Acta Theriologica48(1):101—111,2003.
FLISSN0001-7051
Estimatingthesizeof Europeanrabbits consumedbypredators:
Relationshipbetweenbodymassandtoothdimensions
JavierCALZADA’,Daniel T. HAYDON2and Francisco PALOMARES3
CaizadaJ.,Haydon D.T. andPalomaresF. 2003. Estimatingthesize of European rabbitsconsumedbypredators:Relationshipbetweenbody massand toothdimensions. Acta Theriologica43: 101—111.
A methodfor estimatingbody mass of EuropeanrabbitsOryctolaguscunicnlus (Linnaeus,1758)basedon toothdimensionsisproposed.Regressionmodelsidentified significantrelationshipsbetweenthe body mass of 87 rabbitsand individualtooth length, breadth,productoftoothlengthand breadth, and whetherornot theindividual was infectedwith myxomatosis.Dimensionsof10 of14differentteethexplainedover
80%ofvariationin body mass,andthoseteethwere selectedas adequatepredictorsof
rabbitbody mass.Modelswere testedusingteethfrom 16 additionalrabbitsofknown body mass.Body mass,predictedon thebasis of 9 of the10 selectedteeth, was statistically indistinguishablefromthe observedvaluesfor all 16 individuals. When myxomatosisinfectionstatus oftherabbitwas includedin themodel,aU10 selected teethyieldedpredictionsstatisticallyindistinguishablefromthoseobserved.Prediction errorscan be computedpermitting statistical comparisonof the averagepredicted valueofbody massfromdifferentsamplesofrabbits.Themodel isusefulinestimating
rabbitbody massesfromteethrecoveredfromfeces ofpredatorsanditwill facilitate testingof hypotheseson size-selectivepredation.The methodwas appliedto rabbit
teethfoundin feeal samplesfromtheIberiaalynx Lynxpardinus collectedover a one-yearperiod.Lyaxpreyedpreferentially uponyoungerrabbits duringthe peak breedingperiodofthislagomorph.
DepartmentofApplied Biology,EstaciónBiolOgicadeDonana,OSlO,Avda MariaLuisa s/n, Pabelldndel Peril, 41013Sevilla,Spain,e-mail;(JO, FP); Centrefor TropicalVeterinaryMedicine,Easter Bush,Roslin,Midlothian,Scotland EH259RG, UnitedKingdom(DTH)
Key words: Oryctolaguscuniculus,Lynx pardinus,body mass prediction,preyremains, size selection
Presentaddresses:
‘Departamentode BiologIa Anibiental y Salud Pdblica, Facultad de CienciasExperimentales, Universidadde Huelva,CampusUniversitario de La Rábida,21819 Palosdela Frontera,Huelva, Spain,e-mail:
2 Department of Zoology, University of (Iuelph, Guelph, Ontario Canada N1G 2W1, e-mail:
Towhom allcorrespondenceshouldbeaddressed
Iloli
Introduction
TheEuropean rabbit Oryctolaguscuniculus(Linnaeus, 1758) hasspreadfrom itscenter oforigininSpaintoallcontinents exceptAntarctica andispresent on morethan800islandsworldwide(Flux1994).Asaresult,rabbits arepreyofmore than80vertebratepredators,(DelibesandHiraldo1981,JaksicandSoriguer1981, Soriguer1981,ThompsonandKing1994,Palomares andDelibes1997).Because rabbitsrepresentanimportantresourcebaseforsuchavarietyof predatorspecies, the details oftheir occurrence inpredatordiet isofinterestto ecologistsand wildlifemanagers.
Inthispaper,weproposeamethodforestimating thebodymassofEuropean
rabbitsusingdimensionsoftheirteeththatcanfrequentlyberecoveredintactfrom fecesandpelletsofpredators. Forinstance, CaizadaandPalomares (1996)found teethin69 and64%of thefeceswithrabbitremains,ofIberianlynxLynxpardinus and red foxVulpesvulpes,respectively. Distinctive tooth morphology makesit relativelyeasytodistinguish one toothfromanother (Popeskoetal.1990),so estimatingbodymass basedontooth dimensions couldreduce the difficulty of fragmentidentification,andcanallowidentificationofmultiplepreyitemswithina singlescatorpellet(CarssandElston1996).
Weconstructed andparameterizedastatisticalmodelthat allows estimation of thebodymassofindividual rabbits consumedasprey,andthe average massof groupsofindividual rabbits usingindividual tooth dimensions together withan associated prediction error.We alsoexaminedtheeffectoffactorsthat could potentially influencethebodymassof arabbit(sex,seasonality, andmyxomatosis infection status).We tested theaccuracyandprecisionofthemodelwithan independentsampleofrabbits. Finally,weappliedthetechnique toestimate the bodymassofrabbits consumedbytheendangeredIberianlynxoveroneyearinthe DonanaNational Park.
Materialandmethods
We used87rabbitsselectedfrom theMammalCollectionoftheDonanaBiologicalStation. They werecollectedover all seasonsduringtheyears1995, 1996 and1997froman areaofmediterranean scrublnndcalled Coto del Rey, situatedinthenorthof DoSanaNationalPark (37’09’NS°26’W), southwesternSpain.The areais flat withsandy soils and is coveredwith largeshrubscomprised mostly ofPistacia lentiscusand Halimiwnhalimifoliumscrubs,pulsepastures,and scatteredQucrcus subetfrees.Theclimateismediterraneansubhumid,withwetwintersandhat.,dry summers,withcu. annualrainfallbetween500 and 600 mm.
RabbitsofBananagenerallyweigh less thantheirEuropeancounterparts(Rogersetal. 1994).In
ourstudyarea,meanadultbodymassis1100 gfor malesandnon-pregnantfemales,whereaspregnant femalesareapproximately150 gheavier(Calzada2000).MyxomatosisisprevalentwithintheDofiana populationstovaryingdegrees(Rogersetal.1994)andisassociatedwithweightloss, amongother pathologicalsigns,ininfectedindividuals(Calvete1999).
Rabbitsrepresentedall sex, age, and reproductiveclasses,and includedanimalsinfectedwith myxomatosis(Table1).Myxomatosisdiagnosiswasdeterminedbasedonexternalclinicalsigns:facial skin‘anions(especiallypatentinnoseandears), puswappuatingand swellingeyeaandgenitalorgans
BodymiiiandtoothdimensionsInrabbits108
Table1.Rabbitsusedinthe study.Rabbitswere classifiedasadultif
810gfor males,and >750gfor females(minimumbody massof reproductiverabbitsin Uohana,Villaiherte1994):The presenceor absenceofmyxomatosissignsdetermineddiseasestatus.
HealthYoungAdult status
Diseased
Healthy
(Calvete1999). Rabbitbody mass rangedfrom 105 g atfirstemergencefrom burrowto 1350 g, maximaladultbody mass recordedinthepopulation.
Rabbitteethwere extractedwith laboratoryforcepsafter soft boilingtheskulls.Teethwere cleanedof any dirtwith atoothbrush,and then maximallengthand breadthof eachtoothwere measuredby digitalcalliperwithaprecisionof 0.01 mm. Teethweredefinedas one of14 anatomic types (Popescoet al. 1990;Table2). To identify,which parametersaffectrabbit body mass,a preliminarygenerallinear model (GUM) was constructed examiningthe relationshipbetweenthe naturallogarithmofrabhitbody mass end toothdimensions,effect ofsex, season,and myxomatosis infection.As therewere 14 differenttoothtypes,toothdimensionswere representedby thescores correspondingto thefirst8 principalcomponents obtainedfroma principalcomponentsanalysis (Sharma1996)ofthe dimensionsofall14 toothtypesfor all87rabbits.
The resultsofthisanalysiswere usedto guide theconstructionof an additionalGLM for each toothtypewhich usedthenaturallogoftha ithrabbitbody mass (wO53)asthedependentvariable,and
Table2. Parametervaluesand theirstandarderrorsinbracketscorrespondingto eq. (1)for the14 toothtypes.I— incisor,P— premolar,M— molar.Non-zerovaluesoftheparametersare significantly differentfromzero.
CErrormean
(tooth length) / (tooth breadth) / (interaction) / (the constant) / square(Z2) / R2ji / 4.076 (0.794) / 2327(0.5a1) / —1067(0322) / —t444(L086) / 0.046 / 0.85
7.279(1.352) / 3.772(1.240) / —3.657(1.312) / —0.489(1.232) / 0.081 / 0.73
P2 / 1.155 (0.320) / 0.737(0.176) / 0 / 2.894(0.228) / 0.071 / 0.76
P3 / 4.952(0.810) / 1.717 (0.405) / —0.878(0.225) / —2.717(1.259) / 0.040 / 0.86
P4 / 4.931(0.729) / 1.713(0.423) / -0.892(0.214) / —2.637(1.218) / 0.039 / 0.87
7.152 (0.921) / 3.149(0.484) / —1.614(0.273) / —6.987(1.496) / 0.041 / 0.86
6.022(0.965) / 2.920(0.495) / —1.468(0.305) / —4.836(1.382) / 0.043 / 0.86
M3 / 2.816(0.291) / 0 / 0 / 4.670 (0.207) / 0.110 / 0.53
4.945(0.875) / 3.286(0.634) / —1.479(0.357) / —3.656(1.412) / 0.056 / 0.82
J3 / 0 / 2.835(0.108) / 0 / —0.761(0.258) / 0.058 / 0.80
P4 / 4.734 (1.073) / 3.901(0.739) / —1.417(0.368) / —5.825(1.881) / 0.048 / 0.84
6.185 (1.076) / 5.465(0.894) / —2.036(0.406) / —9.516(2.112) / 0.047 / 0.85
4.593 (0.854) / 3.954 (0.750) / —1.422(0.346) / —5.493(1.667) / 0.044 / 0.85
M3 / 2.125 (0.573) / 3.041(0.988) / —1.016(0.495) / 1.268 (0.908) / 0.074 / 0.76
toothlength(x),breadth(x1,2),theproductoftoothlength and breadth(xj,),and diseasestatus(x4)
asindependentvariablesto model predictedlogbody mass
m+in2+ri,:1 in3+ x14 m+c(1) Equation(1)ismoreconvenientlyexpressedin matrixvectornotationas W=XM.HereXis a matrixof87rows and five columns,thefirstthreecolumnsaretoothlength,breadth,and productof lengthand breadth,thefourthcolumnisequalto 1iftherabbitisdiseasefree,and 0ifdiseased,
whereasthefifthcolumnis equalto1and representstheconstantin aregressionmodel. 14 isa
columnvectorcontainingthefivemodel coefficientsestimatedusingtheleastsquaresmethod(Sokal and Rohlf 1995).Usingthesemodelsitis.possibleto estimatethe ln(mass)ofany additionalrabbit
fromany toothand thediseasestatususingtherelationshipfui1 =x1M,wherexis a row vector
containingtheindependentvariablesorderedinthesameway as asinglerow of X.The standard
deviationoftheithindividualprediction(si)madefrom thesegenerallinearmodelsisgiven by:
(2)
where2istheerrormeansquareassociatedwiththeunderlyinggenerallinearmodel (see Hocking
1985:58).
Estimatesofthemean(LI)and standarderror(SE)ofsetsofteethfromapopulationofNrabbits
can bemadp byaveragingthepredictedweightsforeach individual:
N
“_ 1(8)
w
and addingtheaverageofthevarianceofeach predictionto the samplevariancebeforeproceeding with thecalculationofthestandarderrorintheusualway:
SE=*Zs2+EY
(4)
whereoisthe standarddeviationofLIintheabsenceofuncertaintyintheindividualpredictions.
All modelswereconstructedusingrabbit body masstransformedtothenaturallogs, but the non-transformedmasscan beestimatedusingenapproximateback transformation ofLIwhere:
w
estimatedrabbitmass =a(1+—f-)(5) Becauseitisoftenpossibletoidentifyseveraldifferentteethbelongingtothesamerabbit,wealso
calculatedthevariancesand covariancesbetweenbody massprethctionsobtainedfor thesame
individualusingdifferentteeth,definedas:
—w1,) (iI’5—w1,)l(6)
wherei2i,1and &)j,3areestimatesoftheweightofthesameindividualobtainedusing thejthand kth teeth.Thesecovariancesare usefulbecausetheyindicatetheadditionalinformationabout rabbitbody massthatcan beobtainedfrom additionalteethfound infeces.
The preliminaryGLMindicatedthatlength,breadth,theirproduct,and thediseasestatushave a significantinfluenceinthemodels(see Results).However,at presentmyxoniatosisinfectionstatus cannotbe determinedfromremainsfoundin feces or pellets,and therefore,areducedmodel was constructedthatexcludeddiseasestatus(ie not fittingtheparametert,z4).
Wetestedtheaccuracyand precisionofthe model foreach tooth usingaset of16differentrabbits ofknownbody mass,sex, and diseasestatus(7females,9males;one diseased),and evaluatedboththe reducedmodel (excludingdiseasestatusknowledge)and thefull model (with diseasestatusknowledge). We comparedtheobservedand predictedbody masses(basedon each tooth)ofthe16 rabbitswith a pairedt-test(usingaBonferronicorrected valueofa=0.05/10since we performaseparatetestfor eachtooth).
To determinetheaccuracyand precisionofthecombinedreducedmodelsas afunctionofsample size, acomputerprogramwasdevelopedtorepeatedlysampledifferent-sizedsub-samplesofrandomly
Bodymass and toothdimensionsin rabbits105
selectedteeth?Only thoseteeth,themeasurementsof which could predictover 80% variationin ln(bodymass),wereused.Theresulting pool of 870 teeth wasrandomlysub-sampled with replacement. Predicted meanln(bodymass) from 100 randomlyselectedsub-sampleswas calculated foreach ofseveraldifferentsamplesizes and thepredictionsplottedasafunctionofsamplesize. The reducedmodel (ietheone omittinginformationaboutdiseasestatus)was usedtopredicteachrabbit body mass from thetoothmeasurements.
In orderto show apracticalapplicationofthemodelsproposedhere,we examinedthemonthly
prey selectionofIberianlynx in the studyarea,Lynx feces were collectedmonthlybetweenFebruary and December1995.Feceswere washedona1.5-mmdiametersieve underrunningwater,and their contentswas separatedand identifiedwith thd help ofreferencecollections.Fromtherabbitremains, teeth were separated,cleaned,end measuredwith adigitalcaliper,and thereducedmodelwas used to predictbody mass. Aone-wayANOVA wasconductedtotestfor differencesinthemeanlogarithmic bodymassofpreyeatenbyLynxeachmonth.Sequentialmonthlychangesinbody mass ofpreyselected bylynx was examinedusingpairwisemultiplecomparisonBonferronicorrected(0.05/10)for the 10 testsover the11 months.
PCA analysis,GLMs, ANOVA, and associatedpost-hoctests were performedusingSPSSv.7.5
(SPSS1989). The normalityoftheresidualsinall modelswaschecked byKolmogorov-Smirnovtests,
and homoscedasticityofvariances bythetest suggestedbyGlejser(1969).
Results
Residualsfromallmodelswerenormallydistributed,andtheir variances were homoscedastic.Thefirstprincipalcomponent(PGA1)explained83%ofvariationin toothdimensions,andwasagoodpredictorofrabbitbodymass(Fig.1).Thesecond andthird principalcomponent(PCA 2andPCA3)onlyexplainedanadditional6%
7.5 -
o
7.0
6.5000
6.0
o
80008
0
00
-o
_55.50
0
0
0
0
-10-505101520
1Stprincipalcomponent
Fig.1.Relationshipbetweenthefirstprincipalcomponentandln(bodymass)for103rabbits.Thefirst principalcomponentsummarizes83%ofvariationintoothwidth,breaIth,and productofwidthand breadth.
oftheremainingvariation,thereforeonly the scorescorrespondingto the first principalcomponentwereincludedinthispreliminaryanalysis.
The resultsof the preliminaryGLM analysisindicatedthat therewereno significantinteractions betweenany of the variablesexcepttoothlengthand breadth,nor any detectabledifferencebetweenmaleand femalerabbits(F163
1.422,p 0.237).Diseasestatusexplainedasignificantamountofvariationin rabbitbodymass,diseasedindividualsbeing anaverageof162glighter thannon diseasedindividuals(F159=14.28, p < 0.001). Seasonexplainedasignificant amountofthevariationin body massonly in theabsenceof knowledgeabout diseasestatus.Togetherdiseasestatusand seasonexplainedonlyanadditional6% ofvariationover thatexplainedbyPCA1 alone.
Regressionmodelsthatexcludeddiseasestatusasaparameter,explainedover
80% ofvariationfor 10 ofthe14 individualteeth (Table2). The 4 teeththat explainedtheleastvariation(12,P2,M3,and M3)wereexcludedfromsubsequent analyses.The45covariancesforthe20selectedteethwereallpositive,remarkably uniform,and averaged80%ofthe valueofthe variances.
1200-
U, (I)
E
0
0-
E
C,)
1)
1100- l000i
900-
V
800-
700-
600-
500k
samplemeans
samplestandarderrors
822g
O6
O4
0.2
400-r’
0
0
20406080
Samplesize
Fig. 2. Reliabilityofestimatesofaveragedbody massofrabbits,obtainedfrom samplesofdifferent numbersof individuals.Eachdot istheaveragedpredictionfromasampleof N individuals sub-
-sampledwith replacementfromthedatasetof87rabbitsandtheir870teeth.SampLesizes were3,10,
20, ...SO.One hundredresamplingswereperformedfor each samplesize categorySamerandom
‘jiggle’has been added tothex-axis toallow differentsub-samplestobedistinguished. Theobserved meanbody massof822 gfor the87 rabbitsisindicated.The right-handaxis indicatesthe standard errorofeach sub-samplemean.
Bodymail andtoothdlmeneioniinrabbita107
Predictions ofthe average bodymassofthe16rabbits basedonthereduced modelusingeachofthe10selectedteethseparatelyindicatedthat9ofthe10teeth predictedaveragebodymassesstatisticallSiindistinguishablefromthoseobserved (pairedt-tests: 1,606 t 2.694,allp 0.017).Theexceptionwasthe P3tooth
Ct = 5.174,p 0.001), which significantly underestimated rabbit body mass.
Estimationsofeachindividualrabbitln(bodymass)differedbyanaverageof0.138
(SD=0.116)from the observedvalue,ieby104g(SD=87),using the back
transformationproposedinequation (5).Useofthefullmodelincludingdisease statusresultedinall10pairedt-testsindicatingnosignificantdifferencesbetween predictedandobservedbodymasses(—0.384t 2.823,allp 0.0127).Estimations ofeachindividualrabbitln(bodymass)differedbyanaverageof0.128(SD=0.116) fromtheobservedvalue,ie103g(SD=-84).
Usingthe standarddeviations obtained fromthe resampling procedure, and assuming the distributionofresampled means tobenormal, the probability of obtaining estimatesofaveragebodymasses within100gofthe true mean with samplesizesof3,10,and 20was46,83,and94%respectively.Theprobability of obtaining estimatesofaverage bodymasseswithin 50gofthe true meanwith samplesof20,40,60,and80was 65,75,86,and91%(Fig.2).
Forthepreyselectionstudy,wecollected556lynx feces.Rabbitremainswere foundin99%ofthesamples.Intact rabbitteeth wererecoveredfrom81%ofthe
6.75
6.50
U, U,
2
6.00
0
-Q
C
—I
5.50-
5.25-
Feb MarApr MayJunJulAug Sep OctNov Dec
Month
Fig.3.Monthlyrabbitsizeselectionbylynx.Dotsanderrorbarsindicatemeanln(bodymass)and95%
confidenceinterval.‘ftabbtapredatedbylynxeachavxth.
feces with rabbitremains.Atotalof516bodymassesofrabbitsconsumedbylynx wereestimated. ANOVA showedthat differencesinbody massesof rabbits consumedmonthly were significantlydifferent (F10 28.270,p 0.001). Pairwisemultiplecomparisonamongmonthsindicatedthat rabbits consumed betweenFebruaryandApril were significantlysmallerthatthoseconsumedover therestofthe year(Fig. 3).
Discussion
Ourmodelsallow rabbitbody massto be predictedfrom toothdimensions. Predictionerrorscan becomputedand combinedwith standarderrorsofsample means,so that averagedpredictionsfromdifferentsamplescan be statistically compared.Takenindividually,10out of14teethexplainover80%ofthevariation inthe observedrabbitbody masses,and 9oftheseteethsuccessfullypredictedthe meanmassofadditionalindependentlyacquiredrabbits.Knowledgeofmyxomatosis infectionincreasedthe percentageofvariationexplainedby ca 6%. Resampling simulationssuggestedthatwith samplesizesof20individualweightestimates,the
estimateofthesub-samplemeanhasover a90%chanceoffallingwithin100gof
theoverallsamplemean.Samplesizes 40shouldplace theestimatedsamplemean within50goftheoverallsamplemeanwith probabilityatleast0.75.
Covariancesbetweenpredictionsofindividualbody massesobtained from multipleteethofthesameindividual(possiblyextractedfromthe samejaw bone for example)werepositiveand were typically80% ofthevarianceofestimates obtainedfromsingleteeth. Thisobservaticnsuggeststhat little additional infor,nationis likelyto be availablefrommeasurementsof multipleteeth, and effortexpendedincollectingdataonmultipleteethfromasingleindividualislikely togolargelyunrewarded.Ourmodelssuggestthatitwil] he only slightlyless efficienttoestimatemassfromasingletooth.Furthermore,when teethoccur separatelyin thefeces, attemptingtouse more thanonetoothcould resultinthe mixingupofteethofmultiplepreyitemsofsimilarbodymass. Ourstudyindicated thatanyofthefollowing9teethcan beused toestimaterabbitbodymass;Ii,Pa,
M’,M2,I,P4,M1orN12.The P3toothismore problematic,andthe teethj2 2
P
M ,and M3shouldberegardedaslast choices (Table2).Furthermore,the small
sizeof12,M3,and M3(neverexceeding1.95 mm)willlikelyincreasemeasurement
error.
Individualsfreeofinyxomatosisinfectionwere typically—20%heavierthan infectedones inour studiedpopulation.Myxomatosisprevalenceinthe 87rabbits fromwhichthesemodelswereconstructedwas26.4%.Itisnotcurrentlypossibleto determineinfectionstatusfromrabbitremainsfoundinfecesandpellets,however ifsome idea ofrnyxomatosisprevalenceamongstsampledrabbits isknownthese figuressuggestacrudemultiplicativecorrectiontothemean(unlogged)weightsof groupsof rabbits predictedby themodel.Let pbe the diseaseprevalence.
/1
Multiplying e (1+ bythequantityl.O45—O.l74pwillapproximately correct
formyxomatosisprevalencesother than the26%inbuilt inthemodel.
Themodelcanbeusedtotestfordifferencesbetweentheaveragebodymassof different groups ofrabbits. In the example oflynx prey selection, the rabbits predated duringthefirst monthsoftheyearweresignificantlylighter than those killedfrom June toDecember.The predicted average bodymass ofthe rabbits collectedfromApriltoSeptember differssignificantly(with95%confidence)from theobservedmeanbodymassfortheOctobertoMarchusingalloftheselected teethexceptP3.Differencesobservedinthemeanbodymassesofrabbitsconsumed
bylynx are congruent with rabbit biologyin our study area, En Donana, the
breedingperiodusuallylastsfromOctobertoMay,reachingpeakinMarch,April and May(Delibesand Calderón1979).Onthe otherhand, rabbit growth isfast, reachingadult sizein3.5months (Villafuerte1994).TheIberian lynxisarabbit specialist (Delibes1980),that preysuponthemostavailablerabbit ageandsize classes(Calzada2000).Duringthefirst months oftheyear,the mostabundant rabbitsize classesinthefieldarethelighterones,andlynxpreymoreheavilyupon them.ReproductionisfinishedinMay,andlighterrabbitsbecomerareinthefield untilthenextbreedingperiod.Therefore,lynxpreyuponheavierrabbitsfromJune until thebeginningofthenextreproduction periodofrabbits.
Obtaining information about size-selectivepredation ofrabbits, aswithmost smallpreyitems, isdifficult.Hunting evedts arerarelyobservedand,evenwhen witnessed,sizeofselectedpreyitemsisdifficulttodetermine. Oneapproach isto captureandmarkrabbitswithradiotransmittersandrecordwhenaretheykilled, butthesostsofmaterialsandlabourusedinrabbitradiotracking studiesrenders collaring and monitoring oflarge numbers ofrabbits prohibitively expensive. Another approach is to analyze prey remains left by predatorsat carcasses, skeletons, orbigosseousstructuresofingested preythat can befoundin pile remainsandinbiggerraptor pellets(JanesandBarss1985,DondzarandCeballos
1989,Dickmanetat.1991,Rohneretat.1995,BarryandBarry1996).Incarnivore
feces,longosseousstructuresaremorefragmented becausecarnivores crush the bones before swallowing, and onlythe smallest hard remains of prey canbe recoveredintact (Calzadaand Palomares1996).Sosizeselectionstudieshavebeen basedonsmallhardpiecesofpreysuchasotoliths,scales,nails,vertebralcentrums orteeth, amongothers (Wise1980,Feltbam andMarquiss1989,Heggberget and Moseid1994,Zalewski1996).Theaccurateidentification ofremainssuchasdistal phalanxesorvertebralcentrumscanposeseveredifficulties.Asaresult,whensuch remainsareuseditisusuallyonlypossibletoestimate rabbitweightasbelonging tooneofasmallnumberof sizecategories (Palomaresand Delibes1991, GoszczyfiskiandWasilewski1992).
Whileitmaybedifficultorevenimpossibletostudy size-selectivepredation from the remains ofmanytypes ofsmallormedium sizeprey,itispossibleto
J.Caizadaetal.
examineit usingrabbit toothremains:Accurateestimatesof rabbitsize com
positionin predatordietsispossibJe usingthe methodswedescribed.
Acknowledgements:WethankA.Devenoges,A.Meijers,A.deRoes, A.delCampo,and T.Agudo,who helpedintherabbitteethextractions.WethankM.Delibes,B.Revilla,N.FernándezP.Ferreras,J. Litvaitisand twoanonymousrefereesforconstructivecommentsonpreviousdraftsofthemanuscript. Theresearchwas supportedby DGICYT (projectPB94-0480)and DGES(projectPB97-1163),and sponsoredby LandRover Eapafia. Thefirstauthorwas supportedby apredoctoralgrantfromthe SpanishMinistryofEducationandCulture.DTH wassupportedbyagrantfromtheWeilcomeTrust.
References
BarryR. E. and BarryL. M. 1996.Speciescompositionand age structureof remainsof hyraxes
(Hyracoidea:Procaviidae)atnestsofblack eagles.JournalofMarnmalogy77:702—707.
CalveteC.1999.Epidemiologiadeenfermecladhemorrágica(VHD)ymixomatosis en elconejosilvestre (OryctolaguseuniculusL. i758)enelvalle mediadelEbro.ModelizaciOndeVHD yherrainientas degestiôn.UniversityofZaragoza,Zaragoza:1—287.
CalzadaJ.2000. impactodedepredacionyselecciondepresa delIiLnceibéricoyelzorrosobreelconejo.
UniversityofLeon,LeOn:1—252.
CalzadaJ. and Palomares1.1996.Frecuenciade apariciOn de diferentesrestosde conejoen excrementosdelinceyzorro.DonanaActaVertebrata23:243—252.
CarssB. N. and ElstonD. A. 1996.Errors associatedwith otter Lutralutrafaecalanalysis.II.
Estimatingpreysize distribution frombonesrecoveredin spraints.JournalofZoology, London
238: 318—332.
DelibesM.1980. El linceibOrico.EcologiaycomportsmientoalimenticiosenelCoto Donans,1-luelva, DonanaActa Vertebrata‘73:1—128.
DelibesM.and CalderOn J. 1979.DatossobrelareproducciOndelconejo,Oryctolaguscuniculus(L.),en
Doflans,8.0.deEspafla,duranteun ano sew.DofianaActsVertebrate6:91—99.
DelibesM.and HiraliloF.1981.The rabbitssprey inthemediterraneanecosystem.[In:Proceedings of’theWorld LagomorphConference,Gland,1981.K.Myers and C.B. Macinnes,eds].University ofGuelph,Ontario:600—613.
DickmanC.IL, Predavec1W.arid LynamA.J.1991. Differential predationofsireand sex classesof
micebythebarnowl,Tyto alba.Oikos62:67—76.
Donazar,J. A. and Ceballos0. 1989.Selectivepredationby EagleOwls Bubobubo on rabbits
Oryctolaguscuniculus:Ageand sexpreferences.OralsScnndinavica20:117—122.
FelthamM.J.and Marquiss M.1989. The use offirstvertebraein separating,and estimatingthesize of, trout(SaunaIrutta)and salmon(Salmosalar)in boneremains.JournalofZoology, London
2±9:113—122.
Flux J.B.C.1994.World distribution.[In:The Europeanrabbit:thehistoryand biologyofasuccessful colonizer.H.V.Thompsonand C.1st. King, eds]. Oxford UniversityPress,Oxford:8—21.
GlejserH.1969. Anew testfor heteroscedasticity.JournaloftheAmericanStatisticalAssociation64:
316—323.
GoszczynskiJ.and WasilewskiM.1992.Predationoffoxes on aharepopulationin centralPoland.
ActsTheriologica37:329—338.
HeggbergetT.M. and Moseid K.-E. 1994. Preyselectionin coastalEurasianottersLutralutra.
Ecography 17:331—338.
HockingR.R.1985.Theanalysisoflinearmodels.Brooks/Cole,Monterey:1—385.
JaksicF.M.and SoriguerR.C.1981.Predationupon theEuropeanrabbit(Oryctologuscuniculus)in mediterraneanhabitatsofChile and Spain:acomparativeanalysis.JournalofAnimal Ecology 50:
269-232.
Janes S.W.andBarssJ.M.1985.Predation bythree owl speciesonnorthernpocketgophersof differentbodymass.Oecologia67;76—81.
Palemares F.andDelibesM.1991.AlimentaciOndelmeloncillo.Flerpestesichneumonydelagineta Genettagenetta enInReservaBiolOgicadeDoñnna,S.0.deIaPeninsula Iberica. DoflanaActa Vertebrata 18:5—20.
PalornaresF.andDelibesM.1997.PredationuponEuropeanrabbitsandtheiruseofopenandclosed patchesinMediterraneanhabitats. Oikos80;407—410.
PopeskoP.,RajtovaV.andHordk11990.A colouratlasoftheanatomyofsmalllaboratoryanimals.
Vol1;Rabbit,Guineapig.Wolfe,London:1—255.
RogersP. M.,Arthur C.P.andSoriguer R.C.1994.Therabbit incontinental Europe. [In:The European rabbit: Thehistoryandbiologyofasuccessfulcolonizer.H.V.ThompsonandC.M. King,ecls].OxfordUniversityPress,Oxford:22—63.
Rohner C.,Smith J.N.M.,Stroman J., JoyceM., DoyleF. I. and Eoonstra R.1995. Northern hawk-owlsinthe nearcticborealforest:preyselectionand population consequencesofmultiple preycycles.TheCondor97:208—220.
SharmaS.1996.Appliedmultivariate techniques.NewYork,Wiley:1—512.
SokalH.R.and RohlfF.J.1995. Biometry.W.H.FreemanandCompany,NewYork;1—887.
SoriguerR.C.198L Biologiaydindmicadeunapoblacióndeconejos(Opyctolaguscunicutus,L.)en
AndaluciaOccidental.DoflanaActa Vertebrata8:1—343.
SPSS 1989.SPSSv.7.5.User’s manual.Chicago.
ThompsonhI.V.andKing C.M.1994. TheEuropean rabbit;thehistoryandbiologyofasuccessful colonizer.OxfordUniversityPress,Oxford:1—245.
VillafuerteR.1994.Riesgodepredacidnyestrategias defensivasdelconejo,Oryctolagascuniculus,en elParqueNacionaldeDodana.UniversityofCórdoba,Cdrdoba:1—229.
WiseM.Fl.1980.Theuseoffishvertebraeinseatsforestimatingpreysizeofotters andmink.Journal ofZoology,London192;25—31.
ZalewskiA.1996.Choiceofageclassesof bankvolesClethrionomysglareotusbypinemartenMartes
manes andtawnyowl Strixalaco inBialowiea NationalPark.AetaOecologiea17:233—244.
Received9’i’Jovember2001,accepted 13May 2002.