1

Journal of Mammalogy, 89(2): 398–407, 2008

Responses of a carnivorous marsupial (Antechinus flavipes) to local habitat factors in two

types of forests in south-eastern Australia

Hania Lada, Ralph Mac Nally*, and Andrea C. Taylor

Australian Centre for Biodiversity: Analysis, Policy and Management, School of Biological

Sciences, Monash University, Victoria, 3800, Australia

*Correspondent:

Ralph.MacNally@sci.monash.edu.au

Running head: Response of antechinus to habitat

2

ABSTRACT

Ecosystems around the world have been degraded or destroyed by human activities, including

regulation of river flows, clearance of vegetation and removal of fallen timber. In

southeastern Australia much of the original vegetation was converted to farmland. Remaining

forests such as hilly box-ironbark and floodplain river red gum (Eucalyptus camaldulensis)

are mostly re-growth. The yellow-footed antechinus (Antechinus flavipes) inhabits both types

of forests and is the only small, native, carnivorous mammal on most floodplains in

southeastern Australia. In this region, frequency of flooding has been reduced by regulation of

river flows, which has led to decline in conditions favourable for flood-adapted terrestrial and

aquatic organisms. Here, we compared numbers of A. flavipes in box-ironbark forests, and in

river red gum forests that were deprived of floods; partially inundated with environmental

flows; flooded naturally; and watered in large, artificial floods. We found that abundance of

A. flavipes on floodplains and in box-ironbark forests increased with larger volumes of fallen

timber and with greater numbers of large, old trees. In river red gum forests, numbers of 2nd-

year females increased with proximity to flood locations. For conservation purposes, we

recommend preservation of large trees, restoration of fallen-timber on forest floor, and spring

flooding of floodplains.

Key words: Antechinus flavipes, box-ironbark, Eucalyptus camaldulensis, fallen timber,

flood, hollow, logging, Murray River, river red gum, woodload

3

INTRODUCTION

Ecosystems around the world have been degraded through anthropogenic actions (Sala

et al. 2000). Habitats have been lost due to activities such as clearance of vegetation, logging,

removal of fallen timber, and regulation of river flows (e.g., Cunningham et al. 2007; Maser

and Sedell 1994; Micklin 1988). These actions have led to decline or extinction of many

species (Sala et al. 2000; Wilcove et al. 1998). In Australia during the last 200 years, 18

species of endemic mammals have become extinct most likely because of human-induced

changes to environment (Johnson 2006). These changes principally involved introduction of

alien species, whose detrimental effects were exacerbated by degradation and loss of habitats

(Johnson 2006). Most of the native vegetation in the southeastern part of the continent has

been cleared for farming; for example, only 6% of the original tree cover remains in the

Murray Fans and Victorian Riverina bioregions (Bennett et al. 1998; Department of

Sustainability and Environment 2004). Remaining forests have been or are extensively

logged, and abundance of large, old trees has been reduced substantially (Environment

Conservation Council 1997). Many organisms rely on mature, hollow-bearing trees for shelter

and for foraging opportunities (e.g., Cameron 2006; Erickson and West 2003; Gibbons et al.

2002; Holloway and Malcolm 2007). Fallen timber also provides important ecological

resources (Harmon et al. 1986), but has been removed extensively for firewood over much of

southeastern Australia (Laven and Mac Nally 1998; Mac Nally et al. 2002a). Stripping fallen

timber from the forest floor simplifies habitat structure, which reduces shelter and foraging

opportunities (Harmon et al. 1986). Dependence on fallen timber has been shown for many

organisms (e.g., Gibb et al. 2006; Lindhe et al. 2004; Sippola et al. 2005; Waldien et al.

2006), including Australian saproxylic invertebrates (Grove 2002), and vertebrates such as the

brown treecreeper (Climacteris picumnus) (Mac Nally 2006; Mac Nally et al. 2002b) and the

4

yellow-footed antechinus (Antechinus flavipes) (Lada et al. 2007b; Mac Nally and Horrocks

2002).

In Victoria, despite massive losses of habitats through logging and fallen-timber

removal, box-ironbark forests still support some native fauna, including A. flavipes

(Environment Conservation Council 1997). A. flavipes has a 2-week, synchronized, winter

mating season followed by obligatory die-off of males, and weaning of offspring in summer

(McAllan et al. 2006; McDonald et al. 1981; Woolley 1966). The species also inhabits river

red gum (Eucalyptus camaldulensis) forests of floodplains, where in most cases, it is the only

small, native, carnivorous mammal (Menkhorst and Knight 2001). These 2 types of forests are

markedly different. Box-ironbark forests grow on dry, upland areas, whereas river red gum

forests grow on floodplains and have evolved reliance on inundations. River red gum forests

along the Murray River have been subjected to clearance of vegetation, removal of fallen

timber, and regulation of river flows causing changes in flooding regimes (Close 1990). In

160 years since the beginning of European impacts on the floodplains, frequency of floods

has been reduced, especially in the last 50 years (Close 1990; Murray-Darling Basin

Commission 2005), and much of the native vegetation has been removed (Fahey 1988;

Walker et al. 1993). There is a scarcity of large, old trees in box-ironbark forests, which were

nearly completely cleared during the 19th century gold rush (Environment Conservation

Council 1997). These forests have been logged several times since and large trees were

deliberately removed in the 1930s and 1950s for silviculture (Muir et al. 1995), so that

virtually no old-growth box-ironbark forests remain in Victoria (Environment Conservation

Council 1997).

Lada et al. (2007b) showed that on the Murray River floodplains (in 3 river red gum

forests studied each year from 2003 to 2005) abundance of A. flavipes increased with

proximity to floods, higher fallen-timber loads and the number of large trees. These findings

5

agreed with earlier studies reporting positive effects of fallen-timber loads on the abundance

of A. flavipes in 2 river red gum forests (Mac Nally and Horrocks 2002; Mac Nally et al.

2001). However, similar studies of A. flavipes have not been conducted in box-ironbark

forests, where effects of flooding are not relevant, fallen-timber loads are lower and there are

fewer large trees. Also, the responses of A. flavipes inhabiting river red gum forests on

floodplains of unregulated rivers, such as the Ovens River, remain undocumented. Mac Nally

et al. (2001) attempted to address this in relation to fallen-timber loads only, but failed to trap

A. flavipes on the Ovens floodplain in either 1998 or 1999.

Studies of abundance of mammals other than A. flavipes, before and after floods have

reported either declines or no changes (Andersen et al. 2000; Granjon et al. 2005; Jacob 2003;

Redhead 1979; Williams et al. 2001). Here, we investigated whether evidence for the positive

effects of proximity to floods on numbers of A. flavipes (Lada et al. 2007b) also are found on

floodplains experiencing natural flooding regimes as well as on floodplains inundated more

extensively than those in the study of Lada et al. (2007a). We test whether capture rates of all

individuals, males, females and 2nd-year females of A. flavipes in box-ironbark and in river

red gum forests are related to habitat variables that are of major management concern: fallen-

timber volumes, densities of large trees, and distance from floods (in river red gum forests

only). Capture rates are used here as a relative measure of animal abundances and, in the case

of 2nd-year females, survival of females beyond 1 year. We compare capture rates of A.

flavipes between nearby box-ironbark and river red gum forests, among river red gum forests

differing in current flooding regimes, and among box-ironbark forests.

MATERIALS AND METHODS

Study sites and captures of animals.—Individuals of A. flavipes were trapped once a

year in the austral autumn and winter in 2004 and 2005 at 0.25-ha sites, randomly selected from

6

100 sites within 500 m × 500 m squares, in 7 river red gum forests and in 3 box-ironbark

forests. Minimum distances between sites were 0–150 m. Box-ironbark forests are dominated

by grey box (Eucalyptus microcarpa) and red ironbark (E. sideroxylon) in northern Victoria (E.

tricarpa in central Victoria), with some red box (E. polyanthemos) and red stringybark (E.

macrorhyncha), and with herbs, shrubs, leaves and fallen timber providing groundcover. River

red gum sites (denoted RRG) were in GuttrumRRG (centred on 35°35’S, 144°04’E; n = 32),

KoondrookRRG (35°43’S, 144°15’E; n = 27), GunbowerRRG (35°43’S, 144°13’E; n = 102),

BarmahRRG (35°52’S, 145°01’E; n = 29), MillewaRRG (35°50’S, 145°00’E; n = 30), OvensRRG

(36°13’S, 146°15’E; n = 13) and Reedy LakeRRG (36°42’S, 145°05’E; n = 23) (Fig. 1).

Groundcover is provided by fallen timber and bark in all river red gum forests, along with tall

grass in OvensRRG, and shrubs and reeds in Reedy LakeRRG. Box-ironbark sites (denoted BI)

were in ChilternBI (part of Chiltern-Mt Pilot National Park) (36°08’S, 146°36’E; n = 62), in

RushworthBI (36°39’S, 145°01’E; n = 40) and in the Killawarra section of Warby RangeBI State

Park (36°13’S, 146°11’E; n = 27) (Fig. 1). There were 385 sites in total. Relative and absolute

spatial distributions of sites are available from the corresponding author. GunbowerRRG,

BarmahRRG, MillewaRRG and OvensRRG forests were partially flooded in spring 2003 and 2004,

with the only natural flood occurring in the OvensRRG forest (Ovens River is unregulated).

Management agencies inundated the other forests with environmental flows, which are floods

used to stimulate ecological processes. BarmahRRG, MillewaRRG, GunbowerRRG and

KoondrookRRG are Ramsar-listed wetlands of international importance (Murray-Darling Basin

Commission 2005), yet KoondrookRRG has not been flooded since 1996. The Reedy LakeRRG

Wildlife Reserve is a small forest (730 ha) that fringes a swamp on the Goulburn River

floodplain, so that it has a different character to the other river red gum forests (Fig. 1).

At each 0.25 ha site, 5 or 6 small-mammal Elliott traps (33 × 10 × 9 cm; Elliott

Scientific Co., Upwey, Victoria, Australia), were placed on the ground in a straight line, 12 m

7

apart. This number of traps was considered to be sufficient for sampling most animals in a

0.25 ha area because individuals moved so extensively across whole sites that they were

unlikely to miss locations of all traps (H. Lada, pers. obs.), recapture rates were high from

both the same and other sites, and traps on the same site were never saturated with individuals

of A. flavipes. Traps contained polyester bedding and bait (peanut butter, quick oats and

honey), were wrapped in plastic bags covered with bark and set for 1–5 nights, and checked at

dawn and dusk. Variation in trapping effort was due to early trap collection at sites that were

about to be logged or where individuals were repeatedly recaptured and further trapping was

deemed likely to be detrimental to the welfare of the animals. This variation in effort was

accounted for statistically. Animals were treated in a humane manner, research was

authorized by Monash University Biological Sciences Animal Ethics Committee, and

guidelines approved by the American Society of Mammalogists (Gannon et al. 2007) were

followed.

Recaptured individuals were identified by unique positions of ear-biopsy holes (2

mm), which had been taken for genetic analysis (Lada et al. 2007a). For statistical analysis,

we defined the total number of traps per site for each trapping period as the number of traps

on a site less the number of recaptures from the same site. This accounted for recaptures that

had rendered traps unavailable to new individuals. Red marsupia (pouches) and visible teats

identified 2nd-year females (teats of 1st-year females are not visible until the breeding

season).

Site measurements.—Site measurements for GuttrumRRG, KoondrookRRG and

GunbowerRRG forests were obtained from Lada et al. (2007b) and we followed methods

described therein to characterize each 0.25 ha site in MillewaRRG, BarmahRRG, RushworthBI,

Reedy LakeRRG, Warby RangeBI, OvensRRG and ChilternBI. Volumes of fallen timber (“vol”)

were estimated on each site by measuring length and diameter of each log with a diameter

8

≥0.1 m. The numbers of dead and live large trees with diameter at breast height (DBH) ≥ 60

cm were counted for each site as the “Ltrees” variable. In BarmahRRG, MillewaRRG, OvensRRG

and Reedy LakeRRG forests, Euclidean distances between sites and flood fronts (“FloodDist”)

were obtained from field observations and flood maps (K. Ward, Goulburn-Broken

Catchment Management Authority, pers. comm.). In Reedy LakeRRG, FloodDist was the

Euclidean distance from each site to the exposed edge of the swamp, which had been covered

with water in the spring of 2003 and 2004 (H. Lada, pers. obs.). FloodDist for GunbowerRRG,

KoondrookRRG and GuttrumRRG sites were the same as in Lada et al. (2007a). All flooded

areas were covered with water for at least 1 week. We used 2003 and 2004 annual rainfall for

2004 and 2005 trapping data, respectively (Bureau of Meteorology 2003, 2004) as the

previous-year rainfall variable (“rain”). Given the antechinus life history, unfavourable

conditions for rearing offspring in the previous year would affect abundances of animals in

the following year.

Statistical analysis.—We used logistic regression in WinBUGS 1.4 (Spiegelhalter et

al. 2003) to determine the influence of local habitat factors on antechinus capture rates (of all

individuals, of males, of females and of 2nd-year females) in river red gum and box-ironbark

forests. Capture rate was the proportion of traps at a site that captured a unique individual or,

equivalently, the probability of success for a single trap at a given site in a given year. To

allow for differences in trapping effort, the binomial distribution was used for the response

variable, capture rate. The number of unique individuals rij caught at site i in year j was

treated as the result of nij Bernoulli trials with probability of trapping success pij, where nij was

the number of available traps. Survey year was treated as a categorical factor, with volume of

fallen timber (vol), number of large trees (Ltree) and distance to flood locations (FloodDist)

as linear covariates. For box-ironbark sites, the FloodDist variable was not applicable and was

not included. We added type-level fixed effects to account for possible habitat-type-related

9

differences between river red gum sites and box-ironbark sites, site-level random effects to

account for possible temporal autocorrelations in repeated sampling, and forest-level random

effects to account for possible spatial correlations among sites within forests. The model was:

rij ~ binomial (pij, nij)

logit(pij) = α + yearj + βvol .volij + βLtree .Ltreeij + βflooddist .flooddistij + sitei + foresti + typei,

(1)

where βvol is the volume of fallen timber effect, βLtree is the number of large trees effect, and

βflooddist is the FloodDist effect, and sitei, and foresti are random effects, and typei is a fixed

effect.

To reduce correlation among model parameters and to speed model convergence, all

predictor variables were standardized before model fitting (Spiegelhalter et al. 2003):

X’mk = (Xmk −

€

X k)/sk, where

€

X k and sk are the mean and standard deviation of variable k and

Xmk is the mth value of that variable. We set year2004 = 0, so the intercept α is the mean capture

rate in 2004 and the year effects are deviations from that mean for year 2005. We set typeRRG

to zero. Year effects and all covariate coefficients were assigned uninformative normal prior

distributions, with means 0 and high variances (10,000). The random effects terms sitei, and

foresti, and the fixed effect term typei had normal prior distributions with means 0 and

common variances, σsite2, σforest

2 and σtype2, respectively. We gave uniform priors (0,50) to the

standard deviations of the random effects terms, σsite, σforest and σtype (Gelman 2005).

We sampled posterior distributions for 60,000 Markov Chain Monte Carlo (MCMC)

iterations after a 60,000 burn-in. Three different chains were run and the Gelman-Ruben

statistic was monitored to check for MCMC convergence (Brooks and Gelman 1998).

10

We calculated the posterior probability and hence, posterior odds, that each

regression coefficient or fixed effect was either positive or negative, depending on the sign

of the parameter. The prior odds were set at 1:1 (i.e., the parameter was equally likely a

priori to be negative or positive) because we used an uninformative, normal prior

distribution. This means that the posterior odds are equivalent to Bayes factors (Jaynes

2003). Given the data, the Bayes factor is an odds ratio of superiority of one model or

hypothesis (A) over an alternative (B). In this case, HA is that the model parameter differs

from zero and HB is that the parameter does not differ from zero. Bayes factors > 10

provide strong evidence for one hypothesis over another, while 3 < Bayes factors < 10 are

moderate evidence (Jeffreys 1961), and Bayes factors < 3 are “hardly worth reporting”

(Kass and Raftery 1995).

We ran models with additional terms that potentially might affect capture rates,

namely rainfall, and year × FloodDist interaction. The latter allowed for different FloodDist

effects in 2004 and 2005 (see Lada et al. 2007b). We considered that flooding of river red

gum forests reduced the relevance of rainfall effects, so we added previous-year rainfall

variable (rain) to model (1) for box-ironbark forests only. All models were assessed with

Deviance Information Criterion (DIC, Spiegelhalter et al. 2002) for both sexes, males,

females and 2nd-year females. Models with the smallest DIC are most efficacious, but models

that differ by less than two DIC units are regarded as indistinguishable (Burnham and

Anderson 1998). In such cases we chose the simplest model (fewest parameters). Otherwise,

models with the lowest DIC were accepted as the most parsimonious representations.

When no FloodDist effect was observed, we repeated analyses with values of the

FloodDist variable excluded for one forest (BarmahRRG, MillewaRRG, OvensRRG or Reedy

LakeRRG). This allowed us to determine whether results reported by Lada et al. (2007a) were

unique to the GuttrumRRG-KoondrookRRG-GunbowerRRG area, or whether a particular forest

11

was responsible for a lack of effect. If FloodDist effect is present when a particular forest is

excluded from the analysis, it indicates that the inclusion of the forest had caused

disappearance of the effect.

We compared capture rates among select pairs of forests using equation1. Within any

WinBUGS MCMC run, one can compute probability distributions for any function of model

parameters that are estimated. The pij values are the estimates of capture rates in forest i in

year j, which can be averaged over years to produce pi, the forest i-specific rate. We then

compute the difference between, say, pOvens and pWarby Range as (pOvens – pWarby Range) and

monitor its value in the MCMC run. We can use the same inferential method (Bayes factors)

as described above. Capture rates were compared between closely located pairs of river red

gum and box-ironbark forests: OvensRRG and Warby RangeBI (6 km); and Reedy LakeRRG and

RushworthBI (4 km). We compared all pairs of box-ironbark forests, and pairs of river red

gum forests with different, recent flooding history: KoondrookRRG (not flooded since 1996)

was compared with MillewaRRG (controlled floods in 2003 and 2004). Both are forests

growing on the New South Wales (northern) Murray River floodplains. We also compared

OvensRRG and BarmahRRG, both Victorian forests flooded in 2003 and 2004 by the

unregulated Ovens River and the regulated Murray River, respectively. OvensRRG was

compared also to MillewaRRG.

RESULTS

Site measurements and captures of animals.—In 2004 and 2005 (9,754 trap nights,

excluding traps with recaptures on the same site), 62 individual females were captured on 77

sites, and 90 different males on 130 sites in box-ironbark forests. Over the same period, 205

individual females were trapped on 232 sites and 265 different males on 310 sites in river red

gum forests (numbers of females and males in each forest are available from the

12

corresponding author). Second-year females were not trapped in either Warby RangeBI or

GuttrumRRG forests.

Fallen-timber volumes were very low in box-ironbark forests, but higher in river red

gum forests (Table 1). Numbers of large trees per site were 0–18 in river red gum forests and

0–5 in box-ironbark forests (Table 1). All 0.25-ha sites in Reedy LakeRRG, OvensRRG and

BarmahRRG had at least 1 large tree. Large trees were absent from 3% of sites in both

MillewaRRG and GuttrumRRG , 12% of sites in GunbowerRRG, 15% in KoondrookRRG, 32% in

RushworthBI, 46% in Warby RangeBI and 69% in ChilternBI.

Effects of local habitat factors in box-ironbark and river red gum forests on capture

rates.—Deviance Information Criterion (DIC) values were much smaller for model 1 (3

covariates) than for the model with rainfall (4 covariates), for females, second-year females,

males and all animals. Adding year × FloodDist interaction to model 1 improved the fit only

for males. Therefore, the model with the interaction was chosen for males, and model 1 for

the other classes of animals.

We found strong evidence that capture rates increased with greater volumes of fallen

timber for females, 2nd-year females, males and all individuals (Table 2, Fig. 2). There was

moderate evidence for positive effects of abundance of large trees on survival of females to a

2nd year, which was suggested by the increase in capture rates of this class of animal with the

greater number of large trees (Table 2). Large trees also had a positive effect on capture rates

of females and all animals (Table 2, Fig. 2). There was moderate evidence that capture rates

decreased with the distance to flood locations (the FloodDist effect) for males in 2004 and for

second-year females (Table 2, Fig. 2).

There was no evidence for a flood-distance effect for females. However, a negative

response (i.e., a decrease in numbers of females with distance from inundations) was observed

after FloodDist variables for BarmahRRG were removed. In their absence, the evidence for the

13

negative effects of FloodDist on animal abundance was moderate for females, and strong for

2nd-year females and moderate for all individuals (data not shown). Removing FloodDist

variables for one of OvensRRG, MillewaRRG or ReedyRRG reduced evidence for FloodDist

effects because of the influence of BarmahRRG.

Comparisons of capture rates among forests.—We found strong evidence that capture

rates of all animals, females, 2nd-year females and males were higher in Reedy LakeRRG

forest than in nearby RushworthBI forest (Table 3). Capture rates of females were substantially

higher in OvensRRG (natural flood) than in nearby Warby RangeBI (flood irrelevant), in

GuttrumRRG (no flood since 1996), BarmahRRG (controlled large flood), and MillewaRRG

(controlled large flood) (Table 3). In contrast, males were substantially more abundant in each

of GuttrumRRG and BarmahRRG than in OvensRRG forests (Table 3). We found strong evidence

that capture rates of all individuals, females, males and 2nd-year females were higher in

ChilternBI than in either RushworthBI or in Warby RangeBI forests, and were higher for males

in RushworthBI than in Warby RangeBI (Table 3).

DISCUSSION

Our hierarchical analysis of the effects of local habitat factors in box-ironbark and

river red gum forests on capture rates of A. flavipes in 2004–2005 gave partly similar results

to those reported by Lada et al. (2007b) for flood-deprived and partially inundated river red

gum floodplains along Murray River, visited between 2003 and 2005. Data from Lada et al.

(2007a) for 2004 and 2005 were part of both studies. In both studies, there was strong

evidence for an increase in number of females with higher fallen-timber volumes and greater

number of large trees. We found that abundance of males increased with larger fallen-timber

volumes and proximity to floods, whereas Lada et al. (2007a) also found positive effects of

large trees. Here, evidence was moderate rather than strong for increased survival of females

14

to a 2nd year with proximity to floods. In both studies evidence was moderate and strong for

positive effects of large trees and of fallen-timber on 2nd-year female numbers, respectively.

From comparisons between the 2 studies, we infer that the most consistent evidence was for

increases in animal abundances with larger fallen-timber volumes.

Fallen timber is likely to provide shelter and may be particularly important in habitats

with little vegetative groundcover, such as in river red gum forests along the Murray River.

Stokes et al. (2004) showed that foraging activities of A. flavipes and common dunnarts

(Sminthopsis murina) increased when refuges (wire netting on the ground) from predators

were provided. Hollow logs provided den sites for spot-tailed quolls (Dasyurus maculatus)

more often than did any other features (Glen and Dickman 2006). In addition to structural

complexity, fallen timber offers foraging opportunities. Pouncing birds most frequently

swooped on prey from logs (Antos and Bennett 2006). Mac Nally (2006) showed that, in river

red gum forests, the vulnerable brown treecreeper Climacteris picumnus preferred fallen-

timber volumes of ≥ 67 m3 /ha (reported as woodloads of ≥ 40 Mg /ha). It is unknown

whether similar volumes are required for the positive response of the treecreeper in box-

ironbark forests. In the current study, individuals of A. flavipes were observed foraging on and

inside, sheltering in, and sunning on, logs and trees. Also for antechinus, hollows in trees

(including coppiced ones) provided communal nesting sites, while logs served as highways,

providing routes for movement.

Box-ironbark forests in this study are depleted of fallen timber (only 4 of 129 sites had

volumes ≥ 20 m3 /ha), and the range was narrower (0.3–33.7 m3 ha-1) than in river red gum

forests (5.6–212.3 m3 /ha). Only ChilternBI has no ongoing logging/timber removal. Pre-

European volumes of fallen timber in box-ironbark forests are unknown, but most likely were

much higher than the current volumes. This is because logs and large trees (which would

have contributed fallen timber) have been removed systematically from the forests for the last

15

160 years. In the absence of old-growth box-ironbark remnants, the maximum volume of

fallen timber estimated for a site in this study (i.e., 33.7 m3 /ha in much changed RushworthBI

forest) may offer some guidance for pre-European volumes. Given repeated degradation of

box-ironbark forests, the volume of 33.7 m3 /ha is more likely to be an underestimate than an

overestimate of historic volumes. Laven and Mac Nally (1998) suggested that because of

infertility of Australian soils (Muir et al. 1995), pre-European fallen-timber volumes were

likely lower than the volume estimates in North America of 30–1421 m3 /ha and 46–132 m3

/ha for coniferous and deciduous forests, respectively (Harmon et al. 1986). In the current

guidelines for vegetation quality assessment of box-ironbark forests, 0.1 ha sites are assigned

the highest possible value for logs criterion if there are ≥ 10 m of logs (diameter ≥ 0.1 m) of

which ≥ 5 m are large logs (diameter ≥ 0.35 m) (Department of Sustainability and

Environment 2004). Only 10% of the 129 box-ironbark sites in our study fitted this criterion.

This benchmark is equivalent to the minimum fallen-timber volume of 5.2 m3 /ha, which is

6.4 and 12.8 times lower than the minimum amounts to which floodplainRRG A. flavipes and

C. picumnus respectively, responded positively (Mac Nally 2006; Mac Nally and Horrocks

2002; Mac Nally et al. 2002b). To achieve best outcomes for biodiversity and to reach a

compromise between faunal requirements and the massive amounts needed to replenish

fallen-timber, Mac Nally et al. (2001) recommended that heterogeneity in fallen-timber

volumes be considered in habitat restoration, that is, some areas are provided with very high

volumes of fallen timber. Our study supports these recommendations.

Evidence for flood-distance effects on capture rates of A. flavipes were less

pronounced than that reported by Lada et al. (2007a) and seemed to be due to the inclusion of

data for BarmahRRG. The reasons for the exceptional (negative rather than positive) response

to floods in BarmahRRG are currently unclear. If a similar response had been observed in

MillewaRRG (which lies across river from BarmahRRG), then we may have concluded that large

16

controlled floods were less beneficial to antechinus than small, controlled floods. However,

responses of A. flavipes in MillewaRRG differed from BarmahRRG and were similar to those on

other floodplains, where abundances increased with proximity to floods. The numbers of

males and 2nd-year females increased with the proximity to flood locations, as did the

number of females and all individuals after BarmahRRG FloodDist variables were removed

from analyses. These increases in numbers were not due to philopatry. Through genetic

analyses Lada et al. (2007a) inferred that females were not absolutely philopatric, and that

some females in GunbowerRRG forest dispersed towards flooded areas. Males, the

predominant dispersing gender (Coates 1995; Marchesan and Carthew 2004), moved

throughout the whole forest. Emergence of macroinvertebrates on flooded sites following

water recession (Ballinger and Mac Nally 2005; Ballinger et al. 2005) may be beneficial to

the survival of the carnivorous A. flavipes (Lada et al. 2007b). Considering that the increased

abundance of beetles lasted for more than 2 years after the BarmahRRG flood (Ballinger et al.

2005), some antechinus females may have increased their chances of survival to the mating

season and past weaning by residing in areas that had been flooded the previous year.

Complete die-off of antechinus males after a mating season means that persistence of a

population depends on at least some females being able to wean male offspring each year

(and female offspring if all mothers perish). Therefore, the evidence for the positive effects of

larger fallen-timber volumes, number of large trees and proximity to floods on the abundance

of females are critical when conservation decisions are made.

Females were substantially more abundant in OvensRRG forest (natural flooding

regimes) than in BarmahRRG, MillewaRRG (large, controlled flooding) or GuttrumRRG (no flood

since 1996). This result does not necessarily indicate superiority of natural flooding regimes

over environmental flows because only 13 sites were sampled in OvensRRG forest, all near to

inundations, because the floodplain is relatively narrow. Opportunities to study responses to

17

natural flooding regimes are limited because most river red gum floodplains inhabited by A.

flavipes are along rivers with substantial regulation. Given this, we recommend that the Ovens

River remains unregulated in order to maintain the relatively high abundance of A. flavipes in

OvensRRG forest. A substantially higher abundance of females in Reedy LakeRRG forest than in

nearby RushworthBI forest suggested that the former was a better habitat than the latter, at

least over the 2 years sampled. This may be due to the combination of greater availability of

fallen timber, large, hollow-bearing trees and proximity to “inundated areas” in Reedy

LakeRRG forest. However, this forest is relatively small and, even though designated as a

wildlife reserve, is being degraded by illegal removal of trees and logs (H. Lada, pers. obs.).

We were unable to determine which factors make ChilternBI forest particularly suitable for A.

flavipes compared to WarbyBI and RushworthBI forests, both in animal abundances and

survival of females to a 2nd breeding season. Although not measured for each site, ChilternBI

harboured a huge abundance of golden orb-weaving spiders (Nephila species; H. Lada, pers.

obs.), which are possibly a food source for A. flavipes. A. agilis and A. swainsonii, which are

closely related to A. flavipes, mostly ate Araneae (spiders), Coleoptera and Hymenoptera

(ants), in a study at Mumbulla State Forest (Lunney et al. 2001). Future studies should include

assessment of prey availability and identification of dietary components of A. flavipes from

scats in each of the 3 box-ironbark forests.

We have shown the effects of local habitat factors on the abundances of A. flavipes in

river red gum and box-ironbark forests in the bioregions of Victorian Riverina, Murray Fans,

Inland Slopes and Goldfields in Victoria and Riverina in New South Wales. Despite being

much degraded, these forests are of great importance for conservation (Bennett et al. 1998;

Environment Conservation Council 1997). They support native fauna, and together with black

box (E. largiflorens) woodlands, roadside strips, linear riparian zones and some small isolated

patches are the only native vegetation left in the area, which now consists mostly of

18

production land-uses. Preserving, if not improving, the quality of remaining habitats will

likely benefit many native species, including A. flavipes. The shortages of hollow-bearing

large trees may be partially mitigated by, for example, installing nest boxes for wildlife

shelter in the short-term, and allowing medium-sized trees to grow large in the long-term.

Fallen-timber volumes should be restored in at least some parts of forests, and floodplains

need to be flooded more regularly in spring. It is possible that these three factors also play

important roles on animal abundances in floodplain ecosystems around the world and may

need to be considered together for restoration of habitats.

ACKNOWLEDGEMENTS

The work was partially supported by a grant from the Monash Research Fund: New

Research Areas scheme. Thanks to Anna Lada, Peter Lada and Quentin Lang for assistance

with fieldwork, to Jon Cuddy, Greg Horrocks and Keith Ward for information, to James

Thomson for discussions about data analyses and to two anonymous reviewers for comments

how to improve the manuscript. This work was carried out under Monash University

Biological Sciences Animal Ethics Committee permit BSCI/2003/02, Department of

Sustainability and Environment permit 10002325, Animal Care and Ethics Committee of the

Director-General of New South Wales (NSW) Agriculture Certificate of Approval, Animal

Research Authority issued by the Director-General of NSW Department of Primary

Industries, Forestry Commission of NSW permit XX23517 and NSW National Parks and

Wildlife Service research permit S10252. This is publication number 110 from the Australian

Centre for Biodiversity: Analysis, Policy and Management.

19

LITERATURE CITED

ANDERSEN, D. C., K. R. WILSON, M. S. MILLER, AND M. FALCK. 2000. Movement

patterns of riparian small mammals during predictable floodplain inundation. Journal

of Mammalogy 81:1087-1099.

ANTOS, M. J., AND A. F. BENNETT. 2006. Foraging ecology of ground-feeding woodland

birds in temperate woodlands of southern Australia. Emu 106:29-40.

BALLINGER, A., AND R. MAC NALLY. 2005. Flooding in Barmah-Millewa Forest:

Catastrophe or opportunity for non-aquatic fauna? Proceedings of the Royal Society of

Victoria 117:111-116.

BALLINGER, A., R. MAC NALLY, AND P. S. LAKE. 2005. Immediate and longer-term

effects of managed flooding on floodplain invertebrate assemblages in south-eastern

Australia: generation and maintenance of a mosaic landscape. Freshwater Biology

50:1190-1205.

BENNETT, A., G. BROWN, L. LUMSDEN, D. HESPE, S. KRASNA, AND J. SILINS.

1998. Fragments for the future. Wildlife in the Victorian Riverina (the Northern

Plains). The Department of Natural Resources and Environment, East Melbourne.

BROOKS, S. B., AND A. GELMAN. 1998. General methods for monitoring convergence of

iterative simulations. Journal of Computational and Graphical Statistics 7:434-455.

BUREAU OF METEOROLOGY. 2003. Australian monthly rainfall gridded data (electronic

resource): 1900 to 2003. Bureau of Meteorology, National Climate Centre,

Melbourne, Vic.

BUREAU OF METEOROLOGY. 2004. Monthly Weather Review. Bureau of Meteorology,

Victorian Regional Office.

20

BURNHAM, K., AND D. ANDERSON. 1998. Model selection and inference. Springer, New

York.

CAMERON, M. 2006. Nesting habitat of the glossy black-cockatoo in central New South

Wales. Biological Conservation 127:402-410.

CLOSE, A. 1990. The impact of man on the natural flow regime, Pp. 61-74 in The Murray

(N. MacKay and D. Eastburn, eds.). Murray-Darling Basin Commission, Canberra.

COATES, T. 1995.Reproductive ecology of the yellow-footed antechinus, Antechinus

flavipes (Waterhouse), in north east Victoria. Ph.D., Monash University, Melbourne.

CUNNINGHAM, S. C., J. READ, P. J. BAKER, AND R. MAC NALLY. 2007. Quantitative

assessment of stand condition and its relationship to physiological stress in stands of

Eucalyptus camaldulensis (Myrtaceae). Australian Journal of Botany 55:692–699.

DEPARTMENT OF SUSTAINABILITY AND ENVIRONMENT. 2004. Vegetation Quality

Assessment Manual - Guidelines for applying the Habitat Hectares scoring method.

Victorian Government Department of Sustainability and Environment.

ENVIRONMENT CONSERVATION COUNCIL. 1997. Box-ironbark. Forest and woodlands

investigation. Resources and issues report. Environment Conservation Council

(Victoria).

ERICKSON, J. L., AND S. D. WEST. 2003. Associations of bats with local structure and

landscape features of forested stands in western Oregon and Washington. Biological

Conservation 109:95-102.

FAHEY, C. 1988. The Barmah Forest - a history. Department of Conservation, Forests and

Lands, Melbourne.

GANNON, W. L., R. S. SIKES, AND THE ANIMAL CARE AND USE COMMITTEE OF

THE AMERICAN SOCIETY OF MAMMALOGISTS. 2007. Guidelines of the

21

American Society of Mammalogists for the use of wild mammals in research. Journal

of Mammalogy 88:809-823.

GELMAN, A. 2005. Prior distributions for variance parameters in hierarchical models.

Bayesian Analysis 1:1-19.

GIBB, H., ET AL. 2006. Conservation-oriented forestry and early successional saproxylic

beetles: Responses of functional groups to manipulated dead wood substrates.

Biological Conservation 129:437-450.

GIBBONS, P., D. B. LINDENMAYER, S. C. BARRY, AND M. T. TANTON. 2002. Hollow

selection by vertebrate fauna in forests of southeastern Australia and implications for

forest management. Biological Conservation 103:1-12.

GLEN, A. S., AND C. R. DICKMAN. 2006. Home range, denning behaviour and

microhabitat use of the carnivorous marsupial Dasyurus maculatus in eastern

Australia. Journal of Zoology 268:347-354.

GRANJON, L., J.-F. COSSON, E. QUESSEVEUR, AND B. SICARD. 2005. Population

dynamics of the multimammate rat Mastomys huberti in an annually flooded

agricultural region of central Mali. Journal of Mammalogy 86:997-1008.

GROVE, S. J. 2002. Tree basal area and dead wood as surrogate indicators of saproxylic

insect faunal integrity: a case study from the Australian lowland tropics. Ecological

Indicators 1:171-188.

HARMON, M. E., ET AL. 1986. Ecology of coarse woody debris in temperate ecosystems.

Advances in Ecological Research 15:133-302.

HOLLOWAY, G. L., AND J. R. MALCOLM. 2007. Nest-tree use by northern and southern

flying squirrels in central Ontario. Journal of Mammalogy 88:226-233.

JACOB, J. 2003. The response of small mammal populations to flooding. Mammalian

Biology 68:102-111.

22

JAYNES, E. T. 2003. Probability theory: the logic of science. Cambridge University Press,

New York.

JEFFREYS, H. 1961. Theory of probability. Oxford University Press, Oxford, U. K.

JOHNSON, C. 2006. Australia's mammal extinctions: A 50 000 year history. Cambridge

University Press, New York.

KASS, R. E., AND A. E. RAFTERY. 1995. Bayes Factors. Journal of the American

Statistical Association 90:773-795.

LADA, H., R. MAC NALLY, AND A. C. TAYLOR. 2007a. Genetic reconstruction of

population dynamics of a carnivorous marsupial (Antechinus flavipes) in response to

floods. Molecular Ecology 16:2934-2947.

LADA, H., J. R. THOMSON, R. MAC NALLY, G. HORROCKS, AND A. C. TAYLOR.

2007b. Evaluating simultaneous impacts of three anthropogenic effects on a

floodplain-dwelling marsupial Antechinus flavipes. Biological Conservation 134:527-

536.

LAVEN, N. H., AND R. MAC NALLY. 1998. Association of birds with fallen timber in box-

ironbark forest of central Victoria. Corella 22:56-60.

LINDHE, A., N. ÅSENBLAD, AND H.-G. TORESSON. 2004. Cut logs and high stumps of

spruce, birch, aspen and oak - nine years of saproxylic fungi succession. Biological

Conservation 119:443-454.

LUNNEY, D., A. MATTHEWS, AND J. GRIGG. 2001. The diet of Antechinus agilis and A.

swainsonii in unlogged and regenerating sites in Mumbulla State Forest, south-eastern

New South Wales. Wildlife Research 28:459-464.

MAC NALLY, R. 2006. Longer-term response to experimental manipulation of fallen timber

on forest floors of floodplain forest in south-eastern Australia. Forest Ecology and

Management 229:155-160.

23

MAC NALLY, R., AND G. HORROCKS. 2002. Habitat change and restoration: responses of

a forest-floor mammal species to manipulations of fallen timber in floodplain forests.

Animal Biodiversity & Conservation 25:41-52.

MAC NALLY, R., G. HORROCKS, AND L. PETTIFER. 2002b. Experimental evidence for

potential beneficial effects of fallen timber in forests. Ecological Applications

12:1588-1594.

MAC NALLY, R., A. PARKINSON, G. HORROCKS, L. CONOLE, AND C. TZAROS.

2001. Relationships between terrestrial vertebrate diversity, abundance and

availability of coarse woody debris on south-eastern Australian floodplains. Biological

Conservation 99:191-205.

MAC NALLY, R., A. PARKINSON, G. HORROCKS, AND M. YOUNG. 2002a. Current

loads of coarse woody debris on southeastern Australian floodplains: Evaluation of

change and implications for restoration. Restoration Ecology 10:627-635.

MARCHESAN, D., AND S. M. CARTHEW. 2004. Autecology of the yellow-footed

antechinus (Antechinus flavipes) in a fragmented landscape in southern Australia.

Wildlife Research 31:273-282.

MASER, C., AND J. R. SEDELL. 1994. From the forest to the sea. The ecology of wood in

streams, rivers, estuaries, and oceans. St. Lucie Press, Delray Beach, Florida, USA.

MCALLAN, B. M., C. R. DICKMAN, AND M. S. CROWTHER. 2006. Photoperiod as a

reproductive cue in the marsupial genus Antechinus: ecological and evolutionary

consequences. Biological Journal of the Linnean Society 87:365-379.

MCDONALD, I. R., A. K. LEE, A. J. BRADLEY, AND K. A. THAN. 1981. Endocrine

changes in dasyurid marsupials with differing mortality patterns. General and

Comparative Endocrinology 44:292-301.

24

MENKHORST, P., AND F. KNIGHT. 2001. A field guide to the mammals of Australia.

Oxford University Press, South Melbourne.

MICKLIN, P. P. 1988. Desiccation of the Aral Sea: a water management disaster in the

Soviet Union. Science 241:1170-1176.

MUIR, A. M., S. A. EDWARDS, AND M. J. DICKINS. 1995. Description and conservation

status of the vegetation of the box-ironbark ecosystem in Victoria. Flora & Fauna

Technical Report 136. Department of Conservation and Natural Resources, Victoria,

Research Paper 136.

MURRAY-DARLING BASIN COMMISSION. 2005. The Living Murray Foundation Report

on the significant ecological assets targeted in the First Step Decision. Murray-Darling

Basin Commission.

REDHEAD, T. D. 1979. On the demography of Rattus sordidus colletti in monsoonal

Australia. Australian Journal of Ecology 4:115-136.

SALA, O. E., ET AL. 2000. Global biodiversity scenarios for the year 2100. Science

287:1770-1774.

SIPPOLA, A.-L., M. MÖNKKÖNEN, AND P. RENVALL. 2005. Polypore diversity in the

herb-rich woodland key habitats of Koli National Park in eastern Finland. Biological

Conservation 126:260-269.

SPIEGELHALTER, D., A. THOMAS, N. BEST, AND D. LUNN. 2003. WinBUGS user

manual. Institute of Public Health, Cambridge.

SPIEGELHALTER, D. J., N. G. BEST, B. P. CARLIN, AND A. VAN DER LINDE. 2002.

Bayesian measures of complexity and fit. Journal of the Royal Statistical Society

Series B 64:583-639.

STOKES, V. L., R. P. PECH, P. B. BANKS, AND A. D. ARTHUR. 2004. Foraging

behaviour and habitat use by Antechinus flavipes and Sminthopsis murina

25

(Marsupialia: Dasyuridae) in response to predation risk in eucalypt woodland.

Biological Conservation 117:331-342.

WALDIEN, D. L., J. P. HAYES, AND M. M. P. HUSO. 2006. Use of downed wood by

Townsend's chipmunks (Tamias townsendii) in western Oregon. Journal of

Mammalogy 87:454-460.

WALKER, J., F. BULLEN, AND B. G. WILLIAMS. 1993. Ecohydrological changes in the

Murray-Darling Basin. I. The number of trees cleared over two centuries. Journal of

Applied Ecology 30:265-273.

WILCOVE, D. S., D. ROTHSTEIN, J. DUBOW, A. PHILLIPS, AND E. LOSOS. 1998.

Quantifying threats to imperiled species in the United States. Bioscience 48:607-615.

WILLIAMS, A. K., M. J. RATNASWAMY, AND R. B. RENKEN. 2001. Impacts of a flood

on small mammal populations of lower Missouri River floodplain forests. American

Midland Naturalist 146:217-221.

WOOLLEY, P. 1966. Reproduction in Antechinus spp. and other Dasyurid marsupials.

Symposia of the Zoological Society of London 15:281-294.

26

Figure Legends

FIG. 1.—Location of study sites in river red gum forests (RRG) and box-ironbark forests (BI)

in southeastern Australia. R=river.

FIG. 2.—Fitted relationships between capture rates of yellow-footed antechinus (Antechinus

flavipes) A) females and fallen-timber volumes (vol), B) females and number of large trees

(Ltree), C) males and distances to 2003 floods (FloodDist), D) all individuals and fallen-

timber volumes, E) all individuals and number of large trees, and F) 2nd-year females and

distances to floods. Capture rate is the probability of success (catching a unique individual)

for a single trap at a given site in a given year. Plotted probabilities (capture rates) were

calculated from the posterior distributions of model parameters while holding all covariates,

other than the variable of interest, constant at their means (standardized value = 0). Values at

mean for 2004 were used for all plots. Solid lines show posterior probabilities (capture rates),

broken lines are upper and lower 95% credible intervals.

27

TABLE 1.—Habitat variables in three box-ironbark (BI) and seven river red gum (RRG) forests:

mean fallen-timber volumes (m3 ha-1), numbers of large trees, and distance (km) to flood in

2003 (FloodDist04) and in 2004 (FloodDist05). There were no floods in box-ironbark forests

(N/A).

0.25-ha Fallen-timber large trees

sites volume per 0.25-ha site FloodDist04 FloodDist05

Forest sampled Mean ± SD Mean ± SD Mean ± SD Mean ±SD

ChilternBI 62 4.6 ± 3.2 0.4 ± 0.6 NA NA

Warby RangeBI 27 6.7 ± 5.3 0.6 ± 0.6 NA NA

RushworthBI 40 9.0 ± 7.1 1.4 ± 1.5 NA NA

Reedy LakeRRG 23 24.4 ± 12.6 5.5 ± 2.7 0.8 ± 0.2 0.7 ± 0.2

OvensRRG 13 51.2 ± 20.0 3.6 ± 1.6 0.1 ± 0.1 0.0 ± 0.0

GunbowerRRG 102 59.2 ± 33.5 2.5 ± 2.1 2.1 ± 1.2 1.9 ± 1.3

BarmahRRG 29 68.6 ± 43.4 4.1 ± 3.0 0.9 ± 1.1 2.2 ± 2.5

KoondrookRRG 27 71.1 ± 33.9 2.4 ± 2.1 3.5 ± 0.2 3.3 ± 0.5

GuttrumRRG 32 75.1 ± 39.8 4.5 ± 3.4 14.3 ± 0.2 14.3 ± 0.2

MillewaRRG 30 88.4 ± 31.2 3.6 ± 2.1 0.1 ± 0.1 0.1 ± 0.1

28

TABLE 2.—Bayesian regression analyses of capture rates of the yellow-footed antechinus (Antechinus flavipes) with respect to fallen-timber

volume, number of large trees, and distance from flood locations at sampled sites in river red gum and box-ironbark forests in southeastern

Australia.

Capture ratea — fallen-timber volume Capture ratea — large/very large trees Capture ratea — distance to floods

Posterior Posterior Posterior

mean mean mean

probability probability probability

Direction of Bayesb Direction of Bayesb Direction of Bayesb

Coefficient for Mean ± SD of response response factor Mean ± SD of response response factor Mean ± SD of response response factor

Both sexes 0.31 ± 0.08 positive 1.000 Infinity 0.10 ± 0.06 positive 0.926 12.51 –0.03 ± 0.22 negative 0.567 1.31

Females 0.33 ± 0.10 positive 1.000 Infinity 0.21 ± 0.08 positive 0.994 166 0.08 ± 0.32 positive 0.567 1.31

2nd-yr females 0.26 ± 0.20 positive 0.913 10.5 0.20 ± 0.18 positive 0.860 6.14 –0.37 ± 0.21 negative 0.918 3.76

Males 0.24 ± 0.08 positive 0.997 332 0.01 ± 0.08 positive 0.556 1.25

Males in 2004 –0.19 ± 0.19 negative 0.831 4.92

Males in 2005 –0.02 ± 0.16 negative 0.559 1.27

Capture ratea — interaction between year and distance to flood

Males 0.17 ± 0.16 positive 0.854 5.85

a Data as natural log of odds b Bayes factors: > 10 indicates very high support for hypothesis; > 3 indicates support for hypothesis

29

TABLE 3.—Mean capture rates of the yellow-footed antechinus (Antechinus flavipes) in box-

ironbark (BI) and river red gum (RRG) forests in southeastern Australia for 2004 and 2005.

Pairwise comparisons of capture rates between forests are shown in lower part of table.

Both sexes Males Females Second-year Females

Forest Mean ±SD Mean ± SD Mean ± SD Mean ± SD

ChilternBI 0.085 ± 0.036 0.045 ± 0.026 0.048 ± 0.042 0.016 ± 0.049

WarbyBI 0.024 ± 0.011 0.012 ± 0.008 0.015 ± 0.014 0.001 ± 0.001

RushworthBI 0.025 ± 0.012 0.018 ± 0.011 0.007 ± 0.006 0.001 ± 0.003

Reedy LakeRRG 0.053 ± 0.012 0.030 ± 0.008 0.023 ± 0.008 0.003 ± 0.002

OvensRRG 0.033 ± 0.009 0.010 ± 0.003 0.025 ± 0.009 0.001 ± 0.001

GunbowerRRG 0.049 ± 0.006 0.028 ± 0.004 0.020 ± 0.003 0.002 ± 0.001

BarmahRRG 0.026 ± 0.007 0.016 ± 0.004 0.011 ± 0.004 0.001 ± 0.001

KoondrookRRG 0.035 ± 0.006 0.020 ± 0.004 0.015 ± 0.004 0.001 ± 0.001

GuttrumRRG 0.030 ± 0.015 0.023 ± 0.010 0.006 ± 0.005 0.001 ± 0.001

MillewaRRG 0.031 ± 0.007 0.019 ± 0.005 0.013 ± 0.004 0.001 ± 0.001

Comparison Probabilitya Probabilitya Probabilitya Probabilitya

ChilternBI, WarbyBI 1.000 C 1.000 C 1.000 C 0.999 C

ChilternBI, RushworthBI 1.000 C 1.000 C 1.00 0 C 0.999 C

WarbyBI, RushworthBI 0.585 0.909 Rus 0.058 0.731

ReedyRRG, RushworthBI 0.965 RL 0.895 RL 0.975 RL 0.914 RL

OvensRRG, WarbyBI 0.797 Ov 0.568 0.866 Ov 0.798 Ov

OvensRRG, BarmahRRG 0.770 Ov 0.869 B 0.974 Ov 0.548

OvensRRG, MillewaRRG 0.573 0.952 M 0.969 Ov 0.700

OvensRRG, GuttrumRRG 0.588 0.911 GT 0.963 Ov 0.745 Ov

KoondrookRRG, MillewaRRG 0.696 0.577 0.686 0.510

a Posterior probability of higher capture rate in: Chiltern (C), Ovens (Ov), Reedy Lake (RL), Rushworth (Rus),

Barmah (B), Millew (M) and Guttrum GT). Difference considered substantial if probability > 0.75 (Bayes factor

> 3).

30

31

FIG. 2