Higher extinction rates of dasyurids on Australo-Papuan continental shelf islands and the zoogeography of New Guinea mammals


Correspondence: Tyrone H. Lavery, School of Agriculture and Food Sciences, The University of Queensland, Gatton, Qld, 4343, Australia.

E-mail: tyrone.lavery@uq.edu.au



Contemporary patterns of mammalian species richness on islands are influenced by well-understood biogeographical variables. Whether or not mammalian orders differ in their rates of persistence, however, remains uncertain. Our aims were to assess the persistence of four mammalian orders on Australo-Papuan continental shelf islands in relation to the faunas within adjacent zoogeographic provinces. We also aimed to define New Guinea's mammalian zoogeographic provinces quantitatively.


New Guinea and 274 Australo-Papuan continental shelf islands.


We compiled 4194 distributional records for 264 of New Guinea's native mammals. Records were allocated to existing mapped bioregions. We used cluster analysis to allocate bioregions to zoogeographic provinces. Using generalized linear models, we determined the persistence of insular mammals as proportions of the species present within adjacent zoogeographic provinces. Persistence rates were calculated for four major orders (Dasyuromorphia, Diprotodontia, Peramelemorphia and Rodentia).


The classification dendrogram grouped New Guinea's bioregions into three areas corresponding to the Oceanic, Tumbanan and Austral provinces. In all but two zoogeographic provinces, the proportions of Dasyuromorphia persisting on islands were lower than other orders. Overall, species of Dasyuromorphia were much less likely to persist on Australo-Papuan continental shelf islands.

Main conclusions

Unlike the other orders considered, dasyuromorphs are carnivorous and insectivorous and require large home ranges relative to body size. We suggest that the resulting low population densities might expose species in this order to higher rates of extinction on islands. Translocations of threatened mammals to predator-free islands are common, but our results suggest that insurance populations of threatened dasyurids on small islands may be less secure than translocations of other taxa. Our results support calls for insurance populations of the rapidly declining dasyurid, the Tasmanian devil (Sarcophilus harrisii), to be established on mainland Australia rather than on islands alone.


Many of the patterns observed in island biogeography stem from heritable traits that influence the ability of species to immigrate to and/or persist on islands (e.g. McNab, 1994; Duncan & Forsyth, 2006). As such, the phylogenetic structures of insular species assemblages often exhibit some predictable regularity. Whether study archipelagos consist of continental shelf islands or oceanic islands also has a significant bearing on the processes that are involved in shaping these phylogenetic structures (Cardillo et al., 2008).

Continental shelf islands are those that were formerly joined to larger landmasses and have since been isolated by rising sea levels. They inherited a mainland fauna and following isolation became extinction limited, as species were lost with diminishing island size and habitat area (Lawlor, 1986). In most instances, the timeframes involved are too short for immigration to influence non-volant assemblages. Rather, it is the ability of inherited species to persist following separation that determines a continental shelf island's species richness (Lawlor, 1986). This results in strong relationships between richness and island area, whereby larger islands support more species (e.g. Woinarski et al., 1999; Okie & Brown, 2009). The suite of species present on a continental shelf island at the time of separation is determined partly by the mainland source pool. These are biologically or geographically distinct areas with common characteristics such as climate, plant and animal communities, landforms and geology. The basic unit commonly used for source pools is the bioregion; however, for the purposes of biogeography, bioregions are often grouped into broader, taxon-specific zoogeographic provinces (e.g. Burbidge et al., 2008). Identification of taxon-specific zoogeographic provinces is valuable for landscape scale ecology, and conservation and management planning (Whittaker et al., 2005).

Australia and New Guinea are surrounded by more than 250 continental shelf islands that support a varied mammalian fauna (Abbott & Burbidge, 1995; Flannery, 1995a). Australia has the world's worst rate of mammal extinction. Twenty-four species have become extinct or are possibly extinct since European colonization (Fisher et al., 2003; Johnson, 2006; IUCN, 2011) and this record would be far worse had not nine species extinct on the Australian mainland clung to survival on offshore islands (Burbidge et al., 1997). However, Flannery (1995a) noted that the diversities of marsupial lineages encountered on large continental shelf islands around New Guinea are often well below that expected for island size. He speculated that marsupials have low rates of long-term persistence on islands. This idea has not been previously quantified or empirically tested.

In a global review of insular mammal assemblages, Cardillo et al. (2008) demonstrated differing rates of extinction and persistence in certain taxonomic guilds. Phylogenetically biased patterns of extinction have also been identified within marsupials on the Australian mainland (Johnson et al., 2002). Confirmation of insular extinction biases in Australo-Papuan taxa is of great relevance to mammalian conservation in the region. The survival of many Australian species depends upon refugial and translocated populations on islands and fenced peninsulas where they are isolated from mainland threats (e.g. Short & Turner, 2000; Moro, 2003; Rankmore et al., 2008). For example, population size of the Tasmanian devil, Sarcophilus harrisii, has declined > 60% in 10 years because of a fatal infectious cancer (devil facial tumour disease), and insurance populations of uninfected devils isolated from wild populations are crucial for future reintroduction, should the species become extinct in the wild (IUCN, 2011).

Given the importance of Australian continental shelf islands to mammal conservation, past research has focused on how physical island characteristics and taxonomic groupings relate to species richness (Burbidge et al., 1997; Burbidge, 1999; Burbidge & Manly, 2002). However, these authors examined post-colonial assemblages, to investigate contemporary impacts rather than underlying patterns in the pre-colonial fauna. Furthermore, Burbidge et al. (1997) assessed the persistence of families Dasyuridae, Macropodidae and Potoroidae but not other major Australo-Papuan families such as Muridae, Peramelidae and Phalangeridae.

Our aim was to determine whether there are differences in the rates of persistence of non-volant orders of mammal on Australo-Papuan islands. We assessed persistence rates of the Dasyuromorphia (carnivorous marsupials), Peramelemorphia (bandicoots), Diprotodontia (possums, kangaroos, wombats, koala), and Rodentia (murid rodents) on Australo-Papuan islands as proportions of faunas encountered within adjacent mainland zoogeographic provinces. Mammalian zoogeographic provinces have been empirically defined for Australia (Burbidge et al., 2008). However, the provinces of New Guinea have only been described qualitatively using expert knowledge (Flannery, 1995b; Helgen, 2007a), and no quantitative definitions are available. Furthermore, prior descriptions have not provided visual representation of provinces. We also aimed to define a quantitative classification of mammalian zoogeographic provinces for New Guinea.

Materials and Methods

New Guinea zoogeography

For the identification of provincial source pools, we required a classification of New Guinea's zoogeographic provinces. We used existing mapping of terrestrial bioregions (Wikramanayake et al., 2001) as our fundamental units for delineating provinces. Terrestrial bioregions are units that reflect substantial differences in climate, vegetation and geomorphology. In New Guinea they consist of the central range subalpine grasslands; central range montane rain forests; Huon Peninsula montane rain forests; northern New Guinea lowland rain and freshwater swamp forests; northern New Guinea montane rain forests; south-eastern Papuan rain forests; southern New Guinea freshwater swamp forests; southern New Guinea lowland rain forests; Vogelkop montane rain forests; Vogelkop-Aru lowland rain forests; New Guinea mangroves; and Trans-Fly savanna and grasslands. The New Guinea mangroves bioregion comprised many small units on the New Guinea coastline and lacked sufficient mammal records to be successfully incorporated into the analysis. We thus disregarded the New Guinea mangroves bioregion as an independent unit and instead merged each individual polygon with its adjacent bioregion.

We identified the presence and absence of all native mammals within these bioregions using data from Flannery (1995b), Hitchcock (1997, 1998), Bonoccorso (1998), Flannery & Groves (1998), Aplin et al. (1999, 2010), Helgen & Flannery (1999), Van Dyck (2002), Helgen (2005a,b, 2007a), Woolley (2005), Helgen et al. (2008, 2010, 2011), Musser et al. (2008), Helgen & Helgen (2009) and Musser & Lunde (2009). All accounts in these sources are backed by preserved specimens held in museums or research institutions. We used the statistical program patn 3.0.3 (Belbin, 1995) to group similar bioregions into zoogeographic provinces according to their mammalian faunal compositions. All species were weighted equally within the presence–absence matrix for the 11 bioregions. The matrix was converted by the Czekanowski dissimilarity index to an association matrix, and the bioregions were classified into zoogeographic provinces using a flexible unweighted pair-group method using arithmetic averages (UPGMA) with a beta value of –0.1 (Sneath & Sokal, 1973; Belbin, 1995; Burbidge et al., 2008). To test the suitability of our boundaries we used the statistical procedure analysis of similarity (ANOSIM; Clarke & Green, 1988) to compare differences in the F-ratio between our identified provinces with 10,000 iterations in which bioregions were randomly reallocated between provinces.

Source pools and islands

We assessed the persistence of the four orders on islands by calculating the proportions of source species that have persisted on each island. Tasmania and New Guinea were treated as part of mainland source pools rather than study islands. Continental shelf islands were defined as those connected to the Sahul landmass during the most recent glacial maxima (i.e. if intervening ocean depths between the island and either New Guinea or Australia did not exceed 130 m) (Lambeck, 2002). Intervening ocean depths for Australian islands were available from Abbott & Burbidge (1995) and Burbidge & Manly (2002) and were determined for New Guinean islands using nautical charts (Australian Bureau of Mineral Resources Geology & Geophysics & Papua New Guinea Geological Survey, 1986) and Topex global bathymetry data (Becker et al., 2009).

In this study, the source pool is defined as the fauna of the adjacent zoogeographic province. We partitioned New Guinea and Australia into provinces according to the distributions of all native mammals. Australian bioregion mapping prepared by Commonwealth of Australia (2005) and the presence–absence matrix of Burbidge et al. (2008) were combined with the bioregion mapping of Wikramanayake et al. (2001) and our own New Guinea presence–absence matrix. We used pre-colonial distributions, disregarding local extinctions that followed European arrival.

We used the statistical program patn (Czekanowski dissimilarity index with a flexible UPGMA with a beta value of –0.1) to group similar bioregions from both Australia and New Guinea. From the dendrogram, we identified a level of grouping that was suitable for use as our faunal source pool areas.

Persistence modelling for species in each order

From the literature, we compiled a second presence–absence matrix of non-volant terrestrial mammals on Australo-Papuan continental shelf islands (Flannery, 1995a; Woinarski et al., 1999; Burbidge & Manly, 2002; Gibson & McKenzie, 2012; Lavery et al., 2012). Records of native mammals deliberately translocated to new islands were excluded from the matrix. Data for monotremes (Order Monotremata) and marsupial moles (Order Notoryctemorphia) were sparse and were therefore removed from the matrix and not further analysed. Each remaining species was classified as belonging to the orders Dasyuromorphia (28 species), Diprotodontia (51 species), Peramelemorphia (11 species) or Rodentia (42 species), giving a total of 765 records of 132 species on 274 islands.

Our ‘potential’ pool consisted of all species in a mammalian order found within the zoogeographic province closest to an island. We assumed that the proportion of species present on the islands at the time of separation by rising sea level did not differ between orders. This model does not incorporate the potential effects of interspecific competition or occupation by indigenous people. Previous analyses of post-colonial distributions found no improvement in predicted species number by including these variables (Burbidge et al., 1997). Island area (ha) for Australian islands was provided by Burbidge & Manly (2002) and determined for New Guinea's islands using ArcMap 9 (ESRI Inc., San Diego, CA). Areas were transformed to a logarithmic scale before analyses. The data were analysed with generalized linear modelling using the procedure GENMOD in sas 9.3 (SAS Institute Inc., Cary, NC).

We used the extreme value function (Williams, 1995) in our generalized linear models. This function employs a complementary log–log distribution to estimate intercept (k), slope (y) and maximum likelihood (ML). This model has been used successfully to examine species richness on Australian islands in relation to mainland source pools (Burbidge et al., 1997). We modelled proportions of species (SP) present on an island relative to a source pool as a binomial response variable on island area (A):

display math

The regression intercept and slope behave as follows: a change in the value of intercept shifts the curve to the left or right and a change in the slope parameter results in a steeper or slower increase in SP with island area.

Our model tested for interactions between mammalian order and province; island area and province; and island area and mammalian order. Regression slopes for island area did not differ between mammalian orders (χ23 = 2.83, = 0.419) or between zoogeographic provinces (χ26 = 11.21, = 0.082). Only an interaction between mammalian order and province was evident (χ218 = 51.87, < 0.001). We thus estimated a common slope for all orders across all provinces and estimated individual intercepts for each order within each province. We used orthogonal Helmert contrasts to assess variation in intercepts derived from our generalized linear models. The first contrast tested for differences between Dasyuromorphia and remaining orders; the second, Peramelemorphia against Diprotodontia and Rodentia combined; and the third, Diprotodontia against Rodentia. Contrasts were calculated using intercepts derived from individual provinces and the combined Australo-Papuan dataset.

Diet, home range and body mass

Broad categories of diet (insectivore/carnivore, omnivore, herbivore) differ between orders so we did not consider this trait separately from order. The Dasyuromorphia is an entirely carnivorous and insectivorous lineage; species of Diprotodontia are largely herbivorous (over 65% of species) and the remainder are omnivorous species whose diet includes fungi, seeds, fruit, nectar and invertebrates; Peramelemorphia are also omnivorous; and the Rodentia exhibit a diverse array of dietary preferences including granivorous, frugivorous, omnivorous and carnivorous species.

We tested whether (log-transformed) absolute home range size varied according to order and (log-transformed) body mass, and whether there was an interaction between order and body mass affecting home range size using an ANOVA from R 2.10.0 (R Development Core Team, 2009). We obtained published short-term home range estimates of 77 Australian mammal species (based on radio telemetry) from Diana Fisher's marsupial ecology database (Fisher et al., 2001), the PanTheria database (Jones et al., 2009), and additional primary literature in the case of rodents. We further examined the relationship between body mass and absolute home range size by performing a regression between these two variables and calculating residual values for the 77 species from this regression line.


New Guinea zoogeography

The reviewed literature provided 4194 distributional records for 264 of New Guinea's native mammal species. Partitioning of the dendrogram broadly identified a three-group zoogeographic classification of New Guinea's mammals (Fig. 1). The southern New Guinea lowland rain forests, southern New Guinea freshwater swamp forests, and Trans-Fly savanna and grasslands formed one province across southern New Guinea. Vogelkop montane rain forests and Vogelkop-Aru lowland rain forests formed a second province, and the remaining bioregions (northern New Guinea lowland rain and freshwater swamp forests, central range subalpine grasslands, central range montane rain forests, northern New Guinea montane rain forests, Huon Peninsula montane rain forests, and south-eastern Papuan rain forests) formed the third.

Figure 1.

Distribution of the zoogeographic provinces of New Guinea. Boundaries of New Guinea's terrestrial bioregions were provided by Wikramanayake et al. (2001) and distributional records for all native mammals were used to generate a species presence–absence matrix according to these. Cluster analysis using flexible UPGMA (unweighted pair-group method using arithmetic averages) with a beta value of –0.1 classified bioregions into three zoogeographic provinces. Provinces consist of: the Oceanic province shown in shades of blue; the Tumbanan in shades of orange/brown; and the Austral in shades of green. Mollweide projection.

Source pools

A noticeable separation in the classification dendrogram grouped Australo-Papuan bioregions into nine zoogeographic provinces (Fig. 2). For Australia, the same seven-group level provinces were apparent as identified by Burbidge et al. (2008). These were the Arid Eyrean, East Torresian, West Torresian (here termed the Arafuran), Mainland Bassian, Tasmanian, South-west semi-arid, and South-east semi-arid. At this level New Guinea was partitioned into the Oceanic-Tumbanan and Austral provinces.

Figure 2.

Australo-Papuan zoogeographic provinces and partition structure of the dendrogram derived by classifying bioregions according to their mammalian species composition. Numbers 1–11 correspond to bioregions presented by Wikramanayake et al. (2001); numbers 12–30 correspond to the 19-group classification of Australia's bioregions presented by Burbidge et al. (2008). Our defined zoogeographic provinces are: Oceanic-Tumbanan in shades of orange/brown; Austral-East Torresian in shades of green; Arafuran in shades of grey; South-west semi-arid in shades of purple; South-east semi-arid in shades of red; Arid Eyrean in shades of yellow; and Bassian in shades of blue. Mollweide projection.

For the purposes of comparing the persistence of four mammal orders on islands, we determined that four provinces required further amalgamation for our definition of source pools. First, we amalgamated the Austral and East Torresian provinces to form the Austral-East Torresian province. Although they did not cluster together in our results, many authors have remarked upon the similarities in mammal assemblages present within these two provinces (Schodde & Calaby, 1972; Ziegler, 1982; Flannery, 1995b; Helgen, 2007a). Deficiencies in the faunal inventory for the Austral province compared to the East Torresian are likely to have contributed greatly to their disparity in the dendrogram. Furthermore, a large number of islands occur in the Torres Strait and the late Pleistocene faunas on these islands were inherited from both adjacent provinces (Schodde & Calaby, 1972; Flannery, 1995b; Lavery et al., 2012). We also amalgamated the Tasmanian and Mainland Bassian provinces to form the Bassian province. Once again, a large number of islands occur in Bass Strait and the late Pleistocene faunas of these islands are derived from both the Tasmanian and Mainland Bassian provinces (Hope, 1973). This finalized our zoogeographic provinces into the seven-group classification presented in Fig. 2.

Species persistence modelling

The island dataset comprised 274 continental shelf islands and 765 records of 132 native terrestrial mammals from the four orders. In all provinces, the intercept for Dasyuromorphia was the lowest of the four mammalian orders (Fig. 3a–g). Lower intercepts indicate a shift in the curve to the right and correspond to a greater island area requirement for any arbitrary proportion of the species pool to be present. When contrasted against intercepts for all other orders, those for Dasyuromorphia proved significantly lower in all but the Bassian and Oceanic-Tumbanan provinces (Table 1). Our Helmert contrasts of intercepts derived from the combined Australo-Papuan dataset demonstrated that overall, the Dasyuromorphia require a greater island area for any arbitrary proportion of species to persist than do other orders (χ23 = 56.94, < 0.01). In the Austral-East Torresian and Oceanic-Tumbanan provinces, our contrasts also revealed that the order Diprotodontia requires greater island areas for any arbitrary proportion of species to persist than does Rodentia (Table 1).

Table 1. Chi-square and probability values calculated from orthogonal Helmert contrasts of intercepts from our generalized linear models. We calculated three contrasts for the combined Australo-Papuan dataset and for each of our zoogeographic provinces. The first contrast compared Dasyuromorphia against remaining orders; the second, Peramelemorphia against Diprotodontia and Rodentia combined; and the third, Diprotodontia against Rodentia. A P-value ≤ 0.05 indicates a significant difference in the island area required to support arbitrary proportions of mainland source pool faunas. Significant contrasts are shown in bold (d.f. = 1)
Combined Australo-Papuan Dasyuromorphia vs. Others 56.94 < 0.01
Peramelemorphia vs. Diprotodontia & Rodentia0.110.73
Diprotodontia vs. Rodentia0.030.86
BassianDasyuromorphia vs. Others1.690.19
Peramelemorphia vs. Diprotodontia & Rodentia0.510.47
Diprotodontia vs. Rodentia1.560.21
South-east semi-arid Dasyuromorphia vs. Others 6.50 0.01
Peramelemorphia vs. Diprotodontia & Rodentia0.210.64
Diprotodontia vs. Rodentia1.280.25
South-west semi-arid Dasyuromorphia vs. Others 3.83 0.05
Peramelemorphia vs. Diprotodontia & Rodentia0.380.54
Diprotodontia vs. Rodentia0.100.74
Arid Eyrean Dasyuromorphia vs. Others 24.24 < 0.01
Peramelemorphia vs. Diprotodontia & Rodentia1.400.23
Diprotodontia vs. Rodentia0.570.45
Arafuran Dasyuromorphia vs. Others 18.63 < 0.01
Peramelemorphia vs. Diprotodontia & Rodentia0.100.74
Diprotodontia vs. Rodentia0.470.49
Austral-East Torresian Dasyuromorphia vs. Others 23.08 < 0.01
Peramelemorphia vs. Diprotodontia & Rodentia2.150.14
Diprotodontia vs. Rodentia 8.44 < 0.01
Oceanic-TumbananDasyuromorphia vs. Others3.130.07
Peramelemorphia vs. Diprotodontia & Rodentia2.290.12
Diprotodontia vs. Rodentia 15.14 < 0.01
Figure 3.

Estimated proportions of source pool species (SP) present on an island plotted as a binomial response variable against island areas (A): SP = 1 – exp(–exp(k + y × log(A))). The graphs (in a–g) represent the seven Australo-Papuan zoogeographic provinces and curves within these correspond to the four taxonomic mammalian orders. Slope (y) is consistent between orders and provinces, curves shifted to the right reflect lower intercept values (k) and greater island areas required to support given proportions of source pool species.

Order, body mass and home range

Mean home range size differed marginally between mammalian orders (Dasyuromorphia, mean ± SE = 136.18 ± 100.63 ha; Diprotodontia, 48.50 ± 15.48 ha; Peramelemorphia, 4.02 ± 1.10 ha; Rodentia, 1.17 ± 0.47 ha), and was strongly related to body mass (= 44.68, < 0.001). The interaction between mammalian order and body mass was significant (= 6.16, < 0.001), indicating that the effect of body mass on home range differs between orders. Visual examination of the mean residual values of home range after body mass had been accounted for showed that dasyurids require the largest home range areas for their body size (Fig. 4).

Figure 4.

Associations between home range size and mean adult body weight for four orders of Australo-Papuan mammal species. The figure is based on residual values calculated from a regression of absolute home range size (ha) and mean adult body weight (g) for 77 species of mammal available from the literature. Bars represent the range of residual values calculated from the regression for each mammalian order (= number of species) and lines within these represent the mean residual values.


New Guinea zoogeography

Studying the zoogeography of New Guinea's mammalian fauna is a difficult task. The taxonomy of most groups is uncertain, and distributional data are rudimentary (Helgen, 2007b). The distributions of insectivorous bats are especially poorly known, thus our boundaries are likely to be biased towards monotremes, marsupials, fruit bats and rodents. Furthermore, there is evidence to suggest that the distributions of many larger-bodied species have been greatly altered through human hunting pressure (Aplin et al., 1999; Helgen, 2007a). Where possible, we have included data from the sub-fossil and fossil record; however, these data are sparse and our zoogeographic conclusions should thus be considered a better reflection of modern distributions. Appendix S1 (in Supporting Information) shows point records used to generate bioregion presence–absence data for this study and highlights large gaps, particularly in West Papua, where mammal inventories remain unavailable.

The zoogeographic provinces identified here for New Guinea correspond closely with those suggested by other authors (Schodde & Calaby, 1972; Flannery, 1995b; Helgen, 2007a). The Tumbanan province was a term first coined by Schodde & Calaby (1972): there it was depicted as consisting of New Guinea's central mountain range, high-elevation areas of the Vogelkop and Huon peninsulas, and the Bewani/Torrecelli Range. The authors also identified a separate area of the Trans-Fly Plain comprising biotas of Australian woodland and sclerophyllous vegetation formations. Flannery (1995b) provided a more comprehensive description tailored to the mammalian fauna. In addition to his modified version of the Tumbanan province, he provided detailed descriptions of what he termed the Austral and Oceanic provinces. Province boundaries identified by Flannery (1995b) largely conform to those presented here with three main differences. First, Flannery (1995b) placed the Cyclops and Bewani/Torrecelli ranges of the north coast within the Oceanic province whereas surrounding northern lowlands were described as being part of the Tumbanan province. In the present study both the north coast ranges and surrounding lowlands fall within the Tumbanan province. Second, Flannery (1995b) described the interface between the Austral and Tumbanan provinces as the border where woodland with gallery rain forest meets denser rain forest. Our analyses have instead defined this interface as the border between southern lowland rain forest and central range montane rain forest. Third, in the present study, the Austral province ceases on the western side of the Gulf of Papua and does not continue along the margins of the Papuan Peninsula beyond Port Moresby as far as Popondetta. This is largely an artefact of the bioregional mapping of Wikramanayake et al. (2001) that was used for the analysis; much of the Papuan Peninsula was mapped as a discrete unit.

Detailed bioregion mapping and continued advances in the taxonomy of New Guinea's mammals have allowed us to delineate zoogeographic provinces objectively. However, it is evident that the bioregion mapping produced by Wikramanayake et al. (2001) is produced at a coarser scale than that used for similar exercises in other regions (e.g. Commonwealth of Australia, 2005). With more detailed mapping and advanced taxonomic and distributional data, province boundaries might shift marginally from those presented here.

Patterns of persistence: orders and zoogeographic provinces

Our comparisons of independent mammalian orders do not support the contention that marsupials are generally less likely to persist on Australo-Papuan continental shelf islands than are other taxa (rodents). Instead, our results show that the continental shelf islands of Australia and New Guinea are depauperate in Dasyuromorphia compared with the other mammalian orders. Burbidge et al. (1997) alluded to similar patterns in the post-colonial distribution of Australia's insular mammals, with islands in two of their three defined provinces (northern Australia and south-western Australia) being relatively depauperate in carnivorous mammals. As their research examined only post-colonial distributions, variables such as the introduction of exotic species and habitat destruction obscured any phylogenetic differences in persistence. Here we demonstrate that the Dasyuromorphia were relatively depauperate on Australo-Papuan islands prior to European arrival. We find that proportions of all orders are comparable only within the Bassian province (as per Burbidge et al., 1997).

Our results suggest that dasyurids have disappeared disproportionately from islands that formed after the Last Glacial Maximum. This contrasts with both the earlier Pleistocene mammal extinctions on the mainlands (which were primarily large, slow-breeding herbivores including Diprotodontia) (Johnson, 2006); and extinctions in the 19th and 20th centuries on mainland Australia. In these recent extinctions, species of Dasyuridae (carnivorous and insectivorous marsupials) are least affected, and extinction rates of Potoroidae (bettongs and rat kangaroos), Peramelidae (bandicoots) and Hydromyinae (water rats) were higher than expected under randomly generated null models (Cardillo & Bromham, 2001).

A second, less-pronounced pattern was revealed by our analyses. Diprotodontia were less likely to persist than Rodentia in two zoogeographic provinces. Both provinces (Austral-East Torresian and Oceanic-Tumbanan) are situated in the north of the study area and wholly or partly include areas of New Guinea.

Which traits explain taxonomic differences in persistence rates?

The four mammalian orders differ in trophic group, body size, metabolic rate, life history, and home range size (Van Dyck & Strahan, 2008). Species of Diprotodontia for example have a far greater mean body size (mean ± SE = 5443 ± 694 g) than Australo-Papuan species of Rodentia (181 ± 32 g) (Jones et al., 2009). Such disparity in body size alone could explain our observed differences in the persistence of species from these two orders in northern zoogeographic provinces. McNab (1994) demonstrated that resource limitation on continental shelf islands impacts larger-bodied mammals, often resulting in extinction. Burbidge et al. (1997) suggested that the carnivorous and insectivorous diets of dasyurids were the primary cause for their rarity on islands, because they need proportionally more energy for their body mass than herbivores or omnivores of the same size. It is difficult to disentangle the roles of trophic level, home range and body size, as these traits are correlated (Fisher et al., 2001). In a comparison of home range and trophic level using 279 mammal species worldwide, Kelt & Van Vuren (2001) demonstrated no difference in slopes for the scaling of home range and body size between herbivores, omnivores and carnivores (including insectivores). However, a higher intercept for carnivores and insectivores reflects lower densities of food for carnivorous groups, and thus a requirement for them to have greater home ranges than omnivores or herbivores of the same mass. Large home range requirements increase the extinction rate of carnivores and insectivores on the smallest islands (Heaney, 1984), because these species will have reduced densities.

Although the Dasyuromorphia require larger home ranges than equivalent sized Peramelemorphia, Diprotodontia and Rodentia, their mean body size is smaller than other marsupial orders considered here (Dasyuromorphia 796 ± 507 g; Diprotodontia 5443 ± 694 g; Peramelemorphia 970 ± 230 g) (Jones et al., 2009). The order Dasyuromorphia also has the lowest basal metabolic rate of all orders (McNab, 2008). McNab (1994) proposed that smaller size and lower rates of metabolism are favoured on islands because food is limited. Some small dasyurids do persist on islands and in some cases, home ranges appear adaptable to island pressures, and body sizes of island populations are smaller than their mainland counterparts, e.g. Antechinus minimus (Sale et al., 2009) and Parantechinus apicalis (Mills et al., 2004).

It is also possible that other factors, such as inflexible reproductive strategies, render dasyurid species unable to respond rapidly to favourable conditions, and short lifespans hinder their ability to endure unfavourable conditions (Morris et al., 2008). We were unable to test this because dasyurids possess unique reproductive strategies that cannot be compared across orders. In addition, it is possible that microclimate and vegetation changes that occur when continental shelf islands are created might have a disproportionately negative effect on dasyurids in comparison with other orders, but data to test this are unavailable.

Persistence of threatened dasyurids on islands and conservation implications

Several threatened species of large dasyurid have been found in post-glacial sub-fossil deposits on Bass Strait islands (between Tasmania and mainland Australia) (Hope, 1973). The Tasmanian devil (Sarcophilus harrisii, 8000 g) is known from sub-fossil deposits on Flinders Island (c. 134,000 ha). Spotted-tailed quolls (Dasyurus maculatus, 5000 g) occurred in sub-fossil deposits on Flinders Island, Deal Island (1722 ha) and Cape Barren Island (46,220 ha) and Eastern quolls (Dasyurus viverrinus, 1500 g) were also present in deposits on Flinders Island. Thus, in most cases these species did not persist on islands following isolation by rising sea levels that commenced c. 11,500 years ago (Hope, 1973). Such cases demonstrate that dasyurids were not simply more sparsely distributed in the landscape and less likely to be present in lowland areas that became islands. Instead, they support the conclusion of our biogeographical analysis that dasyurids are less likely to persist on continental shelf islands. They also imply that translocated populations of large dasyurids may be less secure on islands than translocations of species from other orders.


We suggest that diet is a key element influencing the low persistence of the order Dasyuromorphia on continental shelf islands. Carnivorous and insectivorous diets increase the foraging area requirement of a species, resulting in lower densities and smaller population sizes on islands than for herbivores or omnivores.

Threatened dasyurids translocated to islands to create insurance populations have shown great short-term success, for example northern quolls (Rankmore et al., 2008) and dibblers (Moro, 2003). However, our results suggest lower long-term viability of insular populations of dasyurids. Future species recovery planning that includes the establishment of island refuges should include a taxon-specific assessment of rates of persistence on islands of different sizes within applicable zoogeographic provinces.

Multiple ex situ insurance populations of disease-free Tasmanian devils are urgently needed. Devil facial tumour disease will potentially cause extinction of the species on mainland Tasmania and the disease has already reached most parts of the state (Hamede et al., 2011). Populations on islands can be controlled and more easily monitored than on the mainland. However, this species is a large, carnivorous dasyurid that does not appear to persist in restricted areas in the long-term. We argue that translocation sites should include mainland Australia, where top dasyurid predators are also likely to have conservation benefits for other species (Johnson et al., 2007; Ritchie et al., 2012).


We thank Andrew Burbidge, Bryan Manly, Matthew Williams and Ian Abbott for access to raw data on the distributions of mammals within Australian mainland bioregions and continental shelf islands. Allan Lisle provided advice for statistical analyses. Tyrone Lavery was supported by an Australian Postgraduate Award. Diana Fisher was supported by an Australian Research Council Fellowship (FTll0100191).


Tyrone Lavery is a PhD candidate at the University of Queensland. His research focuses on biogeography and systematics of mammals from northern Melanesia and involves a large field component conducting baseline surveys in the region.

The research team is focused on biogeography, systematics and conservation of Australo-Papuan mammals.

Author contributions: T.H.L. conceived the idea for the study, collated and analysed data and wrote the manuscript; D.O.F. provided body size and home range data and analysis and edited/wrote the manuscript; T.F.F. provided guidance regarding the concept of the study; L.K.-P.L. provided guidance regarding the concept of the study and edited the manuscript.