OUP user menu

Impact of anthropogenic habitat fragmentation on population health in a small, carnivorous marsupial

Christopher P. Johnstone , Richard D. Reina , Alan Lill
DOI: http://dx.doi.org/10.1644/10-MAMM-A-034.1 1332-1341 First published online: 16 December 2010


Habitat fragmentation is a major cause of population reduction and loss, and increasing evidence suggests that effects of fragmentation on populations vary as a function of the life history and autecology of species. We investigated the effect of anthropogenic habitat fragmentation on several indicators of population health in the agile antechinus (Antechinus agilis), a small Australian marsupial with an unusual life history. We examined relative abundance, body condition (mass-size residuals [MSR]), and ectoparasite load. Abundance was 2.3-fold higher in continuous than in fragmented Eucalyptus forest, and ectoparasite loads were higher in fragmented than in continuous forest in March, April, May, and July, but not in June or August. Unexpectedly, MSR also was higher in what would usually be considered the less-favorable fragmented habitat than in continuous forest (means: fragmented: +0.6 g; continuous: −0.9 g). Other body-condition indicators did not differ consistently between fragmented and continuous forest populations. Results suggest that an apparently common and secure small mammal species could be declining in anthropogenically fragmented and degraded habitat. Although habitat fragmentation has been associated with nutritional stress in vertebrates, food availability probably was not contributing to the lower abundance of agile antechinus in habitat fragments. The findings indicate that care is needed when generalized expectations about the response of a species to anthropogenic habitat fragmentation are used to inform conservation management.

Key words
  • abundance
  • Antechinus
  • anthropogenic effects
  • body condition
  • degradation
  • ectoparasite
  • fragmentation

Ecological and conservation biology studies on small mammals have used various measures of population well-being (health), including population density (Johnson 2007), individual growth rates (Karels et al. 2000), body condition indices (Schulte-Hostedde et al. 2005), parasite loads (Barnard et al. 2003), wound or injury occurrence (Dunford 1977; Singleton 1989), and percentage of females lactating or successfully weaning young (Karels et al. 2000). These indicators are typically assumed to be correlated with physical environmental factors, most commonly habitat quality (Johnson 2007). When examining vertebrate populations in multiple sites with differing levels of disturbance, fragmentation, or degradation (Chapman et al. 2006; Keesing 1998; Lada et al. 2007; Nupp and Swihart 1996; Suorsa et al. 2004), measuring several independent indicators rather than one indicator of population health can provide a more comprehensive and informative description of the response of a species to disturbance.

Multiple factors can contribute to population decline in fragmented habitat, including isolation, habitat degradation, the creation of novel ecotones, an increase in anthropogenic disturbance (e.g., grazing by livestock and hunting), invasion by generalist species, stochastic threats, and Allee effects (i.e., where an inherent positive relationship exists between population density and per capita growth rate—Ewers and Didham 2005; Fischer and Lindenmayer 2007; Hobbs 2001; Saunders et al. 1999). Increasing evidence suggests that effects of fragmentation on population health vary as a function of life history, autecology, diet (specialists are more severely affected than generalists), and body size (larger taxa are more severely affected than smaller taxa—Debinski and Holt 2000; Ewers and Didham 2005; Turner 1996), and therefore the use of single indicator or model species to gauge community well-being in complex natural ecosystems is contentious (Carignan and Villard 2002; Lindenmayer 1999).

Life history in the genus Antechinus is very unusual, involving a rare case of mammalian semelparity (Braithwaite and Lee 1979). An annual synchronized female estrus is followed by stress-hormone-mediated senescence and mortality of all males in late winter (Barnett 1973; Braithwaite and Lee 1979), leaving populations composed solely of pregnant females. The breeding rut is competitive, with larger males experiencing better breeding success than smaller ones, and multiple paternity within litters is common (Kraaijeveld-Smit et al. 2002, 2003). Only a small proportion of females live to breed in a 2nd year (Wood 1970).

The nocturnal agile antechinus (Antechinus agilis) is native to southeastern Australia (Dickman et al. 1998; Sumner and Dickman 1998) and restricted to native Eucalyptus forest for its foraging, nesting, and breeding (Banks et al. 2005b; Shimmin et al. 2002; Sumner and Dickman 1998). The species was considered to be part of the brown antechinus (A. stuartii) species complex until 1998 (Dickman et al. 1998), and some earlier studies of A. stuartii apparently were conducted on A. agilis. Agile and brown antechinus are unusual in that these species appear to use 2 distinct ranges, a foraging range (<3 ha) and a social range (<5 ha—Cockburn and Lazenby-Cohen 1992; Lazenby-Cohen and Cockburn 1991). Individuals will move more than 500 m outside their foraging range to visit communal nests, and the smaller foraging and larger social ranges do not always overlap (Lazenby-Cohen and Cockburn 1991). This has led to the suggestion that the agile and closely related brown antechinus are unique among mammals in that individuals use a larger habitat area for social interaction than for foraging (Lazenby-Cohen and Cockburn 1991).

The presence of agile antechinus in southeastern Australia presents an opportunity to study a species with a well-documented autecology that possesses many traits (e.g., restriction to forest, reluctance to cross the surrounding matrix, and a specialized life history) usually considered to reduce fitness in circumstances of anthropogenic habitat fragmentation (Turner 1996). Although populations of A. agilis do persist in some anthropogenically fragmented forests (Bennett 1990), the species is generally thought to be negatively affected by habitat fragmentation (Banks et al. 2005a; Bennett 1990). We compared measures of population density (relative abundance), lipid reserves (body condition indices), and ectoparasite load (simplified parasite counts) in populations living in anthropogenically fragmented and relatively undisturbed, continuous Eucalyptus forest (fragments and continuous forest always were separated by >2 km of agricultural fields). We predicted that where native tree cover was anthropogenically fragmented rather than continuous, agile antechinus would have a lower relative abundance, poorer body condition, and higher ectoparasite load.

Materials and Methods

Study area, sites, and population.—The study was conducted in 2007 and 2008 in South Gippsland, Victoria, Australia (Fig. 1). The region has one of the longest histories of European settlement and agriculture in mainland Australia. Forest clearing for agriculture, gold mining, and timber production was triggered by an 1869 Lands Act allowing settlement, and all the fragments in our study have been isolated by anthropogenic land clearing for >50 years. The matrix surrounding all Eucalyptus fragments in this study was pasture, primarily grazed by dairy and beef cattle, but with some sheep farming. Species relaxation, the gradual loss of species after fragmentation of habitat, still could be occurring (Saunders et al. 1999), but we have no good evidence for or against this assumption. Fire is an important part of the ecology of the study area. We used available fire history information and local research to choose study sites where there had been no fires for >30 years, because recent fires could negatively affect abundance in the study species (Catling et al. 1998). However, evidence suggests that abundances of agile antechinus are unaffected by fire history where time since fire is >20 years (Catling et al. 1998; Claridge et al. 2008), so we consider the >30-year criterion to be sufficient to avoid any confounding effects.

Fig. 1

Study region in South Gippsland, southeastern Australia. White areas represent cleared agricultural land, and gray areas represent native tree cover. Approximate locations of fragment study sites are indicated by white boxes (□). Comparison pseudofragment sites were situated within the 2 areas delineated by a heavy line and labeled C. Map is based on the Department of Sustainability and Environment interactive forest-explorer online maps (Forest-Explorer Online maps, http://www.dse.vic.gov.au).

Livetrapping was carried out from April to August 2007 and March to August 2008. Teat number varies among female agile antechinus and is known to be genetically regulated and associated with local environmental conditions (Beckman et al. 2007; Cockburn et al. 1983). However, homogeneity of teat number in our study area (all 8 teats—C. P. Johnstone, pers. obs.) strongly suggested that all sites were environmentally similar prior to fragmentation of the native vegetation by land clearing for agriculture.

A control pseudofragment in an area of relatively similar, undisturbed, continuous forest was matched for size and shape to each fragment (Mac Nally and Bennett 1997; Fig. 2). Pseudofragments were nonoverlapping but were otherwise randomly distributed in suitable continuous forest. Trapping in a fragment was immediately followed by trapping in the paired pseudofragment. All forest fragments and pseudofragments studied were occupied by agile antechinus, ensuring that estimates of relative abundance were not confounded by including unoccupied sites. The 30 fragments were 4.8–293.6 ha in area and were dispersed in an anthropogenically disturbed agricultural landscape 2.1–38.6 km from any area of continuous forest (defined as > 1,000 ha of continuous native tree cover; Fig. 1). Fragments were within an area bounded by the coordinates 38°35′26”S, 145°41′41”E; 38°21′55”S, 146°06′10”E; 38°37′19”S, 146°28′20”E; and 38°45′12”S, 146°01′33”E. All pseudofragments were situated in an area bounded by the coordinates 38°28′03”S, 146°18′45”E; 38°36′50”S, 146°32′40”E; 38°34′35”S, 146°19′12”E; and 38°32′41”S, 146°16′26”E.

Fig. 2

Hypothetical example of a fragment-pseudofragment design. Trapping in a fragment was followed immediately by trapping in the paired pseudofragment. In landscape ecology studies this design (or similar) is needed to separate the effects of habitat fragmentation from habitat loss (Mac Nally and Bennett 1997).

Possible spatial autocorrelation of study sites is a concern in research on the effects of fragmentation. The underlying problem is that where landscapes have a level of anthropogenic fragmentation that is suitable for study, typically only a few areas of continuous habitat suitable for control sites remain. Consequently, comparative control sites often must be restricted to a few large blocks of habitat (usually reserves). In the present investigation fragments were always >500 m from another study fragment, and pseudofragments always had >1 km of native forest buffer between them. Although we cannot completely discount the possibility that spatial autocorrelation confounded some aspects of our results, an examination of site longitudes and latitudes suggested that fragments and pseudofragments probably were not substantially different in their spatial distribution (general linear model: latitude: F1,58 = 1.58, P = 0.217; longitude: F1,58 = 0.56, P = 0.459). Some agile antechinus could have dispersed from other neighboring fragments or continuous forest to the study sites prior to sampling, but sampling commenced approximately 2 months after the annual male-biased dispersal (Cockburn et al. 1985) in January-February, which should have allowed time for immigrants to acclimate to local conditions.

Native tree cover was identified using online native vegetation cover maps (1:75,000) available from the Victorian Department of Sustainability and Environment (Forest-Explorer Online maps, http://www.dse.vic.gov.au/). Habitat similarity among study sites was achieved by restricting sites to stands of forest composed of the 3 ecological vegetation classes (Davies et al. 2002), Lowland Forest, Wet or Damp Forests (Wet), and Wet or Damp Forests (Damp). Most sites contained a mixture of the first 2 ecological vegetation classes, but some also contained small amounts of Riparian Forests or Woodlands or Rainforests.

Trapping protocols.—Trapping and data collection were approved by Monash University Biological Sciences Animal Ethics Committee (permits BSCI/2008/03 and BSCI/2006/05) and conducted under Department of Sustainability and Environment permit 10003798. All research involving live animals followed guidelines approved by the American Society of Mammalogists (Gannon et al. 2007). Trapping was conducted with weather-proofed Elliott live traps (Elliott Scientific, Upwey, Victoria) baited with a mixture of rolled oats, peanut butter, water, and artificial vanilla essence. A plastic tube wrapped in bedding provided an additional insulated refuge for trapped animals. Trapping was conducted overnight for 3 successive days at each fragment and pseudofragment. Trapping in a pseudofragment was started the day after trapping in its paired fragment to incorporate a temporal control in the study. All trapping in a fragment-pseudofragment pair was completed in 8 days (2× trap-setting days and 6× trap nights). In each fragment 1 trapping grid was placed <60 m from the edge and a 2nd in the interior (>80 m from the edge). The spatial positioning of trapping grids in a fragment was replicated as closely as possible in its paired pseudofragment.

Relative abundance.—Because of differences in morphology, physiology, and behavior (Barnett 1973; Cockburn and Lazenby-Cohen 1992; Marlow 1961), males and females were treated as separate populations for all measured variables. In pseudofragments trapping rates of agile antechinus and mammalian bycatch were much higher than expected (up to 100% of traps were occupied per night), and therefore trapping effort was scaled down to reduce trap mortality. In each fragment 2 trapping grids of 21 traps each were used, both being arranged in 3 lines of 7 (20 × 20-m spacing). Pseudofragment grids had 9 traps in 3 lines of 3 (60 × 20-m spacing). Thus, pseudofragment grids covered the same area as fragment grids but had fewer traps. Varying trapping effort can confound comparative measures of abundance (Lada et al. 2007), but we used a binomial distribution of trapping rate (Lada et al. 2007) to allow for the discrepancy. A trapping success was defined as the capture of 1 agile antechinus in a trap at a site on any given night, but the binomial variable was based only on the 1st night of trapping at a site, with each trap set being defined as a trial (i.e., the number of trials was 42 for fragments and 18 for pseudofragments). By using data only from the 1st night, the possibility that learning could confound results was reduced. Binomial variables were calculated with the Excel (Microsoft, Redmond, Washington) function BINOMDIST and used a 0.5 probability of success. Cumulative distribution functions were used rather than probability mass functions. The use of a binomial variable gave more weight to sites that had a greater trapping effort (Lada et al. 2007); that is, the fragments. Trapping rate (TR) is the binomial variable, and it is considered to be a measure of relative abundance (a surrogate for population density). We predicted that the relative abundance of agile antechinus would be greater in undisturbed pseudofragments than in more-degraded fragments.

Body condition indices.—All captured individuals were sexed by visual inspection. Typically, 6 individuals were sampled for body mass, morphometries, and ectoparasite variable (s) in each grid over the 3 trapping days, making a total of 12 individuals per study site. The measurements taken were mass (g to the nearest 0.5 g) and nose-vent length (in mm to the nearest 0.1 mm). All measurements were taken by the same researcher. Repeated measurements are important for reducing the incidence of type II error when constructing small mammal body condition indices (Blackwell et al. 2006). However, when taking body measurements on individuals, we took blood samples for a related research project. Because handling was already lengthy (up to 15 min) and potentially stressful, we decided that it was preferable to reduce handling time by taking single measurements. This meant that comparing body condition indices in the 2 habitat types entailed a greater risk of obtaining a false-negative result. However, sample sizes of ≤12 individuals per site and the use of site means in the analyses should have helped to mitigate the error associated with using single measurements. Body condition index was calculated as the residuals of body mass as a function of nose-vent length (mass-size residuals [MSR] in g—Schulte-Hostedde et al. 2005). Ordinary least-squares regressions were used to generate residuals for MSR (Schulte-Hostedde et al. 2005).

Ectoparasite load.—Permanently attached mites (order Acariformes) and ticks (superorder Parasitiformes) on both ears of each antechinus were counted, but we did not adjust for intensity of parasite infection (i.e., larger parasitiforms, such as Ixodes spp., were not weighted differently than smaller acariform species). Restricting the count to the ears reduced handling time and lessened the risk that parasites might be obscured by fur and overlooked.

Data analysis.—Data were analyzed using R 2.8.1 (R Development Core Team 2010). They were checked for normality and homoscedasticity. The binomial abundance variable, TR, was log10-transformed (logTR) to achieve normality, but no other transformations were needed. Analyses of variance (ANOVAs) were applied to predictor (environmental) and response (abundance, body condition index, and ectoparasite load) variables to identify significant relationships. ANOVA predictor factors were sampling year (2007 and 2008), month (months were allowed to vary from calendar months by 1–3 days, so fragment-pseudofragment pairs were kept in the same month), sex (2 levels), and fragmentation (2 levels). Where the assumption of homogeneity of regression slope of linear models was violated, data were separated so that fragmentation, the factor of primary interest, could be examined (i.e., a post hoc simple main effects test—Engqvist 2005; Quinn and Keough 2002). This treatment of interaction terms in linear model analyses avoids using the more-involved Wilcox modification of the Johnson-Neyman procedure and is preferable if data can be partitioned easily (Engqvist 2005). Simple main effect testing has the usual problems associated with multiple testing, but we used a sequential Bonferroni adjustment of P-values to account for this where needed (Quinn and Keough 2002). We followed the recommendation of Quinn and Keough (2002) that simple main effects tests use the residual sum of squares and degrees of freedom from the full ANOVA. Pseudoreplication was avoided by using site means in analyses, unless otherwise stated. Values are shown as means and standard errors (SEs). Where means for months are given, they are based on all sites sampled in the particular month. Simple main effects testing by month was necessary to examine possible seasonal differences in ectoparasite counts. However, only 2–4 fragments were sampled in a given month. Thus, to obtain a reasonable estimate as to whether biologically significant differences between fragment and continuous forest populations could have occurred in a given month, analyses of ectoparasite data at this level used individual animals as the sampling unit.


In 2007, 181 male and 196 female adult agile antechinus were captured in 15 forest fragments and 248 males and 146 females were captured in 15 pseudofragments. The corresponding numbers for 2008 were 178 males and 183 females and 201 males and 93 females, respectively. In 2007, 73 males and 60 females and 110 males and 52 females were sampled for morphometries, body mass, and ectoparasites in fragments and pseudofragments, respectively. The corresponding numbers for 2008 were 70 males and 55 females and 81 males and 50 females, respectively (Table 1).

View this table:
Table 1

Means, SEs, and ranges for mass, morphometric, trapping-rate, and ectoparasite variables for Antechinus agilis in fragmented and pseudofragmented forest. Means and SEs are shown for trapping rate (unmodified %TR) averaged over 3 trapping nights, binomial distribution of trapping rate (logTR binomial, 1st night only), mass (g), nose-vent length (NV), and mass-size residuals (MSR). Ectoparasite values do not include 1 site pair that was excluded from analyses due to ectoparasite count in the fragmented part of the pair being an outlier (an order of magnitude higher than the mean for all sites).

Fragment (X̄ ± SE)Pseudofragment (X– ± SE)
Unmodified %TR
Females 20070.11 ± 0.030.22 ± 0.04
Females 20080.08 ± 0.020.11 ± 0.02
Males 20070.10 ± 0.020.26 ± 0.04
Males 20080.08 ± 0.010.21 ± 0.03
logTR binomial
Females 2007−7.36 ± 1.00−2.21 ± 0.38
Females 2008−8.86 ± 0.85−3.46 ± 0.31
Males 2007−6.79 ± 0.71−1.78 ± 0.31
Males 2008−8.25 ± 0.47−2.17 ± 0.32
Mass (g)
Females 200716.83 ± 0.5816.88 ± 0.65
Females 200816.82 ± 0.5516.71 ± 0.56
Males 200725.19 ± 1.6520.88 ± 0.65
Males 200825.23 ± 1.0924.15 ± 0.73
NV (mm)
Females 200779.43 ± 1.4880.80 ± 1.46
Females 200879.22 ± 0.8979.50 ± 1.06
Males 200788.21 ± 1.7085.05 ± 1.00
Males 200887.98 ± 0.9688.99 ± 0.95
MSR (g)
Females 2007−0.65 ± 0.65−1.53 ± 0.90
Females 2008−0.52 ± 0.51−0.82 ± 0.41
Males 20071.76 ± 1.33−0.40 ± 0.56
Males 20081.96 ± 0.690.20 ± 0.39
Ectoparasite count
Females 20072.70 ± 0.780.49 ± 0.20
Females 20086.24 ± 3.933.06 ± 1.42
Males 20072.83 ± 0.840.54 ± 0.10
Males 20086.37 ± 3.702.47 ± 0.80

Relative abundance.—Relative abundance was significantly smaller in fragments (TR < 0.001 ± < 0.001) than in pseudofragments (TR = 0.077 ± 0.023; Tables 1 and 2). We observed no difference in logTR between the sexes, but logTR was significantly greater in 2007 (TR = 0.058 ± 0.022) than 2008 (TR = 0.022 ± 0.011). The interaction term YEAR × MONTH was significant in this analysis, but when the data were partitioned by year, logTR was not affected by the month in which sampling occurred in either 2007 (F1,108 = 2.70, P = 0.103) or 2008 (F1,108 = 2.79, P = 0.097).

View this table:
Table 2

ANOVA results for population well-being (health) indicators for Antechinus agilis as a function of the predictor variables Fragmentation (fragment or pseudofragment), Sampling Year (2007 or 2008), Sex, and Month. Values have been averaged by study site and sex, and the sexes have been treated as separate populations. All nonsignificant interactions were removed from the final analyses. All degrees of freedom are the same for all factors within a given F-test. An asterisk (*) indicates P < 0.05. TR = trapping rate; MSR = mass−size residuals.

FRAGMENTATION188.49< 0.001
YEAR × MONTH8.670.004*
SEX13.59< 0.001*
YEAR × MONTH7.670.007*
Ectoparasite count

Body condition: MSRs.—Mean MSR was greater in fragments (0.62 ± 0.44 g) than in pseudofragments (−0.61 ± 0.29 g; Tables 1 and 2). It also was greater for males (0.87 ± 0.41 g) than females (−0.91 ± 0.31 g), but no significant differences were found between years. The interaction term YEAR × MONTH was significant in this analysis; when the years were analyzed separately, MSR increased in 2007 from −1.61 ± 0.59 g (April) to 3.16 ± 4.32 g (August; F1,108 = 16.64, P < 0.001) but did not vary during the sampling period in 2008 (F1,108 = 0.81, P = 0.370). Although our results imply a negative relationship between relative abundance and body condition in free-living agile antechinus, this was significant only for males (females: n = 55, F1,108 = 1.57, P = 0.214; males: n = 59, F1,108 = 9.67, P = 0.004).

Ectoparasite load.—Although mean ectoparasite counts per individual were higher in fragments (4.60 ± 1.41) than in pseudofragments (1.68 ± 0.44; Table 1), this finding was confounded by a significant FRAGMENTATION × MONTH interaction term (Table 2). When the 2 habitat types were analyzed separately, pseudofragment populations showed no seasonal variation in ectoparasite counts (F1,108 = 0.22, P = 0.639), whereas ectoparasite counts in fragment populations decreased from March (13.2 ± 7.2) to August (1.7 ± 0.4; F1,108 = 19.65, P < 0.001). Partitioning data by month revealed a possible trend where fragment populations had higher ectoparasite infection counts than pseudofragment populations in March and May, but ectoparasite counts were the same or similar in both habitat types in April, June, July, and August (March: F1,22 = 11.13, P = 0.018; April: F1,65 = 5.18, P = 0.104; May: F1,146 = 7.90, P = 0.030; June: F1,135 = 3.75, P = 0.110; July: F1,126 = 4.52, P = 0.105; August: F1,54 = 1.79, P = 0.186).


Consistent with our initial prediction, relative abundance of agile antechinus was significantly lower in anthropogenically fragmented Eucalyptus forest than in undisturbed, continuous forest. This difference was evident for both sexes, at all sampling times, and in both study years. The likelihood is that direct (e.g., barriers to dispersal) or indirect (e.g., habitat degradation) factors, or both, were causing lower recruitment, or poorer survival, or both, in anthropogenically fragmented habitat (Fig. 3). We do not know whether the difference in abundance between the 2 habitat types was stable, or if a gradual, ongoing decline in abundance was occurring in fragments. Because extinction of vertebrate populations after anthropogenic fragmentation of habitat can take years or decades (Diamond et al. 1987; Pimm et al. 1993), it can be particularly difficult to detect in an apparently common and secure species (i.e., one that has a broad distribution across multiple sites where it is often locally common) such as the agile antechinus (Menkhorst and Knight 2004). It is thus a serious and often overlooked conservation concern.

Fig. 3

Conceptual flow diagram of possible relationships among environmental variables, life-history traits, and population well-being (health) indicators for Antechinus agilis. Positive relationships are indicated by an arrow. Negative relationships are indicated by a line. The line at the left margin connects 2 sets of degradation effects.

Two potentially confounding aspects relate to the finding that relative abundance of agile antechinus was lower in fragments than in pseudofragments. First, we had no reason to anticipate the trap saturation that occurred in the pseudofragments, which meant that population density was probably underestimated in those sites. This saturation could have led to either of 2 false conclusions: that fragment and pseudofragment abundances were similar, when pseudofragment abundance was actually greater; or that abundance was greater in fragments, when it was actually similar in both types of site. However, abundance of agile antechinus was greater in pseudofragments than in fragments as initially predicted, so the only possible effect of trap saturation in the pseudofragments was simply a reduction in the apparent magnitude of this detected disparity.

Second and more equivocal, where relative abundance is estimated from trapping rates, it is not always clear to what extent behavior is affecting trapping rate in the target species. Trapping rates of agile antechinus potentially could have been affected by differences in movement rates in the different types of site (Diffendorfer et al. 1995). Anthropogenic habitat fragmentation is often associated with habitat degradation (Knight and Fox 2000; Turner 1996), as was the case in our study area. In particular, fragments on average probably had a less-complex habitat structure than did pseudofragments (C. P. Johnstone, pers. obs.). If this resulted in agile antechinus moving less frequently or for shorter distances in fragments than in pseudofragments because the shrub and understory cover was sparser, it would have reduced their probability of encountering a trap. Differences in feeding motivation also could have affected trapping rates. Thus, if food availability limited antechinus population density in pseudofragments, individuals would have been very likely to explore a baited trap. However, if other factors limited population density in fragments, as seems likely, individuals might have been less likely to explore the novel foraging resource provided by a baited trap. Clearly, further research is needed to determine whether such behavioral effects on the trapping rate reduce the resolution of this method of determining the relative abundances of agile antechinus populations in different habitats.

The measure of body condition, MSR (Schulte-Hostedde et al. 2005), also had a consistent relationship with fragmentation, but it was contrary to our initial prediction. Counterintuitively, agile antechinus had apparently greater lipid stores where habitat was fragmented anthropogenically and degraded. Thus, nutritional stress probably was not contributing to the lower relative abundance of agile antechinus in forest fragments. This is not consistent with field studies of agile and other antechinus species that have reported consistent, positive relationships between population density and indicators of food resource abundance, particularly leaf-litter depth and woody debris density (Bennett 1993; Kelly and Bennett 2008; Knight and Fox 2000; Lada et al. 2007). Moreover, Mac Nally and Horrocks (2002) conducted a large-scale experimental manipulation of woody debris load in woodland sites and showed that the relationship between log density and the abundance of yellow-footed antechinus (A. flavipes) appeared to be causal.

Leaf litter and fallen timber also add to habitat structure (Garden et al. 2007; Mac Nally and Horrocks 2002), so conceivably their effects on relative abundance of antechinus are related more to predator avoidance (logs—Stokes et al. 2004) or availability of nesting material (leaf litter) or nesting sites (logs—Cockburn and Lazenby-Cohen 1992), or both, than to food resources. We recorded an index of leaf-litter depth and spatial extent (on a scale of 1–5) and the density of logs (per 400 m2) in our study sites. The observed trend was toward a lower leaf-litter index in fragments than in pseudofragments (mean leaf-litter index: fragments = 2.7 ± 0.1, pseudofragments = 4.1 ± 0.1) but a higher log density in fragments (C. P. Johnstone, pers. obs.), perhaps due to edge effects and windfall (mean log density: fragments = 13.1 ± 1.7, pseudofragments = 9.4 ± 1.3). Thus, the lower relative abundance of agile antechinus in fragments could have been related to less leaf litter than in pseudofragments, but if log density had any positive effect on abundance of agile antechinus, it presumably was less influential than other environmental variables. The higher estimated fat reserves of agile antechinus in fragments could have reflected the greater abundance of woody debris in that type of site; manipulation of debris loads might reveal whether a causal relationship is involved (Mac Nally and Horrocks 2002).

Ectoparasite counts for agile antechinus differed among months. In March and May the results agreed with our initial prediction that individuals in fragments would have higher ectoparasite loads than those in continuous forest. Although this would not be surprising if members of fragment populations were in poorer condition (Beldomenico et al. 2008), they actually had a higher mean MSR than those in pseudofragments, casting some doubt on differences in body condition as an explanation for the disparity in parasite load. Environmental quality possibly was more impaired by habitat degradation or ecological changes in fragments than in pseudofragments (e.g., fewer tree hollows for nesting, poorer quality nesting material, more invasive exotics forming larger parasite reservoirs, or a combination of these), such that transmission of parasites was more likely. Regardless, the link between habitat fragmentation and ectoparasite load was not consistent, because fragment and pseudofragment loads were the same in April, June, July, and August. This was apparently due to loads in fragments (but not pseudofragments) decreasing during the trapping season in both years. Why this decrease occurred is unknown.

Negative population density × body condition (PD × BC) relationships have been observed in small mammals during population cycling in subarctic and semidesert landscapes (Korpimäki et al. 2004). However, agile antechinus has been studied extensively, and no suggestion of population cycling exists in the literature. In stable or temperate environments small mammal PD × BC relationships are typically positive or neutral (Keesing 1998; McGuire et al. 1993; Nupp and Swihart 1998; Sale and Arnould 2009; Sale et al. 2009; but see Lada et al. 2008). Our study area conformed with this stable environment profile, because long-term mean annual rainfall was 800–1,200 mm and mean daily temperature ranged from a minimum of 9–12°C to a maximum of just 18–21°C (30-year means—Australian Bureau of Meteorology 2009).

Experimental studies on small mammals using artificially high population densities usually have attributed negative PD × BC relationships to intraspecific competition for limiting resources needed for survival and maintenance (Ostfeld and Canham 1995; Warnock 1965). However, we observed very few wounds on trapped agile antechinus (C. P. Johnstone, pers. obs.) and have no evidence that strong intraspecific fighting affected body condition. Nonetheless, given what is known about semelparous breeding behavior of agile and brown antechinus and metabolic demands during male lek behavior (Lazenby-Cohen and Cockburn 1988; Woollard 1971), it is possible that intraspecific competition for food could have contributed to the negative PD × BC relationship among males. Agile antechinus adult mass can vary considerably within a population, particularly in males (Table 1). In our study adult male mass varied from 13 g (10 June) to 41 g (26 July), a 3.2-fold difference, and adult female mass from 12 g (25 June) to 26.5 g (24 June), a 2.2-fold difference (C. P. Johnstone, pers. obs.). Conceivably, individuals could attain quite different prebreeding mass, depending on food availability (Dickman 1989). Females have sperm-storage crypts (Shimmin et al. 1999), and mating order influences paternity success, with larger males and those that mate closer to ovulation siring more offspring (Kraaijeveld-Smit et al. 2002, 2003). A clear fitness advantage seems to exist for a male to be larger at the outset of the breeding season (due to both lipid and protein reserves). Sufficient metabolic fuel reserves could allow a male to persist with reproductive effort when smaller competitor males are in the early stages of the negative nitrogen balance, development of digestive tract lesions, and escalating parasite loads that eventually cause postbreeding male mortality (Barker et al. 1978; Beveridge and Barker 1976; Woollard 1971). Strong competition for food could result. Brown and agile antechinus occupy defined foraging ranges (Banks et al. 2005a; Lazenby-Cohen and Cockburn 1991), but the possibility that males actually defend foraging territories has not been suggested in the literature and warrants investigation.

Negative PD × BC relationship also could be explained by differences in physical environmental variables in fragmented and continuous forest. Agile antechinus are seldom or never caught in the cleared matrix surrounding occupied Eucalyptus forest fragments (Bennett 1990), but matrix arthropods could have been wind-blown into fragments, providing a superabundance of prey (spiders in hedgerows—Landis et al. 2000). That wind-carried nutrients, trace elements, or subsidies of detritus from the surrounding matrix could affect native vegetation reserves has been advanced by Scott et al. (1999), although in the context of matrix subsidies generally, this possibility has not been widely discussed or studied. If such a subsidy benefited forest-fragment populations, food presumably was not the most important limiting resource for population density of agile antechinus, otherwise fragment populations would have reached or exceeded the densities observed in continuous forest.

Another possible explanation for the significant PD × BC relationship was that higher levels of interspecific competition in continuous forest reduced the food resources available to agile antechinus. Certainly, 2 native, small mammals, the dusky antechinus (A. swainsonii; 38–170 g) and the bush rat (Rattus fuscipes; 50–225 g), occurred at much higher relative abundances in pseudofragments than in fragments (logTR of A. swainsonii in fragments = −6.00 ± 0.08, in pseudofragments = −2.41 ± 0.08; logTR of R. fuscipes in fragments = −5.37 ± 0.15, in pseudofragments = −1.68 ± 0.10–C. P. Johnstone, pers. obs.). Both species are competitively dominant over agile antechinus (Banks and Dickman 2000; Dickman 1986), and it is possible that in pseudofragments these competitors reduced foraging opportunities for agile antechinus. Another possible food competitor, the ground-foraging, insectivorous southern brown bandicoot (Isoodon obesulus), could have been present in the pseudofragments. Our traps were not large enough to capture adults of this species, but significantly we did not capture any juveniles and saw no indirect evidence (e.g., conical diggings) of bandicoots, so it seems unlikely that they were competing with agile antechinus at any sites. Also, some ground-foraging bird species could have occurred at higher densities in continuous than in fragmented forest and reduced the overall availability of invertebrate leaf-litter prey, but we have no data to evaluate this possibility.

Factors other than food availability, such as higher predation rates or fewer nest sites in fragments than in pseudofragments, likely limit population density in fragmented forest. Exotic terrestrial predators occur at higher densities in fragmented than in continuous forest in Australia—in particular, European red foxes (Vulpes vulpes) and feral cats (Felis catusMay and Norton 1996). More-intense browsing pressure from livestock and consequently a sparser shrub understory in fragmented than in continuous forest (Knight and Fox 2000) could enhance both feral terrestrial mammal and native avian predation pressure on antechinus by removing foraging cover (Hobbs 2001; Knight and Fox 2000; Stokes et al. 2004). Riskier male than female foraging behavior is frequently observed in lek-breeding mammals (Ruckstuhl and Neuhaus 2000) such as the agile antechinus (Lazenby-Cohen and Cockburn 1988); for example, males have a greater tendency to forage away from shrub cover. High levels of predation could have selectively removed agile antechinus in poorer condition from the fragment populations; however, if this was the only underlying cause of the higher MSR in fragment than in continuous forest populations, a shift in the mean, but not the range, of mass and MSR should have occurred, but this was not observed.

The findings indicate that care is needed when generalized expectations about the response of a species to anthropogenic habitat disturbance are used in conservation theory or practice. Anthropogenic habitat fragmentation or degradation can have unexpected effects on a species. Where relationships of populations and their environments are examined, several independent indicators of health are likely to produce more-comprehensive and informative results.


This research was made possible by funding from the Holsworth Wildlife Fund and access kindly granted by private landowners throughout the South Gippsland region. Field accommodation was provided by Parks Victoria (with particular thanks to M. Hoskins), J. and S. Bell, G. and J. Wallis, D. and M. Hook, and D. Farrar. We also thank C. Rankin for access to South Gippsland Shire council reserves. The support, cooperation, and enthusiasm of many individuals and groups helped to facilitate this project, notably the South Gippsland Conservation Society, Venus Bay Landcare, and Anders Inlet Landcare. Special thanks go to K. Achkar-Kerbaji for assistance with fieldwork. Two anonymous referees provided valuable feedback on the manuscript.


  • Associate Editor was Madan K. Oli.

Literature Cited

View Abstract