Conservation Genetics

, Volume 15, Issue 5, pp 1085–1094

Persistence of a potentially rare mammalian genus (Wyulda) provides evidence for areas of evolutionary refugia within the Kimberley, Australia


    • Research School of BiologyAustralian National University
    • Australian Museum Research InstituteAustralian Museum
  • Dan Rosauer
    • Research School of BiologyAustralian National University
  • J. Sean Doody
    • School of Environmental and Life SciencesThe University of Newcastle
  • Myfanwy J. Webb
  • Mark D. B. Eldridge
    • Australian Museum Research InstituteAustralian Museum
Research Article

DOI: 10.1007/s10592-014-0601-4

Cite this article as:
Potter, S., Rosauer, D., Doody, J.S. et al. Conserv Genet (2014) 15: 1085. doi:10.1007/s10592-014-0601-4


Understanding the evolutionary and ecological processes that have shaped current patterns of biodiversity is crucial in the planning and implementation of broad scale conservation management. The temporal and spatial pattern of diversity across a landscape can help identify areas that have acted as climatically stable refugia historically, or do so currently. This has important implications for conservation efforts that try to maximise the evolutionary potential of species, as well as maintaining existing biodiversity. Northern Australia has recently reported catastrophic species decline, particularly in mammals, due to a series of threats. Here we apply an integrative approach utilising molecular analyses and spatial modelling to determine whether disjunct populations of a potentially rare, endemic mammal, the scaly-tailed possum (Wyulda squamicaudata) exhibit differentiation associated with biogeographic barriers or a recent decline. Significant but low genetic differentiation between the east and west Kimberley populations was detected. Principal component analyses indicate potential climatic niche differences that could support recent localised adaptations. Climatic reconstructions back to the last glacial maximum (LGM) indicate areas of suitable habitat have substantially shifted through time for W. squamicaudata and suggest multiple areas of refugia across the Kimberley since the LGM. Further comparative research is required to establish a biogeographical framework that will assist in our understanding of processes that have shaped biodiversity in northern Australia and assist in conservation planning.


Wyulda squamicaudatamtDNANorthern AustraliaConservationGeneticsRefugia


Conservation planning hinges on an understanding of the evolutionary and ecological processes that shape biodiversity, as these are the building blocks with which species and ecosystems will persist and adapt (Klein et al. 2009). Understanding the patterns and processes responsible for generating diversity helps to (i) identify areas of endemism (biological uniqueness), species richness and phylogenetic diversity, (ii) understand how biogeographic barriers influence genetic connectivity across the landscape, and (iii) predict refugia (areas of climatic/ecological and evolutionary stability); all of which are crucial in conservation planning and policy (see Cadotte et al. 2012). Identifying refugia that retain necessary niche and habitat requirements is important for preserving the natural processes that generate diversification and can be identified from assessment of biogeographic patterns as well as environmental processes (see Keppel et al. 2012). Within Australia, refugia are often associated with more mesic locations associated with complex topography (Mackey et al. 2008) such as parts of monsoonal northern Australia (Woinarski et al. 2007). However, northern Australia is currently facing a worrying trend of species decline while the true diversity in the region remains unknown (see Moritz et al. 2013).

The tropical monsoon region of northern Australia, however, has recorded major declines in mammals over the past three decades (see Fitzsimons et al. 2010; Woinarski et al. 2011). Previously this relatively intact landscape was thought to provide security and refuge against threats, but the region now faces challenges associated with cane toad invasion, changed fire regimes, feral cat predation, disease and pastoralism (Fitzsimons et al. 2010). Characterized by extensive sandstone plateaus which are set amongst a range of habitats, including savanna woodlands, rainforest, grasslands and wetlands (Russell-Smith et al. 1993; Woinarski et al. 2007), the region experiences a strongly seasonal climate with rainfall primarily during summer (wet season) and very little winter rainfall (dry season). The monsoonal climate has been developing since the late eocene and early oligocene (40–30 Ma) and climatic fluctuations along with a long history of anthropogenic influences (e.g., fire) has shaped the current distribution of taxa (see Bowman et al. 2010).

To date, limited phylogenetic and phylogeographic studies of this region have been conducted (e.g., Jennings and Edwards 2005; Fujita et al. 2010; Toon et al. 2010; Oliver et al. 2010, 2012; Melville et al. 2011; Potter et al. 2012; Catullo et al. 2013). However, concordant biogeographic patterns are supporting hypotheses developed earlier last century (reviewed in Eldridge et al. 2012) that indicate the Kimberley (Western Australia; see Fig. 1) and the Top End (Northern Territory) have acted as refugia during aridification of the continent throughout the Miocene and Pliocene (reviewed in Byrne et al. 2008, 2011), as well as glacial climatic cycles throughout the Pleistocene (Bowman et al. 2010; Fujita et al. 2010). Such studies are providing insight into the broad-scale historical biogeographic processes that have shaped diversification and persistence of taxa across this region. There is still a lack of knowledge about most taxa in this rugged and remote region. In particular, understanding of finer scale ecological and evolutionary refugia within these broader regions is extremely limited (Doody et al. submitted). An assessment of genetic differentiation within taxa across this region is crucial for determining true species diversity, which is currently seriously underestimated (e.g., Fujita et al. 2010; Potter et al. 2012; Catullo et al. 2013).
Fig. 1

Map of Wyulda squamicaudata tissue sample localities in the Kimberley, Australia. Black circles indicate genetic samples; and white squares together with black circles indicate samples used in spatial modeling analyses. Places of importance are outlined on the map, as well as the east–west (E–W) Kimberley Divide

The scaly-tailed possum (Wyulda squamicaudata) is a poorly known species, endemic to the Kimberley region of northern Western Australia (Doody et al. 2012). Recent studies of other rock-dwelling mammals (e.g., rock-wallabies—Petrogale) have revealed deep phylogenetic structure across the Kimberley and northern Australia (Potter et al. 2012), suggesting similar patterns may emerge in co-distributed taxa. As a monotypic genus that is phylogenetically divergent from other rock-dwelling mammals (Osborne and Christidis 2002), Wyulda is an excellent model for assessing how biogeographic and ecological processes have shaped the diversification of co-distributed mammals in the region. Wyulda was recently rediscovered in the east Kimberley, and 200–300 km separates the east and west Kimberley populations. The geographic isolation of the recently identified east Kimberley population has caused some speculation about current taxonomy and conservation priorities (Doody et al. 2012).

Wyulda squamicaudata was first described in 1917 from a specimen collected in the east Kimberley of Western Australia (Alexander 1919). It is a medium-sized (900–2,000 g) rock-dwelling mammal that inhabits low open woodland and vine thickets (Humphreys et al. 1984; Runcie 1999; Burbidge and Webb 2008), but utilises rocky crevices as dens (Runcie 1999). Wyulda belongs to the marsupial family Phalangeridae, which includes the brush-tailed possums (Trichosurus) and cuscuses (Phalanger, Strigocuscus, Spilocuscus, Ailurops) endemic to the forests of Australia, New Guinea and the islands of eastern Indonesia (Flannery et al. 1987). Wyulda squamicaudata is the most poorly known Australian phalangerid (Flannery 1994) and is patchily distributed in areas of high annual rainfall (in excess of 900 mm) in the Kimberley. Since it was first described, few specimens have been recorded and these have been restricted to the coastal west Kimberley (Burbidge and Webb 2008; McKnight 2008) (see Fig. 1). The elusive nature of this species and lack of further specimens from the east Kimberley led some to question the origin of the type specimen (e.g., Kerle 2001) or to speculate on declines and extinction in the east of its distribution (Burbidge and Webb 2008; McKnight 2008). However, W. squamicaudata was rediscovered in 2010 in the east Kimberley at Emma Gorge, El Questro Station, in the Cockburn Ranges (Doody et al. 2012) (see Fig. 1). This discovery confirms the presence of W. squamicaudata in the east Kimberley and the highly disjunct nature of its current distribution. Wyulda squamicaudata is listed internationally as ‘Data Deficient’, with a decreasing population trend (IUCN 2011) and at the state level as a high priority species requiring “urgent survey and evaluation of conservation status before consideration can be given to declaration as threatened fauna” (Department of Environment and Conservation Western Australia 2013).

The recent discovery of deep phylogenetic diversity within taxa from the Kimberley (e.g., Potter et al. 2012) necessitates a better understanding of genetic diversity within Wyulda to clarify future conservation requirements. Here we use mitochondrial and nuclear sequence data together with spatial modeling of current and past distribution to: (i) assess whether the disjunct east and west Kimberley populations are genetically differentiated, and if this differentiation coincides with the biogeographic barriers reported for other rock-dwelling mammals, (ii) determine whether suitable habitat is present between the disjunct populations, (iii) understand the evolutionary and ecological processes that have shaped current diversity within W. squamicaudata, and (iv) identify any refugial areas and niche requirements associated with the persistence of this rare endemic mammal. Broad-scale conservation initiatives forming in the area (e.g., Kimberley Science Conservation Strategy) will benefit from broader understanding of the evolutionary and ecological refugia across this landscape.


DNA extraction, amplification and sequencing

Sample size was unfortunately constrained by the inherent rarity of this species. Tissue samples were obtained from all available museum specimens (South Australian Museum, n = 4; Western Australian Museum, n = 2; Australian Museum, n = 1) and from ethanol preserved ear biopsies from field captures (n = 16) (see Fig. 1; Supplementary Table S1 for sample localities). One common brush-tail possum (Trichosurus vulpecula) was used as an outgroup, since Trichosurus is the sister genus to Wyulda (Osborne and Christidis 2002).

Genomic DNA (gDNA) was isolated from tissue samples using the ‘salting-out’ method of Sunnucks and Hales (1996) and the Qiagen DNeasy Blood and Tissue Kit. The mitochondrial control region (CR) was amplified using primers L15999 and H16498 (Fumagalli et al. 1997) and nicotinamide dehyrdrogenase subunit 2 (ND2) using mmND2.1 and mrND2c (Osborne and Christidis 2001). In addition, two nuclear regions were amplified for a subset of individuals; the non-coding nuclear ω-globin (intron 2; omega-globin gene using G314, Wheeler et al. 2001a, b; and G424, Blacket et al. 2006); and protein coding portions of exon11 of the breast and ovarian cancer susceptibility gene (BRCA1) using F-498MAC together with R11 and F9 with R-1MAC9-20 (Meredith et al. 2008).

PCR-amplifications were carried out in 25μL reactions with ~100 ng gDNA, 10× PCR Buffer (Qiagen), 0.20 mM dNTPs, 2 mM MgCl2, 2 pmol primers, 100× bovine serum albumin, 5× Q-solution and 0.5 U Taq DNA polymerase (Qiagen). Thermocycling conditions included initial denaturation at 94 °C for 2 min; 35 cycles of 20 s at 94 °C, 40 s at 55 °C (CR, ω-globin)/48 °C (ND2), 48–50 °C (BRCA1), and 50 s at 72 °C; and a final extension for 5 min at 72 °C. PCR products were purified using USB® ExoSAP-IT® (Affymetrix) and sequenced on an AB 3730xl DNA Analyzer (Applied Biosystems) by the Australian Genome Research Facility (AGRF).

Genetic analyses

DNA sequences were edited and aligned using Geneious (version 6.0.5; Biomatters). DnaSp (v5.10; Librado and Rozas 2009) was implemented to estimate the number of haplotypes (H), polymorphic sites and nucleotide diversity (π) (Rozas et al. 2003). Intra-specific sequence divergence (Dxy) among mitochondrial DNA (mtDNA) and nuclear DNA (nDNA) haplotypes was estimated using DnaSp. Nuclear sequence ambiguities at heterozygous sites were phased using DnaSp and 100 permutations. Due to absence of polymorphisms in ω-globin it was omitted from any further analysis.

Maximum likelihood (ML) and Bayesian inference analyses were first performed on individual mitochondrial regions using RAxML (version 7.0.3; Stamatakis 2006; Stamatakis et al. 2008) and MrBayes (version 3.1.2; Ronquist and Huelsenbeck 2003; Huelsenbeck and Ronquist 2005) to ensure concordance. Analyses were then performed on the concatenated mitochondrial alignment, partitioned by gene. The Hasegawa-Kishino-Yano (HKY; Hasegawa et al. 1985) model was applied, with rate variation among sites modeled with a discrete gamma distribution (G) and proportion of invariant sites (I) for MrBayes, based on Modeltest 3.06 (Posada and Crandall 1998) results based on the AIC in PAUP* (version 4.0b10; Swofford 2002). The GTRGAMMA model was applied for RAxML based on model selection options and user guidelines.

Likelihood analyses were started from a complete random starting tree using the rapid Bootstrap analysis, with 1,000 pseudo-replicates and 100 searches per replicate. Default settings as priors were implemented in MrBayes, using random starting trees and four Markov chains (three hot, one cold), sampling every 1,000 generations, for each of two independent analyses run simultaneously. Analyses were terminated when the average standard deviation of split frequencies for the simultaneous analyses fell below 0.01 (~2–10 million generations). Tracer (version 1.5; Rambaut and Drummond 2009) was used to check that the convergence of parameter estimates and log likelihood values had occurred. Posterior probabilities were calculated after discarding the first 25 % of the sampled trees as burn-in. A haplotype network was run using TCS to assess the pattern of genetic differentiation as relationships within low variability loci are often better visualized using phylogenetic networks (Posada and Crandall 2001; Cassens et al. 2005).

To determine if there was genetic differentiation between the geographically disjunct east and west Kimberley populations we assessed Fst values in Arlequin (version 3.11; Excoffier et al. 2005). Analyses were run on individual mtDNA and BRCA1 data sets with 110 permutations, as well as a combined data set.

Tests for deviation from neutral processes were assessed using Tajima’s D (Tajima 1989), Fu’s Fs (Fu 1997) and R2 (Ramos-Onsins and Rozas 2002) run in DnaSp on separate mtDNA and BRCA1 alignments. These are the most powerful tests for detecting demographic expansion and selection/genetic hitchhiking, particularly when dealing with limited sample size (Ramos-Onsins and Rozas 2002). Significance was assessed with 1,000 coalescent simulations against the null hypothesis of a constant population size model. In addition, demographic history was assessed using Bayesian Skyline Plot (BSP) analysis implemented in *BEAST (version 2.0.2; Bouckaert et al. 2013) for mtDNA and BRCA1 individually. Using a strict clock and default priors, the BSP was run for 10 million generations sampled every 1,000 generations with an estimated gamma and proportion of invariant sites for the mtDNA analysis. Tracer (version 1.5; Rambaut and Drummond 2009) was used to ensure satisfactory convergence, based on effective sample sizes and the BSP analysis was performed using the Stepwise-Constant model and ten groups.

Spatial and niche modelling

We downloaded locations of all 54 confirmed W. squamicaudata captures from NatureMap (, which yielded 34 distinct locations. To assess distribution of suitable habitat for W. squamicaudata we fitted a species distribution model (SDM) using climate and topographic predictors. We used interpolated climate grids for present and paleo-climate based on snapshot simulations using the Hadley Centre Climate model (HadCM3; Singarayer and Valdes 2010), prepared as per Fuchs et al. (2013) at 2.5 arc min (~4.5 km) resolution across Australia for three time periods: the present, 6,000 years ago (6 kya) (mid Holocene optimum) and 21 kya (last glacial maximum). Downscaling of climate surfaces was performed using the climates package (VanDerWal et al. 2012) in R ( to generate eight derived climate variables for each time period, representing a subset of the widely used Bioclim (Houlder et al. 2000) climate surfaces. Two topographic predictors, slope and elevation range were calculated from the Etopo1 digital elevation model (Amante and Eakins 2009) at 1 arc minute resolution (~1.8 km) and summarized to align with the climate grids. To enable prediction to areas currently under sea (for past time periods), slope and elevation range were calculated from combined land and bathymetric elevation data (see Supplementary Table S2 for predictors used). The layers were clipped to a 2.5° radius around the Wyulda sites to constrain the background environment in the models. To reduce a bias in sites towards the coastal northwest of the range, we filtered species sites so there was none in adjacent grid cells, leaving 29 capture sites. A species distribution model for W. squamicaudata was fitted to current climate, and then projected into the climate of 6 kya, the mid-Holocene optimum and 21 kya, the LGM, using clamping to limit extrapolation to climates not present in the model fitting.

We performed a principal components analysis (PCA) using the pixel values for the environmental variables that contributed to the fitted SDM (Supplementary Table S2) and plotted the eastern and western W. squamicaudata sites in this ordination space.


Sequence data

A total of 23 individuals of W. squamicaudata were sequenced for CR and 15 for ND2 due to poor amplification success in some degraded samples. We obtained 1189 base pairs (bp) of mitochondrial sequence data (584 bp CR; 585 bp ND2). The CR had 33 variable sites, include 26 parsimony informative, where as ND2 had seven polymorphic and four parsimony informative sites. One indel was detected within the CR alignment. A total of 13 CR haplotypes were found with nucleotide diversity of 0.017, whereas ND2 had reduced diversity with only five haplotypes and a nucleotide diversity of 0.007. From the 1076 bp of BRCA1 sequence data there were nine polymorphic sites. Of the 26 haplotypes (13 individuals) sequenced a total of seven were unique with nucleotide diversity of 0.002.

Phylogenetic analysis of mtDNA indicated a single lineage for W. squamicaudata with individuals from the east Kimberley forming a cluster of related haplotypes within the more diverse west Kimberley lineage (results not shown). Due to low diversity the relationships amongst mtDNA sequences is illustrated using a haplotype network (Fig. 2a). There were no shared haplotypes between the east and west Kimberley (Fig. 2a), with the east Kimberley individuals forming a cluster with few mutational differences (1–3) compared to the differences of up to 21 found amongst the west Kimberley haplotypes. The haplotype network for BRCA1 revealed two shared haplotypes amongst the east and west Kimberley, as well as five haplotypes unique to each region. All haplotypes differed by 1–6 mutations (Fig. 2b).
Fig. 2

a Mitochondrial haplotype network for Wyulda squamicaudata from the east (grey) and west (black) Kimberley with Trichosurus vulpecula (outgroup) highlighted on the network to indicate its association based on maximum likelihood and Bayesian phylogenetic analysis (not shown). b Haplotype network for BRCA1 for W. squamicaudata from the east and west Kimberley. Circle sizes are proportional to the number of individuals with each haplotype, each branch represents one mutational difference and bars indicate the number of mutational steps between haplotypes

There were two fixed mtDNA differences between the east and west Kimberley populations with a Dxy = 0.015 (1.0–1.7 % sequence divergence; SD). There was 0.1–0.4 % SD between east Kimberley haplotypes and 0.1–1.7 % SD between west Kimberley haplotypes. For BRCA1 there was one fixed difference between the east and west Kimberley populations and Dxy = 0.004 (0.1–0.2 % SD). Despite low overall divergence, genetic differentiation (Fst) between the east and west Kimberley populations was significant for both mtDNA (0.566; P = 0.00) and BRCA1 (0.190; P = 0.02).

The demographic tests for neutrality and population expansion were all non-significant (P > 0.05) when W. squamicaudata was analysed as a single population and when it was separated into east and west Kimberley populations. The BSP plots indicate a recent decline in population size for both mtDNA and BRCA1, however the confidence limits to support this are broad and limited given the upper boundary also indicates a population expansion (Fig. 3).
Fig. 3

Bayesian skyline plots highlighting population trends of Wyulda squamicaudata back in time based on a mtDNA, and bBRCA1. Blue lines indicate confidence limits. (Color figure online)

Spatial and niche modelling

The SDM for current distribution of W. squamicaudata based on seven climatic and two topographic predictors performed well (AUC = 0.94), driven most strongly by temperature seasonality, precipitation (wettest quarter and annual), and slope. The broader scale distribution of habitat was principally driven by climate, whilst the local topography is important in determining the locally suitable areas. As expected, the model shows high suitability for the species along the northwest coast of the Kimberley (Fig. 4a). Relatively high suitability extends further northeast along the coast. The inland, eastern samples are in an area of suitable habitat that covers the heavily dissected terrain in the east Kimberley. We found low habitat suitability for the type locality, despite its inclusion in the data used to train the model. The gap of roughly 240 km between the eastern and western locations was predicted to have only limited suitability as habitat for W. squamicaudata, with the most suitable, yet marginal, linking habitat through the northern Kimberley.
Fig. 4

a A species distribution model for Wyulda squamicaudata based on the present climate; b projected onto modelled climate for the mid-Holocene optimum; c and the LGM

Projecting this model into the past, we infer a history of substantial shifts in the locations of suitable climate for the species with suitable habitat at 6 kya (Fig. 4b) focused in the west Kimberley whilst at 21 kya (during the LGM; Fig. 4c) this area appears to have been quite unsuitable. Instead, areas of the east Kimberley, including part of what is now the Joseph Bonaparte Gulf that at that time was above sea level, appear to provide an area of relatively high suitability.

The first two PCA axes of environment across the region together captured 70 % of variance in environment (Fig. 5). Precipitation (annual and wet season) and temperature seasonality were most strongly associated with PC1, which captured 46.7 % of variance, while PC2 aligned most with temperature (annual, driest quarter, warmest quarter), capturing 23.8 % of variance. The western (circles) and eastern (triangles) sites occupy distinct regions of environmental space. The sampled east Kimberley individuals are environmentally close to the main distribution of the west Kimberley population, whereas the type locality sites in the east Kimberley are far more strongly differentiated.
Fig. 5

Principal components analysis of environmental variables of habitat for Wyulda squamicaudata from the east Kimberley (triangles) and west Kimberley (circles). The first principal component (PC1, 46.7 % of variance) is driven by precipitation (annual and wet season) and temperature seasonality; PC2 (23.8 % of variance) is driven by temperature (annual, driest quarter, warmest quarter)


Our genetic data identified no evidence of deep phylogenetic structure between the disjunct east and west Kimberley populations of W. squamicaudata, elucidated by low mtDNA divergence (1.7 % mtDNA SD) and a lack of monophyly or diversity in BRCA1 (0.2 % SD). This contrasts to the pattern of genetic differentiation found in the co-distributed rock-dwelling brachyotis group of rock-wallabies (Petrogale), where deep divergences (5.5–7.6 % mtDNA SD) dated to around 1.8 million years ago, were observed between the east and west Kimberley (Potter et al. 2012). This suggests that biogeographic hypotheses proposed by Potter et al. (2012) in regards to the east– west Kimberley Divide have not influenced the genetic structure within W. squamicaudata. Despite a lack of deep phylogenetic structure, there was evidence of some genetic differentiation between the east and west Kimberley, with significant Fst values and unique mtDNA haplotypes in both populations.

The limited genetic divergence we have detected between the east and west Kimberley W. squamicaudata populations could be a result of: (i) recent colonisation across the Kimberley, from east to west or vice versa; or, (ii) the presence of unknown (or recently extinct) populations between the sampled populations creating recent broad connectivity (even if populations have recently condensed across the landscape due to changes in habitat suitability since the LGM). Recent colonisations into the east or west Kimberley seem unlikely given there is no evidence to suggest recent population expansion with non-significant R2 results and a lack of support from BSP analyses. Although the lack of samples identified from survey efforts throughout the past 200 years implies that this species is not abundant across the Kimberley, dispersal involving unsampled populations cannot be ruled out.

The SDMs indicate spatial and temporal shifts in the locations of suitable habitat for W. squamicaudata across the Kimberley since the LGM. The model of current W. squamicaudata habitat indicates the most suitable habitat connecting the east and west Kimberley populations is not through the central Kimberley, but rather follows the more mesic coastal regions. The 6 kya historic model indicates the most suitable habitat was in the west Kimberley, which is consistent with greater contemporary mtDNA diversity being found within this population. What is surprising is that the west Kimberley was modelled as largely unsuitable habitat at the LGM, and instead the east Kimberley appeared to provide the best habitat during this period. Heavily dissected areas in the east Kimberley were predicted as suitable habitat in all three time periods despite only limited sampling from this area included in model fitting. This suggests that W. squamicaudata in the east Kimberley may have persisted in this region for over 20,000 years whilst the west Kimberley populations have only more recently moved into the current habitat. As the broader areas of suitable habitat across the Kimberley are likely to have varied, assessment of phylogeographic patterns of other taxa is needed to help elucidate whether areas highlighted from SDMs as suitable habitat through to the LGM (e.g., northeast and northwest Kimberley) have acted as an area of evolutionary or ecological refugia. More broadly the Kimberley has been highlighted as a unique biogeographic region (e.g., Cracraft 1991; Bowman et al. 2010), however some areas of endemism within the Kimberley, such as the northwest have been identified (e.g., land snails—Köhler 2010, 2011; frogs—Rosauer et al. 2009; Doughty 2011) which could indicate that this region has provided important evolutionary refugia. The northeast Kimberley is emerging as another area hypothesized to have potentially provided evolutionary refugia (e.g., Oombulgurri region—Fig. 1; see Moritz et al. 2013).

Since W. squamicaudata is rock restricted, it has been hypothesized to inhabit areas of refuge against current threats like fire, predators and introduced herbivores which have altered the savannas of northern Australia (Legge et al. 2008, 2011; see Woinarski et al. 2011). Therefore identifying the patterns of other species, which require similar habitat, will help to understand how these topographically complex regions have allowed persistence of diversity in other taxa. The regions we see emerging in the northwest and northeast Kimberley as areas of suitable habitat for W. squamicaudata could reflect broader areas of refugia for species in the region. The discovery of only one population of W. squamicaudata in the east Kimberley is consistent with the declining suitability of the area’s climate for Wyulda since the LGM, potentially confining individuals into sheltered areas that maintain the specialised niche requirements which appear to influence the distribution of W. squamicaudata. The drivers in the spatial model being associated most strongly with lower temperature seasonality, greater precipitation (wettest quarter and annual) and slope, may support the finer scale niche requirements for persistence of W. squamicaudata or highlight the most stable and mesic regions across the Kimberley since the LGM. In the east Kimberley, where W. squamicaudata is only found in deep gorges and sheltered areas, the associated microhabitat may provide the necessary temperature and moisture regimes which are available more broadly in the more mesic west Kimberley and explain their local persistence. Given the limited available samples of W. squamicaudata and their elusive nature, it is difficult to assess whether the genetic pattern found here is an artifact of isolation by distance (with or without localized extinction), events of dispersal, or the effect of recent localized adaptation. Currently there is a major lack of sampling for taxa in the northeast Kimberley (see Moritz et al. 2013). Further survey work, particularly in areas highlighted from the spatial modeling along the northern Kimberley coast, would elucidate whether populations exist between the east and west Kimberley and help elucidate whether populations of W. squamicaudata do persist in this northeast region of the Kimberley and determine the diversity associated with this potential refugium. If unsampled populations are present along the coast, further analysis would indicate whether localized adaptation is a true feature within this species or just an artifact of limited geographic sampling. As it stands, there is no evidence to suggest populations in the east and west Kimberley represent distinct taxonomic entities. However the importance of fine-scale environmental requirements and adaptations should not be ignored, as this may be influential into the future.

The role of different habitats in influencing the persistence and connectivity of species across the Kimberley is only now starting to emerge. This study highlights a contrasting pattern to other rock-dwelling mammals in the region highlighting the influence of species’ biology in the evolutionary processes that have shaped current patterns of diversity (e.g., dispersal, niche requirements). Despite these differences, both Petrogale and Wyulda indicate the importance of the northwest and northeast as potential refugia with Petrogale highlighting divergent genetic structure within the east Kimberley and between the east and west Kimberley potentially resulting from different refugial areas in these regions in the past. The current areas of suitable habitat across the landscape for Wyulda that have been identified may reflect common patterns for other small-medium sized mammals and highlight the importance of broad areas in the Kimberley important in conservation. With the increase of anthropogenic threats to the flora and fauna of the Kimberley, it is essential we prioritize such areas for conservation. Our results indicate the importance of understanding evolutionary processes and ecological niche requirements in different co-distributed taxa, since different patterns emerge and are potentially caused by different processes (e.g., species’ biology). As it is emerging, two prominent regions, the northwest and northeast Kimberley appear to be areas of refugia at various times throughout history. Additional data sets will add to our understanding of areas of evolutionary and ecological refugia and assist in defining broad areas for conservation that maximize the evolutionary potential of taxa.


Our integrative study of molecular data and spatial modeling provides evidence that suggest ecological refugia across the Kimberley have shifted through the last 21 kya. Several areas of evolutionary and climatic refugia may have influenced current diversity within W. squamicaudata and have allowed its persistence across the Kimberley. Current patterns imply that the east Kimberley population may be locally adapted inhabiting an ecological niche associated with less wet season precipitation and greater temperature seasonality than western populations, or that fine scale habitat shelters it from such differences. Persistence at a broader scale across the Kimberley may be associated with fine scale microhabitat and further research could elucidate whether this is a pattern similar in other taxa. Comparative studies would indicate whether there are shared areas of refugia between co-distributed taxa and whether the habitat niches required for persistence of diversity across the Kimberley are similar between species. Currently it is unclear whether W. squamicaudata is present in patches of suitable habitat between the east and west Kimberley populations, particularly in areas in the east Kimberley where sampling of most taxa is poor. Future survey efforts need to be focused in areas such as Oombulgurri to understand if the limited differentiation of the east Kimberley is just an edge effect, a consequence of isolation by distance, ongoing dispersal, or if it represents true local adaptation. What we can see is that subtle and divergent differences are starting to emerge across the Kimberley and as we understand evolutionary processes for taxa we will be able to better integrate this knowledge into conservation efforts to ensure loss of adaptive potential is not lost.


We would like to thank the following people and institutions for providing samples or assisting with sample collection: Christina Castellano, Simon Clulow, Rosie Honin, Colin McHenry, David Pearson, Ian Radford, Western Australia Department of Environment and Conservation, Australian Wildlife Conservancy, Western Australian Museum, South Australian Museum and Australian Museum. Paleo-climate data were generously provided by Jeremy VanDerWal. We are also grateful to Craig Moritz for helpful comments on the manuscript. This research was supported by funding from the Chadwick Fellowship (Australian Museum), Australian Geographic Society and Monash University.

Supplementary material

10592_2014_601_MOESM1_ESM.docx (16 kb)
Supplementary material 1 (DOCX 82 kb)

Copyright information

© Springer Science+Business Media Dordrecht 2014