Independent Transitions between Monsoonal and Arid Biomes Revealed by Systematic Revison of a Complex of Australian Geckos (Diplodactylus; Diplodactylidae)

How the widespread expansion and intensification of aridity through the Neogene has shaped the Austral biota is a major question in Antipodean biogeography. Lineages distributed across wide aridity gradients provide opportunities to examine the timing, frequency, and direction of transitions between arid and mesic regions. Here, we use molecular genetics and morphological data to investigate the systematics and biogeography of a nominal Australian gecko species (Diplodactylus conspicillatus sensu lato) with a wide distribution spanning most of the Australian Arid Zone (AAZ) and Monsoonal Tropics (AMT). Our data support a minimum of seven genetically distinct and morphologically diagnosable taxa; we thus redefine the type species, ressurrect three names from synonymy, and describe three new species. Our inferred phylogeny suggests the history and diversification of lineages in the AAZ and AMT are intimately linked, with evidence of multiple independent interchanges since the late Miocene. However, despite this shared history, related lineages in these two regions also show evidence of broadly contrasting intra-regional responses to aridification; vicarance and speciation in older and increasingly attenuated mesic regions, versus a more dynamic history including independent colonisations and recent range expansions in the younger AAZ.


Introduction
The extent and intensity of arid conditions in the Southern Hemisphere has increased through the late Neogene, and expansive deserts are now a prominent feature of most southern continents (Africa, Australia and South America) [1][2][3][4]. These generally young arid zones are characterised by low, unpredictable rainfall and strong seasonal variation in temperature, and this major climatic shift has had profound biological implications; some lineages have adapted to Based on these data we also present a revised taxonomy, formally recognizing seven of the lineages identified by Oliver et al. [29] as species (redefined Diplodactylus conspicillatus sensu stricto, resurrected Diplodactylus hillii, D. laevis and D. platyurus and three newly described species) and thereby add six further species to the diverse Australian lizard fauna.

Material examined
This study utilised specimens and tissues held in the Australian Museum (AMS), National Museum of Victoria (NMV), Northern Territory Museum and Art Gallery (NTM), Queensland Museum (QM), South Australian Museum (SAMA) and Western Australian Museum (WAM). Where possible, specimens included in genetic analyses were also included in morphological analyses. Tissue samples from nominated holotypes for all three newly described taxa were included in assessments of genetic diversity.
Morphological characterisation of the types of D. conspicillatus and its synonyms (as listed in Cogger et al., 1983), D. hillii, Gymnodactylus laevis and D.

Genetic data
Genetic analyses included mitochondrial data from 169 specimens of Diplodactylus conspicillatus sensu lato (see Table 1 for specimen and locality information). DNA extraction and sequencing protocols for most samples are detailed elsewhere [25,31]. DNA from new samples was extracted using a Qiagen high throughput extraction robot at Museum Victoria. A *1200 base pair (bp) region of the ND2 gene and surrounding tRNAs was amplified using one of the following two combinations of primers: 1) AAG CTT TCG GGG CCC ATA CC (L4437) [32] and CTA AAA TRT TRC GGG ATC GAG GCC (Asn-tRNA) [33]; or 2) GCC CAT ACC CCG AAA ATS TTG and TTA GGGTRG TTA TTT GHG AYA TKC G [25]. PCR products were amplified for 40 cycles at an annealing temperature of 55°C. Unpurified PCR products were sent to a genetic services company (Macrogen, Korea) and sequenced in both directions using Sanger sequencing technologies.
New sequences generated in this study were aligned with data presented by Oliver et al. [30] and Pepper et al. [24]. Alignment of sequences was first performed automatically using the software MUSCLE [34], then refined by eye in Se-Al [35]. We translated nucleotide data into amino acid sequences and checked the alignment for internal stop codons and frame-shift mutations. Our final edited alignment included up to 1054 characters. We used the unlinked branch lengths and BIC settings in PartitionFinder [36] to determine the best partitioning strategy and model of nucleotide substitution (GTR+I+G, with all codon positions considered together in a single partition).

Phylogenetic analyses
Phylogenetic relationships were estimated using standard Maximum Likelihood (RAxML v7.2.8) [37] and Bayesian techniques (BEAST v1.8.0) [38]. All unique samples were included in initial analyses (S1 Fig.), however for subsequent phylogenetic analyses we focused on a reduced subset of data from which a number of identical or near identical sequences for the two most extensively sampled major clades were removed. Maximum Likelihood analyses were run using the default settings for RAxML on the CIPRES portal; the GTR+G model of sequence Bayesian analyses in BEAST used models and partitions as suggested by Partitionfinder, the Yule speciation prior (appropriate for analyses including relatively divergent lineages) and a relaxed log-normal clock and with model and partitions applied as above. After initial experimentation with settings and sampling, the final MCMC chains were run for 50 million generations, sampling every 50,000 steps. We estimated a timeframe of divergence using a 3% mean rate of pairwise sequence divergence (with a range between 1-4%) per one million years (see Oliver et al. [16] for justification). Tracer v1.5 [39] was used to confirm stability of parameter estimates and adequate mixing of the MCMC chains, and determine appropriate burn-in and acceptable effective sample sizes (> 200). Maximum clade credibility trees, after exclusion of the first ten million generations (20%), were summarized with TreeAnnotator v1.7.2 [38].

Biome evolution
Ancestral state analyses were coestimated with topology and divergence dates in BEAST. To assess the number and nature of transitions between biomes, each node was coded as to whether the corresponding specimen was from within (1) or outside (0) the AAZ (defined here by a moisture index [mean annual rainfall divided by evaporation] of less than 0.4 [9]; Fig. 1, dotted line). The biome state of all outgroups was coded as ambiguous because: a) basal relationships within Diplodactylus were unresolved, and b) some of these taxa occur in the temperate biome, while this study is focused on transistions between the AMT and AAZ. The pattern of biome evolution was estimated using a simple substitution model for binary data which assumes equal probabilities for transitions between all states [40].

Population genetics
To infer past demographic fluctuations in response to the climate cycles of the Pleistocene, we calculated nucleotide diversity and tested for population size change using the basic population genetic measurements of Tajima's D [41], Fu's Fs [42] and R 2 [43] as implemented in DnaSP v. 5.0 [44]. Estimates and significance were calculated with 1000 coalescent simulations against the null hypothesis of a constant population size model, for each population corresponding to the phylogenetically distinct groups based on the mtDNA gene tree. As these historical inferences were based on data from the mitochondrial genome, they should be regarded as a preliminary framework requiring corroboration with appropriate nuclear loci. For major clades intra and inter-specific genetic distances were estimated using Arlequin v. 3.11 [45].

Morphometrics
All measurements (except SVL) and bilateral counts were recorded from the left side. The following measurements were taken using Mitutoyo electronic callipers: snout to vent length (SVL), tip of snout to anterior margin of cloaca with body straightened; tail length (T), from posterior margin of cloaca to tip of tail; tail width (TW), widest point across original tail; head length (HL), mid anterior margin of ear to tip of snout; head width (HW), widest point of head, usually corresponding with, or slightly posterior to, position of ear opening; head depth (HD), lower jaw to top of head at mid orbit; snout length (S), tip of snout to anterior margin of orbit; eye to ear (EE), posterior margin of orbit to mid anterior margin of ear; length of forelimb (L1) and hindlimb (L2), from insertion to tip of longest digit (claw included), with limb stretched straight perpendicular to body; and (AG) axilla to inguinal region with body straightened.
The following scale counts and characters were recorded: subdigital scales from tip of digit (4 th finger, 4 th toe) to basal junction of 3 rd and 4 th digits (series includes enlarged distal pair); supralabial and infralabial scale rows (beginning immediately posterior to rostral and mental scales, and terminating where there is a noticeable reduction in size, or the labial scales begin to pull away from the lip line [approximately level with mid orbit]); number of small scales contacting the posterior edges of the rostral and mental scales; the number of scales in a longitudinal series along the length of the original tail (along vertebral line); the number of scales across the original tail (transverse count taken across the large scale row closest to widest point of tail); the size of the back, nape and head scales (relative to flank and lateral neck scalation); the presence or absence of a small medially projecting process on the posterior edge of the mental; and finally, the size of the 1 st supralabial in relation to the rest of the supralabial row. Specimens included in the morphometric assessment are listed within the species accounts. Additional material examined is listed in S1 Appendix.

Nomenclatural acts
The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix "http://zoobank.org/". The LSID for this publication is: urn:lsid:zoobank.org:pub: C410144B-EC99-4AA4-8780-A2858356CF32. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: PubMed Central, LOCKSS.

Phylogenetic relationships
Monophyly of D. conspicillatus sensu lato was strongly supported in all analyses ( Fig. 2A-B). Within this clade we identified the nine major lineages corresponding to the candidate species identified by Oliver et al. [30], specifically; D. conspicillatus sensu stricto-widespread in the arid zone and extending into the AMT; lineage A-Gulf region, north Queensland; lineage Bwestern Pilbara and Carnarvon region, Western Australia; lineage C-widespread arid zone; lineage D-western Top End, Northern Territory; lineage E-Kimberley, Western Australia; lineage F-Channel Country, western and central Queensland and far north-west New South Wales; lineage G-around Townsville, Queensland; and lineage H-gulf country, north Queensland ( Fig. 1). Monophyly of all major clades is strongly supported, and mean uncorrected genetic divergence between lineages is relatively high (11.3-22.5%) (S1A Table). Lineages B, C, D and E form a clade that is well supported as sister to another clade comprising D. conspicillatus sensu stricto and lineage A. Collectively these clades (D. conspicillatus sensu stricto and A-E) are well supported as sister to the most divergent clade of the complex which contains lineages F-H from eastern Australia.

Divergence dates and biome evolution
Topology and support values for the phylogeny of the D. conspicillatus complex co-estimated by BEAST were congruent to those from RAxML ( Fig. 2A-B). Age estimates derived from application of a mean pairwise sequence divergence rate of 3% per million years suggests the deepest divergences within the complex (including the majority of the lineages and major clades discussed above) occurred in the late Miocene (*5-10mya). Where dense sampling was avaliable the accumulation of diversity within most candidate species is estimated to have occurred during the Pleistocene, with the exception of the clade comprising lineages F-H which includes a number of deep and relatively poorly sampled lineages distributed across Queensland. Nodes with ML support above 95 and Bayesian support (BEAST) above 99 (respectively) are indicated with an asterisk (*). (B) Chronogram and ancestral biome states for the seven species in the revised D. conspicillatus group estimated using BEAST and calibrated with a 3% pairwise mean rate of molecular evolution. Green lineages are from outside the central Australian arid zone (defined by a moisture index of less than 0.4), brown lineages are from inside the Australian arid zone, and the probability (i.e. percentage of reconstructions that feature the observed state) of the inferred ancestral habitat is indicated for major nodes. Distributional data based on morphotyped samples indicates that of the nine major lineages in the D. conspicillatus complex, five (A, D, E, G and H) are absent from the central arid zone (and are mostly restricted to the AMT), two (D. conspicillatus sensu stricto and F) occur in the both AMT and AAZ, and two (lineages B and C) are restricted to the AAZ (although the range of the former is centred on the comparatively mesic Pilbara [see discussion below]) (Fig. 3). Support for most ancestral state reconstructions was relatively weak, but our results suggest that monsoonal environments are ancestral and also provide strong evidence that there have been multiple transitions between the arid and monsoonal areas (Fig. 2B).

Population genetics
Genetic sampling for a number of lineages was sparse, and only D. conspicillatus sensu stricto (n = 28), lineage B (n = 60) and lineage C (n = 62) were well sampled with good geographic spread across their distributions (S1B Table). Diplodactylus conspicillatus sensu stricto included two divergent sublineages (>10% divergence) with distributions in arid and monsoonal areas of Queensland and the Northern Territory, and in arid South Australia and Western Australia, respectively. These two sublineages also showed evidence of further structure (especially the north sublineage which showed mean genetic divergence of 6.7%). Lineage B showed low structure across its distribution in the western/southern Pilbara and Carnarvon Basin (mean 1.3%), while lineage C was characterised by moderate mitochondrial haplotypic diversity (mean 3.9%), especially in the western edge of its range in the northeastern Pilbara and surrounding regions.
Our analyses of population size change using summary statistics indicate that lineages B and C from the AAZ both had significant and large negative values for Fu's Fs (−33.0 and −14.3, respectively)-consistent with a signature of contiguous range expansion [42]. Tajima's D measurements were significantly negative only for lineage B. The other widespread arid zone lineage D. conspicillatus sensu stricto had small negative values (-2.55), however these measurements were again not significant. There was no evidence of deviation from neutral expectations in the largely AMT lineages E or F-H (considered together), which can be interpreted demographically as stable populations in mutation-drift equilibrium [42] (S1C Table). As sampling for all of these northern lineages was very sparse these results should be interpreted with caution.

Morphology
Morphological analyses of genotyped specimens revealed a suite of characters that diagnosed most of the major lineages identified (see Table 2 Using these characters we were readily able to assign specimens to most of the lineages identified by the molecular analyses, and determine the identity of museum specimens for which molecular data were unavaliable. The exception to this general pattern was the three eastern lineages (F, G, H), which could be readily diagnosed from all other lineages in lacking a distinctively enlarged first supralabial scale and in having no (or a poorly developed) canthal stripe, but aside from evidence of size differentiation did not show such consistent diagnostic morphological features from each other.

Species diversity
Delimiting species boundaries involves integration of independent data sources to identify distinct evolutionary lineages [46], ideally including information from mulitple nDNA loci, morphology, geography, ecology and reproduction [47]. The limitations of using mtDNA alone to infer species boundaries and historical phylogeography are widely recognised [48]. However, six of the nine major mitochondrial lineages we identified (D. conspicillatus sensu stricto and lineages A-E) are deeply divergent from each other and in addition can be readily diagnosed by a suite of morphological characters (distinctive features of scalation on the dorsum and original tails). Thus two lines of evidence support the hypothesis that these represent evolutionarly distinct and diagnosable lineages (species). Three further mitochondrial lineages identified previously (F-H) [30] form a strongly supported clade that can be readily diagnosed from all other members of this complex by their distinctive labial scalation, but are more difficult to diagnose from each other, and are represented by few samples in our analyses. Diplodactylus conspicillatus sensu stricto also includes two moderately divergent sublineages (Fig. 2) that were flagged but not named by Oliver et al. [30]. Genetic divergences between these sublineages are   lower than between the recognised candidate taxa, sampling for one is again sparse, and we did not find diagnostic morphological characters. More detailed sampling and additional nDNA data sources are required to resolve the taxonomic status of these remaining mitochondrial lineages; and for the time being we note their potential significance, but do not recognise any as distinct species. Formal diagnoses and descriptions of the seven species we recognise within the D. conspicillatus complex are provided in the systematics section at the end of this paper, however for the remainder of this discussion we consider each of these seven species as separate entities and use our revised binomial arrangement (see Figs. 1-3 for a summary of phylogenetic and distributional information).

Geographic structuring and bioregions
This study has confirmed that species diversity within the D. conspicillatus complex in both the AAZ and AMT is much higher than previously recognised. While significant sampling gaps remain (especially in northern Australia), based on combined genetic datasets and morphological assessments, we were able to infer the broad geographic distributions of most taxa. Of the seven species, three are endemic to the AMT, two are endemic to AAZ and two occur in both biomes. Our systematic analysis of this previously undetected diversity therefore provides oppurtunities to contrast patterns of diversity within biomes, and examine the potential timing and nature of transitions between them.
The distributions of the species endemic to the AMT broadly correspond to seperate regions of endemism; specifically the Kimberley, Top End and Gulf [16,17,49], and are separated by putative biogeographic barriers [49,50] (Fig. 3). Diplodactylus custos sp. nov. is endemic to the Kimberley region. The moderate genetic diversity of this species contrasts with very high genetic diversity of some saxicoline gecko lineages endemic to the same region [6,16,23] but is similar to Kimberley endemic toadlet lineages more strongly associated with savanna woodlands [49]. Based on current sampling, the eastern extent of the range of D. custos sp. nov. appears to broadly correspond with the putative Victoria River drainage barrier [50,51]. Genetic sampling for the remaining two AMT endemics was threadbare. However, based on diagnostic morphological data, D. hillii is only known from the western Top End (east of the Arnhemland Escarpment). A disjunction in this region (the Mid-Territory Break) has been detected in Uperolia toadlets [49] and may be related to variations in geology and topography around the Arnhem Escarpment. The distribution of D. barraganae sp. nov. along the Gulf of Carpentaria also mirrors that of a number of other lizard and frog clades [16,49]. Another putative biogeographic barrier, the Carpenteria Gap [17,49,52], separates D. barraganae sp. nov. from the easternmost species D. platyurus. In this region the clay plains in the hinterland of the Gulf of Carpentaria may form an important divide between the open woodlands of the Top End and Cape York Peninsula [17].
Diplodactylus platyurus has a distribution centred on Queensland, ranging from subhumid areas in the east and north and extending into the periphery of the AAZ in the west. This species contains two deeply divergent lineages from the AMT (Fig. 2, lineages G and H), while samples from a wide region along the eastern periphery of AAZ (lineage F) cluster together in a third lineage. The distribution of lineage F corresponds with the periodically flooded Channel country in western Queensland, a region that provides a set of microhabitats that are not typical of the AAZ, and is home to a suite of taxa that are absent from less watered areas to the west [16,53,54].
Diplodactylus conspicillatus also occurs in the AMT and AAZ, although the vast majority of its range is in the latter biome. This species includes divergent sublineages distributed to the north and west of the Lake Eyre Basin, respectively, and does not show a strong signal of range expansion (although sampling for the north lineage was sparse). These data suggest that this clade has persisted and diversified within or close to the edge of the AAZ since the late Miocene. The potential roles of the vast lower Lake Eyre Basin and highly mobile sand dunes of the Simpson Desert in shaping phylogeographic patterns within this region warrant further investigation [55,56]. The distribution of these two major sublineages also broadly corresponds with the transition from the slightly higher and more reliable summer rainfall deserts in the north, to the drier and more winter rainfall deserts to the south [57,58]; these gradients of seasonality and precipitation may provide climatic axes over which taxa could diversify within the AAZ.
Only two species (Diplodactylus bilybara sp. nov. and D. laevis) have distributions entirely confined to the AAZ, and both show a strong signal of population expansion and relatively shallow intraspecific genetic diversity over most of their ranges. Low genetic diversity has been detected in a number of widespread arid zone taxa, and is thought to reflect relatively recent and major demographic shifts through severe glacial cycles of the Plio-Pleistocene [7,9,14,18,59]. The distribution of the two AAZ endemic taxa in this gecko complex is also outwardly contrasting; Diplodactylus bilybara sp. nov. is restricted to the southern and coastal west Pilbara and Carnarvon regions, while Diplodactylus laevis has a vast distribution across central Australia. However, mitochondrial diversity within the latter species is concentrated along the the westernmost portion of its range, close to the range of the former [24]. This distribution of phylogenetic diversity supports previous work suggesting that the Pilbara and nearby areas have been an important zone of persistence and diversification at the western periphery of the arid zone [6,7,15].

Contrasting diversification in the AMT and AAZ
The overall timeframe and pattern of divergences in the D. conspicillatus complex implies that intensifying aridity since the late Miocene has played a central but at times contrasting role in shaping diversification in the AMT and AAZ [9]. Lineages in older and shrinking mesic zones (such as the AMT) are restricted to relatively small and largely allopatric patches of habitat, a distribution indicative of long-term persistence, but with increasing attenuation and potentially non-adaptive diversfication [6,8,9]. In contrast, the vast sandy plains of the central AAZ appear to have a more dynamic recent history of ecological adaptation, colonisations and large scale range shifts, but less in the way of intra-regional diversification and speciation [9,14].
Bayesian reconstruction of biome shifts within the D. conspicillatus complex suggests that mesic biomes are ancestral. While support for many ancestral state reconstructions in the tree is low, this overall pattern is consistent with the widely held idea that the Australian arid biota is largely derived from peripheral and more mesic biomes [9,14,60]. Furthermore, as intimated above, our simple binary classification of arid vs not arid is also probably overly simplistic; while peripheral regions such as the Pilbara (D. bilybara sp. nov.) and much of the Channel country (D. platyurus lineage F) are technically within the AAZ, compelling arguments can be made as to why they could be viewed as mesic refugia [24,53,61]. Under this interpretation only two widespread species (D. conspicillatus and D. laevis) would be considered successful  colonists of the AAZ, and the diversity in this zone would be rendered more clearly depaurate, recent and derived than that of the AMT.
Our phylogeny also strongly indicates there have been repeated, independent transitions between the AMT and the AAZ; detectable at both interspecific (D. barraganae sp. nov. and D. conspicillatus) and intraspecific levels (D. conspicillatus and D. platyurus). In an analysis of the plant biota of the Southern Hemisphere (including Australia), Crisp et al. [60] found that transitions between biomes were relatively rare in general, and transitions into arid biomes from monsoonal (savanna) environments were particularly rare. However, at least in Australia, the AMT remains the least studied of the major biomes [17,62] and this pattern may to some extent have reflected a lack of data. Even in this study, our sampling from the AMT is also sparse, and additional material will likely refine understanding. However, this and other broadscale analyses increasingly suggest that the history and evolution of many lineages in the the AMT and AAZ has been intimately linked since at least the late Miocene [14,16].
A final notable pattern is that the two most widespread arid zone taxa (D. conspicillatus and D. laevis) have broadly overlapping distributions in the southern and eastern arid zone (Figs. 1  & 3); the only instance of widespread sympatry within the D. conspicillatus complex. Relatively closely related congeners with overlapping distributions in the AAZ have been found in other widely distributed Australian lizard radiations [12,21,63]-sympatric diversity in these closely related lizard taxa may be further evidence of a relatively dynamic recent history of range expansion and ecological diversification in the vast but young arid biome [9].

Hyper-diverse species complexes and evolutionary biology
Nearly 100 new and widely accepted Australian squamate species have been described since 2000; indeed 2007 was the most 'productive' year on record for Australian reptile taxonomy (30 well-characterised species) [64]. While some of this new biodiversity represents singletons or other novelties uncovered by fieldwork, many 'new' species have been detected within morphologically cohesive nominal 'species' that actually comprise a larger number of unrecognised taxa (five or more). In the Australian context such complexes are particularly well documented in geckos [7,8,14,16,25,61,65,66], but have also been detected in blindsnakes [67], skinks [68][69][70], Donnellan pers com], and dragons [27]. While in some cases these complexes appear to comprise genuinely morphologically cryptic taxa, in others (such as the D. conspicillatus group) careful work oftens reveals a suite of diagnositic morphological characters.
There is little sign that the rate of discovery is slowing down, and if anything, it may increase in the short-term as sampling across northern Australia becomes more comprehensive, and researchers assemble increasingly large genomic datasets and develop new analytical methods [23,71]. Even our assessment of the D. conspicillatus group is likely to be an underestimate; there is further deep genetic diversity in D. conspicillatus and D. platyurus, many areas remain poorly sampled, and there are two morphologically distinct specimens from Cape York in north-Queensland for which no genetic data is available (see below). Thus, as with so many Australian lizard groups, further analyses will likely show that species diversity still remains underestimated.
Many arguments for the importance of continued efforts to properly understand this biodiversity for conservation and management purposes have been outlined in a compelling fashion elsewhere [22,72]. However we would like to conclude by further re-emphasising that systematic work on these complexes also has a tendency to reveal interesting macro-evolutionary patterns. For example, using the complexes of Australian lizards listed earlier in this section as examples; systematic work has revealed parthenogenesis [56], ancient vicarience and longterm persistence [8,61], rapid radiation [70], morphological conservatism or parallelism [27,67,68] and provided insight into the comparative history of biomes or regions of endemism ( [14,24], this study). Further work to resolve other species complexes will continue to provide a framework for broader insights into macroevolutionary processes.
Diplodactylus hillii (as D. hilli) was placed in the synonymy of D. conspicillatus by Kluge [76]. In this work Kluge only examined type material held in Australian museums and no consideration was given to the taxonomic status of G. laevis or D. platyurus. However, when Kluge revisited D. conspicillatus for his revision of the genus [77], both G. laevis and D. platyurus were also listed with D. hillii in the synonymy of D. conspicillatus (although the G. laevis type material is not listed amongst the specimens examined). Two of these synonyms (D. hillii and D. platyurus) were resurrected from the synonymy of D. conspicillatus by Wells and Wellington [78] but, as no justification was given, this action was widely ignored. Kluge's D. conspicillatus synonymy was followed by Cogger [29] who examined the type specimens of all the listed synonyms.
Based on a combination of morphology and genetics the available names can readily be assigned to the various taxa under consideration here. Key diagnostic characters are discussed in detail in the species accounts.

Species group diagnosis
Our concept of the D. conspicillatus group includes only species that are part of a strongly supported clade of related forms that have previously been synonymised or confounded with the nominate species. This is contra Kluge [77] and Storr et al. [79], who included the nominate species and some or all of D. kenneallyi, D. pulcher and D. savagei; the phylogenetic relationships of which, based on available data, remain unclear. However, they do not show any evidence of a strong or close affinity to the D. conspicillatus group [25].
All species in the Diplodactylus conspicillatus group can be distinguished from their congeners by the following combination of characters: all or most supralabials small and granular, at most only one enlarged anterior (1 st ) supralabial; terminal lamellae on fingers at most only slightly wider than digit; other prominent enlarged subdigital lamallae absent; tail short, as wide or wider than body, depressed with heterogenous scalation, usually bearing large platelike scales and/or conical tubercules arranged in transverse rows; and dorsal colouration extremely variable, but never consisting of large clearly defined bands or blotches. Comparisons in the following species accounts are restricted to taxa in the D. conspicillatus species group only.
The order of authorships for the three new species herein do not follow that of the paper as a whole.

Diplodactylus conspicillatus Lucas & Frost 1897
Variable Fat-tailed gecko Mid-dorsal scales on trunk plate-like and markedly larger than smaller dorsolateral scales. Scales on nape granular and only slightly larger than granules on side of neck. Original tail spade-like and lacking an acute attenuated extension at tip. Scales on dorsal surface of tail arranged in transverse rows (which usually include rows of both large and small scales). Pattern generally spotted and often with numerous dark blotches that contrast strongly with base colour ( Fig. 7A-B).
Pattern (in spirit). Variable. Most specimens tan to mid-brown and heavily chequered with small dark blotches that may coalesce to produce a reticulated appearance (lighter individuals more uniform; mid-brown, finely peppered with darker markings and bearing pale spots on dorsal and lateral surfaces). Pale spots generally present, most prominent on flanks. In some specimens there is reduced pigmentation on the vertebral zone producing as a ragged-edged vertebral stripe (one specimen, WAM R110770, has a well-defined dark vertebral stripe bordered on either side by a pale paravertebral stripe). Head generally with darker crown but paler towards periphery. A prominent, pale canthal stripe present, extending from anterior edge of orbit to tip of snout and producing a distinctive 'v' shaped marking which contrasts with the darker dorsal and lateral head markings. A broad dark zone on side of face extends posteriorly beyond eye to temporal region. A pale zone below eye extends to ear. Limbs mottled or spotted and inner digits of forelimb with reduced pigmentation. Ventral surfaces offwhite, immaculate.
Comparisons. Diplodactylus conspicillatus is readily distinguished from D. platyurus in possessing an enlarged first supralabial that contacts the ventral edge of the nasal scale (vs 1 st supralabial small and not differentiated from the rest of the supralabial row). It is distinguished from D. barraganae sp. nov. and D. hillii in having enlarged, plate-like mid-dorsal scales that are conspicuously larger than the dorsolateral scales (vs mid-dorsals small and granular, only slightly larger than dorsolateral scales). It is separated from the remaining three species in this complex (D. laevis, D. bilybara sp. nov. and D. custos sp. nov. by the shape of its original tail which is spade-like and lacks an acute attenuated tip (vs original tail bearing a short attenuated tip).
Distribution and Ecology. Very widely distributed throughout much of the arid zone; extending west to the eastern edge of the Pilbara and Western Australian Goldfields, east to Cunnamulla in south central Queensland, and south to the northern edge of the Nullarbor Plain in South Australia (Fig. 3). There are also scattered records from the Australian Monsoonal tropics in the Northern Territory, including two specimens from a high rainfall zone in north-eastern Arnhemland. Throughout this broad region this species inhabits a very wide range of habitats ranging from sparsely vegetated Gibber plains to open woodlands, but is generally associated with harder stony, clay and compacted sandy substrates.
Comments. Examination of the lectotype (NMV D7535: Fig. 8) shows it to be a poorly preserved specimen that looks to be slightly dessicated, is lacking a tail but has a rusted pin protuding from the tail base. Despite its poor condition, it clearly exhibits large plate-like scales on the vertebral region of its back accompanied by small, granular scales on the nape. The only other morphotype that occurs in the vicinity of the Northern Territory/South Australian border (Diplodactylus laevis) has large plate-like scales on both the vertebral region and the nape and is clearly not conspecific with NMV D7535.
The wide range of this species and observed deep phylogenetic structure suggests additional taxonomic investigations are necessary.  Pattern (in spirit). Variable. Most specimens tan to mid-brown and heavily marked with dark, irregular bands that form a broad reticulum on upper lateral / paravertebral zone and may extend to lower flanks. Vertebral zone with a ragged dark edge; generally free of pattern but sometimes the dark flank pattern may bridge this zone or the vertebral line may carry a row of small dark blotches (some individuals with a finer, lighter reticulum over entire dorsal surface which is marked with numerous small pale spots). Head with a pale crown that is continuous with the vertebral zone. A pale canthal stripe present, extending from anterior edge of orbit to tip of snout and producing a distinctive 'v' shaped marking which has dark edging. A poorly defined pale zone below eye extends to the ear. Limbs finely spotted. Inner digits with reduced pigmentation. Original tail with little pattern or with darker bars similar to those on flanks. Ventral surfaces off-white, immaculate.
Comparisons. Diplodactylus hillii is readily distinguished from D. platyurus in possessing an enlarged first supralabial that contacts the ventral edge of the nasal scale (vs 1 st supralabial small and not differentiated from the rest of the supralabial row). It is distinguished from D. conspicillatus, D. laevis, D. bilybara sp. nov. and D. custos sp. nov. in having small mid-dorsal scales that are only slightly larger than the dorsolateral scales (vs mid-dorsal scales conspicuously larger than the smaller dorsolateral scales). It is further separated from D. laevis, D. bilybara sp. nov. and D. custos sp. nov. in lacking an acute attenuated tip to the original tail (vs attenuated tip present). Diplodactylus hillii most resembles D. barraganae sp. nov. with which it shares small mid-dorsal scales and a blunt, spade-like original tail. These two species differ most in the configuration of the scales on the original tail (enlarged scales not in clearly defined transverse rows and mostly subequal for D. hillii vs clearly defined trows of both large and small scales for D. barraganae sp. nov.).
Distribution and Ecology. Found in eastern and central "Top End", from close to Darwin south as far as Elsey National Park (Fig. 3). Its habitat preferences within this area have not been determined.
Comments. The holotype of D. hillii, (QMJ 1994; Fig. 9), was examined and exhibits a unique scale configuration found only in D. conspicillatus sensu lato populations from the N.T. occurring above 15°S (i.e. scales on dorsal surface of original tail all large and not arranged in clearly defined transverse rows).
Pattern (in spirit). Variable. Most specimens tan to mid-brown with a darker reticulated pattern of fine to moderate wavy lines that extend over the entire dorsum. Many specimens exhibit fine pale spotting that is most evident on the flanks. Head, as for body with dark reticulations on crown. A pale canthal stripe present, extending from anterior edge of orbit to tip of snout and producing a distinctive 'v' shaped marking which has dark edging. A broad dark zone on side of face extends posteriorly beyond eye to temporal region. A poorly to well-defined pale zone below eye extends to the ear. Limbs weakly mottled or spotted and inner digits with reduced pigmentation. Tail marked with small dark flecks. Ventral surfaces offwhite, immaculate.
Comparisons. Diplodactylus laevis is readily distinguished from D. platyurus in possessing an enlarged first supralabial that contacts the ventral edge of the nasal scale (vs 1 st supralabial small and not differentiated from the rest of the supralabial row). It is distinguished from D. conspicillatus, D. laevis, D. bilybara sp. nov. and D. custos sp. nov. in having enlarged, platelike scales on the nape and top of head that are appreciably larger than those on the sides of the neck (vs scales on nape granular and not appreciably larger than those on sides of neck). It is most readily distinguished from Diplodactylus hillii and D. barraganae sp. nov. by the shape of its original tail which bears an acute attenuated extension at the tip (vs tail blunt, spade-like without an attenuated tip) and further distinguished from these species by its mid-dorsal scales (mid-dorsals enlarged and plate-like, conspicuously larger than the dorsolateral scales in D. laevis vs mid-dorsal scales small, only slightly larger than the dorsolaterals).
Distribution and Ecology. Widely distributed over much of the Australian arid zone, occurring from the Dampier Peninsula, Pilbara and Great Victoria Desert in the west, through much of north-western South Australia and the southern half of the Northern Territory, with an apparently isolated eastern population in the Channel Country around north-eastern South Australia (Fig. 3).
Comments. A black and white photographic image of the lectotype of Gymnodactylus laevis (SMF8242; Fig. 10) was kindly provided by Dr Harold Cogger. The specimen is damaged (partially digested) having been removed from the gut of a Varanus gouldii specimen from Hermannsburg Mission, NT. Despite its poor condition, it is possible to determine from the image that the specimen has an enlarged 1 st supralabial, enlarged scales on its mid-dorsum, nape and head, an acute attenuated extension at the tip of its original tail and some indication of a reticulated dorsal pattern. Cogger notes from his examination of the specimen that 'colour pattern is light brown or creamish with a series of irregular dark brown spots and patches forming a vague reticulum' (Cogger, unpublished data). This suite of characters fit the specimens examined above, whose distribution encompasses the central Australian region from which SMF8242 was collected. It remains unclear why Mertens [80] chose a partially digested specimen as the lectotype.
Pattern (in spirit). Variable. Most specimens tan to mid-brown with varying degrees of spotting; most prominent on flanks. Dorsum with an overlay of fine, dark reticulations or a more solid dark pattern. Vertebral zone with reduced pigment but often broken by transverse bars, isolating a series irregular pale blotches along back. In some specimens the vertebral zone is largely unpatterned and has a wavy edge where it borders the darker paravertebral zone. Head, as for dorsal ground colour with scattered dark flecks or blotches. Canthal stripe absent or very weak without sharply defined edges and not contrasting strongly with other facial markings. Limbs with fine reticulations, inner digits of forelimb with reduced pigmentation. Ventral surfaces off-white, immaculate.
Comparisons. D. platyurus is readily distinguished from D. conspicillatus, D. laevis, D. hillii, D. bilybara sp. nov., D. custos sp. nov. and D. barraganae sp. nov. by the condition of the 1 st supralabial (small and not differentiated from the rest of the supralabial row in D. platyurus vs greatly enlarged and contacting ventral edge of nasal scale) and by the absence of a well-defined canthal stripe (vs canthal stripe well-developed).
Distribution and Ecology. Occurs over much of eastern and central Queensland, from the Normanton and around Cairns in the north, south to around Rockhampton in the east, and throughout much of the channel country to west of the Great Dividing Range, extending south as far as north-west New South Wales and north-east South Australia (Fig. 3). Occurs in subhumid to arid woodland habitats on a range of sand and clay based substrates (A. Emmott pers. com).
Comments. A black and white photographic image of the holotype of Diplodactylus platyurus (BMNH 1946.8.11.38; Fig. 11) was kindly provided by Dr Harold Cogger. The specimen, from Torrens Ck, Qld (21°25'S, 145°14'E) has an undifferentiated supralabial row (i.e. no enlarged supralabials) a character that is only found in the most easterly populations of the D. conspicillatus group occurring in Queensland and NSW. A specimen from Torrens Ck, QM J47527 (the type locality), displaying this character is included in the material examined.
The taxonomic assignment of two specimens from the Edward River region on western Cape York Peninsula (QM J58251 Melon Yard, Strathgordon H, 14°43'12"S, 142°18'E and QM J81110 Edward River, 14°24'36"S, 142°09'36"E) remains unresolved. Whilst this population is geographically most proximate to D. platyurus, these specimens have an enlarged 1 st supralabial and may represent an additional taxon not included in our limited genetic sampling.  Pattern (in spirit). Tan to mid-brown, suffused with darker pigment on back and flanks. Pattern incorporates diffuse spotting and obscure reticulations and a pale, continuous or broken, vertebral zone. Head with numerous dark scales that often form a fine netted pattern. A moderately well-developed pale canthal stripe present, extending from anterior edge of orbit to tip of snout and producing a distinctive 'v' shaped marking. A diffuse dark zone on side of face extends posteriorly beyond eye to temporal region. Limbs obscurely marked with vague spotting or netted pattern and inner digits of fore and hindlimb with reduced pigmentation. Ventral surfaces off-white, immaculate.
Comparisons. Diplodactylus barraganae sp. nov. is readily distinguished from D. platyurus in possessing an enlarged first supralabial that contacts the ventral edge of the nasal scale (vs 1 st supralabial small and not differentiated from the rest of the supralabial row). It is distinguished from D. conspicillatus, D. laevis, D. bilybara sp. nov. and D. custos sp. nov. in having small mid-dorsal scales that are only slightly larger than the dorsolaterals (vs mid-dorsals enlarged and plate-like, conspicuously larger than the dorsolaterals) and further distinguished from D. laevis, D. bilybara sp. nov. and D. custos sp. nov. by the shape of the original tail (tail blunt, spade-like without an acute attenuated extension at tip in D. barraganae sp. nov. vs tail with an acute attenuated extension at tip).
Distribution and Ecology. Occurs over a broad band along the southern edge of the Gulf of Carpentaria, from the Roper River region in the northwest, east and south as far as Mt Isa (Fig. 3) edge of nasal scale). Dorsal scales on trunk plate-like and markedly larger than smaller dorsolaterals. Scales on nape granular and only slightly larger than granules on side of neck. Original tail with a short to moderate, acute attenuated extension at tip; scales on dorsal surface of tail arranged in transverse rows (often in a pattern of one large row followed by two small rows; scales in the small rows * ¼ the size of the scales in the adjacent large rows. Pattern variable; reticulated or with obscure transverse bands and generally incorporating numerous small pale spots. Dark pigment on crown and snout contrast markedly with pale canthal stripe and lower jaw colour which extends posteriorly as a pale bar towards the ear opening.  small and undifferentiated from adjacent chin scales; eye large, pupil vertical with crenulated margin; ear small, round to horizontally elliptic. Neck: broad with small granular scales on dorsal surface that are only slightly larger than the adjacent scales on the lateral surfaces. Trunk: moderate and somewhat stout; scales of mid-dorsum plate-like and markedly larger than dorsolateral scales; granules small on ventral surface but increase in size on pectoral region; preanal pores absent; a small cluster of postanal tubercles present in both sexes but larger and more prominent in males. Limbs: moderate; forelimb 24.87-35.85% SVL (n = 28, mean = 31.13, SD = 0.02); hindlimb 27.46-38.83% SVL (n = 28, mean = 33.73, SD = 0.02); digits moderate with no or only slight distal expansion; subdigital lamellae granular (not a clearly defined series except for small distal pair which tend to be long and narrow); 8-15 lamellae beneath fourth finger (n = 28, mean = 11.46, mode = 13, SD = 1.69);10-17 lamellae beneath fourth toe (n = 28, mean = 13.18, mode = 13, SD = 1.52). Original tail: short, wide 41.16-58.1% tail length (n = 22, mean = 46.66, SD = 0.04), with a short to moderate, acute attenuated extension at tip (Fig. 6E); scales arranged in clear transverse bands which incorporate rows of both large and small scales (often in a pattern of one large row followed by two small rows, of which the scales in the small rows are much less than ½ the size of the scales in the large rows (Fig. 6E); each large scale bears a short blunt to sharp medial tubercle); 31-49 (n = 23, mean = 40.09, mode = 41, SD = 4.73) medial scale rows on tail from fracture plane (1 st autotomy septum) to tip; 10-15 (n = 24, mean = 13.21, mode = 14, SD = 1.25) rows of scales across original tail (large row at maximum width); ventral scales considerably smaller than dorsal scales. Regrown tail: with rounded distal end and more uniform scalation that is not arranged in clear transverse rows.
Pattern. Variable. Generally reddish-brown or grey. Most specimens with a series of irregular, dark wavy bands across back that usually extend across the vertebral zone (only one specimen, WAM R110027 has an unbroken, paler vertebral zone). There is usually some degree of fine spotting on back and flanks and in some specimens the spots extend across the dorsum in transverse rows. The delineation between the base colour and darker dorsal patterns ranges from moderate to sharply contrasting. Head generally with dark crown. A prominent, pale canthal stripe present, extending from anterior edge of orbit to tip of snout and producing a distinctive 'v' shaped marking which contrasts with the darker dorsal and lateral head markings. A broad dark zone on side of face extends posteriorly beyond eye to temporal region. A pale zone below eye extends to ear. Limbs mottled or spotted and inner digits of forelimb with reduced pigmentation. Ventral surfaces off-white, immaculate.
Comparison. Diplodactylus bilybara sp. nov. is readily distinguished from D. platyurus in possessing an enlarged first supralabial that contacts the ventral edge of the nasal scale (vs 1 st supralabial small and not differentiated from the rest of the supralabial row). It is distinguished from D. conspicillatus, D. hillii and D. barraganae sp. nov. by the shape of its original tail (tail with short to moderate, acute attenuated extension at tip in D. bilybara sp. nov. vs tail blunt, spade-like without an attenuated tip). It is distinguished from D. laevis by the condition of the scales on the nape and top of head (scales granular and not appreciably larger than those on sides of neck in D. bilybara sp. nov. vs scales plate-like, appreciably larger than those on the sides of the neck). D. bilybara sp. nov. is most like D. custos sp. nov. but differs from this species in the following respects: distal half of original tail with alternating rows of large and small scales (generally 1 large row followed by 2 small rows)-scales in the small rows * ¼ the size of the scales in the adjacent large rows vs tail scalation generally more uniform; if smaller scale rows present, these rarely form a double row and the small scales are * ½ the size of the scales in the adjacent large rows for D. custos sp. nov.; dark pigment on crown and snout contrast markedly with pale canthal stripe and lower jaw colour which extends posteriorly towards the ear as a pale bar vs dark pigment on crown and snout generally not contrasting sharply with pale canthal stripe and lower jaw colour for D. custos sp. nov., trunk heavily pigmented and pattern usually incorporating numerous small pale spots vs body pattern often diffuse and generally without numerous pale spots, usually with wavy, dark transverse bands across back for D. custos sp. nov..
Distribution and Ecology. Occurs in the Carnarvon, west Pilbara and west Gascoyne regions along the central west coast of Western Australia (Fig. 3)  Etymology. From the latin for guard, with reference to the Australian Wildlife Conservancy (AWC) and their ambitious and effective conservation and research programs in the Kimberley (where this species is endemic) and elsewhere in Australia. Used as a noun in apposition.
Diagnosis. A large member of the D. conspicillatus group (max SVL 61 mm) with a well-defined canthal stripe and a greatly enlarged first supralabial (first supralabial contacts ventral edge of nasal scale). Mid-dorsal scales on trunk plate-like and markedly larger than smaller dorsolaterals. Scales on nape granular and only slightly larger than granules on side of neck. Original tail with a short, acute attenuated extension at tip (Fig. 6F); scales on dorsal surface arranged in transverse rows generally of uniform size but if smaller scale rows are present, these rarely form a double row and the small scales are * ½ the size of the scales in the adjacent large rows (Fig. 6F). Dark pigment on crown and snout generally not contrasting sharply with pale canthal stripe and lower jaw colour. Body pattern often diffuse and generally without numerous pale spots; may incorporate wavy, dark transverse bands.
Description 71) all small and undifferentiated from adjacent chin scales; eye large, pupil vertical with crenulated margin; ear small, round to horizontally elliptic. Neck: broad with small granular scales on dorsal surface that are only slightly larger than the adjacent scales on the lateral surfaces. Trunk: moderate and somewhat stout; scales of mid-dorsum plate-like and markedly larger than smaller dorsolateral scales; granules small on ventral surface but increase in size on pectoral region; preanal pores absent; a small cluster of postanal tubercles present in both sexes but larger and more prominent in males. Limbs: moderate; forelimb 26.79-35.62% SVL (n = 13, mean = 31.06, SD = 0.03); hindlimb 28.13-39.82% SVL (n = 13, mean = 33.5, SD = 0.04); digits moderate with slight distal expansion; subdigital lamellae granular and not a clearly defined series (except for small distal pair which tend to be broadly oval and displaced laterally by claw); 9-16 lamellae beneath fourth finger (n = 15, mean = 11, mode = 11, SD = 1.60); 10-15 lamellae beneath fourth toe (n = 15, mean = 11.93, mode = 12, SD = 1.16). Original tail: short, wide 35-54.72% tail length (n = 10, mean = 46.0, SD = 0.05) with a short, acute attenuated extension on tip (Fig. 6F); Original tail with scales arranged in clear transverse rows which are largely of uniform size (where rows of smaller scales occur, they are ½ the size of the scales in the larger scale rows) each scale bearing a bluntly-tipped medial tubercle (Fig. 6F); 28-41 (n = 12, mean = 33.42, mode = 32, SD = 3.90) medial scale rows on tail from fracture plane (1 st autotomy septum) to tip; 12-14 (n = 12, mean = 12.92, mode = 12, SD = 0.90) rows of scales across original tail (large row at maximum width); ventral scales considerably smaller than dorsal scales. Regrown tail: with rounded distal end and more uniform scalation that is not arranged in clear transverse rows.
Pattern. Variable. Tan to grey with darker overlay. Flanks and dorsum not strongly contrasting with ground colour and with or without pale spotting. Vertebral zone broken by dark, obscure to well-formed transverse bars. A pale canthal stripe present, extending from anterior edge of orbit to tip of snout and producing a distinctive 'v' shaped marking that does not contrast sharply with other facial markings. A dark zone on side of face extends posteriorly beyond eye to temporal region. Limbs mottled or spotted and inner digits of forelimb with reduced pigmentation. Ventral surfaces off-white, immaculate.
Comparison. Diplodactylus custos sp. nov. is readily distinguished from D. platyurus in possessing an enlarged first supralabial that contacts the ventral edge of the nasal scale (vs 1 st supralabial small and not differentiated from the rest of the supralabial row). It is distinguished from D. conspicillatus, D. hillii and D. barraganae sp. nov. by the shape of its original tail (tail with short attenuated tip in D. custos sp. nov. vs tail blunt, spade-like without an attenuated tip). It is distinguished from D. laevis by the condition of the scales on the nape and top of head (scales granular and not appreciably larger than those on sides of neck in D.custos sp. nov. vs scales plate-like, appreciably larger than those on the sides of the neck). D. custos sp. nov. is most like D. bilybara sp. nov. but differs from this species in the following respects: scalation of original tail reasonably uniform; if smaller scale rows present, these rarely form a double row and the small scales are * ½ the size of the scales in the adjacent large rows vs distal half of original tail with alternating rows of large and small scales (generally 1 large row followed by 2 small rows)-scales in the small rows * ¼ the size of the scales in the adjacent large rows for D. bilybara sp. nov.; dark pigment on crown and snout generally not contrasting sharply with pale canthal stripe and lower jaw colour vs dark pigment on crown and snout contrast markedly with pale canthal stripe and lower jaw colour which extends posteriorly towards the ear as a pale bar in D. bilybara sp. nov.; body pattern often diffuse and generally without numerous pale spots, usually with wavy, dark transverse bands across back vs trunk heavily pigmented and pattern usually incorporating numerous small pale spots for D. bilybara sp. nov. Additionally, D. custos sp. nov. usually has a shorter and less pronounced acute attenuated extension on the original tail tip than D. bilybara sp. nov.
Distribution and Ecology. Known from widespread but scattered localities from across the Kimberley region of north-western Australia, ranging from Kununurra in the north-west, south to Purnululu National Park, to around Derby in the south-west, with additional records from the Yampi Peninsula and high rainfall zone of the north-west Kimberley. Has been recorded from Koolan Island off the west Kimberley, the only insular record of a member of the D. conspicillatus complex (Fig. 3).
A single specimen from Ellenbrae Station in the central Kimberley was collected from open Eucalyptus woodland on the top of a stony rise with heavy clay soils, while specimens from around Kununurra were collected from rocky hillsides vegetated with open grassy woodland (P. Oliver pers obs).

Key to the Diplodactylus conspicillatus species group
Supporting Information S1 Fig. Maximum Likelihood Phylogeny for complete dampling of the Diplodactylus conspicillatus complex and outgroups. Estimated from mitochondrial ND2 data using RAxML with Maximum Likelihood support Boostrap supports shown for key nodes. (PDF) S1 Table. A. Measures of inter-specific genetic diversity and divergence. B. Intra-specific measures of genetic diversity and divergence. C. Diversity and demographic summary statistics.