RESEARCH doi: 10.1111/emr.12402 REPORT Conserving the endangered woylie (Bettongia penicillata ogilbyi): Establishing a semi-arid population within a fenced safe haven
By Michael Smith, Georgia Volck, Nicola Palmer, Chantelle Jackson, Carly Moir, Raquel Parker, Bryony Palmer and Adele Thomasz
Michael Smith, Georgia Volck, Nicola Palmer, Summary Measuring and monitoring population size and growth are critical to assess- Chantelle Jackson, Carly Moir, and Raquel ing the progress and ultimately the success (or failure) of a reintroduction. The Woylie (Bet- Parker a Regional Ecologist, Senior Field tongia penicillata ogilbyi) is one of Australia’s threatened critical weight range mammals. To Ecologist, Senior Field Ecologist, Senior Field increase the species’ area of occupancy, extent of occurrence, number of sub-populations and global population size, in addition to creating a source population for future reintroduc- Ecologist, Field Ecologist, and Field Ecologist, tions, a new population has been re-established into a safe haven located within the spe- respectively, with the Australian Wildlife cies’ former range – Mt Gibson Wildlife Sanctuary. In this paper, we document the first Conservancy (PO Box 8070 Subiaco East, Perth, 3 years of the reintroduction programme, over which time 162 individuals were translocated WA 6008, Australia; Email: Michael.Smith to Mt Gibson. Specifically, we (i) provide information on survivorship, (ii) estimate changes @australianwildlife.org). Bryony Palmer and in critical population metrics (density, population size and distribution) and (iii) look for any major habitat preferences. Survivorship of collared animals was complete (i.e. zero mortal- Adele Thomasz are both former Field Ecologists ity). The most recent population estimate was in the order of 750 individuals, reflecting with the Australian Wildlife Conservancy. strong growth in population size and density. The woylie has occupied the majority of the Bryony is now a PhD candidate at the safe haven and is well represented in all major vegetation communities. The translocation University of Western Australia (Ecosystem is on track to meeting the key success criteria: a self-sustaining population of woylies with Restoration & Intervention Ecology Research a minimum of 300 individuals. Group, School of Biological Sciences, The Key words: population size, reintroduction, survival, Woylie. University of Western Australia, 35 Stirling Highway, Crawley, Western Australia, Australia, 6009). Adele Thomasz, Formerly Field Ecologist with the Australian Wildlife Conservancy (PO Box 8070 Subiaco East, WA 6008, Austarlia; Email: [email protected]) This project arose from a need to assess the success of a reintroduction programme for the Woylie.
fecundity; Royle et al. 2013) are considered fundamental metrics in threatened species Implications for management (Williams 2002). Addi- Introduction et al. managers tionally, by monitoring population abun- o assess the success of a translocation dance, managers should be able to programme, it is particularly important correlate change with monitored pro- This work provides information on T to monitor initial survival, population cesses. Managers then have a pathway to the establishment of a new growth and the dispersal of individuals, enact management, whether it be to population of Woylies in a semi- with a view to ultimately understanding reduce population size of key predators arid setting, with an additional the capacity of a management area to sup- or competitors, increase emigration rates, emphasis on describing a port a particular population size (i.e. carry- fence off areas to allow vegetation recov- monitoring approach (and ing capacity) and its ‘normal’ oscillations ery, genetically supplement the population subsequent analysis) that can be (i.e. Robert et al. 2015). For these and and so on. It is, therefore, also very impor- applied to species that are other reasons, population density and tant to identify key risk factors for monitor- commonly monitored via cage abundance (and underlying processes such ing (Burgman 2005; Metcalf & Wallace trapping. as immigration, emigration, survival and 2013; Smith et al. 2015) and to, over time,
ª 2020 Ecological Society of Australia and John Wiley & Sons Australia, Ltd ECOLOGICAL MANAGEMENT & RESTORATION 1
Ecological Society of Australia RESEARCH REPORT develop targets and evidence to inform a populations and three translocated popu- Release site theory of management, such as a carrying lations (Yeatman & Groom 2012; Woi- Mt Gibson Wildlife Sanctuary is managed capacity and limits of acceptable change narski et al. 2014). by the not-for-profit organisation, the Aus- (Rogers et al. 2013; Gell et al. 2016). In this study, we document the estab- tralian Wildlife Conservancy (AWC). The The Brush-tailed Bettong (Bettongia lishment and dispersal of a new popula- property is approximately 132,500 ha penicillata) is one of Australia’s threatened tion of woylies released into a 7,838 ha and is located about 350 km north-east critical weight range (CWR) mammals safe haven (Fig. 1) and report on initial of Perth (Fig. 1). Mt Gibson is in a transi- (Burbidge & McKenzie 1989; but see Cardi- survival, changes in population density tion zone between the wetter south-west llo & Bromham 2001). Once distributed and abundance, and habitat utilisation and the more arid Eremean region. It fea- across much of the continent, populations over the first 3 years post-reintroduction. tures a mix of Acacia shrublands, various of the Brush-tailed Bettong collapsed fol- In addition to providing a new source pop- Eucalyptus woodlands and Callitris-domi- lowing the introduction of the Red Fox ulation for future translocations, the pri- nated vegetation communities. A 1.8 m (Vulpes Vulpes) and Cat (Felis catus; Yeat- mary aim of this reintroduction was, as high fence surrounds 7,832 ha and pro- man & Groom 2012). By the 1970s, only a per IUCN guidelines (IUCN 2012), to vides a barrier to cats and foxes and other few populations of the Western sub- increase the overall area of occupancy, similarly sized or larger terrestrial verte- species, the Woylie (Bettongia penicillata extent of occurrence, number of sub-pop- brates. Control efforts ensured that the ogilbyi) persisted in south-western Wes- ulations and global population size of woy- safe haven was free of cats, foxes and tern Australia. With a concerted manage- lies (Ruykys & Kanowski 2015). By goats (Capra aegagrus hircus), after ment effort, primarily broad-scale control developing a robust monitoring pro- which a reintroduction programme for of the fox and translocations, the species gramme, ongoing management can pro- ten regionally extinct mammal species was downgraded to ‘low risk’ (conserva- ceed in an adaptive manner with a view was instigated (Kanowski et al., 2018). tion dependent) in 1996 (Start et al. to assessing the success of the transloca- Before the second survey (October– 1998). However, between 1999 and 2006 tion and developing an understanding of November 2017), 71 Banded Hare-walla- the woylie declined again by c. 90% carrying capacity and its management. bies (Lagostrophus fasciatus), a macro- (Wayne et al., 2015) and in 2008 was re- pod herbivore, had been translocated to listed as ‘fauna that was rare, or likely to Mt Gibson. Other than Banded Hare-walla- become extinct’ under the Western Aus- Materials and Methods bies, translocations for Numbats (Myrme- tralian Wildlife Conservation Act 1950, Translocations cobius fasciatus), Greater Bilbies and as critically endangered in the IUCN (Macrotis lagotis), Red-tailed Phascogales Red List (Woinarski & Burbidge 2016). In Woylies were reintroduced to Mt Gibson (Phascogale calura), Greater Stick-nest 2009, it was listed as endangered under from August 2015, with translocations Rats (Leporillus conditor), Shark Bay the Environment Protection and Biodi- being completed by July 2019. Animals bandicoots (Perameles bougainville) and versity Conservation Act 1999 (Groom were translocated from the Karakamia Shark Bay Mice (Pseudomys fieldi) had 2010). Clearly, developing a robust moni- Wildlife Sanctuary, Perup Sanctuary, and begun. However, most of these species toring programme for every woylie popula- Whiteman Park (Yeatman & Groom are not herbivores and all were considered tion is particularly important if managers 2012); these source locations were to be in low enough numbers to not have are to identify and respond to unwanted selected to maximise the genetic diversity had sufficient time to have made signifi- changes in population size in a timely man- of the reintroduced population (Pacioni cant impact upon food availability or veg- ner. et al. 2013). A total of 162 woylies (90 etation structure. Predation by introduced foxes and cats males:72 females) from the three sources is considered an important factor in the were translocated to Mt Gibson and Survival declines of CWR mammals, including the released at a series of pre-determined woylie (Woinarski et al. 2014). One release sites (Fig. 1). Fifty individuals were To assess their survival, 40 individuals (14 approach to manage the impacts of foxes released into Mt Gibson (from Karakamia males:26 females) translocated between and cats is the creation of fenced or island Wildlife Sanctuary) in 2015. A further nine September 2015 and September 2016 ‘safe havens’ from which introduced individuals were translocated from Kara- had radio-tracking collars (Sirtrack V5C predators (and other unwanted species) kamia and 32 from Perup Sanctuary in 163E 130–260 mm or ZV6C 163D 130– are removed (Legge et al. 2018) to which 2016. Fifteen individuals were translo- 260 mm) attached at release. Collars were predator-vulnerable species can be rein- cated from Whiteman Park to Mt Gibson all below 5% of the animal’s body weight. troduced. Reintroductions of the woylie in 2017 and a further 56 from Perup Sanc- Animals were tracked daily for the first to such areas have made a major contribu- tuary to Mt Gibson in 2018. Thus, only 15 2 weeks post-release, and twice weekly tion to securing the species, with ten pop- individuals were translocated to Mt Gib- for the following 2–4 weeks and then ulations established within island or son after the 2017 survey, but before the weekly until either the collars fell off mainland safe havens. Outside of safe 2018 survey from which data are reported or reached approximately 25% of their havens, there are only two ‘natural’ herein. expected battery life, at which point
2 ECOLOGICAL MANAGEMENT & RESTORATION ª 2020 Ecological Society of Australia and John Wiley & Sons Australia, Ltd RESEARCH REPORT
processed. Animals were scanned and microchipped when necessary and had their sex, age, reproductive status, weight, pes length and general condition mea- sured. Animals were released, by approxi- mately 08:00 AWST. When required, a microchip was inserted (following Depart- ment of Biodiversity, Conservation, & Attractions 2017). A combination of Avid, Biomark and Trovan microchips had been used at the source sites but, on recapture at Mt Gibson, only Trovan 9-mm micro- chips were used. The technique of having multiple traps per site and checking traps twice per night in 2017, or having four traps per site in 2018 (but single-trap clearing), ensured that open traps generally remained avail- able to woylies for the entire trap night and thus reduced the effects of the spe- cies’ trap happiness. Given the different Location of the Mt Gibson Wildlife Sanctuary (solid black line) and the fenced safe Figure 1. sampling approach employed, each haven (dashed line). year’s trapping results were analysed sep- arately. The SECR approach is designed to esti- animals were caught and their collars two (2017) and four (2018) wire-mesh mate density and alleviate the issues with removed. Tracking involved confirming cage traps (20 cm 9 20 cm 9 56 cm) the definition of effective trapping area the survivorship of each individual (as were placed at 40 sites in the northern that are associated with more standard determined by pulse frequency, with a third of the safe haven for three trap mark–recapture approaches (Efford & doubling of the same if no movement nights (Fig. S1). The cages were then Fewster 2013; Royle et al. 2013). had been detected by the collar for 10 h). moved to 40 sites in the centre of the safe Although these are not major issues here haven for the subsequent three trap nights because the study area is defined by the and, finally, to the southern end of the safe safe haven fence, the approach allowed Estimating density and population size haven for the last three trap nights us to readily model spatial variation in den- Woylies are easily trappable, so cage trap- (Fig. S1). Sites were located by first creat- sity, an explicit goal of the monitoring. In ping is an inexpensive, well-established ing a 50 m buffer around the vehicle particular, SECR models the spatial distri- and potentially effective approach to mea- tracks within the safe haven and then ran- bution of latent individual activity centres sure population size. However, the spe- domly assigning 120 points within the buf- and distant-dependent detection (Efford & cies can also present a monitoring fered area. This approach was taken: (i) to Fewster 2013). The approach is essentially challenge with single-trap cages because allow sampling across the entire safe spatial in nature and is based on the pre- individuals are particularly ‘trap happy’; haven (as we expected considerable spa- mise that each individual animal is increas- this can then introduce bias to estimates tial heterogeneity in density) and (ii) as it ingly detectable with its proximity to its of density (Royle et al. 2013). To reduce was critical for staff and animal welfare activity centre (Efford & Fewster 2013; the bias associated with the ‘trap-happy’ that cages could be quickly and effectively Royle et al. 2013). nature of woylies, a spatially explicit cap- checked and cleared. Within each site, the The use of SECR techniques with sin- ture–recapture (SECR) modelling two or four cages were positioned within gle-catch traps has the potential to intro- approach (Efford 2016) was used to esti- 10 m of each other. Each trap was duce additional bias to estimates of mate population size. The SECR approach weighted down, half-covered with a hes- population size; this is because the statistically accounts for biases relating to sian bag and baited with universal bait approach assumes all traps are always ‘trap happiness’ (Royle et al. 2013) and (peanut butter and rolled oats). Traps available to individuals of the target spe- can be used to easily model spatial varia- were set between 16:00 and 18:30 AWST cies over the sampling period (Royle tion in density. each evening (i.e. just prior to sunset). et al. 2013). Although this is clearly not Trapping was conducted over a period Traps were cleared twice per night in the case for single-catch traps, some initial of nine days in July 2017 and July 2018, 2017 and once per night in 2018. All traps research has found that ‘reasonable’ popu- respectively. In each trapping session, were finally cleared, and all animals were lation size estimates can nevertheless be
ª 2020 Ecological Society of Australia and John Wiley & Sons Australia, Ltd ECOLOGICAL MANAGEMENT & RESTORATION 3 RESEARCH REPORT made for some species (Distiller & Borch- centres, trap deployment details, animal Trapping ers 2015). Including multiple traps at each capture details and the area that is associ- site and/or checking traps twice per night ated with each potential home range cen- In 2017, 121 woylies (75 males:46 (i.e. effectively mimicking multiple catch tre pixel. A trap response, spatial or non- females) were captured and the maximum cages), allowed an improvement in capac- spatial model, and either a half-normal or distance an individual moved between ity of the SECR approach to estimate woy- negative exponential detection function capture sites was 7.3 km. All models per- lie population size. (Gopalaswamy et al. 2012) can be chosen. formed well and converged (Table S1). Package SPACECAP (Gopalaswamy Finally, as the analysis is a Bayesian data For all models, there was significant indi- et al. 2012), initiated through R software augmented model (Royle et al. 2013), cation of a positive trap response by woy- (R Core Team 2013), was used to analyse the user must enter the number of MCMC lies (95% credibility intervals did not the data. SPACECAP is a simple but robust iterations, the burn-in period, the thinning include zero), as reflected in the increased R package developed for SECR analysis by rate and the augmentation level (Gopalas- detectability of an individual after its first Gopalaswamy et al. (2012) that performs wamy et al. 2012). Models with different capture (Table S1). The density and popu- the Bayesian data augmented models that detection functions and home range spac- lation size estimates from each model were originally developed by Royle et al. ing (25, 50 and 100 m) were run to assess were similar, regardless of pixel density; (2009). Gopalaswamy et al. (2012) pro- their impact on the estimated population however, the negative exponential detec- vide a detailed description of SPACECAP, size. Note that 25 m is equivalent to tion function consistently produced its use and its statistical underpinnings. 0.000625 km2, 50 to 0.0025 km2 and slightly smaller estimates (Fig. 2). The package uses a hierarchical model 100 to 0.01 km2. In 2018, 253 woylies (165 males:88 that incorporates an observation and a N(S) (the population size) is an esti- females) were captured and the maximum state process (Royle et al. 2009). The state mate of the number of activity centres distance an individual moved between process models density and individual located within S. Density is calculated as capture sites was 14.6 km, with mean activity centres, where the location of an D = N(S)/||S||. ||S|| is the state-space area. movement of 2.25 km (SD = 3.32 km). individual (i) is defined by si ~ Uniform k0 is the expected encounter rate of an All models converged (Table S1) well (S). si (1, 2,..,N) represents individual ani- individual [i,j,k] with an activity centre but the Bayesian P-values for the models mal activity centres that are distributed exactly at the site. r (or 1/b2) can be inter- with a half-normal detection function indi- randomly over region S. As described by preted as a mean ‘range’ parameter that is cated a poor fit. The population and den- Gopalaswamy et al. (2012), SPACECAP scaled according to Borchers and Efford sity estimates were consistent across incorporates binary observations (yijk) col- (2008). pixel densities for each detection func- lected from multiple individuals (i) at mul- SPACECAP automatically generates out- tion. As with 2017, the negative exponen- tiple sites (j) across multiple repeated puts that allow assessment of the model; tial models predicted a smaller population visits (k). yijk ~ Bernoulli(pijk), where pijk these include a Bayesian P-value (Gopalas- size than did the half-normal models. is the probability of detecting individual i wamy et al. 2012 provide details relating Given the models with the negative expo- at site j on visit k. Also, SPACECAP uses to the generation of the Bayesian P-value) nential detection function had better fit in the complementary log–log link transfor- that should be around 0.5 for an adequate 2018, we report on the models with a mation (cloglog(pijk) = b0), where b0 is model fit. To evaluate convergence, negative exponential function and a 25- the inverse of the cloglog transformation: SPACECAP provides the Geweke’s diag- m grid spacing (or 0.000625 km2 pixel