Variable offspring provisioning and fitness: a direct test in the field


*Correspondence author. E-mail:


  • 1Variation in the quality of the offspring environment can lead to the evolution of variable offspring provisioning. For variable offspring provisioning to evolve, the magnitude of provisioning per offspring must have effects on offspring and parental fitness.
  • 2Females of the quacking frog, Crinia georgiana, produce clutches of eggs in which egg size varies between individuals in the population and also within clutches, independent of female size or condition. A trade-off between egg size and number exists.
  • 3Using microsatellite markers to trace offspring to parents, and therefore clutch type, we tested the performance of tadpoles from clutches of large, small and variable-sized eggs in ponds in the field.
  • 4Clutches of large eggs resulted in the highest parental fitness and clutches of small eggs resulted in lower parental fitness values.
  • 5The parental fitness indicated that conditions in these ponds were harsh. Clutches with variable egg sizes had intermediate parental fitness but may be of benefit as a bet-hedging strategy when the qualities of ponds are unpredictable.
  • 6This study provides new empirical evidence of important offspring and parental fitness consequences of variable maternal provisioning under natural conditions in the field and demonstrates the importance of moving experimental life-history studies out of the laboratory into the field where realistic selection pressures act on developing offspring.


Theoretical models predict that in a constant offspring environment, a fixed optimal amount of provisioning per offspring should evolve, assuming offspring fitness increases monotonically with size, and a trade-off for parents in offspring size and number exists (Smith & Fretwell 1974; Lloyd 1987; Winkler & Wallin 1987). This optimal offspring size does not necessarily result in maximum offspring fitness but rather maximizes reproductive success or efficiency: parental fitness. Variation in the quality of the offspring environment can result in alterations to the relationship between offspring size and fitness: the offspring fitness curve. Because of this variation, different fitness curves and resulting fitness functions can exist. Since parental fitness is based on these offspring fitness functions (Smith & Fretwell 1974) multiple offspring size optima can exist if there is variation in the quality of the offspring environment (Kaplan & Cooper 1984; McGinley, Temme & Geber 1987). Variation in the quality of the offspring environment may lead to variation in mean offspring size evolving between individuals in a population. Alternatively, a bet-hedging strategy of producing variable-sized offspring in a clutch may evolve under certain conditions: for example, temporal variation in the quality of the offspring environment (Capinera 1979; Crump 1981; McGinley et al. 1987).

The amount of provisioning that an offspring receives can be critical to the fitness and survival of the offspring (Crump 1984; Parichy & Kaplan 1992; Einum & Fleming 2000a,b; Walker, Rypstra & Marshall 2003; Dziminski & Roberts 2006). The consequences of the amount of provisioning on offspring fitness can be crucial in organisms that have a larval stage that occurs in complex, heterogeneous environments (Fox & Mousseau 1996; Einum & Fleming 1999, 2000b; Moran & Emlet 2001; Czesak & Fox 2003). For anurans breeding in temporary freshwater environments, major risks include predation (Woodward 1983; Werner 1986; Kats, Petranka & Sih 1988; Wilbur 1988; Kaplan 1992), competition (Rugh 1934; Savage 1952; Wilbur & Collins 1973; Steinwascher 1978; Morin & Johnson 1988) and desiccation (Crump 1989; Alford 1999; Doughty & Roberts 2003). Offspring provisioning may modulate vulnerability to these risks.

Females of the myobatrachid frog, Crinia georgiana, produce clutches of eggs in which yolk volumes vary both within clutches and between females, but this variation is independent of female size (Dziminski & Roberts 2006). A trade-off in egg size and number also exists (Dziminski & Roberts 2006). In offspring raised in the laboratory, the amount of yolk strongly affects offspring fitness and this translates to strong influences on estimates of parental fitness (Dziminski & Roberts 2006). In many studies, the effects of offspring provisioning have been examined under benign laboratory conditions (Crump 1984; Reznick 1991; Dziminski & Alford 2005; Dziminski & Roberts 2006) with effects commonly disappearing later in the larval stage (Crump 1984; Tejedo & Reques 1992; Heath & Blouw 1998; Laugen, Laurila & Merila 2002; Loman 2002). In the field, conditions may be much harsher and more complex because of density-dependent fitness loss in offspring and predation risks (Alford 1999), so a true test of the value of maternal provisioning demands a direct field test. Under natural conditions, dynamic features of the amphibian larval stage, such as plasticity in larval period and size at metamorphosis will be expressed as a response to both the biotic (Wilbur & Collins 1973; Van Buskirk & Relyea 1998) and physical (Crump 1989) aspects of the offspring environment. We tested the consequences of maternal provisioning on offspring fitness, and tested the performance of variable provisioning strategies. To achieve this in the field, we used microsatellite markers to trace offspring to parents, and thus known maternal investment patterns.

The objectives of this study were to determine, in natural ponds in the field,

  • 1the parental fitness of producing clutches of large, small and variable-sized eggs, and
  • 2the relationship between offspring size and fitness and the relationship between offspring size and parental fitness in the uniform-sized egg clutches.

Crinia georgiana females deposit egg clutches in small ponds at breeding sites and up to six egg clutches may be deposited in a single pond, which can result in high larval densities. This species provides an opportunity to determine and compare the success of clutches directly in the field at natural densities. Crinia georgiana females produce clutches of 158 (± 55 SE) eggs, enabling the accurate measurement of total egg size and number, and allowing an accurate calculation of the trade-off between egg size and number which exists in this species (Dziminski & Roberts 2006). Thus an accurate measure of total investment and strategy (large, small, variable eggs) of the female can be obtained. Collection of all surviving offspring from natural ponds in the field, and matching of offspring to parents, enables the success of these strategies and investments to be determined (Einum & Fleming 1999; Dziminski & Roberts 2006).

Materials and methods

egg collection and experimental ponds

Non-amplexed, gravid female C. georgiana were collected together with an equal number of calling males from a breeding site in the Darling Ranges, c. 50 km southeast of Perth, western Australia over three nights in June 2002. Each night frogs were transported to the laboratory in Perth. Each female together with a randomly chosen male was placed in a clear round plastic container (8·5 cm diameter × 11 cm) with a lid and 1 cm depth of tap water purified by a reverse osmosis filter. Frogs were left in a dark room at 15 °C until eggs were deposited. Every female amplexed with her assigned male and egg deposition was usually complete within 10 min.

Upon completion of egg deposition, the yolk volume of every egg in every clutch was measured using methods in Dziminski & Alford (2005). Females were candled using a bright light to ensure all ova were deposited then weighed to the nearest unit (mg) using an electronic balance. No females retained any ova. To ensure mean egg sizes among females are actually manifestations of strategies rather than simply of fitness differences and constraints, a trade-off between egg size and number must exist. Controlling for allometry (female mass), there was a significant negative relationship between egg size and number (partial correlation: r = −0·562, d.f. = 30, P < 0·001). To illustrate this relationship the residuals from regressions of egg size and number, with female size, were plotted (Fig. 1). Egg size was not simply a function of female size (Pearson correlation: r = 0·150, N = 33, P = 0·406). Furthermore, we estimated a body condition index by calculating residuals from the regression of female snout-vent length and mass (linear regression: slope = 0·218, r2 = 0·835, P < 0·001, N = 33). Egg size was not a function of estimated body condition (Pearson correlation: r = −0·145, N = 33, P = 0·421). There also was no relationship between egg size and overall yolk investment per clutch, controlling for allometry (partial correlation: r = 0·214, d.f. = 30, P = 0·239), further indicating that the differences in eggs size between females were not due to fitness differences between females.

Figure 1.

The trade-off between egg size and number. Residuals from regressions of egg size and number with female size are plotted.

Thirty-six egg clutches were used in the experiment, 12 clutches that had large eggs, 12 clutches that had small eggs and 12 clutches that had variable-sized eggs. Clutches were chosen visually then confirmed by analysis of egg sizes. One-way anova revealed a significant difference in yolk volume between the three clutch types (F2,33 = 16·603, P < 0·001; Fig. 2a), LSD tests revealed that large egg clutches had a significantly larger mean egg size than variable-sized (P < 0·001) and small (P < 0·001) egg clutches. Variable-sized egg clutches had a significantly larger mean egg size than small egg clutches (P < 0·01). One-way anova revealed a significant difference in coefficient of variation (CV) of egg size between the three clutch types (F2,33 = 8·379, P < 0·01; Fig. 2b), LSD tests revealed that variable-sized egg clutches had a higher CV than large (P < 0·01) and small (P < 0·01) egg clutches but the CVs of large and small egg clutches were not significantly different (P = 0·711).

Figure 2.

(a) Mean yolk volume of clutches of large, variable-sized and small eggs used in the experiment. (b) Mean CV of clutches of large, variable-sized and small eggs used in the experiment. All bars represent ± 1 SE.

A toe clip from both parents of every clutch used in the experiment was preserved in 100% ethanol. Egg clutches were kept in plastic containers (8·5 cm diameter × 11 cm) and 1 cm depth of purified tap water until hatching. All eggs hatched. Groups of three egg clutches (one clutch of large eggs, one clutch of small eggs, one clutch of variable eggs) deposited on the same night were randomly assigned to one of 12 ponds. This stocking density falls within the range in natural populations, where usually one or two, but up to occasionally six clutches have been found deposited in a pond (M. Dziminski, pers. obs.). All offspring were put into ponds 12 h after hatching.

Ponds at a breeding site in the Darling Ranges were enclosed with rigid 40 cm high aluminium mesh fence (mesh size = 1 mm2) with an inward inverted lip before the breeding season. This prevented resident frogs in the area breeding in these ponds and the small mesh size prevented metamorphs escaping. Crinia georgiana are ground frogs and adults and metamorphs were never seen climbing the fence. When some larvae developed advanced hind limbs, ponds were monitored daily for metamorphs that had emerged from the water. Emerged metamorphs were collected and placed in rectangular plastic containers (15 × 10 cm, height = 6 cm) with lids, some moss and a little pond water, and kept at 15 °C until (Gosner 1960) stage 46 (complete tail resorbtion), which usually took 1–2 days. Metamorphs were then blotted dry with a cotton towel and weighed to 0·1 mg precision on an electronic balance. The larval period was defined as commencing the day of fertilization and ending at (Gosner 1960) stage 46. All offspring were then euthanized by immersion in and preserved in 100% ethanol. Ponds were monitored until either all offspring emerged or the pond dried.

offspring and adult genotyping

Genomic DNA was extracted from whole limbs of preserved larvae and adult toe clips using the EDNA HiSpExTM tissue kit (Fisher Biotec). Genomic DNA extracts were used as templates for amplification of six microsatellite loci (Dziminski 2006) via the polymerase chain reaction (PCR). Each PCR reaction (10 µL total volume) contained 1 µL of genomic extract, 0·25 U Taq polymerase (HotStartTaq®, Qiagen), 1 µL of 10× PCR buffer (Qiagen), 2 µL Q-solution (Qiagen), 0·2 mM dNTPs and one group of three multiplexed sets of primers (see Table 1). Forward primers were 5′-labelled with a fluorescent dye (Table 1). PCR amplifications were performed in ABI 9700 (Applied Biosystems) and PTC-200 (MJ Research) thermocyclers using the following thermotreatment: 15 min initial polymerase activation step at 95 °C; followed by 30 cycles of 1 min at 95 °C for denaturation, 1 min at 55 °C annealing temperature, and 1 min at 65 °C for extension. Amplified products were diluted (×16) and subjected to capillary electrophoresis on an ABI 3730 DNA Analyser (Applied Biosystems). Fragment sizes were determined with the GeneScan-500 (LIZ) size standard using the software package abi prism GeneMapper version 3.0 (Applied Biosystems). All processes were carried out in 96-well plates using a Biomek® 2000 Laboratory Automation Workstation (Beckman Coulter).

Table 1.  Multiplexed primer labelling and amounts used
LocusDyeAmount per PCR reaction (pmol)
Multiplex 1
 Cg2Ca24Ned 1·0
 Cg3Ca8Vic 2·5
 Cg2Ca66-Fam 2·5
Multiplex 2
 Cg1Ca2Ned 2·5
 Cg1Ca5Vic 0·5

parentage analyses

Offspring in a pond could only come from three mutually exclusive pairs of parents. The computer program parente (Cercueil, Bellemain & Manel 2002) was used to assign maternity and paternity, simultaneously, to an offspring from the three possible combinations of parents in a pond, resulting in 100% confidence in correct assignment.

statistical analyses

anova was used to compare the effects of clutch type on mean size at metamorphosis, the mean length of larval period and survival of clutches. Night of egg deposition does not have any effect on offspring fitness in naturally provisioned egg clutches collected from females in the field (Dziminski & Roberts 2006) so was not included in the model. Ponds were included in the model as a random, blocking factor, and we neither report the statistics of this factor nor the interaction of this factor with clutch type. All post hoc tests were least significant difference (LSD) pairwise multiple comparisons. Data were normally distributed and the assumption of homogeneity of variance among treatment means was satisfied. The proportions of larvae surviving from each clutch type were arcsine transformed (Zar 1999) into degrees before analysis.

To examine the relationship between offspring size and fitness, the mean masses at metamorphosis of offspring from uniform egg size clutches only (small and large clutch type treatments) were plotted against the initial mean egg size and a logarithmic regression was used to describe this relationship. This relationship represents how offspring fitness relates to yolk input (the estimated offspring fitness curve), since size at metamorphosis has been demonstrated to reflect the relative probability of survival and reproduction of metamorphs of two other anuran species (Altwegg & Reyer 2003). Females differ in the amount of energy available to convert to yolk for provisioning offspring. Parental fitness efficiency, or fitness return per reproductive effort, of each clutch was calculated using the following formula:


where eWp is parental fitness, M is total biomass of all offspring at metamorphosis in mg, Iy is the mean amount of yolk per offspring (effort per offspring) in mm3, a is the slope and b is the intercept of the logarithmic regression described above, and E is the total yolk volume (total reproductive effort) produced by the female in mm3. The inclusion of the equation of the relationship between offspring fitness and yolk input in the calculation ensures that there is a distinct parental fitness difference between producing, for example, one offspring of size 1 and two offspring of size 0·5. The anova model described above was used to compare the fitness efficiencies of each clutch type. Since the Smith & Fretwell (1974) fitness set assumes a uniform offspring size distribution within clutches, the fitness efficiencies of uniform egg size clutches only (small and large clutch treatments) were plotted against the initial mean egg size and a logistic regression was used to describe this relationship, which is the parental fitness curve (Smith & Fretwell 1974).


All 72 parents and 1172 offspring that survived were genotyped, and every offspring was assigned parentage successfully. Clutch type had a significant effect on size at metamorphosis (Table 2a, Fig. 3a), larval period (Table 2b, Fig. 3b) and survival (Table 2c, Fig. 3c). Offspring from large egg clutches were significantly larger than offspring from variable-sized (LSD: P < 0·001) and small (LSD: P < 0·001) egg clutches. The mean difference in size at metamorphosis between variable-sized and small egg clutches exceeded the α = 0·05 level (LSD: P = 0·088). Offspring from large egg clutches had a significantly shorter larval period than offspring from variable-sized (LSD: P < 0·01) and small (LSD: P < 0·01) egg clutches. Offspring from large egg clutches had significantly higher survival than offspring from variable-sized (LSD: P < 0·01) and small (LSD: P < 0·01) egg clutches. The mean difference in survival between variable-sized and small egg clutches just exceeded the α = 0·05 level (LSD: P = 0·061).

Table 2. anova of (a) size at metamorphosis, (b) length of larval period, (c) survival and (d) parental fitness
  • *

    Four small egg clutches and one variable egg sized clutch had zero survival, therefore d.f. = 17.

(a) Mass at metamorphosis
  Pond11  5·361  
  Clutch type 2 72·39326·865< 0·001
  Residual17*  2·695  
(b) Larval period
  Clutch type 2311·301 8·050< 0·01
(c) Survival
  Clutch type 22559·81914·820< 0·001
(d) Parental fitness
  Clutch type 2630·73513·618< 0·001
Figure 3.

(a) Least squares (LS) mean size at metamorphosis of offspring from the three clutch types. (b) LS mean length of larval period of offspring from the three clutch types. (c) LS mean proportion of larval survival of offspring from the three clutch types. (d) LS mean parental fitness efficiency (see text for calculations) of the three clutch types. All bars represent ± 1 SE.

In summary, offspring from clutches of large eggs had higher survival, metamorphosed at a larger size and in less time. Variable egg size clutches had an intermediate value for each dependent variable except larval period which was equal to small egg clutches. There was a positive, curvilinear, monotonic relationship between mean egg size and mean offspring size at metamorphosis (Fig. 4a). There was a significant difference in parental fitness efficiency between clutch types (Table 2d, Fig. 3d). Large egg clutches had a significantly higher parental fitness efficiency than variable-sized (LSD: P < 0·01) and small (LSD: P < 0·001) egg clutches. Clutches of large eggs had the highest parental fitness and clutches of small eggs had the lowest. There was a positive sigmoid relationship between mean egg size and parental fitness (Fig. 4b).

Figure 4.

(a) The offspring fitness curve. Logarithmic regression: F1,18 = 63·04, P < 0·001. (b) The parental fitness curve. Logistic regression: F1,18 = 75·22, P < 0·001. See text for calculation of eWp.


The results from this study provide direct evidence of a causal relationship between offspring provisioning strategies and offspring and parental fitness traits in natural conditions in the field. The reproductive effort a female invests, how this is partitioned, and the success or reproductive return of this strategy by accessing all surviving offspring were all measured directly in this study.

Offspring from clutches of large eggs were larger at metamorphosis, had a shorter larval period, and higher survival. Mortality was probably a result of intraspecific competition, predation (although no resident predators were identified in the experimental ponds) and desiccation due to metamorphosis not being completed before the pond dried. Fitness loss due to competition suggests that offspring from clutches of larger eggs had a competitive advantage and that ponds in the field are resource limited. The finding that offspring from larger eggs were at an advantage is further reinforced by examining the relationship between offspring size and egg size which is positive and increases monotonically (Fig. 4a). Since size at metamorphosis of anurans is a direct determinant of the probability to survival and reproduction (Altwegg & Reyer 2003), this relationship can be interpreted as the offspring fitness curve (Smith & Fretwell 1974).

In the study ponds, the most economical strategy resulting in the most efficient fitness return was for a female to produce clutches of large eggs. The shape of the parental fitness curve (Fig. 4b) shows that it had not peaked and declined (Smith & Fretwell 1974). This indicates that an optimal egg size had not yet been reached, and that this offspring environment is on the harsher quality end of the gradient of offspring environments that C. georgiana use for reproduction, and not optimal, as in laboratory rearing experiments examining maternal effects where benign conditions often do not reduce survival and fitness of offspring from small eggs (Reznick 1991). The fact that clutches of small eggs persist in the population suggests that these must also be an effective strategy in some offspring environments experienced by C. georgiana, as may be the case if the offspring environment becomes more favourable in quality.

To compare what would happen if the offspring environment became more favourable, we calculated parental fitness assuming offspring from all the clutches in our field experiment experienced a high quality, benign environment using the equation detailed in the methods. To simulate these conditions we used data from Dziminski & Roberts (2006) on survival and size at metamorphosis of offspring from large and small eggs, raised under benign laboratory conditions, with adequate food. We calculated M for each clutch used in our field experiment by multiplying the number of eggs produced by the mean size at metamorphosis for the corresponding clutch size category (large = 23·75 mg; small = 19·81 mg) and survival (0·96 for both). We used the slope (a = 10·098) and intercept (b = 4·649) from the logistic regression of these offspring sizes and their egg sizes reared in the laboratory. The value for E was from each corresponding field experiment clutch. These calculations are shown in Fig. 5. The mean for the variable-sized strategy could not be estimated accurately because the distributions of egg sizes were not strictly bimodal reflecting the above size categories. They would, however, lie somewhere in between the means of the small and large size categories, as represented by the bars, because these clutches contain some larger eggs than just the small size category but also some smaller eggs than the large egg size category. From our estimates of the parental fitness of the clutches used in the field experiment under a benign, higher quality, offspring environment, the relationship between clutch type and parental fitness becomes inverted (Fig. 5). The optimal strategy now becomes the production of smaller eggs. This is the result of the trade-off in egg size and number where the production of smaller eggs allows the production of more eggs and in a higher quality environment results in a greater return per unit of reproductive effort expended.

Figure 5.

Mean parental fitness efficiency of the three clutch types under benign laboratory conditions. Calculated from data from Dziminski & Roberts (2006). Bars for small and large eggs represent ± 1 SE, bars for variable-sized eggs represent the range in which the mean would exist.

Although not tested in our field experiment, such high quality offspring conditions may occur when ponds flood and offspring disperse into surrounding flooded habitat, releasing them from high density-dependent fitness loss and allowing access to more resources. Ponds flooding and offspring dispersing have been observed in normal years of higher rainfall, and within breeding sites there is also spatial variation where some ponds flood and others do not (M. Dziminski, pers obs.; unpublished data). However, during the experimental period, rainfall at the breeding site was low, sixth lowest annual rainfall on record (Station: Roleystone, WA; data since 1964; BOM 2008) and none of the experimental ponds flooded. Consequently, some of the experimental ponds dried, desiccating any remaining offspring (Table 2). Ponds flooding and offspring dispersing into higher quality habitat could provide the conditions under which producing clutches of small eggs may result in the highest parental fitness, as simulated by laboratory studies (Dziminski & Roberts 2006). Confirmation of this in the field would require the collection of newly metamorphosed offspring from known small clutches that had dispersed from flooded ponds. To gain values of parental fitness all surviving offspring would need to be accessed (as in this study), a feat that is close to impossible due to the substrate complexity of breeding sites (vegetation and leaf litter) and minute size of metamorphs of C. georgiana (20 mg metamorph approximately 7 mm snout-vent length).

In our experiment, clutches of variable-size eggs had a parental fitness intermediate between fitness of clutches of large and small eggs. It was not possible to determine the exact size of individual eggs that surviving offspring originated from. This would require individual marking of offspring at hatching which is again near to impossible with anuran tadpoles (Donnelly et al. 1994). Crinia georgiana tadpoles are only 7·0 mm snout to tail length at hatching, reaching 20·5 mm before metamorphosis (Main 1957). The advantage of the variable yolk allocation strategy could lie in ‘bet-hedging’ (Philippi & Seger 1989). If the parent is uncertain of the offspring environment, for example if the pond will persist, desiccate or flood, then a mixed size strategy would ensure survival and some parental fitness return even in poor quality environments. The variable strategy gives a higher fitness than producing just small eggs in poor quality ponds (Fig. 3d), but leads to increased parental fitness in a higher quality environment (Fig. 5) because of the trade-off in egg size and number and the production of extra, smaller eggs (Capinera 1979; Crump 1981; Kaplan & Cooper 1984). Theoretically, the maintenance of this strategy depends intrinsically on the frequency of good offspring conditions, and the magnitude of the difference in the quality of the offspring environments, spatially and temporally (McGinley et al. 1987). To address this, further fine scale surveys of the frequencies of confined, flooded and early drying ponds, and their predictabilities, at the breeding sites of C. georgiana is required.

As well as the amount of yolk in an egg, the quality of the yolk could also be important in determining offspring fitness. In birds and reptiles, differences in hormone concentrations in the yolk can affect offspring fitness (Rutkowska, Wilk & Cichon 2007; Uller, Astheimer & Olsson 2007; Cucco et al. 2008). Similarly, antioxidant levels in the yolk also have important effects on offspring fitness in birds (Saino et al. 2003). Nevertheless, our study shows that egg size per se clearly has important fitness consequences under natural conditions.

In this study we found that fitness consequences of offspring provisioning are clearly manifest in the larval stage, a stage in the life cycle where mortality is high and selection strongest. In natural ponds in the field that did not flood during the experimental period, clutches of large eggs resulted in the highest parental fitness. The ‘bet-hedging’ strategy of producing variable-sized eggs in clutches may become a benefit when the condition of the offspring environment cannot be predicted by the parent.

Our study illustrates the critical importance of moving experimental life-history studies out of the laboratory into the field where realistic selection pressures act on developing offspring. We have also demonstrated the need for an understanding of the larval environment and its influence on the fitness of variable offspring provisioning strategies.


This research was funded by grants to M. D. from the Society of Wetland Scientists, Australian Geographic, the Linnean Society of New South Wales and the School of Animal Biology, University of Western Australia. The Centre for High-throughput Agricultural Genetic Analysis and the State Agricultural Biotechnology Centre at Murdoch University provided facilities for molecular work. C. Christophersen, K. Munyard, K. Gregg, M. Beveridge, T. Garner, P. Van Eeden, D. Edwards, A. Hettyey, F. Brigg and D. Berryman provided extremely helpful advice, or assisted with field or laboratory work. Authors thank R. Black, D. Reznick, R. Shine and R. Kaplan for helpful comments on the initial manuscript. All animals were collected and maintained according to the standards of the Animal Ethics Committee of the University of Western Australia (approval number UWA 01/100/150) and the Department of Conservation and Land Management, Western Australia (permit numbers SF004187 and CE000319).