Introduction

Reintroductions are a key weapon in the fight against global defaunation; aiming to re-establish viable, free-ranging populations of target species within their indigenous range following their extirpation (IUCN 2013). Their success hinges on a population’s passage through the establishment (where post-release effects drive population dynamics), growth (characterised by high rates of expansion), and regulation (where density dependence limits survival and recruitment) phases of translocation (Sarrazin 2007). These mechanisms can have critical demographic (e.g., survival) and genetic consequences (e.g., selection), and so must be monitored closely throughout the reintroduction process (White et al. 2018).

Establishment initially relies on the survival and dispersal of founders (individuals translocated to an area with no conspecifics), and in later reintroduction stages, reinforcers (individuals translocated to an area with resident conspecifics, with the aim of reinforcing genetic, demographic, and behavioural diversity in a population); the rates of which are likely to change over time (Støen et al. 2009; le Gouar et al. 2012; Wilson et al. 2020). For species with discrete home ranges and territoriality, movement can provide valuable information on habitat suitability and indicate the progress of restoration actions (le Gouar et al. 2012; Bennett et al. 2013). As per the exploration–exploitation dilemma (trade-off between learning and using knowledge to improve performance), an animal will adjust its movement (e.g., foraging, predator avoidance) as it becomes more familiar with its environment (i.e., post-release behavioural modification, PRBM, Berger-Tal et al. 2014). Translocated individuals are expected to move along the PRBM continuum from exploratory to more knowledge-based movements within an established and familiar home range. This sequence of initially high daily movements, followed by home range establishment, has been observed in several species including translocated raccoons (Procyon lotor, Mosillo et al. 1999), dormice (family Gliridae, Bright and Morris 1994), grey wolves (Canis lupus, Fritts et al. 1984), and swift foxes (Vulpes velox, Moehrenschlager and Macdonald 2003). Monitoring movement can, therefore, indicate an individuals’ progression along the PRBM continuum, and alert managers when these movements do not reflect a gradual accumulation of knowledge and acclimatisation to the recipient environment (Berger-Tal and Saltz 2014).

After its establishment, a reintroduced population will tend to grow until its size approaches the carrying capacity of the recipient environment, thereby limiting its demographic rates (e.g., births, recruitment) due to density dependence (Sibly and Hone 2002; Armstrong et al. 2005). Density-dependent movement patterns emerge when animals assess habitat suitability by direct interaction with the physical environment (Danchin et al. 2004) and conspecific cues (presence, density, and health, Danchin et al. 2004; Dall et al. 2005; Seppänen et al. 2007). The well-known ecological mechanisms of conspecific attraction (where animals seek and benefit from associations with conspecifics, Stamps 1988) and conspecific exclusion (where animals are displaced by conspecifics, Muriel et al. 2016) are often amplified as a population approaches density-dependence (Armstrong et al. 2005). In translocations, conspecific attraction can manifest in the social integration of translocated individuals, thereby limiting dispersal, as observed in female yellow-bellied marmots (Marmota flaviventris, Snijders et al. 2017), Iberian ibex (Capra pyrenaica, Garnier et al. 2021), and eastern wild turkeys (Meleagris gallopavo silvestris, Sullivan et al. 2022), however, this may also limit population expansion (Richardson and Ewen 2016). Conversely, conspecific exclusion from territories held by residents can lead to hyperdispersal (Bilby and Moseby 2023) and geographic isolation from the intended release site (i.e., priority effects, Fraser et al. 2015). Since direct assessment of habitat suitability can be risky and energetically expensive, translocated individuals can use indirect cues from conspecifics that takes advantage of residents’ experience or familiarity with the site (Stamps et al. 2005; Valone 2007).

Consideration of these behavioural mechanisms is especially crucial for isolated populations such as those on islands or in fenced havens, where immigration is only possible through reinforcements, and emigration results in permanent removal from the population (Ringma et al. 2017; Legge et al. 2018), with the notable exception of volant species for which haven barriers are more permeable (Innes et al. 2022). Conspecifics can anchor reinforcers and promote genetic mixing (Richardson and Ewen 2016) or drive reinforcers away from the intended establishment area (Clarke and Schedvin 1997). Monitoring and support for dispersed individuals can be difficult from an operational viewpoint and, at worst, those individuals may not contribute to the effective population size (le Gouar et al. 2012).

Here we investigated how movement, habitat use, and conspecific association differed between residents and reinforcers, using the model system of an endangered and solitary mesopredator (the eastern quoll, Dasyurus viverrinus) which was reintroduced to a conservation-fenced haven. We used GPS collars to quantify distances travelled per night, home and core ranges, nocturnal (activity) and diurnal (denning) habitat use and preference (using the Australian National Vegetation Information System, and LiDAR), and conspecific associations over three periods: baseline (residents only), release (residents and reinforcers), and settlement (reinforcers only).

We hypothesised that (1) movement (i.e., distance travelled per night and home and core ranges) would be significantly greater for reinforcers as compared to residents, (2) the movement of reinforcers would change between the release and settlement periods as they progressed along the PRBM continuum, (3) habitat use would differ between animal cohorts but trend towards similarity over time (since resource availability and movement is often inversely related, e.g., Mosnier et al. 2015), (4) conspecific associations would be greatest for reinforcers during the release period, when they would make most use of conspecific cueing to determine suitable habitat to establish themselves (e.g., in songbirds, Ahlering et al. 2010, but see Walker et al. 2009 and Jungen 2018 where habitat lacking in conspecifics was preferable for rattlesnakes Crotalus spp.), and (5) movement and associations between the two distinct pelage colours (fawn- and dark-morphs) would differ, in response to anecdotal differences in their abundance and behaviour (B A Wilson pers obs).

Methods

Study area

Our study was conducted at Mulligans Flat Woodland Sanctuary (MFWS), a publicly accessible 485-ha haven containing critically endangered box-gum grassy woodland (McIntyre et al. 2010; Manning et al. 2011; Shorthouse et al. 2012) on the northern border of the Australian Capital Territory (ACT, -35.167, 149.158). MFWS is surrounded by an 11.5 km conservation fence which excludes invasive species such as feral cats (Felis catus), red foxes (Vulpes vulpes), European rabbits (Oryctolagus cuniculus), and European hares (Lepus europaeus). The fence design includes an overhang which prevents entry by these species (based on successful trials at Arid Recovery, Moseby and Read 2006, but adapted for local conditions, Shorthouse et al. 2012), but it does not prevent climbing animals inside the haven from dispersing over the fence and into the surrounding landscape. While invasive vertebrates have been eradicated within the site, environmental conditions are like those of other unfenced woodlands in the region, allowing most terrestrial species (excluding large macropods, which have been excluded from some areas using tall fences to aid restoration) to access all parts of the haven (Manning et al. 2011; Shorthouse et al. 2012).

MFWS and the adjoining Goorooyarroo Sanctuary form the location of the Mulligans Flat–Goorooyarroo Woodland Experiment, where restoration techniques are trialled to promote biodiversity in temperate woodlands (e.g., supplementing coarse woody debris and controlling grazers). As part of the Experiment, the fence has enabled reintroductions of locally extinct native species to restore biodiversity and ecosystem function (Shorthouse et al. 2012).

Study species

The eastern quoll ('murunguny' in the Indigenous Ngunnawal language) is a critical weight range (Australian mammal between 35–5500 g that suffers the greatest attrition due to predation by invasive species, Burbidge and McKenzie 1989) dasyurid (carnivorous marsupial, Stannard and Old 2013). It is the only quoll species which exhibits two distinct pelage colours, which vary independently of sex and other morphometric features: fawn-morphs (sandy-coloured with white spots) and dark-morphs (black with white spots). The species is sexually dimorphic, with females (x̄ 0.7 kg) being two-thirds the size and weight of males (x̄ 1.1 kg). Populations fluctuate seasonally, driven primarily by high juvenile mortality (annual survival for juvenile females 64.17% ± 19.92 and males 64.93% ± 19.87 in Tasmanian populations), with highest densities observed in early summer following juvenile weaning, decreased densities during the juvenile dispersal period in late summer and autumn, and minimum densities in August due to some die-off of males following the breeding season (Godsell 1983; Wilson et al. 2023).

Eastern quolls are nocturnal, becoming active around dusk for eight hours regardless of day length (Jones et al. 1997) to hunt invertebrates, small mammals, birds, and reptiles, or consume carcasses and vegetation (Blackhall 1980; Godsell 1983). During the day they tend to den underground, in logs, or in rocky outcrops, often in areas that are proximal to foraging grounds, with a preference for ecotones between forest and open grassland (Godsell 1983). Den sharing was considered rare (Godsell 1983; Jones et al. 2001) until frequently observed between reintroduced females (Wilson et al. 2020).

Eastern quolls are solitary but tend to form loose neighbourhoods. Individuals may have overlapping home ranges but maintain large interindividual distances (> 200 m, Godsell 1983), suggesting that they avoid their neighbours. Male home ranges tend to be larger (x̄ 44 ha) and vary more in size than those of females (x̄ 35 ha), though female home ranges increase while weaning their young (Godsell 1983). Home ranges are typically only shared between related females and mothers and their litters, and female aggression is normally only directed to other mothers supporting large young (Godsell 1983).

Once irruptive and broadly distributed throughout southeastern mainland Australia (Godsell 1983; Peacock and Abbott 2014), the eastern quoll was extirpated from the mainland in the 1960s due to habitat degradation, predation by invasive species, disease, and human persecution (Peacock and Abbott 2014). The species is listed nationally as endangered (Environment Protection and Biodiversity Conservation Act 1999, Australian Government 1999) and was restricted to the drier eastern half of the island state of Tasmania (Rounsevell et al. 1991; Jones and Rose 1996) until its successful reintroduction to Mt Rothwell, Victoria in 2003, MFWS, ACT in 2016, and Barrington Wildlife Sanctuary, New South Wales in 2018. It is now listed as endangered in the ACT (Nature Conservation Act 2014s 90C, ACT Parliamentary Counsel 2014). After a pilot translocation to MFWS revealed elevated mortality in male reinforcers, only female (preferring maternal) reinforcers were translocated in later trials (Wilson et al. 2020, 2021). This tactic, which we adapted using the Translocation Tactics Classification Scheme (Batson et al. 2015), allowed us to reintroduce male and female pups via the mothers’ pouches, and since litters can have multiple sires (B Brockett unpublished data), this could have contributed to increased genetic diversity in the establishing population.

Study design

Our study took place in 2018 during the third eastern quoll reintroduction to MFWS (Wilson et al. 2020). We designed our study to compare the effect of cohort on the movements, habitat use, and conspecific associations of female residents (n = 8) and Tasmanian-born female reinforcers (n = 8). We intended to collect baseline movement data from the residents for 21 nights prior to the reinforcers arriving and follow this with 2–3 months of data collection across all individuals. Due to unforeseen GPS battery performance issues, no unit collected data for more than 31 nights (x̄ 25 nights), and one unit deployed on a resident did not collect any locations (“Frost”, Table 1). This meant that (a) eastern quolls were collared and monitoring for various amounts of time between June and August 2018, and (b) the greatest period of overlap between cohorts (where there were ≥ 3 individuals per group) was 11 nights (Table 1). In response, we redesigned our study to compare movements between cohorts (residents n = 7, reinforcers n = 8) in three distinct study periods: baseline (residents only, nights 3–21), release (both cohorts, nights 22–32 [to take advantage of the 11 nights where ≥ 3 members of each cohort were tracked concurrently to explore interactions between them]), and settlement (reinforcers only, nights 33–52), and between pelage colour morphs (fawn n = 9, and dark n = 6, Table 1).

Table 1 Timeline for GPS-deployment and VHF-radiotracking of ‘resident’ (n = 8) and ‘reinforcer’ (translocated from Tasmania to reinforce genetic, demographic, and behavioural diversity in the population, n = 8) eastern quolls (Dasyurus viverrinus) at Mulligans Flat Woodland Sanctuary, Australian Capital Territory between June and August 2018. (Color figure online)
Full size table

Residents and reinforcers

To source resident eastern quolls, we first monitored the existing eastern quoll population at MFWS by trapping across 92 sites using wire cage traps (31 × 31 × 70 cm) in May 2018 (as per Wilson et al. 2023). Using the locations of the 17 females caught, in June 2018 we targeted and captured 8 females (preferring those that were mothers, i.e., carrying pouch young, to match the reproductive status of the reinforcers in the study) and fitted them with 38 g (< 5% of each animal’s body weight) GPS collars (LiteTrack 30 RF, Sirtrack Ltd, Hawkes Bay, New Zealand). We then released them at their point of capture by 0200 h, giving them time to adjust to their collar and find their den by first light; this marked the first study night (Table 1).

Three weeks later, we captured eight wild-born reinforcers (preferring mothers) from free-ranging populations across Tasmania across three non-consecutive nights (four reinforcers on study night 22, one on night 23, and three on night 29, Table 1). We transported reinforcers by air and road to MFWS in animal carrier crates, and on arrival conducted veterinary checks (as per Portas et al. 2020) and fitted them with GPS collars. We allowed reinforcers to recover in a wooden box until their release. Two hours after placing animals in their box in the centre of MFWS at last light, we opened the door and allowed the reinforcers to leave of their own accord to minimise stress (as per Wilson et al. 2020).

Monitoring

Since eastern quolls are predominantly nocturnal, we configured GPS units to record a location (henceforth fix) every 15 min from 1700 to 0700 h (≤ 56 fixes within a 14 h period). To improve accuracy and reduce horizontal dilution of precision (HDOP), we configured GPS units to abandon a fix attempt if they could not detect satellites and compute a location within 75 s (time-out). We also aimed to radiotrack every animal to their diurnal den using the VHF beacon for their first 42 days post-release (as per Wilson et al. 2020), however unforeseen battery performance issues precluded this for some individuals (see Table 1). We captured animals regularly to check their body weight, condition, and collar fit (weight gain can cause collar injury), and were able to retrieve all GPS collars and data at the end of the study period. Overall, GPS collars were deployed for 25–42 nights (dependent on each unit’s functioning) which produced 18,578 raw fixes, and we conducted 77 captures over the study period.

Data analyses

Data curation

Errors can be present in GPS data when fixes are missing or when the location of an acquired fix is erroneous, and these must be screened prior to analyses (Frair et al. 2010). For example, an eastern quoll denning underground will limit the ability of a GPS unit to communicate with satellites, leading to error (Graves and Waller 2006; Körtner et al. 2015). To deal with this, we first removed fixes where the GPS unit (1) was not deployed, (2) timed out before recording a location, (3) was not working correctly (i.e., was in ‘recovery’ mode), and (4) was deployed on a quoll which was caught in a trap, which removed a total of 10,330 fixes (55.6% of total fixes).

Next, we classified fixes according to whether they were inside or outside the conservation-fencing. Rather than simply removing outside fixes (which does not account for potentially erroneous fixes inside the fence), we compared generalised linear models (GLMs) using Akaike’s Information Criterion corrected for small sample sizes (AICc, Mazerolle 2017) to test whether a fix could be predicted as being inside or outside the fence by the GPS variables collected simultaneously with each fix. We used duration (time taken to acquire a fix, < 75 s), ambient temperature (°C), number of satellites used to calculate the position, and HDOP (an index of GPS coordinate precision where lower values are considered more precise, D’Eon and Delparte 2005). The GLM including GPS variables performed significantly better than the null model (> 2 △AICc, 343.85 △log-likelihood), and every variable significantly predicted whether a fix was inside the fence (p < 0.0001), so we filtered fixes that were in the first or third quartile of each variable as follows: (1) ≤ 54 s duration, (2) ≥ 17 °C, (3) ≤ 2.6 HDOP, and (4) ≤ 5 satellites, which removed 4,207 fixes. Of the remaining fixes, 21 were outside the fence, which we manually removed to produce a curated dataset of 2,950 fixes for subsequent analyses. We note that since GPS units could not determine locations while eastern quolls were underground, movement measures using these data describe eastern quoll nocturnal ‘activity’ only.

Distance

To calculate the distances each eastern quoll travelled per night, we summed the linear distance between consecutive fixes for each night. In addition, we calculated the distances per fix by standardizing the distance travelled per night by the number of fixes, to account for the varying number of fixes per individual, and the fact that accuracy generally increases with increasing number of fixes (Piedallu and Gégout 2005).

Ranges

We used the kernel utilisation distribution (KUD) model, which incorporates distance and time lag (i.e., autocorrelation) between consecutive fixes, to calculate home range (eliminating outlying, exploratory locations, 95% contour) and core range (area used with greater intensity, 50% contour) for each individual (a) over the entire study, (b) for each study period, and (c) for each night (to detect progress along the PRBM continuum), in hectares. We also calculated the area of conspecific overlap (or static interaction) in home and core ranges between dyads (pairs of individuals) using methods described by Pebesma (2018).

Habitat use

We determined habitat use when eastern quolls were active at night (‘nocturnal activity’ determined using GPS fixes, animals n = 15) and when denning during the day (‘diurnal denning’ determined using daily VHF locations, animals n = 16), and compared these to habitat types and attributes that were available across the site using two data sources.

The first source was the Australian National Vegetation Information System (NVIS version 6.0, (NLWRA 2001), which we used to delineate extant native vegetation types at MFWS. Under the NVIS, MFWS contained eight major vegetation groups, but for simplicity we aggregated these into five broad vegetation types (henceforth habitat type) based on similarity: Eucalypt woodland (representing 52.5% of the site), regrowth (16.1%), Eucalypt forest (15.2%), grassland (14.5%), and aquatic (1.74%, Fig. 1). We tested for habitat preference in nocturnal activity and diurnal denning compared to a random distribution using the chi-square test for given probabilities, using frequencies of locations in each habitat with the proportional area of each habitat available in MFWS.

Fig. 1
figure 1

For Mulligans Flat Woodland Sanctuary (MFWS), Australian Capital Territory: (a) National Vegetation Information System version 6.0 (NLWRA 2001) major vegetation groups and their percent cover over, and (b) aggregated vegetation groups representing broad MFWS habitat types, mapped using the ggplot2 (Wickham 2014) package in R version 4.2 (R Core Team 2022)

Full size image

The second data source for determining habitat use was overstory and understory (cover fraction to 3.2 m resolution), and aspect (orientation in degrees, derived from a digital elevation model) from Terrestrial Ecosystem Research Network (TERN) airborne LiDAR and hyperspectral products (van Dijk et al. 2018). We selected these metrics since they predicted appropriate habitat for another mesopredator reintroduced to the site (bush stone-curlew Burhinus grallarius, Rapley 2020), and had the potential to contribute to both food availability and foraging success for the eastern quoll. We fitted GLMs to test significant differences in overstory, understory, and aspect values between nocturnal locations, diurnal dens, and available values across the site.

Conspecific associations

To quantify conspecific associations for each individual with any other collared eastern quoll each night, we calculated their (1) proximity index (proportion of simultaneous fixes that are proximal to a conspecific, based on a distance threshold of 50 m, Bertrand et al. 1996), (2) movement correlation coefficient (a Pearson product-moment correlation statistic, Shirabe 2006), (3) coefficient of sociality (between two moving objects using a signed significance Wilcoxon-rank test, Kenward et al. 1993), and (4) coefficient of association (dynamic interaction comparing the observed with the total number of fixes where two moving objects are observed together, where > 0.5 indicates affiliation or fidelity and < 0.5 indicates no association, Cole 1949; Bauman 1998). We also quantified the proportion of days each eastern quoll was detected den sharing with another collared eastern quoll while radiotracking them to their diurnal dens as an additional measure of conspecific association. Note that den sharing with and between uncollared eastern quolls could not be detected or accounted for, so this behaviour may have been more common than observed.

Modelling

We modelled distance, home and core ranges, habitat use, and conspecific associations by fitting generalised linear mixed models (GLMMs) with cohort, study period, and morph as fixed effects, while incorporating individual as a random effect (to control for variation driven by the individual). For distance, where replication was at the night level, we also incorporated minimum overnight temperature (°C), precipitation (mm), and moon illumination (%) as random effects since these can influence the energetic costs of activity (Linley et al. 2020). We fitted models using a gaussian (normal) error distribution based on visual inspection of the data for all tests except den sharing, where we fitted the model using a binomial distribution with a log-link function. We selected the most parsimonious model(s) according to AICc (< 2 ΔAICc, Burnham and Anderson 2002). We reported means, standard errors, and p-values (α = 0.05).

We conducted analyses in R version 4.2 (R Core Team 2022) using the following packages: adehabitatLT for linear distances and adehabitatHR for home and core ranges (Calenge 2006), ggplot2 (Wickham 2014) and ggpubr (Kassambara 2020) for plotting, lme4 for GLMs and GLMMs (Bates et al. 2015), MuMIn for model selection (Bartoń 2016), raster for loading LiDAR products (Hijmans and van Etten 2015), sf for handling spatial vector data (Pebesma 2018), and wildlifeDI for calculating correlation coefficients (Long et al. 2014).

Results

We monitored the nocturnal activity (GPS fixes, n = 2,950) for 15 eastern quolls, and diurnal denning (VHF fixes, n = 51 unique dens) for 16 eastern quolls, over a maximum of 31 nights (x̄ 25 nights).

Distance and ranges

We found study period had a significant effect on the distances travelled per night (p = 0.008), with the greatest distances being travelled during the release period (2.15 km ± 0.18, Fig. 2a). Over the whole study, eastern quolls travelled an average of 1.75 km (± 0.08) per night.

Fig. 2
figure 2

Violin–boxplots and linear regressions for GPS-tracked ‘resident’ (n = 7) and ‘reinforcer’ (translocated from Tasmania to reinforce genetic, demographic, and behavioural diversity in the population, n = 8) female eastern quolls (Dasyurus viverrinus) at Mulligans Flat Woodland Sanctuary, Australian Capital Territory: (a) distances travelled per night (km) per study period (baseline with residents only, nights 3–21; release with both cohorts, nights 22–32; and settlement with reinforcers only, nights 33–52), (b) home ranges (95% kernel utilisation distribution, in hectares) per study period and cohort, (c) home ranges per study night, and (d) home range overlap per cohort. Distances and ranges were calculated using the adehabitatLT and adehabitatHR packages (Calenge 2006), respectively, in R version 4.2 (R Core Team 2022)

Full size image

We found reinforcers’ home (249.4 ha ± 53.2, p < 0.0001) and core (47.3 ha ± 11.3, p = 0.0001) ranges were significantly larger than those of resident home (89.9 ha ± 11.5) and core (19.9 ha ± 2.5) ranges (Figs. 2b, 3). Home (251 ha ± 58.4, p = 0.001) and core (48 ha ± 12.1, p = 0.001) ranges were also significantly larger during the release period (Fig. 2b). In addition, when we assessed nightly home and core ranges, we found the most parsimonious model included both cohort (home range p = 0.0002, Fig. 2c, core range p < 0.0001) and study night (home range p = 0.01, core range p = 0.0074), with residents exhibiting relatively constant ranges (home range R2 = 0.02, core range R2 = 0.01), and reinforcers exhibiting significantly negative trends (home and core ranges R2 = 0.07) over the study period.

Fig. 3
figure 3

Map of home ranges (95% kernel utilisation distribution, KUD) and core ranges (50% KUD) for GPS-tracked ‘resident’ (n = 7) and ‘reinforcer’ (translocated from Tasmania to reinforce genetic, demographic, and behavioural diversity in the population, n = 8) female eastern quolls (Dasyurus viverrinus) at Mulligans Flat Woodland Sanctuary, Australian Capital Territory. Home ranges were calculated using the adehabitatHR package (Calenge 2006) in R version 4.2 (R Core Team 2022)

Full size image

In addition, we found reinforcer home (115.1 ha, ± 15.3, p > 0.0001) and core (8.5 ha ± 2, p = 0.005) ranges overlapped with those of other collared eastern quolls (i.e., static interaction) significantly more than residents’ home (46 ha ± 4.2) and core (3.9 ha ± 0.7) ranges (Figs. 2d, 3). Note that we present here minimum percent overlap since we could not account for other uncollared eastern quolls across the site.

Habitat use

Nocturnal activity

We found nocturnal activity locations in each habitat type varied significantly from random distribution (χ2 = 1455.8, df = 3, p < 0.0001), with eastern quolls preferring to spend their nocturnal activity in grassland (61.47%, after accounting for habitat availability), followed by Eucalypt woodland (29.90%), regrowth (7.58%), and Eucalypt forest (1.05%, Fig. 4a). These habitat preferences were not driven by differences in cohort, study period, or morph (p > 0.05).

Fig. 4
figure 4

For ‘resident’ (n = 8) and ‘reinforcer’ (translocated from Tasmania to reinforce genetic, demographic, and behavioural diversity in the population, n = 8) female eastern quolls (Dasyurus viverrinus) at Mulligans Flat Woodland Sanctuary, Australian Capital Territory: adjusted frequencies (per km2) and percentage of (a) nocturnal activity (GPS locations, n = 15), and (b) diurnal denning (VHF locations, n = 16) spent in Eucalypt woodland, Eucalypt forest, regrowth, and grassland habitat types. Habitat types were aggregated from eight National Vegetation Information System (version 6.0) major vegetation groups (NLWRA 2001) in R version 4.2 (R Core Team 2022)

Full size image

For habitat attributes, we found eastern quolls were active at night in areas with a mean overstory cover of 12.7% (± 0.3, significantly lower than that which was available throughout the site [13.7% ± 0.02], p < 0.0001, Fig. 5a), understory cover of 2.2% (± 0.1, significantly greater than that which was available [1.9% (± 0.004], p < 0.0001, Fig. 5b), and aspect of 203° (± 1.3, i.e., south-southwest-facing, not significantly different from that which was available [205.1° ± 0.1], p = 0.13, Fig. 5c). However, we did find that eastern quolls spent time in a significantly greater understory cover during the release period (2.6% ± 0.1, p < 0.0001, Fig. 5d) and more southwest-facing aspects during the settlement period (222° ± 2.4, p < 0.0001, Fig. 5e).

Fig. 5
figure 5

Violin–boxplots for ‘resident’ (n = 8) and ‘reinforcer’ (translocated from Tasmania to reinforce genetic, demographic, and behavioural diversity in the population, n = 8) female eastern quolls (Dasyurus viverrinus) reintroduced to Mulligans Flat Woodland Sanctuary, Australian Capital Territory: proportion of (a) overstory and (b) understory (layer cover percent, 3.2 m resolution), and (c) aspect (orientation in degrees, derived from a digital elevation model) associated with nocturnal activity (GPS locations, n = 15), diurnal denning (VHF locations, n = 16), and availability of these variables, (d) understory associated with nocturnal activity per study period (baseline with residents only, nights 3–21; release with both cohorts, nights 22–32; and settlement with reinforcers only, nights 33–52), (e) aspect associated with nocturnal activity per study period, (f) overstory associated with diurnal denning per study period, and (g) aspect associated with diurnal denning per study period. Data were sourced from Terrestrial Ecosystem Research Network (TERN) airborne LiDAR and hyperspectral products (van Dijk et al. 2018)

Full size image

Diurnal denning

Of the 51 unique den sites, we found that 13 dens were used once, 28 were used on two–nine occasions, nine were used on 10–39 occasions, and one den was used on 96 occasions. For three days in a row, this den was used by five collared eastern quolls at once (three residents and two reinforcers).

We found diurnal dens in each habitat type varied significantly from random distribution (χ2 = 45.63, df = 3, p < 0.0001), with eastern quolls preferring to den in grassland (37.45%, after accounting for habitat availability), followed by Eucalypt forest (29.14%), Eucalypt woodland (19.32%), and regrowth (14.11%, Fig. 4b). These den preferences were also not driven by differences in cohort, period, or morph (p > 0.05).

For habitat attributes, we found eastern quolls were active at night in areas with a mean overstory cover of 20.9% (± 0.9, significantly greater than that which was available throughout the site, p < 0.0001, Fig. 5a), understory cover of 2.6% (± 0.1, significant greater than that which was available, p < 0.008, Fig. 5b), and aspect of 207.3° (± 2.4, i.e., south-southwest-facing, not significantly different to that which was available, p = 0.28). Similarly, when we compared these attributes between nocturnal locations and diurnal dens, we found overstory (p < 0.0001) and understory cover (p < 0.02) were significantly greater for diurnal dens, but were not different for aspect (p = 0.22, Fig. 5a). Finally, we found that during the settlement period eastern quolls denned in locations with significantly lower overstory cover (17.2% ± 1.5, p = 0.046, Fig. 5f), and significantly more southwest-facing aspects (222° ± 4.7, Fig. 5g).

Conspecific associations

For conspecific associations, we found Pearson movement correlation coefficients for each individual each night were significantly lower during the settlement period (− 0.11 ± 0.03, p = 0.013) and for fawn- (− 0.08 ± 0.02) compared to dark-morphs (0.03 ± 0.03, p = 0.001, Fig. 6a). We also found significantly lower coefficients of sociality during the settlement period (− 0.0004 ± 0.005, p = 0.026, Fig. 6b).

Fig. 6
figure 6

Violin–boxplots of conspecific association measures derived from nocturnal activity (GPS locations) for ‘resident’ (n = 8) and ‘reinforcer’ (translocated from Tasmania to reinforce genetic, demographic, and behavioural diversity in the population n = 8) female eastern quolls (Dasyurus viverrinus) at Mulligans Flat Woodland Sanctuary, Australian Capital Territory: (a) Pearson correlation coefficient by study period (baseline with residents only, days 3–21; release with both cohorts, days 22–32; and settlement with reinforcers only, days 33–52) and morph (i.e., pelage colour; fawn n = 9, and dark n = 6), (b) Coefficient of sociality by study period, and (c) probability of den sharing by study period and cohort. Correlation coefficients were quantified using the wildlifeDI package (Long et al. 2014) in R version 4.2 (R Core Team 2022)

Full size image

For den sharing, two reinforcers den shared only once with another collared eastern quoll, six individuals den shared on two–six occasions (five of which were reinforcers), six individuals den shared on 11–20 occasions (four of which were residents), and two individuals den shared on 22 occasions each (often which each other, both residents).

Overall, we found significantly lower probabilities of den sharing during the settlement period (33.4% ± 10.1) compared to the baseline period (54.7% ± 7.7, p < 0.0001), and for reinforcers (28.6% ± 7.7) compared to residents (51.7% ± 7.3, p < 0.0001, Fig. 6c).

Discussion

By building an understanding of the movement, habitat use and preference, and associations between members of an endangered species following their reintroduction, we can develop informed strategies for species recovery. To the best of our knowledge, this study presents the first empirical exploration of eastern quoll spatial behaviour and conspecific associations using GPS-telemetry. While our use of this technology enabled finer temporal resolution than any previous study, we were limited in temporal longitude by rapid battery depletion. While we acknowledge the short-term nature of our results, they nonetheless provide novel information for the species and highlight the value of post-release monitoring during reinforcement translocations to improve decision-making.

Distance and range

We found eastern quolls travelled the greatest distances each night during the release period (2.15 km ± 0.2 compared to x̄ 1.75 km ± 0.1, Fig. 2a), acknowledging that since GPS units could not determine locations while animals were underground (i.e., denning, described below), this result may not account for distances travelled to and from dens. Indeed, eastern quolls were observed via remote camera visiting multiple dens repeatedly each night (B A Wilson pers obs).

Despite residents and reinforcers travelling comparable distances each night (rejecting our hypothesis that reinforcers would travel greater distances), reinforcers exhibited significantly greater ranges (home = 249 ha ± 53, core = 47 ha ± 11) than did residents (home = 90 ha ± 12, core = 20 ha ± 3, agreeing with our hypotheses, Figs. 2a, 3). Home (251 ha ± 58) and core (48 ha ± 12) ranges were also significantly larger during the release period (Figs. 2b, 3), and these ranges decreased significantly over the study period for reinforcers (home and core R2 = 0.07) when compared to the relatively consistent ranges of the residents (home R2 = 0.02, core R2 = 0.01, Fig. 2c). Decreases in ranging behaviour with time post-release (i.e., post-release behavioural modification, PRBM) has been similarly observed in the related Tasmanian devil (Sarcophilus harrisii, Thalmann et al. 2016) and other carnivores including grey wolves (Fritts et al. 1984) and swift foxes (Moehrenschlager and Macdonald 2003). Our short-term results revealed that reinforcer eastern quolls progressed along the PRBM continuum over two weeks, suggesting an encouraging accumulation of knowledge and acclimatisation to the recipient environment (Berger-Tal and Saltz 2014).

The home ranges of the female eastern quolls in our study (residents 90 ha ± 11, reinforcers 249 ha ± 53) were considerably larger than those of females in a previous study (100% minimum convex polygon = 34.7 ha ± 5.9, 95% anterior–posterior confidence ellipse [AP] = 43.3 ha ± 5.6, 95% non-parametric minimum area versus probability [MAP] = 34.9 ha ± 7.2, Godsell 1983). This is curious, considering our study took place in a temporally and spatially limited (by conservation-fencing) area, while Godsell (1983) studied eastern quoll movements in unfenced regions across Tasmania (Huon Peninsula, albeit over a similarly short period, 8.58 days ± 4.11). While comparing home ranges determined using different tracking technologies, rates of fixes, and analytical methods should be done with extreme caution, the contrast between our home ranges and those of Godsell (1983) could also be attributed to the distribution of food resources and productivity of available habitat (see Oakwood 2002). Tasmania’s sclerophyll forests were wetter between 1971 and 2000 (Bureau of Meteorology 2000) and likely to have been more productive when compared to the dry, temperate woodlands of the Australian Capital Territory. In addition, Australia suffered rainfall deficiencies between 2017 and 2020 (Bureau of Meteorology 2022), and these severe climatic conditions likely affected available prey for eastern quolls during our study period (similarly observed in Tasmania by Fancourt et al. 2018), thereby necessitating eastern quolls in the latter to range further to meet their energetic needs. Indeed, the eastern quoll population at MFWS underwent a decline in abundance prior to the current study (from N = 32 in Austral summer 2018 to N = 26 in autumn), though this may have been driven by juvenile dispersal between these seasons (Wilson et al. 2023).

Considering the larger home ranges of reinforcers, we were not surprised to find that their home ranges overlapped significantly more with other collared eastern quolls (home = 115 ha ± 15, core = 9 ha ± 2) compared to residents (home = 46 ha ± 4, core = 4 ha ± 1, Fig. 2c). While we acknowledge that we could not account for other uncollared eastern quolls across the site, this static measure of interaction suggests territoriality among the female residents. Intrasexual territoriality has been observed in other dasyurids including the northern quoll (Dasyurus hallucatus, Oakwood 2002), chuditch (Dasyurus geoffroii, Serena and Soderquist 1989), and spotted-tailed quoll (Dasyurus maculatus, Belcher and Darrant 2006). Most of the female eastern quolls in our study were either pregnant or carrying pouch young (i.e., maternal, B A Wilson pers obs), so they may have partaken in offspring-defence to prevent infanticide from conspecifics (Wolff and Peterson 1998). While considered rare in marsupials, infanticide has been observed in the MFWS population on one occasion (B A Wilson pers obs via remote camera).

Despite this evident intrasexual territoriality, none of the reinforcers left the site (as occurred in previous trials, Wilson et al. 2020, 2021). With the short-term nature of our results, we cannot differentiate whether the considerable home range overlap by reinforcers was a product of their being at the beginning of the PRBM continuum, or being outcompeted for space by residents (as observed in common brushtail possums Trichosurus vulpecula, Pietsch 1994). If the latter were true, we would have expected reinforcers to disperse over the conservation-fencing in search of territory. Notably, the mean eastern quoll population size at MFWS was still growing in 2018 (N = 29) and reached peak abundance in 2021 (N = 51, Wilson et al. 2022). Based on this, it is possible that the population was still in its growth phase, and could absorb reinforcers without displacing residents. However, since the two established mainland populations of the eastern quoll at the time of this study (MFWS at 485 ha and Mt Rothwell at 473 ha) are too small to halt a continued loss of genetic diversity (Weeks et al. 2011), these and future populations must be managed as a metapopulation with continued reinforcements (Wilson et al. 2023). When these populations reach their regulation phase (i.e., maximum carrying capacity), it will be important to monitor the survival of these reinforcers to ensure they contribute to each population’s genetic, demographic, and behavioural diversity.

Habitat use

During nocturnal activity, we found eastern quolls preferred grassland habitats (61%, after accounting for habitat availability), followed by Eucalypt woodland (30%), regrowth (8%), and Eucalypt forest (1%, Fig. 4a). They were also active in areas of lower overstory (13% ± 0.3) and understory (2% ± 0.1) than that which was available across the site (Fig. 5a–b). For diurnal denning, we found eastern quolls preferred to den in grassland (37%) and Eucalypt forest (29%), followed by Eucalypt woodland (19%), and regrowth (14%). Coupled with the preference for significantly greater overstory (21% ± 1, Fig. 5a) and understory cover (3% ± 0.1, Fig. 5b) than that which was available throughout the site, we ascertain that both grassland and Eucalypt forest offer appropriate foraging and denning conditions for eastern quolls.

Eastern quolls studied in Tasmania were often associated with forest–pasture ecotones that provided open grasslands for foraging (prey including invertebrates and occasional birds, small mammals, reptiles, fruit, and carrion) during the night, and forest habitat where they can den in hollow logs, rocky outcrops, and underground burrows during the day (Godsell 1983). Similarly, a recent dietary study of the eastern quolls at MFWS revealed they also favoured several species of invertebrates and showed opportunistic scavenging of small–medium mammals (Shippley et al. in review). It is surprising that the eastern quolls at MFWS avoided regrowth for both nocturnal activity and diurnal denning. This could indicate that this mid-succession ecosystem offers suboptimal habitat for the species. However, it may play a role in maintaining functional connectivity (i.e., acting as movement corridors) between patches of appropriate habitat, as observed in European pine martens (Martes martes) within agricultural landscapes (Pereboom et al. 2008). Since the eastern quoll fills a similar ecological niche to martens (i.e., as mesopredators), a mosaic of these habitat types is likely to be important for eastern quolls within fragmented landscapes. It is possible that floristic succession at MFWS may be truncated by its agricultural history (McIntyre et al. 2010), so these regrowth and Eucalypt forest habitat types may not progress (without intervention) to more suitable habitat types for eastern quolls in our lifetime. Our results suggest eastern quolls can thrive in a mosaic of recently disturbed (e.g., grasslands derived from agricultural clearing) and undisturbed (e.g., remnant woodlands) sites, and reinforces the need for reintroduction sites to contain suitable habitat from the outset.

We found an effect of study period on habitat use, where eastern quolls preferred to spend their nocturnal activity in areas of greater understory during the release period (2.6% ± 0.1, Fig. 5d). This may have been driven by increased competition upon the arrival of the reinforcers, causing all animals to seek open grasslands for increased foraging efficiency (Godsell 1983). Curiously, nocturnal activity and diurnal denning was more frequent on southwest-facing (rather than the mean south-facing) aspects during the settlement period (222° ± 2.4, Fig. 5e). For the predominantly north-westerly winds of the ACT, southwest-facing slopes are less sheltered than those to the south (and may be less preferred by eastern quolls). Assuming west-facing aspects are suboptimal, reinforcers may have been forced to select suboptimal foraging grounds and dens through conspecific exclusion. Further, we found denning locations had lower overstory cover during the settlement period (17.2% ± 1.5, Fig. 5f). Since we found a significant preference for greater overstory across all eastern quolls compared to that which was available over the site, this lends weight to the suggestion that reinforcers may have been outcompeted for preferred dens with canopy cover. We posit that the eastern quoll population at MFWS in 2018 may have been approaching density-dependence, given this evidence of conspecific exclusion (Armstrong et al. 2005; Muriel et al. 2016) and priority effects (Fraser et al. 2015).

Conspecific associations

We explored conspecific association in eastern quolls using correlation coefficients and den sharing. We found that coefficients of movement correlation and sociality were significantly lower during the settlement period (early July), but this was not driven by cohort (Fig. 6a). We acknowledge the nested nature of the reinforcer cohort within the settlement period, and that we could not account for uncollared eastern quolls within the site, so we therefore report on minimum movement correlation only. In addition, our results could have been driven by mothers transitioning from usual associations to territorial defence associated with parturition and offspring growth in mid-winter (Wolff and Peterson 1998).

Den sharing with other collared eastern quolls occurred frequently across the study, but was less frequent during the settlement period and for reinforcers (Fig. 6c). This behaviour was consistent with den sharing observed during earlier reintroduction trials at MFWS, where the propensity for den sharing had a positive effect on site fidelity (Wilson et al. 2020). However, the prevalence of den sharing suggests den sites may be limited across the site, so it is crucial that potential reintroduction sites contain enough of this habitat feature to support eastern quolls. It is worth noting that the majority of dens observed at MFWS were abandoned warrens that had been excavated by European rabbits prior to their eradication (Macpherson 2023). While our results have improved our knowledge of the extrinsic habitat types and attributes eastern quoll prefer, research on the intrinsic characteristics of dens (i.e., structure, soil type) could be used to design artificial dens in sites that are lacking, but otherwise provide appropriate habitat for reintroduction. On a broader scale, incorporating eastern quoll occurrence data from established populations, like MFWS, and habitat attributes into species distribution models (e.g., maximum entropy modelling) could be used to identify appropriate sites for reintroduction across the species’ former range.

Surprisingly, fawn-morph eastern quolls had significantly lower levels of movement correlation than dark-morphs (Fig. 6b). To our knowledge, this is the first evidence of morph affecting eastern quoll movement, and we chose to investigate this effect following anecdotal evidence that abundance and behaviour differed between these morphs (B A Wilson pers obs). For example, fawn-morphs tended to be more hesitant and agitated during capture and handling, while dark-morphs tended to be calmer and more curious, and were more commonly captured during the current study. The lower rates of capture for fawn-morphs may be linked with lower breeding success and survivorship through the 2019–20 drought (B Brockett, unpublished data). The significantly lower movement correlation of fawn-morphs suggests they avoid foraging in similar patterns to conspecifics compared to their dark-morph counterparts, which could be beneficial in avoiding competition or aggression, but could be disadvantageous if they fail to perceive habitat suitability using conspecific cueing. While we were unable to discern whether this effect was restricted to fawn-morph reinforcers due to small sample size (n = 4), our result lends to the consideration of morph in future species recovery efforts.

Limitations

While we acknowledge that our study only involved maternal eastern quolls, this demographic group was likely to have had higher energetic costs associated with parturition, thereby facing an increased trade-off between resource acquisition and offspring safety. This has been similarly observed in stone martens (Martes foina) where females selected more food-rich areas and less disturbed sites than did males (Santos and Santos-Reis 2010), and in jaguars (Panthera onca) where females’ preference toward intact forest and against roads led to their habitat being more fragmented than those of males (Conde et al. 2010). Based on this, we believe site selection that meets the restricted needs of female eastern quolls is likely to meet the needs of males as well. Further, in the absence of information on relatedness between eastern quolls, we could not determine whether this factor influenced their conspecific associations (e.g., home range overlap, den sharing) and will be an important addition to future work. Finally, although our parsimonious experimental approach (maximising learning with the fewest number of individuals) limited our sample size (e.g., Wilson et al. 2020, 2022), our findings nevertheless offer valuable insights into the factors influencing movement and associations between reintroduced mesopredators in reserves with high levels of threat (i.e., predation) control (as is appropriate for threatened species). However, it is important to view these reserves as “stepping-stones back to the wild, rather than reservoirs of threatened biota” (Batson 2015). We recognise that to prevent the ‘locking-in’ of the current shifting baseline (where we accept native species as being permanently absent from the wild, Manning et al. 2006), future research must explore innovative solutions to drive or enable adaptive evolution of threatened species and invasive predators alike (i.e., ‘coexistence conservation’, Evans et al. 2022).

Conclusions

Here we have demonstrated that short-term movements in a reintroduced mesopredator can be dynamic in time and shaped both directly by habitat and indirectly through conspecific interactions. While conspecific attraction can encourage reinforcers to settle early in a translocation program (as we hypothesised, Stamps et al. 2005; Valone 2007), this effect can also be amplified simply by the abundance and ubiquity of conspecifics present in the landscape (Armstrong et al. 2005) in later phases (i.e., when the population is in its growth phase, Sarrazin 2007). At some point, however, density dependence may trigger conspecific exclusion, especially for territorial species (e.g., by priority effects, Fraser et al. 2015). For the eastern quoll, we found high levels of den sharing across the site and the use of potentially suboptimal habitat by reinforcers, suggesting density-dependent mechanisms were active during our study. This juxtaposition reinforces the need to consider a species’ life history and monitor reinforcers’ progress throughout a translocation program to ensure practitioners can manage these mechanisms adaptively.

Our findings offer important insights into potentially appropriate habitat structure for future reintroduction sites, in contexts where invasive predator impacts are maintained below the population-level upper threshold of tolerance of the species (Evans et al. 2022). We also reinforce the need for intensive post-release monitoring to inform adaptive management interventions during the establishment period. We recommend that tactics pertaining to site selection should be made carefully with specific regard to the current population’s reintroduction phase, and adaptively within a structured framework to ensure decisions are made with best available knowledge to increase the likelihood of positive conservation outcomes.