Megafires attract avian scavenging but carcasses still persist

The effects of fires on vertebrate scavengers have not been characterized despite the importance of scavenging in shaping food web dynamics. We assessed whether the 2019/2020 megafires in Australia shifted the species richness, carcass detection and feeding times of vertebrate scavengers, and whether the fire affected carcasses persistence times.


| INTRODUC TI ON
Extreme disturbance events can influence the patterns and processes that shape and drive ecosystem dynamics (Turner, 2010).
They can lead to major shifts in resource availability for animal populations, in turn altering their movements and foraging behaviours (Doherty et al., 2021). Such effects have been well documented following fires (Nimmo et al., 2019). Fire causes direct animal mortality, but it also incinerates vegetation, altering plant communities and the availability of food and shelter for biota (Haslem et al., 2011;Smit et al., 2010). This can affect the distribution and abundance of animals, as, for example, they starve, move to find more optimal habitat or are left more vulnerable to predation. On the other hand, some animals may benefit from fires, changing their movement patterns to take advantage of fleeing prey and newly cleared landscapes (Nimmo et al., 2019). Indeed, avian predators including rock kestrels (Falco tinnunculus) and jackal buzzards (Buteo rufofuscus) have been observed selectively hovering over burnt patches in search of food (Barnard, 1987). Terrestrial vertebrate predators such as red foxes (Vulpes vulpes) and feral cats (Felis catus) are also attracted to recently burnt areas (Geary et al., 2020;McGregor et al., 2016). But in addition to making prey easier to catch, fires could also provide foraging opportunities for vertebrate scavengers if fire related deaths cause animal carcasses to be abundant.
The effects of fires on vertebrate scavenging dynamics are not well known. This is despite knowledge that scavengers can perform an important ecosystem service by removing dead animal matter from the landscape (Barton et al., 2013;Benbow et al., 2019;Cunningham et al., 2018). In post-fire environments, this may be critical to reduce the spread of diseases or pathogens that can proliferate on or around carcasses. For example, Clostridium botulinum proliferates in carcasses after death and produces botulinum toxin, which is lethal to vertebrates that ingest it (Espelund & Klaveness, 2014). Although obligate scavengers such as vultures are most well known for their ability to rapidly remove carcasses from the landscape, there is increasing awareness that facultative scavengers can rapidly consume and regularly take advantage of animal carcasses (Wirsing & Newsome, 2021). This reflects the conclusion that more energy is transferred within terrestrial food webs as a result of scavenging compared to predation (Wilson & Wolkovich, 2011) and that the role of carrion in nutrient cycling and community dynamics is disproportionately large compared with plant detritus (Barton et al., 2013;Benbow et al., 2019).
The role of carrion in shaping terrestrial food web dynamics is known to vary depending on its quantity and dispersion within an ecosystem , habitat context (Pardo- Barquin et al., 2019), season (Turner et al., 2017), climate , anthropogenic impacts (Sebastián-González et al., 2019) and scavenger community composition (Sebastián-González et al., 2016). However, scavenging dynamics could shift in unexpected ways following fires. For example, large apex scavengers may be impacted by fires if they cannot escape the burnt areas, in turn diminishing their influence on smaller mesoscavengers (O'Bryan et al., 2019) or carcass persistence times (Hill et al., 2018). Habitat is also known to influence scavenging dynamics, with some scavengers preferring open habitats compared to more closed canopy habitats (Roen & Yahner, 2005). For example, monitoring of scavengers on carcass piles in south-central Spain identified that griffon vultures (Gyps fulvus) scavenged more in open habitats, while the red fox and wild boar (Sus scrofa) scavenged more in vegetation-covered habitats (Carrasco-Garcia et al., 2018). Fires could therefore influence carcass detection and feeding times if there is a shift in vegetation structure, with potentially more favourable conditions for avian scavengers if the fire-affected areas become more exposed.
Examining the effects of fires on scavenging dynamics ideally requires monitoring of scavenger populations and carcass persistence times before and after a fire. Such studies are difficult to plan in the absence of management burn scenarios. However, as part of a scavenging study on the edge of the Greater Blue Mountains National Park, in New South Wales south-eastern Australia, we were uniquely placed to study the impacts of a widespread fire event. Specifically, we were monitoring vertebrate scavengers using experimentally placed kangaroo carcasses and remote wildlife cameras, before a large-scale wildfire went through the study site, thus providing an opportunity to replicate the monitoring after the fire. The fire itself was part of the 2019/2020 Australia black summer megafire which burnt approximately 18.6 m hectares of land, of which 11.5 m hectares was in forest and woodland areas. The fire was estimated to have killed or displaced 143 million mammals, 2.46 billion reptiles, 180 million birds and 51 million frogs (van Eeden et al., 2020). The Greater Blue Mountains National Park (and surrounding areas) was one of the worst affected areas, with around 80% of the National Park burnt (Merson, 2020). Consequently, it is likely that overall vertebrate scavenger densities declined post-fire, especially ground dwelling species. Carcass densities may have also increased because of fire related deaths or starvation of animals following the fires.
Understanding how this massive disturbance event influenced scavenger populations is critical in a world that is now experiencing increases in the intensity and frequency of natural disasters (Bradshaw et al., 2021;Ripple et al., 2020) that can result in mass animal mortality events and thus an influx of carcasses into the ecosystem (Fey et al., 2015).
Although Australia has no obligate scavengers, it does host a diverse suite of facultative scavengers (Cunningham et al., 2018;Forsyth et al., 2014). In our study site, the largest native/long established vertebrate scavengers include the dingo (Canis dingo) and wedge tailed eagle (Aquila audax). Both species have the potential to rapidly consume carcass biomass including bones and could feasibly exclude smaller mesoscavengers from accessing carcasses through fear effects (Cunningham et al., 2018;O'Bryan et al., 2019). Native meso-scavengers in the study site include ravens (Corvus spp.) and magpies (Gymnorhina tibicen), as well as goannas (Varanus spp.). The study site also has introduced facultative scavengers, namely the red fox and feral cat. The relative use of carcasses by the red fox and feral cat are not well documented in Australia, but red foxes fed on 60% of deer carcasses monitored in the state of Victoria in Australia, and their scavenging patterns were consistent with the idea that red foxes avoid dingoes (Forsyth et al., 2014). Scavenging by feral cats in the same study was low (10% of carcasses) (Forsyth et al., 2014), but feral cats increased their scavenging rates in the absence of Tasmanian devils (Sarcophilus harrisii) on the island state of Tasmania (Cunningham et al., 2018). Whether red foxes and feral cats increase or decrease their scavenging rates after fires has implications for understanding if carcasses support their populations, and thus potentially exacerbate their impacts on native species (Newsome et al., 2015).
In this study, we compared vertebrate scavenger species richness, carcass detection, feeding times and carcass persistence times before and after a fire. The data were derived from monitoring vertebrate scavengers on experimentally placed eastern grey kangaroo (Macropus giganteus) carcasses, which were on average 30 kg in mass. Because of the potential effects of habitat on scavenging dynamics (Roen & Yahner, 2005), we monitored carcasses in both open and closed canopy habitats before and after the fire. We tested the following predictions: (1) species composition of vertebrates utilizing carcasses will differ as a function of the fire, with decreased scavenger species richness anticipated following the fire event (due to the possible decline in animal numbers; van Eeden et al., 2020); (2) carcass discovery time by vertebrates will differ as a function of fire and habitat type, with carcasses in pre-fire environments (where animal numbers are expected to be higher; van Eeden et al., 2020) and in open habitats (where carcasses are more conspicuous ;Roen & Yahner, 2005) having shorter detection times; (3) vertebrate scavenging activity will decrease post-fire for mammals and reptiles (due to their possible decline post-fire; van Eeden et al., 2020) but less so for birds (who are mobile and can typically escape fires; Woinarski & Recher, 1997); and (4) carcass persistence (time to carcass removal) will increase post-fire (due to an overall decline in carcass visitations by scavengers, especially large mammalian scavengers such as dingoes; . Our results provide the first assessment of how vertebrate scavenger dynamics and carcass persistence times change in relation to a fire event. The fire reached the study site in December 2019. It was connected to a larger mega-fire in the Blue Mountains world heritage area (i.e. the Gosper's Mountain fire, which burnt more than 510,000 ha in total; Boer et al., 2020). The fire burned approximately 90% of the entire study site (which comprised woodland and grassland habitat) over a period of one and a half months (between the 16th of November and the 22nd of December, although spot fires were present before and after these dates). There was virtually no unburnt ground cover left in the woodland or grassland habitats after the fires, although subsequent rains approximately one month (in February 2020) after the fire did result in the grassland habitats re-growing quickly (See Figure 1b

| Carcass monitoring
Carcass monitoring was conducted over three 1-month periods in the warm seasons of summer and autumn. The first two monitoring periods were conducted before the fire, in January 2018 and 2019, and the third was conducted in March of 2020, which was the first possible opportunity (due to study site access restrictions) to conduct follow-up research after the fire. Temperatures are similar between January and March (Bureau of Meteorology, 2021), and no major shifts in scavenger activity are likely. Comparisons between the two periods before the fire and one period after fire allowed for an assessment of whether the fire drove larger differences in scavenging dynamics than those observed between years before the fire. This approach reflects the fact that scavenging rates and species interactions around carcasses are influenced by anthropogenic impacts (Sebastián-González et al., 2019) and/or scavenger community composition (Sebastián-González et al., 2016), which can vary from year-to-year irrespective of background environmental conditions. Indeed, there was active poison baiting and shooting programmes targeting dingoes in the study site before the study began. Thus, our approach of including data from two monitoring periods before the fire allows us to capture and assess any variability in scavenging dynamics before the fire. This is arguably more robust than only including one monitoring period before the fire or lumping the two monitoring periods before the fire together.
During each monitoring period, we distributed 20 adult eastern grey kangaroo carcasses in an equal mix of open grassland (hereafter "open canopy"; n = 10 sites) and forest (hereafter "closed canopy"; n = 10 sites) habitat ( Figure 1). Open canopy sites were at least 50 m from any stands of trees. The closed canopy sites had more than 20% canopy cover before the fire, and in the post-fire period carcasses were placed in forested locations where there was previously canopy cover of more than 20%. Estimated tree canopy cover within the closed canopy sites differed across periods, with the greatest variation occurring between the post-fire period (mean ± SE; 31.1 ± 4.6% coverage) and the pre-fire periods (first pre-fire period: 56.4 ± 3.1% coverage, second pre-fire period: 48.8 ± 2.9% coverage). Within monitoring periods, carcasses were separated by at least 1 km to mitigate scent travel between carcasses. Between monitoring periods, however, carcasses were positioned at least 100 m from any previous carcass placement. Carcasses from previous placements were not expected to influence animal behaviour, as they all were removed by scavengers within weeks of their placement and carcasses are generally common across this landscape. We used dead kangaroos sourced from nearby management culls, so no animals were killed for the purpose of this study. Any carcass displaying evidence of disease (e.g. heavy parasite loads) was not used. All carcasses were placed into the field without freezing within 24 hr of collection (i.e. all at the same time, with 24 hr being the longest time between the first and last carcass placement; January 2018; 9:00-23:00, January 2019; 8:00-13:00, and March 2020; 23:00-6:00). Scientific licences/permits were obtained to relocate the kangaroo carcasses (SL 101901), and research was approved by the University of Sydney Animal Ethics Committee (Project number: 2017/1173).
To allow for ongoing monitoring and detection of scavengers visiting and feeding on each carcass, we used a Reconyx PC800 Hyperfire™ camera trap (Professional Reconyx Inc., Holmen, WI, USA) attached to a free-standing star picket 3-4 m away from each carcass.
The cameras were programmed to take continuous photographs when triggered by thermal movement around the carcass (rapidfire, no wait period). To prevent complete removal of the carcasses away from the remote camera monitoring frame, we secured carcasses to the ground by wire attaching the neck and achilles tendon of the animal to two metal stakes spaced ~0.6 m apart. Cameras were used to monitor carcasses for 30 days to capture the main period of vertebrate scavenging activity and because the majority of carcass biomass (including meat, skin and bones) were removed from the environment in this time.

| Data collection
Camera images were tagged according to each new carcass visitation event by a different species, the number of individuals of a species present, whether any of the individual species fed on the carcass or not, and the date and time that the observation was recorded. A visitation event was considered new if it occurred ≥10 min from the previous visitation event by the same species. We defined scavenger species as species that were observed feeding on at least one carcass over the three study periods. All other animals captured on the remote cameras were excluded from our analyses. Finally, using a combination of in-person visual inspection of the carcasses and inspection of camera images, we determined the number of days until complete carcass consumption. A carcass was defined as completely consumed when less than 5% meat biomass and only skin, hair and/ or bone remained.

| Statistical analysis
We analysed differences in species richness and composition, detection time, feeding time and carcass persistence across study periods and habitats. We conducted analyses on all vertebrate scavengers grouped together and then separately on all avian, mammal and reptile scavenger groups when assessing carcass discovery and feeding times. We conducted our analyses in R Version 4.0.2 (R Development Core Team, 2020).
We assessed scavenger species composition by (1) Therneau, 2021). Survival analyses work well with censored data (Hosmer et al., 2008). Carcass discovery data were right-censored because some carcasses were not discovered by species groups by the end of monitoring periods. We ran three separate analyses investigating how long carcass discovery took for all scavenging vertebrates and then separately for the three scavenger groups (avian, mammalian and reptile). For all analyses, period (first pre-fire period, second pre-fire period and post-fire period) and habitat (open, closed) were used as the predictor variables. Again, we interpreted competitive models (ΔAICc < 2), which we selected from a suite of models fitted with all possible parameter and the interaction: habitat × period. Post hoc analyses were used to investigate differences in the total feeding times across the three different periods, and the interaction term habitat × period (if it was included in the model), and were calculated using Holm-Bonferroni log-rank tests for post hoc analysis (Package survminer; Kassambara et al., 2021). For all models, we tested the proportional hazards assumption by visualizing the survival curves and by testing the non-zero slope for the Schoenfeld residuals versus time (Therneau & Grambsch, 2000

| RE SULTS
We monitored 60 kangaroo carcasses (including 40 kangaroo carcasses before and 20 carcasses after the fire event), resulting in a total of 1,800 observation days and more than 689,000 photo images taken between January 2018 and April 2020.

| Scavenging activity
For the mammalian, avian and reptilian scavengers, two models  Post hoc tests also showed greater feeding between the post-fire period (394.7 ± 95.4 min) compared to the first (102.6 ± 32.6 min; Holm-Bonferroni test: z = 3.51, p = .001) and to the second
There were two models of carcass persistence time that were competitive (ΔAICc < 2) (Appendix S5). The first model included period test; p = .470), but they did show that carcasses persisted longer in the post-fire period compared with the second pre-fire period (p < .001), and in the first pre-fire period compared to second prefire period (p < .001) (Appendix S5; Figure 5).

| D ISCUSS I ON
We present the first study examining the effects of wildfire on vertebrate scavenging dynamics and carcass persistence times. Our results reveal substantial variation in how fire influences different scavenger groups and indicate that avian scavenging was most strongly affected by the fire event. Our findings also highlight the potential importance of other factors such as habitat in influencing scavenger dynamics. Against Prediction 1, we did not find an overall decrease in scavenger species richness following the fire. Similarly, against Prediction 2, overall scavenger detection times did not decline from pre-to the post-fire period. Instead, we found that avian scavengers detected carcasses at a faster rate following the fire event. In support of Prediction 2, we did, however, find that scavengers detected carcasses faster in open compared to closed canopy habitats, although it was only birds and not mammals or reptiles that showed faster carcass detection in the open habitats. Overall, scavengers increased their feeding times in the post-fire period, against Prediction 3. However, this trend was likely driven by avian scavengers, as mammals and reptiles did not change their feeding times across the three study periods. Finally, in partial support of Prediction 4, we found that carcasses persisted longer in the postfire period, but only when compared to the second pre-fire period.
We interpret these findings and explore potential implications of the study below.

| Scavenging in a post-fire landscape
The 2019/2020 fires in eastern Australia were estimated to have killed or displaced billions of animals (van Eeden et al., 2020). At our study site, we therefore expected to see an overall decline in scavenger species richness, carcass detection rates and scavenging activity post-fire, as animal species were supressed or completely removed due to the impact of the fire event. Yet, despite approximately 90% of woodland and grassland habitat in our study site being burnt, our results indicated that the fire did not influence overall scavenger species richness and had little effect on carcass detection rates and It is also possible that some scavenging animals increased in number following the fires. Indeed, we found that fire had a positive effect on avian scavengers who detected carcasses faster and increased their scavenging activity during the post-fire period. Corvids and large predatory birds including eagles and raptors are known to travel to fire-affected areas to hunt fleeing prey and to take advantage of burnt landscapes that provide reduced cover for their prey (Nimmo et al., 2019). This could translate to localized increases in avian populations and higher scavenging activities post-fire. This may explain why avian scavengers, and particularly Australian ravens and wedge-tailed eagles (see Appendix S1), scavenged more in the post-fire period during our study. Some scavenging mammals, including red foxes and dingoes, may also respond positively to fire (Geary et al., 2020), but these species are probably slower to disperse when compared to large birds with the capacity of soaring flight. This might at least partially explain why bird but not mammal scavenging activities increased in our study following the fire.
Animal scavenging in post-fire periods may have also been influenced by decreased opportunities for preferred food sources, such as live prey. In Australia, there are no obligate scavengers, and most carnivores that feed on carcasses function also as predators and therefore feel it is unlikely that this influenced our results.
Fires might also influence scavenging dynamics by burning and reducing vegetation cover, which could increase the conspicuousness of carcasses and, in turn, how easily they are detected by scav-

| Species interactions and carcass persistence
Scavenger dynamics are influenced by a range of abiotic and biotic factors . Indeed, at a global scale, patterns in vertebrate scavenging have been shown to differ according to season, latitude, rainfall, habitat, presence of apex scavengers/predators and human impact (Sebastián-González et al., 2019;Turner et al., 2017). It is likely that some of the variation we observed across study periods was caused by factors other than fire. For example, red fox occurrence on carcasses was highest in the first pre-fire period (95%), lowest in the second pre-fire period (40%), and still low in the post-fire period where red foxes only visited 50% of carcasses.
Scavenging activity by red foxes was also variable, and in the second pre-fire period red foxes did not feed on any carcasses, while in the post-fire period they scavenged for an average of only ~8 min compared to their average feed time of ~42 min during the first prefire period. While it is possible that red foxes were supressed by the fire, a recent global review of predator responses to fires indicated F I G U R E 5 Persistence (days) of kangaroo carcasses monitored in (a) open (yellow) and closed (green) canopy habitat, and (b) the first prefire period during January 2018 (blue), the second pre-fire period during January 2019 (purple) and the post-fire during March 2020 (black) in the Wolgan Valley. Light shading shows 95% confidence intervals that this predator species mostly responded positively to fire (Geary et al., 2020) and fire suppression does not provide an adequate explanation for their low scavenging activity in the second pre-fire period. Instead, red fox scavenging in our study site may have been influenced by competition with larger co-occurring scavenging species, specifically the dingo. Dingoes were detected (100%) and fed (90%) on most carcasses during the second pre-fire period, when red fox scavenging was at its lowest. Dingoes were also relatively active in the post-fire period (detected and fed on 85% of carcasses), when red fox scavenging was again low, and showed the lowest scavenging activity in the first pre-fire period (dingoes detected 85% of carcasses but only fed on 65%), when red foxes were most active. The ability of larger scavengers to outcompete and exclude smaller scavengers from carcasses has been shown in multiple studies (Prugh & Silvy, 2020) and is incorporated into theory (O'Bryan et al., 2019), and our results are consistent with the idea that red foxes avoid dingoes at carcasses (Forsyth et al., 2014).
The high occurrence of dingoes visiting and feeding on carcasses in the second pre-fire period potentially also explains why there were higher persistence times for carcasses in the first pre-fire period and to a lesser extent in the post-fire period. The ability of larger scavengers to rapidly consume carcasses is well recognized, but most research has focused on the scavenging efficiency of obligate scavengers such as vultures (Hill et al., 2018). Concurrent research in the study area and other parts of Australia has identified that dingoes can act as efficient scavengers and decrease carcass persistence times under some circumstances .
Low scavenging activity by dingoes in the first pre-fire period may have been due to active poison baiting and shooting programmes that took place before the study began. Our carcass persistence results might therefore reflect both the cascading effects of removing dingoes, but also that the fire itself had little impact on dingo scavenging. However, more detailed analyses would help to determine the exact relationship between dingoes, red foxes and carcass persistence in our study area.
A key factor that we did not examine was the responses to fire by insect scavengers. Insects can accelerate decomposition, and flies, beetles and ants can double the rate of biomass loss from small vertebrate carcasses in only a few days (Barton & Evans, 2017). Future studies assessing the effects of fire on scavenging and carcass persistence should therefore assess patterns in insect scavenging. Our study was also only conducted in the warmer seasons when insect activity would have been at its height and therefore did not consider the effects of fire on scavenging activity or carcass persistence times during cool periods. This is important to note because in cool periods burnt landscapes may influence carcass removal differently, because vertebrates typically play a greater role in carcass removal during these times due to reduced competition with insects and microbes (DeVault et al., 2003). In addition, our study did not assess whether carcass state (i.e. as burnt or unburnt) influences scavenger dynamics or carcass persistence. Previous studies have shown that insect scavenging activity and decomposition rates are increased on burnt carcasses (McIntosh et al., 2017). Determining how burnt carcasses are scavenged post-fire would contribute to understanding on whether animals killed and burnt in the fires decompose differently to animals that die without any significant burn damage.
We also acknowledge that our study was opportunistic in the sense that we happened to be undertaking scavenging studies in an area that was severely affected by the 2019/2020 bushfires.
Greater insights could be gained from using a more standard beforeafter-control-impact experimental design, but that would also likely require a study that is designed around controlled management burning. Our study site was also on the edge of a much larger fire that burned 510,000 hectares of forest (Boer et al., 2020) and eventually joined four other fires to burn over 1 million hectares of land.
The fire that impacted our study site was slow moving and was low intensity in some places, and our study site also was relatively close to unburnt habitat (<10 km). Major shifts in bird, mammal and reptile scavenging rates may therefore be expected in areas where the fire intensity was greater, or towards the core or middle of fire affected area where animals may have little refuge to escape into. An additional factor that we were not able to assess was the immediate impacts of the fire. Our post-fire survey occurred >2 months after the fire went through because the study site was not accessible straight after the fires. The local activity of predators such as dingoes and feral cats can increase within weeks rather than months post-fire (Leahy et al., 2015). Future scavenging studies should therefore aim to monitor carcasses straight after the fires, if feasible. Such studies would coincide with the period when the carcasses of animals directly killed by the fires might be present, albeit potentially burned.
Later studies (months post-fire) would coincide with the period when animals might be dying from starvation depending on the extent of disturbance created by the fire. In our study, however, there were no signs of excess carcasses in the landscape or kangaroos in poor condition due to the follow-up rains that resulted in rapid grass growth in open areas.

| CON CLUS IONS
Our study identified that carcass use by avian scavengers increased post-fire, but that a slow-moving fire may not have a major influence on overall vertebrate scavenger species richness or carcass persistence times. Although the results are likely to be different in areas more severely affected by fires, we suggest that other factors such as the presence or absence of apex scavengers may have contributed to some of the trends we uncovered. Nonetheless, our study demonstrates that carcass monitoring can be used to assess species responses to disturbance events such as fires. Given the importance of scavenging in shaping ecosystems (Wilson & Wolkovich, 2011), combined with our limited understanding of the impacts of fires on scavenging dynamics, there is a need to better understand how fires influence ecosystem processes linked to decomposition in order to inform how post-fire landscapes should be managed.

ACK N OWLED G EM ENTS
We are indebted to the Emirates One & Only Wolgan Valley Resort for providing access, accommodation and other resources during field studies in the Wolgan Valley, Blue Mountains. The assistance provided by local landholders and landholders in the Mudgee region, central New South Wales, who helped provide materials for this project was very much appreciated. We are also very thankful to the many co-workers and volunteers who provided support and assistance in the field. Funding was provided by the Holsworth

Wildlife Research Endowment and The Australia & Pacific Science
Foundation.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13390.