The unexpected survival of an ancient lineage of anseriform birds into the Neogene of Australia: the youngest record of Presbyornithidae

Presbyornithids were the dominant birds in Palaeogene lacustrine assemblages, especially in the Northern Hemisphere, but are thought to have disappeared worldwide by the mid-Eocene. Now classified within Anseriformes (screamers, ducks, swans and geese), their relationships have long been obscured by their strange wader-like skeletal morphology. Reassessment of the late Oligocene South Australian material attributed to Wilaru tedfordi, long considered to be of a stone-curlew (Burhinidae, Charadriiformes), reveals that this taxon represents the first record of a presbyornithid in Australia. We also describe the larger Wilaru prideauxi sp. nov. from the early Miocene of South Australia, showing that presbyornithids survived in Australia at least until ca 22 Ma. Unlike on other continents, where presbyornithids were replaced by aquatic crown-group anatids (ducks, swans and geese), species of Wilaru lived alongside these waterfowl in Australia. The morphology of the tarsometatarsus of these species indicates that, contrary to other presbyornithids, they were predominantly terrestrial birds, which probably contributed to their long-term survival in Australia. The morphological similarity between species of Wilaru and the Eocene South American presbyornithid Telmabates antiquus supports our hypothesis of a Gondwanan radiation during the evolutionary history of the Presbyornithidae. Teviornis gobiensis from the Late Cretaceous of Mongolia is here also reassessed and confirmed as a presbyornithid. These findings underscore the temporal continuance of Australia’s vertebrates and provide a new context in which the phylogeny and evolutionary history of presbyornithids can be examined.

, as well as the holotype of Presbyornis mongoliensis and the collection of Presbyornis sp. specimens from the early Eocene of Mongolia [20] were examined at PIN. Telmabates antiquus was assessed based on casts at PIN, as well as the original description and images [13] and on Ericson's diagnosis of the species [17]. Lithornithidae (Palaeognathae) were also considered for comparative purposes.

Extant comparative material
The following specimens were examined: Anhimidae: Anhima cornuta ( Fossil and living phoenicopteriforms (flamingos and allies) were also examined.

Remarks on the material previously attributed to Wilaru tedfordi by Boles et al.
Paratype humeri AMNH 11407 and AMNH 11406 of the type species W. tedfordi are part of the same bone and now joined. Coracoid AMNH 11414 is a left one, contra Boles et al. [36]. Right coracoid SAM P.23625 is now a paratype of a new species (see below) and is here removed from the referred material of W. tedfordi 3. Systematic palaeontology 3. 1. Order Anseriformes Wagler, 1831 3. 1. 1. Family Presbyornithidae Wetmore, 1926 In his revision of the New World fossil record of the Presbyornithidae, Ericson [17] recognized four species, Telmabates antiquus Howard, 1955 [13] from the early Eocene of southern South America (Chubut, Argentina), and three from the late Palaeocene to early Eocene of North America: Presbyornis pervetus Wetmore, 1926 [12], Presbyornis isoni Olson, 1994 [14] and Presbyornis recurvirostra (Hardy, 1959 [38]). Two species were later described from Mongolia: the Late Cretaceous Teviornis gobiensis Kurochkin, Dyke, & Karhu [28], and the early Eocene Presbyornis mongoliensis Kurochkin & Dyke [20

Remarks on charadriiform affinities
The skeletal morphology of burhinids is possibly derived within Charadriiformes [39], and some of the features mentioned by Boles et al. [36] do occur in burhinids but not in most other charadriiforms. A similarity between the postcranial skeleton of presbyornithids and burhinids had been noted in the past (e.g. [40]). The ecological disparity within Charadriiformes has resulted in a wide array of morphological variation, although it is unlikely that burhinids represent the ancestral charadriiform morphology [41,42]. Presbyornithids have been linked, primarily based on postcranial morphology, to both flamingos (Phoenicopteriformes) and shorebirds (Charadriiformes) (e.g. [13,43]). The type species P. pervetus was originally considered to be a recurvirostrid [12] and it was not until the mid-1970s when additional cranial material was assessed that anseriform affinities were proposed [24,40,44] (see [19] for a review).

Amended diagnosis
We here focus on features that differentiate presbyornithid genus-level taxa. Most presbyornithid characters mentioned by Howard [13] and Ericson [17] are not repeated here but newly defined or identified features are. Apart from those mentioned above for Presbyornithidae, Wilaru is further characterized by the following combination of features: Humerus with (1) dorsal crus continuous with margo caudalis (figure 1c; as in Telmabates); (2) incisura capitis undercuts caput humeri and tuberculum ventrale (figure 1c; as in Telmabates, only caput humeri            a Specimen SAM P.42004 is overall worn and has a slightly broken trochlea carpalis so its proximal width is likely to have been greater. The processus extensorius is nevertheless well developed.

Description and comparisons
Coracoid. As in all presbyornithids, a small foramen nervi supracoracoidei nearly adjacent to the cotyla scapularis is present ( figure 1i,l,o), the impressio lig. acrocoracohumeralis is distinctly excavated on its medial side ( figure 1i,l,o), there is an elongated depression for a ligamental attachment on the ventral side of the brachial tuberosity ( figure 1j,l,o), and the cotyla scapularis is round and very deep ( figure 1i,l,o). In P. pervetus, the facies articularis clavicularis markedly overhangs the shaft ventrally and encloses therein a distinct fossa ventral to the sulcus m. supracoracoidei (figure 1p). This fossa is absent in species of Wilaru (figure 1m). It is also absent in some of the Eocene presbyornithid specimens from Mongolia [20], and may only be very shallow in T. antiquus [13]. As in presbyornithids, the sulcus m. supracoracoidei is excavated under the dorsal part of the facies articularis clavicularis (figure 1o,p). The ventral shaft margin of the sulcus is thickened and rounded, ventrally a ridge is also present but contrary to species of Presbyornis, it overlaps the ventral profile in medial view (figure 1m). Transverse linear ridges within the impressio sternocoracoidei are present (figure 1i), as in W. tedfordi and all Anseriformes. As in W. tedfordi and T. antiquus, the ventral surface of the sternal end lacks the depression that is observed in species of Presbyornis (figure 1n).
The coracoid of W. prideauxi is stouter than that of W. tedfordi (cf. figure 1i,j with figure 1k), which in turn is only slightly stouter than that of P. pervetus (figure 1n). Presbyornithids are characterized by having the neck of the shaft narrow [17], whereas it is much broader in the superficially similar phoenicopteriforms (flamingos and palaelodids). Similar to species of Presbyornis, Palaelodus ambiguus bears a distinct fossa ventral to the sulcus m. supracoracoidei, which, as noted, is absent in species of Wilaru. Palaelodids are further distinguished from species of Wilaru by having a cranially directed processus procoracoideus (right angles to axis in Wilaru), a foramen n. supracoracoidei that is closer to the margin of the cotyla scapularis, and a much smaller impressio sternocoracoidei restricted to the sternal third of the length from the cotyla scapularis.
Carpometacarpus. The bone (figure 2a,b) is overall more robust than that of W. tedfordi (figure 2c,d), but many features are worn. The area immediately cranial of the processus pisiformis is shallowly excavated and the fossa infratrochlearis is shallow and limited to the area proximal of the processus pisiformis (figure 2c). In P. pervetus there is a proximally directed ledge separating the processus extensorius from the fossa infratrochlearis, which extends caudally past the processus pisiformis (figure 2h). The fovea carpalis cranialis is deeper in P. pervetus (figure 2l) than in species of Wilaru (figure 2j,k). As in W. tedfordi and P. pervetus, both facies of the trochlea carpalis are smaller than in the Cretaceous T. gobiensis, in which they extend more cranially (figure 2f ,g). As in all presbyornithids, caudally the dorsal and ventral rims of the trochlea carpalis have equal distal extension (figure 2j-m), and the sulcus tendineus is elongate, proximally nearly reaching the scar for the insertion of m. extensor carpi ulnaris ( figure 2c,i). The sulcus tendineus is also particularly elongated in T. gobiensis and screamers (Anhimidae), whereas it is somewhat intermediate in length between these taxa and anatids (ducks, geese and swans) in the magpie goose Anseranas semipalmata.
The synostosis metacarpalis distalis ( figure 2a,b) is relatively shorter than that of W. tedfordi, being considerably longer in P. pervetus (figure 2e). The facies articularis digitalis minor has equal distal extent to the facies articularis digitalis major and so it projects further distally (figure 2b) than in W. tedfordi (figure 2d). In W. tedfordi and T. antiquus, the facies articularis digitalis minor ends slightly proximally of the facies articularis digitalis major ( figure 2c,d). Further detail of the distal end is obscured by wear but does not seem to differ from that of W. tedfordi. Tarsometatarsus. The proportions of the tarsometatarsus of species of Wilaru [36] differ greatly from those of P. pervetus in being much shorter and stouter [12]. The length of the tarsometatarsus is, however, not known for other species of presbyornithids. Within anseriforms, the tarsometatarsus of species of Wilaru most closely resembles that of anhimids, which they resemble in overall relative proportions ( figure 2q,r), morphology of the hypotarsus ( figure 2a ,b ,d ) and configuration of the distal trochleae ( figure 2u-y).
At least one sulcus hypotarsi is present in species of Wilaru; the crista medialis hypotarsi is missing from specimens of both W. tedfordi and W. prideauxi (figure 2a ,b ). The hypotarsus is overall reduced compared to that of P. pervetus (figure 2c ), which has four hypotarsal ridges (as does T. antiquus [13]), with the crista medialis hypotarsi being well marked and the other three cristae less so. In species of Wilaru, other than the missing crista medialis hypotarsi, there is one well-marked crista intermedia hypotarsi (figure 2s) and the rest have been reduced to a flat embossment in the lateral portion of the hypotarsus. A similar condition can be observed the anhimid C. torquata, although in A. cornuta the large crista medialis hypotarsi is separated from a much smaller crista lateralis hypotarsi by two very low cristae intermediae hypotarsi which altogether form a triangular hypotarsus. As in P. pervetus, the cristae lateral to the medial crest are of similar small size and reach equally distally (figure 2s). The eminentia intercotylaris is not especially prominent proximally in species of Wilaru, but it is more prominent dorsally (figure 2p,q,a ). The mid-shaft depth equals its width in W. prideauxi, whereas in W. tedfordi its width slightly exceeds its depth.
At the distal end, the trochlea metatarsi II is lacking so it is not known whether a medial groove was present. Within Anseriformes, this groove is absent in presbyornithids (including W. tedfordi, figure 2u), anhimids (figure 2v), and anseranatids (magpie goose). In P. pervetus (figure 2w,z) and P. mongoliensis, the trochleae are more narrowly splayed than they are in species of Wilaru, where their divergence is similar to that of anhimids (figure 2v) or A. semipalmata. The incisura intertrochlearis medialis and the incisura intertrochlearis lateralis extend equally proximally in species of Wilaru, whereas in both anhimids and A. semipalmata the medial notch is shallower proximally (figure 2v). Therefore, trochlea metatarsi II has less distal extent than trochlea metatarsi IV (figure 2u). In W. tedfordi, the trochlea metatarsi II is only slightly retracted plantarly so in distal view, most of its depth overlaps the trochlea metatarsi III, and thus is less retracted than in anhimids and A. semipalmata but similar to the terrestrial Cape Barren goose C. novaehollandiae. The fossa metatarsi I is barely perceptible in W. prideauxi (figure 2r), but it is better marked in W. tedfordi (figure 2x), in which it is situated much lower compared with P. pervetus (figure 2z).

Results and discussion
Crown-group Anseriformes comprise three extant family-level taxa: the South American screamers (Anhimidae), the magpie goose (Anseranatidae) of Australia and New Guinea, and the cosmopolitan Anatidae (ducks, swans and geese). A sister group relationship between Anhimidae and the clade (Anseranatidae + Anatidae) is supported by molecular and morphological evidence (e.g. [47,48]). Presbyornithids have been recovered as the sister taxon to Anatidae in cladistic analyses [40,47], but the character evidence supporting this relationship is weak [19]. These studies have been primarily based on the morphology of P. pervetus, and the possibility that some features of this taxon are derived within Presbyornithidae [19] has not been fully explored. Some of the similarities of species of Wilaru (and T. antiquus) to anhimids provide a new context in which the palaeobiology and evolutionary history of presbyornithids can be examined.

Teviornis gobiensis, a Cretaceous presbyornithid
Teviornis gobiensis, known primarily from its carpometacarpus ( figure 2f ,g,m,o), was attributed to the Anseriformes mainly based on its straight os metacarpale minus, and to the Presbyornithidae based on (i) the caudal part of the dorsal rim of the trochlea carpalis being well developed and connecting with the dorsal edge of the os metacarpale majus, (ii) the presence of well-developed scars for lig. ulnocarpometacarpale dorsale (fossa supratrochlearis) and lig. radiocarpometacarpale dorsale (fossa infratrochlearis), and (iii) the presence of a small canalis interosseus distalis in the fossa infratrochlearis [28]. Clarke & Norell [29] challenged presbyornithid, and even anseriform affinities of T. gobiensis, noting that of the diagnostic characters listed a straight minor metacarpal may be a plesiomorphic feature of Neornithes, that features (ii) and (iii) are present in other anseriforms, and (i) is also present in anhimids. Clarke and Norell did not, however, consider all the distinguishing features mentioned in the description. A recent study [49] supported anseriform affinities of this taxon, noting marked differences from other taxa with a non-curved carpometacarpus (e.g. Gallinuloididae, Lithornithidae). Additional features confirm the identity of T. gobiensis as a presbyornithid, namely (iv) the dorsal and ventral rims of the trochlea carpalis extend caudally and distally to about same level, (v) the sulcus tendineus is very elongate, extending just about to the distal end of the scar for M.
extensor carpi ulnaris, (vi) in distal view, the facies articularis digitalis minor is considerably smaller than the facies articularis digitalis major (as in Wilaru and Telmabates), and (vii) the synostosis metacarpalis proximalis is longer than it is craniocaudally wide. Within anseriforms, feature (iv) is present only in presbyornithids, (v) in presbyornithids and anhimids, (vii) is present in presbyornithids, anhimids, anseranatids, and only few anatids, whereas (vi) is widely distributed within Anseriformes but the alternate condition is present in P. pervetus.
Kurochkin et al. [28] noted that T. gobiensis differed from other presbyornithids in having a fossa infratrochlearis stretched markedly craniocaudally. A shallow, craniocaudally elongated fossa is nonetheless also present in species of Wilaru (figure 2d). Similarly, the dorsoventrally and craniocaudally widened proximal portion of the os metacarpale minus are present in species of Wilaru but also in anhimids and other anseriforms, suggesting they could be plesiomorphic features for Presbyornithidae that are absent in P. pervetus. T. gobiensis, therefore, displays a combination of features of the carpometacarpus present uniquely in presbyornithid genus-level taxa.

Palaeobiology of species of Wilaru: terrestrial and territorial
The cranially elongated extensor process of the carpometacarpus of some specimens of W. tedfordi forms a conspicuous rugose enlargement ( figure 2c,d,k), known as a carpal knob or spur. Carpal knobs are projecting bony cores used primarily in fighting, which in some taxa may have an outer layer of horn [50] but can be bare in others [51]. These rugose structures arise from the deposition of bone on the extensor process, to which they are fused. Carpal knobs and spurs occur in several anseriforms [51], most notably in steamer ducks [52] and anhimids [50]. Within anseriform species that bear them, they are better developed in males, but are still present in females [50]. Well-developed carpal knobs, similar to those of the male paradise shelduck T. variegata, were present in three specimens of our sample, which were also slightly larger compared to the rest (table 1). Both sexual dimorphism and age probably explain these differences [51]. Some of the specimens with the less protruding extensor process still displayed a form of rugose enlargement. From this we infer that, following the pattern in other anseriforms, those individuals of W. tedfordi with the well-developed carpal knobs are males. Anseriform species that bear prominent structures tend to engage in aggressive behaviour and hold year-round feeding and breeding territories [51]. The lack of well-developed carpal knobs in most of the specimens in our sample suggests that predominantly males may have engaged in aggressive behaviour. Howard [13] observed a 'slight excrescence' on the tip of the extensor process of T. antiquus, and illustrations clearly show [13, fig. 6, p. 16], a small carpal knob. This structure is not present in P. pervetus or the Presbyornis specimens from the Eocene of Mongolia [20], which closely resemble each other.
Compared with other presbyornithids, the morphology of the tarsometatarsus, with mediolaterally splayed trochleae for the articulation of the toes, a less plantarly retracted trochlea for the second digit, and a relatively low hallux, suggests that species of Wilaru were more terrestrial. Screamers, which have a similar tarsometatarsal morphology, are birds of predominantly terrestrial habits, frequenting open savannahs and wetlands (meadowlands, marshes, swamps and lakes with abundant vegetation) [53]. The tarsometatarsus is proportionally longer and more gracile in P. pervetus, in which the narrowly divergent trochleae and the markedly retracted trochlea for the second digit indicate more aquatic adaptations, as in most anseriforms (e.g. [17,44]). The length and most detail of the tarsometatarsus are not known for T. antiquus, but it appears to have resembled that of P. pervetus [17].
Terrestrial habits have evolved independently several times within Anseriformes [54,55], and even possibly within screamers [17,44] ( §4.3). The terrestrial habits of species of Wilaru may, therefore, reflect a trophic specialization (such as herbivory) derived within Presbyornithidae. These differences are not surprising given the apparent temporal separation of ca 25 Ma between species of Wilaru and other presbyornithids, and may have been the key to the longevity of the presbyornithid lineage in Australia ( §4.3).
The younger, larger and more robust W. prideauxi represents a further step from the morphology of W. tedfordi down the path of terrestriality. As such, species of Wilaru provide the first example for Australia of two successive species within an avian lineage in the Oligo-Miocene. Multiple lineages of mammals are known over this time period in the Namba and Etadunna Formations in Australia and form the foundation of the biochronological understanding of the different faunas [1,56]. The occurrence of W. prideauxi in the Ngama Local Fauna from Zone D of the Etadunna Formation at Lake Palankarinna, and in the Kutjamarpu Local Fauna from the Wipajiri Formation at Lake Ngapakaldi, South Australia, provides further evidence of the contemporaneity of these local faunas, otherwise linked by mammals, and supports their early Miocene age and distinction from underlying Etadunnan faunal zones

The role of Gondwana and the evolutionary history of Wilaru
Several species of anatids were described from late Oligocene and early Miocene deposits of the Namba and Etadunna formations in South Australia [57], supporting an already established diversity of crowngroup anatids by the late Oligocene. Having been recovered from the same late Oligocene and early Miocene localities, the survival of presbyornithids in Australia into the Neogene indicates they were living alongside crown-group anatids. Presbyornithids seem to have disappeared from the rest of the world during the Eocene [19], coinciding with the earliest records of stem group anatids. However, the more terrestrial adaptations of species of Wilaru suggest that in Australia, they may not have been in direct resource competition with coeval waterfowl. Similarly, at least two palaelodids and two species of flamingo cornered the wading niche in the lakes in which these deposits were formed [5,58]. The causes of the ultimate demise of species of Wilaru after the early Miocene are unknown, but as in the case of much of Australia's fauna, climate change and the progressive aridification of the continent may have played a role, especially if species of Wilaru were territorial and dependent on specific habitats for breeding.
Fossils from the early Eocene Tingamarra Local Fauna, Queensland, were tentatively referred to the form-taxon Graculavidae ('transitional shorebirds' [59]) [32], but it has been acknowledged that some of the material may in fact be presbyornithid [30,32]. Indeed, the coracoid, fragment of humerus, and one distal tibiotarsus tentatively attributed to the 'Graculavidae' were recognized by Boles [32] as remarkably similar to P. pervetus, and we further note the marked similarity with the corresponding elements of W. tedfordi and other members of the Presbyornithidae. Further assessment of this material will help establish if presbyornithids have been in Australia since at least the early Eocene.
Australia's long period of geographical isolation, from complete separation from Antarctica to a close proximity to the Indo-Malayan region (ca 50-15 [60,61]), has certainly promoted the extended temporal continuance of its fauna. The presence of presbyornithids in the early Miocene of Australia, therefore, ought not to appear all that surprising. Within mammals, marsupials and monotremes have survived in Australia since at least the early Eocene [62] and late Oligocene [63], respectively, long after most lineages disappeared elsewhere in the world (e.g. [64]). A similar pattern can be observed among birds, as the globally distributed Palaelodidae, which first appear in the fossil record during the early Oligocene, survived in Australia until the mid-Pleistocene [5], and the plains-wanderer lineage, which was once more widespread but has been on the continent since at least the late Oligocene [8,11], still remains in Australia with a sole representative, Pedionomus torquatus. Anseranatids, arguably known from the early Eocene and late Oligocene of Europe [19] and nowadays represented only by A. semipalmata, survive in Australia and New Guinea, having been recorded in Australia since the late Oligocene [7]. We note that although members of Anseranatidae are known from similar-aged deposits in Australia, they differ from presbyornithids in the morphology of most skeletal elements, but especially the humerus and coracoid.
Within Presbyornithidae, the postcranial morphology of W. tedfordi and W. prideauxi agrees with the South American T. antiquus in nearly all elements ( §3. 2.1), the tarsometatarsus being the exception. On the other hand, P. pervetus resembles T. antiquus in some features more than it does W. tedfordi, but mainly in the morphology of the tarsometatarsus. The overall close similarity between W. tedfordi and T. antiquus may lend support to the hypothesis that at least some aspects of the morphology of P. pervetus may be derived within Presbyornithidae (see also [19]). However, the uncertainty, at least for the time being, as to whether these similarities are derived or plesiomorphic both within Anseriformes and Presbyornithidae precludes a well-informed phylogenetic hypothesis. Assessing potential presbyornithid material from the Eocene of Australia, the discovery of cranial material of species of Wilaru, and the assessment of early-diverging anhimids (thought to be more Presbyornis-like [40]), may all contribute to clarifying the phylogenetic relationships between the different presbyornithid taxa.
Although conceiving a historical biogeographical scenario with the evidence at hand may be premature, the morphological similarity between species of Wilaru and T. antiquus emphasizes the role of Gondwana during the evolutionary history of the Presbyornithidae. This raises the possibility that members of the genera Wilaru, Telmabates and Presbyornis had a common ancestry on the southern landmasses, or at least that there was one Gondwanan radiation within Presbyornithidae including T. antiquus and species of Wilaru. Because only very few elements are known for the Cretaceous T. gobiensis, assessment of its position within Presbyornithidae, and the role Laurasia played during the early evolutionary history of presbyornithids will need to await the discovery of additional material. The morphology of P. mongoliensis and that of most elements attributed to Presbyornis sp. by Kurochkin & Dyke [20], some of which are probably attributable to P. mongoliensis, agrees well with the morphology

Comments on presbyornithid relationships
There are marked differences between screamers and presbyornithids in major skeletal elements that extend beyond the highly pneumatic nature of anhimid limb bones, e.g. the humerus of anhimids has an inflated shaft and a pneumatic fossa pneumotricipitalis, the coracoid has a very reduced acrocoracoid with an unusually shallow sulcus supracoracoideus, and the carpometacarpus has a unique spurlike development of the processus extensorius. Despite this, many of the postcranial features we used in this study reflect interesting similarities with the Anhimidae, despite the derived aberrant morphology of this taxon [44]. Within anseriforms, characters that are shared between Anhimidae and Presbyornithidae, and not present in other Anseriformes (i.e. Anseranatidae and Anatidae), include (1) the presence of a deep fossa in the incisura capitis of the humerus; (2) both articular rims of the trochlea carpalis of the carpometacarpus extending caudally and distally to about same level; (3) a very elongate sulcus tendineus, extending to the distal end of the scar for m. extensor carpi ulnaris; (4) a very prominent linea intermuscularis cranialis of the femur, separated from the lateral margin of the crista trochanteris (in P. pervetus this line runs closer to crista trochanteris); (5) a much smaller, mediolaterally and proximodistally, medial condyle of the tibiotarsus compared to the lateral condyle; and (6) a hypotarsus with one well-developed sulcus hypotarsi and one well-marked crista intermedia hypotarsi with the rest reduced to a near flat embossment in the lateral portion of the hypotarsus (not present in P. pervetus and T. antiquus). Ericson [40] further noted the presence of pleurocoelous thoracic vertebrae in presbyornithids and screamers, but because of the pronounced pneumaticity in the skeleton of extant anhimids whether this feature is indeed homologous for the two remains to be ascertained.
Some of the characters listed above may be plesiomorphic for Anseriformes (e.g. characters 2, 3, and 5), but whether that is true of all features or whether some could be synapomorphic for a clade including (Anhimidae + Presbyornithidae) needs to be established in a cladistic framework. A yet undescribed Eocene representative of the Anhimidae may shed some light on the subject [66]. Although not formally described, this early Eocene specimen from Wyoming was briefly assessed by Ericson [40], who noted that several elements of the postcranial skeleton closely match those of presbyornithids, including the carpometacarpus, coracoid, furcula and tibiotarsus (see also [32]). Chaunoides antiquus from the late Oligocene-early Miocene of Brazil [67] is known from several fragmentary bones, and despite it being morphologically very similar to extant anhimids, the extreme pneumatisation of the skeleton that characterizes modern anhimids is absent. There is, therefore, a strong possibility that screamers are derived from presbyornithid-like birds or that they had a common ancestor. Although living screamers have a galliform-like hooked bill, it has been proposed that the rudimentary lamellae on the bill of anhimids indicates a secondary loss of the filter-feeding ability [44,68], and therefore the skull of early-diverging anhimids may have been more 'anseriform-like' than that of extant anhimids [40]. The occurrence of lamellae-like structures, however, may not necessarily be linked to filter-feeding, as aquatic herbivores may use these structures for grasping and cutting plants [54].
In any case, the recognition of W. tedfordi as a presbyornithid certainly calls for a reassessment of the phylogenetic position of Presbyornithidae within Anseriformes (see also [69][70][71]), which may be closer to the base of the anseriform tree than previously assumed [40,47,72]. Similarly, the phylogenetic placement of anseriform fossil taxa that have been based on cladistics analyses with limited taxon sampling and including P. pervetus alone, such as that of the Late Cretaceous Vegavis iaai from Antarctica [72], should be revised in the light of these new findings.

Conclusion
In this study we show that, contrary to previous reports, W. tedfordi from the late Oligocene of South Australia was not a burhinid (Charadriiformes), but a representative of the Presbyornithidae (Anseriformes). Additionally, we describe a slightly larger and more robust species of Wilaru, W. prideauxi, from the early Miocene of South Australia. This record extends the temporal continuance of presbyornithids by at least 25 million years, as they were believed to have disappeared from the fossil record by the early middle Eocene.
Unlike other presbyornithids, species of Wilaru were predominantly terrestrial birds, as indicated by the morphology of their tarsometatarsus. This adaptation likely contributed to their long-term survival in Australia, where they may have been present since at least the early Eocene, and where they lived alongside aquatic members of crown-group Anatidae. The presence of a bony excrescence on the extensor process of the carpometacarpus, linked with aggressive behaviour and also present in the presbyornithid T. antiquus and many extant anseriforms, may indicate that species of Wilaru were highly territorial.
The morphological similarity between species of Wilaru and the South American T. antiquus ( §3. 2.1) not only suggests a close relationship between the two, but also emphasizes the previously unexplored role of Gondwana in the evolutionary history of the Presbyornithidae, raising the possibility of a Gondwanan origin for the group, or at least a Gondwanan radiation. Similarly, the skeletal resemblance of the North American P. pervetus and the Mongolian P. mongoliensis (including specimens attributed to Presbyornis sp. from Mongolia) possibly indicates that these Northern Hemispheric species were more closely related to each other than to the Gondwanan T. antiquus and species of Wilaru. The phylogenetic affinities of T. gobiensis from the Late Cretaceous of Mongolia, here confirmed as a presbyornithid, remain obscure.
Although screamers (Anhimidae) may have evolved from presbyornithid-like birds, the uncertainty as to whether skeletal features shared by presbyornithids and anhimids are plesiomorphic within Anseriformes or indicative of a close relationship between the two cannot yet be resolved. In any case, recognition of W. tedfordi as a presbyornithid calls for a reassessment of the phylogenetic position of Presbyornithidae within Anseriformes.