Dietary flexibility and high predator efficacy facilitate coexistence in a novel predator interaction

Abstract

Predator diet can be influenced by competition and intraguild predation, leading to resource partitioning and/or avoidance. For sympatric, endemic predators, these processes form as predator species coevolve, facilitating coexistence. However, when novel predator interactions occur, significant dietary overlap could create acute levels of competition leading to intraguild predation and population extinction, or accelerated changes in diet and/or spatial and temporal avoidance. We measured diet, intraguild predation, and spatial and temporal overlap in two predator species in a novel predator interaction: the western quoll (Dasyurus geoffroii), a small, native carnivore reintroduced to semi-arid Australia, and the domestic cat (Felis catus), a larger introduced carnivore already resident at the release site. Both species exhibited high dietary overlap and fed on mammals, reptiles, birds, and invertebrates. Cats included quolls in their diet. Quoll diet was broader (including carrion, bats, and plant material) and flexible, changing significantly with age, sex, and season. Introduced rabbit was the most common prey item recorded for both species (frequency of occurrence = 40–50%). However, quolls consumed rabbits in relation to their availability while rabbit consumption in cats was unrelated to availability suggesting a stronger dependence on rabbit prey. Quoll diet did not change over time since release and they did not spatially or temporally avoid cats. However, cats were significantly spatially associated with rabbits while quolls were not, suggesting higher predation efficacy in quolls possibly due to their smaller body size enabling them to catch rabbits inside warrens. Despite high dietary overlap and intraguild predation, the quoll’s broad and flexible diet and high predation efficacy appeared to assist in facilitating coexistence and reducing competition in this novel predator interaction. This dietary flexibility may be harnessed to improve conservation outcomes: reducing introduced rabbits in our study area could naturally reduce feral cat populations while having less impact on native quolls.

carnivore, cat, competition, diet, reintroduction, western quoll

Dietary interactions between predator species are an important aspect of predator ecology and can include resource partitioning, interference or exploitative competition, and intraguild predation (Creel and Creel 1996; Mills and Gorman 1997; Fedriani et al. 2000; Glen et al. 2011). For native predators, dietary flexibility and interactions have formed through coevolution with sympatric predators under the same environmental conditions, facilitating coexistence (Glen and Dickman 2005). However, when novel interactions occur between predator species, such as when invasive predators arrive or when a native predator is reintroduced to an area with exotic predators, any significant dietary overlap is likely to create extreme levels of competition. The incidence of novel predator interactions is increasing due to anthropogenic impacts and more intensive conservation management. Understanding novel predator interactions can assist with predator management and provide a unique insight into predator ecology.

Optimal foraging theory suggests that animals should forage to maximize their energy intake and thereby increase their fitness advantage (Pyke et al. 1977). Foraging efficacy can be highest when predators forage in areas where their preferred prey is abundant to increase encounter rates, reduce search effort, and optimize hunting success. However, when multiple predators are present in an ecosystem, competition may lead to intraguild predation or may influence diet and habitat selection by forcing smaller predators to forage in poorer quality areas or feed on suboptimal prey species (Mills and Gorman 1997; Durant 1998; Fedriani et al. 2000; Linnell and Strand 2000). Smaller predators may use spatial or temporal avoidance to minimize interactions and competition with larger predators. In novel predator interactions, these strategies may be accelerated to avoid competition and intraguild predation. Diet of the subdominant predator also could change over time since release as they attempt to reduce competition with resident predator species.

Aside from interspecific interactions, many external factors can influence the diet of a carnivore, including: season (Davidson et al. 2013), habitat structure (Hebblewhite et al. 2005), pack size (Packer et al. 1990; Loveridge et al. 2006), or prey breeding season (Davidson et al. 2013). Intraspecific differences in predator diet also can occur that are related to predator sex (Marlow et al. 2015), age (Litvaitis et al. 1986), experience (Estes et al. 2003), and body size (Kutt 2012; Moseby et al. 2015). Feline predators are particularly flexible, exploitative, and opportunistic. Cheetahs (Acinonyx jubatus) vary in prey selection based on sex (Cooper et al. 2007), mountain lion (Puma concolor) on reproductive status and age (Ross et al. 1997), and domestic cats (Felis catus) on sex and body size (Kutt 2012; Moseby et al. 2015).

We studied a novel predator interaction in semi-arid Australia, where a native carnivore, the western quoll (Dasyurus geoffroii), was reintroduced to an area where an introduced carnivore, the domestic cat, was resident. The western quoll is a 1–2 kg marsupial predator with a generalist diet that has been absent from the region for over 100 years. In the arid zone, the diet of quolls included live mammals, carrion, lizards, frogs, and invertebrates (Johnson and Roff 1982; Burbidge et al. 1988). Domestic cats were introduced to Australia in the 1800s with Europeans (Koch et al. 2015) and have not coevolved with western quolls. Cats are larger than quolls (2.5–6.5 kg; Menkhorst and Knight 2011), prefer to hunt live prey, and consume a wide range of mammals, reptiles, and birds (Read and Bowen 2001; Doherty et al. 2015). There is potential for high dietary overlap between the two species, which could create exploitative or interference competition, leading to possible intraguild predation and population extinction of quolls, or dietary changes and temporal and/or spatial avoidance of cats by quolls as the subdominant predator. We investigated the potential for competition by measuring the diet of each predator and the degree of dietary overlap. We then measured changes in quoll diet after release, and temporal and spatial overlap.

We hypothesized that there would be high dietary overlap between quolls and cats and that this would lead to (i) intraguild predation, (ii) changes in quoll diet over time since release in order to reduce competition with cats, and (iii) that quolls would avoid cats or areas where major cat prey items were in high abundance, to reduce intraguild predation and interference competition from larger feral cats. Quolls exhibit sexual dimorphism with males averaging 1.3 kg and females 0.9 kg (Serena and Soderquist 1988). Because of this, we expected that quoll diet also would differ between the sexes.

Materials and Methods

Research on live animals followed ASM guidelines (Sikes et al. 2016) and was approved by the South Australian Wildlife Ethics Committee approval numbers 52/2013 and 8/2017 and University of Adelaide ethics 17666.

Study region.

—The Ikara-Flinders Ranges National Park (I-FRNP) is a 91,840 ha conservation reserve situated within an area of continuous semi-arid habitat. The majority of land within the national park had not been subjected to direct vegetation clearance for agriculture but all the valleys and low hills were grazed for over 100 years by domestic stock (primarily sheep) until 1984, when the release area for the quolls was proclaimed a national park. Despite the removal of domestic stock more than 35 years ago and the subsequent partial control of feral goats (Capra hircus), rabbits (Oryctolagus cuniculus), and over-abundant kangaroos, the history of degradation has limited the recovery of understory plant cover in many areas of the park. Continued heavy grazing by kangaroos and rabbits has contributed to this lack of recovery. The national park supports a variety of habitats including rocky ranges and gorges, river red gum (Eucalyptus camaldulensis) open woodland and creeklines, Triodia hummock grasslands, chenopod shrublands, undulating Callitris (native pine) woodland, mallee woodlands, and patches of rough-barked coolibah (Eucalyptus intertexta). The foothills support mixed Acacia, Dodonaea, and Senna shrublands with thicker understory (Brandle 2001). The higher elevation of the ranges produces higher rainfall than the surrounding arid plains (Brandle 2001) with Wilpena, within the park, averaging 471 mm per annum, and Blinman, approximately 20 km north of the park, averaging 307 mm (Bureau of Meteorology [BOM] 2020).

Quoll reintroduction.

—A total of 93 quolls were reintroduced to the I-FRNP between 2014 and 2016 with 41 released in 2014, 37 in 2015, and 15 in 2016. Releases occurred in April or May each year and were timed to avoid periods when quolls had pouch young. Most (89) of the quolls were sourced from the wild in Western Australia, with four captive-bred individuals sourced from the Alice Springs Desert Park. Details of the release can be found in Moseby et al. (2021).

Quoll scat collection and analysis.

—Quoll scats were collected opportunistically and through targeted trapping between 2014 and 2018. Targeted collection occurred by trapping radiocollared quolls at regular intervals after release and biannual cage trapping in March and December of each year throughout the release area. During these trapping events, quolls were captured using two bait types: rabbit, and a bait mixture of peanut butter, rolled oats, meat meal, and sardines. Scats were collected from the floor of the cage trap following capture, placed into individually labeled Ziplock bags (S. C. Johnson & Son, Inc., Racine, Wisconsin), and stored in a freezer on site. Only scats collected from traps with the peanut butter bait were used in scat analysis because the source of any rabbit remains found in the scat could be wild caught or from ingestion of bait. Scats from cage traps could be identified as belonging to an individual quoll, therefore data on sex and age also were collected. Age was estimated by means of tooth wear, body weight for first year juveniles (less than 600 g), and time between successive captures after first capture. The time since release (months) also was calculated for each scat from known individual quolls. Scats also were collected opportunistically from outside den sites and from latrines discovered during radiotracking. These scats could not be confidently assigned to an individual and were only used for seasonal diet analysis if they were fresh when collected.

We sterilized samples by oven-drying at 60°C for 24 h before placing them in individual fabric bags (30 denier hosiery stocking) that were then soaked in water and detergent for up to 24 h prior to being rinsed under running water. Once dry, the remains (hair, teeth, bones, scales, claws, feathers, exoskeletons etc.) were sorted macroscopically and with the aid of a dissecting microscope. To identify mammals to the lowest possible taxonomic level, a representative sample of all hair in each scat was examined using cross-sections and whole mounts under a compound microscope (10× and 40× lenses). Identification of hair was made using a reference collection and the hair analysis software package “Hair ID—An Interactive Tool for Identifying Australian Mammalian Hair” (Brunner and Triggs 2002). We identified nonmammalian prey more broadly as invertebrates, birds, or reptiles, and all plant material was classed as vegetation. Quoll hairs are commonly found in their scats, but we chose to remove any records of quolls as a dietary item because it is difficult to distinguish between grooming and cannibalism. Unrealistic dietary items such as small rocks and hessian (which had either been consumed during trapping or had adhered to the scat after it was deposited) were not included in the analysis. The composition of quoll diet was assessed by estimating the percentage volume of prey remains from each major prey category in each scat (to the nearest 5%).

Cat diet.

—Feral cats were shot and leg-hold trapped opportunistically within the release area between 2014 and 2019. Cats also were captured in cage traps but because rabbit carcasses were used as bait, these diet records were not considered in this study. Stomach contents of cats that were shot or leg-hold trapped were removed and examined for content. The presence or absence of invertebrates, small mammals, rabbits, reptiles, and birds was recorded (see Moseby et al. [2020] for methods). The sex of cats also was recorded.

Dietary diversity.

—All analyses were carried out in the R version 3.6.1 software environment for statistical and graphical computing (R Core Team 2017). We converted the data into frequency of occurrence (FO) by dividing the number of detections for each species by the number of scats collected, and calculated the diversity of all species in the diet of quolls and cats using Shannon’s diversity index, H (Shannon 1948):

$${\displaystyle \begin{array}{l}{\displaystyle H=-{\displaystyle \underset{i=1}{{\displaystyle \sum ^S}}}{p}_i\ln p_i}\end{array}}$$

where S is the total number of species recorded in the diet, and p_i is the FO of species i (i.e., sum of all p_i for a given site equals 1).

To consider how sample size affected the dietary diversity, we used resampling to construct diversity rarefaction curves independently for each season (quolls) and overall (cats). This approach is widely used to estimate the minimum number of dietary samples required to describe the diet adequately, which is indicated by a negligible increase in H above a certain sample size (i.e., when the rarefaction curve reaches an asymptote). We fitted a four-parameter asymptotic regression model to the resampled data, and defined adequate sampling as that required to reach ≥95% of the asymptotic H.

Rabbit, cat, and quoll activity index.

—Three camera transects each comprising 10 cameras (Reconyx Hyperfire 600; Reconyx, Holmen, Wisconsin), spaced at least 1 km apart, were established in the study area prior to the quoll release. Cameras were set at approximately 50 cm from ground level along linear features such as creeks and along vehicle tracks. Cameras were set at a 22° angle to the linear feature and the resulting images analyzed monthly. Cameras were set continuously from April 2014 to June 2018 inclusive. Photos of individual quolls, rabbits, and cats that were more than 10 min apart were considered separate detections. Monthly detection rates were calculated as the total detections of each species per month divided by the total camera trap nights multiplied by 100.

Statistical analysis.

—We tested the effect of season, rabbit activity, time since release, and sex, on the FO of prey items in the diet of quolls using binomial generalized linear models (GLMs). Separate, simple, binomial models were built depending on the hypothesis being investigated, with all significant terms retained in the model. We tested whether rabbit activity had an effect on the diet of quolls and cats by building a binomial GLM that included the interaction between predator and rabbit activity index. Model fit was assessed by R² values and evaluation of the deviance explained by the fitted model. For this model, only rabbit activity data from cameras where both quolls and cats were detected were included (n = 20) to enable a direct comparison between the predator species.

Dietary, spatial, and temporal overlap—quolls and cats.

—We calculated the dietary overlap between quolls and cats using Pianka’s index (O_jk) implemented via the “EcoSimR” package (Gotelli et al. 2015):

$${\displaystyle \begin{array}{l}{\displaystyle O_{jk}=\frac{{\displaystyle \sum _i^n}{p}_{ij}{p}_{ik}}{\sqrt{\left({\displaystyle \sum _i^n}{p_{ij}}^2{\displaystyle \sum _i^n}{p_{ik}}^2\right)}}}\end{array}}$$

where p is the FO of species i in the diet of quolls j and cats k (Pianka 1973). Pianka’s index varies between 0 (total separation) and 1 (total overlap). Given the difficulty in discerning the species of reptile in quoll scats and conversely, the relative ease of reptile species identification in cat stomachs, we chose to use only the broader “reptile” metric when undertaking our overlap analysis.

To examine temporal activity patterns of quolls, cats, and rabbits, camera detection records were recorded per hour following the methods outlined in Ridout and Linkie (2009). We estimated the activity pattern of each species using a kernel density estimation, then estimated the degree of overlap between the two estimated densities using a coefficient of overlap “Δ”. To calculate the amount of overlap in activity between species, we used the Δ ₄ estimator, which is suitable for large sample sizes (>75), in the package “overlap” and is suitable for circular distributions (Meredith and Ridout 2017). The overlap measure ranges from 0 (no overlap) to 1 (complete overlap). We calculated 95% confidence intervals (CIs) from 10,000 bootstrap samples.

Spatial overlap between cats and quolls at the release site has been explored previously and quoll occupancy was not found to be related to cat occupancy (Moseby et al. 2021). Spatial overlap was measured by using a linear mixed effects model using the package “lme4” (Bates et al. 2015) to test the relationship between quoll and cat camera detection rates vs rabbit camera detection rates. Time since quoll release (time/quoll in number of weeks) was included as a variable. Site (n = 30) and year (n = 5) were fitted as random effects and season as a fixed effect. We used a backwards stepwise regression and ranked models based on Akaike information criterion corrected for small sample size (AICc; Grueber et al. 2011; Symonds and Moussalli 2011). We calculated the differences in AICc (ΔAICc) between the best model (that with the lowest AICc) and every other model, and considered only models with a ΔAICc <2. To test each variable, we added it back into the final model and recalculated the AICc value.

Results

Quoll diet.

—We analyzed 298 quoll scats collected between the austral summer of 2014 and austral autumn of 2018, covering 15 different climatic seasons (Table 1). Rarefaction curves indicated that sufficient samples were collected to adequately describe the diet of quolls in the I-FRNP during each season. The majority of scats contained mammals (mean FO = 0.83) and invertebrates (mean FO = 0.83) with vegetation (mean FO = 0.03), reptiles (mean FO = 0.015), and birds (mean FO = 0.015) recorded much less frequently (Fig. 1). The average percentage volume of major dietary items followed a similar pattern to FO with scats containing on average 50% mammals, 29% invertebrates, 6% vegetation, 5% birds, and 4% reptile (Supplementary Data SD1). We detected 16 different prey categories in the diet of reintroduced quolls including 12 species of mammal (Fig. 2). Overall dietary diversity was moderate (H = 2.80), showing a minor peak in autumn (H = 3.13; Supplementary Data SD2). Rabbits were the most commonly detected mammal species (mean FO = 0.43; Fig. 2). Overall, mammal remains were more likely to be found in scats than any other dietary category and likely reflected both hunting events and scavenging. The presence of kangaroo, sheep, and goat, in the diet was likely to represent scavenging rather than direct predation because all these species have an average adult body weight over 20 kg (>10 times that of a western quoll; Menkhorst and Knight 2011). Feral cat and possum were present in the diet but at a very low frequency and it is not known if this was scavenging or direct hunting. One radio-collared possum and its young were suspected to have been killed by a quoll that was recorded on camera accessing the carcass within a tree hollow, but could not be confirmed. Microbats were recorded in 17 scats from at least 14 different quoll individuals and were likely obtained from tree hollows, rock crevices, or under bark while bats were roosting. The species of microbat could not be determined but genera such as Chalinolobus spp. and Nyctophilus spp. are present in the study region.

Table 1.

Western quoll (Dasyurus geoffroii) and cat (Felis catus) dietary samples analyzed by season and by year.

Season	2014		2015		2016		2017		2018		2019		Total
	Cat	Quoll	Cat	Quoll	Cat	Quoll	Cat	Quoll	Cat	Quoll	Cat	Quoll	Cat	Quoll
Autumn	1	16	1	27	11	1	26	20	0	48	0	0	39	112
Spring	2	12	5	17	7	18	9	0	0	0	0	0	23	47
Summer	8	15	0	20	15	14	6	29	2	0	3	0	34	78
Winter	1	19	3	18	8	24	4	0	0	0	2	0	18	61
Total	12	62	9	82	41	58	45	49	2	48	5	0	114	298

Fig. 1.

—The frequency of occurrence (FO) of major prey categories in 298 western quoll scats and 108 feral cat stomachs that contained dietary items (empty stomachs were excluded). Samples were collected between 2014 and 2019. Frequency of occurrence was calculated by season (n = 15 for quolls and n = 18 for cats), with the horizontal bars representing the mean for quolls and cats, respectively.

Fig. 2.

—The frequency of occurrence (FO) of mammalian prey remains from 298 western quoll scats and 108 feral cat stomachs that were collected between 2014 and 2019. Frequency of occurrence was calculated by season (n = 15 for quolls and n = 18 for cats), with the horizontal bars representing the mean for quolls and cats, respectively. Dietary items are ordered by average adult weight, from lowest to highest.

There was seasonal variation in the detection of major prey categories in the diet of quolls, with birds and invertebrates being detected significantly more often in spring (β = 3.03, P < 0.01 and β = 1.3, P = 0.3, respectively), and invertebrates significantly more in summer (β = 1.62, P = 0.01; Supplementary Data SD3). The activity of rabbits at the study site also had a strong effect on the diet of quolls. In all seasons except spring, the occurrence of rabbits in the diet of quolls increased significantly with increasing rabbit activity (Fig. 3; Supplementary Data SD3). The relationship between increasing rabbit activity at the study site and consumption of other prey items varied according to season and prey group. During the summer months, higher rabbit activity was associated with a decrease in the incidence of vegetation, bats, and reptiles in the diet of quolls. During autumn and winter, higher rabbit activity was associated with a decline in the consumption of bats, and during summer and autumn, invertebrate consumption increased with increasing rabbit activity.

Fig. 3.

—The probability of dietary items occurring in a quoll scat sample during each season as a function of rabbit activity. This figure plots model-based estimates for each species derived from the fitted generalized linear model. Only the dietary items with significant, or near significant (P < 0.1), relationships are plotted. See Supplementary Data SD3 for the full model summary.

Overall, invertebrates were detected significantly less often in scats of male quolls than in those of females (β = −3.28, P < 0.01). As quolls aged, there was a significant decrease in the occurrence of birds and invertebrates in their diet (Supplementary Data SD4). There also were sex-related age effects: older males consumed reptiles (β = 0.97, P = 0.04) and vegetation (β = 0.87, P = 0.02) more often, while older females consumed less of these dietary items (Fig. 4). Mammals were incorporated more in the diet of quolls as they got older, but this was not significant (β = 0.45, P = 0.15). There was no significant relationship between the diet of quolls and the amount of time that had lapsed since they were released.

Fig. 4.

—The probability of the five major prey categories occurring in a quoll scat sample between 2014 and 2018. The figure plots model-based estimates (SE indicated with shading) for female and male quolls at different ages, derived from the fitted generalized linear model. Asterisks indicate significance. See Supplementary Data SD5 for the full model summary.

Cat diet.

—We analyzed 124 cat stomachs collected between the austral summer of 2014 to austral summer of 2019 (Table 1). Ten samples could not be attributed to a season and were removed from any modeling involving a seasonal component (n = 114). Sixteen of the stomachs were empty so FO of prey items was calculated for stomachs where at least one prey item was present so as to enable comparison with quoll scat data (n = 108). We detected 11 different prey categories, including three mammal species and five reptile families (Supplementary Data SD5). Rabbits were the most commonly detected prey category (mean FO = 0.52 of stomachs with prey present), followed by invertebrates (mean FO = 0.30), house mice (mean FO = 0.27), and reptiles (mean FO = 0.27). Rarefaction curves indicated that sufficient samples were collected to adequately describe the overall diet of cats in the I-FRNP during the study period. Overall dietary diversity at 95% of the asymptote was H = 1.91, lower than quoll diversity (H = 2.8). Although we did not detect any quoll remains in cat stomachs during this study, cats were known to kill and consume approximately one third of reintroduced radio-collared quolls during the reintroduction (Moseby et al. 2021).

Dietary relationships between cats and quolls.

—There was a high degree of overlap in the diet of quolls and cats (O_jk = 0.86). The FO of rabbits, reptiles, birds, and small mammals, in cat stomachs and quoll scats, was similar (Figs. 1 and 2) but quoll scats had a considerably higher FO of invertebrates, vegetation, bats, and large mammals (Supplementary Data SD5). Cats and quolls responded differently to the activity of rabbits, with invertebrates occurring significantly more in the diet of quolls (β = 0.16, P = 0.001) and significantly less in the diet of cats (β = −0.51, P < 0.001) with increasing rabbit activity (Fig. 5). Unlike quolls, cats did not significantly increase their frequency of consumption of rabbits as rabbit activity increased but they did reduce their intake of birds when rabbit activity was high (β = −0.44, P = 0.04; Supplementary Data SD6). Changes in rabbit activity had no significant effect on the occurrence of reptiles, house mice, or dunnarts, in the diets of either predator.

Fig. 5.

—The probability of prey species occurring in a quoll scat sample and cat stomach sample between 2014 and 2019 as a function of rabbit activity. The figure plots model-based estimates (SE indicated with shading) for quolls and cats at different rabbit activity indices, derived from the fitted generalized linear model. Asterisks indicate significance. See Supplementary Data SD6 for the full model summary.

Temporal and spatial overlap between cats, quolls, and rabbits.

—There was significant temporal overlap in activity between rabbits and both quolls and cats (Fig. 6). All three species were predominantly nocturnal and crepuscular. A linear mixed effects model comparing detection rates of cat, rabbit, and quoll on cameras found cat detection rates were significantly positively associated with rabbit detection rates (Table 2). indicating a spatial association between the two species. In comparison, quoll detection rates were not associated with rabbit or cat detections and were negatively influenced by time since release (Table 2).

Table 2.

Results of linear mixed effects models comparing detection rates of western quolls (Dasyurus geoffroii) and cats (Felis catus) on camera with rabbit detections and time since quoll release

Species	Intercept	Cat	Rabbit	Time/quolls	Season	df	logLink	AICc
Cat	0.28		0.10	−0.01	+	9	−658.9	1336.0
Quoll	0.31			−0.01	+	8	−952.9	1921.8
Rabbit	0.26	1.01			+	8	−2148.4	4312.9

All species models are >2 ΔAICc from the next top model.

Fig. 6.

—Estimates of activity patterns for quolls, cats, and rabbits, centered around midnight. The solid line represents the kernel density estimate for quolls, while the dashed line represents the same for cats and the dotted line for rabbits. The shaded area represents the estimate of activity overlap (overlap coefficient Δ ₄ with 95% bootstrap confidence interval in parentheses).

Discussion

We expected high dietary overlap between cats and quolls, potentially leading to high competition in this novel predator interaction. As hypothesized, there was significant overlap in reintroduced quoll and resident feral cat diet, but quolls consumed more bats and invertebrates and scavenged more on large mammals than did cats, which were not recorded consuming carrion during the study. The FO of introduced rabbits was similar for both species, with rabbit the most common prey item consumed. The activity of rabbits had a significant influence both on quoll and cat diet, revealing possible differences in dietary preferences. Cats consumed fewer invertebrates as rabbit activity increased and quolls consumed more, suggesting cats prefer to prey on rabbits and quolls on invertebrates. The occurrence of vegetation, bats, and reptiles in the diet of quolls also decreased with increasing rabbit activity, particularly in summer, suggesting a preference for rabbits over these other prey items.

High dietary overlap was expected to lead to competition and intraguild predation. Although quoll remains were not detected in cat stomachs, more than one third of released quolls were killed by cats after release and the majority were consumed (Moseby et al. 2021). The incidence of intraguild predation, high dietary overlap, and the similar, high proportion of rabbit in the diet of both cats and quolls suggests competition was present in this novel predator interaction. Contrary to our prediction, dietary overlap and competition between quolls and cats did not lead to spatial or temporal avoidance of cats by quolls. However, the different interactions and relationships between each predator and their main prey species, rabbits, likely reduced competition between the novel predator species. Quolls consumed rabbits in relation to their availability but consumption of rabbits by cats was unrelated to rabbit activity. Furthermore, there was high temporal overlap in activity patterns between rabbits and both quolls and cats. But while cats were significantly associated spatially with rabbits, quolls did not show any significant association. The high spatial association of cats and rabbits and the high proportion of rabbit in cat diet irrespective of rabbit activity suggest that cats are more reliant on rabbits as a food source at our site and focus their activity in areas with higher rabbit activity. This result is supported by a previous study in the region that found rabbits remained an important prey item for cats even after rabbit calicivirus disease decimated the rabbit population (Holden and Mutze 2002). Cats’ preference for hunting live prey and their aversion to scavenging means they are likely to have stronger spatial ties to their prey. In comparison, quolls exhibited a more flexible diet that included both hunting and scavenging and thus did not need to focus their activity in high rabbit areas.

The similar frequency of rabbit occurrence in quoll and cat diet despite the absence of a strong spatial association of quolls with rabbits suggests quolls are more efficient predators of rabbits than are cats. Quolls are smaller than cats and can catch rabbits within rabbit warrens. Quolls were heard killing live rabbits in their warrens, and in at least two cases quoll radio-collars were broken during successful predation events on rabbits. Freshly killed and eaten adult rabbit carcasses often were observed outside quoll dens, suggesting quolls successfully can hunt animals up to at least 1.6 kg, which is equal to their body weight. These observations are supported by Glen et al. (2010) who found that similar-sized and larger mammals such as common brushtail possums (1–2 kg) and southern brown bandicoots (Isoodon obesulus, 0.4–1.6 kg) featured prominently in quoll diets in the Jarrah forests of Western Australia, although the authors did not discount that this observation may have been based on quolls’ consumption of juveniles. However, personal observations (MAJ, 2018) of western quolls attacking and killing adult burrowing bettongs (Bettongia lesueur, 1.6 kg) at a reintroduced population in northern South Australia supports the suggestion that quolls are capable of killing mammals up to 1.6 kg. In comparison, cats are significantly larger than quolls and may need to wait for rabbits to emerge from their warrens. This factor, coupled with the high intraguild predation of cats on quolls (Moseby et al. 2021) and the quolls’ flexible diet, may enable quolls to avoid areas heavily populated by rabbits but still maintain an intake similar to that of cats, thereby reducing competition and predation. Interestingly, a previous study also found no significant relationship between cat and quoll occupancy on cameras (Moseby et al. 2021), suggesting that quolls may reduce their interactions with cats by avoiding areas of high rabbit abundance where cats are likely to be present, rather than avoiding cats per se.

To reduce competition, we hypothesized that quolls might change their diet over time since release so as to reduce dietary overlap and competition with larger cats. We found no evidence of a change in diet with time since release, but the diet of the reintroduced western quolls was flexible, varying according to extrinsic (season and prey availability) and intrinsic (age and sex) factors. Male quolls consumed significantly fewer invertebrates than did females, possibly due to sexual dimorphism in this species. Larger males may need a higher intake of protein to meet higher energy demands, or the smaller females may find it more difficult to catch larger mammal prey. This theory is supported by the fact that younger individuals also consumed more invertebrates than older individuals in our study and studies on the related spotted tail quoll (Dasyurus maculatus) that found males consumed larger prey than females (Jones and Barmuta 1998). Younger western quolls also consumed more birds than older quolls possibly because they are more agile than older animals and can climb trees more easily to access roosting or nesting birds. Both sexes increased the mammal component of their diet as they aged, suggesting either more reliance on scavenging as their hunting abilities waned or increased experience in catching mammalian prey such as mice and rabbits that arguably are harder to catch than invertebrates. Older males consumed significantly more reptiles and vegetation compared with older females, which consumed fewer of these items. This opposing trend may relate to the smaller body size of females enabling them to obtain sufficient protein from invertebrates, whereas larger males require more protein and thus might target larger prey. Intraspecific differences in diet are found in other carnivores that vary in prey selection according to sex (e.g., cheetahs and domestic cats; Cooper et al. 2007; Moseby et al. 2015), age (cougars; Ross et al. 1997), and body size (domestic cats; Kutt 2012; Moseby et al. 2015).

Western quoll diet also showed temporal flexibility. Birds and invertebrates occurred significantly more in spring and invertebrates significantly more in summer. Invertebrate consumption also was found to increase in summer in the Jarrah forests in Western Australia (Glen et al. 2010). These changes likely are due to increased availability of prey species at these times: invertebrates are more common in the warmer months and nestlings and juvenile birds are more common in spring. Interestingly, occurrence of reptile remains in scats was not dependent on season despite this group being more active in the warmer months, suggesting that quolls are able to locate this prey item when they are inactive or within shelters. Rayner et al. (2012) also recorded similar results for reptiles but Glen et al. (2010) found an opposing seasonal trend with increased reptile consumption in summer.

Western quolls in our study had a catholic diet that included invertebrates, mammals, reptiles, birds, and vegetation. Invertebrates were found in 80% of scats, which is similar to the frequency found in the diet of quolls in mesic Jarrah forests in Western Australia (84%; Soderquist and Serena 1994) and semi-arid Western Australia (81%; Rayner et al. 2012). Furthermore, our results support previous studies that found vertebrate remains at similar or higher frequency than invertebrates, suggesting the western quoll is a generalist predator (Soderquist and Serena 1994; Glen et al. 2010; Rayner et al. 2012). Quolls in our study consumed mammal species ranging in weight from 12 to over 20 kg, suggesting they hunted smaller prey but scavenged on carcasses of large mammals such as goats and kangaroos. Scavenging also has been recorded (Soderquist and Serena 1994) or inferred at other sites (Glen et al. 2010; Rayner et al. 2012).

Vegetation was recorded in 30% of quoll scats, suggesting it is an important component of quoll diet and not ingested incidentally. Vegetation also has been found in western quoll diet in Western Australia, including the red pulp surrounding zamia (Macrozamia reidlei) seeds as well as small fruits and parts of flowers (Hancock 1991; Serena et al. 1991). Ingestion of vegetation is not unusual in small carnivores, with the related northern quoll (Dasyurus hallucatus) also found to consume fruit and other vegetation (Dunlop et al. 2017). The inclusion of vegetation, invertebrate, and vertebrate material suggests western quolls are facultative rather than obligate carnivores.

We used different methods to determine diet in cats and quolls, which could have affected results. Stomach contents were used to determine cat diet, whereas scats were collected from quolls. However, empty cat stomachs were excluded from the FO analyses so as to ensure results were comparable. Furthermore, quoll prey items found in quoll scats but absent or in low occurrence in cat stomachs (such as bat, carrion, and invertebrates) would have been discerned easily in stomach contents if present, providing confidence in the methods used. Although we did not record any mammal species larger than 2 kg in cat stomachs, and thus inferred no cases of scavenging in that species, it is possible that some rabbit consumption included scavenged rabbit carcasses. It is not possible to know what proportion of rabbit in cat and quoll diet included scavenging and we therefore acknowledge that the similar FO of rabbit in the two predators’ diets may have comprised different proportions of live and scavenged rabbit.

Our results suggest that dietary studies can reveal important information regarding the mechanisms underlying novel predator interactions, which then can be used to improve conservation outcomes for native predators. Reintroductions of predator species that have been absent for long periods may be more successful in species with broad and flexible diets. Flexibility would enable them to quickly establish in a novel environment because they can find and adapt to new prey and respond to changes in prey availability. In contrast, predators with more specialist diets such as the Canada lynx (Lynx canadensis), Iberian lynx (Lynx pardinus), and Asian populations of the Eurasian lynx (Lynx lynx), which specialize on lagomorphs (Mengüllüoğlu et al. 2018), may be more difficult to establish in new areas. The Iberian lynx has almost reached population extinction due to prey shortages (Ferrer and Negro 2004), and low density of rabbit prey, among other issues, has hampered reintroduction efforts. Understanding the diet of predators thus can assist with identifying species most suitable for reintroduction into novel environments. In addition, despite high dietary overlap between reintroduced native quolls and resident introduced feral cats, quolls were able to coexist with cats due in part to their flexible diet and higher predation efficacy, thereby enabling them to avoid areas with high abundance of co-preferred prey. The dietary flexibility in the western quoll can be used to improve conservation outcomes: reducing rabbit abundance in our study area may naturally control cat populations while providing less impact on native quolls that can hunt rabbits at low density and vary their diet to include other food sources. We suggest that species with flexible diets and high predation efficacy may be more likely to survive and adjust to novel predator interactions and that studying novel interactions can provide important information regarding predator ecology and conservation.

Supplementary Material

Supplementary data are available at Journal of Mammalogy online.

Supplementary Data SD1.—The average volume (%) of major prey categories in 298 western quoll (Dasyurus geoffroii) scats that were collected between 2014 and 2018. The volume of prey remains was estimated to the nearest 5% in each scat during morphological analysis.

Supplementary Data SD2.—Shannon diversity (H) index for western quoll (Dasyurus geoffroii) diet in each season, and cat (Felis catus) diet overall. Higher H values indicate a higher number of species were recorded in the diet samples more evenly. Ĥ is the estimated diversity within 5% of the asymptote and $\widehat{n}$ is the estimated number of samples needed to reach the asymptote.

Supplementary Data SD3.—Model summary for the probability of prey items occurring in a western quoll (Dasyurus geoffroii) scat sample as a function of season and rabbit (Oryctolagus cuniculus) activity. Model estimates, SEs, and P values are shown from our generalized linear model. Significance is indicated in bold. R² = 0.32. Interactions between predictor variables are indicated with an “x.” Rabbit activity index = RAI.

Supplementary Data SD4.—Model summary for the probability of a major prey category occurring in a western quoll (Dasyurus geoffroii) scat sample as a function of the quoll’s age and sex. Model estimates, SEs, and P values are shown from our generalized linear model. Significance is indicated in bold. R² = 0.5. Interactions between predictor variables are indicated with an “x.”

Supplementary Data SD5.—The different prey categories and species detected in the diet from 298 reintroduced western quoll (Dasyurus geoffroii) scats and the 108 feral cat (Felis catus) stomachs that contained dietary items. Samples were collected between 2014 and 2019. Frequency of occurrence is indicated as a percentage. The asterisk denoted species that were likely to have consumed as carrion by quolls.

Supplementary Data SD6.—Model summary for the probability of a prey item occurring in a western quoll (Dasyurus geoffroii) scat sample and a feral cat (Felis catus) stomach sample as a function of rabbit (Oryctolagus cuniculus) activity (RAI). Only prey items consumed by both predators were included. Model estimates, SEs, and P values are shown from our generalized linear model. Significance is indicated in bold. R² = 0.26. Interactions between predictor variables are indicated with an “x.” Rabbit Activity Index is represented by “RAI.”

Acknowledgments

This project would not have been possible without the support of the Adnyamathanha people, traditional custodians of the land. We thank F. Bernhardt, D. Armstrong, and the Conservation and Wildlife Management Branch of Sporting Shooters SA for their assistance in shooting and trapping feral cats. We thank P. Hodgens, D. Peacock, T. Schroeder, H. Bannister, C. Mills, C. Holt, S. Robinson, R. Ladd, T. Moyle, P. Mitchell, M. Henderson, S. Dorries, K. Smith, M. Le Pla, B. Philp DEW staff for assistance in the field collecting samples, observing predation events, and helping with dietary analysis. We also thank C. Young for assistance with statistical analysis and P. Copley and J. Read for helpful comments on the manuscript. This research was carried out with the approval of the South Australian Wildlife Ethics Committee, approval numbers 52/2013 and 8/2017 and University of Adelaide Ethics 17666.

Funding

Funding for this project was provided by the Foundation for Australia’s Most Endangered Species (FAME) and their generous donors, with assistance from the SA Department for Environment and Water.