INTRODUCTION

Raptors are keystone species that disproportionately influence the health and function of their surrounding environment. Their predatory and scavenging behaviors provide many important ecosystem services such as regulating prey populations (Abramsky et al. 2002, Ydenberg et al. 2017) and maintaining a sanitary environment (Grilli et al. 2019). The rich functional diversity among raptors results in them using a broad spectrum of prey resources and habitat types (Robinson 1994, Montaño-Centellas et al. 2023). Some species are dietary specialists occupying a narrow trophic niche, such as the Snail Kite (Rostrhamus sociabilis; Sykes 1987), while others, such as the Red-tailed Hawk (Buteo jamaicensis) are dietary generalists that occupy a wide trophic niche (Fitch et al. 1946). Furthermore, this information can offer important insights about a species' ecology including its dietary composition, prey selection, and trophic niche (e.g., Navarro-López and Fargallo 2015, Kross et al. 2018, Schoenjahn et al. 2022).

Knowing the degree of diet specialization in a species can have important management and conservation implications. For example, it may enable the prediction of how changes in prey abundance associated with environmental change such as habitat loss or climate change may impact its populations. If a species is known to occupy a narrower trophic niche, it may be less resilient to these types of environmental change processes given its specific ecological requirements (Fargallo et al. 2022).

Determining dietary specialization for raptors can be difficult because they are typically highly mobile and wide-ranging, making in situ observations of these behaviors difficult to obtain in the field. Although diet may vary through the year, the breeding season offers a unique opportunity to determine their dietary habits by direct observation of these behaviors and collection of prey remains surrounding the nest.

Also influencing prey selection within many diurnal raptor species is reverse sexual size dimorphism (RSD), in which females are typically larger than males (Newton 1979). There are numerous competing hypotheses attempting to explain the basis for RSD, although most are difficult to test or fail to provide a general explanation that is consistent across multiple taxa (Selander 1966, Snyder and Wiley 1976, Bildstein 1992). However, there is a clear relationship between RSD and a species' diet among those that catch relatively large and agile prey, such as birds, which exhibit the greatest relative size difference (Newton 1979, Krüger 2005). Empirical evidence shows that consumption of avian prey takes the longest, potentially enhancing the asymmetry in sex roles as females dedicate more time to assist the young with feeding while the male hunts (Sonerud et al. 2014). An alternative nonexclusive theory is that RSD may allow pairs to access a broader array of prey by capturing prey species based on their body size, possibly reducing intersexual competition (Selander 1966), allowing more efficient foraging by the smaller male (Earhart and Johnson 1970, Reynolds 1972, Andersson and Norberg 2008), or better nest defense by the larger female (Schoenjahn et al. 2020).

Australia's Red Goshawk (Erythrotriorchis radiatus) exhibits one of the greatest degrees of RSD among raptors, with males being on average approximately half the mass of females (mean mass males = 588 g vs. females = 1118 g; Snyder and Wiley 1976, Krüger 2005, C. MacColl unpubl. data). This extreme size dimorphism makes it an ideal species to test for compliance with the ecological traits widely associated with RSD such as prey size, type (e.g., birds, mammals), and formidability. Red Goshawks are also a species of conservation concern, classified as endangered (BirdLife International 2022, Australian Department of Climate Change 2023) due to a recent decline, which now largely confines the species to the tropical savanna of northern Australia (MacColl et al. 2023). Knowledge of the species' ecology is limited, largely due to the challenges presented by finding a sufficient sample size of nests for study given their occurrence at very low density and use of forest cover. The species is known to feed almost exclusively on other birds, particularly parrots (Psittaciformes), though it remains unclear whether the prey is partitioned between the sexes, or if the prey is a preferential subset of what is available (Aumann and Baker-Gabb 1991). The species also breeds in two distinct habitat types (woodlands and riparian areas), which may influence dietary preference and the corresponding trophic niche occupied by pairs in these environments.

This study sought to fill some of these knowledge gaps by compiling a comprehensive dataset of Red Goshawk prey records spanning the species' current breeding range across the tropical savanna ecosystems of northern Australia. We quantified its breeding diet by prey species including the number of individuals and relative contributions by biomass, and whether males and females are capturing different sized prey. We investigated regional variation in diet and explored whether dietary composition changes with breeding habitat between woodland and riparian nesting sites. We investigated changes in prey size throughout the breeding season and asked whether nests with male or female nestlings received different sized prey. We explored whether prey selection simply reflects availability or if the goshawks specialized on specific prey species. Lastly, we considered the degree to which the Red Goshawk's diet may have been a driver in its recent range retraction and the up-listing of its conservation status to endangered.

METHODS

Data Collation. Between 2019 and 2023, we obtained data on prey captured by Red Goshawks through observations at nests throughout the species' extant breeding range. This range comprises the tropical savannas of northern Australia within the Northern Territory, and the states of Queensland and Western Australia (Fig. 1). For this analysis we treated Rainbow Lorikeets (Trichoglossus moluccanus) and Red-collared Lorikeets (Trichoglossus rubritorquis) as the same species (hereafter Rainbow/Red-collared Lorikeet). The Rainbow Lorikeet occurs throughout the Queensland range of Red Goshawk, whereas Red-collared Lorikeet only occurs within Red Goshawk's Northern Territory and Western Australian range. Until recently these taxa were considered the same species and treating them as the same species here means our summary statistics better reflect the ecological relevance of our results. The taxonomy of all prey species presented in our analysis can be found in the Appendix.

Figure 1.

Red Goshawk prey sampling locations, and extant (tropical savanna) and extirpated breeding ranges of northern and eastern Australia, respectively.

Prey records were obtained using three methods: (1) observation and/or collection of prey items brought to the nest during monitoring activities; (2) photographic records obtained from online image searches; and (3) photographic records and observations obtained from community scientists either directly via requests for information or indirectly via online database searches including eBird (2023a) and Atlas of Living Australia (2023). Combining these mixed datasets should help minimize potential biases inherent in any one method of data collection on raptor diets (e.g., Mersmann et al. 1992, Real 1996, Redpath et al. 2001). Prey records from community scientists spanned the years 1997–2023.

Information supporting each prey record was recorded, where available, including the date, location, observer, prey species, breeding phase, and habitat type. Because prey collection was sporadic, typically only occurring 2–4 times per season for each nest, prey remains showing advanced decomposition were not assigned to a breeding phase due to uncertainty surrounding the actual time of prey capture and corresponding stage of the breeding cycle. All photographic prey records were cross-referenced by location, date, and prey species to avoid possible duplication of the same prey record by multiple observers because the same known Red Goshawk nests are visited routinely by birdwatchers (MacColl et al. 2023).

Collection of Prey Remains. We searched for prey remains within an approximate radius of 60 m surrounding the nest tree. We collected feathers, skeletal remains, and pellets (regurgitated masses of indigestible matter) and removed them from the site for examination to identify prey species and also to avoid duplicate counts during later sampling events. We used diagnostic body parts to count the minimum number of individuals found among the prey items including the skull or pelvic girdle. In the absence of diagnostic body parts, we counted individuals based on the minimum number of body parts required to constitute one prey individual. For example, if three Rainbow/Red-collared Lorikeet wings were found, then this was counted as a minimum of two individuals. If there was an absence of any skeletal remains but a presence of scattered feathers indicative of a bird being plucked and eaten, we counted this as a unique individual. We also collected pellets and analyzed them for the presence of distinguishable species. If we detected a species inside a pellet, we also counted this as a unique individual, which could lead to duplication if other remains of the same individual were also found and counted separately. We obtained the mean body mass in grams for each prey species as well as their primary habitat use (e.g., woodland, forest, grassland) from Tobias et al. (2022).

Known Prey Captures by Sex. When possible, we noted the sex of the Red Goshawk associated with each observed predation event. These data were typically gleaned from records where the predation event was directly observed and described in the observation notes or if the Red Goshawk was photographed clasping the prey species. For only photographic records, if the prey item was apparently fresh and intact (e.g., unplucked bird), we assumed the individual holding the prey was responsible for its kill. This may dismiss some valid records of prey captures by sex if the prey item was prepared (e.g., plucked and decapitated) or partly eaten by the time the photograph was taken. However, including these images could also introduce potential bias (e.g., incorrect assignment of sex responsible for prey capture) given that prey is typically prepared by Red Goshawks prior to nest delivery (i.e., male to female or adults to young; C. MacColl unpubl. data). This known behavior enables reliable sex assignment to images of Red Goshawks holding freshly caught prey. Conversely, images of fledglings holding prey or adults with partly eaten prey were excluded from this analysis given the higher likelihood that the prey could have been captured and delivered by another individual. We used this subset of the prey data to investigate differences in prey species and body size between the sexes.

Fledgling Sex. We recorded the number and sex of fledglings produced at a nest in the year that the prey sample was collected or observed if known. We determined sex only after the nestlings had fledged or during the late stages of the nestling phase, when the significant RSD in the species enabled reliable identification of males versus females. This information was used to assess potential differences in the size of prey being provided to nests that produced male versus female young. We excluded nests where pairs produced two fledglings from this analysis irrespective of sex (i.e., single sex or mixed broods) so as to remove the potential bias of brood size influencing the results.

Breeding Habitat. We recorded the broad habitat type of each nest alongside its associated prey records. All nests in this study were located in tall Darwin stringybark (Eucalyptus tetrodonta)-dominated woodlands or on freshwater creek (riparian) systems in tall stands of weeping paperbark (Melaleuca leucadendra). We assigned each nest to one of two fundamental breeding habitats: woodland or riparian. We used this subset of the data to assess prey differences between breeding habitats used by Red Goshawks across northern Australia.

Prey Use vs. Availability. We determined the assemblage of bird species available as potential prey for Red Goshawks based on their primary habitat, body mass, and abundance within 10 km of their nest locations only in the Cape York Peninsula, Queensland. Based on the traits identified in our prey data, we chose species primarily using woodland and forest (Tobias et al. 2022) and those falling within the Red Goshawk's prey size range. Note that woodlands and forest are both tree-dominated habitats though the former differs by supporting lower tree height and density with a grassy understory given its more open canopy structure (Tobias et al. 2022). We determined the relative rank abundance of these birds based on their total counts from complete checklists on eBird (2023b) between the years 1997 and 2023 to align with our prey dataset. We then spatially assigned subsets of these data to only checklists conducted within a 10-km radius of Red Goshawk nests that contributed prey data. The size of this buffer was determined by preliminary information on male foraging range during the breeding season by Aumann and Baker-Gabb (1991) and C. MacColl (unpubl. data). This analysis was restricted to the Cape York Peninsula region of Queensland because it had consistent breeding habitat (i.e., woodlands) across all nests, represented the greatest proportion of our prey data, and provided a more complete assessment of key prey species whose geographic range did not extend to other study regions (e.g., Rainbow/Red-collared Lorikeet, Laughing Kookaburra [Dacelo novaeguineae]). The total counts of each bird species were summed and used to rank their relative abundance and subsequent availability as potential prey. To determine whether Red Goshawks were selecting specific species or just capturing prey in proportion to its availability in the environment around the nests, we compared relative abundance of prey in the environment to the frequency and biomass in prey remains.

Statistical Analysis. We quantified the diet of the Red Goshawk using the minimum number of prey individuals identified in our data. This enabled dietary composition to be evaluated by the number of individuals by prey species, and the proportional biomass these prey species contributed to diet based on their average body mass. These two variables were used to compare potential differences in diet by taxonomic class, breeding habitat type (i.e., woodland and riparian), breeding phase (i.e., pre-breeding, incubation, nestling, fledgling), and region (i.e., state or territory). For the latter, we combined data from the Northern Territory and Western Australia given their similarities in prey species' assemblages and bio-geography. We used an analysis of variance to investigate if there were differences in the number of prey individuals and dietary biomass in different habitats and regions. To reveal the most significant prey species' contributions across each of these classes we assessed those contributing ≥5% of dietary biomass.

To determine whether prey was partitioned between the sexes based on size, we pursued two lines of inquiry. First, we assessed changes to mean prey body mass throughout the breeding season according to breeding phase. We expected mean prey body mass and range to be lowest and most restricted during incubation phase, which is the period when the male hunts exclusively. Conversely, we expected mean prey body mass and range to be highest and widest during the fledgling phase, which is the period when both adults hunt to provide for the fledged young. The pre-breeding and nestling phases may provide intermediate results as the female hunts occasionally during these times. Second, we compared body mass differences between the prey captures that could be allocated to a specific sex of the Red Goshawk. These analyses were used to elucidate patterns of male and female prey body mass given the direct and inferred evidence.

Finally, we investigated prey use based on prey availability, and the sex of fledglings produced from the nest. The relative abundance of the avian community identified as available prey was ranked by species and compared with their ranking as known prey based on their frequency and proportional biomass in the Red Goshawk diet. We used Kendall's rank correlation to determine whether the relative rank of a bird's abundance differed from its rank as prey species. We also investigated prey use (via t-test) based on fledgling sex to determine whether nests with female young are brought larger prey compared to nests with male young (given the much smaller size of the males). All data formatting and analyses were conducted in R version 4.3.3 (R Core Team 2021).

RESULTS

We compiled a dataset of 323 prey items representing a minimum of 253 prey individuals from Queensland (n = 161), the Northern Territory (n = 84), and Western Australia (n = 8). We collected data on prey individuals primarily via nest monitoring activities (n = 216), with information from community scientists (n = 27) and online image searches (n = 10) making up the rest of the dataset. The main type of prey item we used to identify individuals was skeletal remains (n = 147) followed by direct observation (n = 46), feathers (n = 36), and pellets (n = 24). In total, data informing the breeding diet of Red Goshawks across northern Australia comprised at least 35 individual nests across 30 breeding territories and 63 breeding events between the years 1997 and 2023 (nests can be used repeatedly).

Diet Composition by Region and Habitat. We found that Red Goshawks preyed exclusively on birds (29 different species, 253 individuals). The most numerous prey species were the Rainbow/Red-collared Lorikeets (n = 103), Blue-winged Kookabura (Dacelo leachii; n = 61), Sulphur-crested Cockatoo (Cacatua galerita, n = 19), and Laughing Kookaburra (n = 9; Fig. 2A). In total, we identified 192 prey individuals belonging to these four species, which represented 75.9% of the dataset.

Figure 2.

Prey species taken by Red Goshawks across northern Australia by (A) the total number of individuals and (B) proportional dietary biomass.

When assessed by mean body mass, we found some larger-bodied prey species provided more in dietary biomass than smaller, more frequently captured species. The species contributing at least 5% of dietary biomass were Blue-winged Kookaburra (33.2%), Sulphur-crested Cockatoo (24.2%), Rainbow/Red-collared Lorikeets (20.6%), and Laughing Kookaburra (5.3%; Fig. 2B). In total, these four species provided 83.3% of the Red Goshawk's dietary biomass.

Regionally, the four most important prey species made up 89.8% of total dietary biomass in Queensland compared to 77.6% of dietary biomass in the Northern Territory and Western Australia. The relative contribution of mutual prey species appeared reasonably consistent between states, with Blue-winged Kookaburra composing 31.9% of dietary biomass in Queensland and 35.8% in the Northern Territory and Western Australia. The two congener lorikeet species showed a similar pattern in their respective states, with Rainbow Lorikeet composing 19.3% of dietary biomass in Queensland and Red-collared Lorikeet composing 23.3% in the Northern Territory and Western Australia. Conversely, the Sulphur-crested Cockatoo appeared to contribute a higher proportion of dietary biomass in Queensland (30.6%; as compared to 11.4% in the Northern Territory and Western Australia) alongside Laughing Kookaburra (8.0%), which only occurs in Queensland. The Straw-necked Ibis (Threskiornis spinicollis) also accounted for 7.1% of dietary biomass in the Northern Territory and Western Australia despite not being reported as a prey species in Queensland.

Overall, the prey in woodland breeding habitats tended to be smaller in mass than prey in riparian breeding habitats (F_1,33 = 3.32, P = 0.07). Although there was a large difference in sample size between woodland (n = 221) and riparian (n = 29) breeding habitats, pairs breeding in woodland habitats appeared to rely on four key prey species dominated by parrots (i.e., lorikeets and cockatoos) and kookaburras, which together represented 88.5% of the overall dietary biomass (Fig. 3A). Conversely, riparian breeding sites appeared more diverse, with breeding pairs relying on seven key prey species that account for 92.8% of their overall dietary biomass (Fig. 3B). Prey taken by birds breeding in these riparian areas included some wetland and grassland species (e.g., Pied Heron [Egretta picata]; Cattle Egret [Bubulcus ibis]), which are likely responsible for the large average prey size in riparian habitats. However, these prey species are only represented by a few individuals and their relative contributions to the Red Goshawk's diet may be skewed, given the small sample sizes associated with this breeding habitat.

Figure 3.

The proportional biomass of prey species contributing at least 5% to Red Goshawk's diet across northern Australia in (A) woodland breeding habitats and (B) riparian breeding habitats. Note that prey records from woodland habitats were only obtained in Queensland and the Northern Territory, whereas prey records from riparian habitats were only obtained in Western Australia and the Northern Territory, despite both habitat types occurring across all states but in differing proportions.

Prey Size and Partitioning. The size distribution of prey taken by Red Goshawks ranged from 37.3 g (Forest Kingfisher [Todiramphus macleayii]) to 1346.7 g (Straw-necked Ibis) (Fig. 4). Mean prey size was 246.1 g, median prey size was 113.3 g, and species in the 100–200 g size class were the most common prey. This size class includes prey species such as Rainbow and Red-collared Lorikeets (both 113.3 g), Red-winged Parrot (Aprosmictus erythropterus; 136 g), and Bar-shouldered Dove (Geopelia humeralis; 128.4 g). There was a notable lack of prey species in the 200–300 g size class, but a larger number in the 300–400 g size class, which included Blue-winged Kookaburra (308 g) and Laughing Kookaburra (333.8 g). The Sulphur-crested Cockatoo, another important prey species, was within the 700–800 g size class.

Figure 4.

Distribution of Red Goshawk prey individuals by mean body mass arranged into 100-g bins and spanning the prey size range.

We found clear differences in prey size between the known prey captures of male vs. female Red Goshawks (Table 1). Females captured larger prey (x̄ = 365.5 g, n = 12) than males (x̄ = 151.4 g, n = 21; F_1,33 = 7.245, P = 0.01) based on 24 direct observations of predation events and nine photographic records. Although there was some overlap in prey body mass between the sexes (e.g., both sexes hunted lorikeet and kookaburra species), males appeared to prey primarily on Rainbow/Red-collared Lorikeets (113.3 g), whereas females were more likely to prey on Blue-winged Kookaburras (308 g).

Table 1.

Body mass of prey taken by male and female Red Goshawks, for directly observed predation events.

Both mean and median prey mass tended to increase as the breeding season progressed, although this was not significant (F_3,170 = 1.72, P = 0.16; Table 2). Lorikeets were the most common prey species for the pre-breeding, incubation, and nestling phases, whereas kookaburras were the most common prey during the fledgling phase. Although there is large RSD in Red Goshawks, for nests with only one nestling we did not find that the size of prey brought to a nest depended on the sex of that nestling. Mean prey size for nests with a single male was 221.7 g (n = 22), and for nests with a single female it was 285.3 g (n = 71; t(91) = -1.2271, P = 0.22).

Table 2.

Body mass of prey taken by Red Goshawks by breeding phase.

Prey Use vs. Availability. We identified 67 species as potential prey for Red Goshawks based on 39,246 individual records from eBird surveys conducted within 10 km of nest sites across Cape York Peninsula, Queensland. Potential prey species' ranks of abundance were not significantly correlated to their ranks as prey species based on frequency (T = 76, P = 0.17, Kendall's rank correlation s = 0.27) or biomass (T = 73, P = 0.27, Kendall's rank correlation s = 0.22). The most frequent prey species (Rainbow Lorikeet) was also the most locally abundant; however, the most important prey species by biomass (Blue-winged Kookaburra) was 14th in abundance out of 67 total species (Table 3). Of the 20 most abundant birds, 12 were confirmed as known prey, whereas eight species were identified as available prey but were apparently not taken by Red Goshawks on Cape York Peninsula (e.g., Little Friarbird [Philemon citreogularis], White-bellied Cuckooshrike [Coracina papuensis], Grey-crowned Babbler [Pomatostomus temporalis], Pale-headed Rosella [Platycercus adscitus], Pied Butcherbird [Cracticus nigrogularis], Spangled Drongo [Dicrurus bracteatus], Australian Magpie [Gymnorhina tibicen], and Silver-crowned Friarbird [Philemon argenticeps]). Moreover, some of the known prey species such as the Galah (Eolophus roseicapilla), Blue-faced Honeyeater (Entomyzon cyanotis), and Torresian Crow (Corvus orru) were preyed upon far less than their relative rank abundance would indicate. This lack of association between relative abundance of potential prey and prey selection suggests that Red Goshawks target specific prey species rather than taking prey in proportion to its availability. This selectivity is further evidenced by the high proportion of the diet made up of just a few key prey species.

Table 3.

The twenty most abundant woodland and forest birds found within 10 km of Red Goshawk nests across Cape York Peninsula, Queensland, and their respective ranks as prey species by frequency and biomass. NA signifies that the species was not identified as a prey item for goshawks in this study but was available to them as a prey item based on its local abundance and body mass.

DISCUSSION

Bird Specialist with Habitat Variation. We found that Red Goshawks consumed only birds and there was substantial variation in prey size. The body mass of the largest prey was 35 times larger than that of the smallest. This ability to access such a broad spectrum of prey appears to be a result of size-based prey partitioning between the size-dimorphic sexes: we found that males were more likely to take smaller prey species than females. We found differences in the average mass of prey taken in different habitats; larger prey was taken in riparian habitats, driven by the occasional capture of larger waterbird species (e.g., Cattle Egret, Straw-necked Ibis, Pied Heron), likely by females.

The reasons for these differences between habitats are not clear. Woodlands appear to support a more uniform prey base compared to the prey species diversity apparent in riparian areas. This difference could be expected, as these riparian areas appear to support higher bird diversity and abundance compared to the intervening dry savannah habitat (Aumann and Baker-Gabb 1991). We note that riparian breeding habitat was only recorded in the Northern Territory and Western Australia, where it is typically associated with more inland creeks and river systems. Red Goshawks require tall trees for nesting, and it may be that these riparian areas provide suitable nesting habitat, and that differences in prey species' contributions between habitats are simply the result of the more varied prey base supported in these riparian areas. Indeed, results from prior work by Aumann and Baker-Gabb (1991) differed across certain species groups with Columbiformes (i.e., pigeons and doves) and Passeriformes (i.e., songbirds) contributing as much or more in dietary biomass than Coraciiformes (i.e., kookaburras and kingfishers), which were much more significant prey items in our study. Although parrots still accounted for the greatest proportion of biomass across studies, differences in prey contributions may be artifacts of sampling with more riparian nests monitored by D. J. Baker-Gabb (pers. comm.), although the analysis was not stratified by habitat. Further work is needed to refine our understanding of the trophic niche occupied by pairs within these riparian systems as it may have important conservation implications given the inland extension to the breeding range these habitats enable.

Prey Use vs. Availability. The comparison of prey used by goshawks versus prey species available within the woodland and forest-bird community suggests prey selection by Red Goshawks. Tau rank coefficients range from 1 (strong relationship between rankings) to 0 (no relationship). Tau rank coefficients in our study were 0.27 and 0.22 (prey frequency and biomass, respectively) suggesting only a weak relationship between prey used versus relative abundance. More behavioral observation or site-based surveys could provide a clearer understanding of local bird abundance, complementing the eBird data (Stuber et al. 2022), and further revealing which species are most available as prey for Red Goshawks based on their foraging ecology.

Males appeared to focus on lorikeets (approximately 113 g), which they were observed flushing from the canopy, followed by rapid pursuit. In contrast, females focus on kookaburras (approximately 308 g), apparently using a sit-and-wait strategy (Aumann and Baker-Gabb 1991). How female Red Goshawks prey upon cockatoos, given the prey's potentially dangerous powerful bill is unclear, although a female goshawk has been observed to aerially attack a flock of Red-tailed Black-Cockatoos (Calyptorhynchus banksii) in flight (Aumann and Baker-Gabb 1991). Although these species are relatively common at a regional scale, there are other species of appropriate size that appear more abundant (e.g., Little Friarbird, Blue-faced Honeyeater, Bar-shouldered Dove) but receive little predation pressure. Behavioral aspects may contribute to this; for example, Peaceful Doves (Geopelia placida) and Bar-shouldered Doves feeding primarily on the ground could limit hunting opportunities for Red Goshawks, although goshawks will occasionally take ground-dwelling prey such as Chestnut-backed Buttonquail (Turnix castanotus; D. J. Baker-Gabb pers. comm.). Although the diet was dominated by few prey species, indicating selection for those species, the remaining diet was diverse, suggesting that Red Goshawks are capable of hunting a wide variety of species.

Prey Partitioning between the Sexes. The extreme size dimorphism between the sexes apparently facilitates a broad prey base to be used when needed, but also enables males to focus on lorikeets and females to focus on kookaburras and cockatoos as staples of their breeding diet. This difference seems to be reflected in the change in average mass of prey being brought to nests as the breeding cycle progressed as well as directly observed prey captures by each sex. We observed smaller prey size during incubation, and larger prey size later in the breeding cycle (i.e., fledgling phase), likely explained by the larger female contributing more to offspring provisioning later in the breeding cycle (Debus et al. 2015). This sex-based delineation in prey size is consistent with other size-dimorphic raptors such as the Eurasian Goshawk (Accipiter gentilis; Grønnesby and Nygård 2000), but it does not appear universal with studies of Peregrine Falcon (Falco peregrinus; Zuberogoitia et al. 2013) and Cooper's Hawk (Accipiter cooperii; Kennedy and Johnson 1986) showing no such partitioning of prey.

Females appear able to access prey unavailable to the male, including large-bodied and quite formidable species such as Sulphur-crested Cockatoo although some historical records suggest males can also prey upon these species (Barnard 1934). However, the steady supply of lorikeets brought to the nest by the male may fill a critical need in the growth cycle of nestlings, given that raptors develop most rapidly in the nest prior to fledging (Donázar and Ceballos 1989). Male fledglings also appear to develop at a faster rate than the larger females, achieving independence and dispersing from their natal territories sooner (C. MacColl unpubl. data). Given these differences in development, we had expected to see larger prey being provided to nests with female nestlings as compared to male nestlings; however, we saw no differences of statistical significance. This result may be due to constraints on the size of prey that males can provide, regardless of whether they are feeding a male or female nestling, or the protracted period in which females are required to remain sedentary at the nest, which limits their ability to hunt and provide larger prey to the young (Sonerud et al. 2014).

RSD Trends and Population Ecology. Numerous competing hypotheses attempt to explain the evolution of RSD in raptors (Selander 1966, Snyder and Wiley 1976, Mueller and Meyer 1985, Bildstein 1992, Schoenjahn et al. 2020) and the consistent morphological and ecological traits associated with it across taxa. For example, RSD is most pronounced in species that hunt relatively large and agile prey capable of escaping or mounting a defense against their predators (Krüger 2005). It is also associated with trophic niche partitioning whereby the sexes utilize distinct resources (Bauld et al. 2022). The extreme size dimorphism we observe in the Red Goshawk is consistent with these general trends. Red Goshawks prey on agile birds predominantly in the medium-large size range (e.g., 100–800 g), and which exhibit considerable escape (e.g., lorikeets) and defense (e.g., cockatoos) potential. Trophic niche partitioning is also clear, given the size-based prey partitioning between the sexes that we have demonstrated.

These traits are also predictors of the population ecology of raptors; species that hunt large and agile prey occur at lower population densities and display lower reproductive rates (Krüger 2000). Moreover, home range size is positively correlated with the proportion of avian prey in the diet (Peery 2000). Again, the diet and hunting methods used by Red Goshawks are consistent with these predictions. The species has low population densities and slow life history traits, and Red Goshawks probably utilize extensive home ranges to meet their avian dietary requirements (Debus and Czechura 1988, Aumann and Baker-Gabb 1991). This underlying ecology likely explains why the species is so seldom encountered (Backstrom et al. 2023) and has long been considered the rarest and most difficult to find of Australia's raptors (Slater 1978, Cupper and Cupper 1981).

Species' Decline. Red Goshawks are in precipitous decline, having been lost from 34% of their breeding range since 1978 (MacColl et al. 2023). The species' breeding range is now likely confined to northern Australia. Here we found its diet was specialized, with just a few key prey species representing a significant proportion of the breeding diet. However, it appears unlikely that the Red Goshawk's decline can be linked to commensurate declines in these preferred prey species across its former range. For example, studies have consistently shown that Rainbow Lorikeet and Sulphur-crested Cockatoo populations have increased throughout eastern Australia (Danson et al. 2005, Burgin and Saunders 2007, Campbell et al. 2022), but this increase may be attributed to these species having high urban adaptability (Catterall et al. 1997, Shukuroglou and McCarthy 2006, Campbell et al. 2022). However, comparison of urban, rural, and forested areas of greater Brisbane found consistent increasing prevalence of Rainbow Lorikeet, Sulphur-crested Cockatoo, and Little Corella (Cacatua sanguinea) in each environment, whereas Laughing Kookaburra tended to decline away from developed areas (Joyce et al. 2018).

Although it appears that prey loss cannot account for the large range contraction of the Red Goshawk, habitat changes may have reduced its ability to hunt effectively. Red Goshawks appear to prefer areas of intact vegetation and may exhibit sensitivity to cleared environments (C. MacColl unpubl. data.). Avoidance of these areas may have restricted their useable space over time as urban/suburban footprints have grown, making goshawks' persistence in these landscapes difficult particularly if they prefer forest cover for hunting.

How sensitive Red Goshawks are to habitat loss, degradation, and fragmentation is uncertain but insights from the Tiwi Islands (Northern Territory, Australia) suggest that >25% of habitat removed from within 4 km of a nest negatively impacts reproductive rate (Queensland Department of Environment and Resource Management 2012). Whether this threshold represents a critical point in the reduction of prey abundance or useable habitat space is uncertain. However, both possibilities point to environmental change as a threatening process across the species' former range in eastern Australia, which has been subject to extensive habitat loss (Fraser et al. 2019, Simmonds et al. 2019, Beyer et al. 2020).

Conclusions. Red Goshawks are specialized avian predators targeting specific prey species across the tropical savannas of northern Australia. Parrot and kookaburra species are staples of the diet, but more information is needed to better understand the trophic niche that Red Goshawks occupy in riparian breeding habitats that extend inland. There is a clear prey size delineation between the two sexes, which enables prey to be accessed across a significant size range. Red Goshawks likely require extensive home ranges to satisfy their bird-eating habits, and as a result, occur at very low population densities across the landscape. These ecological requirements may leave them susceptible to environmental change processes such as habitat loss and degradation, particularly if they require intact forest cover for hunting. The prey species they seem to prefer appear to have remained abundant in the goshawk's former range, suggesting that the loss of these prey populations cannot alone account for the species' significant recent decline.