Habitat in ﬂ uence on activity patterns of brush-tailed rock-wallabies in Queensland

. The brush-tailed rock-wallaby ( Petrogale penicillata ) is a nocturnal refuge dependent herbivore that can be found from south-east Queensland to East Gippsland in Victoria. Brush-tailed rock-wallaby colonies are situated in many different habitat types, and refuge within this habitat is important for their survival. Dingoes coexist with P. penicillata and are perceived to be a threat. This study aimed to determine the link between habitat and P. penicillata behaviour and their interaction with predators. Three P. penicillata colonies were studied within the Little Liverpool Range, Queensland. Images of P. penicillata and dingoes from camera traps were analysed for daily activity patterns and activity overlap between species. Dingo scats were collected and analysed to determine predation. The results indicated that different habitat types change the perceived predation risk and corresponding activity patterns of P. penicillata . Predation risk impacted foraging behaviour, with rock-wallabies in exposed habitat types exhibiting greater predator avoidance behaviours than those in more protected habitat types. The results indicate that brush-tailed rock-wallabies modi ﬁ ed their activity patterns according to different habitat types to reduce their risk of predation.


Introduction
Herbivores need to modify their behaviour if the risk of predation becomes a serious threat to their survival. This is particularly true for native Australian species threatened by introduced predators. The brush-tailed rock-wallaby (Petrogale penicillata), listed as a vulnerable species (EPBC Act Listing and Nature Conservation (Animals) Regulation 2020 (Queensland)), is a refuge dependent species known to display nocturnal behaviour of foraging during the night and sheltering amongst refuge during the day (Carter and Goldizen 2003;Tuft et al. 2011). Studies have shown that P. penicillata alters its foraging behaviour in response to predation risk (Tuft et al. 2011), and this could extend to other behaviours or even times of the day when P. penicillata are most active. This species has historically been widely distributed (Short 1982;Wong and Soderquist 2003) but now inhabits only rocky outcrops, rainforest gorges and mountain ridges throughout south-east Queensland, eastern New South Wales and East Gippsland in Victoria (Short 1982;Eldridge and Close 2008;Egerton and Lochman 2009; Department of the Environment 2018; Piggott et al. 2018).
The red fox (Vulpes vulpes) and dingoes or wild dogs (Canis lupus ssp.) are a major threat to P. penicillata populations throughout Australia (Menkhorst and Hynes 2011; Office of Environment and Heritage 2013; Taggart et al. 2015). Recent studies into the differences between wild dogs and dingoes have found that the majority of wild dogs contain a high proportion of 'pure' dingo DNA and occupy the same ecological niche in terms of diet and mesopredator supression (Letnic et al. 2012;Jackson et al. 2017). For the purposes of this study all dingoes/wild dogs will be referred to as dingoes.
Most studies that have examined predation risk responses have focussed on changes to P. penicillata foraging behaviour in conjunction with the importance of landscape safety features that restrict foraging distance (Tuft et al. 2011). This study aimed to close the knowledge gap in seeking a link between habitat values and predation risk and their effect on P. penicillata activity times. This change in activity is supported through evidence of other animal species altering their behaviour due to predation in different habitat types (Corp et al. 1997;Ellwanger and Gould 2011;Jaman and Huffman 2013;De Oliveira et al. 2014). Within Queensland, P. penicillata populations have been documented in the Little Liverpool, Teviot, Main and McPherson Ranges (Queensland Government 2014). These sites all have similar habitat features, comprising many short ledges, crevices, escape routes and caves (Short 1982;Murray et al. 2008). All of these Queensland sites also have populations of dingoes that have the potential to impact on P. penicillata (Department of Agriculture and Fisheries, The State of Queensland 2017).
Three distinct P. penicillata populations located within different habitat types were studied to determine their activity patterns and activity overlap with that of dingoes at each location, to see if predator avoidance behaviour was evident in P. penicillata in differing habitats. Predation rates by introduced predators on P. penicillata were examined through the collection and analysis of dingo and fox scats. The hypothesis tested in this study was that within different habitat types P. penicillata populations have different activity patterns that are dependent on predation risk.

Study area
The study area was located on Aroona Station, a Queensland Trust for Nature property within the Little Liverpool Range, 41 km south-west of Ipswich, Queensland (27 37.380 0 S, 152 45.600 0 E). The study area comprised 1999 ha on both undulating and steep terrain. Vegetation was a mix of eucalypt woodland and dry rainforest along rocky slopes and plains of native grasses utilised by cattle. Steep cliffs were present throughout the property along the ridgelines. A 2016 pilot study confirmed that P. penicillata, dingoes, and foxes all utilised the property. The study was conducted in two periods: between March and September 2016, and between April and June 2017. Three different habitat types were selected on the property to determine if habitat type influenced activity patterns of resident P. penicillata, dingoes and foxes in those habitats. Site A consisted of a rainforest gully comprising RE (Regional Ecosystem) type 12.8.9, and Sites B and C were on ridges that had exposed cliff faces dominated by Eucalyptus crebra open woodland in mosaics of RE types 12.8. 16, 12.8.17, and 12.8.9 (Table 1) (Jeffers 2016). Each of these Regional Ecosystem types varied in vegetation structure as well as floristics (Table 1). The habitat in Site A was considerably different from that of Sites B and C. Sites B and C consisted of the same mosaic of Regional Ecosystem types 12.8.9, 12.8.16, and 12.8.17 (Table 1). However, the most noticeable habitat element that was different from Site A was that Sites B and C both encompassed open cliff faces and exposed rock shelves (Fig. 1). Therefore P. penicillata in Sites B and C were more exposed to predators and the elements than those in Site A.

Camera trapping
A pilot study was conducted in 2016 in the three study sites (A, B and C). A total of 10 camera traps, comprising eight Reconyx Hyperfire HC600 covert cameras (Reconyx USA) and two UWay VH200HD cameras (Uway Canada), were deployed to monitor P. penicillata, dingoes and other species. In 2017 the number of camera traps deployed was increased to 27, comprising 21 Reconyx Hyperfire HC600 covert cameras and three Reconyx Ultrafire covert (Reconyx USA) and three UWay VH200HD cameras (Uway Canada). Nine cameras were used at each site with similar settings, e.g. per trigger five images were taken with no time delay between images. Distinguishing features, including size of the animal, scars and direction of travel, were used to identify individuals where possible. Cameras were set up 20 m apart around the perimeter of the P. penicillata colonies, focusing along the top and bottom of ridges, gullies, cliffs and gorges (Gowen and Vernes 2014). Camera traps were strapped to trees between 50 and 75 cm above the ground and aimed towards rocks or well used animal tracks where P. penicillata scats had been recorded and activity was expected to be high (Rowcliffe et al. 2008;Gowen and Vernes 2014). Although capture of P. penicllata behaviour was the primary focus the cameras also recorded occurrences of when dingoes and foxes visited these sites.

Scat collection and analysis
Scats from dingoes and foxes were collected opportunistically throughout the study area in 2017. Scats were collected along roads, animal tracks, walking tracks, and around the P. penicillata colony sites within the boundaries of Aroona Station in accordance with ethics approval conditions, following the collection methodology of Lunney et al. (1996). Due to the difficult terrain of the study sites, search methods were adapted to suit the conditions. Adopting the adaptive cluster sampling technique, additional radii searches were conducted around the location of a discovered scat, where the unit of interest was a scat and replacing transects was the opportunistic collection method (Thompson 1991;Keiter et al. 2016). Scats were sent to 'Scats About' for analysis of prey species and their percentage composition within the scat, and confirmation of the scat's source (i.e. dingo or fox) (Triggs 2004).

Statistical analyses
The camera trap data were used to determine the activity overlap between P. penicillata, dingoes and foxes in the three different habitat types. The images from the camera traps were sorted and instances of P. penicillata, dingoes or foxes were recorded. Each picture was analysed to ensure each new individual was recorded as a separate event, along with information about its date, and time and time decimal. Converting image capture times to a time decimal changed the time to a percentage of the overall 24-h period, to standardise for presentation in graphs.
The percentage of activity overlap for P. penicillata, dingoes and foxes was examined using R (The R Foundation 2018). The R studio package 'Estimate of Coefficient of Overlapping for Animal Patterns' was utilised to measure the distance between high densities in the data to measure the percentage overlap (Ridout and Linkie 2009). The overlap coefficient was based on the equation (Weitzman 1970): where f (x) and g(x) are two probability density functions for the two species. The information from the time decimal records was used to form kernel density plots, from which non-parametric estimators were used to label the coefficient of overlap between the different populations and the separate P. penicillata colonies against dingoes and foxes (Meredith and Ridout 2017). The Dhat1 non-parametric estimator was used to explain the percentage of overlap (Ridout and Linkie 2009). The prey species and their percentage composition within the scats, from the dingo and fox scats, were used to compile presence/absence results of different prey species. Animal events per camera were calculated for all species to standardise the different number of camera trap days for each year. Instances for dingoes per camera were lower in 2017 than 2016 across all sites, with average events per camera ranging from 3 to 9.6 in 2016, and 0.78 to 1 in 2017 (Table 2).

Camera trap data
In contrast, instances of P. penicillata events per camera were higher in 2017 than 2016 across all sites, ranging from 6.8 to 61 events per camera in 2016, and 22.9 to 42.1 events per camera in 2017 (Table 2). Activity patterns produced from images at each site showed that P. penicillata had consistent activity patterns (i.e. higher (peaks) or lower (troughs) throughout the different times of the day) in both 2016 and 2017 specific to the site (Figs 2a, 3a and 4a).
At Site A P. penicillata was mostly active at dawn and dusk, which was different to the peak in activity during nighttime hours of P. penicillata at Sites B and C (Figs 2a, 3a and  4a). Dingo activity patterns were most similar for all sites at Site A in 2016 and 2017 (Fig. 2b). Activity patterns of dingoes at Sites B and C had less overlap in 2016 and 2017 (Figs 3b, 4b).
The spike in activity observed at Site C at 0800 hours can be attributed to the low numbers skewing the results and has been included to present data from all sites for both years.

Dingo and fox scat composition
Eighty dingo scats and 18 fox scats were collected on Aroona Station. No P. penicillata remains were found within these scats (Fig. 5). Eastern grey kangaroo (Macropus giganteus) formed a large part of dingo diet, along with northern brown bandicoot (Isoodon macrourus) and swamp wallaby (Wallabia bicolor) (Fig. 5). In contrast, based on the scats, northern brown bandicoot (Isoodon macrourus) was the major source of prey for foxes (Fig. 5).
The scats were collected adjacent to the camera locations. The results from Aroona Station were compared with those from three similar studies that examined dingo or fox predation on P. penicillata. The other studies were of P. penicillata colonies in Armidale, NSW (Lunney et al. 1996), Upper-Macleay and Aspley River Gorges, NSW (Robertshaw and Harden 1985) and the Grampians, Victoria (Lobert 1988) (Table 3). Lunney et al. (1996) and Robertshaw and Harden (1985) examined over 300 predator scats, but both studies showed that less than 3.1% of the scats contained P. penicillata (Table 3).

Discussion
The results from this study support the hypothesis that brushtailed rock-wallabies modify their activity patterns according to different habitat types to reduce their risk of predation. Petrogale penicillata, as a refuge dependent herbivore, is known to alter its foraging behaviour (specifically distance ventured away from refuge) in response to predation risk and the presence of specific landscape habitat features.
At Sites B and C P. penicillata had similar activity times, demonstrating that its activity in these two sites was consistent with current knowledge, with activity times highest between 1800 and 0600 hours for 2016 and confirmed in 2017 with more data (Figs 3a, 4a). However, along with its difference in habitat, P. penicillata at Site A was active mostly during the crepuscular hours of dawn and dusk: the results show two distinct peaks from 1500 to 1900 hours and from 0500 to 0930 hours (Fig. 2).
From the difference in habitat features at Site A, compared with the other two sites, it can be hypothesised that Site A had less protection for P. penicillata, increasing its predation risk, and thus its behaviour was different from that of P. penicillata at Sites B and C. Changes in animal behaviour in response to Dingo daily activity   et al. (1996) 310 1 dog; 0 fox 3.1 Robertshaw and Harden (1985) 358 3-5 depredated A 1-2 A Study did not differentiate between dog and fox predation.
habitat and perceived predation risk have been documented in multiple studies (e.g. Ellwanger and Gould 2011;Tuft et al. 2011;De Oliveira et al. 2014). Foraging is a behaviour that can be indirectly affected by predation pressure; therefore this change in P. penicillata activity times follows this same concept (Tuft et al. 2011).
The assumption that the habitat features were an influencing factor of P. penicillata activity times is strengthened by the fact that the peaks in dingo activity at dawn and dusk (Allen et al. 2013(Allen et al. , 2014 mostly overlapped with the peak activity times of P. penicillata at site A. In contrast, activity patterns of P. penicillata at Sites B and C showed that they avoided periods when the dingoes were most active (Figs 3, 4). The activity patterns from Sites B and C show that P. penicillata activity was highest during the night, whereas dingoes at those sites were generally crepuscular. Predation risk ultimately changes the foraging behaviour of P. penicillata as it is a foraging compromise (Gutman and Dayan 2005;Hammerschlag 2009). Thus two species that coexist in the same area can use temporal partitioning of activity to avoid each other (Gutman and Dayan 2005). At Sites B and C P. penicillata seems to actively avoid periods when dingo activity increases (Figs 3,4) to lower the predation risk, and possibly also due to differences in site structure.
Temporal partitioning may also be reflected in the differing numbers of dingoes and P. penicillata captured on camera traps in 2016 and 2017 (Table 2). In 2016 there were fewer P. penicillata, on average, per camera trap, and greater numbers of dingoes (Table 2). This reversed in 2017, with an increase in P. penicillata numbers recorded per camera trap, and fewer dingoes. Petrogale penicillata could actively be avoiding predators when predators were more abundant. Temporal partitioning was particularly obvious at Sites B and C, where the higher level of exposure of both sites seemed to impact on P. penicillata activity levels when dingoes were present (Figs 3, 4).
Foxes were present within the landscape; however, there were not enough camera trap images to determine their activity and how it compared with that of P. penicillata. The lack of foxes in the same areas where dingoes were caught on camera could suggest that the foxes may be actively avoiding the areas frequented by the dingoes (Letnic et al. 2012;Jackson et al. 2017).
The results of the scat analysis showed no evidence of predation on P. penicillata by dingoes during the study. These findings are similar to those of three studies elsewhere in Australia where rock-wallabies were frequently sighted (Lunney et al. 1996) -Upper-Macleay and Aspley River Gorges (Robertshaw and Harden 1985) and the Grampians (Lobert 1988) (Table 3)with all studies showing low predation rates. Although fewer scats were collected at Aroona Station for analysis compared with the Victorian and NSW studies (Aroona Station: 80 dingo scats and 18 fox scats), the results were similar. In each study, P. penicillata did not form a large part of dingo or fox diets (Table 3). However, the activity of the prey species that were actually present in the dingo scats was not analysed. Thus the overlap in activity of P. penicillata and dingoes may be coincidental, and they may be responding to a landscape of fear.
A study on dingo movement showed that daily average travel distance was 16.7 km (Harden 1985). It is possible that the above studies, including this one at Aroona Station, are potentially missing predator scats that contain P. penicillata, as scats could be deposited outside the search areas from each of the study sites. However, due to the observed regular movements of the dingoes along roads and tracks throughout Aroona Station, it is likely that the scats collected during the study represented the dingo diet well, indicating low levels of predation on P. penicillata. This is further supported by the study of Lunney et al. (1996), which collected scats up to a distance of 60 km from a P. penicillata colony, with very few scats found to contain P. penicillata remains within that range.
Although this study did reveal some interesting behavioural patterns of P. penicillata and its response to predators in different habitat types, several limitations need to be acknowledged. First, it would be valuable to replicate the study over a different period, as seasonality could play an important role in P. penicillata activity and behaviour. The scat collection and analysis covered a period when young P. penicillata joeys were beginning to leave their mothers' pouches. An increase in more vulnerable P. penicillata within the landscape may potentially change predation levels on this species. It is possible that predation levels could change during the dispersal season in spring and summer when young are weaned (Eldridge and Close 2008) and this should be an area of future research. Additionally, increasing the number of predator scats collected would have strengthened the results. Although the study site was intensively surveyed, it would have been preferable to expand the study area to encompass a wider landscape and collect scats that might have been deposited outside of the immediate colony range.

Conclusions
Understanding interactions between introduced predators and P. penicillata is crucial for its conservation. Changes in behaviour through the use of predator avoidance strategies and temporal partitioning, and the lack of remains in dingo scats, indicate that P. penicillata has modified its behaviour in response to predators to decrease predation pressure. The results of this study show that the complexity of its habitat and exposure of colony sites can alter the perceived risk of predators within the landscape, which also influences P. penicillata behaviour and the level of avoidance of the times of the day that predators are most active.

Data availability statement
The data that support this study will be shared upon reasonable request to the corresponding author.