The vegetation structure and condition of contracting lowland habitat for Leadbeater’s possum ( Gymnobelideus leadbeateri )

. Lowland Leadbeater’s possums are on a trajectory to extinction, with fewer than 40 individuals surviving in the wild. Quantification of the vegetation characteristics of their occupied habitat is urgently needed to inform strategies to conserve this genetically distinct population. We surveyed the canopy and midstorey vegetation at all remaining (nine) occupied territories and eleven abandoned territories in lowland swamp forest at the Yellingbo Nature Conservation Reserve. For each territory we quantified canopy and midstorey stem density, basal area (total and live) and vegetation condition (percentage live basal area, tree crown vigour and plant area index) within a 50-m radius of known den locations. The canopy at all locations was dominated by mountain swamp gum ( Eucalyptus camphora ), with most occupied sites supporting dense midstorey dominated by paperbarks, either Melaleuca squarrosa or Melaleuca ericifolia . Occupied territories had higher stem densities and better vegetation condition than abandoned territories. Stem density alone was able to predict occupancy vis-a`-vis abandoned sites with a high (80%) degree of accuracy. Lowland Leadbeater’s possums occupy swamp forests characterised by high stem density, particularly in the midstorey, structural complexity and healthy vegetation. These findings can help guide habitat restoration and translocation projects currently underway to expand the area of lowland habitat for this critically endangered species.


Introduction
Globally, anthropogenic land use is the major driver of biodiversity loss, with agriculture the leading contributor to deforestation, forest fragmentation and degradation (Chaudhary et al. 2016).Often, a species' ability to survive in a fragmented or degraded landscape is influenced by a combination of the: (i) ecological and demographic attributes of the species, such as population size, movement capacity and habitat requirements; and (ii) the suitability of the remaining environment for the species, such as the quality, size and connectivity of habitat, and the threatening process(es) acting upon it (Henle et al. 2004;Bennett and Saunders 2010;Ashman and Watchorn 2019).
Leadbeater's possum (Gymnobelideus leadbeateri) is one of ten mammal species in Australia listed as 'Critically Endangered' under the Environment Protection and Biodiversity Conservation Act 1999.The species has a restricted distribution confined to a 70 Â 95 km area centred on the Central Highlands of Victoria, Australia.Across its range, the species occupies three distinct forest types spanning a broad elevation gradient (100-1500 m Australian Height Datum) (Harley 2004a;DEPI 2014).These include one outlying lowland population that is disjunct from highland populations.Approximately 96% of the remaining habitat known to support extant populations of Leadbeater's possum is montane ash forest (600-1300 m), typically dominated by mountain ash (Eucalyptus regnans) or alpine ash (Eucalyptus delegatensis).Subalpine woodland (1300-1500 m) dominated by snow gum (Eucalyptus pauciflora) accounts for ,4% of the remaining available habitat within the species' range.The third forest type occupied by the species is lowland swamp forest (100-120 m) dominated by mountain swamp gum (Eucalyptus camphora), which currently contributes ,0.1% of available habitat.The vegetation floristics differ markedly between these three forest types, albeit there are several features in common to each, including a predominance of gum-barked eucalypts and a cold, wet climate (Harley 2004a).
The type specimen of Leadbeater's possum was collected from lowland forest near Westernport Bay ,60 km to the south of the species' current range (McCoy 1867).Prior to extensive clearing for agriculture early last century, lowland swamp forest and shrub thicket (e.g.Melaleuca ericifolia) covered extensive areas within Leadbeater's possum's historic range (i.e.$10 000 ha; (Yugovic and Mitchell 2006).The widespread clearing of these floodplain forests for agriculture early last century (e.g.Spencer 1921) is the principal reason that Leadbeater's possums are now confined to a single lowland site, Yellingbo Nature Conservation Reserve, which protects 180 ha of remnant swamp forest along sections of the Cockatoo and Macclesfield Creek floodplains (McMahon and Franklin 1993;Harley et al. 2005).
Molecular analyses indicate that the extant lowland population of Leadbeater's possum at Yellingbo is genetically similar to historic, now-extinct lowland populations and distinct from extant highland populations (Hansen and Taylor 2008;Hansen et al. 2009).On this basis, the lowland population at Yellingbo represents the last remnant of the once extensive lowland distribution for the species.Research on population dynamics and demographic parameters on the extant lowland population has revealed a substantial population decline (,70%) since 2003, with the total wild population now comprising fewer than 40 individuals (Harley 2016).The risk of extinction for this population is extremely high without effective interventions, and it now constitutes the most threatened mammal population in Victoria, Australia.
Extensive research has investigated habitat attributes at sites occupied by highland populations of Leadbeater's possum.Part of the reason for this is tied to major disturbances from fire and timber harvesting across the species' highland range, both of which severely impact habitat quantity and quality (Lindenmayer and Sato 2018;Lindenmayer et al. 2012Lindenmayer et al. , 2020)).In montane ash forest, Smith et al. (1985) and Smith and Lindenmayer (1988) found the basal area of Acacia species to be one of the strongest factors influencing the presence and abundance of Leadbeater's possum.Similarly, Lindenmayer et al. (1991aLindenmayer et al. ( , 1991b) ) found the presence of Leadbeater's possum to be positively correlated with the basal area of Acacia species, and tree hollow occupancy to be associated with the quantity and proximity of vegetation surrounding denning locations.These findings highlight the importance of vegetation structure to Leadbeater's possum.In contrast to montane ash forest, quantitative data examining Leadbeater's possum's habitat requirements for the other two forest types occupied by the species are lacking.
Leadbeater's possum has three broad ecological requirements: hollow-bearing trees for diurnal denning, suitable plant species to support foraging and dense vegetation structure in the canopy or midstorey to facilitate movement (Harley 2004a).We examined two of these variables, floristics and vegetation structure, that are strongly tied to the quality of habitat for foraging and movement.Hollow-bearing trees for denning were not assessed in this study, as numerous nest boxes have been installed specifically for Leadbeater's possum throughout the study area since the mid-1990s, and therefore den availability was not considered a limiting resource (Harley 2004b).
Leadbeater's possum is an exudivore that supplements its diet with invertebrates (Goldingay and Jackson 2004).Our understanding of the species' diet in the wild is based on a single study of a highland population at a single locality.Based on faecal analysis, Smith (1984) identified the major dietary items as the gums of Acacia spp., manna, honeydew and arthropods.To date there has been no assessment of dietary breadth in relation to variables such as the dominant eucalypt, forest age class or elevation.Foraging observations have been made opportunistically for individuals being radio-tracked in lowland swamp forest at Yellingbo (Harley 2005).Most feeding observations involved possums licking, for short duration (typically ,10 s), the trucks and major branches of E. camphora, and occasionally midstorey stems including scented paperbark (Melaleuca squarrosa), swamp paperbark (M.ericifolia) and woolly tea-tree (Leptospermum lanigerum) (D.Harley, unpubl.data).Possums were not observed feeding in canopy foliage, and foraging animals would frequently change trees in which they were feeding rather than spend prolonged periods in particular trees.Notably, there were no observations of possums feeding in wattles at Yellingbo, and wattle gum is not likely to form part of the species' diet in this forest type (Harley et al. 2005).
We characterised the vegetation at all currently occupied lowland territories as well as those abandoned by Leadbeater's possum since 2003 when population declines commenced, with two related aims: (i) to provide a quantitative description of the vegetation floristics and structure within occupied territories to guide habitat restoration and the identification of translocation sites; and (ii) to examine whether there were differences in vegetation structure and condition at these occupied territories compared to territories that have been abandoned.The results reported here provide the most detailed description of vegetation floristics and structure at sites occupied by the geneticallydistinct lowland population of Leadbeater's possum.

Study area
This study was conducted at Yellingbo Nature Conservation Reserve (37851 0 S, 145829 0 E, ,110 m above sea level), which is located ,50 km east of Melbourne and supports the only extant lowland population of Leadbeater's possum (Fig. 1).The reserve comprises ,680 ha of remnant, riparian and swamp forest stretching along four watercourses, the Woori Yallock, Sheepstation, Macclesfield and Cockatoo Creeks.The area receives an average of ,1100 mm of rainfall annually, and experiences mean daily maximum temperatures ranging from 13.68C in winter to 25.68C in summer.
Previous research has found that within Yellingbo, the Leadbeater's possum population is confined to low-lying floodplains with a canopy dominated by mountain swamp gum with, in places, a dense midstorey of tea-tree (Leptospermum spp.) or paperbark (Melaleuca spp.) (Harley et al. 2005).This forest type, and hence the possum's distribution within the reserve, was previously restricted to a narrow, linear section of floodplain, ,6 km long and rarely exceeding 120 m in width when monitoring commenced in the mid-1990s (Harley et al. 2005).
At the time of this study, this distribution had contracted to ,4 km of this floodplain (D.Harley, unpubl. data).
Large areas (.30%) of E. camphora swamp forest at Yellingbo are dieback-affected, putatively due to impeded drainage and prolonged flooding/waterlogging resulting from erosionderived sedimentation following land clearing and stream channelisation upstream of the reserve (McMahon and Franklin 1993;Greet 2016;Zacks et al. 2018).Drier forests dominated by messmate (Eucalyptus obliqua), green scentbark (Eucalyptus fulgens) and narrow-leaved peppermint (Eucalyptus radiata) occur on the well-drained slopes adjacent to the floodplain where tree cover remains.However, this is limited in extent as the reserve is surrounded by extensively cleared farmland.

Study sites
We surveyed 20 locations within the reserve, comprising all nine territories currently known to be occupied by lowland Leadbeater's possums, and 11 formerly occupied territories that have been abandoned in the years since 2003.The location and status of territories at Yellingbo were determined from annual surveys conducted since 1996 (Harley et al. 2005;Harley and Lill 2007;D. Harley, unpublished data).All the suitable foraging habitat in the reserve has been surveyed using a combination of camera traps, nest boxes, stag-watching and call imitation surveys (i.e.multiple survey techniques applied) and sampling is thought to comprise the total wild population.
The social organisation of Leadbeater's possum is based around a single monogamous breeding pair and multiple generations of their young (Smith 1984;Harley 2005).This group of individuals, typically referred to as a 'colony', den together each day and represent a stable social unit (Harley 2005).Each colony occupies a defined territory that does not overlap with that of neighbouring colonies (Smith 1984;Harley 2005;Harley et al. 2005;D. Harley, unpubl. data).For the purposes of this study, each occupied territory supports a resident colony.Abandoned territories previously supported resident colonies.
Given past difficulties in trapping Leadbeater's possum at highland locations, nest boxes were installed at Yellingbo as an alternative capture technique to support population monitoring.During the mid-late 1990s, large numbers of nest boxes (,160) were installed throughout potential foraging habitat and were readily colonised by the species (Harley 2004b).This included targeted provision of nest boxes for colonies that were identified through stag-watching and call imitation (Smith et al. 1989;Harley et al. 2005;Harley 2015).Leadbeater's possum colonies use multiple dens within their territory, periodically shifting between them, albeit the frequency at which this occurs is highly variable (Lindenmayer and Meggs 1996;Harley 2005).Multiple nest boxes ($3) were installed for each colony that was identified, providing animals with multiple denning options.Nocturnal radio-tracking indicates that individuals typically forage within a 100-m radius of the den (D.Harley unpubl.data).

Vegetation surveys
All vegetation assessments were completed in 2019, including for abandoned territories irrespective of when they were abandoned (i.e. after 2003).This meant that the interval between site abandonment and vegetation assessment was variable.This shortcoming arose from the lack of sustained vegetation monitoring to accompany the long-term possum population monitoring.
We assessed the floristics, structure and condition of the canopy and midstorey vegetation in which lowland Leadbeater's possums forage at Yellingbo (Harley et al. 2005).Vegetation sampling was conducted within a 50-m radius of preferred den locations within territories that were currently occupied or had been previously occupied and were subsequently abandoned.Preferred denning locations (nest boxes) were selected based on several criteria, including: (i) the presence of a large Leadbeater's possum nest having been constructed in the nest box, (ii) a high frequency of possum denning records in that nest box relative to neighbouring nest boxes, and (iii) for occupied territories, nest boxes that showed a recent pattern of usage.A 50-m radius was selected to cover the core area of each colony's territory, including both the movement pathways surrounding den locations and key foraging areas, whilst avoiding overlap with neighbouring colonies.
Within the 50-m radius circular plot at each territory, six random sampling locations were selected.At each sampling location, all midstorey plants were identified to species level within a subplot of radius 2.5 m, and all canopy trees identified within a subplot of radius 5 m (the larger subplot size due to the sparser distribution of canopy trees).For each woody plant surveyed, the diameter at breast height (DBH) of each stem .1.3m in height was measured, and its status recorded as dead or alive, based on whether it supported live foliage.Also, for each canopy tree, we visually assessed crown vigour (percentage of existing branch structure supporting live foliage) to the nearest 5% (sensu Cunningham et al. 2007).
At each location we took a fish-eye lens photograph of the canopy, with the camera mounted atop a tripod at 1.3 m above the ground, to measure plant area index (PAI).All hemispherical images were processed in ImageJ using the Hemispherical 2.0 plugin.This plugin applies the minimum threshold algorithm to the blue colour plane, which has been found to be the most accurate for separating sky and vegetation (Glatthorn and Beckscha ¨fer 2014).All binary images were then processed in Gap Light Analyser to calculate plant area index (PAI) using a consistent image mask.
Average vegetation structure and condition measures were calculated for each territory.To characterise the vegetation structure, we calculated stand density (density of stems) and basal area (cross-sectional area of woody stems), including both live and dead stems.To account for the different sized subplots used to survey canopy and midstorey plants, we first converted stem density and total basal area to a per hectare measure.For each subplot we then calculated the numbers of woody stems and basal area for canopy trees, and midstorey, separately, and combined.To characterise vegetation condition, we determined percentage live basal area (percentage of basal area accounted for by live stems), tree crown vigour and plant area index, all of which have been shown to be reliable, objective measures of floodplain stand condition (Cunningham et al. 2007(Cunningham et al. , 2009)).

Data analyses
To compare vegetation measures between occupied and abandoned territories, we used generalised linear models assuming either log-normal, beta or negative binomial error, with status (occupied or abandoned) as the sole predictor.We used logistic regression implemented under leave-one-out cross validation (Sammut and Webb 2010) to test the capacity of our vegetation measures to predict occupancy.In principle, we considered that site occupancy would vary according to stem density, live basal area, and vegetation condition (i.e. percentage live basal area, tree canopy vigour and PAI).However, we could not include all the predictors together because there were insufficient data (numbers of occupied and abandoned territories) to estimate their parameters.We thus assessed support for candidate singleor two-predictor models according to Akaike information criteria corrected for small sample size (Burnham and Anderson 2002) and the classification error rate from cross-validation.The models with two predictor variables failed to converge in the cross-validation context, thus we deemed them unsuitable.
All analyses were conducted in R (R Core Team 2020).Generalised linear models of vegetation measures were fit using the glmmTMB package (Brooks et al. 2017) for ease of implementation of models assuming beta and negative binomial distributions.Cross validation was implemented using the rsample (Kuhn and Wickham 2019) and broom (Robinson and Hayes 2019) packages.All data and code used for analyses are freely available via an Open Science Foundation project page: osf.io/ecfht.

Floristic composition
Eucalyptus camphora was by far the dominant canopy tree across both occupied and abandoned territories.All other canopy trees combined, E. fulgens, E. obliqua, E. ovata and E. radiata, comprised ,16% of all trees recorded (Table 1).Greater numbers of E. camphora trees were recorded at occupied than abandoned territories (785 c.f. 538 stems/ha; Table 1).Furthermore, a mean of 78% of E. camphora trees recorded at occupied territories were alive compared to 50% at abandoned territories.
Melaleuca squarrosa accounted for more than half (53%) of all midstorey trees and shrubs recorded across all territories.The second most common midstorey species was M. ericifolia (18%), with Coprosma quadrifida, Kunzea leptospermoides, Leptospermum scoparium and L. lanigerum also commonly occurring, but typically at lower densities.Similar percentages of midstorey plants recorded were alive at occupied and abandoned territories.However, stem densities of the dominant Melaleuca spp.were several times greater at occupied territories than abandoned territories, with this difference typically attributable to dense paperbark thickets (M.squarrosa, and sometimes M. ericifolia; Table 1).

Predictors of occupancy
Logistic models of lowland Leadbeater's possum occupancy which included either combined stem density (AICc ¼ 21.31), midstorey stem density (22.59) or average tree crown vigour (23.00) were equally the most plausible single-predictor models (i.e. had AICc values ,2 different; null model AICc ¼ 29.75) and all were statistically clear predictors (P , 0.05).Combined stem density was the best predictor of occupancy with an error rate of 0.2, so it correctly predicted occupancy 80% of the time.Models using midstorey stem density and crown vigour were equally plausible on AICc values but had less favourable cross validation error rates of 0.25.All territories surveyed with .10000 stems/ha were occupied, whereas territories with a mean crown vigour of ,36% were all abandoned (Fig. 3).

Discussion
Our study presents the first quantitative data detailing the vegetation at sites inhabited by lowland Leadbeater's possums.There were clear differences in the stem density and condition of vegetation in territories currently occupied by the species at Yellingbo compared to territories that have been abandoned.Abandoned locations had relatively poor stand condition indicated by lower percentage live basal area, tree crown vigour and plant area index scores.Currently occupied territories had higher stem densities, particularly of midstorey plants, with more than three times the number of midstorey stems at occupied territories.Our findings suggest that high stem density is a critical component of lowland Leadbeater's possum habitat, and stem density alone was able to predict occupancy vis-a `-vis abandoned sites with a high (80%) degree of accuracy.The functional role of high stem density is likely to be twofold, both facilitating the possum's arboreal movements through the forest and increasing the substrate available for foraging.
The key finding of this study, that lowland Leadbeater's possums occupy sites supporting high stem densities, corresponds with extensive nocturnal observations of the possum's habitat use made at the study locality over two decades (D.Harley, pers. obs).Given that the possum is unable to glide and rarely descends to the ground, dense vegetation structure greatly facilitates the movement of this species through the forest.This finding concurs with our broader ecological understanding of this species from highland montane ash forest, where the presence and abundance Leadbeater's possum is positively correlated with basal area of Acacia spp. in the midstorey (Smith et al. 1985;Smith and Lindenmayer 1988;Lindenmayer et al. 1991a).Investigations are underway to see if this finding also applies to populations of Leadbeater's possum inhabiting subalpine woodland.Previous studies conducted in montane ash forest occupied by highland populations of Leadbeater's possum indicate that Acacia spp.provide an important food resource and movement pathways for the possum (Smith 1984;Lindenmayer et al. 1991a).Acacia spp.are virtually absent from the floodplain forest inhabited by lowland Leadbeater's possums at Yellingbo, indicating that the species does not rely on Acacia spp. as a food resource in lowland forests.Many of the occupied territories at Yellingbo are characterised by high stem density of Melaleuca spp. in the midstorey, and it is likely that they provide important foraging substrate (Harley et al. 2005).A consistent pattern observed across lowland and highland sites within the possum's range is that high rates of site occupancy coincide with the presence of dense stands of Acacia spp., Melaleuca spp., or Leptospermum spp. in the midstorey (Lindenmayer et al. 1991a;D. Harley, unpubl. data), suggesting these plants may provide an important source of exudates and/or invertebrate prey.Similarly, Goldingay et al. (2020) observed higher rates of nest box occupancy by the brush-tailed phascogale (Phascogale tapoatafa) in areas with greater stem densities, potentially reflecting the increased area of foraging substrate.
The differences in vegetation condition between occupied and abandoned territories at Yellingbo suggest that deteriorating vegetation condition has been a major driver of Leadbeater's possum population decline there.The majority of the 180 ha floodplain at Yellingbo Nature Conservation Reserve does not currently provide suitable habitat for foraging and movement by lowland Leadbeater's possum owing to the open vegetation structure and/or poor stand condition (Harley 2016).The possum's area of occupancy in the reserve has contracted by more than half, and its population size has declined by ,70% since 2003 (D.Harley, unpubl. data).In this forest type, nest boxes can provide adequate den sites (Harley 2004b), so suitable habitat is largely determined by the structural density and health of the canopy and midstorey vegetation to support foraging and movement.Quantifying food availability, as a means of determining site quality and carrying capacity, has not yet been possible owing to uncertainty surrounding specific dietary items and their relative contributions in lowland swamp forest, coupled with difficulties in reliably measuring the production of certain exudates such as honeydew.In the absence of such resolution, the floristic composition and stand density are considered to provide suitable surrogates to inform assessments of site suitability for Leadbeater's possum.Previous research on another critically endangered species at Yellingbo, the helmeted honeyeater (Lichenostomus melanops cassidix), found similar preferences for structurally complex forests characterised by high stem density (Pearce et al. 1994).
One of the major threats to forest condition at Yellingbo is altered flooding regimes (Greet 2016;Zacks et al. 2018).Within the reserve, progressive dieback of E. camphora has been documented since the early 1990s at sites now subject to near permanent inundation and waterlogging (McMahon and Franklin 1993;Greet 2016).Many of the abandoned possum territories that we surveyed were in or adjacent to these dieback-affected areas and were dominated by standing and fallen dead trees that presumably provide little or no exudate food resources for possums.Conversely, in other areas of the reserve, floodplains have been disengaged via levee banks or stream channelisation, and now flood infrequently.At these sites, reduced flooding appears to be promoting a more open forest structure less suitable for Leadbeater's possum (Harley 2016).This may be partly attributable to reduced swamp forest regeneration in the absence of flood disturbance.

Implications for lowland Leadbeater's possum conservation
The Leadbeater's possum population at Yellingbo is the last remnant of a once extensive lowland range for this species.There are no detailed habitat descriptions from other lowland sites historically inhabited by Leadbeater's possum, and notably, E. camphora is absent from these locations.Based on remnant vegetation, these are considered likely to have supported a canopy of swamp gum (E.ovata) with a dense midstorey of M. ericifolia (Yugovic and Mitchell 2006), thus bearing strong similarities to the habitat currently occupied by the possum at Yellingbo.The current lack of high-quality swamp forest dominated by E. camphora and/or E. ovata and Melaleuca shrub thickets due to historic clearing is the key impediment to recovering populations of lowland Leadbeater's possum.Habitat restoration, with the specific intent of increasing the area of available habitat for the possum and a sympatric threatened species with overlapping habitat requirements, the helmeted honeyeater, is currently underway at sites within Yellingbo Nature Conservation Reserve and several locations beyond Yellingbo.This study provides estimates of stem densities that can inform revegetation programmes aiming to provide suitable habitat for the possum at lowland sites.Our results are also being used to inform the selection of translocation sites for this critically endangered species, with an emphasis on stem density.
Based on vegetation dynamics and trajectory, the extent and condition of habitat available for Leadbeater's possums at Yellingbo are likely to continue to decline without appropriate intervention.Sites that remain occupied by the possum may be at an earlier stage of decline than those that have become abandoned.Interventions to reinstate appropriate flooding regimes are likely required to arrest further dieback of swamp forests and reconnect disengaged floodplain areas to promote vegetation regeneration.Hydrological interventions led by Melbourne Water are currently underway to assess the potential for such actions to promote habitat recovery.We suggest regular but variable flooding is essential for maintaining vegetation health and appropriate habitat structure for swamp-adapted fauna such as the lowland Leadbeater's possum (Greet et al. 2020).
Urgent interventions are also required to address the small population size and loss of genetic diversity.Evidence of inbreeding depression in this population has recently been documented (Zilko et al. 2020).A genetic rescue strategy is currently being implemented, whereby mixed pairs of lowland and highland individuals will be established in captivity and subsequently translocated to the wild in order to achieve outbreeding.
Our case study of Leadbeater's possum at Yellingbo exemplifies limitations that often beset research on critically endangered wildlife species, including (1) restricted sample sizes and study locations (in this case, there is only one study location available given a single extant lowland population), (2) data collection that commenced after the population decline commenced, and (3) vegetation data from a single point in time.The last remaining occupied habitat may not be a good guide to a species' realizable niche (Kerley et al. 2012;Simmonds et al. 2017;Lentini et al. 2018), and it is possible that Leadbeater's possums occupied a broader suite of lowland vegetation associations previously, however, appropriate historic baseline data are lacking (e.g.McClenachan et al. 2012).
Despite the inherent limitations of our study, we consider the data presented here provide insights into the key vegetation variables influencing site occupancy by Leadbeater's possum at this location, information that should be useful to managers.The alternative is to make management decisions in the absence of such data, which is not an outcome we support.Rather we encourage threatened species programmes to invest in welldesigned monitoring programmes and use insights from those data to inform decision-making.Initiating the collection of habitat condition data early in any recovery programme is strongly recommended.
This study also highlights a more general management challenge faced by conservation practitioners -simply reserving land for conservation will not always ensure the persistence of threatened species (e.g.Franc ¸oso et al. 2015).Active management may also be necessary to maintain or restore ecosystem structure and function and/or address threatening processes (e.g.climate change, altered hydrology, novel predators, browsing pressure).

Conclusion
Our findings highlight that vegetation structure and condition are strongly associated with territory occupancy by lowland Leadbeater's possum.The limited and contracting extent of high-quality lowland habitat (i.e.characterised by dense, healthy vegetation) is the major threatening process for lowland Leadbeater's possum, historically at a landscape scale and currently within Yellingbo Nature Conservation Reserve.The extent of suitable habitat is so restricted that extinction of Leadbeater's possum from lowland habitats, and the loss of this unique gene pool, is highly likely without effective management intervention.

LegendFig. 1 .
Fig. 1.The distribution of the critically endangered Leadbeater's possum in Victoria, Australia, based on Victorian Biodiversity Atlas (VBA) records.Montane forest and sub-alpine woodland habitats within the Victorian Central Highlands comprise the vast majority of extant populations of Leadbeater's possum (VBA records indicated with small black circles).Lowland populations of Leadbeater's possum, once extensive, are now confined to a single site, Yellingbo Nature Conservation Reserve (black area).

Fig. 3 .
Fig. 3. Prediction curves from logistic regression models with (a) combined canopy and midstorey stem density or (b) tree crown vigour as sole predictors of occupancy by lowland Leadbeater's possum.Grey translucent circles represent observed data.

Table 1 . Floristic composition of occupied and abandoned territories
Mean number and range of stems per hectare and percentage of stems recorded that were alive for canopy trees and midstorey trees and shrubs within territories that were either occupied (n ¼ 9) or abandoned (n ¼ 11) by lowland Leadbeater's possums at Yellingbo Nature Conservation Reserve