Abstract
Invasive predators, land clearing and altered fire regimes have been implicated in species decline and extinction worldwide. Enhanced knowledge of how these factors interact and influence medium-sized mammals is warranted. We tested three hypotheses using occupancy data for a diverse mammal assemblage including three threatened species, five common species, two introduced mesopredators and an apex predator in eastern Australia. We hypothesised that occupancy of mammal species within the assemblage would be influenced by (i) the physical environment (rainfall, vegetation type and elevation), (ii) habitat disturbance (number of fires and habitat fragmentation) and (iii) mesopredator release, whereby occupancy and/or detection of medium-sized mammals are influenced by mesopredators, the feral cat (Felis catus) and the red fox (Vulpes vulpes), which are influenced by an apex predator, the dingo (Canis familiaris). We utilised camera-trapping data from 173 sites (692 camera locations) across a north–south gradient spanning ~ 1500 km in eastern Australia. Although hypotheses i (physical environment) and ii (habitat disturbance) are not mutually exclusive, we show that the variables considered in each were only weakly correlated. We conducted occupancy modelling to investigate the physical environment and habitat disturbance hypotheses. We conducted co-occurrence modelling to investigate interactions between species. The physical environment hypothesis best supported occupancy models for six mammal species: red-necked pademelon (Thylogale thetis), bandicoots (Isoodon macrourus and Perameles nasuta), swamp wallaby (Wallabia bicolor), red-necked wallaby (Macropus rufogriseus), eastern grey kangaroo (Macropus giganteus) and feral cat. The disturbance hypothesis best supported occupancy models for four mammal species: long-nosed potoroo (Potorous tridactylus), red-necked pademelon and both mesopredators. Support for the mesopredator release hypothesis was equivocal. Large macropods showed site avoidance towards the red fox. Four species showed higher detection at sites where mesopredators were not detected. The fox showed a negative detection interaction to the dingo and the cat did not. Our study highlights how factors such as rainfall, land clearing, elevation and number of fires influence the occupancy of species within a diverse mammal assemblage at the macroecological scale. Our findings have implications for the conservation of threatened species in managed landscapes and suggestions for further research following the recent 2019–2020 wildfires.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
The world is facing a biodiversity crisis with the continuing extinction and decline of species globally (Johnson et al. 2017; Rosenberg et al. 2019). While human activities have increased the rate of species extinctions to more than 1000 times the natural background rate (Pimm et al. 1995), managed conservation areas are experiencing secondary legacy effects from human activities long after initial threats have subsided. Threats such as habitat disturbance, introduced predators and altered fire regimes impact biodiversity within protected areas and these factors are likely to be exacerbated outside of conservation lands. The plight of many ground dwelling mammals is such that their persistence requires rewilding efforts where populations are translocated to “safe-havens” such as fenced enclosures and islands where introduced predators and land clearing are absent (Legge et al. 2018). However, gaining a better understanding of how threats and environmental factors influence species persistence outside of such safe havens will be ongoing and important for biodiversity conservation.
Australia’s diverse assemblage of terrestrial mammals is amongst the most unique and distinctive in the world (Holt et al. 2013). In the past two centuries Australia’s mammals have experienced a higher rate of extinction compared to the mammal fauna of any other nation (Woinarski et al. 2015). Thirty-four of Australia’s 273 endemic mammal species have become extinct in the past 200 years (Woinarski et al. 2019) and at present, there are 30 species listed as endangered under Federal legislation with a further 46 listed as vulnerable (Environment Protection and Biodiversity Conservation Act 1999). Among those particularly affected are medium-sized terrestrial mammals (Burbidge and McKenzie 1989; Cardillo and Bromham 2001). Although the broad factors that drive global extinctions are well known (Gibbons et al. 2000; Stuart et al. 2004; Skerrat et al. 2007; Szabo et al. 2012; Sánchez-Bayo and Wyckhuys 2019), further investigation of factors causing declines in managed landscapes are paramount to inform conservation and management and arrest further declines.
The single largest threat to Australian mammals at the continental scale is predation by introduced mesopredators, primarily the European red fox (Vulpes vulpes) and feral cat (Felis catus) (Short and Smith 1994; Kinnear et al. 2002; Woinarski et al. 2015; Radford et al. 2018). At smaller spatial scales, major threats implicated in declines include habitat loss and fragmentation (Law and Dickman 1998; McAlpine and Eyre 2002; McAlpine et al. 2006; Reside et al. 2019) and altered fire regimes (Woinarski et al. 2015). Recent studies suggest that there may be a synergistic effect between the presence of introduced predators and disturbance factors such as fire (Robley et al. 2016; Hradsk et al. 2017). It is unknown how introduced mesopredators respond to disturbance factors associated with landscape fragmentation, however, it is suspected that they may benefit (Catling and Burt 1995a).
Apex predators are widely acknowledged to shape ecosystem processes through direct lethal effects that influence prey populations (Ripple et al. 2014) and through non-lethal effects that alter prey species behaviour (Shrader et al. 2008; Laundré et al. 2014). Apex predators may also influence species assemblages indirectly through the lethal and non-lethal effects they have on mesopredators (Gordon et al. 2015). Understanding the role of apex predators is fundamental to maintaining biodiversity in terrestrial ecosystems (Ritchie and Johnson 2009). However, maintaining or restoring the ecological function of apex predators is often contentious due to the direct threat they pose to humans and threatened species (Allen and Fleming 2012; Augusteyn et al. 2020) and due to predation on livestock and associated economic impacts (Fleming et al. 2006; Muhly and Musiani 2009; Chetri et al. 2019). Indeed, these conflicts with human interests frequently lead to population control of apex predators (van Eeden et al. 2018).
The mesic forests of south-eastern Australia provide refugia for many mammalian taxa that are congeneric with species that have become extinct or suffered major range contractions in arid ecosystems. The challenges of understanding the factors that influence mammal populations over this large geographic region is the presence of overlapping gradients in environmental factors that may influence populations and ultimately species persistence. For example, dingoes (Canis familiaris) and mesopredators are widespread and subject to varying levels of population control but there are pronounced gradients in topography, rainfall and temperature that may also influence endemic mammal populations. In addition, disturbances which influence mammal populations such as wildfire and land-clearing are widespread (Cox et al. 2004; Bradstock 2010; Reside et al. 2019). Consequently, large-scale studies that span these environmental gradients are needed to compliment fine-scale approaches as they encompass much wider gradients and heterogeneity in such factors.
In this study, we use data from “Wildcount” a government funded large geographic-scale camera-trapping program. We investigate three broad hypotheses concerning the relative influence that ‘bottom-up’ factors (the physical environment and habitat disturbance) and ‘top-down’ factors (predators) have on occupancy patterns of medium–large (1.3 to 19 kg) mammals (Fig. 1), including three threatened species, in diverse habitats of eastern New South Wales (Tables 1 and 2). Our first hypothesis was that many of our study species will be influenced by factors arising from the physical environment, namely, rainfall, elevation and/or vegetation type (Catling and Burt 1995b; Claridge and Barry 2000; McHugh et al. 2019). We predict that rainfall will drive site productivity and this may enable higher reproduction and therefore persistence of species. In contrast, elevation may also represent site productivity but it will also be associated with topographic complexity which may facilitate persistence (e.g. McDonald et al. 2017). Vegetation type was also investigated as it reflects species food and shelter requirements.
Our second hypothesis, which is not mutually exclusive of the physical environment hypothesis, was that habitat disturbances such as fire history and habitat fragmentation/land clearing will drive patterns in the occupancy of mammals (Claridge and Barry 2000; Norton et al. 2015). The disturbance factors we chose to investigate were the number of fires over 50 years, and several metrics relating to habitat fragmentation; perimeter/area ratio (P/A ratio) of the reserve surrounding each study site, and the percentage of cleared land surrounding each site with buffers of 5 km for endemic mammals and 10 km for introduced predators.
Our third hypothesis was that mesopredators would have a strong influence on mammal assemblages and that occupancy of mesopredators in turn would be influenced by the presence of dingoes, the apex predator in our system (Colman et al. 2014; Hunter et al. 2018; Rees et al. 2019). We tested our hypotheses in two steps. First, we conducted multi-season occupancy modelling to investigate whether any site-based covariates relating to the physical environment or habitat disturbance can explain variation in the detection of our mammal species. Second, we conducted co-occurrence modelling to investigate interactions between each of the predators (dingo, fox, cat) and threatened and non-threatened mammals. We also investigated whether the three predators showed patterns of co-occurrence that are dependent or independent of each other.
Methods
Study area
This study was conducted across the eastern portion of New South Wales (NSW) in south-eastern Australia (Fig. 2). Marked climatic differences across the study area reflect changes in latitude and elevation. The coast extends 1460 km north to south. Elevation ranged from 0 m on the coast up to 1809 m on the tablelands west of the coast. The highest rainfall occurred in sub-tropical northern NSW (annual average = 2912 mm) and generally decreased along a latitudinal gradient south-ward, with lower rainfall in the temperate south-east corner (annual average = 1523 mm). The lowest rainfall occurred in the south-western slopes region (annual average 509–675 mm).
The study area contained a range of different vegetation types with site numbers proportional to their extent: dry sclerophyll forest was present at 102 sites, wet sclerophyll forest was present at 29 sites, grassy woodland was present at five sites), heathland communities was present at seven sites, semi-arid was present at seven sites, rainforest communities were present at four sites, wetlands were present at six sites and alpine complex were present at two sites.
Survey design
Wildcount is a state-wide fauna monitoring program established by the New South Wales Office of Environment and Heritage in 2012 (EPA 2018). The project covers a large spatial gradient across eastern NSW (Fig. 2) that includes a broad range of habitats, climatic zones and fire histories. Approximately 200 monitoring sites have been established across 146 conservation reserves. Sites were selected using a stratified random selection of 1 × 1-km grid cells overlayed by a 20 × 20-km grid to prevent site clustering and to maximise dispersion. Sites were considered candidates for inclusion if they (i) occurred on NSW reserves and (ii) were intersected by a 4WD access track. Each site contained four camera traps (Reconyx HyperFireTM PC 800 (Infra-red flash) spaced within a 1 × 1-km grid approximately 500 m apart, which were surveyed annually over a 2-week period. Each camera was fixed to a tree at a height of approximately 1 m, directed at a stainless-steel tea strainer which contained a mixture of peanut butter and oats as a lure. The lure was positioned at a distance of 2 m from the camera trap at a height of approximately 20 cm. Cameras operated for 24 h per day and were set to take 3 images per trigger with no quiet time between triggers.
Images were identified to species by trained staff and trained volunteers using an image tagging system in the program Exifpro. Images were assigned an identification confidence value (definite, probable or possible). Images tagged as definite were used for modelling to reduce false positives, however, this may result in a more conservative detection history for species that are difficult to identify such as the Parma wallaby (Notamacropus parma). We accessed detection data for 17 species for the years 2012, 2013 and 2014. The 2012 monitoring period occurred between 13/2/2012 and 11/6/2012 and comprised 791 camera locations. The 2013 monitoring period occurred between 18/2/2013 and 31/5/2013 and comprised 804 camera locations. The 2014 monitoring period occurred between 25/2/2014 and 19/6/2014 and comprised 816 camera locations. We removed all data from each of the four camera locations that were not monitored for three consecutive years, where four camera-traps were not deployed and we also removed sites within the Brigalow and South-western slopes bioregions due to limited distributions of target species through these regions. This left us with 692 camera locations across 173 sites. Each reserve had 1–3 sites which comprised four camera trap locations at each (4–12 camera traps per reserve). The four camera traps at each site were pooled to meet the assumption of site independence (MacKenzie et al. 2018).
Landscape covariates
We selected six landscape covariates hypothesised to influence the occupancy of our target species. These were vegetation type, fire history (number of fires), rainfall, elevation, perimeter/area ratio of conservation reserves and % foliage projective cover (FPC) of woody vegetation within a 5-km and 10-km radius surrounding sites. We developed correlation coefficients for each pair of landscape covariates to determine the relatedness of these variables (Table 1, Supplementary Material). Our correlation matrix (Table 1, Supplementary Material) shows that our physical environment and habitat disturbance covariates are at most weakly correlated.
Bioregions
Bioregions are large geographic areas characterised by common geophysical, hydrological and climatological features that influence the distribution of plant communities, faunal assemblages and ecosystem processes at large spatial scales (Thackway and Cresswell 1995). Under the interim biogeographic Regionalisation of Australia (iBRA) framework, 85 bioregions are recognised of which 17 occur across NSW. The Wildcount dataset cover nine of these Bioregions (Fig. 3A). Many of our species do not occur uniformly across the broader study area. Attempting to model species where they do not occur is likely to produce spurious results by inflating the sample size of non-detection sites. Therefore, we restricted our species data to the Bioregions where species predominantly occurred.
Vegetation
We retrieved vegetation data from the NSW Office of Environment and Heritage (OEH) (https://data.nsw.gov.au/data/dataset/vegetation-classes-of-nsw) in December 2018. The Vegetation Formations and Classes of NSW (version 3.03) displays the major vegetation formations of NSW in raster format at a resolution of 200 m. The data were compiled from a number of vegetation maps across the state of NSW [see Keith (2004) for further information]. Vegetation formation types were extracted to each monitoring site using ArcGIS (ESRI 2011. ArcGIS desktop: Release 11. Redlands CA: Environmental systems Research Institute). Vegetation formations were then pooled based on similar characteristics e.g. where woodlands had both shrubby and grassy components they became woodland. Where 3 of 4 camera traps were located within one vegetation formation, the dominant vegetation formation took precedence. Where 2 of 4 camera traps were located within one vegetation formation, a new vegetation category ‘transition’ was created (refer Fig. 3B for vegetation types).
Rainfall
Rainfall is highly variable between years in eastern Australia due to the El Niño-Southern Oscillation (ENSO) and it is well appreciated that rainfall can drive substantial fluctuations in the abundance of forest mammals. For this reason, we used rainfall in the 3 years preceding each year of camera trapping. We retrieved gridded (5 × 5-km) average monthly rainfall data across Australia from the Australian Bureau of Meteorology (www.bom.gov.au/jsp/awap/rain/archive) in March 2018. Using ArcGIS, monthly rainfall data were firstly merged into three years of average rainfall and then each year was merged to form an average rainfall raster for the survey years. The raster was then clipped to the state of NSW and continuous values of average rainfall were extracted to each monitoring site within the study area (Fig. 3C).
Elevation
We retrieved digital elevation data from Geoscience Australia (www.ga.gov.au/scientific-topics/national-location-information/digital-elevation-data). The Digital Elevation Model (DEM) were derived from data generated from the 1 s Shuttle Radar Topography Mission (SRTM) acquired by NASA in February 2000. We extracted continuous elevational data to each camera monitoring site within the study area using ArcGIS (Fig. 3D). We derived an elevation value for each site by calculating the average of elevation from the four camera trap locations.
Fire history
We retrieved fire history data from OEH (http://datasets.seed.nsw.gov.au/dataset) in August 2016 (Fig. 3E). The data comprise a number of attributes that relate to each fire boundary polygon including fire type (prescribed burn or wildfire), year of fire, area and perimeter of fire. Due to high correlation between these fire variables, we retrieved a single fire history covariate: number of fires that have occurred over a since 1970 which ranged between zero and nine.
Land clearing/habitat fragmentation metrics
We retrieved woody vegetation and Foliage Projective Cover (FPC) spatial data from the NSW Department of Planning and Industry (DPIE) (https://datasets.seed.nsw.gov.au/dataset/landsat-woody-extent-and-foliage-projective-cover-fpc-ver-2-1-25m-20087355d). These data are derived from a time series of Landsat FPC images from 1988 to 2008 and use time series decision tree statistics to identify areas of woody vegetation and assign FPC values to them at a resolution of 25 m. We generated circular buffers around the NSW Wildcount sites at 5 km and 10 km from which we extracted FPC values. FPC values were then divided into two discrete categories (i) non-woody vegetation and (ii) woody vegetation. These values were used to generate a percentage value for cleared (grassy) land within each buffer area as a surrogate for cleared/farm land (Fig. 3F). We also generated a perimeter area ratio covariate based on the perimeter and area of each conservation reserve that Wildcount sites were located within. This was achieved by calculating the perimeter and area of reserves containing survey sites and dividing the perimeter by the area.
Single species occupancy modelling
We conducted multi-season occupancy modelling to test the physical environment and disturbance hypotheses. Modelling was implemented in program PRESENCE version 9.3 (USGS Patuxent Wildlife Research Centre, Laurel MD, USA). Multi-season models include four parameters representing different probabilities: initial occupancy (psi), detection (p), colonisation (gamma) and local extinction (epsilon) (MacKenzie et al. 2003). Our primary focus was on covariates that influenced species occupancy so we allowed colonisation and local extinction to be constant over time. For three species, gamma converged on zero so this parameter was fixed to zero. We allowed detection to either remain constant over time or to be year specific.
We conducted occupancy modelling for 12 species (Table 1; Fig. 1) including several recognised as threatened: the long-nosed potoroo, red-legged pademelon and Parma wallaby. We constructed weekly detection histories showing whether each species was detected (1) or not (0) across the three 2-week monitoring periods.
We firstly compared models for the probability of detection with occupancy set to constant. We then retained the top detection model and examined the influence of covariates, representing our hypotheses, on the probability of occupancy. Covariates used were drawn from the physical environment and the disturbance variables described above. Models were ranked using Akaike’s Information Criterion (AIC) from lowest to highest. The difference in AIC (∆AIC) was calculated between the top model and other models. A model was considered most plausible in explaining the data if it differed to the next model by > 2.0 ∆AIC (Burnham and Anderson 2004). Increasing values of ∆AIC > 4.0 indicated models had poor support (Burnham and Anderson 2004).
We commenced by fitting univariate models that aligned with our hypotheses and proceeded in the following sequence of covariates: (i) elevation, (ii) rainfall, (iii) number of fires, (iv) vegetation type, (v) cleared vegetation within a 5-km buffers for endemic mammals and 10-km buffers for predators surrounding sites and (vi) perimeter to area ratio for conservation reserves surrounding sites. We constructed dummy variables (0 or 1) to represent each vegetation type comprising the four forest and woodland types and uncommon types pooled into an ‘other’ group. This was conducted due to the way Presence reads variables within the design matrix. Rainfall, elevation and number of fires were standardised by subtracting their mean and dividing by their standard deviation. We then fitted models with two or more covariates by including those from models in which model weight (w) was > 0.1 (see Eyre 2007). We did not assess model fit because methods to assess this have not yet been devised for multi-season models (MacKenzie et al. 2018).
Co-occurrence modelling
We employed the mesopredator release hypothesis to help explain co-occurrence patterns within our occupancy data. Co-occurrence modelling can be used to investigate the interactive effects between pairs of species (MacKenzie et al. 2018). We constructed stacked detection histories of paired species, including each of the predator species with subordinate species, and with each other. We used Presence version 9.3 (USGS Patuxent Wildlife Research Centre) to model the species pairs using the formulation of Richmond et al. (2010). In this formulation one species is designated as the dominant species (A) and the other as a subordinate species (B). We designated a predator as the dominant species and the potential prey as the subordinate species. The probability of occupancy is estimated for the dominant species (psiA) and the subordinate species when the dominant species is present (psiB/A) and absent (psiB/a). The probability of detection is estimated for each species when the other is absent (pA, pB), when each of the species is present and detected (rA, rB/A), and for the subordinate species, when the dominant species is present and not detected (rB/a).
Co-occurrence modelling followed a two-step process. We firstly fitted models to investigate influences on occupancy and then retained the top model and investigated the influence of changing the detection parameters. We used landscape covariates for individual species identified as influential from single-species modelling. We compared models with and without the covariates previously identified to be influential. Covariates were retained only where they improved model fit by > 2.0AIC (see below). We then investigated whether the prey species occurred at sites independently of the predator (psiB/A ≠ psiB/a). To investigate the detection parameters all detection parameters were initially estimated separately (pA, pB, rA, rB/A, rB/a). We then investigated whether detection of the predator differed when the prey species was present or not present (rA ≠ pA); whether the detection of the prey at sites with the predator differed depending on whether the predator was detected or not (rB/A ≠ rB/a); and whether detection of the prey differed when the predator was present or not (pB ≠ rB/A). Whether the estimated parameters that were compared were equal or different was used to provide an initial finding about interactions between species pairs.
The modelling produces two derived parameters, phi (the species interaction factor) and delta (the detection interaction factor) (MacKenzie et al. 2018). Phi represents interaction between species as a function of occupancy, whilst delta estimates interaction between species as a function of detection. These parameters do not demonstrate that interaction is occurring but that occupancy or detection of two species is not random with respect to each other. If phi or delta are > 1.0, it suggests a positive association between the two species (targeting of prey by a predator or that both prefer the same habitat); if phi or delta are < 1.0, it suggests a negative association (avoidance). If phi or delta are ~ 1.0 then species associate independently. If the 95% confidence intervals of these parameters overlapped 1.0 we viewed this as lack of clear evidence of an interaction. The estimation of delta in Presence does not account for the difference between pB and rB so we also rely on whether pB, rB/A and rB/a differ to infer a detection response to the predators.
Co-occurrence models were ranked using AIC. The top model was treated as the most plausible if the second model differed by > 2.0AIC. Where the second model was within 2.0AIC of the top model we still used the top model to estimate parameters because it had fewer parameters and the choice made little difference to the parameter estimates. If individual models did not converge they were removed.
Results
Single species occupancy models
Long-nosed potoroo
The long-nosed potoroo was detected on 46 occasions across 13 sites and had a naïve occupancy of 0.12. Detection was estimated as constant across occasions (0.84 ± 0.07). The most plausible occupancy model was one that included the number of fires (Table 3). Based on model weight this model had 2.9 times more support than the null model. Gamma was estimated at 0.03 ± 0.01 and epsilon at 0.22 ± 0.12. The probability of potoroo occupancy decreased with an increasing number of fires (Fig. 4B).
Red-legged pademelon
The red-legged pademelon was detected on 66 occasions across 11 sites and had a naïve occupancy of 0.14. Detection of the red-legged pademelon was estimated as constant across occasions (0.77 ± 0.08). The most plausible occupancy model was the null model (Table 3). The addition of covariates showed no reduction in AIC. Gamma was estimated at 0 and epsilon at 0.26 ± 0.11.
Red-necked pademelon
The red-necked pademelon was detected on 80 occasions across 20 sites and had a naïve occupancy of 0.26. Detection was estimated as constant across occasions (0.81 ± 0.04). The most plausible occupancy model was one that included cleared vegetation within a 5-km radius, rainfall and number of fires (Table 3). This model had 3.5 times more support than the next model. Gamma was estimated at 0 and epsilon at 0.25 ± 0.08. The probability of occupancy decreased with increasing number of fires (Fig. 4B), increased with increasing rainfall (Fig. 4A) and increased with increasing percentage of cleared land (Fig. 4B).
Parma wallaby
The Parma wallaby was detected on 29 occasions across 11 sites and had a naïve occupancy of 0.14. Detection was estimated as constant across survey occasions (0.68 ± 0.10). The top occupancy model included the elevation covariate and was equally plausible to the null model (Table 3). In the null model initial occupancy was estimated as 0.09 ± 0.04. Gamma was estimated at 0.03 ± 0.02 and epsilon at 0.42 ± 0.14.
Bandicoots
Bandicoots were detected on 333 occasions across 103 sites and had a naïve occupancy of 0.59. Long-nosed bandicoots and northern brown bandicoots were pooled due to uncertainty in distinguishing the two species in camera trapping imagery. Detection was estimated as constant across survey occasions (0.63 ± 0.03). The most plausible occupancy model included annual rainfall (Table 3). This model had a model weight of 1.0. Gamma was estimated at 0.10 ± 0.03 and epsilon at 0.15 ± 0.04. The probability of occupancy increased with increasing rainfall (Fig. 4A).
Red-necked wallaby
The red-necked wallaby was detected on 253 occasions across 70 sites and had a naïve occupancy of 0.41. Detection was estimated as constant across occasions (0.75 ± 0.03). The top occupancy model included rainfall and elevation (Table 3). This model had a model weight of 0.99. Gamma was estimated at 0.06 ± 0.02 and epsilon at 0.17 ± 0.04. The probability of occupancy decreased with increasing rainfall but increased as elevation increased (Fig. 4A).
Swamp wallaby
The swamp wallaby was detected on 640 occasions at 161 sites and had a naïve occupancy of 0.93. Detection varied across survey years (2012 = 0.94 ± 0.02, 2013 = 0.92 ± 0.02, 2014 = 0.29 ± 0.03). The top occupancy model was one where the probability of occupancy decreased with an increase in rainfall (Table 3). This model had a model weight of 1.0. Gamma was estimated at 0.34 ± 0.09 and epsilon at 0.10 ± 0.03.
Eastern grey kangaroo
The eastern grey kangaroo was detected on 246 occasions across 72 sites and had a naïve occupancy of 0.42. Detection was estimated as constant across occasions (0.63 ± 0.03). The most plausible occupancy model included rainfall (Table 3). This model had a model weight of 1.0. Gamma was estimated at 0.02 ± 0.02 and epsilon at 0.09 ± 0.03. The probability of eastern grey kangaroo occupancy decreased with increasing rainfall (Fig. 4A).
Feral cat
The feral cat was detected on 105 occasions across 66 sites and had a naïve occupancy of 0.38. Detection was estimated as constant across occasions (0.19 ± 0.02). The top model contained elevation and Perimeter/Area ratio (Table 3). This model had a model weight of 0.98. Gamma and epsilon were fixed at 0.0. The probability of occupancy increased as elevation increased (Fig. 4A) and decreased as the Perimeter/Area ratio of conservation reserves increased (Fig. 4B).
Red fox
The red fox was detected on 287 occasions across 95 sites and had a naïve occupancy of 0.55. Detectability was estimated as constant across occasions (0.63 ± 0.03). The top occupancy model included the percentage of cleared land within a 10-km radius (Table 3). This model had a model weight of 1.0. Gamma was estimated at 0.16 ± 0.03 and epsilon at 0.12 ± 0.04. The probability of occupancy increased as the percentage of cleared land increased (Fig. 4B).
Dingo
The dingo was detected on 34 occasions across 25 sites and had a naïve occupancy of 0.14. Detectability was estimated as constant across occasions (0.30 ± 0.11). The top occupancy model included woodland vegetation however this model was equally plausible when compared to the null model (Table 3). Gamma was estimated at 0.08 ± 0.03 and epsilon at 0.74 ± 0.16.
Summary of hypothesis evaluation
The physical environment hypothesis had support for most generalist species whereas the habitat disturbance hypothesis was supported for sensitive habitat specialists and introduced mesopredators. Of the covariates representing the physical environment hypothesis, rainfall had the strongest overall influence over endemic mammal occupancy with two small forest dwelling species (red-necked pademelon and bandicoots) showing a positive response to rainfall and three herbivorous large macropod species (eastern grey kangaroo, red-necked wallaby and swamp wallaby) showing a negative response. Vegetation type was not influential over any species and elevation was influential over the red-necked wallaby and the feral cat. Of the covariates representing the habitat disturbance hypothesis, the long-nosed potoroo and red-necked pademelon were negatively influenced by number of fires and the red-necked pademelon was negatively influenced by % of cleared land in 5-km radius of sites. The two introduced mesopredators benefited from habitat disturbance where the red fox showed a very strong response to % cleared land in 10-km radius of sites and the feral cat showed a strong negative response to the perimeter area ratio of reserves.
Co-occurrence occupancy models
Red fox with other species
All the top models except that for the potoroo, the red-necked wallaby and feral cat included separate parameters for psiB/A and psiB/a (Table 4). However, only the swamp wallaby and the eastern grey kangaroo appeared to show site avoidance of the red fox as indicated by values of phi < 1.0 (Fig. 5). The red-legged pademelon, Parma wallaby and bandicoots showed positive site associations with the fox however only the red-legged pademelon showed phi values > 1.0 (Fig. 5).
Four species including the red-legged pademelon, Parma wallaby, red-necked pademelon and bandicoots showed much higher detections at sites without the fox (pB) compared to sites with the fox (rB/A) (Table 4). Estimates of delta for these species overlapped or were approximately 1.0 suggesting no detection interactions where both species occurred (Fig. 6).
Feral cat with other species
The parameters for psiB/A and psiB/a were estimated as equal in the top models for potoroos and foxes (Table 5). None of the estimates of phi suggested avoidance of cats with values at or above 1.0 (Fig. 5). The red-legged pademelon and eastern grey kangaroo showed phi values above 1.0 suggesting positive a site interaction. None of the species showed a detection avoidance at sites where cats occurred though red-legged pademelon, Parma wallaby, bandicoots and red-necked wallaby had higher detection at sites where cats did not occur (Table 5). The estimates of delta for the red-legged pademelon, and bandicoots showed a positive association with cat detection (Fig. 6).
Dingo with other species
The parameters for psiB/A and psiB/a were estimated as equal in the top models for long-nosed potoroo, red-legged pademelon, swamp wallaby, eastern grey kangaroo and the red fox suggesting there was no site interaction with these species (Table 6). Three species (Parma wallaby, red-necked pademelon and red-necked wallaby) showed higher occupancy at sites occupied by dingoes compared to where dingoes were absent with Parma wallaby and red-necked wallaby showing phi values > 1 suggesting positive site interactions (Fig. 5). Bandicoots showed a negative site association with the dingo (Fig. 5). Several species showed detection responses to the dingo (Table 6). The long-nosed potoroo and eastern grey kangaroo had higher detection when dingoes weren’t detected whereas Parma wallaby and red-necked wallaby had higher detection at sites where dingoes did not occur. At sites where foxes and dingoes occurred foxes showed lower detection when dingoes were detected. Both the eastern grey kangaroo and red fox had delta estimates < 1.0 suggesting avoidance in the presence of dingoes (Fig. 6).
Discussion
We investigated the extent to which three non-exclusive hypotheses can explain occupancy patterns of medium and large endemic mammals and their introduced predators across a broad geographic scale in eastern Australia. We did not expect uniform responses by species to the factors representing each hypothesis due to differences in species’ life histories. Our results supported the physical environment hypothesis for six mammal species (Table 7), the habitat disturbance hypothesis was supported for four mammal species (Table 7) and support for the mesopredator release hypothesis was equivocal (Table 7). Findings from our large-scale study present both opportunities and constraints which can guide further research that aims to untangle relationships between our endemic mammal species, introduced predators and the biotic and abiotic environment. The varied responses of species to our hypotheses outlines some important findings regarding single-species occupancy modelling over the broad geographic gradient in this study and also outlines some challenges with conducting co-occurrence occupancy modelling over such gradients. We now discuss each hypothesis in turn and conclude with implications for further research, conservation and management.
Physical environment hypothesis
Our study area encompassed a very broad geographic scale so we expected that environmental factors would have a strong influence on occupancy of most of our study species. Thus, it was not surprising that rainfall was an influential predictor of mammal occupancy. Rainfall is a well-known factor driving geographic patterns in mammal assemblages (Heaney 2001; Olff et al. 2002). Forest specialists such as bandicoots and red-necked pademelons showed a positive response to rainfall while grassland and open habitat specialists such as eastern grey kangaroos, and red-necked wallabies were negatively influenced by rainfall. We equate this response to a broad influence on plant community type that reflects the rainfall gradient. The lack of influence of rainfall on the three threatened species suggests that productivity may not be a primary determinant of their persistence in the landscape.
Although elevation can have a pronounced influence on species’ distributions (e.g. Ramírez-Bautista and Williams 2019; Campera et al. 2020) we found only one endemic species, the red-necked wallaby, responded (positively) to a broad gradient (1800 m) in elevation. The lack of response from other endemic species may reflect that our sampling was mostly focused on forest habitats and that most of our focal species are habitat generalists. The Parma wallaby generally occupies areas of high elevation across our study area, however we only found a weak effect of elevation on this species which may have been hampered by a small sample size. One important relationship we identified was that the probability of occupancy of the feral cat increased with elevation. The mechanism behind this relationship may be that it is a reflection of higher productivity at high-elevation sites that induces higher abundance of prey such as small mammals (Bateman et al. 2010). Small mammals comprise approximately 70% of feral cat diet in Australia (Murphy et al. 2019) and high elevation sites where small mammals are abundant may support higher number of cats.
Habitat disturbance hypothesis
We hypothesised that some mammal species would be influenced by fires, percentage of cleared land and the perimeter/area ratio of reserves. Fire is a frequent disturbance in Australia’s forested ecosystems and plays a significant role in altering the composition and structure of vegetation communities and habitat (Bradstock 2010). We found a negative relationship between the number of fires over the past 50 years and the occupancy of the threatened long-nosed potoroo and the non-threatened, red-necked pademelon. This is a novel finding for the red-necked pademelon but not for the long-nosed potoroo. For the long-nosed potoroo, the model that included number of fires showed a modest improvement to model fit compared to the null model. However, previous studies that have investigated the post-fire response of the long-nosed potoroo suggest that occupancy will increase over time following wildfire, with > 15–20 years being optimal, as this fire interval allows for the development of dense ground cover habitat (Claridge and Barry 2000; Norton et al. 2015). The red-necked pademelon showed a stronger negative response to number of wildfires when compared to the long-nosed potoroo which may have resulted from a larger sample size. The ability of species to recover post-fire when vegetation structure is redeveloping will depend on species mobility and the availability of unburnt refugia in the surrounding landscape (e.g. Leonard et al. 2014). Recent studies in temperate Australia suggest that there may be a synergistic effect between fires and introduced predators, whereby fire implemented in the presence of introduced predators have negative effects on medium-sized mammals including potoroos (Robley et al. 2016; Hradsk et al. 2017). However, this finding may not be uniform across the distribution of this species. Long-nosed potoroo and predator activity did not change following prescribed burns in sub-tropical Australia (McHugh et al. 2020). Greater caution may be required in using fire in this species’ habitat in the southern but not northern part of its distribution due to the varied abundance of foxes.
One limitation of our study was that we were not able to ascertain the response of species to prescribed burns due to a smaller sample of sites where prescribed burns had occurred. Wildfires will generally encompass large areas and have a high intensity/severity (Leonard et al. 2014; Chia et al. 2016), whereas prescribed burns will generally be small in scale and conducted during cool weather with low-moderate intensity/severity (McHugh et al. 2020). Consequently, wildfires can be expected to be more detrimental to potoroos compared to prescribed burns. Of particular concern is that the 2019–2020 wildfires in eastern Australia overlapped approximately 40% of the potoroo’s NSW location records (DPIE 2020).
Habitat loss and fragmentation due to clearing of vegetation is considered a threat to biodiversity at a global scale (Pimm and Raven 2000). Native habitats in eastern Australia have been extensively cleared historically (Cox et al. 2004) and this continues in contemporary times (Reside et al. 2019). Understanding the influence of this on biodiversity is critical for conservation. We identified a negative response by the red-necked pademelon to the extent of clearing within a 5-km radius of survey site. Other native species were not influenced perhaps because our survey sites were contained within conservation reserves. Some important findings relating to habitat disturbance are that we found that introduced predators were influenced by land clearing/habitat fragmentation covariates; red fox occupancy increased as clearing increased within a 10-km buffer of sites, whilst feral cat occupancy decreased as the P/A ratio of conservation reserves increased. These are important findings to help understand factors that influence these introduced predators in fragmented landscapes of eastern Australia. These findings can guide conservation efforts targeting these species regarding population control.
Mesopredator release hypothesis
We predicted that dingoes would have a negative association with foxes and cats, which in turn would have a positive association with their prey. We found no evidence of a site–scale interaction between dingoes and feral cats or red foxes. However, we found a negative detection interaction whereby fox detection was significantly lower when dingoes were also detected compared to when they were absent. This suggests that foxes show temporal avoidance of dingoes. Previous studies have indicated that smaller predators may avoid larger predators due to the risk of intra-guild predation (Palomares and Caro 1999; Linnell and Strand 2000; Brawata and Neeman 2011). Lack of site interaction between dingoes and feral cats is consistent with studies finding little support for the idea that cats modify their use of habitat in response to dingo activity (McHugh et al. 2019; Fancourt et al. 2019).
There is an emerging body of literature that is questioning the inference that can be drawn from multi-species occupancy modelling with regard to survey design i.e. spatiotemporal extent, types of detectors used, covariates used, choice of field methods and statistical tools (Devarajen et al. 2020). Experimental evidence and observation support the notion that species interactions influence co-occurrence patterns, however Blanchet et al. (2020) question to what extent the signal of interaction can be retrieved from observational data. A limitation in the present study is that we used weekly detection periods to improve the detection of co-occurrence signals which in-turn may weaken the inference that can be drawn, a challenge that presents itself where dominant and subordinate subject species have limited spatial overlap (Letnic et al. 2009; Brawata and Neeman 2011; Colman et al. 2014, 2015). For example, the detection window of one week may provide opportunity for species to co-occur without directly experiencing each other through using risk averse behaviours. Finding a trade-off between opening the detection window and deriving strong inference from field data remains a challenge for co-occurrence modelling and we hope that through identifying this challenge in the present study, further research may benefit when considering survey design and statistical methods (Devarajen et al. 2020).
In addition, covariates that may be influential but are not considered may weaken inference drawn from co-occurrence modelling (Devarajen et al. 2020). For example, habitat preference may mediate co-occurrence of species (Estevo et al. 2017; McHugh et al. 2019) and many species in this study show a preference towards habitat structure at finer scales (McHugh et al. 2019). Fine scale habitat structure data were not available across the broad scale of this study though it is likely to influence several species in this study and potentially mediate co-occurrence patterns. In support of our findings that did not find a site interaction between the dingo and the red fox, previous studies investigating dingo–fox interactions have found that fox abundance/activity tends to be lower at sites where dingo populations are not controlled and that there is little spatial overlap between them (Letnic et al. 2009; Brawata and Neeman 2011; Colman et al. 2014, 2015).
We found no evidence that foxes avoid dingoes at the site scale, however, this result may be because there was relatively little spatial overlap between these species in our dataset. The naïve occupancy of the dingo was 0.14 whilst that of the red fox was 0.55. The relatively low occupancy of dingoes may reflect ongoing control of populations over much of the study region (Colman et al. 2014; Ballard et al. 2020). Consistent with the idea that there is limited spatial overlap between dingoes and foxes, a previous study which deployed cameras at 298 sites across nine forested conservation reserves in NSW detected dingoes at 18% of sites but detected foxes at just 2% of sites (McHugh et al. 2019).
We found no evidence of a negative site association of the mesopredators and the smaller macropods and bandicoots. Foxes had a positive site-based association with the red-legged pademelon and a negative association with larger macropods (swamp wallaby and eastern grey kangaroo). An explanation for this may be that small macropods have the ability to co-exist with foxes through displaying risk averse behaviours i.e. concealment within dense habitat (Signorell et al. 2010) whereas larger macropods cannot and are more inclined to use rapid movement to evade predation (Creel et al. 2005). Furthermore, detection of bandicoots and small macropods was much higher at sites without foxes compared to sites with foxes (Table 2), supporting the idea that behavioural avoidance may be taking place.
Similarly, the feral cat had positive site-based occupancy associations with several species (red-legged pademelon, Parma wallaby and red-necked wallaby) and all of these species had higher detection values at sites without cats (Table 2). Dingoes showed a positive site-based occupancy association with some small macropods (Parma wallaby and red-necked pademelon) and a negative association with one large macropod: the eastern grey kangaroo, however, bandicoots also displayed a negative response to sites where dingoes were present (Fig. 5). Detection avoidance of the dingo was displayed by the eastern grey kangaroo which is frequently preyed upon by the dingo (Wright 1993) and interestingly this species showed a positive detection interaction with the smaller red fox. Previous literature suggests that red foxes pose a significant threat to eastern grey kangaroo joeys that can limit population growth of kangaroos (Banks et al. 2000) though its likely adult eastern grey kangaroos can avoid fox predation. Despite the apparent co-existence of endemic mammals and introduced mesopredators within the mesic conservation reserves of eastern Australia, concern arises of the impact of feral cats in high-elevation areas (Fig. 4A) and foxes in fragmented areas (Fig. 4B) where their occupancy is relatively high and therefore the threat to endemic species is elevated.
Implications for further research, conservation and management
Our study has provided knowledge and insights on the factors the drive occupancy for endemic mammal species and introduced predators. Rainfall is a key driver of endemic mammal occupancy in eastern Australia. The wetter regions of the study area support habitat for small macropods including threatened species; Parma wallaby, red-legged pademelon, long-nosed potoroo. Although these species have distributions that surpass the NSW state border, high rainfall regions eastern Australia will be important in providing habitat refugium for these threatened species and also can guide decision making with regard to prioritising recovery sites.
Habitat disturbance remains an insidious threat to threatened mammal species in eastern Australia, in particular the threat of wildfires. During late 2019 and early 2020 south-east Australia experienced some of the largest wildfires on record. The 2019–2020 Australian bushfire season occurred during an extended period of low rainfall and record-breaking temperatures (Filkov et al. 2020). It is estimated that the mega-fires burnt ~ 97,000 km2 across southern and eastern Australia, which is considered habitat for 832 species of native vertebrate fauna (Ward et al. 2020). The present study demonstrated the negative influence of number of fires on the long-nosed potoroo and the red-necked pademelon. The 2019–2020 wildfires in eastern Australia overlapped approximately 40% of the potoroo’s NSW location records (DPIE 2020) which is alarming given the impacts that introduced predators may have in temperate regions on long-nosed potoroo distribution following fire (Robley et al. 2016; Hradsk et al. 2017), and we recommend that immediate and further research is required on long-nosed potoroo populations effected by recent wildfires.
In addition, the present study found that landscape fragmentation directly influences introduced mesopredators: the red fox and feral cat across the study area. Our finding that occupancy of the red fox increased with percentage of cleared land is consistent with a study conducted 25 years previously (Catling and Burt 1995a). Given the paucity of fox records in forested areas of north-east NSW and the diversity and abundance of mammal species (Catling and Burt 1997), this region can clearly be identified as refugium for many threatened mammals in addition to the long-nosed potoroo, Parma wallaby and the red-legged pademelon. Threatened species in fragmented south-east temperate regions will benefit from ongoing efforts to control fox populations through lethal baiting.
There is a view that mesic ecosystems of eastern Australia support lower feral cat densities compared to their xeric counterparts (Legge et al. 2017) and therefore the impact of cats on biodiversity may be lower in mesic ecosystems (Doherty et al. 2017; Woinarski et al. 2017). However, our finding that feral cat occupancy increased with elevation has not been identified previously and raises concern that cats may potentially have more severe impacts on fauna in forested high-elevation areas than previously recognised. We suspect that the mechanism that allows for such high cat occupancy in high-elevation forested areas across our study region is that they provide ample food resources for feral cats (small mammals, reptiles and birds) and also habitat structural complexity that allows cats to avoid potential predators. Given that invasive mesopredators have had devastating impacts on biodiversity in xeric systems (Burbidge and McKenzie 1989; Cardillo and Bromham 2001; Woinarski et al. 2015) we recommend that further research is conducted on the ecology and impacts of cats in high elevation forests.
Our study and its design was not without its limitations, including that lure type used was suboptimal for detecting predators (Hanke and Dickman 2013) and our camera traps were located off road within vegetated habitat rather than on tracks, which may result in suboptimal detection of predators (Read et al. 2015; Geyle et al. 2020). However, lure type was standardised across all sites and each site had four widely spaced camera traps, which undoubtedly increased detectability of predators (Stokeld et al. 2015). Also, sites were intersected by a 4WD access track and therefore roads/tracks were in close proximity to camera trap locations. Further modelling on predator interactions within this mammal assemblage may benefit from considering these limitations.
The present study has outlined some important findings regarding single-species occupancy modelling over a broad geographic gradient in eastern Australia and also some challenges associated with co-occurrence occupancy modelling. The very broad study area gives rise to a wide range of variables that could influence our mammal assemblage and by necessity, some of our predictor variables are coarse. Also, alternative hypotheses may be influential at finer scales when considering single-species occupancy modelling, for example habitat structure and resource availability. Similarly, alternative hypotheses may be considered for two-species co-occurrence interactions (Blanchet et al. 2020). However, the benefit of this large-scale approach is that it has evaluated mammal responses to predictor variables over a range of values that is often not possible in studies conducted over smaller geographic scales, and it has encompassed many species not possible at finer scales. Further considerations regarding co-occurrence occupancy modelling should be given to the spatiotemporal overlap of species which will remain a challenge when considering the dingo and the red fox.
Data availability
Raw data will be made openly available following publication. All data will be made available in Dryad following publication.
References
Allen BL, Fleming PJS (2012) Reintroducing the dingo: the risk of dingo predation to threatened vertebrates of western New South Wales. Wildl Res 39(1):35–50
Augusteyn J, Rich M, Story G, Nolan B (2020) Canids potentially threaten bilbies at Astrebla Downs National Park. Australian Mammalogy. https://doi.org/10.1071/AM20034
Ballard G, Fleming PJS, Meek PD, Doak S (2020) Aerial baiting and wild dog mortality in south-eastern Australia. Wildl Res 47(2):99–105. https://doi.org/10.1071/WR18188
Banks PB, Newsome AE, Dickman CR (2000) Predation by red foxes limits recruitment in populations of eastern grey kangaroos. Austral Ecol 25:283–291. https://doi.org/10.1046/j.1442-9993.2000.01039.x
Bateman BL, Kutt AS, Vanderduys EP, Kemp JE (2010) Small-mammal species richness and abundance along a tropical altitudinal gradient: an Australian example. J Trop Ecol 26(2):139–149. https://doi.org/10.1017/S0266467409990460
Blanchet FG, Cazelles K, Gravel D (2020) Co-occurrence is not evidence of ecological interactions. Ecol Lett 23:1050–1063. https://doi.org/10.1111/ele.13525
Bradstock RA (2010) A biogeographic model of fire regimes in Australia: current and future implications. Glob Ecol Biogeogr 19(2):145–158. https://doi.org/10.1111/j.1466-8238.2009.00512.x
Brawata RL, Neeman T (2011) Is water the key? Dingo management, intraguild interactions and predator distribution around water points in arid Australia. Wildl Res 38:426–436
Burbidge AA, McKenzie NL (1989) Patterns in the modern decline of western Australia’s vertebrate fauna: causes and conservation implications. Biol Cons 50(1):143–198. https://doi.org/10.1016/0006-3207(89)90009-8
Burnham KP, Anderson DR (2004) Multimodel inference: understanding AIC and BIC in model selection. Sociol Res Methods 33(2):261–304
Campera M, Santini L, Balestri M, Nekaris KAI, Donati G (2020) Elevation gradients of lemur abundance emphasise the importance of Madagascar’s lowland rainforest for the conservation of endemic taxa. Mammal Rev 50:25–37
Cardillo M, Bromham L (2001) Body size and risk of extinction in Australian mammals. Conserv Biol 15(5):1435–1440
Catling P, Burt R (1995a) Why are red foxes absent from some eucalypt forests in eastern New South Wales? Wildl Res 22(4):535–545. https://doi.org/10.1071/WR9950535
Catling PC, Burt RJ (1995b) Studies of the ground dwelling mammals of eucalypt forests in south-eastern New South Wales: the effect of habitat variables on distribution and abundance. Wildl Res 22:271–288
Catling PC, Burt RJ (1997) Studies of the ground-dwelling mammals of eucalypt forests in north-eastern New South Wales: the species, their abundance and distribution. Wildl Res 24:1–19
Chetri M, Odden M, Devineau O, Wegge P (2019) Patterns of livestock depredation by snow leopards and other large carnivores in the Central Himalayas, Nepal. Global Ecol Conserv 17:e00536. https://doi.org/10.1016/j.gecco.2019.e00536
Chia EK, Bassett M, Leonard SWJ, Holland GJ et al (2016) Effects of the fire regime on mammal occurrence after wildfire: site effects vs landscape context in fire-prone forests. For Ecol Manag 363:130–139
Claridge AW, Barry SC (2000) Factors influencing the distribution of medium-sized ground-dwelling mammals in southeastern mainland Australia. Austral Ecol 25(6):676–699
Colman NJ, Gordon CE, Crowther MS, Letnic M (2014) Lethal control of an apex predator has unintended cascading effects on forest mammal assemblages. Proc R Soc B 281:20133094
Colman NJ, Crowther MS, Letnic M (2015) Macroecological patterns in mammal abundance provide evidence that an apex predator shapes forest ecosystems by supressing herbivore and mesopredator abundance. J Biogeogr 42:1974–1985
Cox MP, Dickman CR, Hunter J (2004) Effects of rainforest fragmentation on non-flying mammals of the Eastern Dorrigo Plateau, Australia. Biol Cons 115:175–189
Creel S, Winnie J, Maxwell, Hamlin K, Creel M (2005) Elk Alter habitat selection as an antipredator response to wolves. Ecology 86(12):3387–3397
Devarajan K, Morelli TL, Tenan S (2020) Multi-species occupancy models: review, roadmap, and recommendations. Ecography 43:1612–1624. https://doi.org/10.1111/ecog.04957
Doherty TS, Dickman CR, Johnson CN, Legge SM et al (2017) Impacts and management of feral cats Felis catus in Australia. Mammal Rev 47(2):83–97. https://doi.org/10.1111/mam.12080
DPIE (2020) Wildlife and conservation bushfire recovery; immediate response January 2020. Department of Planning, Industry and Environment, Sydney
EPA (2018) NSW State of the Environment report. Available at https://www.soe.epa.nsw.gov.au/all-themes/biodiversity/native-fauna
EPBC Act (1999) retrieved from: https://www.environment.gov.au/cgi-bin/sprat/public/publicthreatenedlist.pl
Estevo CA, Nagy-Reis MB, Nichols JD (2017) When habitat matters: habitat preferences can modulate co-occurrence patterns of similar sympatric species. PLoS ONE 12(7):e0179489. https://doi.org/10.1371/journal.pone.0179489
Eyre TJ (2007) Regional habitat selection of large gliding possums at forest stand and landscape scales in southern Queensland, Australia. II. Yellow-bellied glider (Petaurus australis). For Ecol Manag 239:136–149
Fancourt BA, Cremasco P, Wilson C, Gentle MN (2019) Do introduced apex predators suppress introduced mesopredators? A multiscale spatiotemporal study of dingoes and feral cats in Australia suggests not. J Appl Ecol 56(12):2584–2595. https://doi.org/10.1111/1365-2664.13514
Filkov AI, Ngo T, Matthews S, Telfer S, Penman TD (2020) Impact of Australia’s catastrophic 2019/20 bushfire season on communities and environment. Retrospective analysis and current trends. J Saf Sci Resil 1(1):44–56
Fleming PJS, Allen LR, Lapidge SJ et al (2006) A strategic approach to mitigating the impacts of wild canids: proposed activities of the Invasive Animals Cooperative Research Centre. Aust J Exp Agric 46(7):753–762. https://doi.org/10.1071/EA06009
Geyle HM, Stevens M, Duffy R, Greenwood L et al (2020) Evaluation of camera placement for detection of free-ranging carnivores; implications for assessing population changes. Ecol Solut Evid 1(1):e12018. https://doi.org/10.1002/2688-8319.12018
Gibbons JW, Scott DE, Ryan TJ, Buhlmann KA et al (2000) The Global Decline of Reptiles, Déjà Vu Amphibians: Reptile species are declining on a global scale. Six significant threats to reptile populations are habitat loss and degradation, introduced invasive species, environmental pollution, disease, unsustainable use, and global climate change. Bioscience 50(8):653–666. https://doi.org/10.1641/0006-3568(2000)050[0653:tgdord]2.0.co;2
Gordon CE, Feit A, Grüber J, Letnic M (2015) Mesopredator suppression by an apex predator alleviates the risk of predation perceived by small prey. Proc R Soc B Biol Sci 282(1802):20142870
Hanke PU, Dickman CR (2013) Sniffing out the stakes: hair-snares for wild cats in arid environments. Wildl Res 40(1):45–51. https://doi.org/10.1071/WR12210
Heaney LR (2001) Small mammal diversity along elevational gradients in the Philippines: an assessment of patterns and hypotheses. Glob Ecol Biogeogr 10(1):15–39. https://doi.org/10.1046/j.1466-822x.2001.00227.x
Holt BG, Lessard JP, Borregaard MK, Fritz SA et al (2013) An update of Wallace’s zoogeographic regions of the world. Science 339(6115):74–78. https://doi.org/10.1126/science.1228282
Hradsk BA, Mildwaters C, Ritchie EG, Christie F, Di Stefano J (2017) Responses of invasive predators and native prey to a prescribed forest fire. J Mammal 98(3):835–847
Hunter DO, Lagisz M, Leo V, Nakagawa S, Letnic M (2018) Not all predators are equal: a continent-scale analysis of the effects of predator control on Australian mammals. Mammal Rev 48(2):108–122. https://doi.org/10.1111/mam.12115
Johnson CN, Balmford A, Brook BW, Buettel JC et al (2017) Biodiversity losses and conservation responses in the Anthropocene. Science 356(6335):270–275. https://doi.org/10.1126/science.aam9317
Keith D (2004) Ocean shores to desert dunes: the native vegetation of New South Wales and the ACT. New South Wales Government, Sydney
Kinnear JE, Sumner NR, Onus ML (2002) The red fox in Australia—an exotic predator turned biocontrol agent. Biol Conserv 108:335–359
Laundré JW, Hernández L, Medina PL, Campanella A et al (2014) The landscape of fear: the missing link to understand top-down and bottom-up controls of prey abundance. Ecology 95(5):1141–1152
Law BS, Dickman CR (1998) The use of habitat mosaics by terrestrial vertebrate fauna: implications for conservation and management. Biodivers Conserv 7(3):323–333
Legge S, Murphy BP, McGregor H, Woinarski JCZ et al (2017) Enumerating a continental-scale threat: how many feral cats are in Australia? Biol Cons 206:293–303. https://doi.org/10.1016/j.biocon.2016.11.032
Legge S, Lindenmayer DB, Robinson NM, Scheele BC, Southwell DM, Wintle BA, Woinarski JCZ, Bayraktarov E (2018) Introduction: Making it count In Legge, N. Robinson, D. Lindenmayer, B. Scheele, D. Southwell, B. Wintle (Eds.) Monitoring Threatened Species and Ecological Communities. (pp. 1–41). CSIRO Publishing, Clayton South, Victoria
Leonard SWJ, Bennett AF, Clarke MF (2014) Determinants of the occurrence of unburnt forest patches: potential biotic refuges within a large, intense wildfire in south-eastern Australia. For Ecol Manage 314:85–93. https://doi.org/10.1016/j.foreco.2013.11.036
Letnic M, Koch F, Gordon C, Crowther MS, Dickman CR (2009) Keystone effects of an alien top-predator stem extinctions of native mammals. Proc R Soc B 276:3249–3256
Linnell JDC, Strand O (2000) Interference interactions, co-existence and conservation of mammalian carnivores. Divers Distrib 6(4):169–176. https://doi.org/10.1046/j.1472-4642.2000.00069.x
MacKenzie DI, Nichols JD, Hines JE, Knutson MG, Franklin AB (2003) Estimating site occupancy, colonization, and local extinction when a species is detected imperfectly. Ecology 84:2200–2207
MacKenzie DI, Nichols JD, Royle JA, Pollock KH et al (2018) Occupancy estimation and modelling: Inferring patterns and dynamics of species she, 2nd edn. Academic Press, San Diego
McAlpine CA, Eyre TJ (2002) Testing landscape metrics as indicators of habitat loss and fragmentation in continuous eucalypt forests (Queensland, Australia). Landsc Ecol 17(8):711–728
McAlpine CA, Rhodes JR, Callaghan JG, Bowen ME et al (2006) The importance of forest area and configuration relative to local habitat factors for conserving forest mammals: a case study of koalas in Queensland, Australia. Biol Conserv 132(2):153–165. https://doi.org/10.1016/j.biocon.2006.03.021
McDonald PJ, Griffiths AD, Nano CEM, Dickman CR (2015) Landscape-scale factors determine occupancy of the critically endangered central rock-rat in arid Australia: The utility of camera trapping. Biol Conserv 191:83–100
McHugh D, Goldingay RL, Link J, Letnic M (2019) Habitat and introduced predators influence the occupancy of small threatened macropods in subtropical Australia. Ecol Evol 9(11):6300–6317. https://doi.org/10.1002/ece3.5203
McHugh D, Goldingay RL, Parkyn J, Goodwin A, Letnic M (2020) Short-term response of threatened small macropods and their predators to prescribed burns in subtropical Australia. Ecol Manag Restor 21(2):97–107. https://doi.org/10.1111/emr.12407
Muhly TB, Musiani M (2009) Livestock depredation by wolves and the ranching economy in the Northwestern U.S. Ecol Econ 68(8):2439–2450. https://doi.org/10.1016/j.ecolecon.2009.04.008
Murphy BP, Woolley LA, Geyle HM, Legge SM et al (2019) Introduced cats (Felis catus) eating a continental fauna: the number of mammals killed in Australia. Biol Cons 237:28–40. https://doi.org/10.1016/j.biocon.2019.06.013
Norton MA, Prentice A, Dingle J, French K, Claridge AW (2015) Population characteristics and management of the long-nosed potoroo (Potorous tridactylus) in high-quality habitat in the Southern Highlands of New South Wales. Australian Mammalogy 37:67–74
Olff H, Ritchie ME, Prins HHT (2002) Global environmental controls of diversity in large herbivores. Nature 415(6874):901–904. https://doi.org/10.1038/415901a
Palomares F, Caro TM (1999) Interspecific killing among mammalian carnivores. Am Nat 153(5):492–508. https://doi.org/10.1086/303189
Pimm S, Raven P (2000) Extinction by numbers. Nature 403:843–845. https://doi.org/10.1038/35002708
Pimm, SL, Russell GJ, Gittleman JL, Brooks TM (1995) The future of biodiversity. Science 269(5222):347–350
Radford JQ, Woinarski JCZ, Legge S, Baseler M (2018) Degrees of population-level susceptibility of Australian terrestrial non-volant mammal species to predation by the introduced red fox Vulpes vulpes) and feral cat (Felis catus). Wildl Res 45(7):645–657. https://doi.org/10.1071/WR18008
Ramírez-Bautista A, Williams J (2019) The importance of productivity and seasonality for structuring small rodent diversity across a tropical elevation gradient. Oecologia 190:275–286. https://doi.org/10.1007/s00442-018-4287-z
Read JL, Bengsen AJ, Meek PD, Moseby KE (2015) How to snap your cat: optimum lures and their placement for attracting mammalian predators in arid Australia. Wildl Res 42(1):1–12. https://doi.org/10.1071/WR14193
Rees MW, Pascoe JH, Wintle BA, Le Pla M et al (2019) Unexpectedly high densities of feral cats in a rugged temperate forest. Biol Cons 239:108287. https://doi.org/10.1016/j.biocon.2019.108287
Reside AE, Cosgrove AJ, Pointon R, Trezise J et al (2019) How to send a finch extinct. Environ Sci Policy 94:163–173. https://doi.org/10.1016/j.envsci.2019.01.005
Richmond OMW, Hines JE, Beissinger SR (2010) Two-species occupancy models: a new parameterization applied to co-occurrence of secretive rails. Ecol Appl 20(7):2036–2046. https://doi.org/10.1890/09-0470.1
Ripple WJ, Estes JA, Beschta RL, Wilmers CC et al (2014) Status and ecological effects of the world’s largest carnivores. Science 343(6167):1241484. https://doi.org/10.1126/science.1241484
Ritchie EG, Johnson CN (2009) Predator interactions, mesopredator release and biodiversity conservation. Ecol Lett 12(9):982–998. https://doi.org/10.1111/j.1461-0248.2009.01347.x
Robley A, Moloney P, Moxham C, Neave G et al (2016) The effects of interaction between planned burning and foxes on the presence of native mammals. Arthur Rylah Institute of Environmental Research. Department of Environment, Land, Water and Planning. Heidelburg, Victoria
Rosenberg KV, Dokter AM, Blancher PJ, Sauer JR et al (2019) Decline of the North American avifauna. Science 366(6461):120–124. https://doi.org/10.1126/science.aaw1313
Sánchez-Bayo F, Wyckhuys KAG (2019) Worldwide decline of the entomofauna: a review of its drivers. Biol Cons 232:8–27
Short J, Smith A (1994) Mammal decline and recovery in Australia. J Mammal 75(2):288–297
Shrader AM, Brown JS, Kerley GIH, Kotler BP (2008) Do free-ranging domestic goats show ‘landscapes of fear’? Patch use in response to habitat features and predator cues. J Arid Environ 72(10):1811–1819. https://doi.org/10.1016/j.jaridenv.2008.05.004
Signorell N, Wirthner S, Patthey P, Schranz R (2010) Concealment from predators drives foraging habitat selection in brood-rearing Alpine black grouse Tetrao tetrix hens: habitat management implications. Wildlife Biol 16(3):249–257
Skerratt LF, Berger L, Speare R, Cashins S, McDonald KR, Phillott AD, Kenyon, N (2007) Spread of chytridiomycosis has caused the rapid global decline and extinction of frogs. EcoHealth 4(2):125. https://doi.org/10.1007/s10393-007-0093-5
Stokeld D, Frank A, Hill B, Choy J, Mahney T, Stevens A, Young S, Djelk Rangers, Warddeken Rangers, Gillespie G (2015) Multiple cameras required to reliably detect feral cats in northern Australian tropical savanna: an evaluation of sampling design when using camera traps. Wildlife 42(8):642–649
Stokeld D, Frank ASK, Hill B, Choy JL et al (2016) Multiple cameras required to reliably detect feral cats in northern Australian tropical savanna: an evaluation of sampling design when using camera traps. Wildl Res 42(8):642–649. https://doi.org/10.1071/WR15083
Stuart SN, Chanson JS, Cox NA, Young BE et al (2004) Status and trends of amphibian declines and extinctions worldwide. Science 306(5702):1783–1786. https://doi.org/10.1126/science.1103538
Szabo JK, Khwaja N, Garnett ST, Butchart SHM (2012) Global patterns and drivers of avian extinctions at the species and subspecies level. PLoS ONE 7(10):e47080
Thackway R, Cresswell ID (1995) An interim biogeographic regionalisation for Australia: a framework for establishing the national system of reserves. Australian Nature Conservation Agency, Canberra
van Eeden LM, Eklund A, Miller JRB, López-Bao JV et al (2018) Carnivore conservation needs evidence-based livestock protection. PLoS Biol 16(9):e2005577. https://doi.org/10.1371/journal.pbio.2005577
Ward M, Tulloch AIT, Radford JQ, Williams BA, Reside AE, Macdonald SL et al (2020) Impact of 2019–2020 mega-fires on Australian fauna habitat. Nat Ecol Evol. https://doi.org/10.1038/s41559-020-1251-1
Woinarski JCZ, Burbidge AA, Harrison PL (2015) unravelling of a continental fauna: decline and extinction of Australian mammals since European settlement. Proc Natl Acad Sci 112(15):4531–4540
Woinarski JCZ, Murphy BP, Legge SM, Garnett ST et al (2017) How many birds are killed by cats in Australia? Biol Cons 214:76–87. https://doi.org/10.1016/j.biocon.2017.08.006
Woinarski JCZ, Braby MF, Burbidge AA, Coates D (2019) Reading the black book: the number, timing, distribution and causes of listed extinctions in Australia. Biol Conserv 239:108261
Wright SM (1993) Observations of the behaviour of male eastern grey kangaroos when attacked by dingoes. Wildl Res 20(6):845–949
Acknowledgements
We thank Natalie Izquierdo, Alison Foster and the NSW Department of Planning, Industry and Environment for facilitating access to the Wildcount data. We thank the NSW National Parks and Wildlife Service Wildcount team of hundreds that developed and implemented the Wildcount fauna monitoring program.
Funding
Open Access funding enabled and organized by CAUL and its Member Institutions. No Funding was provided for the researchers to conduct this study.
Author information
Authors and Affiliations
Contributions
DM collated data, analysed data and wrote the manuscript, RG assisted in analysing the data and assisted in writing the manuscript, ML conceived the project and gave editorial advice.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest to disclose.
Ethical approval
The authors obtained publicly available data from the NSW government for analysis and therefore ethics approval was not required.
Consent to participate
This study did not involve human subjects therefore consent to participate was not required.
Consent to publish
The Authors give Biodiversity and Conservation authority to publish this manuscript.
Additional information
Communicated by Karen E. Hodges.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
McHugh, D., Goldingay, R.L. & Letnic, M. Occupancy and co-occurrence patterns of endemic mammals and introduced predators across a broad geographical gradient in eastern Australia. Biodivers Conserv 31, 989–1021 (2022). https://doi.org/10.1007/s10531-022-02374-0
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10531-022-02374-0