Carcass use by mesoscavengers drives seasonal shifts in Australian alpine scavenging dynamics

Keywords: alps, apex scavenger, breeding, brushtail possum, carrion, food source, mesoscavenger, raven, scavenger guild.

Carrion is a valuable ecosystem resource, which in contrast to plant-based detritus, is high in energy and nutrient rich (Barton et al. 2013). Although carrion is spatially and temporally patchy (Carter et al. 2007; Wilson and Wolkovich 2011), it is exploited by species that have evolved to scavenge. Historically, scavengers have been viewed as ‘bottom-feeders’, owing to associations with rotting matter, disease, and death (DeVault et al. 2003). However, scavenging is present in many taxa, from obligate and facultative large vertebrate scavengers capable of consuming a whole carcass in one feeding event (Mateo-Tomás et al. 2017), to invertebrate scavengers that can aggregate around carrion in the thousands (Forbes and Carter 2015). Together, these species form scavenger guilds which, in addition to acting as ‘natures clean-up crew’ (Grilli et al. 2019), also support critical linkages, structure, and stability in food webs (Wilson and Wolkovich 2011); distribute nutrients within and among ecosystems; and provide economic and human health benefits related to carcass disposal and sanitary measures (Beasley et al. 2019).

Carrion biomass within an ecosystem fluctuates in response to key modes of death such as predation, but also in response to environmental factors (Forbes and Carter 2015; Moleón et al. 2019). Seasons are one of the strongest governing environmental factors, and can dictate the life histories of many herbivorous and/or migratory species that form a major component of the carrion pool (Skellam 1967; Boyce 1979; Forbes and Carter 2015; Moleón et al. 2019). This is because, in highly seasonal environments, carrion biomass can become cyclically pulsed towards the ends of harsh and/or prolonged seasons when such species incur increased mortality rates (Moleón et al. 2014). This is evident in tropical and subtropical Africa following wet- and dry-season cycles (Knight 1995; Dudley et al. 2003; Jones et al. 2015). For example, each year over 1 million wildebeest (Connochaetes taurins) migrate through the Serengeti Mara ecosystem (Hopcraft et al. 2015), including through the Mara River, where mass wildebeest drownings occur during its crossing (Subalusky et al. 2017, 2020). This sudden annual influx of carrion provides many terrestrial and aquatic scavengers with an abundance of available energy and nutrients (Handler et al. 2021). Similar trends are also common in the higher northern latitudes where many ungulate species face increased mortality rates towards the ends of harsh northern winters (Clutton-Brock et al. 1997; Wilmers et al. 2003; Flint et al. 2010), or in arid Australia where drought can cause food shortages resulting in rapid declines in kangaroo populations (~40%; Caughley et al. 1985; Robertson 1986).

In response to seasonal fluctuations in the availability of carrion, scavengers have adapted and evolved to exploit food resources when they are in abundance. Facultative scavengers have considerable flexibility in their diets regarding both the relative contribution of predated and scavenged food, but also of meat, compared with other food sources (Pereira et al. 2014). For example, the Białowież Forest (Poland) scavenger guild increased scavenging rates during winter in response to greater energy expenditure (i.e. keeping warm, traversing snow), carcass availability, and lack of other food sources (Jedrzejewska and Jedrzejewski 1998; Selva et al. 2003, 2005). These dynamics can become even more complex when considering the contrasting seasonality in the life histories of each scavenger species. For instance, during peak chick-rearing season, the energy requirements of herring gulls (Larus argentatus) and lesser black-backed gulls (Larus fuscus) are inherently greater. Consequently, both species are less selective of fish species when scavenging fishery discards (Sotillo et al. 2014). A plethora of other environmental, life-history, and inter/intra specific factors can also affect species specific scavenging and, thus, add further complexities to scavenging dynamics (Rahmstorf and Coumou 2011).

Despite the obvious potential for seasonal trends in scavenging dynamics, the experimental designs of field-based scavenging studies often overlook the impact of all four seasons (Schoenly et al. 2015). Indeed, it is common to monitor scavenging dynamics surrounding carrion only during two seasons of interest (i.e. hot and cold, or wet and dry, or breeding and non-breeding; Read and Wilson 2004; López-López et al. 2013; Forsyth et al. 2014; Turner et al. 2017; Pardo-Barquín et al. 2019; Sebastián-González et al. 2019; Inagaki et al. 2020; Rød-Eriksen et al. 2020; Enari and Enari 2021). However, such an approach potentially oversimplifies the ways in which the various dimensions of scavenging can change not only between each of the seasons, but over the course of a year, and even among years (López-López et al. 2013). For instance, season, along with daily temperatures, humidity, and moisture levels can influence carrion persistence rates via the effects they have on regulating microbial and insect activity (Matuszewski et al. 2010; Turner et al. 2017; Barton and Bump 2019). Moreover, some studies completely ignore seasonal effects by monitoring scavenging dynamics only during one season (Peisley et al. 2017; Bingham et al. 2018; Cunningham et al. 2018; Rees et al. 2020; Fielding et al. 2022). Therefore, we acknowledge that although the successful design of any robust field study in scavenging ecology is underpinned by temporal aspects (i.e. diel, seasonal, yearly), spatial aspects (i.e. representative of ecosystem, spatially independent), and sample size (Schoenly et al. 2015), in conjunction with a suite of other considerations (i.e. carcass size and weight; Moleón et al. 2015; Sebastián-González et al. 2019), seasonal aspects arguably have the most potential to consistently affect scavenging dynamics.

Here, we exploited the highly seasonal nature of the Australian Alps to monitor the use of carcasses by vertebrate scavengers. We undertook carcass monitoring across four seasons to test hypotheses related to (1) scavenger species richness/composition; (2) the time to first detection and scavenging at a carcass; (3) scavenger activity (i.e. probability of a scavenger investigating versus scavenging a carcass); and (4) the time spent investigating and scavenging a carcass. Accelerated decomposition of carrion during summer as a product of increased invertebrate and microbial activity is known to produce stronger carcass-linked odours (Cammack et al. 2015). Consequently, we predicted that increased olfactory cues during summer would result in greater detectability of carrion by vertebrate scavengers and, thus, shorter time to first arrival and scavenging at carcasses. Conversely, despite carrion being less detectible during winter, it was predicted that the probability of vertebrate scavenging would increase, and time spent scavenging would be longest. This hypothesis was based on vertebrate scavengers potentially being more heavily reliant on carrion during winter when other food resources are scarce (Jedrzejewska and Jedrzejewski 1998; Selva et al. 2003, 2005). Last, we predicted that our findings would highlight the importance of undertaking field-based scavenging studies across all four seasons, so as to fully understand the complexities of scavenging dynamics and the interactions that take place surrounding carrion.

This study was conducted in Kosciuszko National Park, in southern New South Wales, Australia. This region includes many of Australia’s highest peaks and ranges, including the highest peak, Mount Kosciuszko (2228 m asl). The landscape is delineated into three altitudinally distinct ecological communities, namely, montane (500–1500 m), subalpine (1500–1850 m), and alpine (1850–2228 m) (Bear et al. 2006). Each of these communities are highly seasonal, with the temperatures experienced throughout the course of a year differing by up to 50°C (−10 to 40°C). During winter much of the landscape can be covered in snow, but by the following summer the same landscape can experience intense bushfires (Ward et al. 2020). This work was undertaken within the montane zone (between approximately 1000–1500 m) which is characterised by forest stands dominated by snow gums (Eucalyptus pauciflora) and various other Eucalyptus species.

Ethics approvals (Project number: 2019/1640, The University of Sydney) and scientific permits (Licence number: Sl102334, New South Wales Office of Environment and Heritage) were obtained to undertake the fieldwork. Carcasses of eastern grey kangaroos (Macropus giganteus) were sourced fresh and locally from legally approved shooting operations that are conducted to control or harvest overabundant kangaroo populations.

A 15 km transect was established through Kosciuszko National Park, along which all carcass monitoring took place (Fig. 1). This transect ran from north-east to south-west and was selected because of its accessibility (i.e. road access), because it is an undisturbed area with little human activity, and because consistent vegetative structure allowed carcass-monitoring sites with similar canopy cover (~75%) to be established along the transect.

Fig. 1.  Location of the monitoring transect, along which all 60 carcass sites were established within Kosciuszko National Park.

Monitoring periods coincided with the four seasons: March–May 2020 (autumn, average min/max: 1°C/13°C); July–September 2020 (winter, average min/max: −3°C/8°C); September–November 2020 (spring, average min/max: 2°C/15°C); January–March 2021 (summer, average min/max: 6°C/21°C). During each season, 15 sites were established along the transect, separated by approximately 1 km from the nearest sites monitored within the same season, and approximately 250 m from the nearest sites monitored during other seasons (60 different sites in total; Fig. 1). The separation of sites ensured a level of spatial independence, and it prevented habituation of scavengers to a carrion source location (Inagaki et al. 2020; Spencer and Newsome 2021; Spencer et al. 2021; Newsome and Spencer 2022).

Within each season, one fresh eastern grey kangaroo carcass was placed at each carcass-monitoring site (15 in total). The carcasses ranged in weight from 12 kg to 70 kg and, on average, were 28.3 kg (±1.498 standard deviation; Supplementary material Table S1). Each carcass was secured, using wire ties, to star pickets driven into the ground to ensure that they remained in situ (methods used were as per Spencer and Newsome 2021; Spencer et al. 2021; Newsome and Spencer 2022).

Vertebrate scavenger activity was monitored at each site for 60 days by using a Reconyx PC800 Hyperfire™ remote camera. Each camera was placed on a free-standing star picket 3 m north of the carcass; the southern aspect of the cameras decreased exposure to direct sunlight, which would otherwise reduce image quality. The cameras were calibrated to take photographs continuously (approximately one image per second) when triggered by thermal movement (i.e. rapid-fire, no wait period). These approaches and methods follow those previously used in field-based scavenging research (Kostecke et al. 2001; King et al. 2006; Flint et al. 2010; Inagaki et al. 2020; Rød-Eriksen et al. 2020; Spencer and Newsome 2021; Spencer et al. 2021; Newsome and Spencer 2022).

Remote-camera images were analysed for species presence and the number of individuals of a species present. An ‘event’ was characterised as a visitation by a species that occurred more than 10 min after the last visitation by that same species (methods as per Spencer and Newsome 2021; Bragato et al. 2022; Newsome and Spencer 2022). Only species-specific events could be characterised because identification of individuals for most species was not possible. An event was characterised as a ‘scavenging event’ if the species present scavenged on the monitored carcass in at least one of the remote-camera images consisting of that event, otherwise the event was characterised as an ‘investigation event’. Data from species recorded to have scavenged at least once were included in the statistical analyses.

The R statistical environment (R Core Team, ver. 4.2.1; Vienna, Austria; https://www.R-project.org/) was used for all statistical analyses. To determine whether species composition differed among the seasons, we used a permutational multivariate analysis of variance (PERMANOVA) within the R Package ‘vegan’ (Oksanen et al. 2022). The PERMANOVA was used in conjunction with an analysis of similarities (ANOSIM; also within R Package ‘vegan’; Oksanen et al. 2022) to also test whether the magnitude of difference among seasons was greater than that within seasons (Somerfield et al. 2021). To determine which scavenger species were driving any differences in species composition between the seasons, a similarity percentages (SIMPER) analysis was used, also within the R Package ‘vegan’ (Oksanen et al. 2022).

To characterise the seasonal nature of scavenging, the following four response variables were considered: scavenger species richness (Conway–Maxwell Poisson distribution); time to first detection of and scavenging at a carcass (in hours; gamma distribution); scavenger activity (i.e. probability of a scavenger investigating vs scavenging a carcass; binomial distribution); and investigation and scavenging event duration (in minutes; gamma distribution). Each of the response variables was modelled against the explanatory variables of season, altitude, and carcass weight.

Models were constructed using either generalised linear models (GLM) within the R Package ‘lme4’ (Bates et al. 2015), or using generalised linear mixed models (GLMM) within the R Package ‘glmmTMB’ (Brooks et al. 2017). We utilised GLMs when modelling response variables that consisted of only one measurement at each independent carcass-monitoring site, namely, scavenger species richness and time to first detection/scavenging. GLMMs were used for response variables that yielded repeat measurements at each independent carcass-monitoring site (i.e. scavenger activity and event duration) to account for the random effect of site. Model assumptions were checked using residual diagnostics within the R Package ‘DHARMa’ (Hartig and Hartig 2017). This also included testing model residuals for spatial autocorrelation. If detected, a spatial autocorrelation covariate was created using the R Package ‘spdep’ (Bivand et al. 2013) and subsequently included as an explanatory variable in the model (Dormann et al. 2007). These models were constructed twice, once using only investigation events and once using only scavenging events. Only one model was constructed for scavenger activity because it is a binomial response variable (either investigation (0); or scavenging (1)) designed specifically to determine the probability of either an investigation event or scavenging event.

To determine the most parsimonious model(s), corrected Akaike information criterion (AIC_c; Sakamoto et al. 1986) was used to consider the different combinations of explanatory variables (i.e. combinations of season, altitude, and carcass weight; with a ΔAIC_c level of significance <2). Model selection was facilitated by the dredge function within the R Package ‘MuMIn’ (Bartoń 2022).

Significance testing (level of significance P < 0.05) was undertaken using the base model (i.e. where the scavenging response variable was modelled against all three explanatory variables, namely, season, altitude, and carcass weight). This determined which explanatory variables (season, altitude, and/or carcass weight) and/or their levels (with autumn used as the reference level) were important in explaining the scavenging response variable. Finally, a Tukey’s honest significance test within the R Package ‘emmeans’ was used to determine which seasons were significantly different from one another regarding the modelled scavenging response variable (Lenth 2016). This pair-wise test approach could not be used when modelling species richness because it followed a Conway–Maxwell–Poisson distribution, which does not support post hoc analyses.

Of the 60 carcass sites monitored, remote-camera data was gathered for 58; remote-camera data for two sites (one during winter and one during summer) were lost due to theft and camera failure. The remote cameras took 745 599 images of 34 different species including both scavengers and non-scavengers. Of these taxa, nine were considered scavenger species on the basis of recorded scavenging of the monitored carcasses (Fig. 2). These were spotted-tail quoll (Dasyurus maculatus), feral cat (Felis catus), dingo (Canis dingo), pied currawong (Strepera graculina), wedge-tailed eagle (Aquila audax), brushtail possum (Trichosurus vulpecula), raven spp. (Australian raven (Corvus coronoides) and little raven (Corvus mellori), which were indistinguishable from one another in the remote camera images, and so were grouped together as ‘raven(s)), red fox (Vulpes vulpes), and feral pig (Sus scrofa; Fig. 3).

Fig. 2.  The total number of carcass sites visited by each scavenger species per season. Carcass-monitoring sites where scavenging was recorded (solid fill) are distinguished from those that were only investigated (shaded fill).

Fig. 3.  Remote camera images for each of the members of the Australian alpine scavenger guild, determined by recorded scavenging of carcasses in Kosciuszko National Park. (a) Spotted-tail quoll (Dasyurus maculatus), (b) feral cat (Felis catus), (c) dingo (Canis dingo), (d) pied currawong (Strepera graculina), (e) wedge-tailed eagle (Aquila audax), (f) brushtail possum (Trichosurus vulpecula), (g) raven spp. (Australian raven (Corvus coronoides) and little raven (Corvus mellori), indistinguishable from one another in the remote camera images), (h) red fox (Vulpes vulpes), and (i) feral pig (Sus scrofa).

In total, 6857 distinct events were recorded, of which, 2680 were investigation events and 4177 scavenging events (Fig. 4a). Brushtail possums and ravens accounted for 88% of the total recorded events, whereas spotted-tail quolls accounted for the fewest events (Fig. 4b).

Fig. 4.  (a) The total number of events per season and (b) the percentage of total events attributed to each scavenger species where brushtail possums accounted for 61%, raven spp. 27%, pied currawongs 5%, and all other species <2% each.

There were weak differences in species composition among seasons for both investigation events (pseudo-F = 2.898, d.f. = 57, residual d.f. = 54, P = 0.002, nperm = 999) and scavenging events (pseudo-F = 2.991, d.f. = 53, residual d.f. = 50, P = 0.004, nperm = 999; Table S2). Species composition for investigation events was significantly different between autumn and spring (P = 0.030) and between spring and winter (P = 0.018; Table S3). Whereas for scavenging events, species composition was significantly different between autumn and winter (P = 0.036) and between spring and winter (P = 0.012; Table S3). Additionally, there was very little difference when comparing within-season species composition with between-season species composition for both investigation events (R = 0.109, P = 0.001, nperm = 999) and scavenging events (R = 0.109, P = 0.002, nperm = 999). Brushtail possums and ravens were found to be the primary drivers of the observed differences in species composition among the seasons (Fig. 5, Tables S4, S5).

Fig. 5.  The percentage contribution of each scavenger species (contributed most to least, top to bottom of each respective legend) to the observed differences in species composition between each of the seasons (summer, sun; autumn, leaf; winter, snowflake; spring, flower) for (a) investigation events and (b) scavenging events. The percentage contribution was adapted from the SIMPER analyses cumulative sum results (CUMSUM; Tables S4, S5).

Seven models demonstrated equivalent levels of support (ΔAIC_c < 2) for explaining the variation in scavenger species richness during investigation events (Table S6). Scavenger species richness was affected by season, with summer (P = 0.023) and spring (P = 0.038) being significantly different from autumn, the reference level (Table S7). Four models demonstrated equivalent levels of support (ΔAIC_c < 2) for explaining the variation in scavenger species richness during scavenging events (Table S6), but the base model was not significant (Table S7).

Spatial autocorrelation was detected in the model residuals for time to first detection; therefore, a spatial autocorrelation covariate (SACC) was included. Three models demonstrated equivalent levels of support (ΔAIC_c < 2) for explaining the variation in time to first detection of carcasses (Table S8). The time to first detection of carcasses was affected by season (Table S9). Specifically, time to first detection of carcasses (in hours) was 5.344 and 9.622 times longer during summer than during spring (P = 0.012) and winter (P = 0.001) respectively (Fig. 6; Table S10).

Fig. 6.  The time to first detection (hours) of monitored carcasses by vertebrate scavengers during each of the seasons (mean is indicated by a cross).

Three models demonstrated equivalent levels of support (ΔAIC_c < 2) for explaining the variation in time to first scavenging of carcasses (Table S8), but the base model was not significant (Table S9).

Three models demonstrated equivalent levels of support (ΔAIC_c < 2) for explaining the variation in scavenger activity (Table S11). Scavenger activity was affected by season (Table S12). Specifically, scavenging was 2.186 and 3.086 times more likely than investigation during spring than during autumn (P = 0.014) and summer (P ≤ 0.001) respectively, and 2.683 times more likely during winter than during summer (P = 0.002; Table S13).

Four models demonstrated equivalent levels of support (ΔAIC_c < 2) for explaining the variation in the duration of investigation events (Table S14). The duration of investigation events was affected by carcass weight, whereby larger carcasses were investigated for shorter durations (P = 0.031; Fig. S1, Table S15). Four models demonstrated equivalent levels of support (ΔAIC_c < 2) for explaining the variation in the duration of scavenging events (Table S14). The duration of scavenging events was affected by season (Table S15). Specifically, the duration of scavenging events (in minutes) was 1.916 and 1.556 times shorter during spring than during autumn (P = 0.001) and winter (P = 0.027) respectively (Fig. 7; Table S16).

Fig. 7.  The average duration of scavenging events (minutes) by vertebrate scavengers for each carcass monitored during each of the seasons (mean is indicated by a cross).

In this study, we investigated how seasons affect vertebrate scavenging dynamics in an alpine environment. Our findings demonstrated that scavenging dynamics were highly seasonal, and this trend was consistent (i.e. did not significantly change) across an altitudinal gradient for carcasses of all weights. Australia has no strictly obligate vertebrate scavenger, but our study recorded carcass use by nine species of facultative scavenging vertebrates. Scavenging rates by these species should be highly seasonal, linked to factors such as fluctuations in carrion biomass within the landscape, availability of other food sources, and/or the life histories of the scavenger species (Knight 1995; Jedrzejewska and Jedrzejewski 1998; Dudley et al. 2003; Selva et al. 2003, 2005; Wilmers et al. 2003; Wilmers and Getz 2004; Selva and Fortuna 2007; White 2011; Pereira et al. 2014; Jones et al. 2015). Such observations were made in this study, but the trends uncovered were overwhelmingly dictated by the scavenging rates and activities of mesoscavengers, specifically brushtail possums and ravens. Together, these species accounted for 88% of all recorded events. The scavenging rates observed by ravens were not unusual, with many studies both within Australia and globally describing raven species as prolific scavengers (Selva et al. 2003; O’Brien et al. 2010a; Mateo-Tomás et al. 2015; Spencer et al. 2021; Bragato et al. 2022; Newsome and Spencer 2022). However, the brushtail possum has received little recognition as a regular scavenger (but see Brown et al. 1993; Heinsohn and Barker 2006; O’Brien et al. 2010b).

Scavenging dynamics surrounding carrion are not typically dictated by the species-specific scavenging rates of mesoscavengers, but that of larger apex scavenger species (Wilmers et al. 2003; Wilmers and Post 2006; Krofel et al. 2012; Allen et al. 2014). This is because larger species are generally more competitively dominant surrounding carrion (Moleón et al. 2014; Pereira et al. 2014; Inagaki et al. 2020), and are able to open up carcasses, and, thus, provide access to smaller scavengers unable to do so (Selva et al. 2003, 2019; Moleón et al. 2015). However, in this study, larger scavengers such as dingoes, wedge-tailed eagles, and feral pigs accounted for only 5% of all recorded events, suggesting that they were either not abundant during the study period or were not scavenging frequently. Notably, in one case, a dingo was observed to have consumed an entire kangaroo carcass within a 24-h period. The relative absence of dingoes at carcass sites, along with similarly low rates of scavenging by red foxes and feral cats, is likely to have especially influenced the use of carcasses by brushtail possums who regularly feature in the diets of these three predators (Jones and Coman 1981; Newsome et al. 1983; Dickman 1996; Vernes et al. 2001; Kinnear et al. 2002; Brook and Kutt 2011; Pascoe et al. 2012; McComb et al. 2018; Fleming et al. 2021; Moseby et al. 2021). Further still, recent findings in Australia have demonstrated that ravens, in the absence of top scavengers, scavenged across all four seasons and not only during the cooler months when other resources were low (Fielding et al. 2021). Therefore, whether the relative absence of larger scavengers at carcass sites in this study aided mesoscavenger access to the food source is unknown; however, it could be reasonably expected if this resulted in reduced predation risk for the mesoscavengers and/or less competition for the food source.

The time it takes scavengers to detect a carcass, and subsequently scavenge it, is intrinsically linked with carcass decomposition rates and persistence within the ecosystem (Santos et al. 2016). Carcass detection by scavengers is dependent on a number of factors including olfactory cues, visual cues, inter/intra specific cues, and search effort (López-López et al. 2013; Cortés-Avizanda et al. 2014; Cammack et al. 2015). We, therefore, expected that olfactory cues would be the primary mode of detection, given that closed-canopy forested ecosystems, such as the montane zone of Kosciuszko National Park, make visual detection and certain forms of inter/intra specific cues difficult (Barton and Bump 2019; Moleón et al. 2019). Consequently, it was predicted that, during summer, greater olfactory cues owing to increased temperatures, and increased invertebrate and microbial scavenging activity, would facilitate shorter time to first detection of carcasses than during the other seasons (Mann et al. 1990; Carter et al. 2007; Forbes and Carter 2015; Barton and Evans 2017; Barton and Bump 2019). However, there was no evidence to support this hypothesis and, instead, carcasses took longer to be detected during summer, especially when compared with spring and winter.

It is possible that the same mechanism as that expected to facilitate shorter time to first detection of carcasses, invertebrate and microbial scavenger activity, in fact, hindered detection of carcasses by the vertebrate scavengers. During summer (warmer months), carcasses are rapidly colonised (within minutes) en masse by invertebrates and a plethora of microorganisms (Ray et al. 2014; Crippen et al. 2015; Merritt and De Jong 2015; Barton and Evans 2017; Anderson et al. 2019; Evans et al. 2020). The intense scavenging activity that follows accelerates carcass decay through the different stages of decomposition, potentially at such a rate that vertebrate scavengers were given too little time to detect carcasses; i.e. the invertebrate scavengers outcompeted vertebrate scavengers (Ray et al. 2014; Barton and Evans 2017; Evans et al. 2020). Indeed, previous observations of carcass persistence rates within the study site have indicated that eastern grey kangaroo carcasses take at least twice as long to reach complete consumption/decomposition (only skin and bones remaining) in cool compared with warmer periods (Spencer and Newsome 2021). During the winter monitoring period, we observed that some carcasses had not reached complete consumption/decomposition, even after 60 days.

Many global studies have demonstrated that vertebrate scavenging rates are lowest during summer when other food sources are more abundant, and vice versa during winter (Jedrzejewska and Jedrzejewski 1998; Selva et al. 2003, 2005). In our study, brushtail possums accounted for 81% of all recorded events during winter when they scavenged three times more often than during summer. Generally, the diet of brushtail possums consists of leaves, flowers, fruit (Eucalyptus and Acacia) and insects (Evans 1992; How and Hillcox 2000; Wiggins et al. 2003), most of which are only seasonally available during warmer months. Therefore, this marsupial may be exhibiting a greater dependence on carrion during winter that is similar to that of other scavengers in the northern hemisphere (Jedrzejewska and Jedrzejewski 1998; Selva et al. 2003, 2005). This greater dependence on carrion by the brushtail possum during winter is likely to have influenced our analyses of the ‘scavenger activity’ response variable, which recorded scavenging to be 2.683 times more likely than investigation of carcasses during winter than in summer. This may also explain why carcasses took six times longer to be detected during summer (144 h; 6 days) than during winter (24 h), a finding that was contrary to our prediction that time to first arrival would be shortest in summer. Collectively, during autumn, winter, and spring, 93% of the first detections of a carcass were by either brushtail possums or ravens. Conversely, during summer, only 57% of the first detections of a carcass were by either brushtail possums or ravens. Therefore, it is possible that the brushtail possums and ravens may have a disproportionately greater bearing on the time it takes the collective scavenger guild to first detect a carcass.

The species-specific breeding seasons of scavengers can also have profound impacts on their respective scavenging behaviours (Camphuysen 1995; O’Brien et al. 2010a; White 2011; Coppola et al. 2020). The Australian raven and little raven breed from late winter into spring (Moon 2005; Whisson et al. 2015; Ekanayake et al. 2018). Initially, nest construction is prioritised in this early breeding season, and the associated activities are characterised by frequent and short visitations between the nest and sources of nesting material (Moreno-Opo and Margalida 2013). In our study, during the spring monitoring period, numerous remote-camera images captured ravens collecting fur from the kangaroo carcasses, presumably for nest construction (Fig. 8).

Fig. 8.  A raven sp. (Corvus coronoides or Corvus mellori) breeding pair collecting fur from an eastern grey kangaroo (Macropus giganteus) carcass during the spring monitoring period, presumably for nest construction.

Following nest construction, chick rearing often requires breeding pairs to divide time between foraging, feeding chicks, and being vigilant and protective of the nest (O’Brien et al. 2010a). Inherently, during this time, ravens incur greater energy costs associated with these activities, and, thus, must supplement their diets with protein-rich sources and/or greater quantities of food, such as carrion (Camphuysen 1995; O’Brien et al. 2010a). Of all recorded raven scavenging events during this study, 67% were during spring, and this suggests that ravens may heavily rely on carrion to supplement their diet, and/or that of their chicks, during the breeding season (O’Brien et al. 2010a). Further still, this flurry of raven scavenging during spring (a 170% increase on the average of annual raven scavenging events) accounted for 73% of all scavenging events for the collective scavenger guild. Therefore, given their dominance during spring, it is likely that the frequent but brief scavenging events, characteristic of the raven breeding season (i.e. breeding pairs frequently flying back and forth between the nest and nesting materials and/or food sources), were also deterministic of the shorter scavenging event duration recorded for the collective scavenger guild during spring.

The findings here regarding ravens are also indirectly linked to our initial prediction for longer scavenging-event duration during winter; our models indicated that scavenging-event duration during winter was 1.556 times longer than during spring. Although raven scavenging behaviours during their breeding seasons (spring) are likely to have determined this result and supported our prediction, it is juxtaposed to the evidence we initially used to form our hypotheses. That being, previous studies in the northern hemisphere demonstrating scavenging rates to increase during winter in response to a lack of other available food sources (Jedrzejewska and Jedrzejewski 1998; Selva et al. 2003, 2005). This raises several questions that need be addressed regarding the degree to which the Australian alpine winter affects food sources, species diets, and associated flow-on effects to scavenging dynamics versus other alpine areas in the world, and indeed non-alpine ecosystems that experience harsh winters.

Scavenging dynamics in this study were highly seasonal but were dictated by the scavenging activities and behaviours of mesoscavengers, namely, the brushtail possum and ravens. The high rate of scavenging by these species drove the seasonal trends in scavenging dynamics, but the direction in which the seasonal effects affected the scavenging response variables was not always as we predicted. This exemplifies the unexpected influence that seasons can have on ecological processes linked to scavenging and highlights the need for seasonally replicated experimental approaches in field-based scavenging research.

Although carcass weight was demonstrated to have a significant effect on the duration of investigation events, it is possible that this result was skewed by four carcass-weight outliers (55 kg, 55 kg, 57 kg, and 70 kg, compared with an average of 28.3 kg; Table S1). A cursory look at these four carcasses demonstrated relatively short investigation-event durations when compared with all other carcasses, resulting in an instance where potential statistical outliers represent an ‘extreme’ in the response variable (Supplementary material Fig. S1). The removal of said ‘outliers’ from the dataset was considered, but residual analyses did not demonstrate any significant outliers in the model, and, therefore, doing so could have potentially introduced bias. For exploratory purposes and transparency, we re-ran the model removing only the furthest outlier (carcass site 42; carcass weight 70 kg; Table S1) and this returned a model in which carcass weight was not significant (Table S17). It remains unknown if this trend was indeed a real ecological finding, or outlier skew as predicted. We, therefore, recommend that further research is needed to explore the relationship between carcass weight and event duration.

More generally, the methods used herein can be utilised to monitor scavenging dynamics surrounding carrion, and even be further developed to improve field-based designs and yield additional ecological information. Such improvements could include utilising a remote-camera grid in tandem with the carcass monitoring to determine the activity and diversity of scavenger species in the given study region. This direct comparison of scavenger species at carcasses and more generally in the landscape would allow for more informed conclusions surrounding scavenging dynamics, such as, for example, whether species-specific scavenging rates are a product of scavenging behaviour or species density in the study region. Additionally, comprehensive altitudinal gradients that may detect elevational differences in scavenging dynamics could lead to particularly novel findings, insights missed by this study given the small altitudinal gradient used (500 m). This is especially important in scavenging studies that experience much higher altitudes than those present in Australia. Because of geographic limitations, in this study, we were restricted to only 15 carcass-monitoring sites per season. An improved sampling design that includes more within-season replications, multiple transects, and replicated studies over multiple years may help detect more subtle seasonal differences missed by this study, as well as account for within-site (i.e. more representative of the ecosystem) and annual variability (Schoenly et al. 2015).

Our observations of frequent scavenging by the brushtail possum are highly novel, given that they mainly feed on leaves, flowers, and fruit (Eucalyptus and Acacia) and are commonly considered generalist herbivores/folivores (Evans 1992; Owen and Norton 1995; How and Hillcox 2000; Cochrane et al. 2003; Wiggins et al. 2003; Sweetapple et al. 2004). The results of this study, at the most, suggest that the high rates of brushtail possum scavenging were facilitated by reduced predations risk from, and competition with, larger dominant scavengers. Notably, brushtail possums have been previously observed to, on occasion, eat insects, raid birds’ nests, and scavenge on carrion (Brown et al. 1993; Nugent et al. 2000). Therefore, it is also possible that other factors characteristic of the alpine environment, (i.e. less alternative food sources, harsher winters) in which this study was undertaken, resulted in our unique findings regarding brushtail possum scavenging. With this in mind, we recommend that more in-depth dietary studies are undertaken that specifically compare the diets of brushtail possums in the Australian Alps with those of brushtail possums in other Australian ecosystems. This will help quantify the amount of carrion brushtail possums consume compared with other food sources and provide further insights into the ecosystem factors that determine such dietary shifts.

We also observed that ravens were somewhat dependent on carrion both as a source of energy, and nesting material during the breeding season. Very little work has been conducted to determine how carrion biomass within an ecosystem (year to year) may affect the breeding success of facultative scavengers (Beasley et al. 2019). This is an area that warrants future attention, given the results of this study and also those elsewhere regarding ravens (O’Brien et al. 2010a).

Lastly, quantifying carrion biomass in the broader landscape would help determine whether it is a limited resource, or in abundance (Barton et al. 2019). Understanding scavenging dynamics in relation to the availability of carrion would help explain how scavengers react to changes in food supplies and, thus, further develop our knowledge of the role of carrion in supporting biodiversity and ecosystem function.

Supplementary material is available online.

Data available on request.

Thomas Newsome is an Associate Editor of Wildlife Research but was blinded from the peer-review process for this paper.

Funding for this work was provided by the Australian Alps National Parks Cooperative Management Program.