Niche differentiation occurred on 3 dimensions in a size-structured guild of marsupial carnivores (Marsupialia: Dasyuridae) that included the Tasmanian devil (Sarcophilus laniarius), spotted-tailed quoll (Dasyurus maculatus), and eastern quoll (D. viverrinus). Diet was partitioned on a body size–prey size axis, but substantial overlap occurred between adjacent species. Complementary niche differentiation also occurred on horizontal (vegetation type and structure) and vertical (arboreal) habitat dimensions. Eastern quolls separated completely by using grasslands more than the other 2 species. Spotted-tailed quolls were distinguished by a greater degree of arboreal activity that reflected a large proportion of arboreal prey species in their diet. A strong relationship was found between body size, arboreal activity, and morphologic adaptations for climbing between predators and prey. Coevolution in phenotype among predators and prey and consequent constraint on performance in different habitat types were the most easily measured explanations for habitat differentiation among these species, particularly between spotted-tailed quolls and devils. The “ghost of competition past” is an alternative and nonexclusive explanation that can equally explain arboreal habitat use by spotted-tailed quolls and separation between the eastern quoll and the spotted-tailed quoll. Risk of predation and prey availability also may be important factors.
Competition is an important force structuring the marsupial carnivore (Marsupialia: Dasyuridae) guild in Tasmania, Australia (Jones 1997; Jones and Barmuta 1998). This 3-species assemblage comprises the medium-sized Tasmanian devil (Sarcophilus laniarius; mean weight, males 8.4 kg, females 5.4 kg), the spotted-tailed quoll (Dasyurus maculatus maculatus; males 3.2 kg, females 1.7 kg), and the small eastern quoll (D. viverrinus; males 1.1 kg, females 0.7 kg; Jones and Barmuta 1998). The guild is size-structured such that no intra- or interspecific overlap in body mass occurs among adult males and females of any species. However, subadults of the larger of a species pair overlap in body weight with adult males of the next smaller species (Jones and Barmuta 1998). Competitive character displacement and equal spacing in canine tooth strength and temporalis muscle size, the trophic characters that most closely relate to the way in which these carnivores kill their prey, has occurred among sexes and species of quolls in Tasmania (Jones 1997). This suggests that competition has been important on an evolutionary time scale and has influenced size relationships among species. Current competition also is implicated in patterns of diet overlap and relative abundance and behavioral dominance at carcasses of large prey species (Jones 1995; Jones and Barmuta 1998). Some diet separation along a body size–prey size axis occurs, with equal spacing in mean prey size taken by each sex and species correlating with equal spacing in canine strength (Jones 1997). However, patterns of overlap in diet composition reveal 2 groups within which diets overlap: devils and male spotted-tailed quolls that eat large prey and female spotted-tailed quolls and eastern quolls that eat small prey (Jones and Barmuta 1998). Devils also are behaviorally dominant, displacing both species of quolls from carcasses of large prey in most instances (Jones 1995). The middle-sized species, the spotted-tailed quoll, may experience more interspecific competition, both exploitation and interference, than the other 2 species, and devils may experience the least (Jones 1995; Jones and Barmuta 1998).
In nearly all guilds of animals studied, niche differentiation occurs on several dimensions, and the number of dimensions increases with species richness (Schoener 1974, 1986). Niche differentiation is generally complementary; when species are similar on 1 dimension, they differ on another. Habitat is the most common niche parameter partitioned in animals, followed by food (Schoener 1986). Recent studies support the hypothesis of temporal partitioning of rapidly renewable resources (Kotler et al. 1993, 1994; Ziv et al. 1993), but the few predator studies have failed to confirm that differences in time of activity result in the taking of prey that are differentially available during different hours of the day (Schoener 1986).
Habitat selection is the result of a trade-off in performance between different habitat types. Performance is affected by morphologic and physiologic phenotype in relation to quality and type of requirements provided by each habitat type, risk of predation, and competition, including social dominance factors related to density (Brown 1990; Hobbs and Hanley 1990; Van Horne 1983). Habitat partitioning may result from present or past competition (Connell 1961, 1980). Coevolution between predators and prey or between competitors may result in a restricted niche and in evolution of phenotype, perhaps resulting in further habitat specialization (Brown 1990, 1996).
We explored niche differentiation among the sympatric dasyurid carnivores in Tasmania to determine if complementary niche partitioning occurred. We predicted that among age–sex classes of different species where diet overlap was high, partitioning would occur on ≥1 habitat dimensions. We examined alternative ecological forces and evolutionary phenomena that may explain habitat partitioning in this guild. If competition was an important force, we expected habitat separation to reduce competition for the same prey individuals or species. If habitat separation resulted from differential performance in different habitat types, we expected morphologic specialization of each predator to its preferred habitat type. To this end, we assessed tree climbing ability in devils. The potential influence of 2 alternative mechanisms for habitat separation, predation risk and prey availability, could not be tested directly in our study but was considered. We were particularly interested in examining species and population classes where a high degree of overlap occurred in body weight and diet and, therefore, potentially higher levels of competition occurred between, for example, subadult devils and adult male spotted-tailed quolls (Jones and Barmuta 1998).
Materials and Methods
We conducted this study in a 20-km2 area at the northern end of the Cradle Mountain–Lake St. Clair National Park in the central highlands of Tasmania, Australia, in 1991–1992. This site was especially suitable for studying habitat partitioning because it supported a mosaic of small blocks (100-m2 to 1-km2 diameter) of structurally and floristically different communities (Kirkpatrick and Mackie 1991). Five vegetation communities were defined (Jones 1998): rainforest (cool temperate rainforest dominated by myrtle beech, Nothofagus cunninghamii), wet eucalypt forest (Eucalyptus >35 m with a closed shrub layer at 2–6 m and a moss ground layer), dry eucalypt forest (20-m canopy with a closed, dense shrub layer from 0 to 2 m), mixed forest (mixed rainforest and eucalypt forest, usually dense at ground level), and grassland (low vegetation <1 m, either tussock grasslands of Poa, low herbfields, or buttongrass moorland dominated by hummock-forming Gymnoschoenus sphaerocephalus). The ratio of area of these habitat types was about 10:7:5:2:10, respectively.
We used spool-and-line tracking to assess habitat use. That technique provided detailed information on both horizontal and vertical use of the habitat and movement patterns at a resolution that was not possible to achieve with radiotracking. Center-wound spools of fine white thread (240 m long) were individually wrapped in plastic food wrap, strapped together in a single row with pharmaceutical sticking plaster, and the thread tied end to beginning between spools to form a continuous line. The plastic wrap served to prevent the thread from adhering to the sticking plaster and permitted free flow (Jones 1995).
We trapped devils and quolls in wire-cage traps (Jones 1995) baited with about 200 g of beef liver and covered with plastic against rain. Sixty traps were spread over 20 km2 to maximize capture success. Each trap was set for 3 nights every 2nd month from October 1990 to April 1993. Individuals selected for tracking, including only those that were accustomed to being trapped and handled (to reduce effects of stress on subsequent foraging behavior), were kept in situ. Tracking devices were glued to the fur on the flat, sacral region of the rump of the devil or quoll so that the running thread fed out clear of the hind legs, using small drops of quick-setting viscous glue. The slick guard hairs needed trimming for secure adhesion and long fur was trimmed to 0.5 cm in length to prevent the package swaying as the animal moved. After tying the end of the thread to a bush, the individual was released at point of capture.
Tracking devices were very light (<1% body mass) but were bulky. To prevent disturbance of normal behavior, number of spools and hence length of thread attached was limited to 5 spools (1,200 m) for individuals >3 kg in mass (a package 4 by 7 cm and 1.5 cm thick), 4 spools (960 m) for 2- to 3-kg individuals, 3 spools (720 m) for 1.5- to 2-kg animals, and 2 spools (480 m) for individuals <1.5 kg. Individuals carrying spool-and-line tracking devices were observed immediately, at 30 min (1 eastern quoll and 1 spotted-tailed quoll for 1 min each), 1.5 h (1 spotted-tailed quoll for 10 min), and 3 h (1 devil for 15 min) after attachment of the devices. No individuals looked at, sniffed, or touched the devices during those observations, and none were chewed or removed before the thread was finished. Remains of the package were always chewed off in 1–2 days. Sightings of individuals carrying spool-and-line tracking devices and radiotracking of eastern quolls and devils (Jones et al. 1997) indicated that the devices lasted for about 1 and 3.5 h for 480-m and 1,200-m devices, respectively.
Sampling and statistical analysis
Comparisons were made among adults of the 3 species and among species and population classes (adult and subadult devils and male spotted-tailed quolls) where interspecific body size overlap occurred. Adult males of all species were classed as full-grown adult if their body weight exceeded 8.0 kg for devils, 2.6 kg for spotted-tail quolls, and 1.0 kg for eastern quolls. Adult females were defined as: 5 kg for devils, 1.5 kg for spotted-tailed quolls, and 0.5 kg for eastern quolls. Subadult devils were independent juveniles that were less than adult weight. For statistical independence, we sampled each individual only once. Habitat data and spools were collected within 2 days because they became progressively difficult to follow as other animals moving through the area broke the thread. The entire distance, starting at the release point, was used for data collection because we were particularly interested in the use of and behavior on forest edges, which was where most traps were placed. Because only individuals that were accustomed to being regularly trapped were used, many did not flee from the trap site and most seemed to recover within 30 m. Every 10 m (estimated by pacing) along the length of the thread, data on vegetation type, movement pattern, visibility, slope, and arboreal movement, representing coarse- and fine-scale components of both horizontal and vertical aspects of the habitat, were recorded (Table 1). Data for each individual were summarized as percentage of the total number of recorded points in each habitat type, percentage of records where the thread was in a foraging pattern in each habitat type, mean of the rank scores for visibility and slope, number of trees climbed and number of logs encountered per kilometer of thread, percentage of logs encountered that were climbed, percentage of distance traveled above the ground (on logs or in trees), and percentage of the total distance of the thread where the individual was ≤15 m from a forest edge.
We constructed 2 polytomous stepwise logistic regression models to identify differences among species and age–sex classes in habitat use (Hosmer and Lemeshow 1989; SPSS Inc. 1997). The 1st model separated eastern quolls (coded 1), spotted-tailed quolls (coded 2), and devils (coded 3); the 2nd model separated adult devils (coded 1), subadult devils (coded 3), and male spotted-tailed quolls (coded 2). We used a hierarchical approach in model construction. Each variable was initially subjected to univariate analysis, and those with univariate probabilities of <0.25 were entered into the model (Hosmer and Lemeshow 1989). In the species analysis, complete separation was found between eastern quolls and the other 2 species on 1 variable (grassland). Because it was of interest to consider roles and interactions of other variables in separating the species and consequent competing models in the absence of the known strong variable (grassland), we analyzed separation between the 3 species on the remaining variables. Stepping was conducted interactively, with tolerance set at 0.01 to control for multicollinearity. We examined fit of models using log-likelihood ratios and McFadden's rho-squared (Hosmer and Lemeshow 1989; SPSS Inc. 1997). We retained variables with P < 0.25 in the model or if their inclusion made a significant difference to the fit of the model and values of the coefficients (Hosmer and Lemeshow 1989). To determine differences between species or population classes in variables that remained in the model, we used separate Mann-Whitney U-tests (SPSS Inc. 1997) on each of the 3 species pairs. To adjust probabilities for the set of related tests for each variable, we applied the Bonferroni procedure (Sokal and Rolf 1981) that assumed complete correlation between tests and the Rice procedure (Rice 1990) that assumed complete independence. In practice, the difference between the 2 procedures was minor.
To examine for a correlation between the extent of arboreal activity and proportion of arboreal species in the diet, we extracted the diet composition data for arboreal prey species from a concurrent diet study (Jones and Barmuta 1998). We conducted a Pearson correlation analysis (SPSS Inc. 1997) on total percent biomass of arboreal prey species (both mammals and birds) and percent distance spent above the ground for adult and subadult devils, adult and male spotted-tailed quolls, and adult eastern quolls.
Devil climbing ability
To evaluate climbing capability among devils, we devised hair sampling tubes (Suckling 1978) for devils, which were constructed from PVC pipe (11 cm in diameter, 15 cm long). Bait (raw or cooked liver mixed with bacon) was placed behind a metal mesh by use of a screw cap at one end. Double-sided adhesive tape was placed in 4 wide strips inside the open end to collect hairs from devils that had reached the tube. We wired tubes 0–4 m above the ground on trunks of 11 different trees (all eucalypts, Eucalyptus delegatensis, except 1 tea tree, Leptospermum) in an area where regular food provisioning ensured devils would find tubes. Different climbing situations were tested, ranging from vertical to 50° sloping trunks, with diameters ranging from 15 to 60 cm, and both with and without small side branches. We checked hair tubes at daily (within field trips) and weekly (between field trip) intervals for 2 months.
We spooled 24 devils (6 adult male, 9 adult female, 9 subadult), 7 spotted-tailed quolls (5 adult male, 2 adult female), and 7 eastern quolls (4 adult male, 3 adult female). Subadult spotted-tailed quolls were too rarely trapped to enable study of their overlap with adult male eastern quolls. Variables that best separated the 3 species were use of (percentage of records in) grassland, use of rainforest, extent of arboreal activity, and visibility in the 1st meter above the ground (Fig. 1; Table 2) Complete separation occurred in the use of the grassland habitat between eastern quolls and both spotted-tailed quolls (U = 49.0, P = 0.002) and devils (U = 105.0, P < 0.001) (Fig. 1; Tables 3 and 4). As a result, that variable was not entered in the model. No other differences in habitat use were found between eastern quolls and devils. Eastern quolls further separated from spotted-tailed quolls on the percentage of records in rainforest (eastern quolls had fewer records in rainforest; U = 3.0, P = 0.005), visibility (eastern quolls used habitats with greater visibility; U = 9.0, P = 0.047), and extent of arboreal activity with number of logs climbed approaching significance (eastern quolls fewer; U = 10.0, P = 0.063; Fig. 1; Tables 3 and 4). Spotted-tailed quolls were quite arboreal, much more than either eastern quolls (U = 100.0, P = 0.001) or devils (U = 102.0, P < 0.001). The extent of arboreal activity was the only variable that separated spotted-tailed quolls and devils, although results for both the percentage of total records in rainforest (spotted-tailed quolls spend more time in rainforest; U = 80.0, P = 0.052) and visibility approached significance (spotted-tailed quolls use habitats with lower visibility = higher visibility score; U = 77.0, P = 0.084; Fig. 1; Tables 3 and 4).
Means and SE (vertical bars) of proportions of records in rainforest, grassland, or arboreal (on logs or in trees) and visibility scores for sympatric eastern quolls, spotted-tailed quolls, and devils at Cradle Mountain, Tasmania, 1991–1992
Number of logs encountered per kilometer was retained in the model because its inclusion made a significant difference to the fit of the model and the estimated values of the coefficients (Hosmer and Lemeshow 1989), although statistical results for this variable were not significant (Table 2). The final model had a good fit (log-likelihood ratio = 24.66, d.f. = 6, P < 0.001, McFadden's rho-squared = 0.414) and correctly predicted 63% of all records, 54% of eastern quoll data, 66% of spotted-tailed quolls, and 65% of devils. Several variables were collinear with variables that were retained in the model. Arboreal use of habitat (percentage of data points where the animal was on a log or in a tree) correlated with use of rainforest (percentage in rainforest). Whether or not the animal was in a foraging movement pattern also was correlated in each habitat.
Only 1 variable, percentage of records arboreal (on logs or in trees), remained in the model constructed to describe habitat separation between adult devils, subadult devils, and male spotted-tailed quolls (Table 2). The final model had a good fit (log-likelihood ratio = 16.59, d.f. = 2, P = 0.00025; McFadden's rho-squared = 0.284) and correctly predicted 66% of adult devils, 41% of subadult devils, and 61% of male spotted-tailed quolls, which is 57% correct predictions. Only the difference between adult devils and male spotted-tailed quolls (spotted-tailed quolls more arboreal; U = 72.0, P = 0.003) was significant (Fig. 2; Tables 3 and 4). Arboreal habitat use of subadult devils was intermediate between adult devils (U = 104.0, P = 0.029) and male spotted-tailed quolls (U = 36.0, P = 0.072).
Means and SE of proportion of records arboreal for sympatric male spotted-tailed quolls, subadult devils, and adult devils at Cradle Mountain, Tasmania, 1991–1992
No species used the forest–grassland ecotone extensively; most activity was within either habitat. Of the locations in forested habitats (rainforest, wet and dry eucalypt, and mixed forest), only 8.8% (range = 0–46, SD = 12.5) of observations for devils, 17.2% (range = 0–41.2, SD = 15.3) of observations of spotted-tailed quoll, and 6.3% (range = 0–12.8, SD = 4.4) of observations of eastern quoll were within 15 m of the forest edge. Of observations in open habitats (grassland, buttongrass moorland), 4.9% (range = 0–16, SD = 6.0) of devil locations, 6.1% (range = 0–15.1, SD = 6.5) of spotted-tailed quoll locations, and 18% (range = 0–42.6, SD = 14.9) of eastern quoll locations were ≤15 m from the forest edge.
Arboreal activity and diet
The diet of adult male spotted-tailed quolls contained almost twice the proportion of arboreal prey species by biomass (31.7%) as adult devils (16.2%; Table 5). That difference was consistent across mammalian and avian categories. The diet of adult female spotted-tailed quolls was higher in arboreal species (51.4%) than adult eastern quoll diet (37%); a higher proportion of birds in the diet accounted for most of the difference. Diets of adult male spotted-tailed quoll comprised twice the proportion of arboreal species (63.7%) than the diet of subadult devils (35.8%) during the months of the year (February–July) in which they overlapped in body weight (Table 5; Jones and Barmuta 1998). Most of the difference was in the proportion of arboreal mammals, which comprised most of the arboreal component for both species. During winter (July–October), female devils had 10% more arboreal species in their diet than male spotted-tailed quolls, with which their diet significantly overlapped. Female devils consumed more (31.1%) and male spotted-tailed quolls consumed less (15.7%) arboreal mammals during this period.
Tests of climbing ability of devils showed that devils are both willing and capable of climbing large trees (trunk diameter >40 cm; large trees generally do not have small side branches) if the trunk was leaning at an angle of 60–70° or less (deep scratches in the bark indicated difficulty in climbing an 80° slope) and vertical small diameter trunks (trunk diameter <25 cm) with small side branches to a height of 2–3 m. During those trials, devils found (scratch marks as evidence of attempts at jumping) but did not climb small vertical tree trunks (diameter < 25 cm) that had no side branches, although adult devils (n = 3) carrying spool-and-line tracking devices independently climbed (without bait) vertical trunks of 10- to 25-cm diameter, lacking side branches and with smooth bark, to a height of 2.5 m. Subadult devils (n = 4) were recorded climbing bushy, many-branched shrubs to 4 m and a leaning trunk to 7 m. Spotted-tailed quolls (n = 4) carrying spool-and-line tracking devices climbed 8.5 m up large-diameter, steeply angled (50–70°) tree trunks with no lower branches. Eastern quolls carrying spool-and-line tracking devices were recorded scrambling over low fallen timber but not climbing trees. They are capable of balancing on limbs but are not agile climbers.
Niche differentiation occurs on at least 3 dimensions among sympatric dasyurid carnivores in Tasmania, in accordance with Schoener's (1974, 1986) predictions. Eastern quolls clearly separated from the other 2 species on vegetation type, traveling and foraging in grasslands to a greater extent and in rainforest to a lesser extent. Spotted-tailed quolls separated from eastern quolls and devils in their greater use of arboreal habitats. Spotted-tailed quolls also had a tendency to use rainforest more and habitats with lower visibility in the 1st meter above the ground than eastern quolls or devils. The ontogenetic dynamics of habitat differentiation in devils parallel those found in diet (Jones and Barmuta 1998). Both arboreal habitat use and diet of subadult devils were intermediate between those of adult devils and male spotted-tailed quolls.
Niche differentiation was complementary. Where diet overlapped significantly, such as between adult male spotted-tailed quolls and adult devils and between adult female spotted-tailed quolls and adult eastern quolls (Jones and Barmuta 1998), habitats were partitioned. Horizontal spatial separation resulted in complementary niche partitioning because the 2 species are unlikely to encounter the same prey individuals. Vertical habitat partitioning was pronounced, both between devils and adult male spotted-tailed quolls and between spotted-tailed quolls and eastern quolls. Spotted-tailed quolls were more arboreal than devils and eastern quolls, and that was reflected in a higher proportion of arboreal prey species in diet of the spotted-tailed quoll. A detailed breakdown of the pattern of diet overlap between male spotted-tailed quolls and devils reveals that, although overall diet overlap is significant, male spotted-tailed quolls consume almost twice the proportion of arboreal prey than do subadult devils for the 6 months in which these 2 population–species classes overlap in body weight (Jones and Barmuta 1998). Although subadult devils were intermediate between adult devils and spotted-tailed quolls in their use of the arboreal habitat, like adult devils, they lack any specialized adaptations for climbing and their climbing abilities do not match that of spotted-tailed quolls. Inconsistent with those patterns of arboreal habitat use and diet, male spotted-tailed quolls actually had a slightly lower arboreal diet component than female devils during the 4-month period when overall diet overlapped significantly (Jones and Barmuta 1998). In general, however, the greater arboreal activity of spotted-tailed quolls correlated with a higher arboreal prey component in the diet. This indicates that vertical habitat partitioning promotes encounters with different prey individuals, which would function to reduce competition and therefore represents niche complementarity.
Could exploitation or interference competition for food be important in driving habitat separation among these dasyurid carnivores? The clear spatial separation, both horizontal and vertical, between spotted-tailed quolls and eastern quolls, which overlap substantially in diet, would certainly reduce competition for the same prey individuals and may represent the “ghost of competition past” (Connell 1980:137). Although evidence for horizontal spatial separation between devils and male spotted-tailed quolls is weak, the different proportions of arboreal species in the diet of spotted-tailed quolls and devils suggested that the strong vertical habitat separation resulted in a reduction in competition for the same prey individuals. Observations suggested that spotted-tailed quolls, which climb trees of all sizes, move across the canopy (Burnett 2000), and catch arboreal possums and birds in trees, are more adept at tree climbing than are devils. Although subadult devils have been reported catching birds at roost (Guiler 1983), they have more limited climbing abilities. Yet, spotted-tailed quolls are not able to monopolize the arboreal prey resource. The major arboreal prey species, the ringtail possum (Pseudocheirus peregrinus), and brushtail possum (Trichosurus vulpecula), and many birds, also frequent the ground and lower branches, where they are available to and may be ambushed by devils or eastern quolls.
Spotted-tailed quolls stand a significant risk of losing kills to devils because they kill prey too large to consume quickly. Vertical separation in habitat use may reduce this sort of interference competition if spotted-tailed quolls consume their prey in trees, although no evidence of this exists. Two observations of quolls hunting in trees reported the prey brought to the ground (A. Watt and P. Bell, pers. comm.). Most logs, which are used extensively by spotted-tailed quolls, also are within climbing reach of devils. Interference competition from spotted-tailed quolls or devils is unlikely to be an issue for eastern quolls because most of their diet comprises small prey items that can be devoured quickly (Jones and Barmuta 1998).
Overlap in vertical habitat use and diet (Jones and Barmuta 1998) between subadult devils and adult devils, and subadult devils and male spotted-tailed quolls, does have implications for current competition and its effects on survival of young devils. Overall, spotted-tailed quolls experience higher levels of diet overlap than do devils, and we suggested that this may explain the low population density of spotted-tailed quolls (Jones and Barmuta 1998). Still, subadult devils would be competing with high population densities of adult devils. Intraspecific competition among devils is intense, with adults dominating subadults (Jones 1995). Adults also are likely to be competitively superior in terms of hunting experience, which affects prey catchability. Prey availability may be more important for survival and thus recruitment to the population of young devils than for young spotted-tailed quolls or eastern quolls, in which habitat does not overlap significantly with another species and intraspecific competition is less (Jones 1995).
Habitat selection of each species may be strongly influenced by morphologic constraint. The correlation between body size, arboreal adaptations, and arboreal habitat use of both dasyurid predators and their mammalian prey suggests that phenotypic coevolution between predator and prey, resulting in morphologic specialization to this habitat and subsequent constraint on performance in different habitat types (Brown 1990, 1996) might have occurred in this guild. The smallest (eastern quoll) and the largest (devil) predators and their respective primary prey species, small mammals (rodents and Antechinus) and medium- to large-sized mammals (mostly macropods and wombats, Vombatus ursinus), respectively (Jones and Barmuta 1998), are all largely terrestrially active and lack special adaptations for climbing. Both spotted-tailed quolls and possums, the middle-sized predator and prey, are arboreally adapted and active. Like possums, spotted-tailed quolls have ridges on the pads of all their feet and a well-developed hallux, both of which aid climbing. In the case of the spotted-tailed quoll, the predator may have evolved to follow their possum prey up trees, an opportunity that the heavy-bodied devil, their competitor, could not exploit.
The spotted-tailed quoll has specialized morphologically to an arboreal habitat, which means that this species is less adept at foraging on the ground. The spotted-tailed quoll has limb ratios indicative of a slow-running, ambush predator of closed habitats (Jones and Stoddart 1998). It is well adapted to foraging in trees and the rainforest habitat it uses at Cradle Mountain, which usually has large numbers of logs and log tangles on the ground. Likewise, the devil, a heavy-bodied predator, with a massive skull and jaw musculature, is adapted to the niche of predator of large terrestrial prey and scavenger. Even young devils, which are the same body weight as a male spotted-tailed quoll, cannot climb well. Their limb ratios indicate faster running speeds, typical of pounce–pursuit predators of moderately open habitats (Jones and Stoddart 1998). Devils use wet eucalypt forest that has a clear understory with high visibility and few obstructions such as logs. Eastern quolls also have limb ratios that are indicative of faster running speeds and moderately open habitats (Jones and Stoddart 1998), so they are well adapted to the open grasslands.
Complementary niche differentiation occurred among the 3 species of dasyurid carnivores at Cradle Mountain. Some separation occurred in diet (prey species and prey size; Jones and Barmuta 1998), but where diet overlapped significantly, the species partitioned habitat both horizontally (vegetation type and structure) and vertically (arboreal). As has been found among otters (Lutra canadensis) and mink (Mustela vison) in Alaska (Ben-David et al. 1996), morphologic and physiologic constraints on the ability to forage in different habitats may explain a substantial part of the habitat differentiation in this guild. This is further suggested by the maintenance of habitat associations at Cradle Mountain in parts of Tasmania where environmental correlates of distribution favor spotted-tailed quolls and not eastern quolls or devils, reducing competitive pressure on spotted-tailed quolls (M. E. Jones and R. K. Rose, in litt.). However, arguments of coevolution and competition are not exclusive. The “ghost of competition past” also would explain arboreal activity of spotted-tailed quolls and may promote habitat partitioning between eastern and spotted-tailed quolls. Current competition may play a role in maintaining these patterns.
Predation risk and prey availability also may play currently little understood roles in habitat separation in this guild. Predation risk may foster niche differentiation if the risk of mortality and, therefore, fitness varies between habitat types. Eastern quolls are at greater risk of predation than spotted-tailed quolls or devils by virtue of their small body size. Their antipredator behavior in response to devils (Jones 1998; Smith 1999) and the clear habitat separation between eastern quolls and devils, an aggressive competitor and potential predator, suggests that predation risk is a potential influence on habitat selection by eastern quolls. Differential availability of prey in different habitats also may drive niche differentiation. However, available evidence provides little support for this hypothesis. Although the arboreal habitat preferences of spotted-tailed quolls are consistent with availability of their arboreal prey, open habitat preferences of eastern quolls do not parallel the fairly restricted forest distribution of most of their prey species at Cradle Mountain (Watts 1994), or the generally higher densities of small mammals in forests in temperate Australia (Fox 1985; Godsell 1983). Likewise, 2 of the 3 major prey species of the forest-dwelling devil, Bennett's wallabies (Macropus rufogriseus) and wombats, forage primarily in open habitats, although daytimes shelters and dens are in forest.
We thank M. Ben-David, M. Stoddart, C. Dickman, and H. Kruuk for helpful comments on drafts of this manuscript. We are grateful to the Tasmanian Parks and Wildlife Service staff at Cradle Mountain, who assisted us in innumerable ways. This study was made possible by grants from the National Geographic Society (D. M. Stoddart and M. Jones), the Ingram Trust (M. Jones), and an Australian Postgraduate Award to M. Jones.
2000. The ecology of a suite of tropical Australian rainforest predators with particular reference to the spotted-tailed quoll, Dasyurus maculatus gracilis.Ph.D. dissertation, James Cook University of North Queensland, Townsville, Queensland, Australia.