Insuring woylies (Bettongia penicillata ogilbyi) against extinction: establishment of Perup Sanctuary

Perup Sanctuary

Perup Sanctuary is largely characterised by Yerraminnup and Bevan habitat classes (Mattiske and Havel 1998) and receives ~680 mm of annual rainfall on average (Bureau of Meteorology 2022). The haven is 423 ha and is surrounded by a predator-proof fence. The fence is built from 30 mm hexagonal mesh creating a vertical barrier of 190 cm, a 200 cm fixed apron at the base (to resist digging under the fence), a 70 cm outward-facing floppy top, and outrigged electrical wires (to resist animals climbing over) (Wayne et al. 2013b). Foxes, cats, emus (Dromaius novaehollandiae), and the native predator chuditch (Dasyurus geoffroii) were removed or confirmed absent from the Perup Sanctuary at the time of establishment. While most were herded out, a few yongka (western grey kangaroo, Macropus fuliginosus) and one kwara (western brush wallaby, Notamacropus irma) are known to remain within (Yeatman and Wayne 2015). There have been ad hoc opportunistic attempts to control yongka from 2010 to 2023. The sanctuary also contains populations of quenda (Isoodon fusciventer), koomal (Trichosurus vulpecula), ngwayir (Pseudocheirus occidentalis), numbat (Myrmecobius fasciatus), mandarda (Cercartetus concinnus), wambenger (Phascogale tapoatafa), dunnart (Sminthopsis sp.), and mardo (Antechinus flavipes). Originally the objective was to establish an insurance population representing the Kingston and Perup woylie populations in the Upper Warren, which were genetically distinct from the other remnant indigenous population at Dryandra and Tutanning (Pacioni et al. 2011). The objective was later revised in 2013 to have the woylie population in the Perup Sanctuary representative of the genetics across all four remnant indigenous populations (Dryandra, Tutanning, Perup, and Kingston). This was in response to new evidence that demonstrated that contemporary differences between the populations was a modern artefact of genetic isolation resulting from habitat fragmentation, and that historically these fragments were previously part of the same expansive woylie population found across much of southwestern Australia (Pacioni et al. 2015). Based on this, in 2010, 41 founding woylies from surrounding indigenous populations (Upper Warren: Perup and Kingston; Fig. 1) were released into the Perup Sanctuary, with 36 additional individuals later introduced from the Dryandra indigenous population (Fig. 1) in 2013, followed by 23 individuals raised in captivity (orphans from the Upper Warren or Tutanning individuals bred in captivity) introduced between 2013 and 2015 (Table 1). Criteria for the success of the translocation of these founding individuals into the sanctuary determined before the release of animals were as follows:

Fig. 1.

Map of Kingston National Park (Kingston population), Tone-Perup Nature Reserve (Perup population), and Perup Sanctuary within part of the Upper Warren region, Western Australia. State forest and other tenure managed by the Western Australian Government Department of Biodiversity, Conservation and Attractions (DBCA) is also shown.

Since its establishment, the sanctuary has acted as a source for woylie reintroductions and supplementations elsewhere (Table 1). Such harvests have routinely selected individuals that were independent and healthy (preferentially young adults), and only females with pouch young less than 50 mm crown rump length and no evidence of a young at heel (i.e. no large, elongated lactating teat). Harvests also aimed to have parity in sex ratios. Founders or earlier generation individuals and individuals with very high parasite loads (a subjective assessment) were not candidates for translocation. One of the initial Perup founders was inadvertently harvested from the sanctuary in 2013, and one Dryandra individual was found to have escaped in 2014 – these animals have been excluded from the analysis. In some years, harvests of woylies were shortly followed with supplementary introductions, to take advantage of the recently vacated territories (to attempt to improve the prospects of the new individuals).

Population monitoring

Cage trapping to monitor the population has been conducted inside Perup Sanctuary at least annually since 2010 (for dates please see Supplementary Table S1). Over time, the sanctuary has experienced three different configurations for annual routine cage trapping (Table 2; Additional File 1).

Between 2010 and 2013, track-based transects around the internal perimeter and through the middle of the sanctuary were used, but as the population increased, the density of traps was increased in 2013 to avoid trap saturation. From 2014 to 2021, we used three 9 × 9 grids (with reduced spacing in a further attempt to avoid trap saturation) and in 2022 and 2023, a single 10 × 10 grid was trialled. At each of these sessions, traps were baited with peanut butter and oats and opened for four nights (opened in the afternoon and cleared shortly after sunrise the following morning). The sampling design during trapping to harvest individuals for translocation to other populations varied considerably depending on the number of animals required. Although in this study these capture records were only used to supplement our information on demography and the proportion of individuals alive. All captured animals were tagged with uniquely numbered titanium ear tags (Monel 1005-1 ear tags, National Tag and Brand Co., Newport, KY, USA), weighed, sexed, aged (based on the presence of fully formed testes or an active pouch) and released. Each animal was given two ear tags to attempt to ensure animals remained identifiable in the case that one tag was lost.

Extensive surveillance for terrestrial vertebrate predators (including 14 permanently deployed remote sensor camera traps checked fortnightly, opportunistic radio-telemetry of woylies, sign surveys, annual spotlighting, and weekly fence integrity checks) confirmed that there had been no incursions since establishment.

Statistical analysis

Population growth and density

In the first years after the initial release in 2010, the woylie population inside Perup Sanctuary experienced high growth rates and a sharp increase in population density (Fig. 2a). Population density peaked at ~1.2 ha⁻¹ in 2016 and appeared to generally fluctuate around a density of 0.9 ha⁻¹, approximately 380 woylies. There has been a steady downward trend of woylies from 2021 to 2023. The growth rate (λ) in the first 3 years (before supplementation began) was 2.57 (95% confidence interval: 1.86–3.55), 2.12 (1.73–2.61), and 2.09 (1.78–2.46) respectively, before slowing to around 1.0 from 2014 to 2022 (0.67, CI: 0.58–0.76 in 2017; 1.05, CI: 0.95–1.16 in 2018).

Fig. 2.

(a) Density (left axis, solid line) and minimum number of individuals alive (right axis; dashed line). Colours along the dashed lines indicate sampling design: red = 80 trap transect 200 m spacing (2010–2012), orange = 144 trap transect 100 m spacing (2013), green = three 9 × 9 grids 25 m spacing (2014–2021), blue = 10 × 10 grid 100 m spacing (2022–2023). (b) Population growth rate (finite rate of increase, 1 = stable population) of the Perup Sanctuary woylie population. Shading indicates confidence intervals. Vertical dotted lines represent the harvest (−) or supplementation (+) of individuals.

Survivorship

Survival in the founding cohorts varied substantially depending on source population and timing of release. Survivorship was high in individuals sourced locally from the Upper Warren (Kingston and Perup). In the first 2 years, survival was 100% for individuals from both sources, and this gradually declined over time, with the last individuals from these cohorts being recorded alive in 2020, a decade after their initial release (Fig. 3). Individuals from Dryandra, an indigenous remnant population 150 km from Perup, experienced around 50% mortality in the first year after their release in 2013, although after this initial loss, the remaining individuals shared similar survival probabilities with the Perup and Kingston individuals. Survival rates for each of these cohorts dramatically dropped in 2020 and 2021, likely as the remaining founders died from old age. Survival in individuals from each of these sources seemed to fluctuate in line with the survival of the sanctuary born animals (Fig. 3b). None of the individuals raised in captivity (released 2013–2015) were ever recaptured or sighted again, and all are assumed to have died within the first year of their release. The effective sampling area after the release of these individuals was 401 ha (in 2013) and between 180 and 270 ha (between 2014 and 2021). This accounts for between 43% and 95% of the area of the sanctuary, and assuming that the probability of new individuals establishing is equal across the entire sanctuary, we would have expected to detect 43–95% (between 9 and 21 woylies) of these released individuals if they had survived. The proportion of Kingston animals known to be alive following the second release in 2022 was 36% 6 months after release.

Fig. 3.

(a) Proportion of founding individuals known to be alive and (b) estimated survival rate per trapping session of woylies inside Perup Sanctuary from various founding cohorts (Perup = green; Kingston (one cohort released in 2011 and one released in 2022) = orange; Dryandra = blue; captive raised = yellow; sanctuary born = black dashed).

Demographic parameters

The Perup sanctuary woylie population has been male biased except in 2012 (parity; 1:1), and 2021 (female biased) (Fig. 4a). The proportion of adult females breeding was lowest in the first 5 years and has remained high (above 95%) since 2015 (Fig. 4b). The mean weight of both adult males and females has slowly declined since the release of the founders (Fig. 4c, Table 4), dropping from an average of 1450 g to 1180 g in males and 1410 g to 1160 g in females between 2010 and 2022. There was a weak, positive relationship between rainfall in the previous 3 months and population density (Fig. 4d). Rainfall in Djeran and Makuru (from the previous year), and net harvest numbers did not affect population density estimates (Table 5).

Fig. 4.

Demographic parameters of Perup Sanctuary Woylie population. (a) Sex ratio (percentage female); (b) percentage of adult females breeding; (c) weight of adult males (blue dashed line) and females (red solid line); and (d) effect of previous 3 months rainfall on population density. Shading in (c) and (d) indicates 95% confidence intervals.

Assessment of translocation success criteria

All three of the demography-based translocation success criteria were met (Table 6).

Population monitoring sampling design

The optimal spacing for all grid types fell between 70 and 140 m. The larger grid sizes were more sensitive to a change in spacing between traps than the smaller grids, which maintained a smaller RSE value across a larger range of trap spacing (Fig. 5).

Fig. 5.

Expected number of new captures (En, light green line) and recaptures (Er, dark green line) compared to relative standard error (RSE, black dashed line) for four grid formations: four 5 × 5 grids, three 6 × 6 grids, two 7 × 7 grids and one 10 × 10 grid.

Based on RMSE, the larger grid sizes performed better than smaller grid sizes, with the 10 × 10 grid (RMSE: 0.0614) being the best performing overall. The 6 × 6 grids had the lowest RB, after which it increased at a slow rate with grid size but this increase was minimal (Fig. 6). The 5 × 5s performed poorly compared to all other grids, with an RSE value almost 3% higher than the best performing, and 2% higher than the next best grid formation (Fig. 6). Though the expected number of captures predicted by the simulations might be slightly inflated (as we assumed multiple captures per trap were possible), this bias is likely to be minimal, and isn’t expected to alter the interpretation of the outcome (Efford 2023b).

Fig. 6.

The relative standard error (RSE; dashed line), relative bias (RB; solid line) and root mean square error (RMSE; dotted line) for each grid formation of traps.

The best performing simulated grid design (single 10 × 10 grid at 150 m spacing) had an RSE of 6.2, performing slightly better than the sampling designs deployed in 2010–2012 (transect of 80 traps at 200 m; RSE = 7.3) and in 2022–2023 (single 10 × 10 grid at 100 m spacing; RSE = 7.4).

Population growth rates and density

Measures of population growth rates are useful for population viability analyses (PVA; e.g. Pacioni et al. 2017, 2020) and other applications relevant to species conservation and recovery (e.g. identifying population response to threats; Duncan et al. 2020). The population growth rates in the initial years after establishment were expected to approximate the intrinsic rate of increase for the woylie, given that the Perup Sanctuary had no terrestrial mammalian predators, and no woylies present at the time of establishment (i.e. not expected to be resource limited at the time of establishment). The finite annual rate of increase in the Perup Sanctuary in the second and third years (before additional supplementation occurred; 2.56, 1.86–3.55 95% CI; 1.91, 1.61–2.28 95% CI), were comparable to other reported maximum values including in the greater Kingston area when the low woylie densities initially responded to the commencement of regular and broadscale fox baiting (Warrup = 1.49 ± 0.4 s.d.; Winnejup = 2.29 ± 0.44 s.d.; Pacioni et al. 2017), and in Wadderin Sanctuary (~1.2; Short 2023). The annual growth rate inside the predator-free haven at Mount Gibson was higher, approximately 3.1 2–3 years after translocations to the site began, although ongoing supplementation in the second and third year (n = 41, n = 15 respectively) more than doubled the number of founders introduced in the first year (n = 50; Smith et al. 2020). It is important to note that growth at Perup Sanctuary may be somewhat inflated due to the demographic skews in the founding population (i.e. all individuals were of reproductive age and most females carried small pouch young).

Woylie densities observed in the Perup Sanctuary peaked in 2016 at around 1.18 woylies ha⁻¹ (95% CI 0.98–1.40) before averaging around 0.9 woylies ha⁻¹. These are comparable to peak woylie density estimates observed across the Upper Warren region in 1999 (1.57 woylie ha⁻¹; Wayne et al. 2013a), prior to their catastrophic declines. These peak estimates provide an indication of the maximum likely densities for this species when it is not limited by predation in this habitat. Such estimates here are higher than estimates from Perup in the 1970s before extensive fox control, 0.45 woylies ha⁻¹ (Christensen 1980), but similar to those of other predator-free populations including Wadderin Sanctuary (427 ha fenced reserve; around 0.7 woylies ha⁻¹; Short 2023) and Karakamia Sanctuary (275 ha fenced reserve; average 0.9 woylie ha⁻¹ 2015–2021; Palmer et al. 2022). The reduction of woylie densities below 0.7 woylies ha⁻¹ in the last 2 years (2022 and 2023) represent the lowest densities observed since the initial years of the establishment of the colony. This may be due to a reduction in rainfall – in 2023, the Sanctuary received only 60% of the long-term mean rainfall for the three months preceding the survey (Bureau of Meteorology 2022). Regular monitoring should continue to determine whether this trend persists, and to identify the drivers of population change.

Drivers of population change

The woylie population in the Perup Sanctuary underwent an approximate 34% decline about 7 years after the population was established in 2010. The timing of this decline is consistent with the observations at Wadderin Sanctuary (peaked at 7.5 years; Short 2023) and predictions of eruptive dynamics (boom-bust cycles) by Duncan et al. (2020), based on a species’ intrinsic rate of population growth and adult body mass. Duncan et al. (2020) also revealed that this phenomenon has occurred in other woylie populations, and that it tends to occur in species with higher maximum rates of population increase, and in populations that increase rapidly and substantially, which was the case here. Where this occurs, it is consistent with some form of density-dependent feedback, such as overconsumption of resources or the build-up of diseases or predators (not the case here), but it does not negate the possibility that population declines could be driven by external factors such as abrupt environmental change or the arrival of new predators or diseases. The population here seems to largely have avoided the ‘bust’ that other havened bettong populations may have faced (Moseby et al. 2018; Short 2023), perhaps owing to the steady number of animals harvested out of the Perup Sanctuary. Nonetheless, this highlights the importance of adequately monitoring populations as well as the factors that may explain population changes including predators and predation rates, body condition and food resource availability, and animal health and signs of disease.

The positive correlation between rainfall and woylie population density in the Perup Sanctuary may also be helpful for predicting population change and likely indicates that the population is driven by the availability of food resources. The diversity, abundance, and biomass of fungi sporocarps, the staple diet of the woylie, have been related to climate and rainfall during the preceding one or more months (Bateman et al. 2012; Danks et al. 2013; Zosky et al. 2017; Buscardo et al. 2021). The decreased average body mass of woylies over time in the Perup Sanctuary is also consistent with food resources increasingly being a limiting factor for the population. Short (2023) also related observed declines in woylie densities, condition, and reproduction at Wadderin Sanctuary with resource limitation and years of low rainfall. The confirmed ongoing absence of terrestrial predators and the results from health and disease monitoring (Skogvold et al. 2017) further strengthen the case that the woylie population in the Perup Sanctuary has been regulated by food. Ongoing monitoring of population size is valuable to allow managers to intervene if the population exceeds carrying capacity and fails to self-regulate to prevent the loss of individual conditions that can result from over-population (Moseby et al. 2018).

Harvest rates of woylies from Perup Sanctuary appear to have been sustainable as there was no significant effect on the population density, and despite the considerable disparities in the rate of introductions compared to harvests (111:430), population growth trends remained positive or neutral. The maximum net harvest rate of approximately 45% of the adult population in 2014, when 182 woylies were removed from the population, was unintended and occurred because of an inaccurate overestimate of the population at the time. While this demonstrates the remarkable breeding capabilities of the woylie, this was greater than the advised safe harvest rates for the species of no more than 30% of the adult (based on the presence of fully formed testes or active pouch) population within a generation (about 2.5 years for the woylie; Pacioni et al. 2019). This threshold is recommended to avoid a significant reduction in the population size and the risks of reducing the genetic diversity of the population. While the large harvest in 2014 did not reduce the population size it is not known what effects this may have had on the genetic characteristics of the population. However, the systematic avoidance of harvesting founders for translocation, should have reduced the risks of losing unique alleles or reducing the genetic diversity in the woylie population within the Perup Sanctuary. Nonetheless, we support using the harvest recommendations of Pacioni et al. (2019) as a conservative measure to ensure that harvest rates from source populations remain sustainable.

Survivorship

The striking differences in the survivorship between the founder cohorts (100%, 50% and 0% of the initial release, 3-year release, and captive born animals released after 3–5 years respectively) could have important implications for the establishment and supplementation of new populations. It appears that after the initial release of founders in 2010, subsequent released cohorts had lower survival probabilities. This may indicate that as the population grew closer to carrying capacity, there were less available territories, and new individuals were less likely to establish and persist. This was the case even when existing individuals were removed prior to the release of new animals, suggesting that the existing population has an advantage in competing for resources. When looking to maximise genetic diversity and population sizes, these findings would suggest that introducing all individuals at once, and giving them equal opportunity to establish territories, may lead to more individuals establishing, compared to staggered releases through time. It is worth noting that despite considerable survey effort, no captive-reared individuals were ever captured after release, which raises the question, whether raising orphaned joeys and releasing them into sanctuaries is the most humane option given their apparent survival was 0. Captive-bred woylies have fared better elsewhere (e.g. Short 2023) and as the release of each cohort into Perup Sanctuary was not completely identical, the possibility cannot be ruled out that additional individual or environmental factors led to the disparity in survival rates between cohorts here. A study of individual survival probabilities (e.g. with radio collars) would be better suited to address this question. Other factors that may affect the survival of translocated animals include population source (e.g. wild, havened [fenced or island], captive, hand-reared); habitat similarity between the source and recipient site; timing in relation to seasonal variation in the availability of important and potentially limited resources; sex, age, condition, health, medical intervention or therapy (e.g. Northover et al. 2019a, 2019b); and the manner in which the animals are released into the recipient population (e.g. staged or ‘soft’ release vs immediate; Tetzlaff et al. 2019; Resende et al. 2021). An improved understanding of the factors that affect the survivorship of founders and subsequent supplementations will substantially improve the existing PVA models (Pacioni et al. 2020) used to inform how best to manage the genetics of existing and future translocated woylie populations.

Demographics

The factors that may be driving the observed changes in sex ratio over time are not clear; however, they provide some indication of the growth potential for the population. A female biased population can grow at a faster rate than a male biased one, though this is unlikely to be the main driver of population fluctuation in Perup Sanctuary as sex ratio does not appear to correlate with population growth (population growth remained relatively stable between 2014 and 2022 despite substantial fluctuations in sex ratio). Periods of female bias are also beneficial for achieving usual harvest objectives of sex ratio parity or female bias for translocations elsewhere. Historically, sourcing enough females for translocation has been particularly challenging owing to the unsuitability of a large proportion of captured females because of their continuous breeding habits and the associated welfare risks to larger pouch young and dependent young ‘at heel’. With the apparent reduction in male bias within Perup Sanctuary in recent years, this may become less of an issue. In other marsupials, such as brush tailed possums, biasing offspring in favour of males occurs in times of limited resources to reduce resource competition with kin (philopatric females are more likely to remain in their mothers territory and compete with her, whereas males disperse) (Johnson et al. 2001). If the same exists in woylies (a similar system), the male bias sex ratio in Perup Sanctuary may be indicative of periods of resource competition, though to support this hypothesis, sex-dependent offspring survival should also be examined. While the change in sex ratio may instead be an artefact of sampling variation, adult body mass has declined over time which is also likely to indicate resource competition within this population. A previous study has further suggested that this may be a response to both resource competition and an absence of selection pressure for larger size without predators present (Harrison et al. 2023). It remains important to continue to monitor adult body mass to account for, and potentially predict, population changes that can inform adaptive management (Page et al. 2019).

Assessment of translocation success criteria

Isolation inside Perup Sanctuary proved a success in protecting woylies from predators and allowing them to establish robust numbers; however, genetic health and representativeness remain to be evaluated. Modern woylie populations have lost between 46 and 91% of the genetic diversity within their mitochondrial DNA (Pacioni et al. 2015), and populations with low gene flow, such as those isolated by fences, are threatened with further and accelerated losses of genetic diversity and consequent inbreeding depression. For a complete appraisal of the success of this translocation, the fourth criterion (retention of >90% of founder genetic diversity) needs to be evaluated. The founder population contained individuals from multiple sub-populations to maximise diversity and representativeness of the extant diversity. Nevertheless, quantifying the genetic diversity within this population is important. Farquharson et al. (2021) found that genetic heterozygosity of the Perup Sanctuary population in 2020 was high and comparable to indigenous source populations (Dryandra, Perup and Kingston). However, they also found some evidence that inbreeding may be occurring here as well as within Karakamia Sanctuary and the indigenous wild populations in Kingston and Perup. While the potential for inbreeding merits further investigation (e.g. to ensure the modelling indicators are not an artefact of sampling), the authors recommend having a metapopulation management strategy that aims to promote increased genome-wide diversity and reduce inbreeding (e.g. through targeted translocations) and maintain overall species heterozygosity by ensuring large population sizes.

Population monitoring sampling design

Design modelling revealed that the current design (10 × 10 grid) was the best performing array. Our current use of 100 m spacing is slightly below the best performing spacing of 130 m, though it remains a robust design (RSE = 7) and is more practical to be run on foot (there are limited roads inside Perup Sanctuary so this grid must be serviced without vehicles). The design modelling was limited by the fact that the ‘secrdesign’ package currently only operates under the assumption that more than one animal can be detected per detector. When capturing woylies using cage traps, this is very rarely the case, and trap saturation is a commonly experienced problem when this species occurs at high densities. While the design modelling is valuable in informing a robust sampling design, it does not take into account the potential issue of trap saturation.

The optimal spacing for all grid types fell between 70 and 140 m and the design implemented from 2014 to 2021 was spaced well below this (three 9 × 9 grids at 25 m spacing). Empirical literature suggests that an effective sampling array should span at least one home range (Bondrup-Nielsen 1983; Efford and Boulanger 2019), which for woylies inside Perup Sanctuary is ~65 ha (Yeatman and Wayne 2015). This was satisfied by the transect design and the 10 × 10 grid, but each of the 9 × 9 grids only comprised approximately 5 ha. Simulation studies have demonstrated that SECR models from sampling designs encompassing less than one animals home range can still perform well (Sollmann et al. 2012), nonetheless, these estimates should be treated with caution as they may have overestimated density. Overestimations of density by the 9 × 9 grids may partially account for the observed decline in woylie density after 2021, and we recommend that density continue to be monitored closely using the current sampling design (10 × 10 grid, 100–150 m spacing) that spans more than one woylie home range.

Ongoing management of the sanctuary and the woylie population

Ongoing monitoring of the population is fundamentally important to ensure effective and adaptive management. The success to date is no guarantee of success going forward. The many failed historic translocations (Threatened Species Scientific Committee 2018; National Environmental Science Program Threatened Species Research Hub 2019), and the recent rapid >90% declines and local extinctions (Wayne et al. 2013a, 2015, 2017), are clear testimony of the ever-present risks to woylie populations. Done well, monitoring can prompt timely management intervention and identify the putative drivers of population change. Our population here appears to be limited by resource availability, and further understanding of such limitation would be valuable in understanding the ability of the landscape to support the woylie population.

It is also important to assess whether sustained high woylie densities have any adverse effects on the habitat and other species within the Perup Sanctuary. As ecosystem engineers and major soil bioturbators, woylies are expected to provide many benefits to the ecosystem (Garkaklis et al. 2004; Murphy et al. 2005; Fleming et al. 2014). While they are vectors in the dispersal of seeds and spores (Beca et al. 2021; Palmer et al. 2021), including those of hypogeal mycorrhizal fungi, they are also predators of plant propagules, roots and tubers, and some small animals (Zosky et al. 2017). Major changes in vegetation have been shown to result from high densities of another ecosystem engineer, the closely-related boodie (burrowing bettong, Bettongia lesueur) (Linley et al. 2017; Moseby et al. 2018), and vegetation changes have also been observed at multiple sanctuaries containing woylies (Wadderin Sanctuary, Short 2023; Scotia and Yookamurra; Kemp et al. 2021). A formal analysis of the impact of woylies on the plant and fungal communities in Perup Sanctuary would be valuable to assess resource availability.