductive condition at least weekly to estimate date ofparturition, along with records of her movements and burrow use. Lactating femaleDicrostonyx typically use one maternal burrow (Brooks 1993; unpubi. data this study),and juveniles are weaned at 15 to 20 d (Brooks and Banks 1973). Termination oflactation, and extensive movements away from a natal burrow prior to the expectedweaning date, were taken as a failure to wean the litter. To assess potential causes oflitter mortality, I monitored natal burrows at each radiotelemetry check for evidence12of flooding or excavation by predators.In 1992 data on juvenile survival and movements immediately after weaningwere gathered using miniature radios (weight 1 g; AVM Inc., model SM-i) glued to asmall patch of shaved skin on the animal’s back. Radio life was 10-14 days, butradios stayed on those lemmings which were not killed, only for an average 7.3 d(S.E. 0.83, n = 7).Predator exciosureOne of the four populations (grids) was chosen for a predator exclosuretreatment (referred to as PE) to test some of the predictions experimentally. Thistreatment covered 11.4 ha of an 18-ha grid. Border fence posts were erected inautumn 1989. In June 1990 a covering, approximately 2 m above ground, was erectedto exclude avian predators. This consisted of a lattice of 0.6-cm diameter nylon rope,suspended from iron poles 20 m apart. This lattice supported parallel lines, 0.7 mapart, of monofilament nylon fishing line (14 kg test strength) tightly strung betweenthe lines of rope. In late June 1990, a fence of 2.5-cm mesh chicken wire fencing waserected around the periphery of the exciosure. This fence stood 1.5 m above ground,with a ground-level apron of 0.3 m either dug into the ground or covered with mudand rocks. The construction successfully lasted three summers and the twointervening winters, with minor repairs.Mesh size was large enough for microtines to pass through, and theyfrequently did so. I live-trapped daily for arctic ground squirrels (Spermophilus parryii)and ermine, which were taken one to eight kilometres away and released. Sign in13spring indicated that fox and ermine entered the exciosure in winter.Predator populationsNotes were made of raptors prospecting for nest sites in spring, along withtheir degree of territorial defence. A thorough search of all cliffs within the 40 km2area in June gave a complete enumeration of breeding pairs initiating incubation, andtheir clutch sizes. However, I could not reach nests of glaucous gulls (Larushyperboreus), Thayer’s gulls (Larus thayeri), and common ravens (Corvus corax) oninaccessible cliffs. Breeding success was quantified by visiting nests in early July toenumerate numbers of young chicks, and in mid-August to count young just beforeor at fledging.I visited known fox dens biweekly to collect scats, and note pelage and size ofall foxes observed. This allowed identification of individual foxes based on colourpatterns, a count of juveniles, and an estimate of whether or not young were weaned.Estimates of ermine activity were based on casual observations, incidental live-trapping and characteristically eaten microtine carcasses, but an ermine populationestimate was not attempted. In 1991 and 1992 numbers of arctic ground squirrelswere indexed by placing microscope slides covered in talcum powder in a fixedsample of burrow mouths and counting the proportion marked within 24 hours. In1992 all ground squirrels were counted on three, adjacent, one-square-kilometre areasusing live-trapping (Tomahawk traps) at all burrows, coloured ear-tagging, radio-collaring, and periodic observations of the areas. Grizzly bear (lJrsus arctos) activitywas based on tracks and casual observations.14StatisticsThe predator exciosure treatment was not replicated in any one year. To testfor significant deviation of the population parameters in the exciosure from those ofthe three concurrent control populations, I used the modification of the Student-t testcomparing a single observation with the mean of a sample (Sokal and Rohif1981:231).RESULTSCharacteristics of study gridsHummock vegetation was most abundant on PE, and least abundant on C3,but Dryas heath showed the reverse pattern (Table 1.1). The hummock gave moreprotective cover but had a lower density of Dryas food than the heath, so theconsequences of this gradient are unclear.C3 was considerably farther than other grids from a rough-legged hawk nest,but was the closest to the natal fox den (Table 1.1). CI probably received moreintensive raptor surveillance with its proximity to a rough-legged hawk nest.Predator communityThe predator community varied little between 1990 and 1991, but changedmarkedly in 1992 when rough-legged hawks did not breed, and red fox breedingfailed, followed by disappearance of two of three known adults (Table 1.2). In spring1992, the carcasses of three other red foxes were found. They had died during the1991-92 winter, probably from starvation judging by the poor quality of their bonemarrow.15Table 1.2. Numbers and breeding success of the principal lemming predators inthe study area.Number of Number of neststerritorial pairs successful (fledglings)Avian predator 1990 1991 1992 1990 1991 1992Rough-legged hawk 4 6 0 2 (3) 3 (5) 0Peregrine falcon 5 5 6 5(10) 2 (2) 4(10)Golden eagle 1 1 1 1 (1) 1. (1) 1 (2)Gyrfalcon 0 0 1 0 0 1 (2)Number of Number of youngadults weanedMammalian predator 1990 1991 1992 1990 1.991 1992Redfox 2 3 3tol 3 2 0Ermine ? ? ? >1 >3 >2Grizzly bear 1 1 2 0 0 016Eleven resident ground squirrels occupied 3 km2 prior to the emergence of young in1992. The proportion of slides marked in squirrel burrows in four sample areaswithin 24 h in late July declined by 55% from 0.38 (S.E. 0.12) in 1991 to 0.17 (S.E.0.06) in 1992.Predation mortality - adultsPredation was the dominant proximate cause of death of radiocollared adultlemmings in summer, accounting for between 33 and 73% of mortalities on controlgrids (Table 1.3). These are probably underestimates as the fate of 4 to 24% oflemmings and voles remained unknown (Table 1.3).Rough-legged hawks and red foxes were the dominant predators in 1990 and1991 (Table 1.3). Predation pressure was much reduced in 1992 because rough-leggedhawks were absent and foxes scarce. Grizzly bears were occasional visitors in allyears, but excavated lemmings on study grids only in 1991. Ground squirrels werepersistent but infrequent predators. Ermine were not observed on the study gridsexcept in late 1992, when juvenile ermine killed juvenile lemmings in PE (Table 1.4).Ermine depredated a mean of 7.6% (S.E. 0.96) of microtine winter nests on the fourstudy grids in winter 1990-91, and 5.3% (S.E. 2.99) in winter 1991-92.The predator exciosure did not exclude all predation (Table 1.3). Rough-leggedhawks killed lemmings when the nylon was in poor repair or depressed by snow.Ground squirrels dug under the fence, and a grizzly bear got in over the fence. Foxeswere excluded. In 1990, 16 of 23 (70%) predation deaths inside the exclosure occurredbefore it had been completely constructed, and the proportion of lemmings lost topredators did not differ between controls (0.58) and PE (0.58)Table1.3.Fatesofresidentadultlemmingsoncontrolgrids(C)withpredatoraccess,andinthepredatorexciosure(PB), duringsummersof1990, 1991and1992,basedontelemetry.Percentagesinparentheses.199019911992CPE*CPECPBTotalradioed67405225611Totaldying44(66)24(60)40(77)12(48)2(33)1(9)Allpredators39(58)23(58)38(73)11(44)2(33)0(0)Allraptors13(19)13(33)25(48)5(20)1(17)0(0)Rough-leggedHawk7817400PeregrineFalcon011000Unidentifiedraptor647110Allmammals21(31)10(25)9(17)6(24)0(0)0(0)RedFox2185000Table1.3.(continued)199019911992CPE*CPECPEGrizzlyBear002400Ermine000000GroundSquirrel022200Unknownpred.5(7)04(8)01(17)0Naturaldeath4(6)01(2)0(0)0(0)0Accidentaldeath1(2)1(3)1(2)1(4)0(0)1.(9)Unknownfate16(24)5(13)7(14)1(4)1(17)1(9)Dispersed1(2)0(0)0(0)0(0)0(0)2(18)Survived**6(9)11(28)5(10)12(48)3(50)7(64)*Raptorskilled11,andmammals5,lemmingsbeforePEcompleted.Seven(18%)predationdeathswereinPE.**Survivedisthenumberofindividualsstillaliveattheendofthesummer.Co19(Chi-square = 0.005, df = 1, P = 0.943) (Table 1.3). However, by the end of thesummer, the proportion of lemmings still alive was significantly higher in PE (0.28)than on controls (0.09) (Chi-square = 6.446, df = 1, P = 0.011).In 1991, the proportion of lemmings killed was significantly lower in PE (0.44)than on controls (0.73) (CM-square = 6.168, df = 1, P = 0.013), and a significantlyhigher proportion was alive at the end of the summer in PE (0.48) than on controls(0.10) (Chi-square = 14.459, df = 1, P < 0.001) (Table 1.3).In 1992, lemming densities were lower than in 1990 and 1991 (Figure 1.1), andpredator densities declined (Table 1.2). The proportion of lemmings killed oncontrols, and in the exciosure, decreased compared with previous years, but samplesizes were too small for statistical tests (Table 1.3).Predation mortality - subadu itsPredation was the predominant proximate cause of death of lemming litters onall grids; I lacked data on fates of vole litters (Table 1.4). Predators killed littersdirectly, by entering or excavating burrows, or indirectly, by killing the lactatingfemale. Direct and indirect predation on controls accounted for the loss of 37% oflitters in 1990, 52% in 1991 and 14% in 1992.The proportion of litters lost to predators in PE was substantially less than thaton controls in all three years (Table 1.4). The proportion of litters successfully weanedin PE in 1990 was higher than on controls, but not significantly so (Chi-square =2.368, df = 1, P = 0.124). The difference was significant in 1991 (Chi-square = 6.139,df = 1, P = 0.013). The effect of the exciosure appeared weaker in 1992 with a higherTable1.4.Fatesoflemminglittersinitiatedoncontrolareas(C)andpredatorexclosure(PE)ineachsummerstudyperiod.Percentagesinparentheses.199019911992NumberoflittersCPECPECPEInitiatedWeanedLostMotherdied-PredationStormInfanticide-Unknowncause35 14(40)8(23)1(3)1.(3)014 9(64)1(7)0 0 023 7(30)8(35)1(4)0 1(4)38 24(63)0 0 3(8)2(5)7 4(57)0 0 0 08 6(75)0 0 0 0t%Z CTable1.4.(continued)199019911992Numberof littersCPECPECPEPredation-Total5(14)1.(7)4(17)5(13)1(14)2(25)-Fox5(14)1.(7)3(13)000-Bear001(4)4(11)00-Groundsquirrel0001(3)1(14)0-Ermine000002(25)Unknownfate6(17)3(21)2(9)4(11)2(29)0FL22proportion of control litters successfully weaned, but sample sizes were too small forstatistical testing.Other known causes of litter mortality included rain storms flooding burrows,thereby forcing mothers to bring neo-nates above ground, and infanticide asevidenced by incisor wounds and partial consumption of the body, starting with thecranium.I have few data on fate of lemmings from weaning to sexually active adultsize. In 1992, when predator densities were relatively low, seven of fifteen weanlingsradioed in the exclosure were killed by predators before their radios fell off. Six ofthe seven, representing three different litters, were killed by ermine at or close to thenatal burrows. The seventh was killed by a raptor well outside the exclosure.Predator exclusion- densityThe density of Dicrostonyx in PE remained relatively high during summers1990 and 1991, and diverged from the rapidly declining densities on controls (Fig.1.1). The null hypothesis that lemming density in PE was not larger than the mean ofcontrol densities holds for estimates at the beginning of summers 1990 and 1991, butcan be rejected for estimates at the end of the two summers (Table 1.5). By latesummer PE densities were significantly higher than the mean of control densities(Fig. 1.2). The instantaneous weekly rate of population change over the summerperiod was higher in FE compared with the mean of controls in both 1990 and 1991,but not significantly so (1990: t = 1.46, df = 2, 0.10 < P < 0.20; 1991: t = 1.69, df = 2,0.10 <P < 0.20) (Table 1.6).23Fig. Li. Jolly-Seber estimates of collared lemming densities in the predator exclosure(PE) and three control grids (Ci, C2 and C3) over the three summer studyperiods. Densities in winter (shaded bars) are interpolated and exact temporalpatterns are unknown. Vertical arrows indicate weeks when densitycomparisons in Table 1.5 were made.C,> (ID C ci) cm ci)ci) UD > 0Time(weeks)junejulyaugjunejulyaugjunejulyaug25Table 1.5. Tests of the null hypothesis that the Dicrostonyx density in the predatorexclosure was not larger than the mean of sample estimates from control grids(d.f. = 2 in all cases; one-tailed test)1990 1991 1992Beginning * t = 0.85 t = 1.97 t = 9.15of summer 0.40 < P < 0.50 0.05 < P < 0.10 P < 0.02End * t = 4.39 t = 8.86 t = 7.68of summer P < 0.05 P < 0.02 P < 0.02* Weeks for which comparisons were made are indicated on Fig. 1.1.Fig. 1.2. Jolly-Seber estimates of collared lemming density, and the 95% confidenceintervals around these estimates, on the predator exclosure (PE) and threecontrol grids (Cl, C2 and C3), in mid-August 1990, and early August 1991.26272.2PE2.01.8PE1.6vuc- 1.4a,1.2 -I—>_10(/)zO.8C3-0.6-T C2 C21.0.4clii02cii. C3 I1990 199128Table 1.6. Instantaneous weekly rates of population change (r) for microtine species onthe predator exclosure (PE) and three control grids (Cl, C2, and C3). N is theperiod in weeks over which the estimates were made. In blank cells the specieswas absent or at densities too low for estimation.Dicrostonyx Microtus1990 1991 1992 1990 1991 1992N= 9 10 11 9 10 11PE -0.03 -0.03 0.04 - - -Cl -0.18 -0.20 - -0.12 -0.14 0.13C2 -0.07 -0.11 0.10 - - -C3 -0.19 -0.29 0.06 0.12 -0.05 0.1429In 1992 the situation was different. Few lemmings survived the previouswinter, and two control grids had no lemmings in spring. The density in PP wassignificantly higher than the mean of controls both at the beginning and end of thesummer (Table 1.5). In 1992 instantaneous weekly rates of population change werepositive in PE and on all controls; populations were able to grow in the relativeabsence of predators (Table 1.6; Fig. 1.1).The mean density of active microtine burrows on control grids in August 1990(1.0 / ha; n = 6 quadrats) was not significantly different from the mean density onquadrats located randomly through the study area (0.6 / ha; n = 11 quadrats) (Mann-Whitney U = 41.5, F> 0.10), indicating that control grids adequately representedmicrotine densities in the study area.Effect of radiocollarsThe Jolly estimates of average probability of survival over four weeks forresident lemmings on C3, without collars in 1991, was higher than survival of radioedresidents on other controls in 1990 and 1991, but fell within or close to the 95%confidence intervals for other estimates (Table 1.7). This difference was not significantin 1991 (t = 2.33, df = 1, 0.20 <P < 0.40, one-tailed test). The proportion of adults onC3 (without collars), retrapped one and two weeks after initial capture, was slightlyhigher than proportions on Cl and C2 (with collars), but differences were notsignificant (Table 1.8). The trends in the data indicate that radiocollars slightlycompromise lemming survival in the first week after receiving the collar. However,densities on C3 declined in a similar fashion to those on other control grids over the30Table 1.7. Jolly estimates of average probability of survival over a standardfour week period for resident microtines (both sexes combined) on thepredator exclosure (PE) and three control grids (Cl, C2 and C3).Extreme values of 95% confidence intervals are in brackets. N is theperiod in weeks over which the data were available. In blank cells thespecies was absent or at densities too low for estimation.Dicrostonyx Microtus1990 1991 1992 1990 1991N= 8 7 12 8 7FE 0.62 0.66 0.79 - -(.47-.76) (.51-.81) (.50 -1.00)Cl 0.33 0.25 - 0.33 0.45(.19-.47) (.05-.46) (.09-.57) (.05-.96)C2 0.35 0.34 - - -(.19-.52) (.17-.51)C3 0.25 0.47 - 0.63 0.44(.00-.50) (.30-.65) (.00-1.00) (.26-.63)31Table 1.8. Proportion of adult lemmings retrapped after one and after two weeksfollowing initial capture, on Cl and C2 (with radiocollars), and C3 (withoutradiocollars) in 1991. N is the number of individuals. Test statistic (Chi-square)tests the null hypothesis that C3 does not differ from Cl, and C3 does notdiffer from C2.Proportion Cl C2 C3retrapped after (N=16) (N=29) (N=28)One week 0.56 0.52 0.61(X2 = 0.084, (X2 = 0.468,P = 0.77) P = 0.49)Two weeks 0.38 0.41 0.43(X2 = 0.121, (X2 = 0.013,P = 0.73) P = 0.91)32full course of the 1991 summer (Fig. 1.1; Table 1.6). I conclude that any negativeinfluence of radio packages on lemming movements and survival is short-lived, anddoes not significantly affect summer population processes.Predator exclusion - survivalThe Jolly estimator of the average probability of survival over four weeksindicates that resident lemmings survived longer under the predator exclosure thanon the three control grids (Table 1.7). This pattern was significant in 1990 (t = 4.803,df = 2, P < 0.05), but not in 1991 (t = 2.118, df = 2, 0.05 <P < 0.10).The Kaplan-Meier estimates showed no significant differences in lemmingsurvivorship curves between PE and Cl (log-rank test statistic = -0.220, P = 0.8), orPE and C2 (log-rank test statistic = -1.499, P = 0.13) in 1990 (Fig. 1.3). Howeversurvivorship in PE stabilized mid-way through July, when construction wascompleted, indicating that the exclosure had a strong effect when it operated (Fig.1.3).In 1991 the Kaplan-Meier lemming survivorship curves for PE and Cl differedsignificantly (log-rank test statistic = -3.432, P < 0.01), as did those for PE and C2(log-rank test statistic = -3.205, P < 0.01)(Fig. 1.4). The exciosure enhancedsurvivorship of adult resident lemmings, and survival in PE was significantly higherthan either in Cl (z = 3.09, P < 0.01), or C2 (z = 2.83, P <0.01), by the end of thesummer (Fig. 1.4).Predator exclosure - recruitment and population growthOnly juveniles from the first and second summer litters would have grownlarge enough to be recruited as adults during the study periods. Recruits to the adult33Fig. 1.3. Kaplan-Meier survivorship functions for collared lemmings in the predatorexclosure (PE) and on two control grids (Cl. and C2) during summer 1990.34PA!AJfl J0U)L)c:1).QHw-c,’JaOO• 0 CD LO C’*J i- 000000000035Fig. 1.4. Kaplan-Meier survivorship functions for collared lemmings in the predatorexclosure (PB) and on two control grids (Cl and C2) during summer 1991.ProbabilityofSurvival000000000IIIIII,x___-_,——-I —//./..—.....I,I’1’CD./3.17cC( C!)//t3iE)CX000937populations were uncommon on all grids (Table 1.9), suggesting substantial mortalityduring dispersal. More juveniles recruited per pregnancy where pregnancies werefew (i.e. where densities were particularly low) (Table 1.9). Perhaps open spaceencouraged settlement, or predation pressure diminished at very low densities. Evenin these circumstances, numbers of recruits per litter were still low, and there musthave been substantial mortality, or emigration of juveniles, prior to recruitment.Recruitment of lemmings to the adult population, after the weaning of the firstsummer litter, was significantly higher on a unit area basis in PE compared tocontrols in all years (Table 1.10). The origin was uncertain for 21 of 25 (84%) recruitson control grids, and 10 of 15 (67%) of recruits to PE. Two recruits to PE were knownto come from Cl. Only three of 15 (20%) recruits to FE, and four of 25 (16%) recruitsto controls, were caught as subadults on the same grid. These data again suggest thatjuveniles often disperse beyond the population scale chosen for this study. Despiteimproved recruitment to PE, this protected population did not grow in 1990 or 1991.Juvenile dispersalTen radioed juvenile lemmings were followed for a mean of 5 d after weaning(S.E. 1.13) until their radios fell off (n=6) or they were killed by a predator (n=4). Allten left the natal burrow within two days of weaning. Rates of travel were highlyvariable, and appeared slow in the first few days after weaning, and sped up forsome individuals thereafter. Since they were followed for varying periods, data ondistances travelled may be somewhat biased. However, the straight-line distancesfrom natal burrow to last location with a radio gave a mean daily distance travelledof 53.1 (S.E. 17.6, n = 10) m/d. Four of nine weanlings radioed inside the exclosure38Table 1.9. The ratio of toW adult recruits divided by total first and secondlitter pregnancies (raw data in parentheses) on the predatorexclosure WE) and two control grids (Cl and C2).Grid 1990 1991 1992PB 0.38 (5/13) 0.17 (5/29) 1.00 (5/5)Cl 0.07 (1/15) 0.50 (3/6) -C2 0.29 (4/14) 0.33 (4/12) a3.00 (3/0)39Table 1.10. Intensity of recruitment (number of recruits per hectare)on predator exclosure (PE) and three control grids(Cl, C2 and C3). Sample sizes of recruits in parentheses.Statistics are for the t-test of the null hypothesis thatrecruitment on PE does not exceed mean recruitment oncontrol grids (one-tailed, d.f.= 2).Grid 1990 1991 1992PE 0.44 (5) 0.44 (5) 0.44 (5)Cl 0.05 (1) 0.16 (3) 0C2 0.22 (4) 0.22 (4) 0.16 (2)C3 0.08 (2) 0.24 (6) 0.08 (2)t 3.076 4.755 3.896P <0.05 <0.05 <0.05—40had left it before their radios fell off, and all four were followed for at least five days.We gathered daily movement data for four weanlings over periods of fivedays or more. The mean distances between daily radiolocations for three males were6, 51, and 126 m, and for one female, 174 m. These juveniles travelled in remarkablystraight lines; the sum of daily distances travelled divided by straight-line distancefrom natal burrow to last location gave values of 1.39, 1.45, 1.38 and 1.20 respectively.The longest daily distance travelled was 600 m.Tundra volesTundra voles existed in appreciable numbers on only two grids (Cl and C3),and were occasional residents on the two other grids (C2 and PB). Vole densitiesdeclined on one grid in 1990, and both grids in 1991 (Fig. 1.5), following a similarpattern to lemmings. However, they exhibited slower rates of decline than lemmings(Table 1.6), and the vole population on C3 actually increased in 1990. In 1992, bothvole populations increased as did lemmings (Fig. 1.5; Table 1.6). The 1992 increase onCl, from zero in spring, resulted from immigration of maturing adults born in thefirst summer litter. Winter density changes in voles were similar to those inlemmings, with population increases in 1990-91, and declines in 1991-92 (Fig. 1.5).Adult voles were radiocollared only in 1990. Seven of nine radioed adults were killedby predators (rough-legged hawk, red fox and golden eagle), and two had unknownfates.The Jolly estimator of vole survival on control grids was slightly higher thanthat for lemmings on the same grids (Table 1.7). However, vole survival exceeded thehighest observed lemming survival only on C3 in 1990 (Table 1.7). Average41Fig. 1.5. Jolly-Seber estimates of tundra vole densities on two grids (Cl and C3) overthe three summer study periods. Densities in winter (shaded bars) areinterpolated, and exact temporal patterns are unknown.JoHy-SeberDensity(#/happpppppp row-c,ncn_IIIIC.D0*CDUJ/—-Hcrm(D< 7VtDtICCDCo0CCDCo-L43probabilities of survival for adult voles without collars in 1991, were intermediate to,and not significantly different (t = 0.22, F> 0.50) than values in 1990, with collars(Table 1.7).DISCUSSIONPredation limitationThere are various approaches for investigating whether predation mortality isa sufficient condition to limit prey population growth. Protecting a population byexperimentally excluding predators, and comparing the demographic responses ofprotected prey with unprotected prey, is the most insightful approach (Krebs 1988),and the one followed here. With experimental reduction of the resident predators’access to lemmings, there were significant reductions in predation mortality, and totalmortality, for adult and neo-natal lemmings. Also with decreased predator access,there was a significant increase in adult lemming survival, and significantly enhancedrecruitment of maturing juveniles into the adult population. Consequently, theprotected population diverged significantly from control populations in its densitytrajectory in 1990 and 1991. Overall I confirmed three of four predictions of thehypothesis that predation mortality is sufficient to limit collared lemmings to lowdensities. The only prediction not satisfied was growth in the protected population.The results also indicate that predation mortality is a necessary condition forlimitation of collared lemmings at low densities during summer. First, no othermortality factor compensated for the reduction of predation mortality in the protectedpopulation. Second, the virtual absence of the principal lemming predators, roughlegged hawks and red foxes, in summer 1992, was, in effect, a partial regional44predator exclusion, and coincided with growth in all populations, in contrast todeclines in other years. Remaining predators continued to keep this growth low. I donot know how much growth might occur before another limiting factor wouldoperate, making predation mortality an unnecessary condition.The lack of population growth in PE in 1990 and 1991 requires explanation. Ifpredation mortality is sufficient to limit lemmings to low densities, experimentalremoval of such limitation should result in strong population growth. Such growthwould be contingent on the following conditions: (i) the experimental setup protectsall population processes from predation and thereby adequately mimics regionaldeclines or extinctions of predators; (ii) there is no other limiting factor sufficient tocurtail growth at low densities; (iii) sufficient time has elapsed for theoreticallyexpected time lags (May 1976) to pass. I doubt that each of these conditions wassatisfied, and so discuss each one in detail.First, the exclosure did not keep out all predators, and was too small toencompass dispersal distances of many weanlings. Consequently potential populationgrowth was curtailed by continued mortality of resident adults, by predation onweanlings leaving the protected area, and by predation on subadults outside theprotected area which might have immigrated and been recruited to the protectedpopulation. I believe that these conditions explain the lack of growth in the exclosure.In summer 1992, with a regional decline in resident predators, two controls and PEshowed limited population growth. This resulted from immigration of maturingjuveniles born in early summer litters. These immigrants in turn reproduced. Thissuggests that a predator removal experiment of sufficiently large scale would lead to45population growth.Second, this is only a study of summer population processes. In winter,limiting factors other than predation mortality might operate. The higher lemmingdensities in PE compared with controls in spring of 1991 and 1992, indicated that PEwas at least partly effective in reducing predation mortality during the precedingwinters. However, overwinter rates of population change on all grids were positive in1990-91 and negative in 1991-92, despite the fact that ermine and red fox bred in both1990 and 1991, and the proportion of microtine nests depredated by ermine did notdiffer significantly between these winters (Chap. 3). Obviously predation continued inwinter but was insufficient to curtail population growth in both winters.Overwinter population growth results from breeding, which occurred in eachyear of this study, and is common in Dicrostonyx (Krebs 1964; Fuller et at. 1975a).Perhaps other factors, such as an effect of habitat availability or thermal cover onsurvival and reproductive success, operate in winter, Whatever the case, winterbreeding is necessary for recovery from heavy summer declines, and is likelyessential for long-term population persistence, as has been indicated by Fuller et a!.(1975a).Third, a time lag of approximately nine months in a first-order driving variablegenerates stable limit cycles of three or four year periodicity (May 1976). A likelyplace for such a time lag to operate is in the low density period. Such a lag has beenidentified in some Scandinavian microtine species, but without clear evidence of thecontributing factor(s) (Hornfeldt 1994). A possible cause is prey behaviour. Underheavy predation risk, theory suggests that individual prey must trade-off food46acquisition with survival (McNamara and Houston 1987). If such behaviouraldecision-making is not flexible, there may be a time lag as successive generations ofprey learn that the world is safer. This may be significant in a predator exclosureexperiment for a prey species where risk sensitivity, expressed in depressed rates ofgrowth, maturation or reproduction, has population consequences. For example, withavian predators flying over an exclosure, prey likely still receive stimuli inducing riskavoidance.I have insufficient data to address this possibility. However, observations of Ci)juveniles maturing and breeding during the summer of birth, (ii) litter sizes at PearcePoint comparable to those reported elsewhere (Chapter 2, Krebs 1964), and (iii)population growth in PE in 1992, all suggest that risk sensitivity does not stronglyaffect potential population growth.DispersalThe distances travelled by juveniles within 10 days of weaning are noteworthy,as natal dispersal distances are generally thought to be less than one or two hundredmetres in microtine rodents (Madison 1985; McShea and Madison 1992). The dispersaldistances in this study at low density, traversed many home-range sized areas ofunoccupied habitat (unpubl. data). Open habitat is therefore not the only conditionnecessary for settlement. Perhaps other necessary conditions are the presence of aconspecific of the opposite sex, and the absence of a competitively dominantconspecific of the same sex.The factors inducing natal dispersal in lemmings are not fully understood. Theshort time between departure of mothers and departure of weanlings from natal47burrows, and the distances weanlings of both sexes travelled, suggest inherentavoidance of local settlement. They have perhaps evolved to avoid inbreeding, or toavoid immediate competition with an adult conspecific of the same sex, especially inmales, whose adult home ranges and frequencies of movement are substantiallygreater than those of females at low densities (Brooks 1993). They may also reducerisk of attack by predators (which acquire knowledge of lemming movements andresidence by sight and smell), as has been suggested by Brooks (1993) for adultfemales.Tundra volesSimilar patterns of population change in voles and lemmings suggest thatsimilar factors limit their populations. Limited data indicate that predation on tundravoles in summer is severe enough to curtail their population growth. However thislimitation is less severe than that on lemmings. Tundra voles are more agile, and livein wetter habitats with more vegetative cover than habitats typically occupied bylemmings. The increase in winter 1990-91 might reflect unmeasured late summer andautumn reproduction in 1990, rather than winter breeding. Vole breeding appeared tolast longer in autumn than lemming breeding. No juvenile-sized voles were caught inspring, but the population in spring did include adult-sized females which had notbred.Reported vole densities were substantially less than lemming densities. Thiswas an artefact of measuring density over all available habitats. Within the wettervegetation communities used by voles, their densities were similar to or higher thanregional lemming densities.48Relevance to population cyclesThe preliminary study by Krebs et a!. (1995), and data presented here, indicatethat lemmings at Pearce Pt. have persisted at densities of less than three per hectarefor six years. I conclude that these populations do not exhibit multiannual cyclicitybecause density fluctuations are annual, with highest densities (one to three perhectare) frequently recurring in successive years. The highest densities wereapproximately an order of magnitude less than peak densities following an irruptionin other studies: 40/ha (Shelford 1943), 15-25 (Brooks and Banks 1971), and 27 (Batzliet al. 1980).The data support the hypothesis of Pitelka et a!. (1955), Maher (1970) andMacLean et al. (1974), that predation mortality can prolong the period of low densityin lemmings, at least for a year or two. However, lemmings at Pearce Pt. areapparently unable to escape this persistent limitation, a fact which needs explanation.First, winter breeding may be insufficient to make up for summer declines andongoing winter losses. Consequently, spring densities might never excede levels thatmigratory, hibernating and resident active predators can limit during the subsequentsummer. Autumn densities are so low that winter density increases likely result frommultiple litters during the winter. These issues are addressed in Chapter 3.Second, Dicrostonyx habitats at Pearce Pt. are comprised entirely of prostratespecies, and lack the the bushy willow (Salix spp.) and birch (Betula spp.) growth ofother regions. This reduced cover increases vulnerability to predation.Third, the summer predator community at Pearce Point includes manygeneralist predators (grizzly bear, arctic ground squirrel, golden eagle, peregrine49falcon, gyrfalcon, common raven and glaucous gull) which rarely occur together inother tundra regions. Densities of these generalist predators changed little whenlemmings were rare in spring 1992. I presume that they broadened their prey base toinclude such species as arctic ground squirrel, or merely concentrated more on theirother foods, such as plants and birds. Only the rough-legged hawk appeared to be aspecialist lemming predator, failing to settle and breed in 1992. Some red fox andermine, though generally thought of as microtine specialists, persisted whenmicrotines were scarce, probably by broadening their prey base. These issues areaddressed in Chapter 4.50CHAPTER TWODO PREDATORS REGULATE LEMMINGS AT LOW DENSITIES IN SUMMER?INTRODUCTIONCollared lemmings (Dicrostonyx spp.) and brown lemmings (Lemmustrimucronatus) exhibit wide fluctuations in population density across arctic NorthAmerica (Elton 1942, Shelford 1943, Krebs 1964, Fuller et a!. 1975b, Batzli et a!. 1980).As densities fall from peak numbers, predation mortality may speed the rate ofdecline, and may push densities to lower levels, increasing the amplitude of thefluctuation (Pitelka et a!. 1955; Maher 1967,1970; MacLean et a!. 1974). The strongnumerical response of predators to high lemming densities, and the lagged responseby some predators, may prolong the low density phase in the population fluctuation(Pitelka et a!. 1955; Maher 1967,1970; MacLean et a!. 1974).Collared lemmings (Dicrostonyx ki!angmiutak) at Pearce Point in the westernmainland Canadian arctic persisted at densities of less than three per hectare, for atleast six years. Density frequently declined in summer and increased in winter (Krebset a!. 1995, Chapter 1). Predation mortality was deemed both sufficient and necessaryto curtail summer population growth and cause summer population declines(Chapter 1). Adult and neo-natal survival were significantly enhanced in a predatorexclosure, and no other mortality factor compensated for the reduction in predationmortality (Chapter 1). However, the lack of growth in the protected populationindicated that other limiting factors might be operating at low densities.Lemming population growth could be limited by numerous factors other than51predation mortality (Stenseth and Ims 1993). Reproduction may be decreased orinhibited by spacing behaviour (Chitty 1960, 1967), infanticide (Mallory and Brooks1978), stress (Christian 1950, 1978, Boonstra 1994), or risk of predation (Ylonen 1994).Food quantity (Batzli 1983, 1985), and quality in terms of chemical defences (Jung andBatzli 1981), may affect maturation rate, litter size or social interactions. Habitatheterogeneity may limit dispersal success (Stenseth 1986; Ostfeld 1992). Risk ofpredation while feeding may limit individual growth and maturation (Desy andBatzli 1989).To determine whether or not predators regulate lemmings, I must address twoparadigms of population regulation. The density-dependent paradigm involves asystematic search for relationships between demographic parameters and populationdensity (Sinclair 1989, Murdoch 1994). In this chapter I investigate the possibility ofdensity-dependent regulation through predation, by assessing per capita mortalityfrom predation as a function of lemming density (cf. Sinclair 1989). The mechanisticparadigm of population regulation holds that regulation can only be inferred whenall plausible alternative limiting factors have been investigated (Krebs 1995). In thischapter I investigate possible limitation of population growth, both in the predatorexclosure and in unprotected areas, through three alternative factors: socialinteractions inhibiting reproduction, predation risk, and food availability.I infer that social factors were not inhibiting reproduction in the predatorexclosure if: (i) the proportion of adults reproductively active was not lower on PEthan on controls, (ii) litter sizes were not lower on PE than controls, and (iii) therewas no significant relationship between lemming productivity and density (a measure52of social crowding).I infer that behavioural sensitivity to varying degrees of predation risk was notlimiting population growth in the exciosure if: (i) litter sizes in PE were not largerthan those in areas exposed to predation; (ii) growth rates in PE were not higher thanthose in areas exposed to predation.I infer that food quantity was not limiting if the following conditions hold: (i)lemmings in PB did not deplete the standing crop of their principal foods, Dryasintegrifolia, Salix spp. and Carex spp. (Bergman and Krebs 1993), more than lemmingsoutside the exclosure, (ii) there was no difference between net above-ground primaryproduction at Pearce Point compared with other similar habitats where Dicrostonyxfluctuates markedly in density, (iii) enhanced primary production by fertilization,significantly increased the standing crop of foods compared with untreated tundra,(iv) changes in food abundance following enhancement of primary production werenot significantly greater in a herbivore exclosure compared with changes outside theexclosure, and (v) pregnant females did not choose maternal burrows with highersurrounding food availability than that at their previous maternal burrow.METHODSStudy AreaThe study was conducted at Pearce Point (69°48’N, 122°40’W), on the southshore of Amundsen Gulf, western mainland Northwest Territories, Canada. Thedominant vegetation communities are: (i) an upland heath of Dryas integrifolia withvarying amounts of Carex rupestris, Kobresia sp., Salix arctica, and Draba spp.; (ii) a53mesic hummock community dominated by D. integrifoha, C. membranacea, C. atrofusca,C. misandra, S. reticulata, and S. arctica; (iii) a wet marsh dominated by C. aquatilis,and Eriophorum angustifolium.Prey populationsFour areas of tundra, each 18 to 25 ha, were chosen for detailed study oflemming demography. Each area (termed a grid) had reference stakes at 30 mintervals, allowing accurate location of traps and lemmings. An 11.4 ha predatorexclosure (referred to as FE) was built on one grid in 1990. This was largelysuccessful in reducing mammalian and avian predators in summers 1990-1992(Chapter 1). Three grids were kept as control populations (Cl, C2 and C3).Densities of collared lemmings were estimated by mark-recapture using theJolly-Seber open population model. Grids were live-trapped with Longworth trapsevery one or two weeks from early June to late August, each year from 1988 to 1992(Krebs et al. 1995, Chapter 1). Because subadults were relatively difficult to trap anddispersed rapidly (Chapter 1), densities were calculated only for residents, defined asadult lemmings and subadults caught at least twice.Prey MortalityAdult lemmings (>35g) were fitted with a miniature radio (Biotrack Inc., modelSS-1) mounted on a cable-tie collar (package weight 3.0 - 3.5 g). Lemmings wererelocated every two or three days. Causes of death were determined from: radiolocation at predator nests, dens or burrows; predator sign such as tracks andexcavations at the lemming’s burrow; lemming remains such as parts of skeleton, peltor abdominal organs near the recovered radio; predator teeth marks on the radio54package; and whitewash or droppings near the radio package. The fate of 18% oflemmings on control grids remained unknown, as the radio was permanently lost(Chapter 1). For analyses presented here, I used only verified predation mortalities.Mortality data from two study grids (Cl and C2) were used for all investigations ofdensity dependence. C3 was excluded because adults were not radio-collared in 1991.I investigated density dependence in predation mortality in two ways. Forcomparisons between summers, I calculated percent adult lemmings killed on a gridduring the entire summer as a function of the lemming density on the grid at thebeginning of the summer (early June of 1988 through 1992). This addressed thecumulative effect of predators over the summer, but data were biased towards thehigher densities most common in early June (Chapter 1). For comparisons withinsummers, and at lower densities, I tallied the percent of resident adults falling preyon Cl and C2 in successive two-week periods through the summers as a function ofdensity in the first of the two weeks. Numbers of two-week periods varied betweenyears: 1988 (4), 1989 (4), 1990 (5), and 1991 (6). In 1992 lemmings were absent fromCl and C2 for most of the summer.Data points on a grid within a summer were not independent. To obtainindependence I randomly sampled one two-week period per grid (two grids) persummer (four summers), giving eight data points for each analysis. Ten analyseswere performed to estimate means and variances of regression coefficients describingthe linear relationship between percentage killed and density. Geometric meanregressions were also calculated (Ricker 1984), because dependent as well as55independent variables were subject to measurement error, whereas least squaresregression assumes no error in measurement of the dependent variable (Krebs 1989).Prey Reproductive Status and ProductivityAt each capture, and at many radiolocations between captures, lemmings werehandled and weighed, and their reproductive status was assessed. Males were classedas reproductively active (testes scrotal) or inactive (testes abdominal). Females wereclassed as reproductively active if at least one of the following conditions held:vagina perforate (oestrous), pregnant (based on palpation, weight gain, or backcalculation from birth date), or lactating (nipples large and white). Birth history wasclassified based on the pubic symphysis: closed (non-parous), slightly open (parous),or open (birth imminent or very recent). Pelage characteristics allowed me todifferentiate subadults from adults. In early June the population consisted oflemmings with either adult or subadult pelage. The latter were classed as spring-bornindividuals, having been born under the snow in April or May (Krebs 1964).Lemming productivity was measured as the litter size within four days ofbirth, as judged by the reproductive status and history of the female. Natal burrowswere not excavated. The sample comprised litters born in traps, in plywood nestboxes which were permanently installed on all study grids (c. 0.5 - 1.0 / ha), or intussocks above ground.The hypothesis that lemming productivity is a function of lemming densitywas tested by correlating each observed litter size with the estimated Jolly-Seberlemming density on the grid in the weeks of birth and conception, which occurs 20 to22 days before birth (Hasler and Banks 1975).56The proportion of the adult population reproductively active was taken as thesum of lemming-weeks in which marked adults were reproductively active, dividedby the total number of lemming-weeks for which data were gathered, over a 13-weekperiod from the beginning of June to end of August in summers 1990-92. These dataare biased towards early summer because lemmings were more abundant in June andpopulations declined during the summers.Prey GrowthLitter growth was the total weight gain per day, for all neo-nates combined,determined over at least six days starting on or after day five post-partum, andending close to weaning (days 15 to 17 post-partum). The sample includes two littersborn and raised in nest boxes in small predator exclosures (1,250 m2 each) onfertilized tundra. Data are limited because females rarely kept litters above ground.Juvenile growth was calculated as the mean weight gain per day forreproductively inactive, weaned lemmings less than 35 g over periods ranging from 3to 21 days.Adult male growth was taken as the mean change in weight per day over lessthan 21 d, not including the first week after receiving a radiocollar. Only one datapoint per male was used in the analysis. Adult female weights vary greatly withpregnancy. I used weights taken within six days after parturition.Food LimitationThe investigation of food limitation follows the five conditions outlined in theIntroduction.Food depletion: Changes in abundance of principal lemming foods were57calculated by comparing visually-estimated percent cover (± 5%, 2.5% for cover <5%)of these foods in randomly located 0.5m x 0.5m plots between mid-August 1990 andmid-August 1992. Sampling areas were inside PE, on Dryas heath communities on Cland C2, and on Dryas heath immediately adjacent to but outside PB.Net above-ground primary production: Changes in above-ground standingcrop of all vascular plants and lichens in the upland Dryas/Carex heath and in themesic Dryas/Carex hummock communities were measured using clip-plots (Wein andRencz 1976; Svoboda 1977). Within one small homogeneous portion of eachcommunity a 6.25 m by 10.0 m grid was established for random location of 20, 0.25 x0.50 m plots, the optimal size for this sampling (Wein and Rencz 1976). In the firstweek of June I clipped 10 plots of all above-ground vegetation, and sorted these byspecies, and by dead and live components. In the first week of August, at peakgrowth, I clipped and sorted an additional 10 plots from each community. No twoplots shared a side. Clipped samples were dried at 60°C until stable in mass, andweighed. The difference in mean dry weights of each component between Augustand June represented the net above-ground primary production of that component,less any consumption and senescence.Differentiating dead and live components in June was straight-forward, exceptfor Dryas on which some leaves remain alive but reddish-brown in winter, and onwhich dead leaves remain attached for a number of years (Svoboda 1977). In June,only leaves with red petioles were classed as alive. This rule was tested by followingleaf green-up, and the fate of marked, red-petioled leaves (total,= 277), on threeDryas plants in each of two herbivore exclosures (one in each vegetation community).58All red-petioled leaves became green by late June, though 10.5% showed partialnecrosis and may have died during the growing season. No leaves without redpetioles became green.In August I was unable to readily differentiate live and dead standing crop ofDryas stems, so I estimated the dry weight increment of stem during the growingseason as 21% of the net green leaf and shoot production (see Svoboda 1977).Enhancement of primary production: The interaction of enhanced productionand grazing was investigated in (1) a small-scale (less than a lemming home range)factorial experiment combining enhanced productivity, by fertilization, withdecreased grazing, by herbivore exciosure, and (ii) a large-scale (larger than alemming home range) enhancement of production by fertilization.For the small-scale experiment I built three 6 m x 21 m herbivore exclosures,each with a perimeter fence of 1.0 cm mesh chicken wire. Two were in Dryas heathclose to grids C2 and C3, and one was in a mesic hummock community near PE. Iestablished four pairs of 4 m x 4 m plots, such that one plot of each pair fell withinthe exclosure, and one immediately outside. Pairs of plots were systematicallyassigned one of following fertilization treatments in the following sequence: control(unfertilized); 50 kg/ha; 125 kg/ha; and 250 kg/ha (19:19:19 N:P:K fertilizer). Plots ina fertilization sequence were separated by 1 m, and were fertilized in early June 1990and 1991. Changes in standing crop in each 4 m x 4 m plot were assessed by visualestimation of percent cover (± 5%, and 2.5% for cover <5%) of key lemming foods(Dryas, Salix arctica and S. reticulata) within four randomly located 0.5 m x 0.5 mplots. Data were gathered at peak growth in early August 1990 and 1991.59For the large-scale enhancement experiment I fertilized two, one-hectareportions of Dryas heath, one on C2 and one on C3. I applied 19:19:19 N:P:K fertilizerat 125 kg/ha in early June of 1990, 1991 and 1992. Adjacent to each fertilized hectareI monitored one hectare of tundra with similar vegetation, as a control area. In eachfertilized and control hectare I randomly located eight, 0.5m x 0.5m plots for visualestimation of percent cover (± 5%, and 2.5% for cover <5%) of key lemming foods atpeak of summer growth in early August.The large-scale treatment effect was measured as the mean change in standingcrop from 1990 to 1992 on the two fertilized hectares compared to changes on the twocontrol hectares. The small-scale treatment effect was similarly measured as the meanof changes in standing crop from 1990 to 1991 on each of the four fertilizer treatmentsinside the three exciosures, compared to changes immediately outside the exclosures.Female burrow choice: I measured percent cover (± 5%, and 2.5% for cover <5%) of live Dryas leaves in 10, 25cm x 25cm plots at each of 1 m, 3 m, and 5 m radiifrom the natal burrows of radiocollared females on PE in 1992, at the end of lactationand beginning of the subsequent lactation. Plot positions at each radius were chosenrandomly from 20 potential compass bearings from the natal burrow mouth, and plotcorners were permanently marked with metal pins.RESULTSDensity Dependent Predation MortalityThe relationship between percent adult lemmings killed during summer andlemming density in spring is positively density dependent, but with wide scatter inthe data (y = 0.102 + 0.369 x, R2 = 0.47, F = 7.014, P = 0.029) (Fig. 2.1). The geometric60Fig. 2.1. The relationship between percent of adult lemmings killed during thesummer and adult lemming density in spring for two populations: Cl(triangles) and C2 (stars) from 1988 to 1992. The trends in the data aresummarized by the geometric mean regression lines (Ricker 1984) for all data(solid line), and all years except 1992 (broken line).(010090zEVA8070on./‘U4-’0)A601VAV50a7A407073077‘5720zC7o72107o70—io-—7IIIII00.20.40.611. regression still indicates positive density dependence when the 1992 data areremoved. The scatter partly reflects variable foraging effort by predators betweenstudy grids, varying numerical responses of migratory predators (e.g., rough-leggedhawks) at nest establishment, and varying breeding success of resident predators(e.g., red foxes). For example, the high percentage killed on Cl at a spring density of0.65/ha (Fig. 2.1) occurred in 1991 when rough-legged hawks initiated six nests, twomore than in any other season (Chapter 1), and occupied a nest adjacent to this grid.There is also considerable sampling error in assessing predation mortality on studyplots which are much smaller than the foraging ranges of predators.Predation mortality tended to be positively density-dependent when assessedover two-week intervals, but in least squares regression analyses the relationship wassignificant (P <0.05) in only four of the ten tests (those with R2.0.59) (Fig. 2.2). Theslope of the geometric mean regression lines was always positive, describing a meanincrease in proportion killed of 32.9% (S.E. = 4.32) for each additional lemming perhectare. The density at which no lemmings were killed was on average 0.29individuals/ha.Some of the scatter can be explained by temporary changes in the predatorcommunity. For example, red foxes did not reach study grids (with relatively highdensities) until early July in 1990, perhaps because they were unable to cross a streamin flood. Also, sporadic grizzly bear predation in mid-summer 1991 (at moderate tolow lemming densities) increased the percent killed substantially. However, much ofthe scatter is still unexplained. Although per capita predation mortality did declinewith declining lemming density during the summer, the relationship was not very63Fig. 2.2. The relationship between percent of resident adult lemmings killed within atwo-week period, and lemming density at the beginning of the two-weekperiod, for two populations: Cl (triangles) and C2 (stars). Each graphrepresents a random choice of one, two-week data point from each populationin each of four summers (1988-1991). Solid lines are geometric mean regressionlines (Ricker 1984), and R2 values are Pearson coefficients of determination.¶15 772 0 46-50:2 0:4 0:8 02 4 ‘.2 i:4C-5 R2 0.8002 0:4 o:e 02 4 1.2 1:4 1:0 1(f)51515MN-5 R2 oat- 02 0:4 0:0 02 4 ‘.2 1:4 i:640 //‘515S:N0:2 0:4 02 0:0 4 ‘.2 1:, ‘.264z /‘5.: / N0 --5/02 0:4 0:o 02 1 12 1 1:0a)‘-;7‘51050 J..5,..-.-2 0.8802 0:4 02 02 4 i2 1:/7R2 = 0.2902 01 0.5 00 1 18 II 1.5Lz**.5 /___ R2 =.15.4_i 1015WE02 0.4 05 05 1 1.5 1.4 1.5-5 = 0.00Density (resident adults / haj65predictable.Social Inhibition of ReproductionAdult-sized lemmings were reproductively active in 86-100% of individuallemming-weeks in June, July and August, and the proportion of lemming-weeks forwhich lemmings were reproductively active was not systematically lower or higheron PE than on controls in any year (Table 21.). Limited data indicate a slightly higherrate of reproductive activity in 1992, when predators and lemmings were scarce. Thesmall sample of weeks when individual lemmings were inactive was comprised ofspring-born females still maturing in early June, and some adults ending breeding inmid to late August.Litter size increased significantly with female weight post partum (Fig. 2.3).Reproductive history explained a considerable part of the residual variation: forprimiparous females litters were less than expected from female weight (meanresidual -0.53, S.E. 0.299); for multiparous females first litters were larger thanexpected (mean residual +1.01, S.E. 0.306), second litters larger than expected (meanresidual +0.45, S.E. 0.556), and third litters less than expected (mean residual -0.99,S.E. 0.246). Sample sizes were too small to compare PE to controls within eachcategory. However, litters born in PE showed no systematic tendency to larger sizethan those from controls (Fig. 2.3).With control and PE litters lumped together, litter sizes at Pearce Pt. tended tobe lower, especially in early summer, than embryo counts reported by Krebs (1964) atBaker Lake in 1962 (a year of low but increasing densities) (Table 2.2). However,Baker Lake counts were only significantly larger for the first summer litter ofFig. 2.3. The relationship between litter size and female weight post partum asobserved in a predator exciosure (squares) and on control grids (stars).666782r 2_0471• P —0.03030 40 50 60 70 80Female mass post partum (g)68Table 2.1. Proportion of lemming-weeks in which individual adult-sized lemmingswere reproductively active. Numbers of lemming-weeks for which datawere available are in parentheses. Sampling period covers 13 weeksin each year, from beginning of June to end of August.PP Cl C21990 1991 1992 1990 1991 1992 1990 1991 1992Over- 0.96 0.96 - 0.91 0.95- 0.98 0.97 -winter (24) (79) (45) (19) (41) (30)femalesSpring- 0.86 0.93 1.00 0.98 0.88 - 0.91 1.00 1.00born (37) (100) (32) (42) ( 8) (43) (22) (32)femalesAll 0.95 0.90 0.97 0.88 0.90 - 0.96 0.86 -males (38) (31) (35) (66) (21) (46) (59)69Table 2.2. Litter sizes of primiparous (spring-born), multiparous (overwintering)and first summer litter females at Pearce Point, N.W.T., and embryocounts of equivalent litters at Baker Lake, N.W.T. in 1962(data from Krebs 1964, Table 16). Litter periods refer to thechronological sequence of successive litters during the summer.Litter sizesPearce Pt. Baker LakeLitterAge class period mean S.E. n mean S.E. nPrimiparous I 4.40 0.24 5 5.35 0.32 17Multiparous I 5.44 0.50 9 7.17 0.54 6II 5.75 0.56 8 5.50 0.42 14III 4.00 0 3 4.67 0.88 3First summer III 3.50 0.50 2 4.00 0 1litter70multiparous females (t = 2.28, 13 d.f., P <0.05).There was no relationship between litter size and the lemming density on thestudy grid where the females gave birth, either at time of conception (r = 0.05, n = 18,P = 0.84), or at birth (r = 0.17, n = 27, P = 0.40).Behavioural Sensitivity to Predation RiskTotal daily growth of neonates increased significantly with female weight postpartum (Fig. 2.4). Limited data suggest that this daily growth was higher in FE (meanresidual +0.41, S.E. 0.23) than controls (mean residual -0.55, S.E. 0.10) at any givenfemale weight.Per capita daily growth of neo-nates showed no clear relationship to femaleweight post-partum (Fig. 2,5), but rates in FE were higher than those on controls atany given female weight.Daily growth rates of juveniles decreased significantly with body weight (Fig.2.6), but this relationship was quite variable probably because of the differing lengthsof time over which data were gathered. Rates in FE were not higher than controls(Fig. 2.6).Daily growth rates of scrotal males were not significantly related to bodyweight (r = 0.20, P = 0.24, n = 38), and growth rates in FE were not higher than thoseon control grids (t = -0.57, 36 d.f., F> 0.30). Weights in FE (mean 46.5, S.E. 2.8, n =10) were higher than those on controls (mean 39.9, S.E. 2.2, n = 28), but the differencewas not significant (t = 1.64, 36 d.f., 0.1 > F> 0.05).In a two-way analysis of variance there was no effect of the PE on post-partum71Fig. 2.4. The relationship between total growth rate of litters and female weights postpartum as observed in a predator exciosure (squares), a predator exclosure withfertilization (triangles) and control grids (stars).725-I030 4b 50Female weight post partum (g)73Fig. 2.5. The relationship between mean per capita neonatal growth rate on a litter bylitter basis and female weight postpartum as observed in a predator exciosure(squares), a predator exclosure with fertilization (triangles) and control grids (stars).741.31.21.1 CC0.80.7 o060.50.4 c! 0.3O.20100 40 50 60 70 80Female weight post partum (g)75Pig. 2.6. The relationship between growth rate of Juvenile lemmings and their bodyweights. IndMduals living In the predator exiclosure are indicated by squares, andcontrol individuals by stars.7614-1.3 r —0.661.2 P <0001C1.11o.g **0.8 C07 C C*Cu.6 C *2 0.5 C *0.4 C0,3 CC0.2*0.1C*0 ib i i 1 ië 20 22 2 26 28 30 32 34Weight (g)77female weights (F = 0.10, P = 0.75, for first and second litter periods; Table 2.3).However, post-partum weights varied significantly between successive litters (F =5.32, P = 0.003), increasing from first (June) to second litters (July), and decreasing forthird litters (August). There was no interaction effect (F = 1.32, P = 0.28).Food LimitationFood depletion: Changes in percent cover of the three principal collaredlemming foods were not substantial in either PE or on three control areas (Table 2.4),and differences between PP and the mean of controls were not significant for anyfood (Dryas: t = 0, 2 d.f., P> 0.9; Salix: t = 0.98, 2 d.f., P> 0.9; Carex: t5 = 2.25, 2 d.f.,0.10 <P < 0.20).Net primary production: The net above-ground primary production of vascularplants in 1992 was 40.0 g/m2 in the drier heath community, and 53.6 g/m2 in thewetter hummock community (Table 2.5). Net primary production of Dryas wasslightly higher in the heath, especially for leaves and flowers. The production ofsedges, willows and forbs was higher in the hummock community. These values maybe underestimates, especially for forb and willow production, since sample areascould have been grazed, most likely by arctic ground squirrels (Spermophilus parryii)and caribou (Rangifer tarandus).Assuming that all the primary production of three principal food groups(Dryas leaves and flowers, Salix leaves, and Carex) is available to lemmings, then asmuch as 88% of the net primary production in both the heath and hummockcommunities is potential lemming food.These estimates of net above-ground production are similar to values obtained78Table 2.3. Mean ± S.E. post-partum weights of adult female collaredlemmings comparing predator exclosure and control gridsfor three litter periods during the summer. “n’ recordssample size.Predator Controlexclosure gridsAge Litterclass period n mean ± S.E. n mean ± S.E.Spring- I 6 43.3 ± 2.4 11 45.3 ± 1.8born II 5 53.6 ± 1.5 4 47.0 ± 1.7III 3 51.7±6.1 0Winter I 7 48.5 ± 2.4 9 50.8 ± 1.8adults II 6 52.4 ± 3.7 8 52.6 ± 1.6III 5 50.6±4.2 1 51.0±0Summer III 1 58.0 ± 0 1 36.0 ± 0born79Table 2.4. Mean ± S.E. percent cover (± 0.5%) of three principal collared lemmingfoods measured in 0.5m x 0.5m plots in PE and three control areas(C2, C3, and immediately outside PE) in early August 1990 and 1992.ttnTt records number of plots.Lemming foodDryas Salix spp. Carex spp.Site Year integrifoliaPE 1990 14 ± 2.8 3 ± 0.5 17 ± 2.6(n=25) 1992 16 ± 2.7 5 ± 0.9 15 ± 2.0C2 1990 14±1.6 2±0.3 7±1.7(n=8) 1992 18± 1.9 2±0.3 7± 1.5C3 1990 11±2.0 0±0 3± 0.4(n=8) 1992 16 ± 3.4 0 ± 0 3 ± 0.5Outside 1990 17 ± 6.2 7 ± 1.3 12 ± 3.0PB 1992 14 ± 4.2 9 ± 2.6 14 ± 3.0(n=7)Table2.5.Meanabove-groundmass(g/0.125m2)ofvascularplantsandlichens,andnetabove-groundprimaryproduction(NPP;growthfromJunetoAugust1992)oftheseplants, intwocollaredlemminghabitats(heathandhummock)atPearcePoint,N,W.T.,basedonclipsamplingoften,0.125m2plotsineachcommunityinearlyJune,aftersnowmelt,andinearlyAugust,atpeakgrowth.Valuesinparenthesesarestandarddeviations.“-“meansthespeciesisabsent;“n.m.meansproductioncannotbemeasured.Plant identificationsfromPorsildandCody(1980)andCody(1992).HeathHummockPlant species,classandcomponentAugustJuneNPPAugustJuneNPPDryasintegrifolia-live*(1.13)(0.41) 1.6(1.76)(0.47)14.611.3(1.33)(1.61)0.80.80(4.48)(3.18)1.10.5(1.12)(0.88)(0.76)(0.35)**,class——-------———---andcomponentAugustJuneNPPAugustJuneNPPForbs0.20.200.300.3(0.10)(0.23)(0.28)(0)Saxifragaoppositifolia3.13.10n.m.n.m.n.m.(1.61)(1.34)Eguisetumarvense--*****0.20.20n.m.n.m.n.m.(0.35)(0.13)Total(g/0.125m2)5.06.7Total(g/m2)40.053.6CoTable2.5(continued)*Liveincludesonlyleavesandinflorescences;stemiscalculatedas21%oflive,followingSvoboda(1977).**Intheheaththisincludesprimarily.rupestris, ,petricosa1,nardina,C.scirpoideaandKobresiamyosuroides;inthehummockitincludesprimarilymembranacea, Catrofusca,,misandra,and.scirpoidea.Intheheathonlyarctica;inthehummockarcticaandreticulata.IntheheathChrysanthemumintegrifolium,DrabacorymbosaandStellarialongipes;inthehummockPolygonumviviparum, OxyriadigvnaandArmeriamaritima.*****Unidentified.0083from other Canadian arctic tundra communities where collared lemmings are knownto undergo substantial increases in population density in some years (Chitty 1950;Fuller et at. 1975b) (Table 2.6).Enhancement of primary production: In the small-scale fertilizer andherbivore exclusion experiment the percent cover of Dryas increased progressivelywith heavier fertilization, both inside and outside the exclosures, but willows showedlittle or no response (Table 2.7). In a two-way analysis of variance, fertilization andexclosure treatments both explained significant amounts of the variance in Dryaspercent cover, but there was no significant interaction effect (Table 2.8).Dryas outside the exclosures responded less strongly to fertilization at sites C2and C4, where I noted that lemmings had fed in and nested under the snow besidethe fertilized plots adjacent to the exclosure during the 1990-91 winter. Some of thepotential increases in Dryas standing crop following fertilization were apparentlyconsumed on these small plots outside the exciosures. However, the effect offertilization on live Dryas outside the exciosures was still significant (F 5.81, P =0.02), indicating that not all the enhanced production was being consumed.In the large-scale experiment, percent cover of fertilized live Dryas increasedsignificantly after one year (t = 3.5, df = 2, P < 0.05, one-tailed; Table 2.9), and to anextent similar to Dryas inside the small-scale exciosures (Table 2.7). This increase wasmore pronounced and strongly significant compared to controls after two years (t =11.3, df = 2, P <0.005, one-tailed; Table 2.9). Fertilization increased ground cover ofDryas by an average 54% after one year, and 99% after two years. Dryas is a perennialcushion plant on which previous growth remains attached for years after it has died.84Table 2.6. Comparison of estimates of net above-ground primary production(g/m2)in Dryas integrifolia / Carex rupestrisheath and.p integrifolia / . membranacea hummock communities atPearce Point, N.W.T., with estimates for similar communitiesat Canadian arctic sites where lemmings are known to undergosubstantial fluctuations in density.Net primaryPlant community Site production (g/m2)integrifolia / Pearce Pt.(70°N) 40.0. rupestris Banks I., Big R.(73°N) 15.2 1heath Victoria I. (71°N) 41.4Devon I., Truelove 26.0 2lowland (75°N)integrifolia / Pearce Pt. (70°N) 53.6C. membranacea Banks I., Thomsen R., 54.8hummock (73°N)Devon I., Truelove 54.3lowland (75°N)Data sources:1 Svoboda 1977, Table 12.2 Svoboda 1977, Table 11. Beach ridge community similar to heathcommunity at Pearce Pt. (Svoboda 1977, Table 8).Muc 1977, Table 4. Frost-boil meadows have species compositionclosest to hummock areas at Pearce Pt. (Muc 1977, Table 1).85Table 2.7. Mean changes in standing crop (% cover) from August 1990 toAugust 1991 for three collared lemming foods (Dryasintegrifolia (D.i), Salix arctica (S.), and S. reticulata()) under four different fertilization treatments,in paired 4m x 4m plots either inside or outside threeherbivore exciosures. Blank cells indicate the specieswas not present.Inside Outsideexciosure exclosureFertilizertreatment D.i. S.a. S.r. D.i. S.a. S.r.Unfertilized 1 0 -1 0control 1 12 -1 1 -1 -1 050kg/ha 4 0 1 07 53 1 0 1 0 0125kg/ha 9 0 3 09 86 0 1 3 1 2250kg/ha 8 0 8 011 69 0 0 5 0 286Table 2.8. Two-way analysis of variance investigating the effects ofherbivore exclosure and fertilization treatments on variancein mean change in standing crop (% cover) of live DryasAugust 1.990 to August 1991. Raw data areintegrifolia fromin Table 2.7.Source of Sum of d.f. Mean F-ratio Pvariance squares squaresExclosure 40.042 1 40.042 11.720 0.003Fertilizer 188.792 3 62.931 18.419 0.000Exclosure *fertilizer 2.458 3 0.819 0.240 0.867Error 54.667 16 3.41787Table 2.9. Mean changes in percent cover of two collared lemming food groups(live Dryas integrifolia, and live Carex/Kobresia), and the exposedsurface of dead Dryas, on paired one hectare control (unfertilized)and fertilized plots in two heath communities (C2 and C3), measuredfrom August 1990 to August 1991 (one-year) and to August 1992(two-years). Blank cells indicate data not available.Change in percent coverLive Dryas Dead Dryas Live CarexFertilization Timetreatment period C2 C3 C2 C3 C2 C3Unfertilized One-year 1 3 -1 -1control Two-years 4 5 0 0 -1 0Fertilized One-year 8 6 -7 -6(125 kg/ha) Two-years 12 13 -7 -7 8 988This dead material covers much of the ground surface in Dryas heath communities,and live growth covers only part of the top of this cushion. Increases in live Dryasstanding crop resulted in significant declines in exposed cover of dead Dryas in thefirst year (t -11.0, df = 2, P < 0.005, one-tailed). However, the exposed cover ofdead Dryas did not decline further in the second year (Table 2.9). The additionalincrease in cover of live Dryas in the second year was comprised primarily of newlateral branches beyond the edges of the established cushion.Increments in standing crop of another set of potential lemming foods, Carexand Kobresia species, were also significant over two years (t = 12.7, df = 2, P < 0.05,one-tailed; Table 2.9), representing an average 90% increase in ground cover. Willowswere too infrequent or small to assess changes.Burrow choice: A limited sample of three pregnant females, each changingmaternal burrows between litters in late July, did not consistently choose newburrows with increased Dryas abundance compared to the Dryas abundance they leftbehind at the end of their previous lactation (Table 2.10).DISCUSSIONDensity Dependence in Predation MortalityThe data indicated positive density dependence between cumulative summerpredation mortality and spring lemming density, and between predation mortality,assessed over two-week periods within summers, and lemming density. However,there was a great deal of scatter in the results, and the relationships had littlepredictive power.The lack of clear relationships may reflect the diversity of summer predators,89Table 2.10. Absolute abundance of Dryas integrifolia (% cover)at three radii from maternal burrows of three females atparturition, expressed as a proportion of j integrifoliaabundance at the previous maternal burrows at the end oftheir previous lactations.Female Radius Change in Dryas(m) abundance(i) 1 + 0.793 +1.945 +0.56(ii) 1 + 1.663 -0.375 -0.49(iii) 1- 0.303 -0.355 +0.3390their varying responses to lemming density, and the small scale of the study gridscompared to predator home ranges. Specialist lemming predators, such as the rough-legged hawk (Potanov 1986, Smith 1987), likely continue to kill many lemmings evenat low densities, perhaps in an inverse density dependent fashion. The red fox,normally thought of as a generalist (Voigt 1987), is expected to change its diet toinclude alternative prey when lemmings are scarce, thereby acting in a positivelydensity dependent fashion. Some generalist predators, such as peregrine falcon andarctic ground squirrel, continued to kill lemmings at very low densities in 1992(Chapter 1). A mix of predators with differing degrees of specialization likely causesincreased scatter in relationships between per capita mortality and density. Studygrids (18-25 ha) were much smaller than the hunting ranges of foxes (probablygreater than 40 km2. Chapter 4), and hawks (probably greater than 15 km2, Chapter4). Therefore, predation rates on grids could easily differ from rates throughout thestudy area, increasing the scatter in the data.Positive density dependence in predation mortality has rarely beendemonstrated with small rodent prey, and does not necessarily result in a stable preypopulation. Erlinge et a!. (1983, 1984, 1988) and Erlinge (1987) provided evidence forpositive density dependence in winter predation mortality on non-cyclic voles, withtotal annual predation approximating the annual production of voles. The systemappeared to be stable, but was fairly unusual in that the generalist vole predatorsprimarily ate other species, notably rabbits (Oryctolagus cuniculus), so voles werepreyed on heavily only when relatively common or vulnerable (Erlinge 1987). Inwestern Finland predation on voles by breeding raptors is positively density9:1dependent with very little time lag (Korpimaki and Norrdahl 1989, 991a,b;Korpimaki 1993), and least weasel predation on voles in winter is delayed densitydependent with a time lag of approximately nine months (Korpimaki 1993). Voledensities still varied cyclically in this system. Peak densities were probably dampenedand declines initiated by the raptor predation, which often exceeded production(Korpimaki and Norrdahl 1991b). Declines may have been exacerbated by the leastweasel predation (Korpimaki 1993). The system was not stable, and was distinctlyseasonal. Sinclair et al. (1990) provided evidence for positive density dependence inpredation on low to medium densities of house mice (Mus musculus) in Australia, butinverse density dependence when mice reached high densities. It was unclearwhether predation exceeded production at low densities, but the principal avianpredators were migratory, allowing mice to escape predator limitation, and, in somewinters with excellent food conditions, reach outbreak densities. Fitzgerald (1977)demonstrated destabilizing inverse density dependence in weasel predation on cyclicvoles in alpine California, as did Pearson (1966) studying feral cats and cyclic voles incoastal California. Positive density dependence does not necessarily result in a stableprey population, often because the predation still exceeds production and operatesonly in some seasons.Inhibition of ReproductionI found no evidence that reproduction was inhibited or decreased in thepredator exciosure. The proportion of adults reproductively active was the same inPB as on control grids, and was high in all summers. Litter size was not lower in PEthan on controls. Production of young was not significantly affected by density. I92conclude that there was no substantial social inhibition of reproduction in theexciosure.Social inhibition of reproduction in lemmings has been inferred to occur athigh densities, and has been demonstrated in some laboratory studies. Birth ratesdecline during and soon after periods of high density in fluctuating populations ofDicrostonyx as a result of slower maturation, shorter reproductive periods and smallerbody sizes (Krebs 1964; Fuller et a!. 1975b; Mallory et al. 1981). Rates at which littersare weaned in the lab may be depressed because of infanticide by strange adultmales and females within a few days post-partum (Mallory and Brooks 1978, 1980).In this study lemmings were always at low densities so social interactions werewould have been relatively few. I could not assess maturation rates because very fewindividuals were followed from birth to adult status. However, juveniles grew asquickly in PE as they did on control grids. I found no shortening of the reproductiveperiod in FE, where densities were higher. Adult female weights were not higher inPE than on controls. Adult males were larger in FE, perhaps because of bettersurvival and continuing growth as adults.Lemmings were occasionally infanticidal, which certainly limits populationgrowth to some extent. A minimum of 4% of litters on control grids and 5% in PEwere killed by infanticide in 1991, judging by sign on carcasses and termination oflactation by the female (Chapter 1). Radiocollared adult males were often located inthe same, or adjacent, burrow as lactating females at or shortly after parturition,which is to be expected in a species with post-partum oestrous. Adult females,however, were generally widely spaced and seldom more than 20 m from their natal93burrow. The low adult survival (Chapter 1) led to frequent turnover in territorialmales during the summer. Consequently surviving females likely encountered strangemales every few weeks.Infanticide in these circumstances is best explained by the sexual selectionhypothesis (Hrdy 1979), where males kill young they did not sire, so as to increasetheir future reproductive potential. I cannot rule out the use of young as a nutritionalresource by the killer, but think that alternative hypotheses such as competition forresources, social pathology or parental manipulation of sex ratio (Hausfater and Hrdy1984) are unlikely. Mallory and Brooks (1978) postulated that females encounterstrange males infrequently at low densities. However, heavy predation mortalitychanges the adult composition of the population frequently at these densities, sopotentially enhances the rate of infanticide. PE did not protect against this in 1991because some territorial males were killed inside, or when they moved outside.Infanticide did occur, but appeared to be insufficiently common to explain the lack ofpopulation growth in PE, and was likely an indirect result of predation mortality.Food LimitationThe results satisfied four of the five predictions of the hypothesis that foodwas not limiting. First, lemmings in PE did not deplete the standing crop of theirprincipal foods more than did lemmings on control grids. Second, net above-groundprimary production at Pearce Point was close to or higher than that in similarvegetation communities in other Canadian arctic situations where Dicrostonyx reachconsiderably higher densities. Since the majority of the primary production was inlemming foods, food quantity appeared sufficient for population growth. Third,94radiocollared lemmings occupied the fertilized one-hectare patches of heath, andcould have consumed the additional production. However, the increases in standingcrop of Dryas on the fertilized patches were similar to the increases in the herbivoreexciosures in the same habitats, indicating that lemmings did not consume anappreciable amount of the enhanced production. Fourth, adult females did notalways choose new maternal burrows with higher food availability than that neartheir previous burrow. Taken together, these results provide strong evidence thatfood abundance was not limiting population growth in summer.The fifth prediction, that the herbivore exclosure would not significantlyenhance the effect of fertilizer treatments compared with fertilized areas withoutprotection from grazing, was not satisfied. At two of the three exclosures, increases instanding crop were greater inside than outside. This reflects a problem of scale in thedesign. The very small treatment areas (4 m x 4 m) provided local hotspots of foodwhich were grazed by lemmings in winter.None of the tests of the hypothesis of no food limitation was completelyadequate. They were conducted at spatial scales smaller than the space occupied by apopulation. However, because of their similar results, I am confident that populationgrowth in PE was not limited proximally by summer food quantity.Food limitation of Dicrostonyx has not been assessed at low populationdensities. Most hypotheses regarding food limitation have addressed high densitypopulations, especially of the brown lemming (Lemmus trimucronatus), whichsometimes consumes the majority of its standing crop of foods in winter (Thompson1955; Schultz 1964; Batzli and Jung 1980).95Behavioural Sensitivity to Predation Risk.Ideally food and predation should be studied together because of the potentialpopulation consequences of the behavioural tradeoffs individuals must make betweenacquiring food and avoiding predation (McNamara and Houston 1.987). Suchdecisions concern when, where and how to feed and reproduce (Lima and Dill 1.990).The results indicate no effect of the predator exciosure treatment (reduced risk)on reproductive rate. Lemmings in PE were not reproductively active for longerperiods, and did not produce larger litters, than did control lemmings. Collaredlemming litter sizes have not varied significantly between years in other studiesdespite substantial annual variation in the abundance of predators (Krebs 1964; Fulleret a!. 1975a).However, changes in both litter size and female post partum weight throughsuccessive litters, follow changes in food availability, with rapid leaf proliferation inJune and July, and senescence in August. This could reflect sensitivity to predationrisk; females might acquire more food within a safe distance of cover as leaf growthprocedes, and less food as leaves senesce.Limited data also indicate that total daily growth of litters, and per capita dailygrowth of neonates, were higher in FE than controls. Higher encounter rates withpredators and their sign may force lactating females to remain stationary or return toa burrow more often, or induce them to stay in a burrow longer or travel shorterdistances to feed. Lactating females in FE might have foraged longer and more oftenthan control females.Whether improved litter growth rates can enhance population growth will96depend on whether weanling females reach maturity faster, and at larger body size,thereby producing relatively large litters earlier in life. However, juveniles in PE didnot grow faster, probably because many of those weaned in PE subsequentlyemigrated, and many of those found in FE were likely immigrants (Chapter 1).Immigrants would have been weaned in high risk areas, and may have been unableto respond quickly to the safer conditions. Adult males did not grow faster in FE,though differences would be harder to detect at the low rates (0.2 - 0.3 g/d)observed. However, males tended to be larger in FE, probably because they survivedlonger. Adult females did not differ in weight between FE and controls.The predator exciosure successfully kept out foxes and to a lesser extent othermammalian predators, but avian predators could still fly over. Lemmings in PEprobably continued to perceive some predation risk, so this treatment was anincomplete test of potential behavioural sensitivity to risk. I tentatively conclude thatrisk sensitivity limits population growth, but with effects too small to explain the lackof population growth in FE or controls.Other Limiting FactorsI now consider some factors which might limit population growth, but which Idid not directly assess. Induced or constitutive defence chemicals might inhibitreproduction and become limiting following heavy grazing (Haukioja 1980, Bryant etat. 1991; Seldal et al. 1994). Secondary plant compounds in graminoids may stimulatereproduction (Berger et a!. 1981). Since I did not observe reduced reproductive effort,in PE or comparing Pearce Ft to Baker Lake, I consider it unlikely that these factorslimited summer population growth at Pearce Point.97Habitat heterogeneity might affect population growth by influencing dispersalsuccess through habitats of differing patchiness and quality (Stenseth and Ims 1993).Densities of microtines on study grids were similar to densities in other parts of thestudy area (40 km2) (Chapter 1), so I do not consider study grids to be inferiorhabitats. Juveniles dispersed an average 53 m/d in the first 10 days after weaning,and many born in PE emigrated. They also suffered substantial predation mortalityeven when predators were relatively uncommon in 1992. Low rates of juvenileimmigration to PE, because of predation mortality, were probably responsible for lackof population growth (Chapter 1). Habitat barriers such as cliffs and a river deltamay have frustrated some potential immigrants.Community DynamicsAt a community level, the summer results are consistent with the generalmodel of Hairston et al. (1960); predators strongly limit herbivore populations whichare then unable to limit vegetative growth. Other resident herbivores at Pearce Pt. aretundra voles, which are also strongly limited by predation mortality in summer(Chapter 1), and arctic ground squirrels.Oksanen et al. (1981) predict that at primary productivities of 50-150 g/m2herbivores should be non-cyclic because they are food limited, and as productivityincreases their densities should periodically support carnivores which induceirruptive dynamics. The results are contrary to these predictions. The predatorcommunity at Pearce Point was relatively persistent from year to year (Chapter 1).Lemmings can show both multi-annual irruptions (Banks I.) and persistent lowdensities, within this same range of productivities. Lemmings appear not to be food98limited at low and relatively stable densities. Evidently the three-level trophicexploitation system can persist at quite low productivities.Adding seasonality to previous models, Oksanen (1990) predicted that for arapidly reproducing herbivore with short generation time, seasonality is adestabilizing force and should result in chaotic populations limited proximally byfood. The results do not fit this model either. Seasonality at Pearce Pt. seems to bestabilizing; strong summer predation limitation is relaxed in winter, allowingpopulation growth and annual density fluctuations of limited amplitude.Predator RegulationModels of predation regulation in the density dependent paradigm posit aprey density below which prey productivity exceeds predation mortality, andtherefore prey population growth is possible (Sinclair 1989, Hanski et a!. 1991).Resulting prey dynamics could be stable, or fluctuate around an equilibrium. I couldnot measure productivity adequately because I lacked data on fates of juveniles.However, recruitment of adult lemmings was very low in summer (Chapter 1), andthe analyses of predator diets indicated that most subadults weaned in summer werekilled by predators (Chapter 4). Therefore, predation of adults exceeded recruitmentof adults throughout the summer. The result is a destabilizing effect of predation onlemmings in summer.The density dependent predation mortality at Pearce Point is not stabilizing forthe following reason. Predation mortality depresses lemming populations at PearcePt. until they reach 0.1 to 0.4 adults per hectare (Fig. 2.2, Chapter 1). At suchdensities, rough-legged hawks do not nest in spring, and experience poor breeding99success in late summer, and red fox breeding fails. Consequently, these two principalpredators fall out of the system, as in 1992. When these two predators are rare orabsent, lemming populations grow, but when they are both present, lemmingscontinually decline through the summer (Chapter 1). The heavy predation mortalityin summer is therefore destabilizing. The persistence of lemmings between yearsresults from winter breeding in the absence of most predators.Rosenzweig and MacArthur (1963) conclude that a stable equilibrium in suchexploitative interactions depends on predators’ access to alternative prey, low preycapture efficiency of predators, and interference or territorial behaviour amongpredators constraining their access to prey. These conditions are not satisfied atPearce Point. Rough-legged hawks are unable to fully compensate for declininglemming availability with alternative prey (Chapter 4). Prey-capture efficiency islikely high because of the lack of cover with low vegetation (see Janes 1985), shallowlemming burrows, and lack of darkness in summer. Mammalian and avian predatorsappear unable to constrain each other’s use of the same space. Lemming persistenceat Pearce Point therefore depends, not on stability through density dependentpredation, which is unstable in summer, but instead on the changing predatorcommunity between seasons, which allows population growth in most winters (Krebset a!. 1995, Chapter 1).I could identify no factor, other than predation mortality, strongly limitingpopulation growth in summer. The small limiting effects of infanticide and predationrisk are most likely side-effects of predation mortality. Habitat heterogeneity remainsto be fully investigated. From the mechanistic point of view, predation mortality100remains the most likely regulatory factor at low densities in summer. Such regulationis contingent on winter population growth, which, during this study, did not producespring densities in excess of those which summer predators could depress. Factorslimiting winter growth are addressed in Chapter Three.101CHAPTER THREEPOSSIBLE FACTORS LIMITING WINTER POPULATION GROWTHIN COLLARED LEMMINGSINTRODUCTIONWinter field studies of collared lemmings (Dicrostonyx spp.) have been limitedin scope. Yet lemmings experience winter conditions for at least eight months of theyear, so winter population processes, including breeding under the snow, are crucialfor their year-round persistence (Fuller et a!. 1975a). Some information is availablefrom laboratory studies and from the field sign these arvicolid rodents leave of theirwinter activity.Collared lemmings exist as summer and winter morphs. Changes from onemorph to the other can be induced in the lab by shifts in photoperiod from summerto winter regimes (Hasler et a!. 1976, Malcolm and Brooks 1985, Mallory et a!. 1986).Individuals raised under winter photoperiods grow faster and reach heavierasymptotic body weights than those under summer photoperiods (Hasler et a!, 1976,Mallory et a!. 1981,1986, Malcolm and Brooks 1985, Reynolds and Lavigne 1989).Winter morphs also assume a more rounded body shape (Malcolm and Brooks 1993).Collared lemmings are the only microtine to moult from a dark summer pelage to awhite winter coat, and they develop bifid claws on the third and fourth digits of theforefeet (Hansen 1957).These changes are adaptive. Summer morphs have to increase their resting\%102metabolic rates significantly faster than winter morphs (198% increase at 0°Ccompared to 16-30% for winter morphs) at temperatures below thermoneutrality(15°C to 20°C) (Chappell 1980, Reynolds and Lavigne 1988). Much of the energeticadvantage for winter morphs derives from the thick winter coat and piloerection(Scholander 1950, Chappell 1980), and these winter morphs have significantlyreduced minimal thermal conductances (Reynolds and Lavigne 1988). The change inbody shape results in a decreased surface area, which also reduces rate of heat loss inwinter (Malcolm and Brooks 1993). Increased absolute energy requirements resultingfrom increased body mass were apparently met in the lab, not by increased foodintake, but by increased length of the gastro-intestinal tract in winter morphs(Reynolds and Lavigne 1989). White coat colour seems unlikely to improve energybalance, but is likely an adaptation to avoid detection by predators (Chappell 1980).The bifid claws are assumed to enhance digging in snow (Hansen 1957).Winter morphs can breed in subnivean temperatures of -20°C, and survivetemperatures as low as -35°C (MacLean et a!. 1974; Fuller et a!. 1975a). The insulationprovided by the nests of dead vegetation they build (MacLean et a!. 1974) is probablycritical for such survival and reproduction (Casey 1981). Chappell (1980) maintainedthe internal temperature of heated copper casts of lemmings, placed in nests at -20°C,with energy inputs equivalent to metabolic rates at thermoneutrality. In a simulationmodel of brown lemming (Lemmus trimucronatus) energetics, Collier et a!. (1975) foundthat larger females had improved survival and reproductive capability in winter, butthat the energy costs of lactation, combined with consequent additional costs whileforaging and rewarming the nest, could often exceed the capabilities of most females.1.03Subnivean temperatures appear critical to an understanding of winter populationprocesses.Judged by distribution of nests in spring, lemmings choose areas of deepestsnow as winter habitat (MacLean et al. 1974, Fuller et al. 1975a). Snow moderatescold ambient temperatures, keeping temperature at or near the ground substantiallywarmer at least through early and mid-winter (MacLean et a!. 1.974; Fuller et a!. 1975a;Chappell 1980). Temperature at the ground gradually declines through the winter toa low in March (e.g., -25°C, Fuller et a!. 1975b), and the temperature gradient isreversed for a period in spring (April, May or June) because snow insulates a colderground from warming air temperatures (Fuller et a!. 1975a; Chappell 1.980).Collared lemmings bred in April or May of every year of population cycles atboth Baker Lake, Northwest Territories (N.W.T.) (n = 4 yr) (Krebs 1964), andTruelove Lowland, Devon I., N.W.T. (n 4) (Fuller et a!. 1975b). Breeding fromNovember through March was less common at Baker Lake (Krebs 1964), andprobably did not occur in all winters on Devon I. (Fuller et a!. 1975b). Weather andsnow conditions may limit winter breeding. Shelford (1943) found that winters withbreeding leading to a peak density at Churchill, Manitoba, were relatively warm withheavy snow, especially in early winter. At the same locality, Scott (1993) found peakabundances to be correlated with years which combined relatively cold autumntemperatures, minimizing the freeze-thaw cycles in October, and warmer thanaverage temperatures and heavier than average snowfall in November and December.Similarly at Baker Lake, Krebs (1964) found that a winter population increaseoccurred when autumn was dry, without surface freezing of rain, and was followed104by relatively heavy snow. At Barrow, Alaska, MacLean et al. (1974) found that winterbreeding leading to a peak was associated with warmer than average springtemperatures. They surmised that deep snow in spring (April to June) keeps theground cold for longer, thereby making spring breeding more costly (see also Fulleret a!. 1975a).Lemming populations do not increase in all winters with apparently suitablesnow and temperature conditions (Krebs 1964, MacLean et a!. 1974). Predation,especially by ermine (Mustela erminea) and least weasels (M. nivalis), may be sufficientto cause overwinter population declines even when lemmings are breeding under thesnow (Maher 1967, MacLean et a!. 1974).At Pearce Point, N.W.T., collared lemmings have bred in spring (April or May)under the snow, and perhaps earlier in the winter, in six consecutive winters (Krebset a!. 1995; Chapter 1). Breeding under the snow was necessary for populations torecover from frequent summer declines driven principally by predation mortality, butin some winters populations declined (Krebs et a!. 1995; Chapter 1).In this chapter I first estimate the extent of reproduction required to achieveobserved rates of winter population change at Pearce Point. Second, I test thehypothesis that ermine predation explains differences in population growth betweenwinters. Third, I assess whether wintering microtines use habitats with the deepestsnow. Fourth, I assess how much of the variance in rates of winter population changecan be explained by a combination of: (i) cold during autumn moult (September andOctober) limiting potential body growth and accumulation of energy reserves; (ii)thermal conditions under the snow (a combined function of ambient air temperature105and snow depth) limiting the proportion of females reproducing and their individualreproductive output during winter (November through March); and (iii) snow depthin spring (April and May) limiting the proportion of females breeding and theirreproductive success.METHODSStudy AreaThe study took place at Pearce Point (69°48’N, 122°40’W), on the south shore ofAmundsen Gulf, western mainland N.W.T., Canada. The coastal tundra is comprisedof dry Dryas/Carex heath, mesic Carex/Salix/Dryas hummock (frost-boil), and wetCarex meadow vegetation communities, all within 20 m a.s.1..The following climate data are from maps in Maxwell (1980), based on years1941-72. Pearce Point lies close to the 100 cm isohyet for annual mean total snowfall.The mean date of snow cover formation is October 1, but snow cover can be delayeduntil November 1. Monthly snowfall is highest in September (17 cm), October (22 cm)and April (12 cm). However, mean snow depth at the end of October is only 5 - 10cm, because of autumn melting. Mean snow depth reaches 20 cm by the end ofNovember, and 25 cm by the end of January, but stays at that depth until rapid meltin May. The mean daily temperature is approximately 2°C in September, -27.5°C inJanuary, and -5°C in May.Climate DataMean daily temperature and daily snow-depth were measured at theDepartment of National Defence radar base at Clinton Point (69°35’N, 120°48’W),approximately 65 km east of Pearce Point, at 101. m a.s.l., but also close to the coast.106Data were made available by the Atmospheric Environment Service of EnvironmentCanada. I use these data to represent winter conditions at Pearce Point, since I lackwinter data from the study area itself. Long-term climate data (Maxwell 1980)indicate that conditions at Clinton Point are very similar to those at Pearce Point.To estimate the effect of temperature and snow depth on lemming populationgrowth, I derived three indices of thermal stress, corresponding to three time periodswith differing degrees of reproduction (Krebs 1964). Autumn cold stress: I used themean of all mean daily temperatures for September and October as an index of thecold stress a lemming encounters while changing from summer to winter morph.These morphological and physiological changes can be induced at thermoneutraltemperatures, by decreasing photoperiod to light available less than 25% of the time(Hasler et al. 1976, Malcolm and Brooks 1985, Reynolds and Lavigne 1988). The roleof colder temperatures in the field is unknown. Between August and November thetime the sun is above the horizon at Point Barrow, Alaska (71°N) decreases from 65to 20%, reaching a mean of 9% for November (Chappell 1980). September andOctober are therefore the critical months for photoperiod-induced changes inmorphology. Winter cold stress: During winter (November through March) Icalculated an index of cold stress as the sum of all daily snow depths over theperiod, divided by the sum of mean daily temperatures over the period. Since allmean temperatures were below freezing, this gave the ratio of centimetre-days ofsnow to degree-days of frost. Higher values of the index actually reflect lower stressunder the insulative blanket of snow. Spring cold stress: During spring (April andMay) I estimated the insulative power of the snow as the mean daily snow depth107over the period.Population EstimationPopulation densities of collared lemmings and tundra voles (Microtusoeconomus) were estimated using mark-recapture and the open Jolly-Seber populationmodel. Animals were caught alive in Longworth traps and individually identifiedwith numbered ear tags. Densities were estimated on four grids, each 18 to 25 ha,from early June to late August in each of six summers (1988 through 1992). A largeportion (11.4 ha) of one grid (PE) was maintained as a predator exclosure, and threeother grids were controls (Cl, C2 and C3). However the exciosure was not successfulat excluding mammalian predators in winter because drifting snow allowed red foxes(Vulpes vulpes) to enter over the fence, and ermine could get through the fence. Also,most winter nests on PE were outside the exclosure. So I consider PE an untreatedpopulation for the winter. Details of trapping and population estimation are inChapter 1.I did not measure birth and death rates during the winter. Instead I useinstantaneous rates of population change, calculated from densities in late summerand early the following summer, as a composite measure of demographic processes.To estimate how many winter litters were necessary to explain the populationchanges observed between summers, I used a simple difference equation model withthree time periods corresponding to the stress indices: autumn, winter and spring. Iconsidered only females of two age-classes, adult (, 35 g) and subadult. I lacked dataon number of subadult females in early autumn, but assume these were at least asfrequent as adults. I assumed that each nest occupied by ermine represented the108death of a female lemming. I assumed a winter litter size of three (Krebs 1964), and a50:50 sex ratio of neonates.Winter Snow DistributionIn late May I mapped the distribution of remaining snow patches on studygrids (1991 on PE, Cl and C2; 1992 on PB and C2). I assume that this remnant snowcoincides with areas of deepest snow during the winter. This is reasonable becausemost remnant patches were in the lee of hills or banks where snow drifts form.Drifted snow is relatively dense because wind-blown crystals break and then packtightly in drifts (Kind 1981). Deep, dense snow melts relatively slowly because uppersnow layers insulate lower layers from melting air temperatures, and compact, finegrained snow reflects a high proportion of incoming radiation (Male and Gray 1981).Microtine Winter NestsThe density of microtine winter nests was estimated by systematicallysearching grids, and plotting nest locations, in early summer after snow-melt (PB, Cl,C2 and C3 in 1991; FE and C2 in 1992). In 1992 I did not search Cl and C3systematically, so cannot estimate nest density. However, I did collect otherinformation on these nests in 1992. I was unable to differentiate lemming nests fromthose of voles.Weasels line nests they occupy with microtine fur, and leave scats and preyremains nearby (Maher 1967, MacLean et a!. 1974). The only Mustela species in thestudy area was the ermine. Each nest was examined for sign of occupancy by ermine,and the vicinity was searched for scats and remains (typically stomachs, feet andtails). All nests with ermine sign were removed to search them for prey remains and109ear tags. All other nests were destroyed to insure they were not recounted thesubsequent year.The association of winter nests with deep snow was tested by comparing theactual distribution of nests, either under snow or in the open in late May, to arandom distribution. The random distribution assumes numbers in proportion to thepercent of ground either covered or clear of snow in late May.RESULTSExtent of winter breedingI use data from 1990-91, a winter of population growth, and 1991-92, a winterof decline, to estimate the extent of winter breeding. The number of overwinter adultfemales in spring 1991 was higher than the number of adults late the previoussummer on three of four grids (Table 3.1). However, only one ear-tagged adultfemale from summer 1990 was found on all grids in spring 1991. Evidently nearly alllate summer adults failed to survive the winter, and therefore nearly all overwinteradults in spring had been born late the previous summer, or had been born in winter.To assess how many litters may have been born in winter, I contrast low andhigh survival scenarios, with data from grid C2 in 1990-91, the grid with the highestrate of increase, and only lemmings present (Table 3.1). In a low survival scenario Iassume only one subadult female for each adult in late summer, and monthlysurvival rates of 0.6, 0.7 and 0.8 for autumn, winter and spring periods respectively.The adults surviving to winter would have to produce at least three litters to reachobserved early summer 1991 numbers. In a high survival scenario, with two:iioTable 3.1. Numbers of females on study grids in late summer (end of August) and inearly summer (early June), after the intervening winter, and minimum numberof females killed by ermine during winter.Late Early summersummer of next year Minimum # femaleskilledby ermineAdult Over-winter Spring-born (i.e. nests occupied)Winter Grid females females males1990-91 PE 8 6 6 5Cl 4 6 2 4C2 3 7 5 3C3 2 3 4 3Total 17 22 17 151991-92 PE 11 0 3 0Cl 1 0 0 3C2 2 0 0 1C3 0 1 2Total 15 0 4 6111subadults for every adult in autumn, and monthly survival rates of 0.8, 0.9 and 0.9for autumn, winter and spring respectively, early summer numbers could beexplained with only one winter litter per female. These latter survival rates are quitehigh for microtine rodents, so it seems likely that the strongest population growthfrom summer to summer is the result of multiple litters in winter, as well as springreproduction.During the 1991-92 declines, there may have been little or no winterreproduction, but some lemmings survived to breed in spring. I found only spring-born lemmings in early summer 1992, indicating that the few adults surviving winterdied during the spring. The severe decline on PE may be attributed to red foxpredation (in the absence of any sign of ermine predation, Table 3.2), emigration, orvery stressful weather conditions.Ermine predationThere was no significant difference between 1990-91 and 1991-92 in proportionof nests occupied by ermine (Mann-Whitney U = 11.5 <U0.10), although all microtinepopulations grew in winter 1990-91 but declined in winter 1991-92 (Table 3.2).The instantaneous weekly rate of population change was not significantlycorrelated with the percentage of nests occupied by ermine (r = 0.21, F = 0.28, P0.62) over both years (Table 3.2). For winter 1990-91 there was an insignificant trendto lower rates of population change on grids with higher percentage of nestsoccupied (r = -0.75, F = 2.55, P = 0.25).Nest DistributionWinter nests were positively associated with areas of deepest snowTable3.2.Instantaneousweeklyratesofchangeofmicrotines,microtinenestdensities,andnestoccupancybyermineonthefourstudygridsovertwowintersofintensivestudy(1990-91and1991-92).SurveysofClandC3wereincompletein1992sonestdensitiescannotbecalculated.1990-911991-92PEClC2C3PEClC2C3Instantaneousweeklyrateofchange:-bothmicrotines0.00080.01650.02770.0433-0.0116-0.0165-0.0446-0.0613-Dicrostonyxonly0.00080.01650.02770.0520-0.01840.0000-0.0446-0.0679Microtinenests:Totalfound5642544861224836Nestdensity(#/ha)3.112.503.231.923.39-2.87-Nestspercapita4. on four of five cases assessed (Table 3.3; Fig. 3.1). The insignificantassociation on C2 in 1992 may have resulted from a late snow-melt, and my mappingof snow cover before many areas of thin snow cover had melted.Snow and Temperature RegimesSnow depth profiles for the five winters studied (Fig. 3.2 and 3.3) followedgenerally the same pattern as found in long-term climate data (Maxwell 1980). Mostsnowfalls in September melted, but snow accumulated rapidly in late October andearly November, after which it was relatively stable for the rest of the winter.However, snow depths towards the end of January were all below the 25 cm long-term mean (Maxwell 1980). Only in 1990-91 did snow depths exceed this level in latewinter.Dicrostonyx populations grew in three of five winters, and declined in two, andthe indices of thermal stress varied substantially between years (Table 3.4). Theautumn stress index explained the highest proportion of variance in rate of change (r2= 0.67, F = 6.124, P = 0.09)(Fig. 3.4). Combining the index of thermal stress in winterwith the autumn index, the explained proportion of variance in rate of populationchange increased substantially (r2 = 0.98, F = 41.09, P = 0.02). The addition of thespring index of thermal stress, mean daily snow depth, increased the explanatorypower of the model only marginally (r2 = 0.99, F = 21.72, P = 0.16).The intensity of cold experienced by lemmings when moulting and changingbody shape in autumn had lasting effects on population growth in some winters.Autumn 1991 was the coldest of five, and was followed by the steepest populationdecline (Fig. 3.4). Autumns 1987 and 1988 were the mildest, and subsequent winters115Table 3.3. The association of microtine winter nests with remnant winter snow.Snow cover was mapped in late May, and nest distribution, though mappedafter snow melt, is related to snow distribution in late May.Spring 1991 Spring 1992PE Cl C2 PE C2Distribution of winternests in late May:Percent under snow 69.8 40.6 27.5 39.6 55.6Percent in open 30.2 59.4 72.5 60.4 44.4Distribution of snow:Percent snow-covered 14.4 8.5 7.2 15.7 48.6Percent snow-free 85.6 91.5 92.8 84.3 51.4Sample size of nests 53 32 51 53 45G test statistic 83.71 24.41 19.05 17.44 0.89P <0.001 <0.001 <0.001 <0.001 >0.10116Fig. 3.1. Map of the distribution of winter nests (dark points) with respect todistribution of remnant snow in late May (hachured line) on PB grid, spring1991.0250,II118Fig. 3.2. Profiles of mean weekly snow depth (cm) over the nine month periods forwinters 1987 through 1990. Data were collected at Clinton Point, approximately65 km east of the study area.11945-— 1987-8840• 19884935—•—•- 1989-90. 3025>, 20I I ISept Oct Nov Dec Jan Feb Mar Apr May JuneWinter monthsFig. 3.3. Profiles of mean weekly snow depth (cm) over the nine month period forwinters 1990 through 1992. Data were collected at Clinton Point.12012145-‘—b’ 40 — 1990-911—• 1991-923025/> 20 //1510-I5. ii0il I ‘1 iiiiiiiiiiii 1111111111111 ‘‘111111Sep Oct Nov Dec Jan Feb Mar Apr May JuneWinter months122Table 3.4. Mean ( S.E.) instantaneous weekly rates of population change (r) from lateAugust to early June for Dicrostonyx alone on four study grids, combined withthree indices of thermal stress which might affect r. These are: Autumn- theaverage of the mean daily temperatures for September and October (°C); Winter- the cumulative cm-days of snow cover divided by the cumulative degree-daysof frost for November through March; Spring- the mean daily snow depth.Indices of thermal stressWinter r Autumn Winter Spring1987-88 0.0326 * -0.66 0.627 6.4(0.0131)1988-89 0.0151 * -1.41 0.521 8.6(0.0075)1989-90 -0.0092 * -2.38 0.579 5.0(0.0073)1990-91 0.0243 -2.82 0.926 15.0(0.0108)1991-92 -0.0327 -4.45 0.720 20.8(0.0149)* data from Krebs et a!. (1995)123Fig. 3.4. Relationship between instantaneous weekly rate of population changeoverwinter and the average of mean daily temperatures (°C) during theautumn (September and October) when Dicrostonyx are changing from summerto winter morphology.124a 0.OB- 20.05 r —0.670.0400.0390-91a 0.020.0188-890‘- -0.01•0 69-900 O.0-0.03-0,04 91-92-0.05 I- -4 - - -1 uMean daily autumn temperature (C)125had high rates of population growth. However, cold autumns could be amelioratedor exacerbated by thermal conditions during winter. For example, averagepopulation growth in 1989-90 was less than expected based on autumn temperatures(Fig. 3.4), because there was little snow in winter (Table 3.4, Fig. 3.2). By contrast,growth in 1990-91 was substantially higher than expected based on autumntemperatures (Fig. 3.4), because of thick snow (Table 3.4, Fig. 3.3).DISCUSSIONErmine PredationThe intensity of ermine predation explained little of the variance in winterpopulation growth of microtines in the two years studied. Only within a winter ofrelatively good thermal conditions (1990-91) did the expected inverse correlation ofpopulation growth and intensity of ermine predation hold. Perhaps in this winter,with a high ratio of snow depth to cold temperatures, thermal conditions under thesnow were not strongly limiting, and the limiting effect of ermine predation could beexpressed. In the colder 1991-92, effects of ermine predation may have been maskedby the stronger proximal limitation of cold. In such a winter ermine predation mayhave been compensatory, or just of low intensity, perhaps because cold puts severeenergetic costs on ermine with their elongate body shape (Brown and Lasiewski1972). The data on ermine predation cover only two winters, so the conclusion thatermine predation explains little of the variance in population growth remainstentative.Theoretical models (Hanski et al. 1993, Hanski and Korpimaki in press) andfield data from Fennoscandia (Korpimaki et al. 1991, Korpimaki 1993) and North126America (Fitzgerald 1977), suggest that delayed numerical responses of least weaselsand ermine to peak microtine populations can cause declines and prolong periods oflow density. North American data from some lemming declines support thishypothesis; highest proportions of nests occupied by Mustela spp. coincided withdeclines on Banks I. (20%, Maher 1967), Devon I. (11-16%, Fuller et a!. 1975a) and atPoint Barrow (34%, MacLean et a!. 1974). The proportion of nests occupied at PearcePt. was lower, indicating a smaller limiting effect of this predation.The role of ermine predation during low lemming densities varied in otherstudies. MacLean et a!. (1974) observed no weasel predation, but relatively low snowcover, in the first winter following a decline. There was no substantial lemmingpopulation growth until the subsequent winter, when snow cover was much deeperbut weasels occupied 5.6% of nests. Fuller et a!. (1975a,b) observed substantiallemming population growth in a summer following a winter with 11 % of nestsoccupied. Evidently factors other than ermine predation can limit winter populationgrowth at low densities, prolonging the low density phase. Also, factors operating insummer sometimes curtail growth and prolong the low phase. At Pearce Point,ermine predation does not explain differences in population growth between winters,though it is a persistent limiting factor.Snow DistributionDistribution of lemmings among habitats in winter was strongly influenced bydistribution of snow, as previously observed (Maher 1967, MacLean et a!. 1974, Fulleret a!. 1975a). Wind redistributes snow in drifts, and may keep upland tundra bare orthinly covered. Areas with little or no snow cover are poor winter habitat because of127high energy costs and predation risk for lemmings foraging in the open.Snow distribution, if it varies little between winters, probably cannot explainmuch of the variance in rate of population growth within a study area. However, itmay explain differences between regions with differing annual snowfalls. Within therange of Dicrostonyx, annual mean total snowfall varies from less than 50 cm togreater than 200 cm, and mean snow depths in late October from 5 to 50 cm(Maxwell 1980). Pearce Point has relatively little snow on average (ca 100 cm/yr, and5-10 cm at the end of October, Maxwell 1980), so the limiting effects of snowdistribution and depth may be more pronounced here than in eastern regions, wheresnow is deeper (Maxwell 1980).Winter Temperature RegimesThe critical factors limiting winter population growth at Pearce Point appear tobe intensity of autumn cold, and depth of snow cover per degree-day of frost inwinter. The association of strong population growth with relatively deep snow,especially in November and December, has been noted before (Krebs 1964, MacLeanet a!. 1974, Scott 1993). However, combining snow depth with ambient airtemperature is a useful index of thermal stress because it relates directly to theenergetic costs faced by lemmings under the snow. This supports earlier suggestions(Shelford 1943, Krebs 1964) that relatively deep snow is a necessary condition for thewinter breeding which fuels a population irruption.The discovery that autumn cold may curtail winter population growth is new,but expected given the increased metabolic costs faced by summer morphs exposedto freezing temperatures (Chappell 1980, Reynolds and Lavigne 1988). Autumn cold128might affect population growth in a number of ways. First, higher metabolic costsinduced by cold should force summer morphs to spend more time foraging, makingthem more vulnerable to autumn predation. Second, higher metabolic costs maydiminish energy for body growth after maintenance requirements have been met,especially for late summer young shifting to winter morphology. This is critical sincegrowth to a larger body mass and more rounded body shape are adaptive for energyconservation (Malcolm and Brooks 1993) and for reproduction in low temperatures(Collier et al. 1975). Third, in severe cold, maintenance energy requirements may notbe met, and lemmings may die of exposure. I have no data to address these threemechanisms, but suspect that the third is rare, and the second most likely.Duration of snow cover in spring explained little of the variance in populationgrowth, probably because spring breeding is the last in a sequence of demographicprocesses starting in autumn. The number of females breeding in spring and theirindividual reproductive efforts depend primarily on their individual histories inautumn and winter. The winters studied did not include any with benign autumnand winter conditions followed by a spring of long-lasting snow. In such a case, theimpact of poor spring conditions on population growth may be stronger, assuggested by MacLean et al. (1974).Other Factors Affecting Winter Population GrowthThe frequency and dispersion of foods could influence population growthbecause distance travelled and time spent away from the nest in feeding affect winterenergy balance and reproductive success (Collier et a!. 1975). I lack data on standingcrops of foods before each winter. However, these crops appeared to change little129between years, so seemed unlikely to strongly influence rates of population changebetween years within this study. Food availability increases as more tundra iscovered in snow. However, I do not know how differing depths of snowfall relate tosnow distribution. I can only say that food availabilky, especially as it is influencedby snow distribution, may have been a limiting factor at Pearce Pt. in any of thewinters studied. I recommend quantifying standing crops of lemming foods, andsnow distribution, in any attempts to explain regional variation in rates of winterpopulation growth.Mallory et a!. (1986) proposed that age structure may influence winterreproductive potential; lemmings born in spring and surviving the summer may notacclimate to winter as well as summer-born young, so an autumn population with ahigh proportion of older individuals may have relatively lower rates of survival andreproduction. I cannot address this idea adequately because I lack data onrecruitment of late summer young as autumn adults. However interannualdifferences in autumn age structure were probably small because populationssuffered heavy mortality in all summers, and there was little recruitment of animalsfrom early summer litters (Chapter 1). A younger age structure, following a summerof population growth, may enhance winter population growth (Mallory et a!. 1986),but only if autumn cold does not impede growth to optimal body size for winterreproduction.Hansson and Henttonen (1985) found a correlation between both snow depthand period of snow cover, and the amplitude of microtine population irruptions innorthern Europe. They hypothesized that deeper snow precludes successful foraging130by generalist predators such as foxes, thereby enhancing winter survival ofClethrionomys and Microtus species. Lindstrom and Hornfeldt (1994) found less ofthese small rodents in fox diets as snow depth increased. Collared lemmings areuniquely adapted to winter with frequent reproduction under the snow (Fuller et a!.1975a, Malcolm and Brooks 1985, Mallory et a!. 1986, Reynolds and Lavigne 1988).Consequently deep snow that protects them from fox predation should have dramaticeffects on population growth, as long as mustelid predators are not abundant. AtPearce Point, snow may get drifted and packed by wind such that it protectslemmings from fox predation, but I lack data on the intensity of this predation. Inconclusion, relatively deep snow may enhance winter population growth byproviding protection from some predators, as well as by providing thermal cover,and access to food over a relatively wide area.131CHAPTER FOURPATIERNS OF PREDATION ON NON-CYCLIC LEMMINGS:THE GENERALIST PREDATOR HYPOTHESISINTRODUCTIONSome microtine rodent populations do not exhibit multiannual cyclicity withhigh amplitude density fluctuations, but show slower rates of change in density oftenwith an annual pattern (Taitt and Krebs 1985). Andersson and Erlinge (1977)proposed that these more constant prey densities over time might result fromlimitation to population growth in two ways: (i) generalist predators, with type-Illfunctional responses, operating in a community with a relatively diverse prey base(generalist predator hypothesis), or (ii) nomadic avian or mammalian predatorsresponding rapidly in local abundance and breeding success to growing rodentpopulations (nomadic predator hypothesis). Hanski et a!. (1991) demonstratedthrough mathematical modelling that increasing the number of generalist predators,in a predator prey model with multiannual fluctuations, tends to reduce theamplitude and period of fluctuations, eventually to a stable equilibrium point.Rosenzweig and MacArthur (1963) also undertook a theoretical investigation ofexploitative predator-prey systems. They concluded that stability in such systemsdepended on the following conditions: (a) sufficient alternative food to sustain thegeneralist predators when preferred prey are scarce, (b) relatively inefficient preycapture by these generalist predators, principally because prey have a secure refuge,and (c) predator population growth limited by some factor other than food, and most132likely by territorial behaviour.Some of the same species exhibiting constant densities, also exhibit highamplitude, multiannual density fluctuations in other portions of their range (Hanssonand Henttonen 1985). The specialist predator hypothesis proposes that an irruptionoccurs when a population escapes limitation by specialist predators, which have fewor no alternative prey when rodents become scarce (Andersson and Erlinge 1977,Hanski et al. 1993). In some other ecological communities rodent irruptions occurdespite the existence of predators with a diverse prey base. As an explanation forthese dynamics, the alternative prey hypothesis proposes that the alternative preyonly partially compensates for declining rodent numbers. Consequently the predatorsdecline, allowing a rodent irruption (Angelstam et a!. 1984, Lindstrom et a!. 1987).The generalist-specialist distinction is not always clear. A true generalist keepsnumbers of a prey type within a narrow range by rapidly compensating for anydeclines in one prey by using other prey. A true specialist can never fully compensatefor declines in its prey, so drives prey to lower levels, ultimately too low to maintainitself. Between these extremes, some predators may compensate for declining primaryprey to varying degrees depending on annual circumstances, or the local diversity ofthe prey base. These I will call hlsemi_generalistsu.At Pearce Point, Northwest Territories, Canada, on the western arctic mainlandcoast, collared lemmings (Dicrostonyx ki!angmiutak) and tundra voles (Microtusoeconomus) remained at low densities over six years, in conjunction with a fairlyconstant and diverse predator community (Krebs et a!. 1995, Chapter 1). Densities ofresident adult microtines rarely exceeded two per hectare, and generally showed133annual fluctuations with summer declines, and winter increases driven by breedingunder the snow. Predation mortality, principally by red fox and rough-legged hawk(Buteo lagopus), was sufficient and necessary to explain summer population declinesof adult microtines, but the fates of subadults remained unclear (Chapter 1). Theconstancy in microtine numbers makes this system a suitable candidate for testing thegeneralist predator hypothesis.In this chapter, I test the generalist predator hypothesis by assessing how welleach of the principal lemming predators adheres to the following predictions: (i) thedominant predators do not show strong numerical responses to variations inlemming density, (ii) the proportion of lemming biomass in predator diets declineswith decreasing lemming abundance, and is compensated for by increasedconsumption of alternative prey, (iii) predators show weak functional responses atlow lemming densities, and stronger functional responses at higher lemming densities(type-Ill response), (iv) at very low lemming densities predators consume virtually nolemmings due to low capture efficiency. I also test an earlier prediction (Chapter 1)that the majority of subadult lemmings born in summer are killed by predators.METHODSStudy AreaIntensive research was conducted in a 40 km2 area immediately inland fromPearce Point, Northwest Territories, Canada (69°49’N, 122°43’W), on the north coast ofthe western Canadian mainland, from late May to early September of 1990, 1991 and1992. Gently rolling, glacially scoured hills are frequently broken by dolomite andbasalt cliffs, both along the coast and inland. Collared lemmings used three upland134habitats: Dryas integrifolia heath, D. integrifolia / Carex rupestris heath, and D.integrifolia / Carex membranacea hummock communities (Krebs et a!. 1995). Tundravoles were restricted to ribbon-like wet Carex aquatilis meadows and wetter hummockcommunities along stream and lake shores (Bergman and Krebs 1993; Krebs et a!.1995). Brown lemmings (Lemmus trimucronatus) were extremely rare, and found onlyin one area of extensive Eriophorum / Carex tussock meadow. A fourth residentvertebrate herbivore, the arctic ground squirrel (Spermophilus parryii) was widespreadthrough the drier upland communities, where well-drained soils provided excellentdenning opportunities. Substantial numbers of caribou (Rangifer tarandus) passedthrough the area in mid to late summer.The numerous cliffs provided excellent nesting habitat for several avianpredators: rough-legged hawk, golden eagle (Aquila chrysaetos), peregrine falcon (Falcoperegrinus), gyrfalcon (Falco rusticolus), raven (Corvus corax), Thayer’s gull (Larusthayeri) and glaucous gull (L. hyperboreus). The following alternative prey specieswere regular non-colonial nesters or summer residents within a 10 km inland radiusof Pearce Pt: tundra swan (Cygnus columbianus), Canada goose (Branta canadensis),common eider (Somateria mollissima), semipalmated plover (Charadrius semipalniatus),golden plover (Pluvialis dorninica), Baird’s sandpiper (Calidris bairdii), pectoralsandpiper (Calidris melanotos), horned lark (Eremophilia alpestris), American pipit(An thus rubescens), Lapland longspur (Calcarius lapponicus) and snow bunting(Plectrophenax nivalis). The following species were numerous on migration: snowgoose (Chen caerulescens), northern pintail (Anas acuta), oldsquaw (Clangula hyernalis),red-breasted merganser (Mergus serrator), sanderling (Calidris alba).135Microtine PopulationsPopulation densities of lemmings and voles were estimated with the JollySeber open population model using data gathered in weekly or biweekly live-trapping (Longworth traps), marking with ear tags, and repeated recapture, on threeareas (grids), each 18 to 25 ha, and each including a range of vegetation communities.Juveniles of summer cohorts were caught infrequently, so densities are of residentmicrotines only (i.e. adults and subadults caught at least twice on the same grid). Apredator exciosure was built on a fourth grid, but data reported here refer only to thethree control grids. For further details on procedures see Chapters 1 and 2.Lemmings over 35 g (adults) were fitted with a radio-transmitter (Biotrack Inc.,model SS-1) mounted to a cable tie as neck collar (total weight c. 3 g) in all threeyears. Voles over 35 g (larger adults) received radiocollars only in 1990. Relocationsof radiocollared individuals every two or three days gave data on cause and timingof death.At each capture individual weight, using a spring-loaded Pesola scale, andreproductive condition were recorded. Date of parturition was estimated usingchanges in weight, degree of closure of the pubic symphysis, and teat size. Lactatingfemale Dicrostonyx typically use one maternal burrow (Brooks 1993; unpubi. data thisstudy), and juveniles are weaned at 15 to 20 days (Brooks and Banks 1973). Iassumed that a litter was not successfully weaned if either lactation ended, or thefemale was located at least twice in a burrow >30 m from the natal burrow, prior tothe expected weaning date (see Brooks 1993). A small sample of litters born aboveground in traps, nest boxes or sedge tussocks gave an estimate of mean litter size136within four days of birth. By combining these data, I estimated the numbers ofjuveniles weaned per unit area (Chapter 1).HabitatsHabitats as classified by Krebs et at. (1995) were mapped to quantify theregional availability of collared lemming habitats. For study grids, mapping was doneon foot. For a regional assessment, digital Landsat data were used in a supervisedclassification of habitats. The area of each habitat type available to lemmings wasassessed in two zones, with obvious differences in representation of the habitats. Acoastal zone (21.82 km2), in which exposed sand and rock was more common,included all study grids. An inland zone (832.96 km2) had relatively even vegetativecover, with little sand or rock.Predator Numerical ResponsesNumerical responses of predators include both the numbers of adultsestablishing breeding territories in spring, and the numbers of young successfullyweaned or fledged in summer. I lack data on predator numbers in winter.In late May and early June all cliffs within the 40 km2 study area, includingareas up to 6 km inland from Pearce Pt. harbour, were searched for raptorsestablishing nests or showing territorial behaviour. I recorded clutch size, hatchingsuccess, and fledging success with subsequent visits to all cliffs in mid-June, earlyJuly, and early to mid August.One red fox natal den was located in the 40 km2 study area. I observed the denfrom a distance periodically, and visited it regularly to collect scats (see below). Fromthese observations, and frequent sightings of foxes, numbers of adults and juveniles137were counted. Individual foxes were readily recognized by their unique coatcolouration and pattern as adults, and after mid-July as juveniles. I captured one non-lactating female in a Novak leg-snare in June 1991, and gave her a coloured ear tag.Ravens, and both species of gulls, nested on inaccessible mainland and islandsea-cliffs, and I was unable to accurately count active nests and breeding success. Idid not accurately estimate populations of arctic ground squirrels, another lemmingpredator (Boonstra et al. 1990), in all years. However, I obtained an index of squirrelabundance in early July of 1991 and 1992 by placing microscope slides covered intalcum powder (see Boonstra et a!. 1992) in all burrow mouths in three areasseparated by at least two kilometres, and counting the proportion of slides tracked bysquirrels within 24 hours. Also in 1992, live-trapping (Tomahawk traps) and colouredear-tagging of squirrels at all burrows on three, one-square-kilometre study blocks,were combined with periodic observations of the blocks, to get an absolute count ofadult squirrels prior to weaning of juveniles. Ermine (Mustela erminea) were summerand winter residents, and bred in the study area each summer judged by casualobservations, but I did not make an accurate population estimate. I estimated grizzlybear (Ursus arctos) numbers based on occasional sightings and track sizes.Predator DietsCollections: I collected pellets of all resident raptorial birds, except the raven,and scats of two mammalian predators, the red fox and grizzly bear. Regurgitatedpellets of raptors and gulls were collected systematically at egg-laying in late Mayand early June (pre-incubation sample), at hatching in late June and early July(incubation sample) and at fledging in mid to late August (nestling sample) (cf. Poole138and Bromley 1988), at a series of obvious raptor perches (cliff tops, glacial erratics,rock outcrops) throughout the study area. The 1990 collections did not includeperches used by gulls or golden eagles, and the collection area was expanded in 1991and 1992 to sample these species. Because eagle eggs hatch in early June at thislatitude (Poole and Bromley 1988) all eagle pellets were lumped together as a nestlingcollection. Nearly all gull pellets were probably from glaucous gulls judged byobservations of perch use by gulls. To minimize disturbance, I did not collect pelletsand prey remains at or below nests during the reproductive period. At eachcollection every pellet was removed, its length and maximum diameter wererecorded, and it was individually bagged and labelledI collected fox scats systematically from the same sites as the raptor pellets,systematically from the natal den and two other intermittently-used dens, andopportunistically whenever fresh scats were found. In 1990 systematic collectionswere monthly, and in 1991 and 1992 biweekly. At each collection, every scat wasremoved, its maximum diameter recorded, and it was individually bagged andlabelled. Most scats were collected at the natal den. This sample likely consisted ofscats from adults and juveniles, which might have had different diets. I used thefrequency distribution of maximum diameters of scats to differentiate juvenile fromadult scats.I collected all grizzly bear scats encountered.Analyses: Fox scats were autoclaved, and soaked in water for 24 to 72 hours toloosen material. To remove soluble material, I washed scats through a series ofseives, with paper towel on top of the lowest seive to catch small undigested139remains. Remains were air dried. All bones (as small as microtine molar teeth), largefeathers, eggshell, insect parts and large pieces of vegetation were separated by hand,leaving a set of remains consisting mostly of mammalian hair sometimes mixed withsmall feathers, small pieces of vegetation and inorganic debris (mostly sand). Thepercentage by weight of each component of this latter set of remains was estimatedby eye, taking account of the higher density of inorganic debris. All sets of remainswere weighed (± O.05g).Remains were identified to species (mammals), or order (birds, insects), usingreference material collected in the field and from museum collections, and usingmolar teeth descriptions in Banfield (1974), hair keys in Kennedy and Carbyn (1981)and Adorjan and Kolenosky (1969), and feather keys in Day (1966).Since molar teeth are the most recognizable microtine remains in fox scats(Lockie 1959), and each of the 12 molar teeth of each of the three microtines in ourstudy area is unique, I recorded the frequency of each molar tooth in each scat. Whenremains included two microtines, I calculated the relative proportions of hair andbone weights based on the relative proportions of molar teeth (Lockie 1959). Whenmicrotine skeletal material was lacking, I differentiated voles from lemmings by haircolour.Raptor pellets and bear scats were separated by hand into the same sets ofremains as fox scats, and components were identified using the same referencematerial and keys.Conversions to biomass: For fox scats I used information in Lockie (1959),Goszczynski (1974) and Reynolds and Aebischer (1991) to provide the following140multiplicative factors to convert total weight of undigested remains by species(mammals) or order (birds) to biomass of prey killed: voles (23), ground squirrels(45), large mammals such as caribou (Rangifer tarandus) and grizzly bear (100),passeriform and charadriiform (“smalli birds (45), anseriform and galliform (“large’t)birds (61), eggshell (9.1). For lemmings I used a factor of 28, because adult lemmingsare substantially larger, with shorter appendages, and a more globular body formwith lower surface to volume ratio, compared to the vole species studied by Lockie(1959) and Goszczynski (1974). Insects were ignored in biomass estimations, as theywere mostly in trace quantities in the scat. I assumed that foxes ate a similarproportion of ground squirrel carcasses as they would lagomorphs, because of thesimilar body sizes of these prey. I used a factor of 100 for large mammals, a figureintermediate to those provided by Lockie (1959) and Goszczynski (1974).An alternative approach to biomass estimation suggested for foxes by P.-O.Palm (unpubl. data), and for ermine by Gamberg and Atkinson (1988) assumes thatmammalian hair is not digested at all. This is a valid assumption judged by themicroscopic appearance of hair in fox scats. To determine the relationship betweenhair and body weights, I removed hair from pelts of 14 Dicrostonyx collected by N.Larter on Banks I., N.W.T., by soaking pelts in water for a few days, and graduallyscraping away the hair (Gamberg and Atkinson 1988). A conversion factor wasderived from the weight of hair as a proportion of body weight.For raptors, conversion factors relating remains in pellets to biomass ingestedwere unavailable. Mandible length is a good predictor of body weight (Hamilton1980), but mandibles were partially digested in most pellets. However, the molar141tooth rows on dentary and maxillary bones were frequently intact, along withnumerous separate molar teeth. I can use these to assess lemming biomass ingested ifI assume that: (i) all lemming heads are ingested, (ii) all lemmings ingested arerepresented by at least one tooth, and (iii) the length of the molar tooth rows aresignificant predictors of body weight. To test the first two assumptions for rough-legged hawks, volunteers fed freshly thawed lemmings, originally from a laboratorycolony, to a captive hawk in a 5 x 5 x 3 m outdoor aviary (Orphaned WildlifeRehabilitation Centre, OWL, Delta, British Columbia), at approximately midday inApril, and at two rates: 2 lemmings/d for seven days, followed by four lemmings/dfor thirteen days. Pellets were collected daily, and the hawk was weighed at thebeginning and end of each period. Light regimes differed from the 24 h daylight ofan arctic summer, but temperature regimes were not much warmer at the southernlatitude of the experiment. To test the third assumption, I measured tooth rowlengths of 17 collared lemmings collected on Banks I., N.W.T., in summer 1993.Conversion to number of individuals killed: For foxes, estimates of dailyconsumption rates (g/d) in each time period are derived from lemming biomass (g)per scat by multiplying by an assumed defecation rate (scats/d) (see below). Thisconsumption rate (g/d) is converted to numbers of individual adult and subadultlemmings consumed each day, based on the proportions of adults and subadults inthe diet for that time period (see below), and the mean live weights of these ageclasses (g) in the population, based on live trapping during the same time period.For foxes, daily defecation rate varies from one to six scats for a poorly fed fox(ingesting 200 g/d) to eight to twelve scats for a well fed fox (600-800 g/d), giving a142reasonable mean of seven for an intermediate diet of 300-500 g/d (P-O. Palm, differentiate adult (>35g) from subadult (<35g) lemmings in scats, I usedthe bony parts most frequently found intact (ulna, and upper and lower molar toothrows), and for which a linear dimension varied with body size. I used a sample of 17collared lemmings from Banks I., to determine the regression relationships betweenbody weight and the lengths of these bony parts. For each collection period Iassumed that the ratio of subadults to adults in scats represented the relativeproportions of those age classes ingested. I also assumed that subadult remainsrepresented only weaned individuals, since bones of suckling juveniles are brittle andunlikely to pass intact through fox digestive tracts.For rough-legged hawks I used the captive feeding experiment (see above) todetermine what proportion of lemmings ingested were represented by unique molarteeth in pellets, and therefore calculate a conversion factor relating number ofindividuals in pellets (maximum number of unique molar tooth) to number ofindividuals ingested. I used this experiment to test the assumption that the ratio ofadult to subadult lemmings in pellets, as determined by molar tooth row lengths,represented the relative proportions of these age classes ingested. Calculating dailyconsumption rates depends on casting rate of pellets, which I also calculated from thecaptive feeding experiment.For foxes and hawks, the sex ratio of lemmings ingested was estimated fromthe incidence of undigested male and female pelvic bones, differentiated on thelength of the ischium and the width of the pubis (Dunmire 1955).143Total impact of predation: For fox predation, daily consumption rates of adultand subadult lemmings were converted to total impact for the time period bymultiplying by the number of foxes (adult or juvenile) present during the timeperiod, and the length of the period (d). The results from each time period weresummed to give total impact over the summer.An estimate of the area foxes used in summer to kill all adult lemmings inscats was based on numbers of adult radio-collared lemmings killed on study gridsover a summer, and the area of collared lemming habitat on study grids and in thetwo regional zones of differing habitat availability. Collared lemmings were located97% of the time in three habitat types (Dryas heath, Dryas/Carex heath, andCarex/Dryas hummock) (Krebs et a!. 1995), which I consider collectively as lemminghabitat. Lemming habitat comprised 36.6 ha of the study grids on which mortalitydata were collected, allowing a calculation of numbers killed/ha of lemming habitat.Foxes hunted throughout the coastal zone, so the number of lemmings they killed inthis zone was the number killed/ha of lemming habitat multiplied by the number ofhectares of lemming habitat in this zone. The remainder of their kill was from theinland zone. The area of inland zone they used was based on the numbers killed/haof lemming habitat on study grids, and the proportion of lemming habitat in theinland zone. The total estimated area traversed by foxes was the entire coastal zoneplus the area of the inland zone required for the remaining kill.To calculate the impact of fox predation on subadult lemmings I assume thatsubadults were weaned in lemming habitat, and killed within the area traversed byfoxes hunting adults. Numbers of subadults available were based on mean litter size,144number of litters weaned/ha of lemming habitat on study grids, and the total area oflemming habitat in the area traversed by hunting foxes.The total impact of adult rough-legged hawk predation was calculated in thesame manner as red fox predation. I lacked data on diets of nestlings.Predator Functional ResponsesI have good estimates for adult lemming densities only. Predator functionalresponses are therefore considered only for predation on adult lemmings. For foxesand rough-legged hawks, daily consumption rates of adult lemmings were takenfrom the scat and pellet analyses (see above) and plotted against the regional adultlemming density for the same time period, estimated as the mean of densities onthree control grids.RESULTSMicrotine DemographyMean densities of resident Dicrostonyx declined sharply during summers 1990and 1991, while Microtus on the same study grids maintained relatively stable orslightly declining densities (Fig. 4.1). Both species virtually disappeared from grids inwinter 1991-92, remaining absent or at extremely low densities through summer 1992until August.Red foxes and rough-legged hawks, together, were the cause of between 57%(1991) and 72% (1990) of all confirmed predation mortalities of adult lemmings (Table4.1). A number of the mortalities attributed to an unspecified raptor could have beenrough-legged hawk kills, but may also have been peregrine falcon or gull kills.145Fig. 4.1. Mean densities (numbers/ha) of resident collared lemmings (solid line) andtundra voles (broken line) on the three study grids. Solid bars representwinter, during which densities are interpolated.1461,2-1.1 1990 1991 1992‘ 09*0.8Cl,0.7Cl) 060.5‘5 A04 1”0.3 “-7 ‘a 0.20.1‘‘%_.“,__0 I I I I I I I I I I I I I I I I I I I I “Junejulyaug June Julyaug june July augTime (weeks)147Table 4.1. Numbers of confirmed predation mortalities of radio-collaredadult collared lemmings attributable to individual predatorspecies. The sex ratio (male:female) of adults killedby the two principal predators is indicated in parentheses.Predator species 1990 1991 1992Red fox 21 (7:14) 4 (2:2) 0Suspected fox 5 4 0Arctic ground squirrel 0 2 0Grizzly bear 0 2 0Ermine 0 0 0Rough-legged hawk 7 (4:3) 17 (11:6) 0Peregrine falcon 0 1 1Unknown raptor 6 7 0Total 39 37 1148Ermine did not kill lemmings on study grids in summer, though they did in winter(Chapters 1 and 3). Ermine appeared to avoid tundra beyond 50 m of rocky cover insummer.For lemmings, mean litter size across all cohorts and all years was 5.0 (S.E.0.27, n = 27). In 1990 lemmings weaned 14 litters on study grids exposed to predation(36.6 ha lemming habitat), and in 1991 seven litters.Numerical Responses of PredatorsMicrotine densities differed little between years 1990 and 1991, but bothmicrotine species were extremely rare in 1992 (Table 4.2, Fig. 4.1). Red fox and rough-legged hawks, the principal lemming predators, showed obvious numerical responsesto low microtine densities in 1992 (Table 4.2; Fig. 4.1). Three red foxes died duringthe 1991-92 winter. The female ear-tagged in 1991 was seen at the den with one pupon 14 June 1992, but not again. The den was virtually unused after 21 June, and onlyone of three adults from June was seen later in the summer. At least two rough-legged hawk pairs were seen in late May 1992, but none nested. By contrast, threeother raptors, and the other mammalian predators, showed little or no numericalresponse to the very low microtine densities in 1992 (Table 4.2). Numbers of nestinggulls of both species were noticeably lower in 1992, perhaps in response to lowlemming densities, but I lack quantitative data.Steep summer declines in lemming abundance in 1990 and 1991 appeared toaffect red fox and rough-legged hawk breeding success. I do not know how many foxpups were born, but in 1991 only two of five pups were weaned. In 1990, roughlegged hawks abandoned two nests (eight eggs) in early July, just prior to hatch, and149Table 4.2. Numbers of raptor pairs establishing breeding territories,numbers of nests successful and young fledged, and numbers ofadult and weaned juvenile mammalian predators, in relation tomean adult Dicrostonyx density (#/ha) and combined mean adultDicrostonyx and Microtus densities (#/ha) in spring (earlyJune) and summer (early July) on three study grids.1990 1991 1992Spring densities:Dicrostonyx 1.03 0.84 0.01Both microtines 1.15 1.09 0.03Summer densities:Dicrostonyx 0.73 0.62 0Both microtines 0.88 0.88 0.01.Raptor territorial pairs:Rough-legged hawk 4 6 0Peregrine falcon 5 5 6Golden eagle 1 1 1Gyrfalcon 0 0 1Successful nests (# fledglings)Rough-legged hawk 2 (3) 3 (5) 0Peregrine falcon 5(10) 2 (2) 4(10)Golden eagle 1 (1) 1 (1) 1 (2)Gyrfalcon 0 0 1 (2)150Table 4.2. (continued)1990 1991 1992Mammal adults (# weaned young)Red fox 2 (3) 3 (2) 3 to 1 (0)Ermine ? (.2) ? (3) ? (2)Grizzly bear 1 (0) 1 (0) 2 (0)151three young fledged from the remaining eight eggs. In 1991, three pairs abandonednests (11 eggs) in late June, close to hatch, and five young fledged from 11 othereggs. Chicks were food stressed judged by aggressive actions of larger chicks aimedat smaller ones, and discovery of chick remains below two nests.By contrast peregrine falcons fledged 10 young from 14 eggs in 1990, and alsoin 1992 when lemmings were scarce. Their low success in 1991 (two fledged from 16eggs) resulted from predation on one nest, and two intense storms, one of whichwashed a nest off a cliff, the other hitting when young were downy but quite large.In 1992, gyrfalcons and golden eagles fledged all young hatched. At least onegyrfalcon used the area in 1990, but I found no evidence of nesting.The mean proportion of microscope slides with ground squirrel tracks on foursample areas was 0.38 (S.E. 0.12) in 1991, and 0.17 (S.E. 0.06) in 1992, indicating adecline between years. In 1992, 11 resident squirrels occupied three km2 prior to theemergence of young. The density of double burrows (cf. Carl 1971) was 9.3/km2 (S.E.1.4, n = 4, one km2 blocks).Grizzly bears occasionally visited the study area, each individual spendingonly from two to ten days in a summer, though one denned in the area in winter1990-91.In 1990 and 1991 one pair of rough-legged hawks (RLH) nested seven km weston the coast, and in all years a pair of peregrine falcons (PF) nested seven km inland,but we lack complete data on these nests outside the intensive study area. Beyondthese the nearest raptor nests (RLH and PF) were 12.5 km inland. One pair of RLHnested at this site in 1992, suggesting that microtine densities inland may have been152higher than at the coast.In summary, and as far as numerical responses are concerned, rough-leggedhawks, red foxes and perhaps glaucous gulls, acted as specialists, whereas grizzlybears, golden eagles, peregrine falcons and gyrfalcons acted as generalists.Predicting Lemming Body WeightUlna length, mandible length (cf. Hamilton 1980), and upper and lower molartooth row lengths were all significant predictors of live body weight for Dicrostonyx(Table 4.3).Hair weight was also a significant predictor of body weight (Table 4.3). Haircomprised a mean 2.8% ±0.14% S.E. of body weight, giving a conversion factor ofhair weight in scats to body weight ingested of 36.Fox dietDifferentiating adult and juvenile scats: There was a shift in the percentfrequency of maximum diameters of fox scats at the natal den in 1991 from June tomid-July, and again from mid-July to August (Fig. 4.2). The frequency distribution ofdiameters of scats collected away from the den in June and July 1991, presumably alladult scats, was most similar to the August distribution at the den (Fig. 4.2). I assumethat juveniles contributed the great majority of scats <1.5 cm in diameter in June andJuly, and use this diameter to differentiate juvenile from adult scats at the natal den.Composition: Few fox scats were collected in both June 1990, when collectionswere not made at the natal den, and July and August 1992, when only one fox wasobserved and regular collections found very few scats (Table 4.4).Fox diets were similar in 1990 and 1991 on a percent frequency of occurrence153Table 4.3. The linear regressions of live body weight (g) of Dicrostonyx (dependentvariable) on lengths (cm) of mandible, ulna, and upper and lower molar toothrows, and from weight (g) of hair.Independent Regression equationvariable n r2 F PMandible length log y = 0.855 + 4.367 log x 16 0.73 37.1 <0.001Ulna length log y = 1.012 + 2.684 log x 17 0.91 146.8 <0.001Upper tooth row log y = -2.626 + 4.954 log x 17 0.85 84.6 <0.001Lower tooth row log y = -1.783 + 4.011 log x 17 0.80 60.9 <0.001Hair weight y = 34.45 x 14 0.99 836.6 <0.001154Fig. 4.2. Frequency distributions (percent) of maximum diameters of red fox scatscollected at (a) the natal den from 29 May to 6 July 1991 (n=83), (b) the natalden from 7-20 July 1991 (n=47), (c) the natal den in August 1991 (n=58), and(d) away from the natal den in June and early July 1991 (n=23).155252015105°1.1 1.2 1.3 1.4 1.5 1.8 1., 1.9 19 2 21 2:2 2:3 2!425>\20 bUCcDj5D11 1. 1.3 1. 1.5 1. 1. 1.8 1.12.2 242520151.5 I.E 1.7 1.H2 2.1 2.2 2.4C 10cI_)UL_ I1Ia)a2520 d15nu105________011 1. 1.4 1. 1.8 1.7 1.8 1.9 2L 23 2:4Maximum diameter (cm)Table4.4.Percentfrequencyofoccurrenceofallpreyremainsinredfoxscats.199019911992JuneJulyAugJuneJune23JulyJuly21AugAugJuneJulyPreytype2-22-July67-20-Aug34-1718-31&AugMammalia:DicrostonyxMicrotusLemmusSpermophilusRan giferljrsusVulpesUnidentified8082201300205381103200092777570648314212932505200005024544562507402019142040000040200020200063503740003460373000300001100 8 0 8 8 0 0 8Table4.4.(continued)199019911992JuneJulyAugJuneJune23JulyJuly21AugAugJuneJulyPreytype2-22-July67-20-Aug34-1718-31&AugAyes:Anseriform15182400700460Galliform020402000610Charadriiform/85561575142110PasseriformUnidentified8811812291914201420Eggshell15854415007340-Insecta:Coleoptera/03114015242317020HymenopteraCrustacea:02000000000SampleSize:1361385052553722463510—------ci——------------—----158and a biomass per scat basis (Tables 4.4, 4.5 and 4.6). Dicrostonyx was by far the mostcommon food item, but was found in fewer scats and contributed less in biomasslater in the summer, as lemming densities declined. The frequency and biomass ofMicrotus and Spermophilus generally increased as the summer progressed; voles bredsteadily through this period, and ground squirrels became more common with theemergence and dispersal of juveniles after mid-July. Caribou was a common preyitem only in late summer 1991, Some caribou calves became separated from theirmothers in the study area on migration in July and August, and foxes may have beenable to kill a weak calf or scavenge from a carcass. There was one adult cariboucarcass and one grizzly bear carcass in the intensive study area in 1990, providingscavenging opportunities for foxes. One adult male ermine was found, with its skullcrushed by fox, at the natal den in June 1991.Birds were taken in all months, but were most frequent and contributed mostin biomass in July and August (Tables 4.4, 4.5 and 4.6) coincident with vulnerableyoung leaving the nest (especially in sandpipers and plovers) and with adult moult(especially in ducks and geese). Unidentified bird remains (bones and feather shafts)could most often be assigned to “large (anseriform and galliform) or “small”(charadriiform and passeriform) bird classes for biomass estimation, even if theycould not be classified to order. Remains of red-breasted merganser, oldsquaw, snowgoose, and snow bunting were found at the natal den. Gallinaceous bird remainswere likely rock ptarmigan (Lagopus mutus), which were very rarely seen in the studyarea, but may have been more common inland. Most insect remains were bumblebees(Bombus sp.), and an unidentified crab species comprised the crustacean remains.Table4.5.Mean±S.E.biomass(g)perscatforredfox,andpercentoftotalbiomass(parentheses)ofprincipal preygroupsinsummers1990and1992.Juvenileandadultscatsarerecordedseparatelywheresamplesizepermits.Allbiomassestimationsarebasedonconversionofallpreyremains,exceptforDicrostonyxa)whichisbasedonhairalone.“n”issamplesizeofscats.19901992JuneJulyAugust1-13June14-28JuneJuly&AugustAdultscatsn:134438221310Dicrostonyxa)66.2±10.8(78)35.3±5.1(40)34.0±6.0(42)45.3±11.2(32)38.5±11.7(42)28.4±12.2(33)Dicrostonyxb)67.6±10.0(80)36.2±5.4(41)34.3±5.9(42)40.9±10.6(29)35.3±9.8(39)27.7±11.6(32)Microtus3.2±3.2(4)11.1±3.4(12)2.8+1.3(3)33.4±12.5(23)16.1±6.9(18)19.3±9.0(22)Spermophilus1.4±1.4(2)9.8±4.0(11)30.1±6.4(37)31.5±10.3(22)24.6±16.9(27)23.4±8.9(27)Largemammal4.6±3.3(5)3.8±1.8(4)3.2±1.7(4)27.7±10.6(19)9.0±8.4(10)3.0±2.0(3)Largebird7.0±6.0(8)25.4±8.9(29)10.2±3.5(12)7.1±3.1(5)3.3±3.3(4)12.2±9.7(14)Small bird<0.1(0)2.7±2.2(3)0.9±0,7(1)1.8±0.9(1)2.4±1.8(3)0(0)Egg0.6±0.5(1)0.1±0.1(0)0.3±0.2(0)0(0)0.1±0.1(0)1.6±1.0(2)Total84.8±11.589.1±9.381.8±6.9142.3±11.290.8±20.187.3±17.7Table4.5.(continued)19901992JuneJulyAugust1-13June14-28JuneJuly&AugustJuvenilescatsn:17Dicrostonyxa)33.5±5.0(65)Dicrostanyxb)34.9±5.5(68)Microtus2.3±2.3(4)Spermophilus6.6±4.3(13)Largemammal0(0)Largebird3.9±3.6(8)Smallbird3.7±3.7(7)Egg0(0)Total51.5±6.8CTable4.6.Mean(jS.E.)biomass(g)perscatforredfoxandpercentoftotalbiomass(parentheses)ofprincipalpreygroupsinsummer1991.Juvenileandadultscatsarerecordedseparatelyforperiodsinwhichtheycouldbedifferentiated.OtherdetailsasinTable4.5.June2-22June23July7-20July21Aug4-17Augl8-31-July6-Aug3Adultscats:n:292733372246Dicrostonyxa)80.2±13.0(67)26.9±6.1(36)35.7±6.9(51)28.1±6.1(47)12.8±3.3(31)26.4±3.5(36)Dicrostonyxb)80.0±12.0(67)29.0±6.2(39)35.0±5.9(50)26.7±5.5(45)13.5±3.4(32)26.6±3.3(37)Microtus5.2±2.8(4)6.6±2.7(9)9.9±3.1(14)6.0±1.8(10)8.6±2.7(21)11.0±2.3(15)Spermophilus17.7±7.1(15)29.3±6.7(39)13.2±3.6(19)23.6±4.5(40)17.4±5.7(42)30.3±6.2(42)Largemammal1.7±1.4(1)0.2±0.2(0)0.9±0.9(1)1.5±0.7(3)0.6±0.5(1)1.7±0.7(2)Largebird10.9±6.8(9)5.4±44(7)7.5±4.7(11)0.7±0.5(1)0.8±0.6(2)2.8±1.3(4)Smallbird3.3±2.5(3)4.8±2.1(6)3.2±0.9(5)0.9±0.5(2)0.6±0.4(1)0.3±0.2(0)Egg0.1±0.1(0)<0.1(0)0.5±0.2(1)0(0)0(0)0.1±0.1(0)Total118.9±15.075.1±6.270.2±9.259.4±6.641.9±5.572.8±6.0a)Table4.6.(continued)June2-22June23July7-20July21Aug4-17Augl8-31-July6-Aug3Juvenilescats:n:212522Dicrostonyxa)25.5±2.6(69)24.5±3.3(57)18.8±4.3(46)Dicrostonyxb)27.5±2.6(74)26.0±3.4(60)19.9±4.4(50)Microtus3.5±1.9(9)2.2±1.2(5)6.1±2.6(15)Spermophilus5.6±2.8(15)11.9±3.7(28)10.2±3.1(26)Largemammal0(0)0(0)0(0)Largebird0.6±0.4(2)0(0)1.9±1.0(5)Smalibird0(0)3.1±2.2(7)1.4±1.1(4)Egg<0,1(0)0(0)0(0)Total37.2±3.843.1±3.239.5±3.5163In June 1992 fox diet was similar to that in late summer 1990 and 1991 (Tables4.4 and 4.5), with a relatively high incidence and biomass of birds, and mammalsother than lemmings. This pattern appears directly related to the extremely lowlemming densities. Foxes still managed to find lemmings to eat, perhaps by travellinginland or by finding pockets of higher density than those on study grids.For adult scats, estimates of lemming biomass derived from combined weightsof bone and hair (conversion factor 28) were very similar to those derived from hairweight alone in 1990 and 1991, but conversion from hair gave slightly higherestimates in 1992 perhaps because of more complete digestion of bone by food-stressed foxes (Tables 4.5 and 4.6). For juvenile scats, estimates derived from hairwere consistently lower, possibly because of poorer digestive efficiency of juvenilefoxes (Lockie 1959).Juvenile diets were less diverse than adult diets for the same time period(Tables 4.5 and 4.6). They had higher proportions of lemmings, completely lackedlarge mammals, and virtually lacked eggs. These patterns are to be expected giventhat nearly all juvenile foods in these periods would be provisioned by adults.However, adult foxes apparently consumed more ground squirrel and large bird preythan they provided to juveniles.Daily consumption: Since the numerical responses of foxes indicated somefood stress in mid-summer 1991 (pup mortality), and severe stress in 1992 (no pupsweaned; disappearance of adults), it is likely that defecation rates were lower in theseperiods. Biomass per scat remained high in 1990, but declined in 1991, anotherindication of food stress. I assumed no food stress in 1990 or June 1991 (defecation164rate of seven scats/day), and some stress in July and August 1991. (six scats/day). In1992 I assumed a rate of four scats/day (Table 4.7).Total daily consumption was highest in early June and comprised mostly oflemmings at a time when lemmings were particularly vulnerable, with snow meltingand flooding burrows (Table 4.7). Daily consumption of lemmings dropped stronglyby July and tended to fall through the rest of the summer.In 1990 foxes compensated for declines in lemming consumption by increasedconsumption in July and August of voles, large birds, and particularly groundsquirrels. Total daily consumption did not decline strongly as summer progressed. In1991, however, declines in lemming consumption were not well compensated for byincreased consumption of alternative prey, until late August, when ground squirrelconsumption increased substantially. Foxes already made quite heavy use of volesand ground squirrels in June and July, and appeared unable to increase this use tomake up for the decline in lemming consumption. Consequently adult consumptionfell well below 1990 levels by early August, and juvenile consumption was less thanin 1990. In early June 1992, total daily consumption was remarkably high despite lowlemming consumption. However, foxes appeared unable to sustain their relativelyhigh use of alternative prey by mid-June and July, and consumption of all speciesdeclined.Impact on the lemming population: The three habitats used by lemmingscomprised 56.05% of the coastal zone, and 90.94% of the inland zone. In 1990 foxesTable4.7.Adultandjuvenileredfoxdefecationrates(scats/d)andbiomassconsumptionrates(g/d)convertedtopercapitaconsumptionrateofadultandsubadultDicrostonyx,andtotalnumbersofDicrostonyxkilledpersummer(a)1990and1992,and(b)1991.NotethatinJune,percapitadailyconsumptionbyjuvenilefoxeswascutinhalfforcalculatingtotalconsumptionbythefoxpopulation,becausejuvenilefoxesarenotfullyweaneduntillateJune.(a)19901992JuneJulyAugust1-13June14-28JuneJuly&AugAssumeddefecationrate(scats/d)Adultfoxconsumptionrate(g/d)DicrostonyxTotalJuvenilefoxconsumptionrate(g/d)DicrostonyxTotalRatioadult/subadultlemmingsinscatsMeanliveDicrostonyxweights(g)AdultsSubadults7473.2593.6192.5260.43/150.029.47 253.4623.7244.3360.5 3/547.919.57240.1572.63/1043.119.64 163.6569.22/350.029.44 141.2363.23/346.522.24 110.8349.2 0/247.919.5cJ’Table4.7.(continued)19901992(a)(continued)JuneJulyAugust1-13June14-28JuneJuly&AugustNumberofDicrostonyxkilledperdayperadultfoxAdults7. Assumeddefecationrate(scats/d)766666Adultfoxconsumptionrate(g/d)Dicrostonyx560.0203.0210.0160.281.0159.6Total832.3525.7421.2356.4251.4436.8Juvenilefoxconsumptionrate(g/d)Dicrostonyx192.5182.0119.4Total260,4301.7237.0Ratioadult/subadultlemmingsinscats8/64/43/73/81/54/17MeanliveDicrostonyxweights(g)Adults50.346.546.744.144.341.0Subadults29. approximately 1430 adult lemmings (Table 4.7). Twenty-one adultlemmings (31% of the adult population) were killed by foxes on the 36.6 ha oflemming habitat on study grids (Table 4.1). This converts to 702 adults killed in thecoastal zone (21.82 km2), with the remainder being killed in 12.68 km2 of lemminghabitat, or 13.94 km2 total area, in the inland zone. In 1990 the fox populationrequired a minimum area of 35.76 km2.In 1991, foxes killed a minimum of four adult lemmings (8% of all adults) onstudy grids. This converts to 134 adults in the coastal zone, the remainder (1,105)being killed on 101.38 km2 of good habitat, or 111.48 km2 of total habitat, inland. In1991, total area hunted by foxes was 133.30 km2. Actual predation impacts and areasused by foxes are probably intermediate between these quite different annualestimates.In 1990 the estimated fox hunting range included 2,491 ha of lemming habitat.Lemmings weaned 0.38 litters/ha of lemming habitat, with a mean litter size of five.This converts to 4,733 weaned subadults. Of these, an estimated 2,189 (46.2%) werekilled by foxes (Table 4.7a). In 1991, hunting range included 11,361 ha of lemminghabitat, and lemmings weaned 0.19 litters/ha lemming habitat, or 10,793 weanlings.Of these, 1,852 (17.2%) were killed by foxes. The percentage killed in either year waslikely intermediate between these estimates.Foxes tended to take more females than males during the summer, based ondeaths of radiocollared adults (Table 4.1), and the ratio of male to female ratio ofpelvic bones in scats: 1990 (0:4), 1991 (12:14).Functional response: The relationship between number of adult lemmings170killed! fox! day and the adult lemming density (Fig. 4.3) is adequately described bya linear model (y = 7.486 x, r2 = 0.87, F=72.556, P < 0.001). However, more of thevariance is explained by an exponential model (y = 10.995 x2, r2 = 0.91). I concludethat the functional response of foxes to adult lemming density approximates a typeIII curve, but without any data at high enough lemming densities to indicatesaturation of fox kill rates (Fig. 4.3).Up to an adult lemming density of 0.65/ha, foxes showed a slow increase innumber of adult lemmings killed!d. These data were from time periods when thediet was comprised largely of prey other than lemmings (e.g., July and August),perhaps indicating prey switching. The 1992 data are plotted at densities on studygrids, which may have somewhat underestimated regional densities. Nevertheless,even when lemmings were extremely rare, foxes continued to consume one or twoeach day, and the lemmings lacked a low density refuge from this predation. Atlemming densities > 0.65/ha, foxes increased their kill rate markedly. These datawere from June.I cannot estimate the functional response to the entire lemming population,because I lack accurate estimates of subadult densities. I suspect that the responsewould be closer to linear, with a weaker rate of increase with density, because foxesconsumed more subadults/day as these became available later in the summer.In summary, foxes can be classed as semi-generalists as far as their diet isconcerned. They fully or partially compensate for declining lemming abundance, andexhibit features of a type- III response.171Fig. 4.3. Functional response of adult foxes to adult lemming density. Data plotted arethe number of adult lemmings ingested per day by each adult fox to feed itselffor each of the time periods in Table 4.4.1721o.8 1990az19917E19926EE— 432E0 : 0:2 0:4 06 08 12Adult lemming density (# I ha)173Grizzly Bear DietOf 16 bear scats collected over three summers, 10 (63%) contained arcticground squirrel remains, and three (19%) contained collared lemming remains.Rough-legged Hawk Feeding ExperimentWhen fed two lemmings per day the captive hawk lost a little weight, digestedlemmings somewhat more thoroughly, and cast pellets at a slightly slower rate thanwhen fed four a day, when it gained weight (Table 4.8). It discarded no lemmingheads when dismembering prey, and generally swallowed lemmings whole, up tofour at a time. It removed the gastrointestinal tract more frequently when fed at thehigher rate. It consumed all food offered, except for one day at the higher rate, whentwo lemmings were not consumed within 24 h. I derived the following correctionfactors, based on the proportion of lemmings fed which were recovered in pellets(Table 4.8), to convert number of individual lemmings in pellets to number ingested:when slightly food-stressed (74 g/d) 1.27; and when well-fed (137 g/d), 1.02.Over the entire experiment, the ratio of adult to subadult lemmings offeredwas 26/28, but the ratio recovered in pellets, based on intact molar tooth rows indentary and maxillary bones, was 9/4. Of adults fed, 34.6% were recovered with atleast one tooth row intact, whereas only 14.3% of subadults were recovered. I usethese proportions to correct observed ratios in pellets collected in the field.Rough-legged Hawk DietComposition: In percent frequency of occurrence, lemmings dominated therough-legged hawk diet throughout the summer (Table 4.9). Lemming biomassingested dropped substantially in late summer (Table 4.10), and hawks consumed an174Table 4.8. Summary of conditions and results of the captive feeding experimentwith a rough-legged hawk, including proportion of lemmings recoveredin pellets, and pellet casting rate.First Secondperiod periodFeeding rate: lemmings/d 2 4g/d (± S.E.) 74.0 (± 2.0) 136.5 ( 6.23)Hawk weight (kg) - start 1.340 1.325- end 1.325 1.505Duration (d) 7 12Number of pellets cast 6 13Number (%) lemmings 11 of 14 (79%) 39 of 40 (98%)recoveredMean lemmings/pellet 1.83 (± 0.31) 3.00 (± 0.42)(S.E.)Casting rate (pellets/d) 0.9 1.1Table4.9.Percent frequencyofoccurrenceof allremainsinrough-leggedhawkpellets.199019911992BeforeIncubationNestlingBeforeIncubationNestlingBeforeRestofincubationincubationincubationsummer(to9June)(to7July)(to25Aug)(to1June)(to5July)(to15Aug)(to3June)(n=13)(n=92)(n=62)(n=42)(n=24)(n=31)(n=8)(n=7)Mammalia:DicrostonyxMicrotusSpermophilusAyes:AnseriformCharadriiform/PasseriformUnidentifiedEggshellInsecta:Coleoptera/LepidopteraVegetationSand/pebbles100988295961001008601313101723132901260013014250081813327 0 0 0 0 0 0000200122060010003000224300224064887413141368393845100100ciiTable4.10. ConversionofDicrostonyxremainsinrough-leggedhawkpelletstoconsumptionratesofindividuallemmingsandlemmingbiomass(g/d)byadulthawksduringtheprincipal timeperiodsofthestudy.199019911992BeforeIncubationNestlingBeforeIncubationNestlingBeforeincubationincubation9June(to9June)(to7July)(to25Aug)(to1June)(to5July)(to15Aug)Conversiontoindividualsingested/d:Mean(S.D.)individualremains/pelletCorrectionfactor1Mean#individualsingested/pelletCastingrate(pellets /d)Mean#individualsingested/dConversiontoadultandsubadultlemmingconsumptionrates:RatioadulttosubadultremainsCorrectedratio2Numberof adultsingested/dNumberofsubadultsingested/d2.15(0.90)1.022.19 1.1 2.4111/832/560.88 1.531.97(1.16)1.272.500.9 2.2558/44168/3080.79 1.461.29(1.32)1.271.640.9 1.4818/2352/1610.361.121.81(1.15)1.021,85 1.1 2.0423/1966/1330.681.362.17(1.27)1.272.760.92.4815/643/421.251.231.84(1.29)1.272.340.9 2.1115/1443/980.641.471.38(0.74)1.271.750.9 1.583/39/210.471.11a)Table4.10.(continued)199019911992BeforeIncubationNestlingBeforeIncubationNestlingBeforeincubationincubation9June(to9June)(to7July)(to25Aug)(to1June)(to5July)(to15Aug)Conversiontototal consumption:Numberof adulthawks88412126Period(d)21(9)284921(0)3441Total(summer)consumptionbyadults:Adultlemmings396(311)838(667)Subadultlemmings804(657)1207(864)Conversiontobiomassingested/d:Meanadultweight(g) biomassingested(g/d)89.082.438.573.893.163.255.9Tocorrectforproportionofingestedindividualsrecognizableinpellets(1.02whenwell-fed; 1.27whenfood-stressed)2Tocorrectfor differentialdigestionofadultandsubadultjaws (34.6%adultjaws recovered;14.3%subadultjaws recovered)Pre-incubationperiodisapproximateandbasedondatainPooleandBromley(1988).FiguresinparenthesesrefertonumbersofdaysinJuneusedtocalculatethesummer(June, JulyandAugust)consumption.178increasing proportion of voles, ground squirrels and birds (Table 4.9).I lack data to convert all alternative prey remains to biomass ingested.However, if I assume the same correction factors and casting rates for pelletscontaining voles as those of lemmings, and use mean vole weights from live-trapping(both age classes combined), then the voles provided between 3.9 and 9.2 g/d, whichdoes little to compensate for the declining rate of lemming consumption. Given thehigh rate of nest abandonment close to hatching, it appears that many rough-leggedhawk pairs were unable to compensate for the declining rate of lemmingconsumption. A few pairs were successful each year, and therefore were able to killsufficient alternative prey to feed themselves and a few young. Prey remains afterfledging of two young from one nest in 1990 included one ground squirrel, twoCalidris sandpipers, one duck, and small passerine feathers.Impact on lemming population: On study grids in 1990, hawks killed at leastseven adult lemmings (10% of all adults), and, in 1991, seventeen adult lemmings(33%). Extrapolating from the 36.6 ha of lemming habitat on grids, hawks removed234 adults from the coastal zone in 1990, and 568 in 1991. The remainder of theirestimated summer kill (77 in 1990, 99 in 1991) was from inland, giving a total huntingarea of 26.25 km2 in 1990, and 24.17 km2 in 1991.In 1990 this hunting range included 1,626 ha of good habitat, and lemmingsweaned 0.38 litters/ha lemming habitat. With litter size of 5, this converts to 3,089weanlings available, of which hawks consumed an estimated 657 (21.3%) (Table 4.10).In 1991, hawk hunting range encompassed 1,437 ha of good habitat, and lemmingsweaned 0.19 litters/ha lemming habitat. This converts to 1,365 weanlings available, of179which hawks consumed an estimated 864 (63.3%).There was a tendency for hawks to take more males than females based ondeaths of radiocollared adults (Table 4.1). Sex ratio (male:female) data from pelletswere limited and showed no clear trend: 1990 (1:2), 1991 (3:1).Functional response: The rate at which adult hawks ingest adult lemmingsgenerally increases with increasing adult lemming density (Table 4.10, Fig. 4.4). Alinear model explains a significant proportion of the variance (y = 1.031 x, r2 = 0.85,F = 34.095, P = 0.001), although there is a lot of scatter. No other meaningfulrelationship clearly fits the data (Fig. 4.4). I conclude that the functional response istype-IT, without any evidence of saturation of kill rates at higher densities.The wide scatter may in part result from inaccurate data. I lack accurateestimates of lemming density in the pre-incubation period (May), but use the highestdensity in early June to represent this period (two data points with the highestdensities in Fig. 4.4). The 1992 pre-incubation data point is plotted at the meandensity on study grids, which may have underestimated regional densities. Acrossthe range of lemming densities observed, the rough-legged hawks did not increasetheir kill rates as fast as foxes (Fig. 4.3), indicating that hawks are less flexible in theirresponses to lemming abundance.In summary, rough-legged hawks were specialists, being unable to compensatefor declining lemming abundance, and having a type-IT functional response.Diets of Other RaptorsLemmings comprised a substantial proportion, on a frequency of occurrencebasis, of the diet of all other raptors and gulls at Pearce Pt. (Tables 4.11 and 4.12),180Fig. 4.4. Functional response of adult rough-legged hawks to adult lemming density.Data plotted are for pre-incubation periods in each year, and incubation andnestling periods in two years (Table 4.10), and represent adult lemmingsingested by an adult hawk to feed itself.1811,4-1.2199o‘1991)E19920.8EE0.6•1.’- 0.4 -0.2EZ Uu 02 0:4 06 08 1.2Adult lemming density (# I ha)Table4.11. Percentfrequencyofoccurrenceofallitemsinperegrinefalconpellets.199019911992BeforeIncubationNestlingBeforeIncubationNestlingBeforeIncubationNestlingincubationincubationincubation(to9June)(to7July)(to25Aug)(to1June)(to5July)(to15Aug)(to3June)(to15July)(to30Aug)n=3n=28n=40n=1On=28nz18n=13n=22n=50Mammalia:Dicrostonyx674325801433231416Microtus0713000054Spermophilus00100017004Unidentified040000000Ayes:Anseriform0211320326151826Charadriiform/1004678507178927784PasseriformUnidentified01101006050Insecta:Coleoptera000040002Vegetation0251560615084112Sand/pebbles04365606144100100100—---------———---------——-----------—-—-----------------00 toTable4.12.Percentfrequencyofoccurrenceofallitemsinpelletsofgyrfalcons,gullsandgoldeneaglesduringthreecollectionperiods(I-beforelaying; II-incubation;III-nestling) (DatesasindicatedinTable4.11).199019921991199219911992AnseriformGalliformCharadriiform/PasseriformEggshellUnidentified55430700025002300537100000000009290411410051402771377366970113643050902330111802301100000003200095860051416140011000160GyrfalconGullGoldeneagleIIIIIIIIIIIIIIIIIIII&IIIAllAll2274151413209117197Mammalia:DicrostonyxMicrotusSpermophilusRangiferVu!pesAyes:014000514071423302291405027001022180015111871151011929CoTable4.12(continued)1990Gyrfalcon19921991Gull1992Goldeneagle19911992IIIIIIIIIIIIIIIIIIII&IIIAllAllInsecta(Coleoptera)EchinodermataMolluscaCrustaceaVegetationSand/pebbles0000000000003243753271100000000005364931000500015330800001011977555606955331000 0 0 14 57 1002143000000897163100185though they were not the dominant prey of these predators, except perhaps for gulls.The incidence of lemmings in pellets generally declined with declining lemmingdensities through the summers. However, even in summer 1992, when lemmingswere rare, they were still remarkably common in gull and peregrine falcon diets (seealso Table 4.1). Although many of the species in gull diets may have been scavenged,I think that lemmings were primary gull prey as no other predator left lemming partsavailable to be scavenged.I lack data on casting rates, and frequency of ingestion of lemming heads forall these predators, so cannot accurately estimate total impact. However, theincidence of individual lemmings in pellets of these predators (Table 4.13) indicatesrelatively high mortality rates by gulls and peregrine falcons, and continuedpredation by these species in 1992 when lemmings were scarce. Casting rate forperegrines, as with other falcons (Balgooyen 1971, Duke et al. 1976), is likely close toone pellet per day. If we assume that peregrine falcons ingested all lemming heads,the number of lemmings per pellet (Table 4.13) is particularly high in early summer,reaching 77% of the incidence in rough-legged hawk pellets in pre-laying period 1991(Table 4.10), and is considerably lower in late summer. Each adult peregrinecontinued to eat one lemming every one or two weeks in 1992.DISCUSSIONGeneralist Predator HypothesisThe generalist predator hypothesis predicts that lemming numbers remainrelatively constant because (i) dominant predators do not respond numerically tolemming density, (ii) declining lemming biomass in the diet is compensated for byTable4.13. Mean(S.D.)numberofindividuallemmingremainsperpellet,basedonmolarteeth,foravianpredatorsotherthanrough-leggedhawksduringpre-incubation(I), incubation(II),andnestling(III)periods.SamplesizesasinTables4.11&4.12.Blankcellsindicatenopelletscollected, exceptallperiodslumpedforeagles, andperiodsII&IIIlumpedforgullsin1992.199019911992IIIIIIIIIIIIIIIIIIPeregrinefalcon0.670.540.201.400. (0.26)(0.55)(0.38)(0.47)(0.24)Gyrfalcon-0.320.57--00.070(0.65)(0.79)(0.26)Goldeneagle--0.470(1.12)Gull-- (1.29)(0.33)(0.67) (0.53)187increased consumption of alternative prey, (iii) predator functional responses are typeIII, and (iv) predators consume very few lemmings at low lemming densities. Thedata from Pearce Point do not provide consistent support for the generalist predatorhypothesis as an explanation for the constancy of lemming densities. Somepredictions were satisfied for some predators, but all predictions were never satisfiedfor any one predator.The rough-legged hawk acts as a lemming specialist, with a strong numericalresponse to changing lemming numbers, an inability to compensate for declininglemming consumption with alternative prey, and a type-IT functional response. Otherstudies also show that small rodents are the dominant prey of rough-legged hawkson arctic tundra, with various birds (principally passerines) and arctic groundsquirrels comprising a minor component of the diet (White and Cade 1971, Springer1975, Smith 1987, Popanov 1988). Rough-legged hawks also show rapid numericalresponses in nesting densities to fluctuating microtine densities (White and Cade1971, Poole and Bromley 1988), so could be considered nomadic specialists (Galushin1974).The red fox is a semi-generalist. In some summers it is able to compensate fordeclining lemming consumption with alternative prey, and exhibits some features ofa type-ITT functional response. However, foxes still rely primarily on lemmings,especially in spring, and fail to breed successfully when lemming densities are verylow in spring. I know of no comparable study of red fox on arctic tundra.The other raptors,- peregrine falcons, gyrfalcons and golden eagles,- react tochanging lemming densities as generalists. Lemmings are not their principal prey, so188they show little or no numerical response to lemming density, and are able tocompensate for declining lemming availability by increased use of their principalprey. I do not know how they responded functionally to changing lemmingabundance. Previous studies of tundra raptors have shown little or no use ofmicrotine rodents by gyrfalcons (White and Cade 1971, Poole and Boag 1988), byperegrine falcons (White and Cade 1971, Hunter et at. 1988), or by golden eagles(Poole and Bromley 1988). The results from Pearce Point are unusual, especially forperegrine falcons. Diet estimation at Pearce Point was based on pellets, but Whiteand Cade (1971) and Hunter et a!. (1988) estimated diet from prey remains at nests.The latter method tends to underestimate the contribution of small mammals toraptor diets, because the entire small mammal is often ingested, or small bones arelost in the nest structure (Marti 1987, Simmons et a!. 1991). This might partly explainthe paucity of lemming remains in previous reports of peregrine falcon diets. Theheavy use of ground squirrels by gyrfalcons probably reflects a regional scarcity ofptarmigan in summer. Glaucous gulls may be lemming specialists in this system, butthe data are inconclusive.Ermine are often thought of as microtine specialists, yet they can breed at verylow lemming densities at Pearce Pt., and likely survive by broadening their prey baseto include various birds. Grizzly bears and arctic ground squirrels probably functionas generalists since they are largely herbivorous, and, in the case of bears, consumesubstantial numbers of ground squirrels (Pearson 1975).None of the predators at Pearce Pt., except gyrfalcons and golden eagle,satisfied the prediction (Rosenzweig and MacArthur 1963) that predation on189lemmings be rare at very low lemming densities, because of low capture efficiency.Lemmings lack secure hiding places, especially from mammalian predators which digto reach adults and juveniles in burrows. When above ground, lemmings have littlecover. Perhaps more than any other factor, this lack of secure refuge allows principalpredators to drive lemmings to such low densities, and allows such a diverseassemblage of predators to persistently prey on lemmings even at low densities.Rosenzweig and MacArthur (1963) also predicted that in a system with stableprey, predator populations should be limited by factors other than food, and likely byterritoriality. However, at Pearce Point, rough-legged hawks and red foxes appearedfood-limited, and there was little evidence of territoriality. Nest abandonment byhawks, and fox pup mortality coincided with the rapid decline of lemmings in lateJune and July. The total daily consumption fell below maintenance levels for adulthawks in late summer, and for foxes in summer 1991. Rough-legged hawks areapparently food-limited in central arctic Canada, with numbers of successful nests,and mean brood sizes correlated with microtine prey abundance (Poole and Boag1988, Poole and Bromley 1988). In north boreal Sweden, where foxes feed primarilyon cyclic microtine prey, Englund (1970) found a positive correlation between indicesof reproduction and vole abundance; foxes appeared food limited. Rough-leggedhawks and peregrine falcons acted aggressively towards one another whenestablishing nests in spring, but their hunting ranges appeared to overlapconsiderably. No raptors nested within 1.55 km of the golden eagles, an avoidancenoted by Poole and Bromley (1988). Red fox regularly hunted throughout huntingranges of all raptors, and ground squirrels were distributed throughout the study190area. Consequently, it seems likely that there was substantial exploitative competitionbetween predators for the declining prey base of lemmings.In summary, the generalist predator hypothesis is not a sufficient explanationfor the relative constancy of lemming numbers at Pearce Point. Although some of thepredators acted as generalists, the dominant predators were specialists or semi-generalists. When these species were present in summer, lemmings declined todensities which could no longer sustain breeding by hawks in the current summer,and which would not allow settlement and successful breeding by hawks and foxesthe subsequent spring. The system was unstable in summer because of theinsufficiency of alternative prey for all predators, the lack of territoriality amongpredators, and the lack of secure refuges for lemmings at low densities.ProceduresDetermination of predator diets and impact depends on numerousassumptions which might bias results. Here I discuss the most importantassumptions. First, I used 28 as the factor to convert undigested remains in fox scatsto an estimate of biomass consumed. This choice was validated by the very similarestimates of biomass derived from another method, conversion of weight of lemminghair in scats. Second, I assumed that the conversion factor does not change with foxage or degree of satiation. Juvenile foxes, and well-fed foxes, digest food lessthoroughly (Lockie 1959, Reynolds and Aebischer 1991). Consequently, consumptionrates for juvenile foxes, and for foxes in early June, may have been overestimated. Ihave no way of correcting this. In most time periods the estimated total dailyconsumption fell within or close to the range of estimated daily requirements: 320-380191g/d for captive foxes (Lockie 1959, Sargeant 1978) and 450-550 g/d for more active,free-ranging foxes (Scott 1941., Ryszkowski et at. 1973). Estimated daily consumptionwas particularly high in early June, perhaps in part because of the above bias.However, lemmings were more vulnerable at this time, as snow melt removed coverand most burrows were flooded. This was the only time foxes cached lemmings,indicating they were satiated. Foxes may eat more in early summer as they recoverbody condition after the winter.Third, I assumed that the proportion of adult to subadult remains in fox scatsrepresented the proportions ingested. I think this is the most problematic assumption.Foxes digest bone. Thinner, weaker subadult lemming bones are likely betterdigested, as observed when feeding lemmings to a captive hawk. Consequently Iprobably overestimated the consumption of adult lemmings by foxes, andunderestimated subadult consumption, but to an unknown degree.Fourth, estimates of total impact on the regional lemming population werederived by extrapolating observed mortality rates from study grids, which were smallcompared with foraging ranges of predators. Grids were from one to four kilometresfrom the fox den, and experienced a wide range of rates of fox predation. This givesconfidence that regional fox predation rates were within the observed range.However, one grid was within 500 m of a hawk nest in 1990, and immediatelyadjacent to a nest in 1991. Hawk predation rates from grids may have overestimatedregional rates, and so I may have underestimated the area used by hunting hawks.Impact on the Lemming PopulationI lack data on the hunting ranges of the principal predators, but estimated the192areas required by foxes, to consume all the adult lemmings indicated by analysis oftheir scats, as 35.8 km2 in 1990 and 133.3 km2 in 1991. The largest home range recordfor red fox in North America, in an alpine tundra habitat in northern BritishColumbia, is 34 2 (Jones and Theberge 1982, Voigt 1987). The foxes werefrequently observed travelling throughout the 40 km2 intensive study area. The natalden was only one km from the border of this area, so foxes undoubtedly used aneven larger area for hunting. Despite the fact that estimates of hunting ranges mayhave been too large, because of potential biases in scat analyses (above), they areclose to reported values and to my direct observations of area used by foxes, so arereasonable.All hawks together needed at least 24-27 km2 to satisfy observed kill rates.These estimates seem reasonable given that one pair of rough-legged hawks, in thenorthern taiga of Finland, used a 10 km2 foraging range to feed nestlings in partiallyforested bog habitats (Pasanen and Sulkava 1971).Recruitment of summer-born lemmings as adults, on grids accessible topredators, was very low (0.12 - 0.21 individuals! ha! summer), and did not reachhigh levels in the predator exclosure (0.44 individuals! ha! summer) (Chapter 1). Thestudy grids, including the exclosure, fell within the foraging ranges of hawks andfoxes, and therefore subadults on these grids probably were killed at near the ratesestimated: in 1990 a combined 68% for the two predators, and in 1991 a combined81%. The various other predators could also have consumed substantial numbers ofsubadults. These results confirm my contention that most subadult lemmings werekilled by predators before maturing, and that the lemming population in the predator193exciosure did not grow because most potential immigrating recruits were killed(Chapter 1).Individual foxes have a substantially higher limiting effect on lemmingpopulation growth than individual rough-legged hawks, because of their higher dailyfood requirements, and their tendency to take more females than males. The latterpattern has also been observed by Brooks (1993) and Krebs et a!. (1995). Predation onfemales removes not only current adults, but also the litters currently suckling and inutero.The Specialist-Generalist Continuum.There is evidence for the generalist predator hypothesis from Europeanstudies. Microtine populations tend to be non-cyclic in southern latitudes (south of60°N). In these areas the rodents belong to a diverse community of herbivoressupporting a number of generalist predators, principally the red fox (Vulpes uulpes)and the common buzzard (Buteo buteo), and winter snow is not deep enough toprevent successful hunting by these predators (Hansson and Henttonen 1985, 1988).In one such community in southern Sweden, Erlinge et a!. (1983, 1984), and Erlinge(1987), demonstrated that predation is the principal factor limiting small rodentpopulation growth. In this system, generalist predators showed strong functionalresponses to rodent densities by switching from their principal prey, the rabbit(Oryctolagus cuniculus), to rodents, with short time delay, when rodents wereparticularly vulnerable and abundant. The numbers of generalist predators werelimited by territoriality. Erlinge (1987) concluded that the conditions proposed byRosenzweig and MacArthur (1963) were met in his system.194Korpimaki and Norrdahl (1989, 1991a,b) tested the nomadic predatorhypothesis with cyclic microtine rodents in western Finland. They demonstrated thatsome raptors, whose principal foods are microtine rodents, show rapid numericalresponses to microtine densities through immigration and enhanced breeding successin areas with irruptive rodent populations. These raptors showed strong type-TIfunctional responses to rodent density (Korpimaki and Norrdahl 1989, 1991a). Theymay have influenced the amplitude of the cycle, by curtailing irruptive populationgrowth, but rodent densities still fluctuated cyclically. The raptors alone wereinsufficient to stabilize prey densities. Rosenzweig and MacArthur’s (1963) conditionswere not assessed for all predators, but probably did not hold for these raptors,which were weakly territorial (Korpimaki and Norrdahl 1991a).Specialist mustelid predators, most notably the least weasel (Mustela nivalisspp.), are the dominant small rodent predators in a number of communities whererodents fluctuate cydically. Their strong, but delayed numerical responses to highrodent numbers, their total reliance on rodent prey in winter, and their adaptationsfor hunting rodents in their burrows, act together to drive the rodents to densities toolow for continued support of the predators (Fitzgerald 1977, Korpimaki et a!. 1991,Korpimaki 1993).The Pearce Point community lies somewhere intermediate on this specialistgeneralist spectrum. In contrast to the northern European systems, from which thespecialist-generalist model is derived, collared lemmings at Pearce Point (i) are facedevery summer by a generalist predator population, and (ii) are able to breed mostwinters, and thereby recover from very low densities in autumn.195The migratory and resident generalist predators at Pearce Point stabilizelemming numbers in summer, because they limit lemming population growth, evenwhen one or other of the principal, more specialist, predators has disappeared. Themore specialist predators (hawks and foxes) destabilize lemming numbers becausethey drive summer lemming declines with no temporal or spatial refuge.Winter and spring breeding under the snow is stabilizing at Pearce Pointbecause it allows lemmings to return to early summer densities high enough tosupport the full complement of summer predators. Winter breeding in northernFennoscandian Microtus, Clethriononiys and Lemmus spp. is generally associated withthe increase phase of a population irruption (Hansson 1984b, Kaikusalo and Tast1984). The lack of breeding by these rodents in most winters is likely destabilizing,because they cannot compensate for continuing winter predation, and may decline todensities which cannot support specialist or semi-generalist predators the subsequentsummer. This pattern occurred in winter 1991-92 at Pearce Point (Chapter 3).Diets and numerical responses of red fox and rough-legged hawks at PearcePoint are similar to patterns observed in some boreal systems. In these regions somegeneralist predators, notably red fox, preferentially prey on small rodents, but take anincreasing proportion of alternative hare and grouse prey when rodents are scarce.However, the predators cannot fully compensate for declining vole consumptionwhen voles are scarce, and therefore decline in abundance themselves; they are semigeneralists (Angelstam et a!. 1984, Lindstrom et a!. 1987). The alternative preyhypothesis has been proposed to explain the resulting cyclic variations in all preyspecies (Keith 1974, Angelstam et a!. 1985). However, lemmings at Pearce Point do196not show multiannual cycles.Pearce Point differed from these boreal systems in that microtine preydynamics were stabilized by winter breeding (Chapter 3), and most or all of thegeneralist predators were absent in winter. Peregrine falcons, eagles, bears andground squirrels were not active residents in winter, but I lack information on thewinter ecology of red fox. Foxes showed high site fidelity between summers, judgedby the presence of recognizable adults. They may have been resident for part of thewinter, though what sustained them, given the low autumn lemming densities andthe hibernation or emigration of key alternative prey, remains a crucial unansweredquestion.TA/hat Allows Generalist Predators to Persist ?Arctic ground squirrels appear to be a key species in understanding how thediversity of semi-generalist and generalist predators persist at Pearce Point. For redfoxes, they are the main alternative prey to lemmings in terms of biomass, and foxesprobably could not persist without them. They are also a principal prey for generalistpredators such as the golden eagle, gyrfalcon, and grizzly bear. They provide asubstantial source of food in late summer for the specialist rough-legged hawks, andthe generalist peregrine falcons. They are also lemming predators themselves(Boonstra et al. 1990, Chapter 1). I therefore hypothesize that, without groundsquirrels, the system would likely lose the three generalist lemming predators and thered fox, leaving a relatively underutilized lemming prey base ready for exploitationby more specialist lemming predators, such as the rough-legged hawk, arctic fox(Alopex lagopus), and jaegers (Stercorarius spp.).197CHAPTER FIVEWHY DON’T ALL LEMMING POPULATIONS IRRUPT?INTRODUCTIONMicrotine rodents, including North American lemmings (Lemmus spp.,Dicrostonyx spp.), are well known for their density fluctuations (Elton 1942, Krebs andMyers 1974). Virtually all populations of Lemmus studied to date fluctuate widely indensity (Stenseth and Ims 1993). Most populations of collared lemmings (Dicrostonyxspp.) studied in North America also irrupt in some years, exhibiting greater than tenfold changes in density (Elton 1942, Shelford 1943, Krebs 1964, Fuller et a!. 1975b,Mallory et a!. 1981, Rodgers and Lewis 1986, Pitelka and Batzli 1993).However, not all collared lemming populations show such wide densityfluctuations. In this study, densities of Dicrostonyx kilangmiutak did not follow amultiannual pattern, with a clear irruptive increase to a peak, followed by a decline.Instead they stayed at consistently less than three individuals per hectare for sixyears, with annual fluctuations (Krebs et a!, 1995, Chapter 1 ). Accurate densityestimates have not been made in all previous studies of collared lemmings, especiallywhen they are rare (e.g., low phase of cycles) (Krebs et a!. 1995). However, estimatesof numbers per hectare at the relatively short-lived peak of an irruption include 40(Shelford 1943), greater than 40 (Watson 1956), 15-25 (Brooks and Banks 1971), and 27(Batzli et a!. 1980). Non-irruptive populations of Dicrostonyx also exist in the northernfoothills of the Brooks Range of Alaska (Batzli and Jung 1980, Pitelka and Batzli1993).198In this chapter I investigate potential reasons why some lemming populationsirrupt while others do not irrupt. I ask the following questions: (i) what are thecharacteristics of a population irruption from low densities in lemmings ? (ii) whatare the geographic patterns of irruptions in North American lemmings ? (iii) what arethe factors necessary to limit lemming populations to low densities ? and (iv) how dothese differ between regions where lemmings irrupt and regions where lemmingpopulations are relatively stable (e.g., Pearce Point) ? I end with a set of alternativehypotheses suggesting why some lemming populations are relatively stable.CHARACTERISTICS OF POPULATION IRRUPTIONSI define a lemming irruption as sustained population growth exceeding a tenfold increase in density, and reaching a high density (>10 lemmings/ha) which is notrepeated annually. This is essentially Taitt and Krebs’ (1985) definition of a cyclicpopulation fluctuation in microtine rodents, where the irruption is the increase phase.Because of the short duration of many lemming studies, I do not assume cyclicitywith attendant increase phases, but speak of irruptions. Ten-fold increases in densitycan occur in a non-irruptive situation, where densities are consistently less than 3/ha,where highest densities are reached almost every year, and where lemmings neverreach the abundance of cyclic peaks (see above)(Chapter 1; Krebs et al. 1995).Therefore a peak density of more than ten individuals per hectare is a necessarycriterion to differentiate irruptive and non-irruptive situations. This represents thelower limit of an order of magnitude difference between maximum densities at anon-irruptive situation (Pearce Point) and at irruptive situations.199The period of low density is frequently the longest of all phases in microtinepopulation cycles (Krebs and Myers 1974, Stenseth and Ims 1993), and this holds formost populations of lemmings studied over many years in North America (Shelford1943, Batzli et at. 1980, Rodgers and Lewis 1986). If the population has declined frommuch higher densities, the length of the breeding periods may be shorter, and therate of maturation slower, at the beginning of the low period compared withsubsequent year(s) (Krebs 1964, Fuller et at. 1975b). However, reproduction continues,and litter sizes at low densities do not differ substantially from those at higherdensities (Krebs 1964, Batzli et at. 1980). Low density populations frequently declinesomewhat in summer, despite reproduction, and often recoup these losses by winterbreeding (Krebs 1964, Fuller et at. 1975b).The scale and speed of population growth in an irruption is obviously theresult of successful reproduction, and repeated recruitment of successive cohorts tothe breeding population (Krebs 1964, Batzli et at. 1980). An irruption often starts withan increasing summer population (Shelford 1943, Fuller et a!. 1975b, Batzli et at. 1980).The shift from low to irruptive densities is always accompanied by winter breeding(Krebs 1964, Fuller et at. 1975b, Batzli et at. 1980). Small increases in survival of adultfemales and their offspring can have dramatic impacts on density, because femalesurvival ensures weaning of the current litter, and with post-partum oestrus,juveniles and subadults rapidly comprise the majority of the population (Batzli et at.1980). There is no indication that immigration plays a key role in initiating irruptions,because low densities generally occur synchronously over quite large regions (Elton1942, Krebs 1964).200Any attempt to discriminate between non-irruptive and irruptive populationsmust address the factors limiting reproduction and recruitment in both summer andwinter, specifically: survival of adults, reproductive output of adults, survival ofneonates to weaning, survival of weanlings through dispersal, and maturation rate ofjuveniles.GEOGRAPHIC PATTERNS IN LEMMING IRRUPTIONSThe most comprehensive attempt to document regional trends in arctic smallmammal abundance was the Canadian Arctic Wild Life Enquiry undertaken from1935 to 1949 by Charles Elton, and Dennis and Helen Chitty of the Bureau of AnimalPopulation at Oxford University. The researchers asked resident observers (mostlyHudson’s Bay Company employees), by questionnaire, to record whether smallrodent, arctic fox and snowy owl abundance had increased, decreased or remainedthe same since the previous year. Records of change in small rodent numbers werebased on observations by the respondent and other local trappers and traders. Datawere summarised annually within 11 regions (Fig. 5.1). Here I summarize the resultsacross all 14 years for the 11 regions, giving the proportion of observers reportingsubstantial change in the small rodent population (i.e. increase or decrease inabundance), and the proportion indicating no change in abundance.In regions 1,2,5,6,7 and 8 (Fig. 5.1) observers infrequently (<20% of reports)classified small rodent populations as “no-change”, indicating that the rodentsfluctuated frequently (Table 5.1). Reports of increase exceeded reports of decrease,often by a substantial margin, suggesting that declines in abundance were oftenprecipitous. Fluctuations of lemmings were remarkably regular, with peak abundanceTable5.1.SummaryofdatacollectedbytheCanadianArcticWildLifeEnquiry, concerningrelativestabilityinlemmingabundance, for1935to1949, inelevenregionsoftheCanadianarctic. Numbersarethepercentagesofallobserverreports, summedoverthefourteenyears,indicatingannual increase,decrease, ornochangeinthelemmingabundance. ForregionlocationsseeFig.5.1. DataaretakenfromdatasummariesinChittyandElton(1937),Chitty(1938),Chitty(1939),Chitty(1940),ChittyandChitty(1941),ChittyandNicholson(1942),Chitty(1943),ChittyandChitty(1945)andChitty(1950).RegionsLemmingabundance1234567891011Increase4963273459594947373531Decrease4035184222373940371931Nochange11.355241941212274638C202Fig. 5.1. Map illustrating jurisdictions with arctic tundra (A, Alaska; Y, Yukon; N,Northwest Territories; Q, Quebec; L, Labrador), the eleven regions (numbers)summarized by the Arctic Wild Life Enquiry (Chitty 1950), and, in northernAlaska, the division (dotted line) between the coastal plain and the foothills(after Hartman and Johnson 1984). Study sites mentioned in the text areindicated with a dot and label (CS, Cape Sabine; A, Atkasook; W, Wainwright;PB, Point Barrow; P, Pitt Point; U, Umiat; BI, Banks Island; PP, Pearce Point;BL, Baker Lake; TL, Truelove Lowland; I, Igloolik; EP, Eskimo Point; C,Churchill; CP, Cumberland Peninsula).C204every three or four years (Elton 1942, Chitty 1950).In regions 3, 10 and 11, reports of no-change were frequent (Table 5.1),suggesting more constancy in rodent numbers. On the western Canadian arcticmainland (region 11), many observers did not report substantial fluctuations inabundance (Chitty and Nicholson 1942). Regions 3 and 10 fall largely south of thetree-line, where the small rodent community probably did not indude Lemmus orDicrostonyx (Hall 1981). Region 11, however, is comprised mostly of tundra habitats,well within the range of lemmings (Hall 1981) (Fig. 5.3).Two other regions, 4 and 9, were intermediate between the other groupings inreports of no-change, and reports of increases did not exceed declines (Table 5.1).A number of field studies confirm a contrast between regions of relativeconstancy and inconstancy. First I consider regions of relative inconstancy. Elton(1942) summarised information from many observers in northern Quebec andLabrador (regions 1 and 2) to show that lemmings and their principal predators,arctic fox (Alopex lagopus) and red fox (Vulpes vulpes), follow definite cycles inabundance with a predominantly four-year period. On the Cumberland Peninsula ofsouthern Baffin Island (region 6), Watson (1956) reported a concurrent irruption ofLemmus and Dicrostonyx, and a snap-trapping success as high as that at the peakdensities of irruptions reported by Krebs (1964) and Mallory et a!. (1981). Fuller et al.(1975b), recorded a greater than ten-fold variation in abundance of Dicrostonyx, over asix-year period, at Truelove Lowland on northern Devon Island (region 7). Manniche(1910, reported in Elton 1942) recorded substantial density changes, though notaccurately quantified, in Dicrostonyx in northeastern Greenland, indicative of an205irruption and precipitous decline. On Banks Island, Dicrostonyx and Lemmusoccasionally irrupt to short-lived high densities (Manning and Macpherson 1958,Larter, unpubi. data). The only exception to this pattern is the relatively constantpopulation of between zero and three lemmings per hectare over five summers atIgloolik (Rodgers and Lewis 1986). Although Igloolik is on an island in region 7 (Fig.5.1), this island is close to the mainland, and may be ecologically more similar toregion 9.Of the regions with intermediate levels of stability, region 4 is the best studied.Shelford (1943) and Brooks and Banks (1971) working at Churchill, Krebs (1964) atBaker Lake, and Mallory et a!. (1981) at Eskimo Point (Arviat), all report dramaticdensity fluctuations greater than ten-fold in amplitude. Shelford (1943) reported aperiod of three or four years between cyclic peaks of Dicrostonyx at Churchill in the1930s. Scott (1993) reports a continued periodicity of four years at Churchill in the1950s, and from 1967 to the present. The period of fluctuations at Baker Lake andEskimo Point were not known. In an ongoing study of lemming demography atWalker Bay, Kent Peninsula, in region 9, both lemming species declined precipitouslyfrom a peak in 1993 to very low densities in 1994 (Krebs and Wilson, unpubl. data),Region 11 was anomalous in the Arctic Wild Life Enquiry data, and is alsoanomalous in field studies. At Pearce Point, on the Northwest Territories (N.W.T.)mainland, Dicrostonyx kilanginiutak have persisted at low densities over six years withno evidence of cyclic irruptions, and densities rarely exceeding two animals perhectare (i.e. similar to low densities reported by Shelford (1943) and Krebs (1964))(Krebs et a!. 1995, Chapter 1). There have been no detailed field studies of lemmings206in northern Yukon (part of region 11), but there are no substantial changes betweenyears in the proportion of fox dens used to rear pups (i.e. natal dens)(Smits andSlough 1993). Because the proportion of natal dens occupied is much higher insummers of lemming irruptions (Macpherson 1969, Eberhardt et a!. 1983), Smits andSlough (1993) speculate that lemmings on the Yukon coastal plain do not irruptregularly, if at allAt Point Barrow, Alaska, the northernmost tip of the coastal plain (Fig. 5.1),Lemmus sibiricus undergoes periodic irruptions, with amplitude of at least threeorders of magnitude, and periodicity of three to six years. Dicrostonyx torquatus israrer at this site, but when present, irrupts in synchrony with Lemmus. (Pitelka 1973,Batzli et a!. 1980). At three other sites on the coastal plain, Wainwright, Pitt Point andInaru River (between Barrow and Atkasook) (Fig. 5.1), Pitelka and Batzli (1993) reportirruptive populations of both lemming species.At Atkasook, on the inland coastal tundra (Fig. 5.1), Batzli and Jung (1980)report a microtine community with consistently low densities over four years (1975-78); average numbers per hectare over all habitats range from 0.3 to 1.6 for Lemmus,0.4 to 0.7 for Dicrostonyx, and 0.03 to 0.9 for Microtus. Densities frequently declinedduring the summer, with low adult survival, and density changes within seasonswere generally greater than those between years (Batzli and Jung 1980). Oneinterpretation of these data is a lack of irruptive dynamics. Pitelka and Batzli (1993)applied an index of cyclicity (s) (Henttonen et a!. 1985) to another data set fromAtkasook (1955-59), and found both lemmings to be cyclic. However, their capturerates were low, even in the year of highest abundance, and peak densities were207substantially less than those found on the north coast (Pitelka and Batzli 1993). Itseems that irruptions at Atkasook are irregular and of low amplitude.At two sites in the foothills, Cape Sabine and Umiat, lemming densities variedwithin less than an order of magnitude, and were mostly non-cyclic, according to theindex of cyclicity applied over only four years (Pitelka and Batzli 1993).In summary, patterns of population fluctuation describe two regions withinarctic North America. Lemmings undergo large, and often regular, irruptions on thetundra of the Ungava-Labrador peninsula, on the islands of the arctic archipelago, onthe Alaskan coastal plain, and on the mainland tundra of the Northwest Territories,Canada, east of approximately the Coppermine River. In the latter area, fluctuationsat some sites may be absent or less regular and extreme. When sympatric, bothspecies irrupt synchronously, and Leminus generally reach higher peak densities thanDicrostonyx (Watson 1956, Krebs 1964, Batzli et at. 1980; but see Rodgers and Lewis1986). The second region includes the tundra of the Canadian mainland west of theCoppermine River, through northern Yukon, and includes the foothills of the BrooksRange of Alaska. In this region Dicrostonyx is generally more common than Lemmus,and populations rarely, if ever, irrupt to the high densities typical in the first region(Krebs et at. 1995, Pitelka and Batzli 1993).FACTORS LIMITING POPULATION GROWTH AT LOW DENSITIESNumerous ecological factors, extrinsic and intrinsic, are sufficient to limitpopulation growth in lemmings at low densities (Table 5.2), but factor effects differ inmagnitude and vary in time and space. The impact of summer predation by a diverseTable5.2.Summaryoffactorswhicharesufficienttolimitlemmingpopulationgrowthatlowdensities, andthosethatarenecessarytocurtail irruptivepopulationgrowth.SufficientNecessaryFactorSeasontolimitgrowth?tocurtailirruption?PredationmortalityWinterYes(MacLeanetal.1974)No(MacLeanetal.1974)(Maher1967)(Batzlietal.1980)(Chap1)SummerYes(Pitelkaetal.1955)Yes(Chap1,2&4)(Maher1970)(Batzlietal.1980)(Chap1)PredationriskWinter??SummerYes(Chap2)No(Chap2)FoodavailabilityWinterYes?(Collieretal.1975)No?(Batzlietal.1980)(Krebs1964)SummerNo(Batzlietal.1980)No(Batzlietal.1980)(Chap2)(Chap2)FoodnutrientqualityWinterYes?(BatzliandJung1980)No?SummerNo?(Batzlietal.1980)No?FooddefencesWinterYes?(Batzli1983)No?&SummerTable5.2(continued)SufficientNecessaryFactorSeasontolimitgrowth?tocurtailirruption?ColdtemperatureAutumnYes?(Fulleretal. 1975b)No?(Chap3)(Reynolds&Lavigne1988)(Chap3)ThinsnowdepthWinterYes(Shelford1943)No?(Shelford1943)(MacLeanetal.1974)(Krebs1964)(Fulleretal.1975b)(MacLeanetal.1974)(Chap3)(Chap3)Flooding(rain)SummerYes(Batzlietal.1980)No(Chap1)(Chap1)IcecoatingAutumnYes(Fuller1967)Novegetation(Batzlietal.1980)AgestructureWinterYes?(Boonstra1994)No?(Boonstra1994)&senescence&SummerReproductiveinhibitionWinterNo(Chap2)No(Chap2)&SummerInfanticideSummerYes(Brooks 1993)No(Chap2)(Chap2)AggressivespacingWinterNo(Krebsetal.1995)No(Krebsetal.1995)behaviour&Summer(Chap2)(Chap2)(Brooks 1993)C210community of migratory and resident predators can be large, affecting adult,subadult and juvenile survival, adult reproductive output, and perhaps maturationrate of juveniles (Chapters 1,2 & 4). The impact of predation risk appears smaller,potentially affecting maturation of juveniles and adult reproductive output, and beingan ancillary condition to predation mortality (Chapter 2). Many factors operate onlyfor short and irregular periods of time (e.g., flooding and icing). Although theirlimiting effects might be strong, affecting survival of all age-classes, they arenegligible in many years. Some factors are likely the legacy of a population irruptionand subsequent decline (e.g., winter predation by specialist predators; adult skew inage structure), so are of marginal interest in understanding a persistent low. Somefactors such as food quality and availability or winter snow depth, are best thoughtof as enabling factors; both an abundance of high quality food, and thick widespreadsnow providing access to food, are likely to facilitate population growth.Only one factor, summer predation mortality, is clearly necessary to curtail apopulation irruption (Table 5.2). At Pearce Point, where lemmings do not irrupt,survival of adults and neo-nates, reproductive output of adults, and recruitment ofadults all increase with removal of predators in summer (Chapter 1). Predatorsconsume the great majority of dispersing subadults (Chapter 4). Without thispredatory impact on all age-classes, population growth would proceed.At sites with irruptions, diverse predators are abundant at the beginning of thelow density phase, but have declined dramatically by the pre-irruptive summer,allowing leniming population growth this summer (Shelford 1943, Batzli et al. 1980).Some resident specialists, such as the least weasel (Mustela nivalis), may disappear211locally (Thompson 1955. MacLean et at. 1974). Most migratory avian predators do notbreed, or even stay in the region (Pitelka et at. 1955). Semi-specialists, such as thearctic fox, are also rare and not breeding (Macpherson 1969). Consequently, limitationby predation mortality is minimal prior to an irruption.However, minimal predation mortality is not the only condition necessary forirruptive population growth. An irruption depends on winter population growth,which requires (i) continued good survival of reproductive adults and their offspring,and (ii) sufficient energy and nutrients for survival and repeated reproduction. Thefirst condition is satisfied by the disappearance of most resident predators theprevious summer. The only exception may be least weasels, which can locate theexpanding population in winter, and respond numerically by breeding under thesnow (MacLean et at. 1974, Batzli et at. 1980). The irruption has already occurred bythis time (late winter), though the peak may be truncated (Batzli et a!. 1980, Pitelkaand Batzli 1993).The second condition stems from the fact that mere survival in winter isenergetically very costly (Reynolds and Lavigne 1988), and depends on the insulativevalue of a nest, and of snow cover (MacLean et at. 1974, Fuller et a!. 1975a, Chappell1980, Casey 1981). The energetic costs of pregnancy and lactation are high, especiallygiven the additional costs of rewarming the nest and sucklings following a foragingbout, and the cost of foraging in cold temperatures (Collier et at. 1975, Batzli et at.1980). Therefore survival and reproduction are likely enhanced where absolutequantity of foods is higher, and where snow is deeper (Collier et at. 1975). Deepersnow provides more insulation from ambient temperatures, and, after drifting by212wind, covers larger areas of tundra, providing access to food over a larger area.Irruptive growth is delayed during winters of unusually low snow accumulation(Shelford 1943, Krebs 1964). Deep snow may also protect lemmings from foxpredation (cf. Lindstrom and Hornfeldt 1994). Therefore, substantial accumulation ofsnow is necessary for winter population growth.In summary, three conditions appear necessary for irruptive populationgrowth: minimal predation mortality (especially in summer), adequate foodavailability, and adequate snow depth. Batzli et a!. (1980) came to the sameconclusion regarding irruptive growth in Alaska. In some areas the latter twoconditions could conceivably be satisfied in all years, but in other areas they mayonly be satisfied sporadically, resulting in less regular and slower population growthin the absence of predation mortality.HOW DO REGIONS OF IRRUPTIVE AND NON-IRRUPTIVE DYNAMICSDIFFER?Summer PredatorsIn regions where lemmings irrupt, the summer predator community isremarkably consistent, and dominated by small rodent specialists (Table 5.3). Theseinclude jaegers (Stercorarius spp.), snowy owl (Nyctea scandiaca), short-eared owl (Asioflammeus), rough-legged hawk (Buteo lagopus), and least weasel. Depending onavailability of alternative prey, the arctic fox may rely to varying extent on lemmingsin summer. Only in coastal areas can they come close to compensating for lowlemming abundance with alternative foods, but do so by becoming more mobile(Macpherson 1969, Chesemore 1975, Riewe 1977, Batzli et a!. 1980). Ermine (MustelaTable5.3.Presence(Y)orabsence(N) oflemmingpredatorsassummerbreedersatsiteswherelemmingshavebeenintensivelystudied.SitesareorganizedbyArcticWildLifeEnquiryregions(Chitty1950),andcodedasfollows:(a)Churchill(Shelford1943),(b)Churchill(Brooks 1993),(c)BakerLake(Krebs1964),(d)BaffinI.(Watson1956),(e)TrueloveLowland(Pattie1977, Riewe1977),(f)PearcePoint(Chapters 1&4),(g)Barrow(Batzlieta!.1980),(h)Atkasook(BatzliandSobaski1980, BeeandHall1956, WhiteandCade1971).Anasterisk(*)meansspeciesistransient,andaquestionmark(?)meansdataarelacking. Region46711AlaskaAlaskacoastfoothillsPredatorabcdefghBIRDSPomarineJaegerNNNNNNYNParasiticJaeger??YNY***Long-tailedJaegerNNYNY**YGlaucousGullNNNNYYYNGoldenEagleNNNNNYN*Rough-leggedHawkY?**NYN*PeregrineFalconNNN*YYN*Table5.3.(cont.)Region46711AlaskaAlaskacoastfoothillsabcdefghGyrfalconNNN*NYNNShort-earedOwlYY*NN*yySnowyOwlYN*Y*NY*CommonRaven???YYYNNMAMMALSGrizzlybearNNNNNYN*WolfN*NN*N*Redfox?N?*NNYNYArcticfoxYY*YYNYNErmineYNYYYYYYLeastweaselNNNNNNYYWolverineY?N*NNNNNArcticgroundsquirrelNNYNNYNY215erminea) may partly compensate for low lemming abundance in summer by feedingon birds, but are lemming specialists in winter (Riewe 1977, Batzli et at. 1980). Allthese predators show dramatic numerical responses to lemming abundance, withvarying degrees of delay. Most importantly, these responses result in all predatorsbeing very scarce or non-existent after lemmings have been at low densities (<2/ha)for a year or more (Elton 1942, Shelford 1943, Pitelka et a!. 1955, Watson 1957,Macpherson 1969, Maher 1970, MacLean et at. 1974, Riewe 1977).Where lemmings do not irrupt, the predator community is dominated bygeneralist predators (Table 5.3). At Pearce Point, the arctic fox is replaced by the redfox, which eats primarily lemmings but is able in some years to compensate for ascarcity of lemmings by feeding on other species (Chapter 4). The other generalistpredators include peregrine falcon (Falco peregrinus), gyrfalcon (Falco rusticolus),golden eagle (Aquila chrysaetos), glaucous gull (Larus hyperboreus), grizzly bear (lirsusarctos) and arctic ground squirrel (Spermophitus parryii) (Table 5.3). The system alsoincludes some specialist or semi-specialist species, the rough-legged hawk andermine, in common with irruptive regions. All these species continue to breed atspring lemming densities of 1-2/ha (Chapters 1 and 4). Pitelka and Batzli (1993) didnot clearly list the predator community at sites with non-irruptive lemmings in thefoothills of the Brooks Range. However, this region includes the breeding ranges ofall the generalist predators at Pearce Point, except glaucous gulls (Bee and Hall 1956,White and Cade 1971, Batzli and Sobaski 1980).The different composition and temporal persistence of predator communities atsites without irruptions requires explanation. One critical species shift is from arctic216to red fox. Competition with red fox may define the southern limit of arctic foxbreeding range, at least in western North America (Chesemore 1975). Ranges of redand arctic fox overlap extensively in the southern tundra and northern taiga zones(Fig. 5.2), but much of this overlap reflects movements beyond the breeding ranges,especially by arctic foxes (Wrigley and Hatch 1976, Garrott and Eberhardt 1987). Thetwo foxes breed within 10 km of each other on the northern coastal plain of Yukon(Smits and Slough 1993). This may be exceptional because red fox are dominant overarctic fox in agonistic encounters and in competition for food (Rudzinski et al. 1982,Schamel and Tracy 1986), and may prey on arctic fox (Frafjord et a!. 1989).The distributions of the two species may be in long term flux, and quitevariable in time and space. Red fox expanded their range onto the tundra west ofHudson Bay in the 1930s (Marsh 1938). They colonized southern Baffin Island in theearly 1900s, and gradually expanded their range northward, eventually reachingCornwallis and Devon Islands in the 1960s (Macpherson 1964). In arctic Russia, asimilar northward range expansion of red fox occurred this century, withdisplacement of arctic fox from some breeding sites. This expansion was attributed toa warming climate (Skrobov 1960, Chirkova 1968).The factors limiting the northern range, and the local distribution, of red foxare not clear. Hersteinsson and MacDonald (1982) suggest that red fox are limited byFig. 5.2. Ranges of (a) red fox (Vulpes vulpes), and (b) arctic fox (Alopex lagopus), innorthern North America (from Banfield (1974) and Hall (1981)).217[‘.3 00219their need for a more diverse prey base than that supporting arctic fox. Red fox,being larger, have higher daily food needs (320-470 g/d; Lockie 1959, Ryszkowski etal. 1973, Sargeant 1978) compared with arctic fox (c. 205 g/d; Riewe 1977).Consequently, they will depress rodent populations faster than arctic fox, and aremore likely to need alternative foods in order to survive without becoming nomadic.Red fox could utilize breeding birds (including seabirds, waterfowl, ptarmigan(Lagopus spp.), shorebirds and passerines), medium-sized mammals such as arcticground squirrels and arctic hare (Lepus arcticus), and carrion (e.g., caribou (Rangifertarandus) carcasses) to supplement a summer diet of small rodents. Not all of theseare available year-round, unless foxes cache food for winter. In northern Yukon, redand arctic fox both rely on microtine rodents in summer, but breeding adult andjuvenile red fox consumed a higher proportion of birds, principally waterfowl (Smitset a!. 1989). At Pearce Point, the red fox can fully or at least partially compensate fordeclining summer lemming abundance principally by feeding on arctic groundsquirrels, but also on tundra voles and birds (Chapter 4). Red fox in alpine tundra ofAlaska and Yukon rely heavily on arctic ground squirrels in summer (Jones andTheberge 1983). A red fox female in northwest Alaska killed at least four squirrels aday in mid-summer to feed a family (Carl 1971). Ground squirrels and birds appearsufficient to sustain red fox breeding when microtine rodents do not suffice, but howred fox sustain themselves in winter is unclear (Chapter 4).Arctic ground squirrels are confined to mainland tundra, excluding Ungava,having radiated from Beringia following the Wisconsin glaciation, and not being ableto establish themselves as breeding populations on islands because of their need to220hibernate (Macpherson 1965). Consequently they are not found in two large regionswhere lemmings irrupt: Ungava, and the arctic islands (Fig. 5.3). Red foxes in theseregions may have other alternative prey such as birds, arctic hare, ptarmigan orcaribou, to supplement a diet of lemmings. For example, red foxes raising young by aseabird colony on Digges Island, off the Ungava coast, raised young on a mixed dietof collared lemmings and thick-billed murres (Uria lomvia) (Gaston et a!. 1985). Tosurvive winter in these regions, foxes likely rely more on cached food, or a nomadicscavenging existence (Andriashek et a!. 1985). However, unless red foxes return tobreed at the same sites in consecutive years, their distribution on the eastern arcticislands is unlikely to affect the frequency of lemming irruptions.The diversity of microtine rodent species also tends to be higher incommunities with red foxes and non-irruptive lemmings. Irruptive populations oflemmings occur in regions with one or both lemming species, but rarely any othersmall rodent (Fig. 5.3)(Krebs 1964, Fuller et a!. 1975b, Batzli et a!. 1980). In nonirruptive situations, Lemmus becomes relatively uncommon, apparently because ofhabitat limitations, and other microtines, such as the tundra vole (Microtusoeconomus), the tundra red-backed vole (C!ethrionomys rutilus) and the singing vole(Microtus miurus), appear in the system (Batzli and Jung 1980, Pitelka and Batzli 1993,Krebs et a!. 1995). If foxes select lemmings (in particular Dicrostonyx) first, reducingtheir abundance, then other microtines may act as alternative prey, as do groundsquirrels and tundra voles at Pearce Point (Chapter 4).Arctic foxes appear adapted to situations with less diverse and less predictablefood resources. In summer, the inland pairs supplement a diet of small rodents with221Fig. 5.3. Ranges in arctic North America of two lemming species complexes, (a)Lemmus, and (b) Dicrostonyx , and ranges of four rodent species potentiallysympatric with lemmings, and acting as potential alternate food for red fox: (c)arctic ground squirrel, (d) tundra red-backed vole, (e) tundra vole, and (f)singing vole.223eggs, birds, insects, carrion and berries (Garrott and Eberhardt 1987). Their breedingsuccess is high when small rodents are abundant (Garrott and Eberhardt 1987), butthey are unable to compensate for a scarcity of small rodents with these other foods(Macpherson 1969). They may try to compensate by feeding more on waterfowl eggs,nestlings and perhaps adults (Anthony et al. 1991, Underhill et a!. 1993), but thesefood sources are short-lived. Consequently arctic foxes become nomadic scavengersand hunters in some summers and most winters (Chesemore 1975, Garrott andEberhardt 1987). Arctic ground squirrels are absent, or rare items in arctic fox diets(Chesemore 1968, Macpherson 1969, Garrott and Eberhardt 1987), probably becausethe squirrels are scarce in most breeding areas. However, arctic fox at Prudhoe Bay,Alaska (where squirrels are present), do eat more squirrels when lemmings are scarce(Garrott et a!. 1983). This might partly explain the less extreme lemming fluctuationsat Prudhoe Bay compared with Barrow, where squirrels are rare (Batzli et a!. 1980).The second major distinction between predator communities (Table 5.3) wasthe greater diversity of generalist predators where lemmings do not irrupt. Onceagain the arctic ground squirrel was a critical prey item. Of the generalists at PearcePoint, the golden eagles fed primarily on ground squirrels, the gyrfalcons replacedtheir primary prey, ptarmigan, with ground squirrels (see also Poole and Boag 1988),and the ground squirrels were found in 63% of grizzly bear scats (Chapter 4). Theprimary protein source for tundra bears in Yukon is ground squirrels (Pearson 1975),and grizzly bears exerted significant late summer predation pressure on arctic groundsquirrels in northwest Alaska (Carl 1971). Peregrine falcons were an exception, takingfew ground squirrels. However, ground squirrels themselves were also one of the224generalist lemming predators (Boonstra et a!. 1990, Chapter 1). At this one well-studied site with non-irruptive lemmings, arctic ground squirrels appear to be thecrucial extra prey item allowing persistence in summer of red fox and a suite ofgeneralist lemming predators.Ground squirrel range overlaps regions where lemmings irrupt, but they areabsent or sparsely distributed at mainland lemming study sites with irruptions,including Baker Lake and Walker Bay (Table 5.3). Ground squirrel distribution islimited primarily to well-drained soils providing an adequate substrate for denningand digging hibernacula (Carl 1971). Their distribution in mainland regions istherefore limited by surficial geology, and confined to areas of relatively sandyalluvium, till or eskers (Carl 1971, Batzli and Sobaski 1980). The dispersion ofsquirrels, and consequently their availability to a predator, varies substantially fromarea to area.Snow DepthAnnual mean total snowfall is not lower in regions without irruptionscompared with regions with irruptions (Fig. 5.4). Therefore the lower snowfallsexperienced by the central Canadian mainland coast, western islands, and Alaskancoastal plain, can be sufficient in some winters to allow irruptive growth. However,inadequate snow may preclude irruptive winter breeding, even when predators arescarce and food abundant (Shelford 1943, Krebs 1964, MacLean et a!. 1974). The resultis a longer period between consecutive irruptions, and potentially less regularity inthe period and amplitude of irruptions. Long-term lemming data are lacking from aregion of heavy snowfall, but the regularity of peaks (mean period 4 ± 0.2 (S.E.)225Fig. 5.4. Map of isohyets (dotted lines) of mean total annual snowfall (cm) in arcticNorth America. Data are taken from Maxwell (1980) and Hartman and Johnson(1984).226227years) in fox pelts from Ungava reported by Elton (1942) suggests a regular cycle inthat region of higher snowfall.Food AvailabilityBoth lemming species occupy a range of vegetation communities. Comparisonsof winter food availability between regions are suspect unless vegetationcommunities are very similar, and unless one can quantify live tissue in standingcrops of lemming foods. This has not been done frequently enough to make anysuitable comparisons between regions of differing constancy in lemming numbers.HYPOTHESESI propose a number of alternative hypotheses, each of which might besufficient to explain the paucity or absence of lemming irruptions on the tundra ofthe western Canadian mainland and the north slope of the Brooks Range in Alaska.Each one stems from the strong limiting effect of generalist predators with sufficientalternative prey. Erlinge et al. (1991) also concluded that heavy predation bygeneralist predators best explained the lack of cyclic irruptions in someFennoscandian populations of the field vole (Microtus agrestis).I hypothesize that a lemming population cannot irrupt when:(a) it exists within the hunting range of a breeding population of red foxes, whichuses this range in consecutive summers.(b) it exists within the overlapping hunting ranges of the following generalistlemming predators: grizzly bear, arctic ground squirrel and peregrine falcon.(c) it coexists with a persistent arctic ground squirrel population.These are not mutually exclusive hypotheses. Based on current knowledge,228each one might be sufficient, or some combination of them might be sufficient, toexplain a lack of irruptions. An arctic ground squirrel population may be the solesufficient condition, because the other two conditions are frequently met whenground squirrels are present, and therefore prey for the generalist predators.However, there is probably some threshold density below which ground squirrels donot exist in sufficient numbers or proximity to lemmings to provide sufficient prey tosustain generalist predators, nor kill a substantial proportion of the lemmingsthemselves. This threshold is likely determined by habitat quality, in particular densite availability and dispersion (Carl 1971). For example, arctic ground squirrels existin the Baker Lake region (Table 5.3), but at low densities because habitats are widelydispersed and associated with eskers. They therefore are likely to have little direct orindirect impact on lemmings in this area.Differentiating among hypotheses will require annual estimation of numbersand breeding status of all summer predators in areas where lemming populationnumbers are monitored. Simulation modelling of community dynamics, usingparameters in this thesis, may help determine threshold densities at which groundsquirrels can support red foxes and other generalist predators during summer. Themost insightful field data regarding the hypotheses are likely to come from areaswhere red and arctic fox distributions overlap, but where ground squirrels are absent.Sites could include Herschel Island in northern Yukon, Baffin Island, or a number ofmoderate sized islands in Hudson Strait between Baffin Island and Ungava.229CONCLUSIONThe population dynamics of collared lemmings (Dicrostonyx kilangmiutak) atPearce Point, N.W.T., are clearly different in summer and winter. In most snow-freeperiods, lemming populations decline, or remain fairly stable, despite production ofat least two, and more often three, litters by each female. In periods of snow cover,populations recover with breeding under the snow in mid-winter and spring. Theprincipal pattern is one of annual fluctuations with highest densities most frequentlyoccurring in spring, but never exceeding three adults per hectare.The snow-free summers expose lemmings to heavy predation mortality whichstrongly limits their population growth. Predators drive summer declines or limitgrowth to modest gains. A diverse assemblage of predators preys heavily on all life-stages: adults, subadults and neonates. When predation mortality is experimentallyreduced, survival of all life stages increases, and no other limiting factor compensatesfor reduced predation mortality. Consequently, predation mortality is sufficient andnecessary to limit growth and maintain lemmings at low densities in summer.A search for alternative factors operating at low densities in summer, revealsthat neither food availability nor social inhibition of reproduction appear to limitsummer population growth. Heavy predation mortality however, appears topredispose the population to additional, though weak, limiting factors. These areinfanticide of neonates, probably by strange males, and suppressed individual growthrates of neonates, because of the lactating female’s sensitivity to predation risk. Whensummer predation mortality is intense, it is destabilizing; predators drive lemmings230to low densities at which the predators can no longer sustain themselves.Winter population growth requires the production of successive litters underthe snow. Suitable winter habitat coincides with the distribution of deepest snow,probably because of the thermal and predation cover provided by snow. Variance inrates of population change among seasons of snow cover is well explained in amultiple regression analysis by a combination of the extent of cold stress in autumn(September and October), when lemmings are changing from summer to wintermorphology, and the cold stress in winter, as indexed by the degree of thermal coverprovided by each cm-day of snow per degree-day of frost. Ermine predation explainslittle of the interannual variation in winter population growth. Lemmings may befood limited in winter. Despite winter population growth, densities by spring havenot been high enough, in six years of study, to exceed those which predators canmaintain at low densities in summer.The summer predator community includes specialist (rough-legged hawk, andperhaps glaucous gull), semi-generalist (red fox, ermine), and generalist (grizzly bear,golden eagle, peregrine falcon, gyrfalcon, arctic ground squirrel) predators. Rough-legged hawks and red foxes are the principal lemming predators. At lemmingdensities greater than one per hectare in early summer, hawks settle and breed, andred foxes breed successfully. At very low lemming densities in early summer, onlygeneralist predators continue to breed successfully. However, these generalistscontinue to exert substantial mortality on lemmings even at low densities, andthereby continue to limit population growth when both specialists and some semigeneralists are absent. The persistence in this community of generalists and semi-231generalists appears directly tied to arctic ground squirrels, which are fairly abundantand evenly distributed. The ground squirrels are the primary prey for golden eagles,gyrfalcons and grizzly bears. They are the primary alternative prey for foxes whenlemmings are scarce. They are also lemming predators themselves.Pearce Point is situated in a region of arctic North America where populationsof microtine rodents, in particular collared lemmings, rarely if ever irrupt to densitiesgreater than ten per hectare. This region includes the Canadian mainland, west ofapproximately the Coppermine River, and the northern foothills of the Brooks Rangeof Alaska. It contrasts with other North American tundra regions where populationsof microtine rodents, notably collared and brown lemmings, irrupt periodically andoften cyclically. Ecological communities in these regions differ substantially in thecomposition of their predator conmnmities. Where lemmings irrupt, most predatorsare lemming specialists and frequently disappear from the community in the preirruptive summer. Where lemmings do not irrupt, most predators are generalists orsemi-generalists. Factors limiting the distribution of these generalists are not fullyunderstood, but appear related to the diversity of prey species, notably groundsquirrels, in the community.In conclusion, predation is a dominant factor in the population dynamics ofcollared lemmings, and probably of other arctic microtine rodents. The distributionsand persistence of tundra predators are worthy of further study in attempts to betterunderstand the regional variation in lemming dynamics, and the dynamics of otherprey which share the same predators.232LITERATURE CITEDAdorjan,A.S. and G.B.Kolenosky 1969. A manual for the identification of hairs ofselected Ontario mammals.Research Report (Wildlife) #90. Ontario Ministry ofNatural Resources, Toronto.Andersson,M. and Erlinge,S. 1977. Influence of predation on rodent populations.Oikos 29:591-597.Andriashek,D., Kiliaan,H.P.L. and Taylor,M.K. 1985. Observations on Foxes, Alopexlagopus and Vulpes vulpes, and Wolves, Canis lupus, on the off-shore sea ice ofnorthern Labrador. Canadian Field-Naturalist 99:86-89.Angelstam,P., Lindstrom,E. and Widen,P. 1984. Role of predation in short-termpopulation fluctuations of some birds and mammals in Fennoscandia.Oecologia 62:199-208.Angelstam,P., Lindstrom,E. and Widen,P. 1985. Synchronous short- term populationfluctuations of some birds and mammals in Fennoscandia- occurrence anddistribution. Holarctic Ecology 8:285-298.Anthony,R.M., Flint,P.L. and Sedinger,J.S. 1991. Arctic fox removal improves nestsuccess of black brant. Wildlife Society Bulletin 19:176-184.Balgooyen,T.G. 1971. Pellet regurgitation by captive sparrow hawks (Falco sparverius).Condor 73:382-385.Banfield,A.W.F. 1974. The mammals of Canada. University of Toronto Press, Toronto,Ontario.Batzli,G.O. 1983. Responses of arctic rodent populations to nutritional factors. Oikos40:396-406.Batzli,G.O. 1985. The role of nutrition in population cycles of microtine rodents. ActaZoologici Fennici 173:13-17.Batzli,G.O. and Jung,H-J.G. 1980. Nutritional ecology of microtine rodents: resourceutilization near Atkasook, Alaska. Arctic and Alpine Research 12:483-499.Batzli,G.O. and Sobaski,S.T. 1980. Distribution, abundance and foraging patterns ofground squirrels near Atkasook, Alaska. Arctic and Alpine Research 12:501-510.Batzli,G.O., White,R.G., MacLean,S.F., Pitelka,F.A. and Collier,B.D. 1980. Theherbivore-based trophic system. In An arctic ecosystem: the coastal tundra atBarrow, Alaska. (Eds. J. Brown, P.C. Miller, L.L. Tieszen and F.L. Bunnell). pp.233335-410. US/IBP Synthesis Series No. 12, Dowden, Hutchinsons and Ross, Inc..Stroudsburg, Pennsylvania.Bee,J.W. and Hall,E.R. 1956. Mammals of northern Alaska. University of KansasMuseum of Natural History, Lawrence, Kansas.Berger,P.J., Negus,N.C., Sanders,E.H. and Gardner,P.D. 1981. Chemical triggering ofreproduction in Microtus montanus. Science 214:69-70.Bergman,C.M. and Krebs,C.J. 1993. Diet overlap of collared lemmings and tundravoles at Pearce Point, Northwest Territories. Canadian Journal of Zoology71:1703-1709.Boonstra,R. 1994. Population cycles in microtines: the senescence hypothesis.Evolutionary Ecology 8:196-219.Boonstra,R., Krebs,C.J. and Kanter,M. 1990. Arctic ground squirrel predation oncollared lemmings. Canadian Journal of Zoology 68:757-760.Boonstra,R., Kanter,M. and Krebs,C.J. 1992. A tracking technique to locate smallmammals at low densities. Journal of Mammalogy 73:683-685.Brooks,R.J. 1993. Dynamics of home range in collared lemmings. In The Biology ofLemmings. (Eds. N.C. Stenseth and R.A. Ims). pp.355-386. Linnean SocietySymposium Series No. 15. Academic Press, London.Brooks,R.J. and Banks,E.M. 1971. Radio-tracking study of lemming home range.Communications in Behavioral Biology 6:1-5.Brooks,R.J. and Banks,E.M. 1973. Behavioural biology of the collared lemming(Dicrostonyx groenlandicus): an analysis of acoustic communication. AnimalBehaviour Monographs 6:1-83.Brown,J.H. and Lasiewski,R.C. 1972. Metabolism of weasels: the cost of being longand thin. Ecology 53:939-943.Bryant,J.P., Provenza,R.D., Pastor,J., Reichardt,P.B., Clausen,T.P. and Toit,J.T. 1991.Interactions between woody plants and browsing mammals mediated bysecondary metabolites. Annual Review of Ecology and Systematics 22:431-446.Carl,E.A. 1971. Population control in arctic ground squirrels. Ecology 52:395-413.Casey,T.M. 1981. Nest insulation: energy savings to brown lemmings using a winternest. Oecologia 50:199-204.234Caughley,G. and Krebs,C.J. 1983. Are big mammals simply little mammals writ large?Oecologia 59:7-17.Chappell,M.A. 1980. Thermal energetics and thermoregulatory costs of small arcticmammals. Journal of Mammalogy. 61:278-291.Chesemore,D.L. 1968. Notes on the food habits of arctic foxes in northern Alaska.Canadian Journal of Zoology 46:1127-1130.Chesemore,D.L. 1975. Ecology of the arctic fox (Alopex lagopus) in North America: areview. In The Wild Canids: their systematics, behavioral ecology andevolution. (Ed. M.W. Fox). pp.143-163. Van Nostrand Reinhold Company, NewYork.Chirkova,A.F. 1968. The relationship between arctic fox and red fox in the far north.Problems of the North 11:111-113. Translated from the Russian by the NationalResearch Council of Canada.Chitty,D. 1938. Canadian Arctic Wild Life Enquiry, 1936-37. Journal of AnimalEcology 7:381-394.Chitty,D. 1939. Canadian Arctic Wild Life Enquiry, 1937-38. Journal of AnimalEcology 8:247-260.Chitty,D. 1940. Canadian Arctic Wild Life Enquiry, 1938-39. Journal of AnimalEcology 9:227-242.Chitty,D. 1960. Population processes in the vole and their relevance to general theory.Canadian Journal of Zoology 38:99-113.Chitty,D. 1967. The natural selection of self-regulatory behaviour in animalpopulations. Proceedings of the Ecological Society of Australia 2:51-78.Chitty,H. 1943. Canadian Arctic Wild Life Enquiry, 1941-42. Journal of AnimalEcology 12:163-172.Chitty,H. 1950. Canadian Arctic Wild Life Enquiry, 1943-49: with a summary ofresults since 1933. Journal of Animal Ecology 19:180-193.Chitty,D. and Chitty,H. 1941. Canadian Arctic Wild Life Enquiry, 1939-40. Journal ofAnimal Ecology 10:184-203.Chitty,D. and Elton,C. 1937. Canadian Arctic Wild Life Enquiry, 1935-36. Journal ofAnimal Ecology 6:368-385.Chitty,D. and Nicholson,M. 1942. Canadian Arctic Wild Life Enquiry, 1940-41. Journal235of Animal Ecology 11:270-287.Chitty,H. and Chitty,D. 1945. Canadian Arctic Wild Life Enquiry, 1942-43. Journal ofAnimal Ecology 14:37-45.Christian,J.J. 1950. The adrenal-pituitary system and population cycles in smallmammals. Journal of Mammalogy 31:247-259.Christian,J.J. 1978. Neurobehavioral endocrine regulation of small mammalpopulations. In Populations of small mammals under natural conditions.(Ed. D.P. Snyder). pp.143-159. Pymatuning Laboratory of Ecology,University of Pittsburgh, Pittsburgh.Cody,W.J., Scotter,G.W. and Zoltai,S.C. 1992. Vascular plant flora of the Melville Hillsregion, Northwest Territories. Canadian Field-Naturalist 106:87-99.Collier,B.D., Stenseth,N.C., Barkley,S. and Osborn,R. 1975. A simulation model ofenergy acquisition and utilization by the brown lemming Lemmustrimucronatus at Barrow, Alaska. Oikos 26:276-294.Day,M.G. 1966. Identification of hair and feather remains in the gut and faeces ofstoats and weasels. Journal of Zoology 148:201-217.Desy,E.A. and Batzli,G.O. 1989. Effects of food availability and predation on prairievole demography: a field experiment. Ecology 70:411-421.Duke,G.E., Evanson,O.A. and Jegers,A. 1976. Meal to pellet intervals in 14 species ofraptors. Comparative Biochemistry and Physiology 53A:1-6.Dunmire,W.W. 1955. Sex dimorphism in the pelvis of rodents. Journal of Mammalogy36:356-361.Eberhardt,L.E., Garrott,R.A., and Hanson,W.C. 1983. Den use by arctic foxes innorthern Alaska. Journal of Mammalogy 64:97-102.Elton,C. 1942. Voles, mice and lemmings: problems in population dynamics.Clarendon Press, Oxford.Englund,J. 1970. Some aspects of reproduction and mortality rates in Swedish foxes(Vulpes vulpes) 1961-63 and 1966-69. Viltrevy 8:1-82.Erlinge,S. 1987. Predation and non-cyclicity in a microtine population in southernSweden. Oikos 50:347-352.Erlinge,S., Agrell,J., Nelson,J. and Sandell,M. 1991. Why are some microtinepopulations cyclic while others are not? Acta Theriologica 36:63-71.236Erlinge,S., Goransson,G., Hansson,L., Hogstedt,G., Liberg,O., Nilsson,I.N., Nilsson,T.,Schantz,T. von, and Sylven,M. 1983. Predation as a regulating factor onsmall rodent populations in southern Sweden. Oikos 40:36-52.Erlinge,S., Goransson,G., Hogstedt,G., Jansson,G., Liberg,O., Loman,J., Nilsson,I.N.,Schantz,T. von, Sylven,M. 1984. Can vertebrate predators regulate theirprey ? American Naturalist 123:125-133.Erlinge,S., Goransson,G., Hogstedt,G., Jansson,G., Liberg,O., Loman,J., Nilsson,I.N.,Schantz,T. von, Sylven,M. 1988. More thoughts on vertebrate predatorregulation of prey. American Naturalist 132:148-154.Fitzgerald,B.M. 1977. Weasel predation on a cyclic population of the montane vole(Microtus montanus) in California. Journal of Animal Ecology 46:367-397.Frafjord,K., Becker,D. and Angerbjorn,A. 1989. Interactions between arctic and redfoxes in Scandinavia - predation and aggression. Arctic 42:354-356.Fuller,W.A. 1967. Ecologie hivernale des lemmings et fluctuations de leurspopulations. Terre et Vie 21:97-115.Fuller,W.A., Martell,A.M., Smith,R.F.C. and Speller,S.W. 1975a. High arctic lemmings(Dicrostonyx groenlandicus) I. Natural history observations. CanadianField-Naturalist 89:223-233.Fuller,W.A., Martell,A.M., Smith,R.F.C. and Speller,S.W. 1975b. High-arctic lemmings,Dicrostonyx groenlandicus. II. Demography. Canadian Journal of Zoology53:867-878.Galushin,V.M. 1974. Synchronous fluctuations in populations of some raptors andtheir prey. The Ibis 116:127-134.Gamberg,M. and J.L.Atkinson 1988. Prey hair and bone recovery in ermine scats.Journal of Wildlife Management 552:657-660.Garrott,R.A., and Eberhardt,L.E. 1987. Arctic fox. In Wild Furbearer management andconservation in North America. (Eds. M. Novak, J.A. Baker, M.E.Obbard and B. Malloch). pp.394-407Ontario Ministry of NaturalResources, Toronto, Ontario.Garrott,R.A., Eberhardt,L.E. and Hanson,W.C. 1983. Summer food habits of juvenilearctic foxes in northern Alaska. Journal of Wildlife Management 47:540-545.Gaston,A.J., Cairns,D.K., Elliot,R.D. and Noble,D.G. 1985. A natural history of DiggesSound. Canadian Wildlife Service Report Series No. 46. Ottawa, Ontario.237Goszczynski,J. 1974. Studies on the food of foxes. Acta Theriologica 19:1-18.Hall,E.R. 1981. The mammals of North America. John Wiley and Sons, New York.Hairston,N.G., Smith,F.E. and Slobodkin,L.B. 1960. Community structure, populationcontrol and competition. American Naturalist 94:421-425.Hamilton,K.L. 1980. A technique for estimating barn owl prey biomass. RaptorResearch 14:52-55.Hansen, R.M. 1957. Remarks on the bifid claws of varying lemmings. Journal ofMammalogy 38:127-128.Hanski,I. and Korpimaki,E. in press. Microtine rodent dynamics in northern Europe:parameterized models for the predator- prey interaction. EcologyHanski,I., Hansson,L. and Henttonen,H. 1991. Specialist predators, generalistpredators, and the microtine rodent cyde. Journal of Animal Ecology60:353-367.Hanski,I., Turchin,P., Korpimaki,E. and Henttonen,H. 1993. Population oscillations ofboreal rodents: regulation by mustelid predators leads to chaos. Nature364:232-235.Hansson,L. 1984a. Predation as the factor causing extended low densities in microtinecycles. Oikos 43:255-256.Hansson,L. 1984b. Winter reproduction of small mammals in relation to foodconditions and population dynamics. In Winter ecology of smallmammals (Ed. J.F. Merritt). pp.225-234. Carnegie Museum of NaturalHistory, Special Publication No. 10.Hansson,L. 1987. An interpretation of rodent dynamics as due to trophic interactions.Oikos 50:308-318.Hansson,L. and Henttonen,H. 1985. Gradients in density variations of small rodents:the importance of latitude and snow cover. Oecologia 67: 394-402.Hansson,L. and H.Henttonen 1988. Rodents dynamics as community processes.Trends in Ecology and Evolution 3:195-200.Hartman,C.W. and Johnson,P.R. 1984. Environmental atlas of Alaska. Institute ofWater Resources, University of Alaska, Fairbanks.Hasler,J.F. and Banks,E.M. 1975. Reproductive performance and growth in captivecollared lemmings (Dicrostonyx groenlandicus). Canadian Journal of238Zoology 53: 777-787.Hasler,J.F., Buhl,A.E. and Banks,E.M. 1976. The influence of photoperiod on growthand sexual function in male and female collared lemmings (Dicrostonyxgroenlandicus). Journal of Reproduction and Fertility 46:323-329,Haukioja,E. 1980. On the role of plant defences in the fluctuation of herbivorepopulations. Oikos 35:201-213.Hausfater,S. and Hrdy,S.B. 1984. Infanticide: comparative and evolutionaryperspectives. Aldine Publishing, New York.Henttonen, H. 1985. Predation causing extended low densities in microtine cycles:further evidence from shrew dynamics. Oikos 44:156-157.Henttonen,H., McGuire,A.D. and Hansson,L. 1985. Comparisons of amplitudes andfrequencies (spectral analyses) of density variations in long-term datasets of Clethrionomys species. Annales Zoologici Fennici 22:221-227.Henttonen,H., Oksanen,T., Jortikka,A. and Haukisalmi,V. 1987. How much doweasels shape microtine cycles in the northern Fennoscandian taiga ?Oikos 50:353-365.Hersteinsson,P. and MacDonald,D.W. 1982. Some comparisons between red and arcticfoxes, Vu!pes vulpes and Alopex lagopus, as revealed by radio tracking.Symposium of the Zoological Society of London 49:259-289.Hornfeldt,B. 1994. Delayed density-dependence as a determinant of vole cycles.Ecology 75:791-806.Hrdy,S.B. 1979. Infanticide among animals: a review, classification, andexamination of the implications for the reproductive strategiesof females. Ethology and Sociobiology 1:13-40.Hunter,R.E., J.A.Crawford and R.E.Ambrose 1988. Prey selection by peregrine falconsduring the nestling stage. Journal of Wildlife Management 52:730-736.Ims,R.A., and Steen,H. 1990. Geographical synchrony in microtine population cycles:a theoretical evaluation of the role of nomadic avian predators. Oikos57:381-387.Janes,S.W. 1985. Habitat selection in raptorial birds. In Habitat selection in birds. (Ed.M.L. Cody). pp.159-188. Academic Press, London.Jarrell,G.H. and Fredga,K. 1993. How many kinds of lemmings ? a taxonomicoverview. In The Biology of Lemmings. (Eds. N.C. Stenseth and R.A.239Ims). Linnean Society Symposium Series No. 15. Academic Press,London.Jones,D.M. and Theberge,J.B. 1982. Summer home range and habitat utilisation of thered fox (Vulpes vulpes) in a tundra habitat, northwest British Columbia.Canadian Journal of Zoology 60:807-812.Jones,D.M. and Theberge,J.B. 1983. Variations in red fox, Vulpes vulpes, summer dietsin northwestern British Columbia, and southwestern Yukon. CanadianField-Naturalist 97:311-314.Jung,H-J.G. and Batzli,G.O. 1981. Nutritional ecology of the microtine rodents: effectof plant extracts on the growth of arctic microtines. Journal ofMammalogy 62:286-292.Kaikusalo,A. and Tast,J. 1984. Winter breeding of microtine rodents at Kilpisjarvi,Finnish Lapland. In Winter ecology of small mammals. (Ed. J.F. Merritt).pp.243-252. Carnegie Museum of Natural History, Special PublicationNo. 10.Keith,L.B. 1974. Some features of population dynamics in mammals. InternationalCongress Game Biology 11:17-58.Kennedy,A.J. and Carbyn,L.N. 1981. Identification of wolf prey using hair and featherremains, with special reference to western Canadian National Parks.Canadian Wildlife Service, Edmonton, Alberta.Kind,R.J. 1981. Snow drifting. In Handbook of snow: principles, processes,management and use. (Ed. D.M. Gray and D.H. Male). pp.338-359.Pergamon Press, Toronto, Canada.Korpimaki,E. 1993. Regulation of multiannual vole cycles by density-dependent avianand mammalian predation? Oikos 66:359-363.Korpimaki,E. and Norrdahl,K. 1989. Predation of Tengmalm’s owls: numericalresponses, functional responses and dampening impact on populationfluctuations of microtines. Oikos 54:154-164.Korpimaki,E. and Norrdahl,K. 1991.a. Numerical and functional responses of Kestrels,Short-eared owls and Long-eared owls to vole densities. Ecology 72:814-826.Korpimaki,E. and Norrdahl,K. 1991h. Do breeding nomadic avian predators dampenpopulation fluctuations of small mammals? Oikos 62:195-208.Korpimaki,E., Norrdahl,K. and Rinta-Jaskari,T. 1991. Responses of stoats and least240weasels to fluctuating food abundances: is the low phase of the volecycle due to mustelid predation ? Oecologia 88:552-561.Krebs,C.J. 1964. The lemming cycle at Baker Lake, Northwest Territories, during 1959-62. Arctic Institute of North America Technical Paper No. 15.. Calgary,Alberta.Krebs,C.J. 1988. The experimental approach to rodent population dynamics. Oikos52:143-149.Krebs,C.J. 1989. Ecological methodology. Harper and Row, New York.Krebs,C.J. 1995. Two paradigms of population regulation. Wildlife Reviews (in press).Krebs,C.J. and Myers,J.H. 1974. Population cycles in small mammals. Advances inEcological Research 8:267-399.Krebs,C.J., Boonstra,R. and Kenney,A.J. 1995. Population dynamics of the collaredlemming and tundra vole at Pearce Point, N.W.T., Canada. Oecologia (inpress).Lidicker,W.Z. 1988. Solving the enigma of microtine cycles”. Journal of Mammalogy69:225-235.Lima,S.L. and Dill,L.M. 1990. Behavioural decisions made under the risk of predation:a review and prospectus. Canadian Journal of Zoology 68:619-640.Lindstrom,E.R. and Hornfeldt,B. 1994. Vole cycles, snow depth and fox predation.Oikos 70:156-160.Lindstrom,E., Angelstam,P., Widen,P. and Andren,H. 1987. Do predators synchronizevole and grouse fluctuations ? - an experiment. Oikos 48:121-124.Lockie,J.D. 1959. The estimation of the food of foxes. Journal of Wildlife Management23:224-227.MacLean,S.F., Fitzgerald,B.M. and Pitelka,F.A. 1974. Predation cycles in arcticlemmings: winter reproduction and predation by weasels. Arctic and AlpineResearch 6:1-12.Macpherson,A.H. 1964. A northward range extension of the red fox in the easternCanadian arctic. Journal of Mammalogy 45:138-140.Macpherson,A.H. 1965. The origin of diversity in mammals on the Canadian arctictundra. Systematic Zoology 14:153-173.241Macpherson,A.H. 1969. The dynamics of Canadian arctic fox populations. CanadianWildlife Service, Report Series No. 8, Ottawa.Madison,D.M. 1985. Activity rhythm and spacing. In Biology of New World Microtus.(Ed. R.H. Tamarin). pp.373-419. American Society of Mammalogists, SpecialPublication No. 8.Maher,W.J. 1967. Predation by weasels on a winter population of lemmings, BanksIsland, Northwest Territories. Canadian Field-Naturalist 81:248-250.Maher,W.J. 1970. The pomarine jaeger as a brown lemming predator in northernAlaska. The Wilson Bulletin 82:130-157.Malcolm,J.R., and Brooks,R.J. 1985. Influence of photoperiod and photoperiod reversalon growth, mortality, and indicators of age of Dicrostonyx groenlandicus.Canadian Journal of Zoology 63:1497-1509.Malcolm,J.R. and Brooks,R.J. 1993. The adaptive value of photoperiod-induced shapechanges in the collared lemming. In The biology of lemmings. (Eds. N.C.Stenseth and R.A. Ims). pp.311-328. Linnean Society Symposium Series No.15.Academic Press, London.Male,D.H. and Gray,D.M. 1981. Snow cover, ablation and runoff. In Handbook ofsnow: principles, processes, management and use. (Eds. D.M. Gray and D.H.Male). pp.360-436. Pergamon Press, Toronto, Ontario.Mallory,F.F. and Brooks,R.J. 1978. Infanticide and other reproductive strategies in thecollared lemming, Dicrostonyx groenlandicus. Nature 273:144-146.Mallory,F,F. and Brooks,R.J. 1980. Infanticide and pregnancy failure: reproductivestrategies in the female collared lemming (Dicrostonyx groenlandicus). Biology ofReproduction 22:192-196.Mallory,F.F., Elliott,J.R. and Brooks,R.J. 1981. Changes in body size in fluctuatingpopulations of the collared lemming: age and photoperiod influences.Canadian Journal of Zoology 59:174-182.Mallory,F.F., Brooks,R.J. and Elliott,J.R. 1986. Variations of skull-body regressions ofthe lemming (Dicsrostonyx groenlandicus) under laboratory and field conditions.Zoological Journal of the Linnean Society 87:125-138.Manniche,A.L.V. 1910. The terrestrial mammals and birds of north-east Greenland.Meddelelser om Gronland 45(1):1-199.Manning,T.H. and Macpherson,A.H. 1958. The mammals of Banks Island. ArcticInstitute of North America, Technical Paper No. 2. Calgary, Alberta.242Marsh,D.B. 1938. The influx of the red fox and its colour phases into the barren lands.Canadian Field-Naturalist 52:60-61.Marti,C.D. 1987. Raptor food habits studies. In Raptor Management TechniquesManual. (Eds. B.A.G. Pendleton, B.A. Millsap, K.W. Cline and D.M. Bird).pp.67-80. National Wildlife Federation, Washington, D.C..Maxwell,J.B. 1980. The climate of the Canadian arctic islands and adjacent waters.Environment Canada, Atmospheric Environment Service. Canadian PublishingCentre, Hull.May,R.M. 1976. Models for single populations. In Theoretical ecology. (Ed. R.M.May). pp.4-25. Blackwell Scientific Press, Oxford.McNamara,J.M. and Houston,A.I. 1987. Starvation and predation as factors limitingpopulation size. Ecology 68:1515-1519.McShea,W.J. and Madison,D.M. 1992. Alternative approaches to the study of smallmammal dispersal: insights from radiotelemetry. In Animal Dispersal: smallmammals as a model. (Eds. N.C. Stenseth and W.Z. Lidicker). pp.319-332.Chapman and Hall, London.Muc,M. 1977. Ecology and primary production of sedge-moss meadow communities,Truelove Lowland. In Truelove Lowland, Devon Island, Canada: a high arcticecosystem.(Ed. L.C. Bliss). pp.157-184. The University of Alberta Press,Edmonton, Alberta.Murdoch,W.W. 1994. Population regulation in theory and practice. Ecology 75:271-287.Oksanen,L. 1990. Exploitation ecosystems in seasonal environments. Oikos 57:14-24.Oksanen,L. and Oksanen,T. 1992. Long-term microtine dynamics in northFennscandian tundra: the vole cycle and lemming chaos. Ecography 15:226-236.Oksanen,L. Fretwell,S.D., Arruda,J. and Niemela,P. 1981. Exploitation ecosystems ingradients of primary productivity. American Naturalist 118:240-261.Ostfeld,R.S. 1992. Effects of habitat patchiness on population dynamics: a modellingapproach. Wildlife 2001: Populations. (Eds. D.R. McCullough and R.H.Barrett).pp.851-863. Elsevier Applied Science, London.Pasanen,S. and Sulkava,S. 1971. On the nutritional biology of the rough-leggedbuzzard, Buteo lagopus Brunn, in Finnish lapland. Aquilo Series Zoology12:53-63.243Pattie,D.L. 1977. Population levels and bioenergetics of arctic birds on TrueloveLowland. In Truelove Lowland, Devon Island, Canada: a high arctic ecosystem.(Ed. L.C.Bliss) pp.413-436. The University of Alberta Press, Alberta.Pearson,A.M. 1975. The northern interior grizzly bear lJrsus arctos L.. CanadianWildlife Service, Report Series No. 34. Ottawa, Ontario.Pearson,O.P. 1966. The prey of carnivores during one cycle of mouse abundance.Journal of Animal Ecology 35:217-233.Pearson,O.P. 1971. Additional measurements on the impact of carnivores onCalifornia voles (Microtus californicus). Journal of Mammalogy 52:41-49.Pitelka,F.A. 1973. Cyclic pattern in lemming populations near Barrow, Alaska. InAlaskan arctic tundra (Ed. M.E. Britton) pp.199-215. Arctic Institute of NorthAmerica, Technical Paper No. 25. Calgary, Alberta.Pitelka,RA. and Batzli,G.O. 1993. Distribution, abundance and habitat use bylemmings on the north slope of Alaska. In The biology of lemmings (Eds. N.C.Stenseth and R.A. Ims). pp. 213-236. Linnean Society Symposium Series No. 15.Academic Press, London.Pitelka,F.A., Tomich,P.Q., and Treichel,G.W. 1955. Ecological relations of jaegers andowls as lemming predators near Barrow, Alaska. Ecological Monographs 25:85-117.Pollock,K.H., Winterstein,S.R., Bunck,C.M. and Curtis,P.D. 1989. Survival analysis intelemetry studies: the staggered entry design. Journal of Wildlife Management53:7-15.Poole,K.G. and Boag.D.A. 1988. Ecology of gyrfalcons, Falco rusticolus, in the centralCanadian arctic: diet and feeding behaviour. Canadian Journal of Zoology66:334-344.Poole,K.G. and Bromley,R.G. 1988. Interrelationships within a raptor guild in thecentral Canadian Arctic. Canadian Journal of Zoology 66:2275-2282,Popanov,E.R. 1988. Diet of the rough-legged buzzard (Buteo lagopus Pontopp.) in theKolmyo tundra: selective depredation by raptors. Soviet Journal of Ecology17:46-48.Porsild,A.E. and Cody,W.J. 1980. Vascular plants of the continental NorthwestTerritories, Canada. National Museums of Canada, Ottawa. 667 pp.Reynolds,J.C. and N.J.Aebischer 1991. Comparison and quantification of carnivorediet by faecal analysis:a critique, with recommendations, based on a study of244the fox, Vulpes vulpes. Mammal Reviews 21:97-122.Reynolds,P.S. and Lavigne,D.M. 1988. Photoperiodic effects on body size andenergetics of the collared lemming, Dicrostonyx groenlandicus. Canadian Journalof Zoology 66:835-841.Reynolds,P.S. and Lavigne,D.M. 1989. Photoperiodic effects on post- weaning growthand food consumption in the collared lemming Dicrostonyx groenlandicus.Journal of Zoology 218:109-121.Ricker, W.E. 1984. Computation and uses of central trend lines. Canadian Journal ofZoology 62:1897-1905.Riewe,R.R. 1977. Mammalian carnivores utilizing Truelove Lowland. In TrueloveLowland, Devon Island, Canada: a high arctic ecosystem. (Ed. L.C.Bliss) pp.493-501. University of Alberta Press, Edmonton, Alberta.Rosenzweig,M.L. and MacArthur,R.H. 1963. Graphical representation and stabilityconditions of predator-prey interactions. American Naturalist 97:209-223.Rodgers, A.R. and Lewis,M.C. 1986. Diet selection in arctic lemmings (Lemmussibiricus and Dicrostonyx groenlandicus): demography, home range and habitatuse. Canadian Journal of Zoology 64:2717-2727.Rudzinski,D.R., Graves,H.B., Sargeant,A.B. and Storm,G.L. 1982. Behavioralinteractions of penned red and arctic foxes. Journal of Wildlife Management46:877-884.Ryszkowski,L., Goszczynski,J. and Truszkowski,J. 1973. Trophic relationships of thecommon vole in cultivated fields. Acta Theriologica 18:125-165.Sargeant,A.B. 1978. Red fox prey demands and implications to prairie duckproduction. Journal of Wildlife Management 42:520-527.Schamel,D. and Tracy,D.M. 1986. Encounters between arctic foxes, Alopex lagopus, andred foxes, Vulpes vulpes. Canadian Field-Naturalist 100:562-563.Scholander,P.F., Hock,R., Walters,V. and Irving,L. 1950. Adaptation to cold in arcticand tropical mammals and birds in relation to body temperature, insulationand basal metabolic rate. Biological Bulletin 99:259-271.Schultz,A.M. 1964. The nutrient recovery hypothesis for arctic microtines cycles:ecosystem variables in relation to arctic microtine cycles. In Grazing interrestrial and marine environments. (Ed. D.J. Crisp). pp.57-68. BlackwellScientific Publications, Oxford.245Scott,P.A, 1993. Relationship between the onset of winter and collared lemmingabundance at Churchill, Manitoba, Canada: 1932-90. Arctic 46:293-296.Scott,T.G. 1941. Methods and computation in fecal analysis with reference to the redfox. Iowa State College Journal of Science 15:279-285.Seldal,T., Andersen,K.J., and Hogstedt,G. 1994. Grazing-induced proteinase inhibitors:a possible cause for lemming population cycles. Oikos 70:3-11.Shelford,V.E. 1943. The abundance of the collared lemming (Dicrostonyx groenlandicus(Tr.) var. richardsoni Mer.) in the Churchill area, 1929 to 1940. Ecology 24:472-484.Simmons,R.E., Avery,D.M. and Avery,G. 1991. Biases in diets determined from pelletsand remains: correction factors for a mammal and bird-eating raptor. Journalof Raptor Research 25:63-67.Sinclair,A.R.E. 1989. Population regulation in animals. In Ecological concepts: thecontribution of ecology to an understanding of the natural world. (Ed.J.M.Cherrett). pp.197-242. Blackwell Scientific Publications, Oxford.Sinclair,A.R.E., Olsen,P.D. and Redhead,T.D. 1990. Can predators regulate smallmammal populations ? Evidence from house mouse outbreaks in Australia.Oikos 59:382-392.Skrobov,V.D. 1960. On the interrelations of the polar fox and fox in the tundra of theNenets National Region. Zoologicheskii Zhurnal 39:469-471.Smith,C.J. 1987. Parental roles and nestling foods in the rough- legged hgawk, Buteolagopus. Canadian Field-Naturalist 101:101-103.Smits,C.M.M. and Slough,B.G. 1993. Abundance and summer occupancy of arctic fox,Alopex lagopus, and red fox, Vulpes vulpes, dens in the northern YukonTerritory, 1984-1990. Canadian Field-Naturalist 107:13-18.Smits,C.M.M., Slough,B.G. and Yasui,C.A. 1989. Summer food habits of sympatricarctic foxes, Alopex lagopus, and red foxes, Vulpes vulpes, in the northern YukonTerritory. Canadian Field-Naturalist 103:363-367.Sokal,R.R. and Rohlf,F.J. 1981. Biometry. W.H.Freeman, New York.Springer,A.M. 1975. Observations on the summer diet of rough- legged hawks fromAlaska. Condor 77:338-339.Stenseth,N.C. 1986. On the interaction between stabilizing social factors anddestabilizing trophic factors in small rodent populations. Theoretical246Population Biology 29:365-384.Stenseth,N.C. and Ims,R.A. 1993. Population dynamics of lemmings: temporal andspatial variation - an introduction. In The Biology of Lemmings. (Eds. N.C.Stenseth and R.A. Ims). pp.61-96. Linnean Society Symposium Series No. 15.Academic Press, London.Svoboda,J. 1977. Ecology and primary production of raised beach communities,Truelove lowland. In Truelove Lowland, Devon Island, Canada: a high arcticecosystem. (Ed. L.C. Bliss). pp.185-216. The University of Alberta Press,Edmonton, Alberta.Taitt,M.J. and Krebs,C.J. 1983. Predation, cover, and food manipulations during aspring decline of Microtus townsendii. Journal of Animal Ecology 52:837-848.Taitt,M.J. and Krebs,C.J. 1985. Population dynamics and cycles. In Biology of the NewWorld Microtus. (Ed. R.H.Tamarin). pp.567-620. American Society ofMammalogists Special Publication No. 8.Thompson,D.Q. 1955. The role of food and cover in population fluctuations of thebrown lemming at Point Barrow, Alaska. Transactions of the Twentieth NorthAmerican Wildlife Conference. 20:166-176.Underhill,L.G., Prys-Jones,R.P., Syroechkovski,E.E., Groen,N.M., Karpov,V.,Lappo,H.G., Van Roomen,M.W.J., Rybkin,A., Schekkerman,H., Spiekman,H.and Summers,R.W. 1993. Breeding of waders (Charadrii) and Brent GeeseBranta bernicla at Pronchishcheva Lake, northeastern Taimyr, Russia, in a peakand a decreasing lemming year. Ibis 135:277-292.Voigt,D.R. 1987. Red fox. In Wild furbearer management and conservation in NorthAmerica. (Eds. M. Novak, J.A. Baker, M.E. Obbard, and B. Malloch). pp.37S-393. Ontario Ministry of Natural Resources, Toronto, Ontario.Watson,A. 1956. Ecological notes on the lemmings Lemnius trimucronatus andDicrostonyx groenlandicus in Baffin Island. Journal of Animal Ecology 25:289-302.Watson,A. 1957. The behaviour, breeding and food-ecology of the snowy owl Nycteascandiaca. Ibis 99:419-462.Wein,R.W., and Rencz,A.N. 1976. Plant cover and standing crop sampling proceduresfor the Canadian high arctic. Arctic and Alpine Research 8:139-150.White,C.M. and Cade,T.J. 1971. Cliff-nesting raptors and ravens along the ColvilleRiver in arctic Alaska. Living Bird 10:107-150.247Wrigley,R.E. and Hatch,D.R.M. 1976. Arctic fox migrations in Manitoba. Arctic 29:147-158.Ydenberg, R. 1987. Nomadic predators and geographical synchrony in microtinepopulation cycles. Oikos 50:270-272.Ylonen,H. 1994. Vole cycles and antipredatory behaviour. Trends in Ecology andEvolution 9:426-430.


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