Diet overlap and relative abundance of sympatric dasyurid carnivores] a hypothesis of competition

Ecology 0887\ 56\ 309Ð310 Diet overlap and relative abundance of sympatric dasyurid carnivores] a hypothesis of competition MENNA E[ JONES and LEON A[ BARMUTA Department of Zoology\ University of Tasmania\ GPO Box 141!94\ Hobart TAS 6990\ Australia Summary 0[ Diet overlap among age and sex classes of sympatric dasyurid carnivores "Mar! supialia] Dasyuridae# at Cradle Mountain National Park in Tasmania\ Australia\ was determined to assess the likelihood of current interspeci_c competition\ which could in~uence and explain the disparate population densities of the three species[ 1[ The carnivore guild divided into two groups based on body size and prey size\ within which diet overlapped] Tasmanian devils "Sarcophilus laniarius# and male spotted!tailed quolls "Dasyurus maculatus#\ which consumed larger prey species\ and female spotted!tailed quolls and eastern quolls "D[ viverrinus#\ which consumed smal! ler prey species[ Male spotted!tailed quolls overlapped in diet with adult devils in winter\ but not in summer[ However\ in summer the small number of male spotted! tailed quolls overlapped in both body weight and diet with a large cohort of young devils[ Too few data were obtained to repeat these analyses with female and young spotted!tailed quolls and eastern quolls\ but results indicated that a similar pattern of overlap may occur[ 2[ Spotted!tailed quolls would experience the highest degree of dietary overlap with another species of carnivore\ with all age and sex classes experiencing overlap for much of the year[ Adult devils and young eastern quolls would both be free of overlap for more than half the year[ 3[ No indications of seasonal food limitation\ when competition is most likely to occur\ were found during this study\ but this may occur over a longer time scale[ 4 If the high degree of diet overlap experienced by spotted!tailed quolls means higher competitive pressure\ this may explain the low density of this species at Cradle Mountain[ 5[ These results\ high levels of interference experienced by spotted!tailed quolls\ and the behavioural and numerical dominance of devils\ the largest species in the guild\ support Brown + Maurer s "0875# and Cotgreave s "0882# ideas that competitive dominance may be more important than energetic equivalence in determining relation! ships between body size and abundance in local assemblages of animals[ Key!words] abundance\ carnivore\ competition\ dasyurid\ diet[ Ecology "0887# 56\ 309Ð310 Introduction A _rst step in investigating the importance of com! petition in structuring guilds is measuring overlap in resource use[ With information on resource require! ments and resource availability\ this may provide indirect evidence for the likelihood of one species a}ecting another "MacNally 0872#\ although a high degree of overlap does not necessarily mean that com! petition is taking place "Wiens 0866#[ Competition may result in reduced fecundity\ growth or energy stores of individuals and reduced density and:or an altered age structure at a population level "Dunham 0879 reviewed in MacNally 0872^ Korpimaki 0876^ Petren + Case 0885#[ While population densities of animals typically decrease with increasing body size over large geo! graphical areas\ an observation from which the {ener! getic equivalence rule was derived "Blackburn\ Har! vey + Pagel 0889^ Gri.ths 0881^ Cotgreave 0882#\ this does not often hold within local communities where the relationship is variable "Gri.ths 0875^ Blackburn et al[ 0889^ Cotgreave + Harvey 0880^ Nee et al[ 0880^ Cotgreave + Harvey 0881^ Cotgreave 309

300 M[E[ Jones + L[A[ Barmuta 0882#[ At the scale of feeding guilds\ the relationships for British birds vary with guild type[ Only insec! tivores show a negative relationship^ carnivores and herbivores are likely to not show any relationship and omnivores a positive relationship "Cotgreave + Harvey 0880^ Nee et al[ 0880^ Cotgreave 0882#[ Find! ing positive relationships between body size and abun! dance in local assemblages of rodents\ birds\ _sh and plants\ and _nding that the larger species accounted for more of the energy ~ow in the local ecosystem\ Brown + Maurer "0875# suggested that the ecological advantages of large body size\ such as greater energy e.ciency\ greater mobility and more e.cient homeo! static mechanisms\ enable large!bodied species to use a greater range of habitats[ Also\ large species can dominate resource use within habitats[ Competitive dominance in interspeci_c interactions and better avoidance of predators enable more individuals of a larger species to be supported on the same amount of energy than smaller species[ These results contradict the {energetic equivalence rule and suggest that while purely evolutionary phenomena may explain patterns at the highest levels of ecological organization\ at the lower levels such as guilds\ ecological forces such as competition may provide explanations "Cotgreave 0882#[ A three!species assemblage of dasyurid carnivores ðcomprising Tasmanian devils Sarcophilus laniarius "Boitard 0730# "see Werdelin 0876# spotted!tailed quolls Dasyurus maculatus "Kerr 0681#\ and eastern quolls Dasyurus viverrinus "Shaw 0799#Ł at Cradle Mountain in subalpine Tasmania\ Australia\ exhibits patterns of relative abundance consistent with both the lack of relationship between body size and abun! dance found in carnivorous British birds "Cotgreave + Harvey 0880^ Nee et al[ 0880^ Cotgreave 0882# and with numerical dominance of the largest species found by Brown + Maurer "0875#[ Since character dis! placement and its basis in diet separation by prey size\ which constitutes evidence of competition operating on an evolutionary time scale\ has been demonstrated for this guild "Jones 0886#\ and the middle!sized spec! ies is very rare\ it is possible that abundance is a}ected by current competition[ The aim of this study was to assess the likelihood that current interspeci_c competition\ which could in~uence the population densities of the three species\ is occurring[ This can be done through examination of diet overlap between age:sex classes of species at di}erent times of year[ Di}erences in feeding ecology\ including prey type and size\ between age groups and sexes have been demonstrated in numerous species "Polis 0873#[ The observed patterns of body weight overlap among age!sex classes of the three species\ where spotted!tailed quolls overlap with both of the other two species in body weight at some stage of their development "Fig[ 0#\ suggest that there may be diet overlap and\ therefore\ the potential for competition[ Interspeci_c competition may be occurring for resources other than food[ In the very wet environ! ment of alpine Tasmania\ fresh water is freely avail! able\ but dry sites suitable for dens might be limited[ However\ there are large di}erences in tunnel dimen! sions\ and siting of dens with respect to habitat type and topography among the three species\ so it appears unlikely that competition for den sites is occurring "Jones 0884#[ Methods FIELD SCAT COLLECTION AND POPULATION CLASSES The _eld study was conducted in a 19!km 1 area at the northern end of the Cradle MountainÐLake St Clair National Park in the central highlands of Tasmania\ Australia[ Part of the South!west Tasmania World Heritage Area\ this nearly pristine site supports sym! patric populations of the three species and is remote from farming areas and\ hence\ from the in~uence of domestic sheep and poultry in the diet[ With elevation ranging from 659 to 0979 m\ and bounded to the south by an alpine plateau with peaks to 0434 m "Cradle Mountain#\ the area experiences cool summers "mean maximum temperature 05>C\ range 5Ð22>C# and cold\ wet winters "mean minimum temperature 9=6>C\ range 7Ð01>C# with continuous rain and frequent sleet and snow from July to October[ Annual precipitation averages 1257 mm "range 1209Ð1392 mm#[ For analy! ses of seasonal di}erences in diet overlap\ winter was de_ned as from July to October\ inclusive\ and sum! mer as November to March[ From November to March\ weather conditions are milder and drier\ and a large cohort of inexperienced juvenile prey species are becoming independent[ Scats for analysis of diet overlap were obtained from trapped animals individually marked with a number tattooed in the ear[ Animals were trapped overnight "and released the next morning# in specially constructed wire cage traps "Jones 0884# baited with beef liver\ which leaves no residue in the scat[ Sixty traps spread over 19 km 1 were placed to maximize capture success of di}erent individuals of all species[ Each trap was set for only three nights each trip\ minimizing the number of recaptures[ Trapping was carried out every second month between October 0889 and April 0882[ Trapped individuals of all species were assigned to a population class based on age\ sex and body size[ Current year young were classed as juveniles if they were caught during the months when females were lactating^ between October "the earliest date that young of the year were trapped# and January for devils[ So few young eastern and spotted!tailed quolls were trapped that accurate weaning dates were not determined for these species\ although the breeding season overlaps with that of devils[ Mating\ birth\ cessation of pouch life and weaning occurs in March\

301 Dasyurid carnivore diet overlap and abundance Fig[ 0[ A simpli_ed growth curve showing body weights of the age:size:sex population classes of dasyurid carnivores at Cradle Mountain\ Tasmania[ Males are represented by solid lines\ females by dashed lines[ April\ August and January for devils^ and in June\ July\ SeptemberÐOctober "for both species of quolls#\ and November "eastern quolls# and December "spot! ted!tailed quolls# "Fleay 0839^ Troughton 0843^ God! sell 0872^ Bryant 0877^ personal observations#[ From February until their body weight reached the mini! mum weight for mature adult females "see below#\ young were classed as sub!adults "Fig[ 0#[ Females of all species were classed as adults "referred to as female# if their body weight exceeded the minimum stable weight for a healthy\ mature female recorded during the study*4 kg for devils\ 0=4 kg for spotted!tailed quolls and 9=4 kg for eastern quolls*or had bred previously "old females in poor condition weighed as little as 3=4 kg^ Fig[ 0#[ Males were classed as small adults "called {small male # if their body weight was within the adult female weight range^ 4Ð7 kg for devils\ 0=4Ð1=5 kg for spotted! tailed quolls "see comment below# and 9=4Ð0=9 kg for eastern quolls "Fig[ 0#[ All males in this category were still growing[ While females reached _nal adult weight within 07 months of independence\ most males took 1 years to attain their greater _nal adult weight[ Male devils and eastern quolls were classed as full!grown adults "referred to as males# if their body weight was greater than the maximum weight for adult females "Fig[ 0#^ 7=9 kg for devils and 0=9 kg for eastern quolls[ A higher minimum weight of 1=5 kg "maximum female weight is 0=8 kg# was set for male spotted!tailed quolls because the extreme sexual size dimorphism in body weight "male weight 199) of female weight# of spot! ted!tailed quolls meant that males smaller than 1=5 kg still had the gracile build typical of females and bore little resemblance to robust full!grown males[ While these weights represented those of full!grown adults\ the male class included some individuals in the _nal stages of growth[ Independent sub!adults overlapped in body weight with males of the next smallest species as they grew for four to six months of the year] sub!adult devils:male spotted!tailed quolls overlapped from February to May "for male devils which grew faster# and July "female devils#^ sub!adult spotted!tailed quolls:male eastern quolls from December to June[ Adult females overlapped in body weight with dependent juveniles of the next larger species for 3 months of the year] female spotted!tailed quolls:juvenile devils over! lapped from October to January[ Too few juvenile spotted!tailed quolls were trapped to determine the period during which they overlapped in body weight with female eastern quolls[ Relative abundance of di}erent population classes during periods of the year was estimated from the mean number of individuals "MNA# known to be alive^ those individuals that were trapped during the period or both before and after[ ESTIMATION OF DIET COMPOSITION Scats were washed through 0! and 9=0!mm stacked sieves\ and fur\ feathers\ bone and invertebrate

302 M[E[ Jones + L[A[ Barmuta remains\ and _ne sludge stored separately in 69) alcohol[ Mammalian prey were identi_ed to genus or species by microscopic examination of cross!sections of fur "Brunner and Coman 0863^ Taylor 0874#[ Possums\ macropods and wombats were classed as adult or juvenile where possible by comparison of bone fragments in the scats with reference skeletons in the Tasmanian Museum and Art Gallery\ and diag! rams in Green and Rainbird "0872#[ Avian prey were classi_ed to order level\ or species if the feathers were distinctive\ using Day "0855# and a reference collec! tion[ Reptiles were classi_ed to species using a key for Tasmanian reptiles "R[ Swain\ unpublished# and reference specimens "by R[ Swain and S[ Jones#[ Crus! taceans were identi_ed to species level "by A[ Rich! ardson# and insects to order "CSIRO 0880#[ Earth! worms were detected in scats by microscopically examining the _ne sludge for chaetae "Kruuk and Parish 0870#[ Forty food types were found in dasyurid carnivore scats from Cradle Mountain[ In decreasing size order\ mammals found were] wombats "Vombatus ursinus tasmaniensis#\ domestic dogs "Canis familiaris#\ Bennett s wallabies "Macropus rufogriseus rufog! riseus#\ Tasmanian devils "Sarcophilus laniarius#\ Tas! manian pademelons "Thylogale billardieri#\ echidnas "Tachyglossus aculeatus setosus#\ spotted!tailed quolls "Dasyurus maculatus maculatus#\ brushtail possums "Trichosurus vulpecula#\ platypuses "Ornithorhynchus anatinus#\ eastern quolls "Dasyurus viverrinus#\ ringtail possums "Pseudocheirus peregrinus viverrinus#\ brown bandicoots "Isoodon obesulus obesulus#\ sugar gliders "Petaurus breviceps breviceps#\ swamp rats "Rattus lutreolus velutinus#\ long!tailed mice "Pseudomys hig! ginsi#\ antechinuses "Antechinus swainsonii swainsonii or A[ minimus minimus#\ white!footed dunnarts "Smin! thopsis leucopus leucopus#\ house mice "Mus musculus#\ and pygmy possums "C[ lepidus or C[ nanus nanus#[ Using the hair analysis technique\ antechinuses "Ante! chinus swainsonii swainsonii\ A[ minimus minimus# and pygmy possums "Cercartetus lepidus\ C[ nanus nanus# could only be distinguished to Genus[ Birds identi_ed were Tasmanian native hens "Gallinula mortierii#\ black currawongs "Strepera fuliginosa# and green rosellas "Platycercus caledonicus#\ with small pass! erines combined in one category\ and bird eggs as a separate category[ Reptile species found were tiger snakes "Notechis ater#\ white!lipped whip snakes "Drysdalia coronoides# and metallic skinks "Niv! eoscincus metallicus#[ Invertebrate prey included two species of terrestrial cray_sh "Parasticoides tasmanicus and Astacopsis tricornus#\ earthworms\ spiders\ mil! lipedes\ moth pupae and larvae "O[ Lepidoptera#\ beetles "O[ Coleoptera#\ grasshoppers "O[ Orthop! tera#\ cockroaches "O[ Blattodea#\ dragon~ies "O[ Odonata#\ wasps "O[ Hymenoptera# and ~ies "O[ Dip! tera\ including larvae#[ Plant material\ food scraps and rubbish were also found[ There were very few unidenti_ed prey items[ Because the importance of each prey type is taken into account\ percentage biomass ingested is a better measure of diet composition than percentage occur! rence\ which over!estimates the importance of small dietary items and under!estimates large items "Corbett 0878#[ Percentage biomass was estimated as follows] the number of individuals of each prey type in a scat were counted or estimated\ and multiplied by the mean mass for that species "Strahan 0872^ Slater\ Slater + Slater 0878#[ An assumption was made for all prey types larger than 4 g "all except invertebrates and skinks# that no more than one individual was present in each scat[ This produces a conservative estimate of biomass for large prey types[ For skinks and invertebrates other than earthworms\ the mini! mum number of individuals present was estimated from numbers of identi_able parts\ such as legs and heads\ which gives a relatively reliable estimate of numbers "Dickman + Huang 0877#[ For earthworms\ an assumption was made that only one earthworm was present per scat[ Gizzards\ which are often used to count numbers in scats\ were not found "Bradbury 0866#\ but the numbers of chaetae in individual scats were low "4Ð29#[ Scavenged human food and plant material were not quanti_ed[ An upper limit on biomass ingested was placed on prey types larger than an individual of an age:sex class could eat in one meal[ These limits were set at] 39) of body mass for devils\ which consume up to this amount in a single meal "Pemberton + Renouf 0882#^ 12) for eastern quolls\ derived from studies of food consumption based on sodium turnover "Green + Eberhard 0872#^ and 12) for spotted!tailed quolls\ for which there are no published data\ but are more similar in feeding habits to eastern quolls than devils[ The mass limits were set at 9=1 kg for female eastern quolls\ 9=2 kg for male eastern quolls\ 9=3 kg for female spotted!tailed quolls\ 9=6 kg for male spotted!tailed quolls\ 1=1 kg for female devils and 2=3 kg for male devils[ These values are consistent with feeding obser! vations in the _eld[ The largest prey species that were counted as being eaten entirely were sugar glider "Petaurus breviceps# for quolls "both sexes and spec! ies#\ adult ringtail possum "Pseudocheirus peregrinus# and juvenile brushtail possum "Trichosurus vulpecula# for female devils and adult brushtail possum for male devils[ Mass of juvenile possums\ macropods and wombat was designated as one quarter of adult mass[ A limitation of this method of estimating biomass is that the number of scats per meal was not known[ All scats "up to _ve# deposited in a trap overnight were treated as one sample and it was assumed that these represented an equivalent meal across species and population classes[ Given the large range of pred! ator size "699 g to 09 kg#\ feeding habits "insectivore: carnivore\ carnivore and specialized scavenger#\ and prey sizes and types consumed by each population class\ the feeding trials required to derive conversion factors such as those used by Floyd\ Mech + Jordan

303 Dasyurid carnivore diet overlap and abundance "0867# and Corbett "0878# for large mammalian prey fed to wolves were beyond the scope of this study[ For example\ eastern quolls scavenging on a large mammal do not consume much fur[ This would lead to an under!estimate in the percentage diet com! position for these species as the quantity of fur in a scat is used in the conversion equations[ Social dominance may also in~uence the proportions of bone\ fur and ~esh consumed by devils "Jones 0884#\ thus rendering a conversion factor inaccurate[ STATISTICAL ANALYSES The chance of bias\ resulting from one individual spe! cializing on a particular prey type\ was signi_cant because the number of individuals per population class was often small[ The small number of prey items in scats precluded the use of a test of heterogeneity "Kruuk + Moorhouse 0889#\ but x 1 tests showed no signi_cant di}erences "all P 9=94# in diet com! position between individuals within population classes[ All data from each individual "the sample unit# were used\ except for adult female devils where\ because a large number of samples were collected\ 099 scats were subsampled in a way that maximized the number of individuals sampled in each month[ Dietary data based on prey taxa usually violate assumptions of normal distribution and show too much individual variation to allow a standard con! tingency table approach "Patterson 0875#[ The Mantel test\ a non!parametric test that overcomes these di.culties\ was used to provide signi_cance tests for the values of diet overlap between population classes or species "Patterson 0875^ Belbin 0882#[ This test calculates the correlation between a matrix containing values for diet overlap between all individuals of two groups "e[g[ species\ population class# and a species identity matrix\ then through random permutations where the elements of one matrix are shu/ed\ con! structs a reference distribution and tests whether the correlation between the original matrices is higher or lower than would be expected if there was no di}er! ence between the two groups "Fortin + Gurevitch 0882#[ A null hypothesis of complete overlap was used\ the opposite tail of the distribution than is usually used in the Mantel test[ As the Mantel test compares only two samples at a time\ the Bonferroni procedure was used to correct for the use of multiple com! parisons "e[g[ between _ve pairs of adjacent age:sex classes#[ Where sample sizes were disparate and a result of no signi_cance was gained\ the weighting procedure devised by Luo "0882# was employed[ Diet overlap between individuals of adjacent species and population classes was calculated using Schoener s "0857# simple index of resource overlap which\ while limited in usefulness for interpreting similarity of resource use "Hurlbert 0867#\ was appropriate because no data on prey species abundance\ which almost certainly varied\ were collected[ With frequently only one prey item per scat and just one or two scats per individual\ there were a large number of zero diet overlap values when 26 food types were used\ so the number of prey categories was reduced to 09[ The mammalian prey assemblage at Cradle Mountain\ which comprised the bulk of car! nivore diet\ was divided into four signi_cantly or near! signi_cantly di}erent size categories using the Body Mass Di}erence Index "BMDI# developed by Holling "0881#[ A signi_cant gap in the size!ranked assemblage exists if a value of the BMDI exceeds one standard error "9=67 in this study# above the mean\ followed by at least two values below the mean[ Two signi_cant size gaps were found\ that between sugar gliders and bandicoots "BMDI 2=1#\ and that between pygmy possums and antechinuses "BMDI 9=81#[ A further gap between pademelons and Bennett s wallabies would have been signi_cant "BMDI 9=85# if the car! nivore species\ which occasionally occur in the diet\ had been excluded from the size distribution[ The four categories used were\ in decreasing size order\ large mammal "wombats and Bennett s wallabies#\ medium! sized mammal "devils to eastern quolls\ including pad! emelons and possums#\ small mammal "sugar gliders to antechinuses# and tiny mammal "dunnarts to pygmy possums#[ Other diet categories were large birds "native hens to rosellas#\ small birds "all small passerines#\ snakes "tiger and whip#\ skinks\ cray_sh and small invertebrates "all invertebrates except cray! _sh#[ Where sample sizes were small\ the number of diet categories were further reduced to six ðlarge\ med! ium and small "including tiny# mammals\ birds\ other "reptiles and cray_sh# and invertebratesł\ to increase the power of the Mantel test in detecting di}erences[ This reduction in categories made no di}erence to any results[ The Mantel test is sensitive to small sample sizes "Fortin + Gurevitch 0882^ Luo 0882# and loses power to detect di}erences at sample sizes smaller than 09[ Sample sizes of 19 or more are recommended[ Mini! mum representative sample sizes for each species were determined following Dickman "0877# using Hur! tubia s "0862# measure of trophic diversity[ Dietary records for individuals were incorporated accumu! latively until trophic diversity\ H k \ stabilized[ Accumulated trophic diversity for devils stabilized after 03 individuals were analysed[ The results for both species of quolls were inconclusive[ Diversity did not increase at all\ remaining stable from the _rst record "a record is the diet composition of accumu! lated scats for one individual#[ The implications for analysis of quoll diet are that large numbers of scats are needed from each individual[ In this study\ very few female quolls were caught "dietary information was available on only six of each species# and many of these were only caught once[ These small sample sizes may have contributed to the lack of signi_cant di}erences in diet overlap found among the quolls[ Another potential limitation of the Mantel test was

304 M[E[ Jones + L[A[ Barmuta encountered in this study] where there were large num! bers of zeros in the overlap matrix "even when only six prey categories were used#\ di}erences that were seemingly apparent in the data were not detected by the Mantel test[ This possible lack of power would arise because the Mantel test relies on comparing the degree of overlap within groups against the degree of overlap between groups[ If within!group overlap is low\ the di}erence between the groups has to be very large "i[e[ nearly zero overlap# to be detected by the test and return a signi_cant result[ Within!group overlap is low if a large number of individuals score 099) of biomass in a single prey category\ resulting in a high proportion of zero overlap values in the diet matrix within a group[ This may have been a problem with female eastern and spotted!tailed quolls[ Very few individuals were trapped and recapture intervals were often long "particularly for spotted!tailed quolls#\ so most individuals were represented by only one or two scats[ With only one or two scats\ the likelihood that only one dietary category is found in the scat\ and therefore 099) biomass is scored in one food cate! gory\ is high[ Where signi_cant di}erences in diet were demon! strated between seasons\ population niche breadth was calculated for each season\ following methods used by Dickman "0875#[ Because no data were col! lected on resource availability\ only the normalized "B# index "Levins 0857# was determined[ Results RELATIVE ABUNDANCE Overall abundance of the largest species\ the Tas! manian devil "Sarcophilus laniarius^ Dasyuridae\ Mar! supialia#\ with 015 individuals "47 male and 57 female# trapped between November 0889 and March 0882\ was two!and!a!half times that of the smallest species\ the eastern quoll "Dasyurus viverrinus] 38 individuals\ 39 male and nine female#\ which was two!and!a!half times more abundant than the middle!sized species\ the spotted!tailed quoll "D[ maculatus] 08 individuals\ 01 male and seven female#[ Even boosting the numbers of female quolls to equal those of male quolls\ to account for possible trap!shyness of female quolls\ does not dramatically alter these ratios[ Therefore\ devils were _ve times more abundant than the rare spotted!tailed quolls[ For the periods of the year when body weight of sub!adults overlapped with adults of another species\ there were always fewer individual spotted!tailed quolls than individuals of the species with which it overlapped[ Adult male spotted!tailed quolls "mean 3# were outnumbered by sub!adult devils "mean 14# and sub!adult spotted!tailed quolls "mean 1# were outnumbered by male eastern quolls "mean 8# "Table 0#[ Too few dependent juveniles of any species were trapped for a similar comparison to be mean! ingful[ DIET OVERLAP BETWEEN SPECIES AND SEX:SPECIES CLASSES All three species of large dasyurids at Cradle Moun! tain are carnivorous\ and include a broad spectrum of prey sizes\ both vertebrate and invertebrate\ in their diet[ When all data from adults "males\ females and small males# of each species were combined\ diet of each of the three species of marsupial carnivores at Cradle Mountain was signi_cantly di}erent from the other two species "Table 1#[ The greatest di}erences were between eastern quolls and devils "Mantel test\ P ³ 9=990#\ and between eastern quolls and spotted! tailed quolls "Mantel test\ P ³ 9=990#[ Detection of the di}erence between spotted!tailed quolls and devils required further reduction of prey categories to six and the di}erence was still only just signi_cant "Man! tel test\ P 9=91\ Bonferroni critical P 9=91#[ The only adjacent age:sex class couplet in which signi_cant di}erences in diet were found was that between male and female spotted!tailed quolls "Man! tel test\ P9=91\ Bonferroni P!value 9=90[ Note] the Bonferroni procedure is conservative^ Table 1#[ Signi_cant overlap was found in the year!round diets of male and female devils "Mantel test\ P9=43#\ female devils and male spotted!tailed quolls "Mantel test\ P9=03#\ female spotted!tailed quolls and male eastern quolls "Mantel test\ P9=14#\ male and female eastern quolls "Mantel test\ P9=01#\ small male devils with both male "Mantel test\ P9=17# and female "Mantel test\ P9=58# devils\ and small male eastern quolls with both male "Mantel test\ P9=53# and female "Mantel test\ P9=50# eastern quolls[ Bonferroni critical P!value was 9=90 for this entire set of analyses[ Sample sizes were small for female quolls of both species "see Methods#[ Devils fed primarily on large mammals such as wall! abies and wombats\ and secondarily on medium!sized mammals "Table 1#[ Small mammals and birds were of minor importance[ Spotted!tailed quolls consumed mammals of all sizes and birds were also important in the diet[ Diet of males and females was quite di}erent] medium!sized mammals followed by large mammals comprised most of the diet of males\ while birds\ fol! lowed by small mammals\ comprised most of the diet of females[ Eastern quoll diet was more diverse than the other two species\ with a representation of at least three percentage in all prey categories[ Medium!sized and small mammals\ birds\ skinks and small invert! ebrates "including insects\ spiders and earthworms# were eaten in fairly equal proportions by mass[ A larger proportion of invertebrates were consumed than by the other two species[ The scats of all three species contained scavenged food scraps and rubbish\ and ingested plant material\ food items that could not be quanti_ed[ Eastern quolls

305 Dasyurid carnivore diet overlap and abundance Table 0[ Relative abundance "MNA# of di}erent population classes of all three species during the period of weight overlap "FebruaryÐMay:July# and during the remainder of the year Adult male Small male Adult female Sub!adult Juvenile Total Feb[ÐMay:Jul[ Devil 6 2 07 14 9 42 Spotted!tailed quoll 3 0 1 1 9 8 Eastern quoll 8 8 4 0 9 13 Jul[:Sept[ÐJan[ Devil 4 4 04 09 00 35 Spotted!tailed quoll 3 0 0 0 9 6 Eastern quoll 2 5 0 9 9 09 These sub!adults are now larger than the adult male of the next smaller species[ Table 1[ Diet composition by percentage biomass of adults "sexes combined# of the three species of sympatric dasyurid carnivores at Cradle Mountain\ and the one adjacent sex:species class comparison "between male and female spotted!tailed quoll# between which a signi_cant di}erence in diet was found[ The number of unidenti_ed prey items was negligible Spotted!tailed Eastern Male spotted! Female spotted! Prey category Devil quoll quoll tailed quoll tailed quoll Large mammal 50 17 3 23 05 Medium mammal 26 22 04 31 04 Small mammal ³0 06 04 09 18 Tiny mammal 9 2 00 3 1 Large bird 0 6 7 2 04 Small bird ³0 00 05 6 11 Snake 9 9 3 9 9 Skink 9 ³0 02 9 9 Cray_sh 9 ³0 2 9 0 Small invertebrate 9 9 00 9 9 Plant 2 44 Food scraps 4 19 6 Rubbish 1 10 19 No[ individuals 43 08 40 00 4 Percentage occurrence[ consumed more plant material than the other species[ It was mostly grass and was not _nely macerated\ so ingestion may be incidental as they forage for cryptic grass!dwelling creatures "Table 1#[ SEASONAL CHANGES IN DIET\ DIET OVERLAP AND NICHE BREADTH There was strong statistical evidence for di}erences in diet between winter and summer in both male "Mantel test\ P 9=991# and female "Mantel test\ P 9=990# devils "Table 2\ Fig[ 1#[ The greater proportion of the diet for both sexes was large mammals in summer and medium!sized mammals in winter[ Male and female diet overlapped in both summer "Mantel test\ P 9=92# and winter "Mantel test\ P 9=11#\ although results for summer approached signi_cance[ Therefore\ comparisons between small male and fully grown adult devils were made only for summer[ Diet of small male devils in summer was very di}erent from that of large adult males "Mantel test\ P9=990#\ but overlapped with that of females "Mantel test\ P9=1#[ There was a trend from small males to females to large males of decreasing proportions of medium!sized mammals and increasing proportions of large mam! mals in the diet\ with only the di}erence between small males and large males being signi_cant "Table 2#[ Bon! ferroni critical P!value was 9=902 for these analyses[ Female devil and male spotted!tailed quoll diet overlapped signi_cantly in winter "Mantel test\ P9=14#\ but di}ered in summer "Mantel test\ P9=990# "Table 2\ Fig[ 1#[ Bonferroni critical P 9=914 for both analyses in this set[ This was prob! ably the result of the switch in female devils to a higher proportion of large mammals in summer[ Diets of male and female spotted!tailed quolls di}ered in summer "Mantel test\ P9=90#\ but not in winter "Mantel test\ P9=11# and there were no di}erences in diet between seasons for either sex "Mantel test\ males] P9=61^ females] P9=49#[ Bonferroni critical P 9=902 for all analyses in this set[ In summer\ diet of males was primarily medium

306 M[E[ Jones + L[A[ Barmuta Table 2[ Diet composition by percentage biomass of adult male and female dasyurid carnivores in summer and in winter[ Data are provided only for those species\ age:sex classes and seasons where signi_cant di}erences in composition were found Male Female Male Male Female Female Small male spotted! spotted! Prey devil devil devil devil devil tailed quoll tailed quoll category "winter# "summer# "winter# "summer# "summer# "summer# "summer# Large mammal 34 66 14 50 31 28 19 Medium mammal 44 10 47 26 45 39 9 Small mammal ³0 ³0 6 0 0 8 18 Tiny mammal 9 9 9 9 9 3 1 Large bird ³0 1 2 0 0 1 12 Small bird 9 ³0 6 ³0 ³0 5 15 Snake 9 9 9 ³0 9 9 9 Skink 9 9 9 9 9 9 ³0 Cray_sh 9 9 9 9 9 9 9 Small invertebrate 9 9 9 9 9 9 9 No[ individuals 00 06 04 23 02 00 3 Fig[ 1[ Year!round diet overlap between male spotted!tailed quoll "seasons combined# and di}erent age classes of devil] in winter with female devil "JulyÐOctober# and in summer with sub!adult devil ðfrom FebruaryÐMay:July "males:females#ł\ but not with female devil "NovemberÐMarch#[ and large!sized mammals\ while that of females was mostly small mammals and birds "Table 2#[ There were too few data for small male spotted!tailed quolls to incorporate this age:sex class into the above analy! ses[ Likewise\ the comparison between female spotted! tailed quolls and male eastern quolls was not analysed[ No signi_cant di}erences in eastern quoll diet were found between seasons\ when sexes were analysed sep! arately "Mantel test\ males] P9=61^ females] P9=20# or when they were combined "including small adult males to boost sample sizes# "Mantel test\ P9=11# or between sexes within seasons "Mantel test\ summer] P9=43^ winter] P9=81#[ In! vertebrate prey were present in the diet only in summer[ Analysis of change in niche breadth with season was conducted only for devils for which strong seasonal di}erences in diet were found[ Results for the nor! malized "B# index for 09 prey categories^ B9=190 for males in winter^ B9=047 for males in summer^ B9=133 for females in winter^ B9=083 for females in summer\ imply that niche breadth was greater in winter[ Close examination of the data reveal\ however\ that it is equitability rather than niche breadth that changes[ The switch to larger mammals in summer represents a specialization of diet\ as a greater pro!

307 Dasyurid carnivore diet overlap and abundance portion of the diet was concentrated in one category "Table 2#[ DIET OVERLAP DURING ANNUAL PERIODS OF INTERSPECIFIC BODY WEIGHT OVERLAP Signi_cant overlap in diet was found between male spotted!tailed quolls and sub!adult devils "Mantel test\ P9=18#\ and between sub!adult and adult devils "Mantel test\ P9=11#\ but not between male spotted!tailed quolls and adult devils "Mantel test\ P9=990#\ during the months of the year when body weight overlapped "Table 3\ Fig[ 1#[ Bonferroni criti! cal P9=905 for these analyses[ The diet of sub!adult devils was intermediate between that of male spotted! tailed quolls and adult devils[ Birds constituted a sub! stantial percentage of the biomass of the diet of both sub!adult devils and adult male spotted!tailed quolls\ but rarely appeared in the diet of adult devils\ and the large mammal component of sub!adult devil diet was intermediate between that of adult devils and male spotted!tailed quolls[ Juvenile devil diet was similar to that of female devils "their mothers# "Mantel test\ P9=31#\ sub! adult devils "Mantel test\ P9=45# and male spotted! tailed quolls "Mantel test\ P9=66#\ and quite di}er! ent from small spotted!tailed quolls "small males\ females and sub!adults^ Mantel test\ P9=990# with which they overlapped in body weight "Table 3#[ Juv! enile devil diet was most similar to that of sub!adult devils\ comprising a smaller size of mammal prey and more birds than female devils\ but a higher proportion of large mammalian species "probably from their mother# and fewer birds than did the larger male spot! ted!tailed quolls[ There was insu.cient data to analyse diet overlap between sub!adult spotted!tailed quolls and adult male eastern quolls during the months of body weight overlap[ Discussion Diets of sympatric devils\ spotted!tailed quolls and eastern quolls at Cradle Mountain were distinctly di}erent[ Mean prey size and therefore prey type was correlated with body size "Jones 0886#\ as is typical of mustelids "Erlinge 0876^ King 0880#[ When diet over! lap between adjacent size!ranked sex:species classes in these sexually size dimorphic species were analysed\ two dietary groups based on body size and prey size emerged] devils and male spotted!tailed quolls\ which consumed mostly larger prey species "medium and large!sized mammals] possums\ wallabies and wombats#\ and female spotted!tailed and eastern quolls\ which consumed mostly smaller prey species "small mammals up to the size of sugar gliders\ birds\ reptiles and invertebrates#[ Finer de_nition of the patterns of diet overlap among the devil:male spotted!tailed quoll group revealed that because devils switched to consuming a higher proportion of large mammalian prey species in summer\ diet overlap between male spotted!tailed quolls and adult devils only occurred in winter[ In summer\ however\ there was a large cohort of young devils "that outnumbered male spotted!tailed quolls by six to one# that overlapped in diet with male spot! ted!tailed quolls\ from when they _rst learnt to hunt as dependent juveniles smaller than the quolls\ until they had grown larger than male spotted!tailed quolls[ The result was that there was no time of year when male spotted!tailed quolls were free of dietary overlap with devils[ Conversely\ adult devils were free from overlap with another carnivore species for more than half the year[ Table 3[ Diet composition by percentage biomass of age:sex:species classes of dasyurid carnivores at Cradle Mountain during annual periods of body weight overlap between sub!adult and juvenile devil and other classes[ All data for year used for small spotted!tailed quolls to increase sample size Sub!adult devil ðfebð May Male All small Male "males#: spotted! Juvenile spotted! Female Sub!adult spotted! Prey July Adult devil tailed quoll devil tailed quoll devil devil "FebÐ tailed quoll category "females#ł "FebÐJuly# "FebÐJuly# "OctÐJan# "All year# "OctÐJan# May:July# "OctÐJan# Large mammal 22 38 3 27 05 43 22 17 Medium mammal 36 36 66 35 06 33 36 22 Small mammal 6 1 3 4 34 0 6 05 Tiny mammal 9 0 ³0 0 2 9 9 2 Large bird 0 0 03 1 ³0 ³0 ³0 6 Small bird 01 ³0 0 4 09 ³0 01 00 Snake 9 9 9 2 9 9 9 9 Skink 9 9 ³0 0 ³0 9 9 0 Cray_sh 9 9 9 9 ³0 9 9 0 Small invertebrate 9 9 9 9 7 9 9 9 No[ individuals 34 31 5 07 02 19 34 00

308 M[E[ Jones + L[A[ Barmuta Too few data were obtained to repeat these analyses on the female and young spotted!tailed quoll:eastern quoll diet group\ but from examination of the data\ I suggest that a similar pattern of overlap may occur[ Diets of female spotted!tailed quolls and adult eastern quolls "small mammals and birds# overlapped\ but the degree of overlap is probably higher in winter than in summer\ when eastern quolls consumed signi_cant proportions of invertebrates[ In summer\ diet of young spotted!tailed quolls probably overlaps with that of adult eastern quolls as their body weight over! laps _rst with female "juvenile spotted!tailed#\ then with male "sub!adult spotted!tailed# eastern quolls[ Juvenile and sub!adult eastern quolls are free of over! lap in body weight and probably free of dietary over! lap with another species[ In summary\ spotted!tailed quolls "sex and age classes combined# would experience the highest degree of dietary overlap with another species of carnivore\ with all age and sex classes experiencing overlap for much of the year[ Adult devils and young eastern quolls would both be free of overlap for more than half the year[ Because no data or model exists of the resource requirements of the carnivores or of resource availability\ at best inferences only can be drawn that spotted!tailed quolls are likely to su}er more from the e}ects of exploitation competition than the other two species[ Intraspeci_c competition or density e}ects\ which may be stronger than relationships between species "MacNally 0872#\ have not been taken into account[ Indications are that these could be strong for devils] interference competition while scavenging on car! casses of large prey species which comprise the bulk of their diet is intense "M[ Jones\ unpublished data# and the diet of young devils resemble that of adult devils more closely than it does that of male spotted! tailed quolls[ This is consistent with other studies which indicate that although body size may overlap\ di}erences between age classes of a species are usually less than those between species "Polis 0873#[ The degree of dietary overlap within species suggests that current intraspeci_c competition may be less impor! tant for spotted!tailed quolls than for the other two species[ Also not considered are the e}ects of intra! guild predation and cannibalism[ Fur of all three car! nivore species was found in devil scats and eastern quoll fur was found in eastern quoll scats\ but as it was usually too cold at Cradle Mountain for ~y maggots\ indicative of scavenged meat\ to develop\ it is not known whether these were predated or can! nibalized[ Competition is only likely to occur if resources are limiting[ Environmental variability means there may be times of year when food is abundant and times when food is scarce "Wiens 0882#[ Resource overlap has generally been found to be high when resources are abundant and to be least during lean times when dietary specialization occurs "Schoener 0871^ Wiens 0878^ Smith 0880#[ If food resources for the carnivores at Cradle Mountain were limited in winter\ the time when plant productivity is lowest\ niche breadth might be expected to decrease[ This was not the case^ niche breadth in devils was greater in winter\ although this represented a change in equitability in the diet rather than specialization\ as adult devils switched to eating larger prey species "probably the naive juveniles of these species# in summer[ No indications of nutritional stress or behavioural or physiological adaptations to nutritional stress were found among the dasyurid carnivores at Cradle Mountain\ that might indicate periods of seasonal food limitation[ Survivorship:residency was actually higher in winter than in summer "Jones 0884#\ although this may re~ect lower mobility\ and therefore higher residency rather than lower mortality[ In summer\ as juveniles become independent and the mating season begins\ residents range widely and tran! sients were trapped in the study area[ There were no indications\ in the rate at which naturally occurring carcasses were eaten\ that the carnivores at Cradle Mountain were short of food in winter[ A small experi! ment\ where _ve feeding stations throughout the study area were provisioned with 04 kg of "road!killed# car! casses of prey species\ six times during 0 year at 1! month intervals\ also showed no di}erences in amount of food consumed in summer and in winter "Jones 0884#[ Another short experiment\ where a site was saturated with food "carcasses of prey species# for two nights and then six brushtail possum carcasses with spool!and!line tracking devices "M[ Jones\ unpub! lished data# attached were placed\ produced no evi! dence that devils cache surplus food^ a behavioural adaptation that may be useful for tiding over food shortages in cold climates "Jones 0884#[ Body fat may be used as an energy source to tide animals over per! iods of food limitation[ No seasonal trends in body condition were found in any species\ except a decline in condition associated with the mating period in males "M[ Jones\ unpublished data#[ Finally\ no evi! dence was found that torpor\ an adaptation for energy conservation especially under food deprivation\ was a normal part of the winter routine of devils or eastern quolls "Jones\ Grigg + Beard 0886#[ The lack of evidence for seasonal nutritional stress may mean that competition among the carnivores at Cradle Mountain is occurring continuously\ or it may re~ect the observation from Jaksic s "Jaksic\ Greene + Yanez 0870^ cited in Wiens 0882# study of vertebrate predators in Chile\ that for guilds of predators that rely on mammalian prey\ as opposed to insect prey\ periods of resource abundance and scarcity are expre! ssed on a scale of years not seasons "Wiens 0882#[ A period of resource scarcity may not have occurred during this _eld study "two!and!a!half years#[ If the high degree of diet overlap experienced by spotted!tailed quolls means higher competitive pres! sure\ this may explain the low density of this species

319 Dasyurid carnivore diet overlap and abundance at Cradle Mountain[ Data on interspeci_c interaction rates suggest that spotted!tailed quolls also experience much higher levels of interference competition at car! casses than either devils or eastern quolls "M[ Jones\ unpublished data#[ In addition to competition\ spot! ted!tailed quolls are at higher risk of predation\ both from owls "Mooney 0882# and devils\ by nature of their smaller body size\ than are devils[ The combined e}ects of higher levels of both exploitation and inter! ference competition\ and predation may in~uence low population densities in spotted!tailed quolls[ Similar patterns of competition\ predation and low popu! lation density have been noted in another mammalian carnivore guild[ High levels of diet overlap\ inter! ference competition at carcasses\ and predation from lions and:or hyaenas correlate with low population densities of African wild dogs "Lycaeon pictus# over their entire range\ and may limit population size of this endangered species "Creel + Creel 0885#[ The possibility that the pattern of abundance among the marsupial carnivores results from di}ering environ! mental preferences and that spotted!tailed quolls sim! ply do not like high altitude\ cold\ wet habitats in Tasmania\ can be discounted[ Cradle Mountain is within the distributional zone of moderate probability of occurrence for all three species of dasyurid car! nivores "Jones + Rose 0885#\ so is probably habitat of intermediate optimality for all the species[ The largest species\ the Tasmanian devil\ is by far the most abun! dant in this three species assemblage[ Devils are com! petitively dominant at large food sources "carcasses^ M[ Jones\ unpublished data# and as the largest car! nivore in the area\ are likely to experience lower pre! dation pressure than the two quoll species[ The high diet overlap and interference experienced by the spec! ies with very low density\ suggestive of competition\ and the high numerical and behavioural dominance of the largest species in the assemblage support Brown + Maurer s "0875# and Cotgreave s "0882# ideas that competitive dominance may be more important than energetic equivalence in determining relationships between body size and abundance in local assemblages of animals[ Acknowledgements I thank Chris Johnson\ Michael Stoddart\ Chris Dick! man\ Hans Kruuk and Tamar Dayan for helpful com! ments provided when reviewing drafts of this manu! script[ Tim Kingston "Queen Victoria Museum and Art Gallery#\ Alastair Richardson\ Roy Swain and Sue Jones "all Zoology Department\ University of Tasmania# assisted with identi_cations\ and reference specimens of earthworms\ crustaceans and reptiles[ Trapping for dietary information was carried out under permits from the Tasmanian Parks and Wildlife Service and the University of Tasmania Animal Ethics Committee[ I am grateful to the Parks and Wildlife sta} at Cradle Mountain who assisted me in innumer! able ways[ This study was made possible by grants from the National Geographic Society "to Professor D[ M[ Stoddart and M[ Jones#\ the Ingram Trust "to M[ Jones# and an Australian Postgraduate Award to M[ Jones[ References Belbin\ L[ "0882# PATN Pattern analysis package\ 2[50[ Divi! sion of Wildlife and Ecology\ CSIRO Australia\ Canberra[ Blackburn\ T[M[\ Harvey\ P[H[ + Pagel\ M[D[ "0889# Species number\ population density and body size relationships in natural communities[ Ecology\ 48\ 224Ð 234[ Bradbury\ K[ "0866# Identi_cation of earthworms in mam! malian scats[ Journal of Zoology\ London\ 072\ 442Ð444[ Brown\ J[H[ + Maurer\ B[A[ "0875# Body size\ ecological dominance and Cope s rule[ Nature\ 213\ 137Ð149[ Brunner\ H[ + Coman\ B[J[ "0863# The Identi_cation of Mam! malian Hair[ Inkata Press\ Melbourne[ Bryant\ S[L[ "0877# Seasonal breeding in the eastern quoll\ Dasyurus viverrinus "Marsupialia] Dasyuridae#[ PhD thesis\ University of Tasmania[ Corbett\ L[K[ "0878# Assessing the diet of dingoes from feces] a comparison of 2 methods[ Journal of Wildlife Manage! ment\ 42\ 232Ð235[ Cotgreave\ P[ "0882# The relationship between body size and population abundance in animals[ Trends in Ecology and Evolution\ 7\ 133Ð137[ Cotgreave\ P[ + Harvey\ P[H[ "0880# Bird community struc! ture[ Nature\ 242\ 012[ Cotgreave\ P[ + Harvey\ P[H[ "0881# Relationships between body size\ abundance and phylogeny in bird communities[ Functional Ecology\ 5\ 137Ð145[ Creel\ S[ + Creel\ N[M[ "0885# Limitation of African wild dogs by competition with larger carnivores[ Conservation Biology\ 09\ 415Ð27[ CSIRO "0880# The Insects of Australia\ 1nd edn[ Melbourne University Press\ Melbourne[ Day\ M[G[ "0855# Identi_cation of hair and feather remains in the gut and faeces of stoats and weasels[ Journal of Zoology\ London\ 037\ 190Ð106[ Dickman\ C[R[ "0875# Niche compression] two tests of an hypothesis using narrowly sympatric predator species[ Australian Journal of Ecology\ 00\ 010Ð023[ Dickman\ C[R[ "0877# Body size\ prey size and community structure in insectivorous mammals[ Ecology\ 58\ 458Ð479[ Dickman\ C[R[ + Huang\ C[ "0877# The reliability of fecal analysis as a method for determining the diet of insec! tivorous mammals[ Journal of Mammalogy\ 58\ 097Ð002[ Dunham\ A[E[ "0879# An experimental study of interspeci_c competition between the iguanid lizards Sceloporus mer! riami and Urasaurus ornatus[ Ecological Monographs\ 49\ 298Ð229[ Erlinge\ S[ "0876# Why do European stoats Mustela erminea not follow Bergmann s rule< Holarctic Ecology\ 09\ 22Ð28[ Fleay\ D[F[ "0839# Breeding of the tiger cat[ Victorian Natu! ralist\ 45\ 047Ð052[ Floyd\ T[J[\ Mech\ L[D[ + Jordan\ P[A[ "0867# Relating wolf scat content to prey consumed[ Journal of Wildlife Management\ 31\ 417Ð421[ Fortin\ M[ + Gurevitch\ J[ "0882# Mantel Tests] Spatial structure in _eld experiments[ Design and Analysis of Eco! logical Experiments "S[M[ Scheiner + J[ Gurevitch\ eds#\ pp[ 231Ð248[ Chapman and Hall\ New York[ Godsell\ J[ "0872# Ecology of the eastern quoll Dasyurus viv! errinus "Dasyuridae] Marsupialia#[ PhD thesis\ Australian National University[ Green\ B[ + Eberhard\ I[ "0872# Water and sodium intake\