88 NEWZEALANDJOURNALOFECOLOGY,VOL.31,NO.1,2007

New Zealand Journal of Ecology(2007)31(1): 88-97©NewZealandEcologicalSociety

88 Availableon-lineat:http://www.newzealandecology.org/nzje/

Selectionofalpinegrasslandsoverbeechforestbystoats(Mustela erminea)inmontanesouthernNewZealand

DesH.V.Smith1,DeborahJ.Wilson2*,HenrikMoller3,ElaineC.Murphy1,YolandavanHeezik31DepartmentofConservation,Research,DevelopmentandImprovementDivision,POBox13049,Christchurch,NewZealand2LandcareResearch,PrivateBag1930,Dunedin,NewZealand3ZoologyDepartment,UniversityofOtago,POBox56,Dunedin,NewZealand*Authorforcorrespondence(E-mail:wilsond@landcareresearch.co.nz)

Publishedon-line:11May2007

___________________________________________________________________________________________________________________________________

Abstract:PredationbyintroducedstoatsisnowconsideredamajorthreattothepopulationviabilityofseveralNewZealandendemicbirdspecies.Historicallystoatresearchandmanagementhasfocusedonbeechforestsandlittleisknownabouttheecologyofstoatsinthealpinegrasslandsoccurringabovethenaturalaltitudinallimitof beech forest. Several stoat control operations in beech forest valley floors in southern New Zealand assume thatadjacentmontaneareasactasabarriertostoatimmigration.Stoatswerelive-trappedandradio-trackedinalpinegrasslandsabovetheBorlandBurn,FiordlandNationalPark,duringthesummerandautumnof2003and2004.Seventeenstoatswereradio-collaredandhomerangeswereestimatedfor11ofthem.Thesehomeranges were used in a compositional analysis which showed that these stoats spent significantly more time in alpinegrasslandthaninadjacentbeechforest.Rangecorescalculatedforsixofthesestoatswerelocatedhighupinalpinegrasslandandcontainedverylittlebeechforest.Thismeansthatmontaneareasthatcontainalpinegrasslandsareunlikelytobebarrierstostoatimmigration;rathertheymaybeasourceofdispersingstoatsthatreinvadecontrolareas.Also,endemicanimalspeciesthatinhabitalpinegrasslandscouldbeatriskfromstoatpredation.___________________________________________________________________________________________________________________________________

Keywords:homerange;Chionochloa;habitatuse

IntroductionStoats(Mustelaerminea)areintroducedmammaliancarnivores(King&Murphy2005)heavilyimplicatedinthedeclineofNewZealand’snativefauna(Elliott1996;Elliottetal.1996;McLennanetal.1996;Wilsonetal.1998).Otherintroducedcarnivoressuchasferalcats(Felis catus),ferrets(M. furo)weasels(M. nivalis vulgaris),hedgehogs(Erinaceus europaeus)andrats(Rattus spp.)alsopreyonNewZealand’snativebirds,reptilesandinsects,butthestoatisthemostwidespreadspeciesofintroducedpredatorinNewZealand’sbeechforests (Nothofagus spp.) (King & McMillan 1982;Murphy & Dowding 1995;Alterio & Moller 1997;Milleretal.2001).

MostofthenationalparksinNewZealand’sSouthIsland are montane. In these protected areas alpinegrasslandsdominatedbysnowtussock(Chionochloaspp.)occurabovethealtitudinallimitsofthesouthernbeeches. Fiordland National Park, New Zealand’s

largest(1272311ha),has17%(221186ha)ofitsareaabovetreeline,andmuchofthisisalpinegrassland.

StoatshavebeenshowntoinhabitalpineareasintheNorthernHemisphere(Fitzgerald1977;Martinolietal.2001),butNewZealand’salpinegrasslandsdifferinthattheydonotsupportthediversesmallmammalcommunities that stoats are known to prey upon intheirnativerange.

LittleisknownabouttheecologyofstoatsinNewZealand’salpinegrasslands.Somestudieshaveassumedthattheruggednatureofsuchareaswillprohibitstoatdispersalbetweenadjacentbeechforestvalleys(Lavers&Mills1978;Dilksetal.2003).Apreviousandmorelimitedradio-trackingstudyintheBorlandandGrebevalleysconsistentlylocatedstoatactivitynearvalleyfloors (Purdey et al. 2004). Cuthbert and Sommer(2002)radio-taggedstoatsinalpinegrasslandsintheKowhaiValley,SeawardKaikouraMountains,NewZealand. Their research was specifically related to the managementofHutton’sshearwaters(Puffinus huttoni),

89SMITHETAL.:SELECTIONOFALPINEGRASSLANDS

which were the main prey of stoats there, identified in 99.6% (n = 788) of scats examined (Cuthbert etal.2000).BecauseHutton’sshearwatersdonot liveanywhereelse,thissituationmaynotbetypicalofstoatecologyinalpinegrasslandsgenerally.

ResearchintheMurchisonMountains(FiordlandNational Park) investigated whether stoats movedbetween alpine grasslands and beech forest valleyfloors, but no such movements were observed between thetwohabitats(Smith2002;Smith&Jamieson2005).However, thestudyranforonly3months,andwasabletocollecthomerangedataforonlytwostoatsinalpinegrassland.WiththeexceptionoftheMurchisonMountains,locationoftheTakaheProgramme(Lee&Jamieson 2001; Smith et al. 2005), stoat control at five mainlandsites inFiordlanddoesnot reachaltitudeshigherthan760ma.s.l.,relyingontheassumptionthatimmigrationacrosshigh-altitudeareasisminimal.

The results reported in Cuthbert & Sommer(2002) and Smith & Jamieson (2005) suggest thatsomestoatsmayberesidentathighaltitudes(atleastduringthesummerandautumn),andmayincorporatealpinegrasslandswithintheirhomerange.Theuseandselectionofalpinegrasslandbystoatscouldmeanthatendemicanimalsthatinhabitalpinegrasslandmightbeatriskfrompredation,andthatalpinegrasslandscouldbeanimportanthabitatthatistypicallyoutsidethefocusofmanagement.Further,thisbehaviourwouldpresumablymeanthatalpineareaswouldnotactasabarriertostoatdispersal.Thisresearchthereforeaimedtoanswerthefollowingquestions:dostoatsinhabitinghighaltitudesincorporatealpinegrasslandswithintheirhome range and, if so, do these stoats select alpinegrasslandsoverbeechforest?

MethodsStudy siteTheBorlandBurn(45°40’S,167°20’E)is63kmsouthofTeAnau,NewZealand,inthesouth-easternpartofFiordlandNationalPark.Theareaconsistsofruggedmountainsrisingfromsealevelto>1600mseparatedby U-shaped (glacially formed) valleys. The mostdistinctive ecotone is the beech forest timberline atbetween900and1000ma.s.l., abovewhichalpinegrasslands dominate. In some places the ecotonalboundarybetweenalpinegrasslandsandbeechforestconsistsofanarrowbandofsubalpinescrub.SubalpinescrubismadeupofwoodyshrubsandintheBorlandBurn these are typically Dracophyllum longifolium,D. uniflorum,Hebe odora,Halocarpus bidwilliiandOlearia colensoivar. argentea.Ashingleroadbuiltforpower linemaintenancecrosses theBorlandSaddleat990ma.s.l.,providingtheeasiestaccesstoalpine

grasslandsinthenationalpark.DuringthisstudytherewasnoheavybeechseedfallinFiordland.

Live trappingLivetrappingandradiotrackingwereundertakenovertwo 3-month field seasons: the first between 6 January and23March2003,thesecondbetween6Januaryand19 April 2004. Because the first field season served partly as a pilot study, some modifications to methods were made in the second field season.

Two lines of live traps were set up in alpinegrasslandsalongridgesoneithersideoftheBorlandSaddle. The first, Study Site 1 (S1), headed away from thesaddleinasouth-easterlydirection,whilethesecond,StudySite2(S2),headedawaytothenorth-west.Thenearestpointsonthetwolineswere2kmapartandtrapswithinlineswere200mapart.In2003,S1was4kmlongandcontained20trapsandS2was2kmlongandcontainedninetraps.In2004S2wasextendedto4kminlength(20traps).Twotypesoftrapswereusedonboth lines: aluminum Elliott B traps (Elliott Scientific EquipmentTM)withanadaptedwoodennestbox,andwoodenEdgartraps(King&Edgar1977).Eachlinecontained 15 Elliot B traps and five Edgar traps (trap numbers1,5,10,15and20).

Trapswerebaitedwithaheneggandachunkofrabbitmeat.In2003samplingcommencedwhenthetrapswerelaidout(6January).However,priortothe2004 field season traps were laid out and pre-baited inNovember2003,almost2monthsbeforesamplingcommenced.The purpose of the pre-baiting was toaccustom stoats to having traps within their homeranges,inordertoincreasetheirprobabilityofcaptureandhencethesamplesizeofradio-trackedstoats.

Traplineswererunfor5nightsatthestartofeachfield season and for 5 nights at the end of each field season, but the final 2004 trapping session was disrupted bysnowstorms.Betweenthese5-nightsessions,traplineswererunfortwonightsineachfortnightinordertocaptureanynewanimalsthatmighthavemovedintothearea,whilestillallowingtimeforradiotracking.

Allcapturedstoatswereaged,sexedandweighed,ear-tagged in both ears, and fitted with a radio-collar. Ageclasswasassessedbythepresenceofenlargedtestesonadultmalesandthepresenceofvisiblenipplesonadultfemales(Grue&King1984).

Radio trackingThreetypesofradio-collarswereused.In2003theywerealltwo-stagetransmittersweighingapproximately10g,witha3-mm-widebrasscollarthatformedtheaerial (Sirtrack, Havelock North, New Zealand). In2004,Sirtrack transmitterswith a6-mm-widebrasscollar, modified to weigh 12 g, as well as two-stage cable-tiecollarswithanexternalwhipaerial,weighing5g(Biotrack,Dorset,UK),wereused.

90 NEWZEALANDJOURNALOFECOLOGY,VOL.31,NO.1,2007

The principal radio-tracking technique wastriangulation, undertaken using hand-held (TR4)receivers with Yagi aerials and mirror sightingcompasses with the declination already adjusted togrid north. Several permanent triangulation stationswere identified, including all trap sites and several other stations that provided right angles to the traplinesorcoverageofnewareas.Theestablishmentoftriangulationstationswashenceaniterativeprocessdeveloped throughout the season, dependent on themovementofstoats.

Stoatlocationswereestimatedfromtwoorthreebearings taken froman individual.Tominimise theimpact of animal movement on location error, allbearings for a given estimate were taken within 20minutes of each other. Field workers would delaycollecting bearings for a particular stoat until theyhad determined a cluster of triangulation stationswith the strongest signals for that animal. Medianerror from triangulationwasestimated tobe157mbytriangulatingtransmittersrandomlyplacedinthefield (Smith, 2006). Estimates fell in the correct habitat 77%ofthetime,withriskofanestimatefallinginthewronghabitatbeinghighestwhentransmitterswereclosetothealpinegrassland–beechforestecotonalboundary (Smith 2006). Sequential locations of thesameindividualswereobtainedatleast2hoursapartinordertoreduceanyeffectsofserialautocorrelation(Rooneyetal.1998).

Analysis of data

TriangulationsTriangulations were calculated with a programdevelopedinMicrosoftAccessbytheDepartmentofConservation’sNationalKakapoteam.Thisprogramhadagraphicalinterfacethatallowedtheremovalofbearings.Bearingswereremovedif(1)theyfailedtointercept with another bearing or (2) three bearingswereinterceptedbuttheydidnotformatriangle.Inthissituationthetwobearingswithanangleofintersectclosestto90°werechosen(White&Garrott1990).

Home range In this study, Minimum Convex Polygons (MCPs)(Mohr1947;White&Garrott1990)wereestimatedfrom100%ofthelocationsofastoat.Thesewereusedtodescribetheextentofmovementsforeachstoat,andto assess each stoat’s available habitat. Incrementalanalysis (Kenward et al. 2003) was undertaken foreachMCPtodeterminewhetheranadequatesampleoflocationshadbeenachieved.Incrementalanalysisinvolvesplottingestimatesofhome-rangeareaagainstthenumberoflocationsusedtoobtaintheestimates,inthe order in which they were obtained in the field. In theory,homerangeareashouldapproachanasymptote

whenanadequate sample size is reached (White&Garrott1990).

Kernelanalysis(Worton1989;Seaman&Powell1996)wasusedtoestimateutilisationdistributionsforstoatswithmorethan40radiolocations.Utilisationdistributionsplottheareainhectareswithincontoursagainstthenumberoflocationsusedtoestimateeachcontour.Contourareaswereplottedfrom20%to100%oftheactualnumberoflocations,at5%intervals.Foreach utilisation distribution the point of inflexion where theslopeofthedistributionsteepenedwasdeterminedby eye. The area below this inflexion has a higher density ofcontoursthantheareaabove,andindicatesacoreareaofuse(Kenwardetal.2003).Rangecoreswerethenestimatedforeachstoatbyplottingthecontouridentified by the inflexion. Adaptive kernels were used instead of fixed kernels because they attach more uncertaintytoouterlocations(Kernohanetal.2001),whichisdesirablebecauseoftheerrorproducedfromtriangulation.Least-squarescrossvalidationwasusedasthesmoothingmultiplieranda‘globalminimum’bandwidthwasselectedtopreventtheanalysisfromfailingifaminimumbandwidthcouldnotbeestimatedfromthedata(Kernohanetal.2001).

All analyses were undertaken using Ranges 6software(Anatrack).Two-dimensionalhomerangesandrangecoresestimatedinRanges6wereimportedintoArcmapTMandclippedagainstbackgroundcontoursfromthestudyareasothatthree-dimensionalsurfaceareacouldbecalculatedforeachhomerangeusingthe‘3Danalyst’toolinArcsceneTM.Thesewereusedtoprovideanindicationofhomerangesizeinthreedimensions.

Habitat useTheproportionofbeechforest,subalpinescrubandalpinegrasslandsineachtwodimensionalMCPwasassessedinArcmapusingdatafromtheNewZealand1:50000Topomapseries.Althoughuseofsubalpinescrub is not central to the hypotheses being tested,thishabitatwasincludedintheanalysisbecausesomeradio-locationestimates fellwithin it.Thesehabitatproportionswerethenusedasestimatesof‘availablehabitat’foreachstoat,whiletheproportionofradiolocations falling in each habitat provided estimatesof‘usedhabitat’foreachstoat(Ericksonetal.2001).‘Availablehabitat’wascomparedwith‘usedhabitat’with compositional analysis, a manova technique(Aebischeretal.1993) that treats theanimalas theexperimentalunitandavailablehabitatasacategoricalcovariate(Ericksonetal.2001).Compositionalanalysisconstructs a matrix of log-ratios of ‘used habitat’versus‘availablehabitat’fromwhichWilk’slambdais calculated.A randomisation test (1000 iterations)isundertakenonWilk’slambda,andiftheobservedvaluecorrespondswithoneofthe50valuesthatfalls

91SMITHETAL.:SELECTIONOFALPINEGRASSLANDS

withinthetailofthedistribution(i.e.0.05),thisshowsnon-randomuseofthishabitat.T-testsarethenusedina post hoc manner to find where the actual difference inuseoccursbetweenhabitats(Aebischeretal.1993).Compositionalanalysismakestheassumptionsthat(1)radio-markedanimalsarespatiallyindependentofeachother,i.e.theydonotaffecteachothersmovements,and (2) compositions (i.e. estimates of availableand used habitat) are equally accurate for differentanimals. Compositional analysis was undertakenusing‘CompositionalAnalysisExcel tool–Version3.2’(Smith2001).

ResultsLive trappingIn2003,ninestoatswerecaught in492.5correctedtrap-nights (Nelson & Clark 1973), consisting ofthree adult males, three juvenile males, two adultfemalesandonejuvenilefemale.Sixofthesestoatswere radio-collared. In2004,19stoatswerecaughtin564.5correctedtrap-nights;17ofthesestoatswereradio-collared.Thesexandageofallstoatstrappedin2004arepresentedinTable1.

Radio trackingIn2003,becauseofaproblemwiththedesignofthe3-mm brass collars, five of these transmitters failed. Thirty-three radio locations were collected for onejuvenilemale(JM1)overthe3-monthperiod.Thisstoathadtoberadio-collaredthreeconsecutivetimesinordertocollectthesedata.In2004,440radiolocationswerecollectedonthe17radio-collaredstoats.Onlytwooftheseradiolocationsfellbelow800ma.s.l.,andnonefellbelow760ma.s.l.

Home rangeIncremental analysis produced mixed results (Fig.1). The incremental analysis plots showed strongasymptotes early in their tracking period of somestoats.However, therewasno consistent asymptoteforsomeofthestoatswiththelargestsamplesizesofradio locationsandtrackedover the longestperiods(i.e.AM1,AM4).Elevenstoatswerejudgedtohaveenough radio-location data over a sufficient period to providereliablehomerangeinformation(thesestoatsaredenotedbyanasterixinTable1).

Thehomerangeofthestoatradio-trackedin2003(JM1)was84hawitharangewidthof1.4km.Thecompositionofhishomerangewas46%alpinegrassland

Table 1.Home-rangeandhabitat-useparametersforstoatsradio-trackedattheBorlandBurnstudysitesbetween6Januaryand19April2004.Homeranges=100%MCPs.%diff=howmuchlarger,proportionately,a3Dhomerangeiscomparedwitha2Dhomerange.Underthecolumn‘Stoat’A=adult,J=juvenile,F=female,M=male,e.g.AF1=adultfemale1.Rangewidthisthedistanceacrossthehomerangeatitswidestpoint.___________________________________________________________________________________________________________________________________Stoat Noof Tracking Home Home % Core Range %MCP %MCP %MCP No. No. No. locations period range range diff range width in in in locations locations locations (days)1 (2D) (3D) (ha) (km) beech alpine scrub in in in (ha) (ha) forest grassland beech alpine scrub forest grassland_________________________________________________________________________________________________________________________

1(AF1)* 47 84 81.0 103 27 18 1.2 35.8 64.0 0.2 8 38 12(AM1)* 57 82 351.0 418 19 100 2.8 57.5 39.3 3.3 27 27 33(AM2)* 46 81 553.0 626 13 62 4.8 51.7 38.0 10.3 14 32 05(AF2)* 46 81 310.0 350 13 64 2.6 9.0 88.1 2.9 1 45 06(AM3)* 49 81 233.0 281 21 74 2.2 0.0 100.0 0.0 0 49 07(JF1)* 27 80 50.0 55 10 1.2 0.0 100.0 0.0 0 27 08(JF2) 9 29 151.0+ 176+ 16 2.0+ 0.0 100.0 0.0 0 9 09(AF3)* 18 77 87.0 106 22 1.8 76.7 23.3 0.0 15 3 010(JF3) 11 31 52.0+ 62+ 19 1.6+ 23.9 51.9 0.0 8 3 011(AF4) 5 5 66.0+ - 1.7+ 0.0 100.0 0.0 0 5 012(AM4)* 43 48 159.0 176 11 110 1.8 8.8 86.2 5.0 3 35 513(JF4)* 18 64 48.0 58 21 1.0 16.7 54.2 29.2 3 10 514(AM5)* 26 47 109.0 127 17 1.8 23.9 61.5 14.7 4 20 215(AF5)* 17 33 153.0 178 16 2.0 32.7 58.3 9.0 4 11 216(JF5) 2 3 - - - - - 0.0 1 1 017(JM2) 9 15 6.4+ 7+ 17 0.9+ 0.0 100.0 0.0 0 9 018(JF6) 8 4 18.0+ - 0.7+ 62.5 37.5 0.0 5 3 0___________________________________________________________________________________________________________________________________

Mean%diff 17 Total 93 327 18___________________________________________________________________________________________________________________________________

*Stoatsusedinthecompositionalanalysis+Estimatesbasedofflowradio-locationsamplesizes(estimatesshouldbeviewedasminimumobservedmovements)1Tracking period is the time between first capture and the last radio location for a given animal

92 NEWZEALANDJOURNALOFECOLOGY,VOL.31,NO.1,2007

A B

Figure 1. Incremental analyses of all stoats radio-tracked in the Borland Burn during summer 2003 and summer 2004.A=males,B=females.

and54%beechforest.Despitethiscomposition,79%(26outof33)ofJM1’sradiolocationswereinalpinegrasslands while 21% (7 out of 33) were in beechforest.Theaveragehome-rangesizeforthe11stoatsradio-trackedin2004,thehomerangesofwhichcouldbeestimated,was194±47.19ha(mean±SE)intwodimensionsand225±53.67hainthreedimensions.Theaveragevaluesforeachsexwere121.5±37.23haintwodimensionsforfemalesand141.67±45.46hainthreedimensions(n=6),and281±79.26haintwodimensionsformalesand325.6±90.16hainthreedimensions(n = 5) (error estimates and 95% confidence intervalsareprovidedinTable2).The100%MCPsforthesestoatsarepresentedinFigure2.

Thereweresixstoats(AF1,AF2,AM1,AM2,AM3andAM4)withlargeenoughsamplesizestoestimaterange cores using kernel analysis. Their utilisationdistributionsareplotted inFigure3, alongwith thepoints of inflexion used to identify the range cores. Rangecoresvariedinsizefrom18to110ha(Table2).Therangecoresofmoststoatsexcludemuchofthe

beechforestintheirMCPsandanybeechforestthatisintheirrangecoresisecotonal(Fig.4).Ingeneral,therangecoresarecentredinthehigh-altitude,alpinegrasslandpartsofeachstoat’sMCP(Fig.4).

Habitat useData from the11 stoatsdenotedwithanasterisk inTable1wereusedinthecompositionalanalysis.Thecomposition(‘availablehabitat’)oftheirhomerangesand the number of locations in each habitat (‘usedhabitat’)withintheirhomerangesareshowninTable1.Becausecompositional analysisdoesnotwork iftheuseofanyhabitatis0%foragivenanimal,0%wasreplacedwith0.1%insevensuchsituations,asrecommended by Aebischer et al. (1993). Overallhabitatusebystoatswasnon-random(RandP=0.043).Alpine grasslands were ranked significantly over beech forest(t8=2.72,P=0.026),buttherewasnodetectabledifference between alpine grasslands and subalpinescrub(t7=1.07,P=0.43)orbetweensubalpinescrubandbeechforest(t7 = −0.58, P=0.63).

Table 2. Error and confidence interval estimates for the 11 stoats considered to have sufficient home-range information (denotedwithanasteriskinTable1).HR(homeranges;ha)=100%MCPs;RW=rangewidth(km);2D=twodimensional;3D=threedimensional.___________________________________________________________________________________________________________________________________

Females(n=6) Males(n=5) Sexescombined(n=11) HR(2D) HR(3D) RW(2D) HR(2D) HR(3D) RW(2D) HR(2D) HR(3D) RW(2D)___________________________________________________________________________________________________________________________________

Mean 121.50 141.67 1.63 281.00 325.60 2.68 194.00 225.27 2.12SD 91.17 111.36 0.56 177.24 201.61 1.25 156.50 178.00 1.06SE 37.22 45.46 0.23 79.26 90.16 0.56 47.19 53.67 0.3295%C.I. 79.91 89.1 0.49 141.82 176.71 1.00 92.48 105.19 0.62________________________________________________________________________________________________________________

93SMITHETAL.:SELECTIONOFALPINEGRASSLANDS

Figure 2.100%MinimumConvexPolygonsforthe11stoatsintheBorlandBurnusedinthecompositionalanalysis(dots=females,solidlines=males).Whiteheavilycontouredareasarealpinegrasslandwhiledarkgreyareasarebeechforest.Scalebar=1km.

Figure 3.UtilisationdistributionsofsixstoatsintheBorlandBurnthathadrangecoresestimatedusingkernelanalysis.Foreach distribution the proportion of locations falling below the first point of inflexion (indicated with a vertical line) was used toestimatetherangecore.AF1=60%,AF2=75%,AM3=80%,AM2=70%,AM4=80%,AM1=60%.

Hec

tare

sH

ecta

res

Hec

tare

s

Hec

tare

sH

ecta

res

Hec

tare

s

94 NEWZEALANDJOURNALOFECOLOGY,VOL.31,NO.1,2007

DiscussionTheseresultsprovideevidencethatstoatsinhabitinghighaltitudesinsouth-easternFiordlandNationalParkincluded alpine grassland within their home rangeduringthesummerandautumnof2003and2004.Thestoatsobservedinthisresearchusedalpinegrasslandsignificantly more often than adjacent beech forest. Their range cores were high up in alpine grasslandandcontainedlittlebeechforest(Fig.4).Rangecoresarelikelytorepresentdensites,orfavouredforagingareas. These results show that these stoats lived inalpinegrasslandsandspentmostoftheirtimethere,but that some of them made occasional forays intobeechforestareas.

Four stoats thatwerepresent in both2003 and2004wereoriginallycaptured(during2003)at trapsites within their 2004 home ranges. Nine of thestoatsreportedintheseresultswerere-trappedatsiteswithintheir2004homerangesduringaculloutofthe

Figure 4.RangecoresofstoatsinhabitingalpinegrasslandsintheBorlandBurn,estimatedusingkernelanalysis.(A)=AM3,(B)=AF2,(C)=AM2,(D)=AM1,(E)=AF1and(F)=AM4.Whiteheavilycontouredareas=alpinegrassland,darkgreyareas=beechforest,lightgreyareas=subalpinescrub,blackuncontouredareasarelakes.Scalebars=1km

populationin2005(Smith2006).Theseobservationsindicate that thisresidencyinalpinegrasslandswasmaintained between years. It is not known whetherthesespatialpatternsweremaintainedinwinter,butitshouldnotbeassumedthatstoatsmovedownintotheforestduringthewintermonths.

Due to difficult field conditions, the sample sizes of radio locations on individual stoats are variable,with asymptotic approaches that were sometimesvague. However, when the period over which the11 stoats (selected forcompositionalanalysis)wereradio-trackedisconsidered(Table1),andgiventheycan move up to 3 km in 3 hours (pers. obs.), thereislittledoubtthesestoatswereresidentintheareasdescribed by their home range estimates during thesummer of 2004. However, given the ambiguity ofincrementalanalysisand theeffectsof triangulationerror,theMCPsshouldnotbeseenasprecise.Instead,theyindicatethescaleofmovementandhomerangeofthestoatsobserved,andprovideausefulframework

95SMITHETAL.:SELECTIONOFALPINEGRASSLANDS

fordevelopinghypothesesabouttheamountsofeachhabitattypeavailabletoeachstoat.

TheMCPestimatesthattookintoaccountreliefofthemountains(i.e.threedimensions)wereonaverage17%larger than thosecalculated in twodimensions(Table1).TheseMCPsshouldbeusedbyresearcherswhowanttocompareresultsfromthisstudytoMCPscalculated for stoats radio-tracked on flat terrain.

Thepower-linemaintenanceroaddidnotfallwithinthehomerangesofmoststoats(Figs2&4)anddidnotappear to influence their movements. It is also unlikely thattheroadpermittedinvasionofthesesitesbystoats,whichoccuratallaltitudesthroughoutNewZealand(King&Murphy2005),includingalpinegrasslandsfarfromanyroad(Smith&Jamieson2005).

The spatial and temporal replication of thisresearch, when combined with Smith & Jamieson(2005),representstwositesandthreesummers.Thissuggeststhatstoatsarelikelytobepresentinalpinegrasslandsinotherareas,whichthereforearenotactingasbarrierstostoatdispersal.TheresearchofCuthbertetal. (2000)representsa thirdobservationofstoatsbeingpresentandactive in thealpineenvironmentsofNewZealand.

Groundweta (Hemiandrus spp.)were themostcommonpreyforstoatsinhabitingalpinegrasslandsintheBorlandBurn(Smith2006)andtheMurchisonMountains (Smith et al. 2005). Brown hare (Lepus europaeus occidentalis)werealsocommonpreyforstoats in the Borland Burn (Smith 2006), but haresareabsentfromtheMurchisonMountains.Inadjacentlow-altitudebeechforest in theBorlandBurn,birdsandmicewerethemostcommonpreyofstoats(Smith2006)andthisisconsistentwithstoatdietstudiesinother beech forest sites in years where there is noheavy beech seedfall (Murphy & Dowding 1995).GroundwetaandharesmayinpartbeplayingaroleintheselectionofalpinegrasslandsbystoatsintheBorlandBurn,butotherecologicalphenomenamayalsobeimportant.

The presence of stoats in alpine grasslandshas important implications for protected speciesmanagementinnationalparksinNewZealand.Thefact thatoutof440radio locations,collectedon17stoatsoverathreeandahalfmonthperiod,nonefellbelow 760 m a.s.l. indicates that potentially manystoatslivewellabovethereachofvalley-basedcontroloperations.Whilemostcontrolfocusesonprotectingforest species such as mōhua (Mohoua ochrocephala)and kākā (Nestor meridionalis)(Dilksetal.2003),allmainlandcontrolareasinNewZealandarelikelytobesubjecttoreinvasion.Theresultsofthisresearchsuggest that nearby high-altitude areas with significant areasofalpinegrasslandsmaybepotentialsourcesofreinvasion.Trappingstoatsinalpinegrasslandswillbelogistically difficult and therefore expensive. However,

managers aiming to control stoats in different partsofFiordlandorinothernationalparksshouldassesswhetherstoatsarepresentinalpinegrasslands.Suchassessmentcouldbemadeusingtrackingtunnels(King&Edgar1977).

Theselectionofalpinegrasslandsbystoatshasthepotentialtobeacomplicatingfactorforconceptualmodels of stoat population biology developed forNewZealand’snationalparks,suchasthebeechmastmodel(King1983).Forexample,whatroledoalpinegrasslands play in the survival and recruitment ofstoatsinnationalparkssuchasFiordland?Isitpossiblethatstoatsinalpinegrasslandsareimportantdriversof landscape-level population dynamics in nationalparks such asFiordland?This research formedpartofamoreintensiveinvestigationintostoatecologyinalpinegrasslandsintheBorlandBurn,whichconcludedthat during these years alpine grasslands were nota sink habitat for beech forest populations (Smith2006).Thereforeyoungstoatsdispersingfromalpinegrasslands may influence landscape-level population dynamicsinnationalparkssuchasFiordland.Further,areindigenousspeciesthatinhabitalpinegrasslandsatriskfromstoatpredation?Suchquestionscanonlybeansweredwithcontinuedresearch.

AcknowledgementsWethankG.PickerellandP.Leiforassistancewithfieldwork, and D. Shanahan for assistance with the use ofGIS.ThisresearchwasfundedbytheDepartmentof Conservation’s Stoat Technical Advisory Group(ResearchContract:3652)andtheZoologyDepartment,UniversityofOtago.Wealso thankC.Kingandananonymousrefereeformanyusefulcomments.

ReferencesAebischer NJ, Robertson PA, Kenward RE 1993.

Compositionalanalysisofhabitatusefromanimalradio-trackingdata.Ecology74:1313–1325.

Alterio N, Moller H 1997. Diet of feral house catsFelis catus, ferrets Mustela furo and stoats M. erminea in grassland surrounding yellow-eyedpenguin Megadyptes antipodes breeding areas,SouthIsland,NewZealand.JournalofZoology243:869–877.

CuthbertR,SommerE2002.Homerange,territorialbehaviour and habitat use of stoats (Mustela erminea) in a colony of Hutton’s shearwater(Puffinus huttoni), New Zealand. New ZealandJournalofZoology29:149–160.

Cuthbert R, Sommer E, Davis LS 2000. Seasonalvariationinthedietofstoatsinabreedingcolony

96 NEWZEALANDJOURNALOFECOLOGY,VOL.31,NO.1,2007

ofHutton’sshearwaters.NewZealandJournalofZoology27:367–373.

Dilks P,WillansM,PrydeM,Fraser I 2003.Largescale stoat control to protect mohua (Mohoua ochrocephala)andkaka(Nestor meridionalis)intheEglintonValley,Fiordland,NewZealand.NewZealandJournalofEcology27:1–9.

ElliottGP1996.Productivityandmortalityofmohua(Mohoua ochrocephala).NewZealandJournalofZoology23:229–237.

ElliottGP,DilksPJ,O’DonnellCFJ1996.Theecologyof yellow-crowned parakeets Cyanoramphus auricepsinNothofagusforestinFiordland,NewZealand. New Zealand Journal of Zoology 23:249–265.

Erickson PW, McDonald TL, Gerow KG, HowlinS,Kern JW2001. Statistical issues in resourceselectionstudieswithradio-markedanimals.In:MillspaughJJ,MarzluffJMedsRadio trackingandanimalpPopulations.SanDiego,CA,USA,AcademicPress.Pp.209–242.

Fitzgerald BM 1977. Weasel predation on a cyclicpopulation of the montane vole (Microtus montanus) in California. Journal of AnimalEcology46:367–397.

GrueHE,KingCM1984.Evaluationofagecriteriain New Zealand stoats (Mustela erminea) ofknownage.NewZealandJournalofZoology11:437–443.

Kenward RE, SouthAB,Walls SS 2003. Ranges 6v1.2:Fortheanalysisoftrackingandlocationdata.Onlinemanual.Wareham,UK,Anatrack.

KernohanBJ,GitzenRA,MillspaughJJ2001.Analysisofanimalspaceuseandmovements.In:MillspaughJJ,Marzluff JMedsRadio trackingandanimalpopulations. San Diego, CA, USA, AcademicPress.Pp.125–166.

King CM 1983. The relationships between beech(Nothofagus sp.)seedfallandpopulationsofmice(Mus musculus),andthedemographicanddietaryresponses of stoats (Mustela erminea), in threeNewZealandforests.JournalofAnimalEcology52:141–166.

KingCM,EdgarRL1977.Techniquesfortrappingandtrackingstoats(Mustela erminea);areview,andanewsystem.NewZealandJournalofZoology4:193–212.

KingCM,McMillanCD1982.Population-structureanddispersalofpeak-yearcohortsofstoats(Mustela erminea)in2NewZealandforests,withespecialreference to control. New Zealand Journal ofEcology5:59–66.

King CM, Murphy EC 2005. Stoat. In: King CMed. New Zealand handbook of mammals. 2ndedn. Melbourne, Oxford University Press. Pp.261–287.

Lavers RB, Mills JA 1978. Stoat studies in theMurchisonMountains,Fiordland.In:Seminaronthe takaheand itshabitat. Invercargill,WildlifeService,NZDepartmentofInternalAffairs.

Lee WG, Jamieson IG 2001. The takahe: fifty years of conservationmanagementandresearch.Dunedin,UniversityofOtagoPress.132p.

MartinoliA,PreatoniDG,ChiarenziB,WautersLA,TosiG2001.Dietofstoats(Mustela erminea)inan Alpine habitat: the importance of fruitconsumption in summer. Acta Oecologica 22:45–53.

McLennanJA,PotterMA,RobertsonHA,WakeGC,ColbourneR,DewL,JoyceL,McCannAJ,MilesJ,MillerPJ,ReidJ1996.Roleofpredationinthedeclineofkiwi,Apteryx spp.,inNewZealand.NewZealandJournalofEcology20:27–35.

MillerC,ElliottM,AlterioN2001.Homerangeofstoats(Mustela erminea)inpodocarpforest,SouthWestland,NewZealand:implicationsforacontrolstrategy.WildlifeResearch28:165–172.

MohrCO1947.TableofequivalentpopulationsofNorthAmerican small mammals. American MidlandNaturalist37:223–249.

MurphyEC,DowdingJE1995.EcologyofthestoatinNothofagusforest:homerange,habitatuseanddietatdifferentstagesofthebeechmastcycle.NewZealandJournalofEcology19:97–109.

NelsonL,ClarkFW1973.Correctionforsprungtrapsin catch/effort calculations of trapping results.JournalofMammalogy54:295–298.

PurdeyDC,KingCM,LawrenceB2004.Agestructure,dispersion and diet of a population of stoats(Mustela erminea)insouthernFiordlandduringthedeclinephaseof thebeechmast cycle.NewZealandJournalofZoology31:205–225.

RooneySM,WolfeA,HaydenTJ1998.Autocorrelateddata in telemetry studies: time to independenceandtheproblemofbehaviouraleffects.MammalReview28:89–98.

SeamanDE,PowellRA1996.Anevaluationof theaccuracyofkerneldensity estimators forhomerangeanalysis.Ecology77:2075–2085.

SmithDHV2002.Themovement,diet and relativeabundance of stoats (Mustela erminea) in theSpecial TakaheArea, Fiordland National Park.Unpublished MSc thesis, University of Otago,Dunedin,NewZealand.

SmithDHV2006.Movements,populationdynamicsandpredatorybehaviourofstoatsinhabitingalpinegrasslandsinFiordland.UnpublishedPhDthesis,UniversityofOtago,Dunedin,NewZealand.

SmithDHV, Jamieson IG2005.Lackofmovementofstoats(Mustela erminea)betweenNothofagusvalley floors and alpine grasslands, withimplicationsfortheconservationofNewZealand’s

97SMITHETAL.:SELECTIONOFALPINEGRASSLANDS

endangeredfauna.NewZealandJournalofEcology29:45–52.

SmithDHV,JamiesonIG,PeachRME2005.Importanceofgroundweta(Hemiandrusspp.)instoat(Mustela erminea)dietinsmallmontanevalleysandalpinegrasslands.NewZealandJournalofEcology29:207–214.

SmithPG2001.CompositionalAnalysisExcel tooluser’sguide.Version3.2.Unpublished:1,BettwsCottage,Bettws,Abergavenny,NP77LG,UK.

WhiteGC,GarrottRA1990.Analysisofwildliferadio-trackingdata.SanDiego,CA,USA,AcademicPress.383p.

WilsonPR,KarlBJ,ToftRJ,BeggsJR,TaylorRH1998. The role of introduced predators andcompetitors in the decline of Kaka (Nestor meridionalis) populations in New Zealand.BiologicalConservation83:175–185.

WortonBJ1989.Kernelmethodsforestimatingtheutilization distribution in home-range studies.Ecology70:164–168.

EditorialBoardmember:CatrionaMacLeod