Conventional wisdom on roosting behavior of Australian flying‐foxes—A critical review, and evaluation using new data

Abstract Fruit bats (Family: Pteropodidae) are animals of great ecological and economic importance, yet their populations are threatened by ongoing habitat loss and human persecution. A lack of ecological knowledge for the vast majority of Pteropodid species presents additional challenges for their conservation and management. In Australia, populations of flying‐fox species (Genus: Pteropus) are declining and management approaches are highly contentious. Australian flying‐fox roosts are exposed to management regimes involving habitat modification, through human–wildlife conflict management policies, or vegetation restoration programs. Details on the fine‐scale roosting ecology of flying‐foxes are not sufficiently known to provide evidence‐based guidance for these regimes, and the impact on flying‐foxes of these habitat modifications is poorly understood. We seek to identify and test commonly held understandings about the roosting ecology of Australian flying‐foxes to inform practical recommendations and guide and refine management practices at flying‐fox roosts. We identify 31 statements relevant to understanding of flying‐fox roosting structure and synthesize these in the context of existing literature. We then contribute a contemporary, fine‐scale dataset on within‐roost structure to further evaluate 11 of these statements. The new dataset encompasses 13‐monthly repeat measures from 2,522 spatially referenced roost trees across eight sites in southeastern Queensland and northeastern New South Wales. We show evidence of sympatry and indirect competition between species, including spatial segregation of black and grey‐headed flying‐foxes within roosts and seasonal displacement of both species by little red flying‐foxes. We demonstrate roost‐specific annual trends in occupancy and abundance and provide updated demographic information including the spatial and temporal distributions of males and females within roosts. Insights from our systematic and quantitative study will be important to guide evidence‐based recommendations on restoration and management and will be crucial for the implementation of priority recovery actions for the preservation of these species in the future.


| INTRODUC TI ON
Fruit bats (Family: Pteropodidae) are animals of extraordinary ecological and economic importance (Fujita & Tuttle, 1991). As longdistance seed dispersers and pollinators, fruit bats play a crucial role in the maintenance and regeneration of forest ecosystems (Hodgkison et al., 2003;Oleksy et al., 2015;Shilton et al., 1999). Moreover, fruit bats are responsible for the propagation of at least 289 plant species across their distribution, 186 of which have economic value, making fruit bats important contributors to the sustainability of human livelihoods (Fujita & Tuttle, 1991). Despite their importance, many fruit bat species are in severe decline. Half are listed as near threatened to extinct according to the IUCN (88 of the 177 species with sufficient data) (IUCN, 2020), with human persecution and habitat loss identified as two of the largest threats imposed on these species (Acharya et al., 2011;Andrianaivoarivelo et al., 2011;IUCN, 2020;Jenkins et al., 2007). While measures have been taken in some countries to reverse this trend-including increased legislative protection (Aziz et al., 2016;Eby & Lunney, 2002a;Thiriet, 2010) and community awareness campaigns (Anthony et al., 2018;Carroll & Feistner, 1996;Trewhella et al., 2005)-conservation and management efforts for the majority of these species remain hindered by an enduring absence of ecological knowledge (Fujita & Tuttle, 1991;Mickleburgh et al., 2002) and ongoing conflict with humans (Aziz et al., 2016;Currey et al., 2018).
These same conservation challenges persist for Australian flying-foxes (Genus: Pteropus) despite improved levels of protection.
Indiscriminate and widespread persecution and killing of flying-foxes were persistent until the ~1990s (Fujita & Tuttle, 1991;Hall, 2002;Ratcliffe, 1931 Government of Victoria, 1988;New South Wales Government, 2016;Queensland Government, 1992). However, loss and degradation of roosting habitat continues to pose a substantial threat, and management of these species must additionally balance conservation outcomes with negative public perception and human-wildlife conflict (e.g., BBC News Australia, 2017; Kohut, 2017;Welle, 2021).
A major challenge for these species is that policies for conservation and conflict management are often in direct contrast. The identification, management, and protection of roosting habitat are listed as priority recovery actions for the vulnerable grey-headed flying-fox (Pteropus poliocephalus) and endangered spectacled flyingfox (Pteropus conspicillatus) (Commonwealth of Australia, 2017a). Yet in direct contrast, roost management policies and guidelines that aim to reduce human-wildlife conflict often promote removal of roost trees to create perimeter buffers between the roost and private properties, which can exceed 50 meters in some cases (State of NSW & Office of Environment & Heritage, 2018). In more extreme cases, flying-fox roost management permits can be granted to disturb, drive away, or destroy flying-fox roosts entirely (Lenson, . Management challenges in Australia are being further compounded by an emerging and accelerating trend of urbanization of flying-fox roost sites, and fragmentation of roost populations (Meade et al., 2021;Tait et al., 2014;Williams et al., 2006). Roost structures are transitioning from large roosts that are seasonally occupied by nomadic individuals into smaller, continuously occupied roosts in urban areas Eby et al. in review). This fragmentation, or fissioning, of roost populations has been attributed to environmental change -both land clearing of winter flowering native species in southeastern Australia (Eby et al., 1999) and the concurrent increase in availability of exotic winter food resources in urban areas (Parry-Jones & Augee, 2001;Williams et al., 2006). As a consequence, increasing numbers of roosts have formed near residential housing, particularly in metropolitan areas such as Sydney, the Gold Coast, and Brisbane, despite overall population declines (Tait et al., 2014). These urban roosts often develop into sites of ongoing conflict with neighbors (Commonwealth of Australia, 2017b), and there has been growing demand to reduce the impact of roosts on local communities through active management of flying-fox camps (Currey et al., 2018). Similar changes with fragmentation and urbanization have been observed elsewhere Peel et al., 2017), suggesting that this occurrence is likely representative of other systems across the range of Pteropodids.
A second major challenge for management of these species is that systematically informed, baseline ecological knowledge is limited, so the impact and effectiveness of efforts to contribute to either conservation (roost restoration) or conflict (roost modification) goals are unknown. Roosting requirements of these species are not well understood (Commonwealth of Australia, 2017a) beyond broadscale trends in roosting patterns (e.g., Tidemann et al., 1999;, migration (Eby, 1991;Eby et al., 1999;Meade et al., 2021), and studies on sociality and behavior (Klose et al., 2009;Nelson, 1965b;Welbergen, 2005). Detailed (fine-scale) spatiotemporal patterns in animal density and tree use remain unquantified crucial for the implementation of priority recovery actions for the preservation of these species in the future.

K E Y W O R D S
camp, conservation, fruit bat, habitat, management, Pteropodidae (Commonwealth of Australia, 2017a), and knowledge on historical usage patterns (e.g., Nelson, 1965b;Ratcliffe, 1931;Tidemann et al., 1999; may be inconsistent with current usage patterns. This lack of detailed information is of particular concern, as current conservation strategies that aim to identify, protect, and restore important roosting habitat, and practices for managing conflict, are necessarily founded on observations that may not fully reflect the habitat requirements of the animals. In this context, the number of flying-fox roosts exposed to programs of vegetation modification is increasing rapidly in Australia, yet the potential impact of modifications to roosting habitat on flying-foxes is largely unknown. More information is needed to provide baseline ecological data in this time of rapid ecological change, and to guide and support vegetation management practices and decision-making criteria to provide a realistic representation of the roosting habitat needs and preferences of flying-foxes. Systematic and comprehensive examination of multiple species in Australia may also help identify whether generalities exist among Pteropodids, and guide understanding in systems where more limited data and resources are available. In this paper, we seek to identify and evaluate commonly held understandings about the roosting ecology of Australian flyingfoxes, focusing on species on the Australian mainland. We first review "gray literature" (management, recovery, and restoration plans or reports published by state government and local groups) to identify commonly held understandings concerning flying-fox roosting structure. We then review the existing empirical literature, to critically evaluate the extent of empirical support for these statements and highlight gaps in empirical evidence. Lastly, we utilize high-resolution spatial mapping techniques and monthly field observations to systematically and quantitatively document spatial and intra-annual temporal patterns in flying-fox roost and tree use in southeast Queensland and northeast New South Wales. This approach allows us to highlight where quantitative information on flying-fox roosting has been missing, and where updated information may be required. Our new dataset is the first to capture fine-scale spatial and temporal dynamics of flying-fox roost use in a structured, repeatable design, and provides baseline information in a time of rapid ecological change. Such a systematic and quantitative study will be important for informing evidence-based recommendations to guide vegetation modification practices and improve roost management strategies for flying-fox conservation. This will be crucial for implementation of effective habitat restoration projects, to successfully balance the management of these threatened, contentious, and urbanizing wildlife, and to guide comparable approaches in other Pteropodid species across their range.

| ME THODS
Four species of flying-fox occur in Australia. These include the greyheaded flying-fox (P. poliocephalus), black flying-fox (Pteropus alecto), spectacled flying-fox (P. conspicillatus), and little red flying-fox (Pteropus scapulatus). Little red flying-foxes are the smallest of the species, ranging in adult weight from approximately 330-550g and with a wingspan of around 0.9 m (Bartholomew et al., 1964). Adult weight of the other species each range between 650 and 1,000 g, and wingspan is around 1.2 m (Eby & Lunney, 2002b). Reproduction is seasonal and synchronous, with each species showing a single birth pulse per year-typically October-November for grey-headed and black flying-foxes, and April-May for little red flying-foxes (Eby & Lunney, 2002b;McIlwee & Martin, 2002).
Flying-foxes are highly gregarious and occur in large communal aggregations known as "roosts" or "camps" (Ratcliffe, 1931). Flyingfoxes roost in the exposed branches of trees, and a single roost community can collectively number hundreds to hundreds of thousands of individuals (National Flying-Fox Monitoring Program, 2017).
Roosts are used as daytime rest stops by animals that forage in surrounding areas or as short-term stopover sites by migrating animals and function as maternity colonies in the breeding season (Eby & Palmer, 1991;Tidemann & Nelson, 2004

| Review of gray literature
Flying-fox management is generally undertaken in line with sitespecific roost management plans (e.g., Council of Ipswich, 2016;EcoLogical, 2014;Scenic Rim Regional Council, 2015), which are adopted by local government councils based on their state's flyingfox camp management policy (e.g., Queensland: SEQ Catchments (2012), State of Queensland Department of Environment and Science (2020); and New South Wales: State of NSW and Office of Environment and Heritage (2018)). We focused on statements made in state-level documents, as these are the primary resource for individual roost plans. We identified common statements/understandings across these documents, with particular emphasis on those that pertain to (1) routine vegetation management activities (weed removal and trimming understory vegetation); (2) creation of buffers (either by clearing/trimming canopy trees, or disturbing animals at the roost boundary); and (3) restoration interventions.

| Review of existing empirical support
We conducted a systematic literature search of peer-reviewed published literature using ISI's Web of Knowledge (27 July 2020).
Keywords were chosen to target studies evaluating the within-and between-roost structure of Australian flying-foxes (Table S1). This included any studies relevant to (1) the physical structure of roosts (e.g., area, tree structure, tree/roost selection), (2) the social structure of roosts (e.g., demographic and species structuring), (3) roosting behavior (e.g., territoriality and fidelity of individuals), (4) movement and migration relating to occupancy and abundance of roosts, and (5) roost microclimate. In addition to the literature search, reference lists and relevant studies already known to the authors were also screened to identify potentially relevant studies not captured by our initial search. We also included empirical support from key unpublished sources (e.g., dissertations).

| Empirical data collection
We collected data on roosting structure at eight sites in southeast Queensland and northeast New South Wales (Figure 1). These sites were chosen to represent a gradient of habitats utilized by flyingfoxes, ranging from metropolitan areas of Brisbane and the Gold Coast, to roosts in peri-urban and rural areas ( Figure 1, Table 1). All sites were previously documented as having a continuous population of grey-headed or black flying-foxes. Little red flying-foxes visited some roost sites intermittently; however, no roost sites occurred within the distribution of spectacled flying-foxes (National Flying-Fox Monitoring Program, 2017).
We mapped the spatial arrangement of all overstory, canopy, and midstory trees in a grid network of 10 stratified random subplots (20 × 20 meters each) per roost site. Subplots were stratified throughout perceived "core" (five subplots) and "peripheral" (five subplots) roosting areas, classed as areas observed to be frequently occupied (core) or infrequently (peripheral) by bats (Welbergen, 2005). Core and peripheral areas were evaluated from regular observations made prior to roost tree mapping, though note that these categories were revised subsequently with the quantitative data. Trees were mapped and tagged using tree survey methods described in the "Ausplots Forest Monitoring Network, Large Tree Survey Protocol" (Wood et al., 2015). To evaluate spatiotemporal patterns in roosting, we revisited all tagged trees and scored the extent of species occupancy using the following tree abundance index: 0 = zero bats; 1 = 1-5 bats; 2 = 6-10 bats; 3 = 11-20 bats; 4 = 21-50 bats; 5 = 51-100 bats, 6 = 101-200 bats, and 7 = >200 bats. For a subset of trees (N = 60 per site, consistent through time), absolute counts and minimum/maximum roosting heights of each species were taken. Overall roost perimeter (perimeter of area occupied) was mapped with GPS (accurate to 10 m) immediately after the tree survey to estimate perimeter length and roost area. Total abundance at each roost was also estimated with a census count of bats where feasible (i.e., where total abundance was predicted to be <5,000 individuals), or by counting bats as they emerged in the evening from their roosts ("fly-out"), as per recommendations in Westcott et al. (2011). If these counts could not be conducted, population counts from local councils (conducted within ~a week of the bat surveys) were used, as the total abundance of roosts is generally stable over short time frames (Nelson, 1965b). Because roost estimates become more unreliable with increasing total abundance, and because our estimation methods were intrinsically linked with total abundance, we converted the total estimated abundance into an index estimate (where bin ranges increase with total abundance) for use in analyses, as per values used by the National Flying-Fox Monitoring Program (2017). Index categories were as follows: 1 = 1-499 bats, 2 = 500-2,499 bats, 3 = 2,500-4,999 bats, 4 = 5,000-9,999 bats, 5 = 10,000-15,999 bats, 6 = 16,000-49,999 bats and 7 = >50,000 bats. Roosting surveys were repeated once a month for 13 months (August 2018-August 2019). More detailed methods of empirical data collection can be found in Appendix S1.

| Statistical analyses
The main statistical comparisons tested with our empirical data were as follows: (1) whether density of occupation is greater for subplots in "core" areas of the roost compared with subplots in irregularly occupied "peripheral" areas (defined by occupation greater than or less than 80% of surveys respectively (Appendix S1); (2) whether bat occupation decreases with distance from the roost center (per species); (3) whether bat species segregate in vertical space; and (4) whether dominant individuals occupy the center of roosts, and subdominant individuals the outer area (per species). We also provide qualitative comparisons of (5) seasonal patterns of abundance and occupancy per species; and (6) whether bat species segregate in horizontal space. Because we do not observe dominance directly, we assume that only dominant reproducing males share their territory with females and their young and use the proportion of males per tree as a proxy for dominance structure in the roost. This is consistent with behavioral studies of dominance and observations of "bachelor male trees" containing entirely nonreproducing males (e.g., Markus, 2002 andWelbergen, 2005).
We utilized generalized additive models for all statistical comparisons to allow for nonlinearity, with random effects modeled with smooth functions. Roost site and subplot were modeled using a standard random-effects smoothing function. Session was modeled using a cyclic cubic regression spline in cases where seasonality in the time series was evident (all comparisons except those involving the proportion of male black, male grey-headed, and combined male bats per tree); otherwise, session was modeled with a standard random-effects smoothing function. We accounted for nonindependence (nesting) of random effects by including an autoregressive model for errors in the model (Laurinec, 2017;Yang et al., 2012). For the comparisons involving evaluation of species, models were run F I G U R E 1 Map of roost sites included in the study. Gray shading indicates urban land cover of dense human habitation (as per Schneider et al., 2009), and gray circles are locations of flying-fox roosts. Circles show 45-km foraging radii from roost study sites (as per Giles et al., 2018). GIS land-cover data were downloaded from Natural Earth (2020)  separately for each species owing to differences in seasonality of occupation (and so differences in the fit of cyclic cubic regression splines). Error distribution for comparisons was specified according to data type and extent of zero-inflation (as per Crawley, 2013).
We fit the models and performed checks of standardized residuals in R (version 4.0.2), using the "mgcv" package (functions "gamm" and "gam.check") (as per Wood, 2017). See Appendix S1 for more detailed information on modeling decisions, a summary table of comparisons, and a breakdown of spatial and temporal replicates of measures. Summarized data are given in Appendix S2, and annotated R code is available on GitHub at: https://github.com/Tamik aLunn/ FF-roost -ecology.

| RE SULTS
From our review of management, recovery, and restoration documents published by state government, we highlighted 31 commonly held understandings relevant to flying-fox roosting structure (Table 2). From our systematic search for empirical literature, we generated a total of 79 search results. Of these, 52 were removed through screening (10 from outside the Australian mainland, 4 on non-Pteropus species, and 38 focused on topics other than roost structure). An additional 18 published studies and 4 honors/PhD theses were included from citations and the author's reference collections, giving 49 included studies in total (Appendix S1). Lastly, we generated an empirical dataset consisting of 13-monthly repeat measures from 2,522 trees across eight roost sites. Roost sites contained 118-474 measured and tagged trees each, with an average of 2 (sparsely structured) to 75 (densely structured) trees per 20x20 meter subplot. Tree roosting height and count were recorded for 9,056 trees out of 32,206 repeat measures. (Note that our total repeat measures were less than 32,786 owing to cases of tree removal through the duration of the survey.) We report model outputs of main interest in the main text, but see Appendix S3 for full model output.
Below, and in Table 2, we synthesize how commonly held understandings compare with existing literature and new data from our study.

| Use of area
"Some areas of permanent camps are more consistently occupied ('core areas') than others" This understanding was widely reported in previous studies, with none contradicting it (Table 2). Consistent with these other studies (e.g., Nelson, 1965b;Welbergen, 2005), we observed some areas of roosts to be more consistently occupied than others ( Figure 2).
"'Core areas' are more densely occupied than 'peripheral areas'" TA B L E 1 Information on roost sites included in the study Flying-foxes adjust the location of "core areas" through time SEQ Catchments (2012) Hall (2002), Pallin (2000) Welbergen (2005) Areas outside of the "core area" are used by more transient animals SEQ Catchments (2012) Welbergen ( Individuals at the periphery of groups act as "guards" State of Queensland Department of Environment and Science (2020) Nelson (1965b); Klose et al. (2009) (2003) The roosting positions of individual males are highly consistent, and animals return to the same branch of a tree over many weeks or months SEQ Catchments (2012) Welbergen (2005) (2002), Williams et al. (2006), Parry-Jones and Augee (2001), Parry-Jones and Augee (1992) Habitat preferences Stager and Hall (1983) The habitat patch must be at least 1ha in size but be large enough to accommodate and sustain large numbers of flying-foxes. For a small roost (10,000 bats), the area needed is approximately 3ha, and for a large roost (50,000), the area needed is 10ha SEQ Catchments (2012), State of NSW and Department of Planning Industry and Environment (2019) as per State of NSW and Office of Environment and Heritage (2018), EcoLogical (2014) Pallin (2000), Roberts (2005) Flying-foxes prefer complex vegetation structure (upper, mid-, and understory layers) SEQ Catchments (2012), State of NSW and Department of Planning Industry and Environment (2019) as per State of NSW and Office of Environment and Heritage (2018) Pallin (2000) (citing report by Buchanan) Flying-foxes prefer dense vegetation SEQ Catchments (2012) Roberts (2005) Flying-foxes prefer a dense understory SEQ Catchments (2012) Roberts (2005) Flying-foxes prefer a closed canopy at least 3−5 m high SEQ Catchments (2012) (2005) Welbergen (2005) The structure of roost-wide vegetation is more important than the characteristics of individual roost trees (e.g., species, canopy cover) SEQ Catchments (2012) Palmer and Woinarski (1999), Pallin (2000), Vardon et al. (2001), Tidemann et al. (1999), Vardon and, Hall and Richards (2000), Roberts (2005) Flying-foxes prefer level topography (<5° incline) SEQ Catchments (2012) (2020) Welbergen (2005), Richards (2002), Pallin (2000), McWilliam (1984), Hall (2002) Some tree species are more resilient to damage by flying-fox roosting than others SEQ Catchments (2012) In small remnant patches, the process of opening the canopy (from tree damage by roosting) will increase the impact of invasive weeds SEQ Catchments (2012), State of Queensland Department of Environment and Science (2020) Pallin (2000), McWilliam (1984), Hall (2002) Where sufficient roosting space is available, flying-foxes shift their roosting areas, which lessens their damage to vegetation over time SEQ Catchments (2012), EcoLogical Pallin (2000), Hall (2002) Note: Note that additional evidence from our 13-month empirical study only addresses questions that require less than one year of data (i.e., intra-annual patterns in roost structure). Statements not addressed with our empirical data are colored gray. An extended version of this table with details on study results is provided in Appendix S3.

TA B L E 2 (Continued)
Existing empirical data broadly supported this statement (Table 2).
In our study, peripheral areas (those occupied less than 80% of the time) generally were less densely occupied than core areas, though density varied substantially across roost site, subplot, and session (all contributed substantially as random effects). Here, lower density refers to both a lower number of bats per subplot in peripheral subplots (−0.581 ± 0.177, p = .001, Figure 3), and a lower proportion of occupied trees (−0.222 ± 0.078, p = .005, Appendix S3). Within subplots, we also note that some trees were more consistently used than others, including trees that were occupied in 100% of surveys where bats were present at the roost (Appendix S1). The number of bats per tree in irregularly occupied trees (occupied <80% of the time) was typically lower than in regularly occupied trees (−0.606 ± 0.034, p < .001).
We observed negative relationships between bat occupation metrics and distance from the roost center, including the number of bats per occupied subplot (−1.639 ± 0.016, p < .001, Figure 4) and proportion of occupied trees per subplot (−0.315 ± 0.034, p < .001, Appendix S3). This decline with distance from the center of subplot was largely driven by little red flying-foxes (Figure 4). Roost site, subplot, and session also all contributed substantially as random effects (Appendix S3).

"Roost area fluctuates with total abundance"
Studies have previously reported changes to total roosting area, but none to date have formally quantified the relationship between area and total bat abundance (Table 2). From our data, we observed substantial fluctuations in total roost area within some roost sites across monthly surveys, and overall, a positive relationship with total bat F I G U R E 2 Occupancy of subplots across survey period, for surveys when at least one bat was present. (a) shows the total number of bats per subplot, and (b) shows the proportion of surveys the subplot was occupied. Facets/color indicates separate roost sites. "(C)" indicates roosts that have features of contemporary roost types (see Table 1). Note that construction works at the "Avondale" roost during the survey caused the bats to shift their roosting location, such that only one subplot was utilized thereafter abundance. The extent of variation was variable across roosts, however ( Figure 5). We note that relationships between total bat abundance and area were likely masked in many roosts by the large span of population values in some index categories (e.g., index six spans 16,000-49,999 bats). It is probable that data of finer resolution may have detected this relationship more strongly for roosts in this size range, but are not available in this dataset.

| Spatial segregation of species
Results from our new dataset included systematic recording of the three species that occur in southeast Queensland-P. alecto, P. poliocephalus, and P. scapulatus. The majority of observations were made of P. alecto and P. poliocephalus, which occupy this region continuously through the year. P. scapulatus was found irregularly at some roosts, which is consistent with the seasonal migration patterns of this species (Nelson, 1965a). Fine-scale spatial overlap between species was evaluated during surveys when multiple species were present (N = 73, 70.2% of surveys). Black and grey-headed flying-foxes co-occurred in 65 surveys (62.5%), black and little red flying-foxes co-occurred in 17 surveys (16.3%), and grey-headed and little red flying-foxes co-occurred in nine surveys (8.7%). We observed roostdependent support for spatial segregation of species.
"Species share roosts sites, but segregate spatially within" Observations from previous studies commonly report co-occupation of roosts by multiple species, with anecdotal observations of inconsistent overlap or separation within and between trees (Table 2). We observed some horizontal spatial segregation of species, with species showing preference for discrete areas in roosts. In the "Lismore"  (Figure 6a). Co-occupation of individual trees by two different species was also relatively low-across surveys where two species were present, 4.6%-7.9% of occupied trees were co-occupied by two species, versus 92.1%-95.4% that were occupied by only one species (Figure 6b). Only six trees were ever observed to occupy all three species at once.

"Large influxes of species into roosts (especially little red flyingfoxes) can displace other species"
Only one previous study had reported displacement by species, reporting an anecdotal observation of black and grey-headed F I G U R E 3 Occupancy of subplots in "core" and "peripheral" areas, shown by average total number of bats per occupied subplot across the survey period. Data are filtered to show numbers of bats when subplots were occupied (i.e., unoccupied subplots are removed). "Core" subplots were identified as those occupied in at least 80% of surveys (when bats present at the roost), and "peripheral" subplots as those occupied less than 80% of the time. (a) Shows areas split by roost site (facet and color), and (b) shows all roosts combined. Area displayed in subplot has been cropped to remove extreme outliers. "(C)" indicates roosts that have features of contemporary roost types (see Table 1) flying-foxes being displaced by little red flying-foxes (Table 2). Our quantitative data document changing distribution of regular species occupants in response to ("invading") irregular species occupants, supporting this prior observation. Little red flying-foxes, in particular, were observed to displace black and grey-headed flying-foxes from their usual roosting locations (most notably at the "Redcliffe" roost: Appendix S4). Black and grey-headed flying-foxes tended to co-occur in roosts without too much impact on each other (Appendix S4).

"Species roost at different heights"
Previously, only one study had formally documented differences in roosting height between species (Table 2). This included a record of black flying-foxes and grey-headed flying-foxes only and did not provide measures of absolute height (rather, roosting in different quadrants of trees) (Welbergen, 2005 (Figure 7). Note, however, that topographical variation (change in ground height, e.g., from creeks and small crests) within roosts was not taken into consideration in measures of height. Differences in heights presented here reflect a relative difference in roosting heights from F I G U R E 4 Distance from roost center and occupancy of bats, shown by the average total number of bats per occupied subplot during the survey period. Data are filtered to show numbers of bats when trees are occupied (i.e., unoccupied subplots are removed). Roost center is calculated for each survey as the centroid of the roost area at the time of the survey. Distance from the center is calculated as the mean distance of trees in each subplot from this centroid, scaled by the maximum observed distance value per session. (a) shows values per species (line type) split by roost (facets), and (b) shows species and roost combined. Trend line is by loess fit (local polynomial regression fit) with standard error bands (gray shading). "(C)" indicates roosts that have features of contemporary roost types (see Table 1) the ground within trees, but may not reflect true, realized height relative to the canopy.

| Demographic/social structure "The majority of roost trees are occupied by mixed groups of adults, with territories comprised of a single male and one or more females and their dependent young"
We commonly observed roost trees to be occupied by mixed groups of sexes, with a single tree occupied by one or more males, and one or more females and their dependent young. This is inconsistent with general knowledge based on historical studies such as Nelson (1965a) and Nelson (1965b), but consistent with more contemporary observations (Table 2). We also observed cases where trees were occupied by entirely male individuals (consistent with reports of "bachelor male" trees in Markus (2002)). We would note here that a single tree may contain multiple male territories (Connell, 2003;Markus, 2002), and the survey methods did not allow inference on the composition of individual territories, only individual trees. The proportion of males per tree appeared to follow seasonal patterns that were mostly consistent between black and grey-headed flyingfoxes within roosts (Appendix S3). Some roosts ("Toowoomba," "Avondale," "Lismore") showed an increase in the proportion of males per tree after parturition in September/October, while other roosts ("Sunnybank," "Canungra") decreased immediately after this time. We also did not observe complete segregation of sexes at any time of the year, in contrast to Nelson (1965b) who noted complete segregation between September until early December, and March to April.

F I G U R E 5
Relationship between total roost abundance (x-axis) and total roost area (y-axis) for each roost site. (a) shows relationship split by roost (facets), and (b) shows relationship with roosts combined. Trend line is by loess fit (local polynomial regression fit) with standard error bands (gray shading). Note that trend lines could not be fitted for all sites and are omitted. "(C)" indicates roosts that have features of contemporary roost types (see Table 1 and peri-urban (>10 km from urban edge; see Table 1)  The small effect sizes observed would suggest that there is no clear spatial structure to reproductive groupings or dominance groupings, however. This can be seen also in maps showing male composition F I G U R E 7 Difference in roosting height per species, over time. Fill shows average roosting height range per species (minimum height to maximum height). Fill boundaries (minimum and maximum curves) are by loess fit (local polynomial regression fit). (a) shows relationship split by roost (facets), and (b) shows relationship with roosts combined. In (a), dashed line represents the average canopy height per site; for roost sites where species occupy distinctly different areas ("Clunes" and 'Lismore"), canopy height is split by areas the species predominantly occupy. "(C)" indicates roosts that have features of contemporary roost types (see Table 1). Note that height data are taken from the tree subset only (up to N = 60 per roost site) and that trend lines could not be fitted for all site-by-species combinations and are omitted by tree relative to the roost perimeter, given in Appendix S5. Prior studies have reported inconsistent spatial patterns in flying-fox occupation (Table 2).

| Roost abundance/occupancy
"Individual roosts have distinguishable seasonal patterns of abundance and occupation," "Intra-and interannual variations in abundance can be extreme," and "Roost abundance peaks in

F I G U R E 8
Proportion of male bats per occupied tree versus distance of tree from the roost center, scaled by the maximum distance value observed per session. (a) shows values per species (row facet) split by roost (column facet); (b) shows combined species value pooled by roost. Trend line is by loess fit (local polynomial regression fit) with standard error bands (gray shading). "BFF" refers to black flying-fox, "GHFF" greyheaded flying-fox, and "LRFF" little red flying-fox. "(C)" indicates roosts that have features of contemporary roost types (see Table 1) F I G U R E 9 Different scales of bat abundance measures through time. (a) shows total roost abundance; (b) shows total roost area; (c) shows the proportion of occupied trees per subplot; and (d) shows the total abundance of occupied subplots. Total roost abundance is measured by an index score of abundance: 1 = 1-499 bats; 2 = 500-2,499 bats; 3 = 2,500-4,999 bats; 4 = 5,000-9,999 bats; 5 = 10,000-15,999 bats; 6 = 16,000-49,999 bats; and 7 = >50,000 bats

| D ISCUSS I ON
The success of efforts to conserve Pteropodid bats across their distribution relies on effective population and habitat management.
Pivotal to this is a baseline understanding of species ecology and behavior, which is currently lacking for the majority of these species (Fujita & Tuttle, 1991;Mickleburgh et al., 2002). Here, we provide a synthesis of all existing literature, as well as an unprecedented empirical dataset, to meet that need for Australian species of Pteropus.
We highlight that many existing beliefs on which conservation and management decisions are based are unsupported or outdated, and suggest that management plans should be updated to incorporate improved knowledge. Most importantly, we highlight that a onesize-fits-all approach to roost management will be inappropriate, given the extent of variation between sites even within a regional area. Roost management guidelines need to be changed to promote a more tailored approach that requires preliminary data acquisition before management plans are formulated and approved.

| Existing understanding not supported "Individual roosts have distinguishable seasonal patterns of abundance and occupation"
All roost sites in our empirical dataset were occupied continuously throughout the year by both sexes. This type of roost occupation has been noted from 1981 onward (Parry-Jones, 1985;Puddicombe, 1981) and has become common in recent decades (e.g., Aston, 1987;Eby, 1991;Larsen et al., 2002;Van der Ree et al., 2006).
This pattern of occupation contrasts to the "summer" and "winter" pattern described historically by Nelson (1965a) andNelson (1965b) and cited in the Flying-fox Roost Management Guideline for Queensland, where "summer roosts" of reproducing individuals would form between ~September/October and April/May, and "winter roosts" of dispersed animals would form between April/May and September (Nelson, 1965a(Nelson, , 1965bParry-Jones & Augee, 1991, 1992Ratcliffe, 1931). For these roost types, overwintering animals at summer roosts were rare and, when present, were documented as being predominantly juveniles or lone adult males (Nelson, 1965b). Specifically, territory formation (from January) and conception (from ~March) (P. poliocephalus and P. alecto) (Welbergen, 2005) coupled with abundant flowering of native flora in these months (Nelson, 1965a) were understood to drive and support aggregative living in summer/ autumn, while decreased food availability and the cessation of mating from ~May triggered the animals to disperse and adopt a lessgregarious living style in winter (Parry-Jones & Augee, 1992). This ecology has changed in more recent decades, where continuous availability of exotic foods in urban areas has reduced the need for migratory behaviors and allows aggregate groups to remain year-  (2020)) are of concern, as recommendations based on historical usage patterns may be inconsistent with current usage patterns, particularly in urban areas where occupation patterns have changed the most (Larsen et al., 2002;Tait et al., 2014), and where human-bat conflict is often the highest (Kung et al., 2015). Roost monitoring prior to management actions should encompass every season, and not assume that bats will disperse in winter. Similarly, contemporary overwintering roosts commonly contain individuals from all age and sex groups and may be consistently utilized through time (Larsen et al., 2002;Tait et al., 2014). These historic perspectives also describe the complete separation of males and females between September until early December (the period immediately before parturition, during lactation, and before conception) and again post-March (after conception) (Nelson, 1965a(Nelson, , 1965b. During these times, animals were historically noted to segregate by tree or height, such that all social contacts were between individuals of the same sex. However, these observations contrast with more recent observations of flying-fox social groupings (Eby et al., 1999;McWilliam, 1984;Puddicombe, 1981;Welbergen, 2005), and observations from this study. In contemporary roosts, mixedsex groups are commonly present all year around, such that males and females co-occur in the roost and within trees year around.
In addition, more recent observations, and results from this study, suggest that there is no clear spatial structure in the distribution of the sexes within the roost. This contrasts with the common perspective that dominant reproducing individuals-particularly reproducing females-occupy the center of roosts, and nondominant individuals-including weaned juveniles-occupy the edges of roosts (Table 2). Puddicombe (1981) notes that reproductive groups (mixed groups of males, females, and their young) were uniformly distributed through the camp and present in peripheral areas (McWilliam, 1984).
Additionally, in this study we observed randomly distributed groups of mixed males and females, and groups of all-male trees. While we did not systematically record age of bats, as estimating age from observations at a distance is not always possible, on few occasions we did observe what looked to be all-juvenile trees. These observations of all-juvenile trees were not at the edges of the roosts. This difference in sex and age structure of roosts potentially reflects the change in occupancy patterns in flying-fox roosts, where aggregative living was historically believed to be driven by strong social drivers (i.e., mating), whereas aggregative living in contemporary roosts is thought to be driven by continuous resource availability in the urban environment (Parry-Jones & Augee, 2001;Williams et al., 2006 A recent tracking study demonstrated a sex-based difference in visitation to major urban areas, with males but not females suggested to roost more often in major urban areas (Meade et al., 2021).

| Existing understanding supported, but conditional on roost site and local conditions "Roost abundance peaks in March" and "Intra-and interannual variations in abundance can be extreme"
March was identified in some management documents as being the time for peak abundance in flying-fox roosts (e.g., State of Queensland Department of Environment & Science, 2020). However, studies on P. poliocephalus and P. alecto identify a typical pattern of increasing abundance from September-October (when females give birth) until a peak in January-February (when the season's young are able to fly independently) (Eby, 1991;Eby & Palmer, 1991;Nelson, 1965b;Parry-Jones & Augee, 2001). Roost sizes then decrease during March-April (the period of mating) to low winter counts in continuously occupied/overwintering roosts, or zero winter counts in seasonally occupied summer roosts (Eby, 1991;Eby & Palmer, 1991;Nelson, 1965b). These studies note that cyclical patterns of occupation are driven by reproductive factors (i.e., timing of birth and independent flight), but highlight that irregular, local dynamics of food availability can superimpose variability into these patterns of abundance (Parry-Jones & Augee, 1992). Indeed, many studies note high intra-and interannual variability in abundance. Parry-Jones and Augee (2001), for example, note that animals from their study roost appeared to migrate away and decrease in abundance in response to a blossoming event, presumably to move to a roost in closer proximity to the blossoming.
In our study, some roosts showed patterns consistent with a total roost abundance peak toward March (e.g., "Redcliffe," "Canungra," and "Clunes"). Others showed either no considerable fluctuation in abundance ("Burleigh') or peaks at other times ("Toowoomba," "Sunnybank," "Avondale," "Lismore"). Drivers of peaks were variable between roosts. For the "Redcliffe" roost, seasonal migration of little red flying-foxes from ~January 2019 contributed to a peak in abundance around March (see species abundance plots in Appendix S3).
For the "Lismore" roost, a blossoming event in winter 2018 triggered an influx of nomadic bats into the population, driving the peak observed in August 2018. Dynamics observed in other roosts were likely the result of local dynamics of food availability.
We note also that estimates of abundance from our study were much smaller than those of historical estimates. Ratcliffe (1931) describes "small" roosts as ~5,000-10,000 animals, "medium" as 10,000-50,000, and "large" as anything over this size. Ratcliffe (1931) also report roosts in northern Queensland with bats "into the millions" (Red River) and "exceeded a quarter of a million, possibly considerably" (Burnett River). Likewise, Lunney and Moon (1997)

| Existing understanding supported
"Some areas of permanent camps are more consistently occupied ('core areas') than others," "'Core areas' are more densely occupied than 'peripheral areas'," and "Roost area fluctuates with total abundance" Variability in the usage and occupation of areas within roosts has been highlighted in management documents (e.g., SEQ Catchments, 2012). This includes more persistent usage of "core" areas, higher occupation of "core" areas, and variability in the roost perimeter (reflecting expansion and contraction from the core area).
All existing literature (to our knowledge) and the new data from our study support these understandings. We would note, however, the distinction between a "core/peripheral" roost area and a "central/ edge" roost area. We defined the core area based on consistency of occupation, not spatial location. Areas identified to be "core" were not necessarily in the center of the roost (see location of roost centroid relative to the roost perimeter and surveyed subplots, in Appendix S4). This distinction has not necessarily been made in literature and management plans to date but has important implications for the interpretation of "core" roosting areas, and management recommendations specific for "core/central" or "peripheral/edge" areas.
For example, it cannot be assumed that buffer creation via vegetation removal from the roost edge will not affect a "core" area of bat roosting, and so will not have a substantial impact on flying-foxes.
Management activities should be prescribed for specific zones in roosts, based on prior monitoring of the roost, and recognizing the ecological importance of different areas (Ku-ring-gai Council, 2018;Pallin, 2000). In addition, prior monitoring of core/peripheral roosting areas will be important to inform the location and potential effectiveness of buffer creation. Given the potential for roost area to fluctuate with abundance, creation of buffers via vegetation removal may reduce the area of normal roost habitat available, and result in an expansion into new areas when flying-fox numbers increase (as noted in Currey et al., 2018). The prescription of buffers should be planned with care to avoid unintended outcomes during periods of high population abundance.
"Species share roost sites, but segregate spatially within," "Large influxes of species into roosts (especially little red flying-foxes) can displace other species," and "Species roost at different heights" The range of black flying-foxes underwent a phase of rapid southern expansion in the late 1990s and early 2000s, increasing the area of overlap with grey-headed flying-foxes (Roberts et al., 2012a). As the two species co-occupy roosts where their distributions overlap, this process has substantially increased the number of roosts occupied by both species, and thereby increased the pertinence of understanding the structure of mixed-species roosts. There has been relatively little formal documentation of species overlap and segregation within roosts. Ratcliffe (1932) noted that sections of roosts were occupied by different species-specifically, that little red flying-foxes and black flying-foxes occupied different areas.
Some horizontal separation has also been noted by Nelson (1965b) and Klose et al. (2009), and notes of displacement by little red flyingfox have been described in Birt and Markus (1999). We contribute quantitative, spatial information on the extent and overlap of little red flying-fox, black flying-fox and grey-headed flying-fox roosting, extending on the predominantly anecdotal observations underlying management documents to date. Findings from our data support common understandings of flying-fox roost structure: species commonly showed preferences for discrete areas of roosts, and even more commonly, preference for occupation of separate trees. We also observed segregation of species by roosting height, with black flying-fox foxes showing the highest roosting, followed by greyheaded flying-foxes and little red flying-foxes. These findings flag the importance of species monitoring of roost sites prior to management interventions. It cannot be assumed, for example, that species occupy areas of the roost uniformly, and management actions need to consider areas that may be more or less important to vulnerable species, such as the grey-headed flying-fox. These results also give interesting insights into understanding disease transmission dynamics within roosts, relating to the extent of mixing of primary host species (e.g., black flying-foxes for Hendra virus) and other species presumed to be incidental hosts (e.g., grey-headed and little red flying-foxes).   -ring-gai Council, 2018), and seek to maintain refuges in the mid-and lower storys at all times. Special care not to disturb bats should be taken in identified core areas of the roost.

| CON CLUS ION
This study takes a thorough, multifaceted approach to better understand the ecology of flying-fox roost use and structure in Australia. We build upon broadscale knowledge of historic roosting occupancy and abundance patterns, and provide updated baseline information on roosting structure in urban and peri-urban roosts by providing fine-scale spatial, and temporal data on roost and tree use. Specifically, we demonstrate high variation in patterns of occupancy and abundance between roosts sites, and provide updated demographic information including the spatial and temporal distributions of males and females within roosts. We also show evidence of sympatry and indirect competition between species, including spatial segregation of black and grey-headed flying-foxes within roosts, and seasonal displacement of both species by little red flying-foxes. The outcomes of this research will be of immediate, practical benefit to management and conservation of flying-fox roosts in Australia, and meet research needs specifically identified in the draft Recovery Plan for the vulnerable grey-headed flyingfox. The level of spatial and temporal detail provided in our empirical study will be important in designing management plans that are sensitive to flying-fox habitat needs, and in identifying and protecting important habitat areas within roosts that are reflective of current movements and preferences. Most importantly, we highlight that a one-size-fits-all approach to roost management will be inappropriate, given the extent of variation between sites even within a regional area. Fine-scale information on roost tree preferences will also improve understanding of the potential impacts of existing conflict management strategies involving vegetation removal, including buffer creation, and can guide vegetation removal efforts to heed these habitat requirements. This information is timely and much needed in advance of the recently announced Environmental Trust grants program for flying-fox habitat restoration, and in the face of continued and increasing urbanization of flying-foxes in Australia.