A new species of Mukupirna (Diprotodontia, Mukupirnidae) from the Oligocene of Central Australia sheds light on basal vombatoid interrelationships

Abstract

The late Oligocene taxa Marada arcanum and Mukupirna nambensis (Diprotodontia, Vombatiformes) are the only known representatives of the families Maradidae and Mukupirnidae, respectively. Mukupirna nambensis was described from a partial skeleton, including a cranium but no dentary, and reconstructed as the sister taxon to Vombatidae (wombats). By contrast, Ma. arcanum is known only from a single dentary, preventing direct comparison between the two. Here, we describe a new species, Mu. fortidentata sp. nov., based on craniodental and postcranial specimens from the Oligocene Pwerte Marnte Marnte Local Fauna, Northern Territory, Australia. Phylogenetic analysis of Vombatiformes, using 124 craniodental and 20 postcranial characters, places these three species within Vombatoidea, wherein Marada arcanum is sister to species of Mukupirna + Vombatidae. Mukupirna fortidentata sp. nov. does not share any robust synapomorphies of the dentary with Ma. arcanum that would support placing them together in a clade to the exclusion of Vombatidae. We therefore maintain separation of the families Mukupirnidae and Maradidae. From a functional perspective, the craniodental specimens of Mu. fortidentata sp. nov. reveal a suite of morphological traits that are unusual among vombatiforms, which we interpret as adaptations for acquiring and processing hard plant material. These include: a short, broad rostrum; large, robust, steeply upturned incisors; and a steep, anteroposteriorly decreasing gradient in cheek tooth size. The dental specimens of Mu. fortidentata sp. nov. also assist in the identification of two further allied taxa: an early vombatid from the younger late Oligocene Tarkarooloo Local Fauna, South Australia; and a possible vombatoid from the earliest Miocene Geilston Bay Local Fauna, Tasmania. The Tarkarooloo Local Fauna taxon indicates that vombatids diverged from other vombatoids prior to 24 million years ago.

Arthur I. Crichton [[email protected]], College of Science and Engineering, Flinders University, Bedford Park, Adelaide, 5042, South Australia;

Trevor H. Worthy [[email protected]], Aaron B. Camens [[email protected]], Adam Yates [[email protected]] Museum and Art Gallery of the Northern Territory, Alice Springs 0870, Northern Territory, Alice Springs, 0870 Australia;

Aidan M. C. Couzens [[email protected]], Department of Ecology and Evolutionary Biology, University of California, Los Angeles, CA 90095, USA;

Gavin J. Prideaux [[email protected]], Flinders University School of Biological Sciences, Palaeontology, Adelaide, 5001 Australia.

Keywords:

• Mukupirnidae
• Maradidae
• Vombatidae
• Vombatiformes
• late Oligocene
• phylogenetic analysis
• Pwerte Marnte Marnte

THREE SPECIES of wombat (Vombatidae) and the koala (Phascolarctidae) are the only extant representatives of a once diverse radiation of vombatiform marsupials. This suborder included the largest Australian marsupial herbivores, with six extinct families recognized, namely: Diprotodontidae, Palorchestidae, Ilariidae, Wynyardiidae, Maradidae and Mukupirnidae (e.g., Beck et al. 2022). Our view of the divergences between these clades, and Australian marsupial families more generally, has been concealed by the temporal gap in the known fossil record from the early Eocene (55 Ma) to late Oligocene (25 Ma) (e.g., Rich 1991, Black et al. 2012b). By the late Oligocene, all eight vombatiform families had already evolved uniquely specialized dentitions, exacerbating the challenge of resolving their interrelationships (e.g., Archer et al. 1999, Black 2007, 2012a, 2012b, Brewer et al. 2015, Gillespie et al. 2016, Beck et al. 2020, 2022).

Maradidae is currently the most poorly understood of vombatiform families, being represented by only one species, Marada arcanum Black, 2007, which was described and remains known from just a single dentary from the late Oligocene Hiatus Site, Faunal Zone A, of Riversleigh World Heritage Area, northwest Queensland (Black 2007). When described, Ma. arcanum was noted as expressing an unusual mix of derived and ancestral traits, though sharing greatest overall affinity to species of supposed wynyardiids within the genera Namilamadeta and Muramura. More recently, the family Mukupirnidae was erected to contain Mukupirna nambensis Beck, Louys, Brewer, Archer, Black & Tedford, 2020, which was described from a distorted cranium and partial postcranial skeleton from the late Oligocene Pinpa Local Fauna (LF) of Lake Pinpa, northeastern South Australia (Beck et al. 2020). The phylogenetic position of Mu. nambensis within Vombatiformes was assessed by Beck et al. (2020) using a morphological character matrix, which resolved it as the sister group to vombatids within Vombatoidea. Importantly, however, Ma. arcanum was not included in this analysis because it is known only from the dentary, which is unknown for Mu. nambensis. However, the similar size of Ma. arcanum and Mu. nambensis (as indicated by molar row length), coupled with similarities in molar morphology with species of Namilamadeta and Muramura, led Beck et al. (2020) to flag the possibility that they might belong within the same family, genus or even species.

A further possibly allied taxon, reported as Vombatomorphia? fam., gen. et sp. nov. by Murray & Megirian (2006) prior to the erection of the families Maradidae and Mukupirnidae, is represented by a lower m3 (NTM P2815-11) from the late Oligocene Pwerte Marnte Marnte fossil locality. It was subsequently suggested by Black et al. (2012b) that the tooth may belong to Marada arcanum.

Here we have described additional specimens referable to this unnamed vombatomorphian from the Pwerte Marnte Marnte LF, including its complete upper and lower cheek dentition, as well as several postcranial specimens. This material provides key information enabling appraisal of the relative phylogenetic positions of Mukupirnidae and Maradidae. Additionally, the functional morphology of the craniodental material from the new Pwerte Marnte Marnte taxon was examined to infer its dietary adaptations and, in turn, to better understand the selective pressures faced by early vombatoids prior to the evolution of hypselodonty. We also use this opportunity to reassess undescribed molar fragments from the late Oligocene Tarkarooloo LF, South Australia, and have reinterpreted the taxonomic affinity of an enigmatic partial dentary from the earliest Miocene Geilston Bay, Tasmania.

Materials and methods

Terminology

Molar position homology follows Luckett (1993). Molar cusp nomenclature follows Rich et al. (1978), with the exception of the structure reported therein as the hypocone, which is referred to as the metaconule following Tedford & Woodburne (1987). Premolar cusp nomenclature follows Pledge (2003). Mandibular terminology follows Stirton (1967). Higher-level systematic nomenclature follows Aplin & Archer (1987), with the exception of the use of the superfamily Vombatoidea for the clade that includes Vombatidae + Mukupirnidae following Beck et al. (2020), and the subordinal placement of Thylacoleonidae as Diprotodontia incertae sedis following Beck et al. (2022). We also use ?Wynyardiidae to refer to species within the genera Namilamadeta, Muramura and Ayekaye, because their inclusion within the family Wynyardiidae has never been robustly demonstrated, following Tedford et al. (1977), Rich & Archer (1979), Pledge (1987), and Megirian et al. (2004). Biostratigraphic nomenclature follows Woodburne et al. (1994), Archer et al. (1997), Travouillon et al. (2006) and Megirian et al. (2010). The age of vertebrate bearing-localities from the Namba and Etadunna Formations follows Woodburne et al. (1994) and Megirian et al. (2010), and those of Riversleigh World Heritage Area follow Archer et al. (1997) and Woodhead et al. (2016).

Specimen preparation and measurements

The specimens from the Pwerte Marnte Marnte fossil site (Fig. 1) that are described herein were recovered from ca 2 tonnes of limestone that was quarried from a small (3 m by 2 m) area on expeditions in 2014 and 2020, led by Aidan Couzens and Arthur Crichton, respectively.

Figure 1. Map of Australia depicting the locations of late Oligocene and early Miocene fossil sites from which vombatoids have been recovered. Abbreviations: NSW, New South Wales; NT, Northern Territory; Qld, Queensland; SA, South Australia; Tas, Tasmania; Vic, Victoria; WA, Western Australia; WHA, World Heritage Area.

The fossil locality preserves heavily fractured and distorted partial skeletal elements in a 0.5–1 m thick calcareous limestone conglomerate of densely concentrated non-diagnostic bone fragments and well-rounded mainly quartz pebbles (see Murray & Megirian 2006). Several poorly defined depositional layers are evident, loosely delimited by changes in pebble and bone fragment density. Within each depositional layer, the fossiliferous material is heavily mixed, with no association between elements. To the extent that it can currently be assessed, there is overlap in faunal composition between the layers, with little indication that the site preserves more than a single local fauna (A. Crichton, pers. obs., January 2023). The fossiliferous rock was processed during 2020 through 2022 at Flinders University using a combination of etching with acetic acid (5–10%) and mechanical approaches, e.g., rock saws and pneumatic micro-jack tools. The only exceptions are NTM P10438, collected by D. Megirian, P. Latz and H. Larson in 2005, and NTM P6371, collected by P. Murray, D. Megirian, J. & I. Archibald in 2003. The Lake Tarkarooloo specimens described herein were collected on an expedition led by Thomas Rich in 1976.

Measurements were made using Mitutoyo digital callipers (model No CD-8ʺC: Takatsu-ku, Kanagawa, Japan) and rounded to 0.1 mm. Morphological comparisons were made directly from the specimens or casts thereof using an Olympus SZX12 microscope, with exceptions drawing upon published descriptions and figures (Supplementary Data 1).

Phylogenetic analyses

Morphological matrix

A morphological dataset of 124 craniodental and 20 postcranial characters (Supplementary Data 2, 3) was constructed for representatives of the suborder Vombatiformes to assess the relationships of the new Pwerte Marnte Marnte taxon, and resolve interrelationships among other taxa, in particular between Mukupirna nambensis and Marada arcanum. The matrix was built primarily upon that of Beck et al. (2020), which, in turn, built on Brewer et al. (2015), Black et al. (2012a) and previous works therein identified. Scoring of cranial characters for Ramsayia magna (Owen 1872) follows Louys et al. (2022). To capture aspects of the dentary morphology considered by Black (2007) in her family-level distinction of Marada arcanum from other vombatiforms, several further characters were here adopted from Black & Archer (1997), Black (2007), Black et al. (2012a), and Brewer et al. (2015). Where possible, the scoring of characters taken from the literature was reassessed to verify the results of others and to ensure consistency. Forty-seven additional craniodental characters were included. The description of all characters and assessed states can be found in Supplementary Data 2 and 3, respectively. These were scored for 46 taxa, with the following amendments to the matrix of Beck et al. (2020): we excluded Wakaleo pitikantensis (Rauscher, 1987) due to the paucity of represented material (known from a damaged maxilla); we replaced ‘Ngapakaldia spp.’ with Ngapakaldia tedfordi Stirton, 1967, replaced ‘Muramura spp.’ with Mur. pinpensis Pledge, 2003 and Mur. williamsi Pledge, 1987; and included ‘Rhizophascolonus spp.’, Mukupirna fortidentata sp. nov., and Marada arcanum. The didelphimorphian Didelphis marsupialis Linnaeus, 1758 was coded as the outgroup, alongside the peramelemorphians Perameles bougainville Quoy & Gaimard, 1824 and Galadi speciosus Travouillon, Gurovich, Beck & Muirhead, 2010 and the burramyid Cercartetus lepidus Thomas, 1888 (after Beck et al. 2020). Multistate morphological characters perceived as representing morphoclines were ordered.

Phylogenetic inference

Undated Bayesian analysis of the morphological dataset was carried out in MrBayes 3.2.7a (Ronquist et al. 2012), using the Markov Chain Monte Carlo (MCMC) approach, with gamma rate variability implemented for morphological data maintaining the assumption that only variable characters were scored. The Bayesian analyses were run for 15 million generations, using four independent runs of four chains (one cold and three heated chains, with the temperature of the heated chains set to the default value of 0.2), sampling trees every 1000 generations and a burn-in fraction of 25%. The post-burn-in trees were summarized using a majority rule consensus of all compatible groups, with Bayesian posterior probabilities as support values.

Maximum parsimony analyses were performed on the morphological dataset, in TNT version 1.5 (Goloboff et al. 2008), following the methods of Beck et al. (2020). The tree search involved an initial ‘new technology’ search with sectorial search, ratchet, drift and tree fusing that was run until the same minimum tree length was found 1000 times. From these saved trees a ‘traditional’ search was applied using the tree bisection resection (TBR) swapping algorithm, with the resulting most parsimonious trees combined into a strict consensus tree. Support values for branch nodes were calculated using 2000 standard bootstrap replicates, implemented using a ‘traditional’ search, which results in output as absolute frequencies.

Estimating body mass

Body mass estimates for Mukupirna fortidentata sp. nov. were calculated for the holotype NTM P11997 (rostral portion of a skull), and the referred specimens NTM P13348 (distal half of left humerus) and NTM P13262 (left P3) (see Table 4). The humeral estimate was derived from minimum humeral circumference, using the regression equations of Richards et al. (2019). Craniodental estimates were calculated using regression equations of Myers (2001) and incorporating the relevant smearing estimates therein recognized. We used the three highest ranking equations from the diprotodontian dataset in Myers (2001), namely: upper molar row length, measured at the widest point of the crown (UMORL); upper molar row length, measured at the alveoli (UMRL); and upper third premolar maximum width (3UPW).

Museum abbreviations

AMNH, Department of Vertebrate Paleontology, American Museum of Natural History, New York, USA; NHMUK, Natural History Museum, London, UK; NMV P, Palaeontology section, Museums Victoria, Melbourne, Victoria, Australia; NTM P, Museum of Central Australia, Museum and Art Gallery of the Northern Territory, Alice Springs, Northern Territory; QM F, Queensland Museum Fossil Collection, Brisbane; SAMA P, Palaeontology section, South Australian Museum, Adelaide, South Australia.

Systematic palaeontology

Order DIPROTODONTIA Owen, 1866

Suborder VOMBATIFORMES Woodburne, 1984; sensu Beck et al. (2020)

Infraorder VOMBATOMORPHIA Aplin and Archer, 1987; sensu Beck et al. (2020)

Superfamily VOMBATOIDEA Kirsch, 1968; sensu Beck et al. (2020)

Family MUKUPIRNIDAE Beck, Louys, Brewer, Archer, Black & Tedford, 2020

Genus Mukupirna Beck, Louys, Brewer, Archer, Black & Tedford, 2020

Type species

Mukupirna nambensis Beck, Louys, Brewer, Archer, Black & Tedford, 2020

Amended generic diagnosis

Species of Mukupirna are distinguished from other vombatiforms, unless otherwise noted, by having: an i1 that is proportionately larger and more steeply inclined; a horizontal ramus markedly deeper below p3 than below m4; a strongly bilobate P3 (except species of Muramura), with numerous prominent crenulations descending the posterolingual and posterobuccal faces; roots on upper cheek teeth that extend out of the maxillae far beyond the alveolar rims; prominent cristae on M1–2 that link stylar cusps C and D, effectively closing off the transverse valley buccally (except early vombatids and Raemeotherium yatkolai); a conical metacone on M1 lacking pre- or postmetacristae (except early vombatids); a very short or absent diastema between I3 and C1 (except Nimbavombatus boodjamullensis); molars with a medially positioned cleft in the transverse lophs (except early vombatids and Ma. arcanum); an anteroposteriorly straight postprotocristid + cristid obliqua (except early vombatids, species of Namilamadeta, and Ma. arcanum); and the protoconid and hypoconid on m2–m4 positioned at roughly one third tooth-width from the buccal margin (except species of Namilamadeta and Ma. arcanum).

Differs from species of Namilamadeta in having: bicuspid rather than tricuspid p3/P3; proportionately shorter p3/P3; and generally more bulbous molars. Differs from vombatids in having: I1/i1 that that are not hypselodont; a bicuspid p3, except Vombatus ursinus (Shaw 1800); a precingulid on the anterobuccal face of protoconid on lower molars; and lacking enamel tracts on cheek teeth. Differs from Marada arcanum in having: an i1–p3 diastema two thirds shorter; a mental foramen positioned directly ventral to the anterior half of p3, rather than more anteroventrally; a posterior lobe on p3 larger than the anterior lobe, rather than considerably smaller; a p3 with lingual and buccal faces that are ridged rather than smooth; a paracristid on m1 that forms a raised crest as it projects lingually; an anteroposteriorly narrower transverse valley that is V-shaped rather than U-shaped on m1; lacking a cuspate lingual cingulid that closes off the transverse valley on m1; a more buccally extensive precingulid on the lower molars; an origin of the ascending ramus that is proximate to m3 rather than posterior to m4; and the posterior extent of mandibular symphysis is ventral to p3 rather than anterior to p3.

Mukupirna fortidentata sp. nov.

Figs 2–9

Figure 2. Mukupirna fortidentata, sp. nov., partial skull (holotype, NTM P11997), in association with annotated line drawings. A, Occlusal view; B, lateral view; C, dorsal view; D, anterior view. Scale bar equals 40 mm.

Figure 3. Occlusal view of the upper cheek teeth of species of mukupirnid. A, Mukupirna nambensis right cheek tooth row (cast of holotype, AMNH FM102646), with M4 digitally repositioned. B, Mukupirna fortidentata right cheek tooth row (holotype, NTM P11997), with an annotated line drawing; and C, the associated left cheek tooth row. Scale bar equals 10 mm. Abbreviations: hy, hypocone; mcl, metaconule; me, metacone; pa, paracone; pas, parastyle; pr, protocone; stB, stylar cusp B; stC, stylar cusp C; stD, stylar cusp D.

Figure 4. Mukupirna fortidentata, sp. nov., referred upper dentition specimens. A, Left premaxilla (NTM P11998), depicted from left right in mesial, occlusal and lateral views. B, Left I2 (NTM P13261), depicted from left right in lingual, occlusal, buccal and posterior views. C, Left I3 (NTM P13264) depicted from left right in lingual, occlusal, buccal and posterior views. D, partial left maxilla preserving M1–M4 (NTM P11999), depicted from left right in mesial, occlusal and lateral views. Scale bar equals 20 mm.

Figure 5. Mukupirna fortidentata, sp. nov., left dentary (paratype, NTM P12000), with annotated line drawings. Depicted from top to bottom in mesial, occlusal and lateral views. Scale bar equals 20 mm.

Figure 6. Mukupirna fortidentata, sp. nov., dentary specimens, depicted from top to bottom: in mesial, occlusal and lateral views. A, right dentary (paratype, NTM P10438). B, left dentary (paratype, NTM P12001). Scale bar equals 20 mm.

Figure 7. Occlusal view of the lower cheek teeth of species of maradid and mukupirnid. A, Marada arcanum left dentary (holotype, QM F42738: image flipped), modified from Black (2007). B, Mukupirna fortidentata right dentary (Paratype, NTM P12000), with annotated line drawings. Scale bar equals 10 mm. Abbreviations: end, entoconid; hyd, hypoconid; lcd, lingual cingulid; med, metaconid; pacd, paracristid; pcd, precingulid; prd, protoconid.

Figure 8. Limb elements referred to Mukupirna fortidentata, sp. nov. Distal left humerus (NTM P13347) in A, cranial, B, lateral, E, caudal, F, medial and I, distal views. Distal left tibia (NTM P13346) in C, cranial, D, lateral, G, caudal, H, medial and J, distal views. Scale bar equals 30 mm. Abbreviations: ca, capitulum; dpc, deltopectoral crest; gtc, groove for the tendon of the m. tibialis cranialis; le, lateral epicondyle; lsr, lateral supracondylar ridge; ltf, lateral talar facet; mm, medial malleolus; msb, medial supracondylar bridge; msf, medial supracondylar foramen; mtf, medial talar facet; of, olecranon fossa; rf, radial fossa; tr, trochlea.

Figure 9. Podial elements referred to Mukupirna fortidentata, sp. nov. Α, E, I, M, Q, U, left talus (NTM P13344); B, F, J, N, R, V, right talus (NTM P13345); C, G, K, O, S, W, partial right calcaneus (NTM P6371); and D, H, L, P, T, X, right pisiform (NTM P13347). Scale bar equals 20 mm. Abbreviations: cf, cuboid facet; CLAJP, continuous lower ankle joint pattern; ct, calcaneal tuber; cut, tuberosity for attachment of the m. flexor carpi ulnaris tendon; ectf, ectal facet; ff, fibular facet; fg, flexor groove; ltf, lateral tibial facet; mtf, medial tibial facet; nf, navicular facet; nn, navicular notch; pit, pisiform tuber; pt, plantar tuberosity; sf, styloid facet; suf, sustentacular facet; trf, triquetral facet.

Vombatomorphia? fam., gen. et sp. nov. Murray & Megirian, 2006

Marada arcanum Black et al. 2012b (in part)

Diagnosis

The upper cheek teeth of Mukupirna fortidentata sp. nov. are distinguished from those of Mukupirna nambensis in having anterior molars that are proportionately wider, wherein M1 anterior width is subequal to length, as opposed to less than length by 12%. Consequently, the anteroposteriorly decreasing cheek tooth width gradient (M4 width/M1 width) of Mu. fortidentata sp. nov. is steeper than that of Mu. nambensis by 20%. The upper cheek teeth are also more strongly bilobed than those of Mu. nambensis. In occlusal view, the longitudinal axis of the P3 is aligned with the buccal cusps on M1, whereas that of Mu. nambensis is aligned with the lingual cusps. Mukupirna fortidentata sp. nov. lacks a diastema between I3 and C1, whereas Mu. nambensis has a very short (3.5 mm) diastema. The molar roots are also splayed outwards from the crown towards the maxilla/dentary, and consequently, total alveolar rim width is greater than molar crown width by 10% on M1, as opposed to being subequal in Mu. nambensis; furthermore, the lingual face of the molar roots has a markedly deeper dorsoventral concavity at mid-length.

Etymology

Derived from the Latin fortis (strong) and dentata (toothed), the name refers to the robustness of the incisors and anterior cheek teeth. The gender of the genus Mukupirna was not specified by Beck et al. (2020). Following article 30.2.4. of ICZN (1999), the genus is to be treated as feminine because the name ends in -a.

LSID of new species

http://zoobank.org/urn:lsid:zoobank.org:pub:2215F286-BBD2-42E7-AF27-0326E251FE4B

Holotype

NTM P11997, a dorsoventrally crushed adult splanchnocranium containing left P3–M3 and right C1–M4 (Figs 2, 3).

Paratypes

NTM P11998, partial premaxilla preserving partial I1 and I2; NTM P11999, partial left maxilla preserving posterior extent of M1 and M2–M3; NTM P12000, left dentary preserving i1–m4; NTM P12001, partial left dentary preserving heavily worn p3–m3; NTM P10438, partial right dentary preserving worn p3–m4; NTM P12002, right dentary preserving p3; NTM P13257, partial left dentary preserving i1, p3, m1 and lingual half of m2.

Referred material

NTM P12003, left maxilla fragment preserving half P3 and half M1; NTM P2815–11, left m3; NTM P12004, left p3; NTM P12005, right m1; NTM P12006, worn right m3; NTM P12007, worn right? m2; NTM P12008, anterior half P3; NTM P12009, posterior half P3; NTM P12010, posterior half p3; NTM P12011, unworn anterior half m2; NTM P13261, left I2; NTM P13262, left P3; NTM P13263, left I3; NTM P13264, right I3; NTM P13348, distal half of a left humerus; NTM P13347, right pisiform; NTM P13346, damaged distal left tibia; NTM P13345, right talus; NTM P13344, left talus; NTM P6371, damaged right calcaneus.

Comments

Several unassociated postcranial specimens recovered from the Pwerte Marnte Marnte fossil site compare most closely in general morphology (though up to 20% smaller) to the equivalent elements of Mu. nambensis. It has not been possible to compare these elements to those of ilariids or ?wynyardiids. However, based on the relative size of the known ilariid and ?wynyardiid material from the site, it is expected that their postcranial skeletons would be markedly larger and smaller, respectively, than the specimens in question. In light of the strong possibility that these specimens derive from Mu. fortidentata, we take this opportunity to describe them.

Type locality, unit and age

Pwerte Marnte Marnte fossil locality (24°21′S 133°43′E), on the southern flank of the James Range, Northern Territory, Australia (Fig. 1), has produced the Pwerte Marnte Marnte Local Fauna (Murray & Megirian 2006). Initial biochronological assessment indicated a probable late Oligocene age for this assemblage (Murray & Megirian 2006, Crichton et al. 2023). The stages of evolution expressed by several marsupial taxa have been taken to suggest that it may predate those of the Etadunna and Namba Formations of the southern Lake Eyre Basin (Murray & Megirian 2006), and thus possibly correspond to an as-yet-unnamed land mammal age immediately preceding the Etadunnan (Megirian et al. 2010).

Descriptions and comparisons

Craniodental

Cranium

The specimen NTM P11997 is the rostral portion of a dorsoventrally crushed adult cranium held together by matrix (Fig. 2A–D). It includes the premaxillae, maxillae, nasals, and partial lacrimals. The left P3–M3 and right P3–M4 are preserved, as is the base of the right C1, and the alveoli for the incisors and left C1. Damage to the posterior end of the palate and left maxilla was caused by a rock-drill while quarrying. Crushing of the specimen in the dorsoventral plane has reduced the minimum distance between the palate and nasal to 15 mm (Fig. 2D). The maxillae and premaxillae are splayed laterally, with their dorsal extremities crushed ventrally towards the nasals. The nasals remain largely undistorted, though bearing some damage to their posterior ends (Fig. 2C). The nasals are very robust (maximum observable thickness = 8 mm), with an anteroposterior length of at least 80 mm. The nasals terminate anteriorly in a blunt point dorsal to the I1 alveolus. Abundant foramina are present on the dorsolateral surfaces of the nasals. In transverse section, the nasals arch ventrally towards their lateral and medial margins. At the anterior-most point of the naso–premaxillary suture, the nasals are relatively broad (combined width of 35 mm), becoming laterally constricted towards the suture with the maxillae (combined width of 28 mm), and, though damaged, appear to subsequently expand laterally towards the nasofrontal suture (inferred combined width of ca 37 mm) before tapering sharply.

What remains of the lacrimals is poorly preserved, with the left side slightly more complete than the right (Fig. 2C). On the left side, fragments of bone between the lacrimal and nasal appear to derive from the premaxilla, suggesting a probable sutural contact between the frontal and maxilla. The lacrimal foramen (5 mm long and 4 mm wide) is preserved on the left side, anterior to the nasofrontal suture by ca 7 mm and lateral to the suture by 8 mm.

Much of the anterolateral portion of the right premaxilla is missing, while the left premaxilla has been distorted laterally. The lateral surface of the premaxillae is densely pitted with tiny foramina. The rostral rim of the premaxillae curves dorsally towards its suture with the nasals. The uncrushed right premaxilla bears a bulbous lateral expansion that derives from a particularly large I1 alveolus. The incisive foramen is situated in a deep, anteroposteriorly extended palatal fossa (Fig. 2A). Intensive fracturing of the palatal region of the premaxilla precludes measurement of the incisive foramen, though the anterior edge is aligned with the middle of the I3 alveolus, and the posterior edge with the middle of the C1 alveolus (Fig. 2A). There is no diastema between I3 and C1. The diastema separating C1 and P3 is quite short (7 mm).

The maxillae are relatively undistorted, with the left maxilla being better preserved than the right. A large infraorbital foramen (anteroposterior diameter of 5 mm, and dorsoventral diameter of 8 mm) is situated dorsal to the P3 mid-length by 13 mm. The region of the maxilla adjacent to the infraorbital foramen is densely pitted by small foramina, as is the lateral surface of diastema between the P3 and C1 alveoli. A weak sheath projects anteriorly from the posterior rim of the infraorbital foramen. On the palatal region of the maxilla, a prominent longitudinal ridge extends posteriorly from the premaxillary–maxillary suture for 16 mm before terminating in line with, though 11 mm medial to, the anterior alveolus of P3 (Fig. 2A).

The rostral morphology of Mukupirna fortidentata appears generally similar to that of Mu. nambensis, with the former possibly somewhat broader than the latter, though both NTM P11997 and AMNH FM 102646 are relatively crushed. The nasals (unknown in Mu. nambensis) are proportionately thicker than in any other vombatiform, as well as thylacoleonids. Unlike Mukupirna fortidentata, in which there is no diastema between I3 and C1, Mukupirna nambensis has a short diastema between I3 and C1 (3.5 mm). The maxilla of N. boodjamullensis (QM F23774) also seems to lack a diastema between C1 and the posterolingual remnant of the I3 alveolus. Mukupirna fortidentata shares with Mu. nambensis and N. boodjamullensis a short diastema between C1 and P3, being much shorter than that of most ?wynyardiids and diprotodontoids. The ?wynyardiid N. crassirostrum also has a short diastema between C1 and P3, though it consequently has a much longer diastema between I3 and C1. As in N. boodjamullensis and Muramura williamsi, the infraorbital foramen is dorsal to P3, while in species of Namilamadeta in it is positioned slightly anterior to P3.

Upper dentition

The upper dental formula is I1–I3 C1 P3 M1–M4. The alveoli of the left incisors, although damaged, are better preserved than those on the right (Fig. 2A–D). The alveoli for the incisors and canines are oriented posteriorly at an angle of ca 50° from the dorsoventral axis. The alveolus for I1 is very large and oblong from anterior view, wherein the anterior rim is directly dorsal to the posterior rim (Fig. 2D). Consequently, the alveolus length for I1, as measured dorsoventrally, is 18 mm and the width is 9 mm. The alveolus for the left I2 of NTM P11997 is also oblong, with a length of 5 mm and width of 9.5 mm, while that of the I3 is rounder and measures 11.5 mm by 10.3 mm, respectively.

A partial left premaxilla (NTM P11999) preserves the I1 and I2 (Fig. 4A). The I1 is missing most of the crown, retaining only a small section of the posterolingual surface. What remains of the I1 shows that it was relatively large (labiolingual length = 15 mm; mediolateral width = 11.5 mm) and crescent shaped, wherein the root is long (40 mm) and curves posteriorly, dorsal to I2 and I3, while the crown is oriented ventrally and may have recurved posteriorly. On I1, enamel is restricted to the portion of the tooth that projects outside of the alveolus. The dentine on what remains of the lingual surface has occlusal wear, which terminates 3 mm ventral to the rim of the alveolus.

The I2 preserved by NTM P11999 is relatively small (anteroposterior length = 5.4 mm; labiolingual width = 6.6 mm) and has an anteroposteriorly compressed root (Fig. 4A). The occlusal surface of the I2 is worn flat and the crown tapers posteriorly in width. Enamel extends down the labial and lingual faces, but is absent from the anterior face. An isolated I2 (NTM P13261) is very similar in morphology to that preserved in NTM P11999 (Fig. 4B).

Two I3 specimens (NTM P13263; NTM P13264) are referred to the taxon based on: their relative size as compared to the alveoli for the I3 preserved in NTM P11997 and NTM P11999; high crown height on the posterolabial face, consistent with being oriented anteriorly at a relatively steep angle from the dorsoventral axis; and their relative abundance in the assemblage, consistent with Mu. fortidentata as the most common large mammal otherwise represented (Fig. 4C). The specimens of I3 are considerably larger than those of the I2 (NTM P13263, anteroposterior length = 6.5 mm, labiolingual width = 7.7 mm; NTM P13264, anteroposterior length = 6.1 mm, labiolingual width = 7.4 mm), with a long root (24 mm: NTM P13264) that is round in cross section. They are high crowned, wherein the enamel extends 14 mm down the labial surface and 7 mm down the lingual face (Fig. 4C). Enamel is absent from the anterior surface. The occlusal surface is worn flat. The crown tapers posteriorly in width. An enamel fold is present on the posterolingual surface.

On NTM P11997, the root of the right C1 is preserved and projects anteroventrally 7 mm from the rim of the alveolus (Fig. 2A). There is no enamel remaining, wherein the lingual face of the crown is broken off and the labial occlusal face is worn down to 3 mm from the rim of the alveolus. At the alveolus, the anteroposterior width of the C1 is 6.4 mm and the labiolingual width is 7.4 mm.

The axis of the P3 to M4 is relatively straight, with a moderate helical trend in the occlusal plane from lingually inclined (P3) to buccally inclined (M4) (Fig. 2A). The tooth row (P3–M4) measures 55 mm at the base of the crown and 61 mm at the alveoli, deriving from an anteriorly projected anterior root on the P3 and a posteriorly projected posterior root on the M4 (also in paratype P11999). The molar roots also splay outwards dorsally from the crown, and consequently alveolar width is greater than molar width at the crown by up to 10% (also in paratype P11999: Fig. 4D). The molar size gradient decreases steeply posteriorly (M4/M1 length = 0.70; M4/M1 anterior width = 0.56). Relative crown height also decreases posteriorly (Figs 2B, 4D). The occlusal surface is relatively worn, wherein the four primary cusps, and the transverse links between them, have dentine exposed. A slight anteroposteriorly decreasing tooth wear gradient is evident along the cheek tooth row.

The alveoli for the C1 and I3 are wider than those of Mu. nambensis by approximately 20%, while those of I1 and I2 cannot be reliably compared. Both Mu. fortidentata and Mu. nambensis have a relative posterior to anterior molar length ratio (M4/M1 length) close to 0.70, while the molar width ratio (M4/M1 anterior width) in Mu. fortidentata is considerably lower (0.56) than Mu. nambensis (>0.68) mainly due to the wider anterior molars of Mu. fortidentata (Table 1; Fig. 3). This represents the steepest molar width gradient among all vombatiforms. Thylacoleonids (Diprotodontia incertae sedis) have a markedly steeper molar size gradient, wherein the posterior molars are strongly atrophied or completely lost. In absolute terms, the cheek tooth row length (measured at the base of the crown) of Mu. fortidentata (55.0 mm) is 8% longer than that of Mu. nambensis (51.4 mm), after digitally repositioning the M4 of AMNH FM 102646 (see Fig. 3). The molar roots are also not splayed outwards towards the alveoli in Mu. nambensis, with alveoli being subequal to molar width at the crown, as opposed to up to 10% wider in Mu. fortidentata.

Table 1. Measurements (in mm) of the upper cheek teeth from Mukupirna fortidentata and Mukupirna nambensis.

Specimen	P3			M1			M2			M3			M4
	L	AW	PW	L	AW	PW	L	AW	PW	L	AW	PW	L	AW	PW
Mukupirna fortidentata
NTM P11997 L	14.0	8.6	11.3	11.9	12.3	11.6	11.0	11.5	10.3	–	9.6	–	–	–	–
NTM P11997 R	14.0	8.7	11.2	11.5*	12.2	11.8	11.3	11.6	10.3	10.4	9.6	7.8	8.7	7.0	5.1
NTM P11999 L	–	–	–	–	–	11.4	10.6	11.1	9.5	9.9	9.0	7.0	8.4	6.8	5.6
NTM P12003 L	>14.8	–	–	>13.4	–	–	–	–	–	–	–	–	–	–	–
NTM P13262 L	15.3	9.3	12.3	–	–	–	–	–	–	–	–	–	–	–	–
Mukupirna nambensis
AMNH FM 102646 R	12.8	7.4	>9.2	>11.5	10.2	10.3	11.1	9.8	9.3	9.6	8.6	7.3	8.0	>6.9	5.6

Measurements of Mu. nambensis taken from a cast of the right maxilla from AMNH FM 102646. Abbreviations: L, length; AW, anterior width; PW, posterior width.

The crown of the P3 is bulbous and has an outline in occlusal view that is bilobed (Fig. 3B), with the posterior lobe 23% wider than the anterior lobe (Table 1). The apices of the cusps on the longitudinal crest spanning between the lobes have been obliterated by wear, forming a deep occlusal facet that slopes lingually. Nonetheless, the bilobed morphology is consistent with a bicuspate crown, wherein the anterior and posterior lobes likely supported the parastyle and paracone respectively. The tooth crown is robustly supported by two large roots that extend ventrally far beyond the alveoli, wherein the base of the crown is 12 mm ventral to the rim of the alveolus on the anterior lobe, and 7 mm ventral on the posterior lobe (Fig. 2B). The anterior root is slanted anteriorly at an angle of 20° from the dorsoventral axis. Enamel does not extend down the roots.

A small and relatively worn posterolingual cusp (hypocone) is present, with a smaller similarly worn cusp directly anterior (protocone) (Fig. 3B). Worn cristae ascend up the lingual face from the hypocone and protocone. From the apex of the parastyle, a crista descends anteriorly to near the base of the crown before bifurcating, with one arm (buccal precingulum) continuing buccally and quickly terminating, and the other arm (lingual precingulum) continuing posterolingually around the base of the anterior lobe. From the apex of the parastyle, a prominent crista also descends each of the lingual and buccal faces, terminating near the base of the crown. Numerous fine ridges characterize the buccal face and, to a lesser degree, the posterolingual face, of the paracone (Fig. 3). The P3 specimens NTM P12003 and NTM P13262 differ from those of NTM P11997 in having ridges that are less prominent on the posterolingual face, and in being approximately 9% larger.

The P3 differs from that of Mu. nambensis in being proportionately wider relative to M1 length, and more strongly bilobed (Fig. 3). The main cusps of the P3 are also aligned with the buccal cusps on M1, rather than the lingual cusps in Mu. nambensis (Fig. 3). We considered the possibility that the alignment of cusps on the P3 relative to the M1 in Mu. nambensis may be an artefact of the fragmented and partially deformed nature of the holotype AMNH FM 102646, but discounted its likelihood on grounds that the posterior end of the P3 crest is fitted into a shallow depression on the anterior face of M1, likely representing the P3–M1 facet. In ?wynyardiids and N. boodjamullensis, the main cusps on the P3 are also aligned with the buccal cusps on the M1. Unlike Mu. fortidentata, the posterolingual cusp on the P3 is reportedly absent in Mu. nambensis (Beck et al. 2020); however, a thin section is missing from the posterolingual face of the crown of the right P3 on the holotype of the latter. Enamel does not extend as far down the root on the anterobuccal face as it does in Mu. nambensis; though this is demonstrably intraspecifically variable in the p3 of Mu. fortidentata (see below). Though similar in relative length, the absolute lengths of P3 specimens from Mu. fortidentata are 9–19% larger than those of Mu. nambensis (Table 1).

Among ?wynyardiids, the P3 is most similar to that of species of Muramura, differing in its markedly greater absolute size, and better-developed ridges on the paracone. The P3 differs from those of the species of Namilamadeta by being more strongly bilobed. ?Wynyardiids all bear three cusps on the P3 in unworn specimens, with the posterior two being weakly differentiated. It is possible that the P3 of Mu. fortidentata also possessed three apices in its unworn state.

The upper molars have bunolophodont crown morphology (Fig. 3B). Anterior width is greater than posterior width. The crown is divisible into four quadrants, each of which has a bulbous outline in occlusal view. The occlusal surface is relatively worn, wherein the four primary cusps, and the transverse links between them, have dentine exposed. The lingual cusps have the greatest wear, being worn flat, with each exposing a C-shaped lingual rim of thick enamel from occlusal view. The buccal margins of the occlusal surfaces of the crowns on the right M1 and M2 of NTM P11997 (Fig. 3B), and the metacone of the left M1 on NTM P11999 (Fig. 4D), have chips of enamel missing either due to post-depositional damage or lifetime wear. The dentine exposed by at least some of these chips is worn, indicating that they occurred during life. The left M1 and M2 of NTM P11997 are more complete and will therefore form the point of reference for the morphological description of these tooth positions, with any differences between the left and right molars identified therein. The anterior two cusps represent the protocone (lingual) and paracone (buccal); and the posterior two cusps represent the metaconule (lingual) and metacone (buccal) (Fig. 3B). Though thoroughly worn, the primary cusps are transversely linked by cristae, which descend lingually from the buccal cusps and buccally from the lingual cusps. The juncture between the cristae is demarked by a medially positioned cleft in the transverse lophs. The stylar cusps C and D, positioned posterobuccally and anterobuccally relative to the paracone and metacone respectively, are integrated into the transverse lophs (Fig. 3B).

The M1 is slightly wider (3%) than it is long and has a slightly greater anterior than posterior width (4%) (Table 1). Though well-worn, the metacone appears larger than the paracone, and the protocone larger than the metaconule. Anterobuccal to the protocone, what remains of the preprotocrista continues buccally along the anterior margin to meet a worn preparacrista that links posteriorly to the metacone. The worn preparacrista is bulbous, seeming to form a cuspate structure in the stylar cusp B position (Fig. 3B). The paracone is lingually displaced, positioned at one third tooth-width from the buccal margin. The paracone is linked posterobuccally to a worn stylar cusp C.

Though relatively worn, the metacone appears to have been conical, wherein the anterior and posterior occlusal margins are rounded with no discernible pre- or postmetacrista. The occlusal outline of the metacone is positioned slightly posterolingual to the outline of stylar cusp D (Fig. 3B). Prominent cristae link between stylar cusp C and D, effectively closing off the transverse valley on the buccal face. A crista also descends posteriorly from stylar cusp D, meeting the postmetaconulecrista.

From what can be ascertained of the relatively worn M1 crown, the morphology is generally similar to that of Mu. nambensis. The M1 differs from that of Mu. nambensis in being approximately 3% wider than it is long, rather than 11% narrower than long. The buccal face is more lingually sloped, such that the stylar cusps and the paracone and metacone are comparatively displaced. The crown is more strongly bilobed from occlusal view. The molar roots are splayed from the crown towards the alveoli, such that tooth width at the alveoli is 10% greater than that at the crown, as opposed to being subequal in Mu. nambensis. The lingual molar root also has a markedly deeper dorsoventral concavity at mid-length on the lingual surface, which may reflect incipient partitioning of the root in Mu. fortidentata.

The M2 is similar in overall morphology to the M1 but is 8% shorter and 6% narrower (Table 1) and the preparacrista is markedly more buccally positioned. The paratype NTM P11999 differs from the holotype in that there is no discernible cusp in the stylar cusp C position (Fig. 3B, 4D). The M2 differs from that of Mu. nambensis in the same attributes as that of the M1.

The M3 is similar in overall morphology to the M2, but is 8% shorter and 17% narrower (Table 1); more strongly bilophodont, wherein the stylar cusp C is atrophied; and stylar cusp D is more transversely in line with the principal cusps (metacone and metaconule) on the metaloph. The buccal face also bears noticeably less lingual sloping. The lingual molar roots splay anteroposteriorly into the alveoli and are strongly partitioned by the palate at mid-length as it rises up the lingual face of the dorsoventral concavity towards the crown. The M3 differs from that of Mu. nambensis in the same attributes as that of the M2, with exception of relative length vs width, wherein both taxa bear an M3 that is longer than wide.

The M4 is similar in overall morphology to M3 but is 16% shorter and 27% narrower. Additionally, the talonid is markedly reduced and, as a result of surface wear, has no clearly discernible metaconule or metacone. It differs from Mu. nambensis in the same attributes as that of the M3, with the exception of the relative proportions as both are subequal in size.

Dentary

This description is primarily based on the most complete dentary specimen (NTM P12000), which preserves the i1, p3 and m1–4, but is missing much of the ascending ramus, including the coronoid, condylar and angular processes (Fig. 5).

The dentary is relatively short and deep, with greatest dorsoventral depth of the horizontal ramus below the p3 (>35 mm), and lowest below the posterior root of m4 (26 mm). Specimens NTM P10438, NTM P12001 and NTM P12002 are slightly deeper posteriorly, though still have greatest depth below p3 (Fig. 6). The diastema between the i1 and p3 is short (14 mm), and is bound by a distinct dorsal ridge along its lateral margin. Four roughly evenly spaced foramina are situated directly buccal to the diastemal ridge, in association with rugose pitting.

Mesial to the diastema, the unfused mandibular symphysis has a smooth though uneven surface (Fig. 5B). The posterior edge of the symphysis is damaged, but it appears to terminate ventral to the posterior root of p3. A large mental foramen (anteroposterior length, 2 mm; dorsoventral height, 3 mm) is situated 11 mm ventral to the middle of the anterior root of p3. One smaller nutrient foramen (diameter, 1 mm) is positioned 16 mm ventral to the mid-length of the posterior alveolus of m1. The digastric sulcus is relatively well developed, extending along the lower third of the mesial side of the horizontal ramus, from ventral to the m2 trigonid to ventral to the posterior margin of the m4 at the anterior end of the pterygoid fossa. The pterygoid fossa is deep, with a maximum depth of 10 mm. A small and damaged foramen is present dorsal to the pterygoid fossa, approximately 18 mm posterior to, and 12 mm ventral to, the posterior border of the m4. This likely represents the pterygoid foramen. The ascending ramus is relatively thick (greatest thickness, 4 mm) with a slightly rugose anterolingual surface. The anterior edge of the ascending ramus is inclined posteriorly at a relatively shallow angle of 60° relative to the horizontal ramus, and originates at a point opposite the mid-length of m3 (Fig. 5C). Posterior to the m4, there is a small postalveolar shelf. The masseteric fossa is deep, though with a poorly defined anterior rim. Possibly owing to distortion, there is no discernible masseteric foramen, though there is an artificial inclusion filled with calcite crystals at the anteroventral border to the masseteric fossa. We believe that this inclusion does not derive from the masseteric foramen, on the basis that the edges are broken and the structure is absent on specimens NTM P12001 and NTM P12002. Remnants of the posterior shelf of the masseteric fossa project laterally from the posteroventral margin of the masseteric fossa (Fig. 5C).

The horizontal ramus of Mu. fortidentata differs from that of Ma. arcanum and ?wynyardiids in a having markedly greater dorsoventral depth below the p3 compared to below the m4. This shape resembles that of thylacoleonids. The relative length of the diastema between the i1 and p3 (diastema length/cheek tooth row length) is much shorter in Mu. fortidentata (0.28) than in Ma. arcanum (0.73) and Warendja wakefieldi Hope & Wilkinson, 1982 (0.50), and slightly shorter than in the ?wynyardiids Na. albivenator Pledge, 2005 (0.35) and Muramura pinpensis (0.31), as well as Raemeotherium yatkolai Rich, Archer & Tedford, 1978 (0.35). In Mu. fortidentata, the anterior mental foramen is positioned ventral to the anterior root of the p3, while in Ma. arcanum, species of ?wynyardiid, and R. yatkolai, it is positioned anteroventral to the p3. In Warendja wakefieldi, the anterior mental foramen is anteroventral to the p3 in NMV P48980 and ventral to it in NMV P48982. The presence of a shallow angle of the anterior border of the ascending ramus relative to the horizontal ramus in Mu. fortidentata (60°) is similar to that of Ma. arcanum (55°), Warendja wakefieldi (56–59°), and, to a lesser degree, species of Namilamadeta (65–70°), Muramura (65–70°) and Raemeotherium yatkolai (70°). The anterior edge of the ascending ramus rises from buccal to the m3 mid-length in Mu. fortidentata and vombatids, buccal to the m4 in ?wynyardiids, and posterior to the m4 in Ma. arcanum and R. yatkolai. Both Mu. fortidentata and ?wynyardiids share a deep pterygoid fossa. Marada arcanum appears to bear a very shallow pterygoid fossa, though much of the posterior extremity of the only known dentary (QMF42738) is missing. The digastric sulcus is deeper and longer than in most vombatiforms, being flat in Marada arcanum and vombatids, and very shallow and posteriorly restricted in ?wynyardiids. The digastric sulcus morphology is most similar to that of the late Oligocene thylacoleonid Wakaleo schouteni Gillespie, Archer & Hand, 2019 (see Gillespie et al. 2019).

Lower dentition

The lower dental formula is i1, p3, m1–m4. The cheek tooth row (p3–m4) measures 53.7 mm, with a steeply posteriorly decreasing molar size gradient (m4/m1 length = 0.79; m4/m1 anterior width = 0.77) (Figs 5–7; Table 2). Relative crown height also decreases posteriorly. Greatest tooth wear is present on the apices of the cusps, particularly on the p3 and m1. On NTM P10438, the cusps on p3–m4 are worn down to roughly half crown height (Fig. 6). The cheek teeth on NTM P12001 are more heavily worn, such that only a sliver of enamel remains on the lingual and buccal faces of p3 and m1, and the crown is worn down to the base of the transverse valley on the m2 and m3 (Fig. 6).

Table 2. Measurements (in mm) of the lower cheek teeth of Mukupirna fortidentata and Marada arcanum.

Specimen	P3		M1			M2			M3			M4
	L	W	L	AW	PW	L	AW	PW	L	AW	PW	L	AW	PW
Mukupirna fortidentata
NTM P12000 L	11.3	8.9	12.2	9.1	9.4	11.7	8.8	8.5	10.5*	–	–	9.6	7.0	6.3
NTM P10438 R	12.9*	>8.8	12.1	9.7	9.5	>12.1	9.3	9.2	11.3	–	8.2	9.5	7.2	6.4
NTM P13257 L	10.5	8.5	10.5	8.3	8.8	–	–	–	–	–	–	–	–	–
NTM P12001 L	11.1*	8.4*	–	–	–	11.3*	8.9	–	–	8.5*	–	–	–	–
NTM P12004 L	11.3	8.3*	–	–	–	–	–	–	–	–	–		–	–
NTM P12002 R	12.0*	8.6*	–	–	–	–	–	–	–	–	–	–	–	–
NTM P12007 R	–	–	–	–	–	>10.4	8.7*	7.8				–	–	–
NTM P12006 R	–	–	–	–	–	–	–	–	10.5*	>7.8	7.5*	–	–	–
NTM P2815-11 L	–	–	–	–	–	–	–	–	10.9	8.9	7.7	–	–	–
Marada arcanum
QMF42738 R	8.7	5.2	11.4	7.4	7.3	10.9	8.0	7.5	10.8	8.4	7.3	10.0	7.2	6.3

Measurements of Ma. arcanum were taken from Black (2007). Abbreviations: L, length; AW, anterior width; PW, posterior width.

Tooth row length is similar in Mu. fortidentata and Ma. arcanum at 53.7 mm and 51.8 mm, respectively, though the posteriorly decreasing molar size gradient is considerably weaker in the latter (m4/m1 length = 0.88; m4/m1 anterior width = 0.97) (Fig. 7; Table 2). As for the upper molar width gradient, the lower molar width gradient in Mukupirna fortidentata is the steepest of all vombatiforms.

The i1 is long (crown height = 37 mm) and deep (labiolingual length = 14 mm; mesolateral width = 8 mm). The crown is projected anteriorly at a steep angle of 50° from the plane of the horizontal diastema (Fig. 5). The labial face curves dorsally such that it is almost vertical near the apex. The enamel surface is punctate and has a thickness of 0.5 mm, restricted to the labial and mediolateral surfaces. Tooth wear is restricted to the lingual surface, decreasing posteriorly in extent from the apex. The mediolateral width of the crown tapers slightly towards the somewhat pointed apex, wherein the lateral margin curves mesially along the occluding edge. The tooth is not hypselodont, wherein the enamel does not extend down the root into the alveolus. The root is long (>35 mm) and curves posteriorly, ventral to the p3.

The i1 is proportionately larger than that of contemporary vombatiform taxa. The angulation at which the i1 projects relative to the plane of the horizontal ramus (50°) is similar to that of species of Ilaria and Phascolarctos cinereus (Goldfuss, 1817), and steeper than in the species of Muramura (20°), Namilamadeta (30°), and Ma. arcanum (ca 30°). The relative size and angulation of the i1 is similar to that of thylacoleonids, though differing in that enamel does not encapsulate the crown, in common with most other vombatomorphians.

Similar to the P3, the crown of the p3 is bulbous and slightly two-lobed, with posterior width 17% greater than anterior width. These lobes abut one another anteroposteriorly, forming a crest that is aligned slightly anterobuccal to posterolingual, meeting the m1 buccal to paracristid (Fig. 6A). The apices of the lobes have been obliterated by wear, forming an occlusal facet that slopes buccally. Nonetheless, the two-lobed morphology is consistent with a bicuspate crown, wherein the anterior and posterior lobes likely supported a protoconid and metaconid, respectively. The enamel is moderately wrinkled, with weak ridges descending from the crest. Particularly well-developed ridges descend the lingual and buccal faces from the apex of each of the protoconid and metaconid. A prominent ridge also continues anteriorly and posteriorly down the crown from the protoconid and metaconid, respectively. The posterior ridge from the metaconid bifurcates into two ridges, descending each of the posterolingual and posterobuccal faces before sweeping anteriorly. The p3 is supported by two large roots that extend dorsally beyond the rim of the alveoli by 4 mm. Thinning enamel extends a moderate length down the roots on the buccal face, and to a lesser degree on the lingual face. The extent to which enamel extends down the roots on the p3 varies noticeably between specimens NTM P12000, P12004 and P12002, but in no specimen does it extend down the root into the alveolus.

In common with Ma. arcanum, the p3 in Mu. fortidentata appears to be bicuspid (Fig. 7), while that of ?wynyardiids is tricuspid. Enamel extends slightly down the anterobuccal face of the anterior root of the p3 in Mu. fortidentata, Ma. arcanum and species of Ilaria. The p3 differs from that of Ma. arcanum in being proportionately longer relative to m1 by 16%; having moderately wrinkled rather than smooth enamel; and a protoconid smaller than the metaconid rather than the former being several times larger than the latter (Fig. 7).

The lower molars have a sub-rectangular outline in occlusal view, with a bunolophodont crown morphology that is composed of four bulbous principal cuspids with rounded outer faces (Fig. 7B). The cuspids are transversely linked by cristids, which together form lophids. Owing to the juncture between the transverse cristids, each lophid is partitioned at mid-width by a cleft. Weak crenulations descend the anterior and posterior face of the lophids, indicative of prominent ridges in their unworn condition.

In occlusal view, the m1 has corners that are rounded, and lingual and buccal margins that are concave at mid-length (Fig. 7B). The trigonid is noticeably taller than the talonid. The tooth is high-crowned. The buccal face is particularly high-crowned, such that the juncture between the postprotocristid and cristid obliqua occurs at two-thirds of the crown height. A precingulid is present anterobuccal to the protoconid. Though the apices of all cusps are somewhat worn, the protoconid is clearly the tallest cusp. The protoconid is positioned centrally, slightly buccal to tooth mid-width, and consequently the buccal face of the trigonid has a gentle slope towards the base of the crown. The paracristid is directed anterolingually from the protoconid to the anterior margin of the tooth, where it then continues lingually to the paraconid position, forming a prominent raised crest. A distinct valley separates the paracristid from the metaconid, with no premetacristid linking them. The metaconid, positioned posterolingual to the protoconid, is the smallest cusp. A worn posterolingually oriented cristid links the protoconid to the metaconid, which together form the protolophid. The postmetacristid descends the posterolingual face of the metaconid to the transverse valley.

The transverse valley separating the protolophid and hypolophid is closed off buccally by the worn postprotocristid + cristid obliqua. The preentocristid is weak and does not meet the posterior terminus of the similarly weak postmetacrista. A small pocket is present at the lingual end of the transverse valley, lingual to the posterior terminus of the postmetacrista and the anterior terminus of the preentocristid. This pocket derives from a weak, and partitioned, lingual cingulid.

Unlike the protolophid, the hypolophid is partitioned at mid-width by a cleft. The hypolophid is oriented anterolingually, reflecting the posterior position of the hypoconid relative to the entoconid (Fig. 7B). The hypolophid is a third wider than the protolophid, reflecting the more lingual position of the protoconid relative to the hypoconid and the more buccal position of the metaconid relative to the entoconid. The posthypocristid descends posterolingually from the hypoconid to near the base of the crown, and then curves and continues lingually to meet the postentocristid at an almost 90° angle.

In common with Ma. arcanum and vombatids, but unlike ?wynyardiids and ilariids, on m1: the four principal cuspids are bulbous and somewhat conical, wherein the buccal faces of the protoconid and hypoconid are rounded rather than anteroposteriorly compressed (also in species of Namilamadeta); the juncture between the postprotocrista and cristid obliqua is more buccally positioned, rather than more centrally positioned; the postprotocristid + cristid obliqua is anteroposteriorly straight; and the lingual face of the hypoconid is expanded such that the cleft in the hypolophid defining the point of contact between the hypoconid and entoconid occurs midway between their apices, rather than much closer to the apex of the hypoconid. The m1 differs from that of Ma. arcanum by having: a proportionally greater maximum width by 12%; a paracristid that forms a raised crest as it projects lingually; thick crenulations that descend the anterior and posterior faces of the hypolophid; a postprotocristid longer than the cristid obliqua; an anteroposteriorly narrower transverse valley that is V-shaped rather than U-shaped; lacking a cuspate lingual cingulid that closes off the transverse valley; and lacking a transverse valley incised into the anterobuccal face of the hypoconid (Fig. 7). It differs from vombatids in having a precingulid on the anterobuccal face of protoconid, and lacking enamel tracts that extend down the roots on the buccal face of the protoconid and hypoconid.

The m2 is similar to the m1, with the exception of being smaller, lower crowned and having subtle differences in the arrangement of structures on the trigonid (Fig. 7). The precingulid anterobuccal to the protoconid is deeper and more buccally extensive, forming a pocket that continues anterolingually to meet the paracristid. The lingual end of the paracristid forms a relatively flat shelf that is closed off lingually by a weak premetacristid. On the m2, the protolophid more closely parallels the orientation, length and general morphology of the hypolophid than does that on the m1. In particular, the protoconid is more buccally positioned, while the metaconid is larger and more anterolingually positioned. The protoconid and hypoconid are also positioned slightly more posteriorly than the metaconid and entoconid, respectively. In turn, the protolophid and hypolophid are both oriented slightly anterolingually to posterobuccally.

Mukupirna fortidentata and Ma. arcanum are more similar in the morphology of the m2 than the m1. The m2 differs from that of Ma. arcanum in having: a transverse valley that is V-shaped rather than U-shaped; a precingulid that is more buccally extensive; and lacking a pocket buccal to the transverse valley (Fig. 7). The m2 differs from that of vombatids, ilariids and ?wynyardiids in the same attributes as does that of the m1. In common with Ma. arcanum, vombatids and ?wynyardiids, the protolophid and hypolophid are each oriented slightly anterolingually to posterobuccally, rather than traversing directly buccolingually as in diprotodontoids.

The m3 on NTM P12000 was broken and distorted by taphonomic processes, with fragments scattered as preserved. These were repositioned during preparation (Fig. 7). From what remains of the tooth, it is evident that considerable natural wear occurred during life, extending almost to the base of the enamel on the buccal face. This likely reflects that, during life, the occluding upper molar was malformed and positioned ventrally far below the other cheek teeth. The comparatively little-worn molar specimen NTM P2815-11, reported by Murray & Megirian (2006), is also considered to represent an m3. The specimen, NTM P2815-11, is generally similar to the m2 preserved in P12000, except in: being smaller, lower crowned, and trapezoidal in occlusal outline owing to a proportionately narrower talonid. As in the m2, the m3 is generally similar in morphology to that of Ma. arcanum. In addition to the differences discussed with respect to the m2, the m3 differs from that of Ma. arcanum in being proportionately smaller, and the talonid is smaller relative to the trigonid (Fig. 7; Table 2).

The m4 is similar to the m3, except in: being smaller; lower crowned; having an even talonid; a hypolophid that traverses buccolingually rather than posterobuccally to anterolingually; and a prominent cristid descending anterobuccally from the apex of the hypoconid to the transverse valley (Fig. 7). The m4 is also less worn than the other molars, with more clearly defined ridges or crenulations on the anterior and posterior faces of the lophids. In addition to the differences discussed with respect to the m3, the m4 differs from that of Ma. arcanum in that the hypolophid traverses buccolingually rather than posterobuccally to anterolingually. The m4 shares with that of Ma. arcanum a cristid descending the anterobuccal face of the hypoconid, which is also present in Muramura pinpensis.

Forelimb

Humerus

The distal half of a left humerus (NTM P13348) is preserved from just proximal to the termination of the deltopectoral crest (Fig. 8A, B, E, F, I). The lateral supracondylar ridge is abraded, and the medial condyle is missing. The fragment of the crest preserved suggests that it was thick, but the degree to which it protruded from the shaft is unknown. The lateral supracondylar ridge is well developed, and extends proximally along the diaphysis to just below the termination of the deltopectoral crest. The medial supracondyloid foramen is large, and the supracondylar bridge is oriented at 55° relative to the diaphysis. This indicates that the distal humerus was markedly wide relative to the diaphysis. The capitulum is hemispherical, with an interruption on the lateral border, and the lateral epicondyle projects a distance approximately equal to the width of the capitulum. The trochlea is gently concave, and the medial extent of the ulnar facet is not preserved but the capitulum and ulnar facets are level. Both the radial and olecranon fossae are extremely deep.

The radial and olecranon fossae are deeper than in any of the compared species. The humerus is similar in morphology to those of early diprotodontoids such as Ngapakaldia tedfordi and Nimbadon lavarackorum Hand, Archer, Godthelp, Rich and Pledge, 1993, though with a better-developed lateral condyle. The diaphysis below the deltopectoral crest is also more slender than in N. tedfordi, and the deltopectoral crest is more laterally deflected than in Ni. lavarackorum. The trochlea is deeper than in the arboreal taxa studied, but not as deep as in extant vombatids. The crushed and distorted nature of the Mukupirna nambensis humerus (AMNH 102646) described by Beck et al. (2020) makes comparison difficult, but it appears similar to the humerus described here except that it is about 20% smaller and the lateral supracondylar ridge extends more proximally in M. nambensis.

Pisiform

The right pisiform (NTM P13347) is robust, with a distinct but thick waist and a bulbous distal end (Fig. 9D, H, L, P, T, X). The tuber is about 40% wider (mediolaterally) than it is deep (craniocaudally). The triquetral facet is somewhat damaged but appears to be flat, triangular, and restricted to the craniomedial portion of the proximal end. The facet for the styloid process of the ulna is large (ca twice the size of the triquetral facet), oval, deeply mediolaterally convex, and near-perpendicular to the triquetral facet. The distal end is enlarged to accommodate attachment of a well-developed m. flexor carpi ulnaris tendon.

Overall, the pisiform is most similar in shape to that of Lasiorhinus latifrons Owen, 1845 (and most unlike those of extant arboreal taxa such as Phascolarctos cinereus, Trichosurus vulpecula Kerr, 1792 and Pseudocheirus peregrinus Boddaert, 1785), but is more robust, with the proximal end being more like that seen in Zygomaturus trilobus Macleay, 1858. It does not display the triangular proximal end seen in Nimbadon lavarackorum, Ngapakaldia tedfordi and extant wombats (e.g., see Munson 1992, Black et al. 2012c). The specimen is similar to that of Mukupirna nambensis, but is slightly smaller, with a more pronounced waist and a rounder distal end (see Beck et al. 2020).

Hind limb

Tibia

A distal left tibia (NTM P13346), consisting of the distal third of the diaphysis and damaged distal epiphysis, is preserved (Fig. 8C, D, G, H, J). The specimen is of the right size and morphology to fit with the left talus (NTM P13345). It is a robust bone, and the shaft shows a degree of mediolateral compression, even after allowing for compaction of the fossil. The cranial and caudal borders of the lateral talar facet are damaged, and the preserved portion is flat and slopes downwards cranially. The medial talar facet is small and occupies the cranial part of the lateral face of the medial malleolus. The malleolus is prominent and forms an angle of 130° with the lateral talar facet. Its medial edge is interrupted by a deep groove, possibly for the tendon of the m. tibialis cranialis.

The distal tibia of Mukupirna nambensis is not known, precluding comparisons. The distal tibia is most similar to that seen in vombatids, but the medial malleolus is craniocaudally shorter and mediolaterally wider relative to the lateral talar facet; the medial talar facet is closer to horizontal and more continuous with the lateral facet; and the groove on the medial malleolus is larger in the former. The fragment preserved suggests that the medial talar facet may continue onto the malleolus in a fashion similar to that seen in the extant arboreal taxa studied, but this cannot be confirmed. The tibial fragment differs from Ngapakaldia and Nimbadon in that the medial malleolus is restricted to the cranial part of the medial border, rather extending along the whole of the medial border.

Talus

Two near-perfect tali are known, one right (NTM P13345: Fig. 9A, E, I, M, Q, U) and one left (NTM P13344: Fig. 9B, F, J, N, R, V). The tibial facets occupy approximately half of the dorsal surface, and the components corresponding to the lateral and malleolar facets form a continuous surface, the shared border being represented by a slight, gently convex ridge. The lateral tibial facet is about three times as large as the malleolar facet, and twice as large as the fibular facet; it is rectangular, and the trochlea is shallow. The malleolar facet occupies the middle third of the medial border of the talus, and is roughly triangular and almost flat. The fibular facet is triangular and flat, occupying the caudal half of the lateral edge of the talus. It is angled at 140° (in NTM P13345) or 120° (in NTM P13344) relative to the main tibial facet, and the two facets are separated by a low, straight ridge. A large ligamental pit at the cranial border of the lateral tibial facet separates it from the navicular facet, but the navicular and malleolar facets are joined for ca 5 mm on the medial edge. The calcaneal facet is mediolaterally concave caudally and convex cranially and is restricted to the lateral half of the plantar talus. The medial plantar tuberosity is large and dome shaped, and forms the medial edge of a deep groove on the caudal border of the talus through which the flexor tendons pass. A large, deep, lozenge-shaped, ligamental pit runs from the medial plantar tuberosity to the medial edge of the cuboid facet, separating the navicular and calcaneal facets. The cuboid facet is continuous with the calcaneal facet and is large, being approximately half the size of the latter facet. The navicular facet occupies the medial half of the cranial face and cranial half of the medial face and is broadly U-shaped.

The two tali (NTM P13345 & NTM P13344) are most similar in overall shape to those of extant vombatids (Lasiorhinus spp., and Vombatus ursinus) but the malleolar facet is closer to horizontal in the former (as is seen in extant arboreal taxa). The tali are less craniocaudally elongate than those of the extant arboreal taxa examined, and the medial plantar tuberosity is larger and rounder than in all other species examined; a feature noted in Mukupirna nambensis by Beck et al. (2020). The Continuous Lower Ankle Joint Pattern (including the sustentacular and ectal facets) of Szalay (1994) is most like that of Ngapakaldia tedfordi, excepting that it extends toward the plantar tuberosity in the latter.

Calcaneus

The right calcaneus (NTM P6371: Fig. 9C, G, K, O, S, W) articulates very closely with one of the tali (NTM P13345). The distal half of the tuber, the accessory sustentacula (viz. Szalay 1994), and the area for attachment of the calcaneofibular ligament are all missing. The proximal part of the tuber preserved is mediolaterally compressed and slants dorsolaterally to plantomedially. The talar facet is oriented near to the transverse plane and mirrors the corresponding facet on the talus with a concave sustentacular component and convex ectal component. The cuboid facet is L-shaped in dorsal view with a strong lateral component. The navicular facet is a deep, V-shaped notch on the medial border of the cuboid facet.

The fragmentary nature of the calcaneus makes it difficult to make comparisons to other taxa. The tuber appears to be relatively slender in cross section and in this sense, it is similar to that of P. cinereus, T. vulpecula and Nimbadon lavarackorum and unlike Ngapakaldia tedfordi. The cubonavicular joint is also most similar in morphology to that seen in T. vulpecula and N. lavarackorum. It also matches the descriptions of the calcaneus of Mukupirna nambensis provided by Beck et al. (2020).

Vombatidae gen. et. sp. indet.

Fig. 10

Figure 10. Vombatomorph molar fragments from the Tarkarooloo Local Fauna, with annotated line drawings in occlusal view. A, Vombatidae gen. et. sp. indet., trigonid of left m2 or m3 (NMV P.157575), photographed from left to right in occlusal, buccal and anterior views. B, Fam., gen. et. sp. indet., partial right ?m4 (NMV P.157576), photographed from left to right in occlusal, lingual, and anterior views. C, Fam., gen. et. sp. indet., partial ?posterior moiety of right ?upper molar (NMV P.157536), photographed from left to right in occlusal, lingual and posterior views. Scale bar equals 4 mm. Abbreviations: co, cristid obliqua; hyd, hypoconid; me, metacone; med, metaconid; pa, paracone; pacd, paracristid; pcd, precingulid; pomed, postmetacristid; poprd, postprotocristid; prd, protocone.

Referred material

NMV P157575, worn trigonid of left m2 or m3.

Type locality, unit and age

Tarkarooloo Local Fauna from Tom O's Quarry, west side of Lake Tarkarooloo (Fig. 1: 31°08′S 140°06′E), Namba Formation, in the Callabonna Sub-basin, South Australia (Rich et al. 1991, Woodburne et al. 1994). Based on biocorrelative and stage-of-evolution comparisons, it is thought that the Tarkarooloo Local Fauna is of late Oligocene age, correlating with the Ngama Local Fauna (Zone D of the Etadunna Formation: ca 24.1 Ma), or is slightly older (Woodburne et al. 1994, Megirian et al. 2010).

Description and comparisons

The partial lower molar NMV P157575 is identified as preserving the trigonid, rather than the talonid, on the basis that: a weak precingulid is present; and the lingual cuspid (metaconid) is positioned more anteriorly than the buccal cusp (protoconid) (Fig. 10A). The specimen compares best in size and general morphology to those of species of vombatid (Figs 6, 10). It shares with those of species of basal vombatids, as well as Mukupirna and Marada: a generally simple and bulbous profile of the primary cusps; enamel that is noticeably thicker on the buccal face than lingual face of lower molars; and relatively large size as compared to species of ?wynyardiid. In addition—though quite worn—the posterior terminus of the postprotocrista is close to the buccal margin; and therefore, the postprotocristid and cristid obliqua were likely quite anteroposteriorly straight, rather than descending posterolingually and anterolingually, respectively.

In common with early vombatids, but unlike species of Mukupirna and Marada, the trigonid of NMV P157575 has: enamel that extends much further down the buccal face of the roots than the lingual face; the apex of the protoconid positioned very close to the buccal margin; and what remains of the crown attests to a strongly bilobed outline from occlusal view. Additionally, NMV P157575 lacks a true precingulid, instead retaining only a slight swelling on the anterior face, buccal to the paracristid. A similar swelling is present on the heavily worn Nimbavombatus boodjamullensis lower molar specimen QM F23773 (Brewer et al. 2015, fig. 5), and to a lesser degree on the Rhizophascolonus ngangaba specimens QM F23768 and QM F23769 (Brewer et al. 2018: fig. 12). The enamel tracts are much shorter than those of Rhizophascolonus spp., and of similar length, or slightly shorter than, those of N. boodjamullensis. To the extent NMV P157575 can be compared to the lower molars of N. boodjamullensis, it differs in its larger size by 20%.

Three further molar fragments (NMV P48996, NMV P157576 and NMV P157563) recovered from Tom O’s Quarry, Lake Tarkarooloo, are worthy of brief discussion. The partial molar NMV P48996 (fig. 5, Rich & Archer 1979), was referred to Vombatidae by Rich and Archer (1979) on the basis that it is bilobed and hypsodont with closed roots. It may derive from the same taxon as the referred vombatid specimen NMV P157575. The whereabouts of NMV P48996 is currently unknown, and consequently, direct comparison could not be made.

The molar fragments NMV P157576 and NMV P157563 may also derive from the same taxon as NMV P157575, though they do not preserve any attributes that would definitively unite them (Fig. 10), and consequently are referred to Vombatomorphia fam., gen. et. sp. indeterminate. On NMV P157576, crown height increases towards the damaged face, indicating that if it is a lower molar, then it is a right molar, and if it is an upper, then it is a left molar. Additionally, one of the lophs is markedly narrower, indicating that the tooth is likely an M4 or m4. The protolophid is also markedly narrower than the hypolophid on m1 in basal vombatids (e.g., Rhizophascolonus ngangaba: see Brewer et al. 2018). We consider it less likely that NMV P157576 represents an m1 on the basis that: the cusps on the wider loph are considerably taller than those on the narrower loph; and the tooth is very low crowned. The specimen NMV P157563 would seem likely to be the lingual side from an upper molar given that the loph transverses buccolingually, rather than posterolingually to anterobuccally, as is typically the case in lower molars (Fig. 10C).

Results of phylogenetic analyses

Parsimony analyses with all taxa included, and ordering of states where morphoclines were inferred, generated a strict consensus of 120 most parsimonious trees, each of 420 steps (Fig. 11A). The strict consensus tree had a consistency index of 0.45 and a retention index of 0.79.

Figure 11. Phylogeny of vombatiforms based on: A, maximum parsimony analysis with numbers at nodes representing bootstrap support values; B, undated Bayesian analysis, presented as a majority rule consensus with numbers at nodes representing Bayesian posterior probabilities.

The Bayesian results, presented as a 50% majority rule consensus topology of the post-burnin trees, were for the most part consistent with that under maximum parsimony, though with higher support values for most nodes (Fig. 11). Marada, Mukupirna and vombatids were moderately strongly supported as a clade (BPP = 0.83 under Bayesian inference, Fig. 11B; BS = 69% in parsimony, Fig. 11A). Monophyly of Vombatidae was strongly supported (BPP = 1.0; BS = 78%) to the exclusion of Marada and Mukupirna. Under Bayesian inference, Mukupirna fortidentata and Mu. nambensis formed a moderately strongly supported clade (BPP = 0.91) sister to Vombatidae (BPP = 0.63), to which Ma. arcanum was basal. This maradid–mukupirnid–vombatid group formed the strongly supported sister clade to Diprotodontoidea under Bayesian inference (BPP = 0.88; BS = 27%), together sister to a paraphyletic ?Wynyardiidae. Ilariidae was supported as the most basally diverging family within Vombatomorphia (BPP = 0.78, BS = 94%). Monophyly of Vombatomorphia + Phascolarctidae was strongly supported (BPP = 1.0, BS = 90%), with low support for Thylacoleonidae sister to this clade (BPP = 0.56).

Results of body mass estimates

Body mass of Mukupirna fortidentata was estimated at 48 kg using minimum humeral circumference of NTM P13348 (see Table 4). The estimates that derive from dental measurements (NTM P11997) were quite varied, at 26, 36 and 84 kg for UMORL, UMRL and 3UPW, respectively (see Table 4). The largest dental specimen from Mu. fortidentata (the left P3, NTM P13262), yielded a body mass estimate 31% higher than the corresponding estimate for the holotype NTM P11997.

Discussion

We have described Mukupirna fortidentata sp. nov., based on a partial skull, and isolated partial maxillae and dentaries, from the Pwerte Marnte Marnte site in the Northern Territory, Australia. The taxon differs from Mu. nambensis principally in having: the longitudinal axis of the P3 aligned with the buccal molar cusps; a proportionally steeper anteroposteriorly decreasing molar width gradient; upper molars that are more strongly bilobed; and no diastema between the I3 and C1.

Implications for Mukupirnidae and Maradidae

The complete upper and lower dentition of Mukupirna fortidentata sp. nov. allows for assessment of the relationship between species of Mukupirna and Marada, and therefore, the families Mukupirnidae and Maradidae, respectively. Phylogenetic analyses in this study support Marada, Mukupirna and vombatids as forming a monophyletic Vombatoidea sensu Beck et al. (2020) (BS = 69%, BPP = 0.83: Fig. 11). Vombatidae is strongly supported (BS = 78%, BPP = 1.0) as monophyletic; the species of Mukupirna form the immediate sister clade (BPP = 0.63), with Ma. arcanum the next most basally diverging taxon. The dentary and associated lower dentition of Mu. fortidentata reveal several morphological differences from Ma. arcanum that we consider to be compelling evidence in support of their generic distinction. In particular, the dentary of Mu. fortidentata is overall more robust than that of Ma. arcanum, wherein it is markedly shorter and deeper, with a shorter diastema. Moreover, the lower dentition of Mu. fortidentata differs most notably from that of Ma. arcanum in: having an i1 that is proportionately larger, projecting at a steeper angle relative to the plane of the horizontal ramus of 50° compared to ca 30°; a p3 with a larger posterior than anterior lobe, rather than a markedly smaller posterior lobe; an m1 with a paracristid that forms a prominent shelf as it projects anterolingually; a more buccally extensive precingulid on the lower molars; and a markedly steeper anteroposterior gradient in molar length and width (Table 2). It could be argued that the lower dentitions of Ma. arcanum and species of Mukupirna are not sufficiently dissimilar from one another to merit family level distinction. However, it also cannot be demonstrated that these taxa form a clade, as they do not share any obvious synapomorphies to the exclusion of Vombatidae.

We note that, contrary to the results of our phylogenetic analysis, Ma. arcanum shows several attributes in cheek teeth morphology that are suggestive of a somewhat intermediate stage of evolution between the species of Mukupirna and vombatids. These include: p3 size reduced relative to the m1, with a posterior lobe reduced relative to anterior lobe (see QM F57966 in Brewer et al. 2018, fig. 14); a cuspate lingual cingulid (metastylid) on the m1 (see QM F57967 & QM F23764 in Brewer et al. 2018, fig. 11); enamel that extends slightly down the root on the buccal face of m1; loss of the precingulid on the m1, and a reduced precingulid on the m2–m4; lower molars with a juncture between the postprotocristid and cristid obliqua that is closer to the buccal margin; and lower molars that are proportionately narrower with a more equal anteroposterior size gradient along the molar row. In this study, the recovered phylogenetic placement of Ma. arcanum as more basal within Vombatoidea, rather than sister to Vombatidae, may reflect in part that the taxon also shares some attributes of dentary morphology (unknown in basal vombatids) with the plesiomorphic diprotodontoid Raemeotherium yatkolai (see Black 2007).

Lake Tarkarooloo vombatid

The vombatid taxon from the Tarkarooloo LF is insufficiently known to formally name. It is tentatively referred to Vombatidae, rather than Maradidae or Mukupirnidae, because the lower molar specimen NMV P157575 has: enamel that extends much further down the buccal face of the roots than on the lingual face; a more strongly bilobed outline from occlusal view; and only a faint remnant of the precingulid (Fig. 10). In particular, the presence of enamel extending down the buccal surface of the roots on the lower molars (and the lingual surface of the roots on upper molars) is thought to represent a stage towards the development of hypselodont cheek teeth, which characterizes the dentitions of later vombatids (Brewer et al. 2015, Brewer et al. 2018, Beck et al. 2020).

The oldest named vombatids (see Stirton et al. 1967, Brewer et al. 2008, Brewer et al. 2015, Brewer et al. 2018), date to 18.5 Ma in the early Miocene (ages given follow Woodburne et al. 1994, Megirian et al. 2010, Woodhead et al. 2016). A partial dentary referred to an indeterminate species of Rhizophascolonus is known from Bone Reef Site, Riversleigh, which was initially considered Faunal Zone A (late Oligocene) on the basis of biocorrelation (Archer et al. 1989, 1997, Travouillon et al. 2006), though more recently stage-of-evolution comparisons seem to place it within Faunal Zone B (Arena et al. 2016), as noted by Brewer et al. (2018). In this study, the referral of the partial lower molar NMV P157575 to Vombatidae would seem to corroborate Rich & Archer (1979) in their familial referral of the (?lost) molar fragment, NMV P48996, which was also recovered from Tom O’s Quarry, Lake Tarkarooloo. Together, these specimens provide support that early vombatids were present by at least 24.1 Ma in the late Oligocene.

Geilston Bay dentary

The taxonomic affinity of a dentary (NHM UK PV OR 40157) preserving the worn m1–m4 from the purportedly very early Miocene Geilston Bay Local Fauna, Tasmania, has been the subject of some uncertainty since it was initially reported by Tedford et al. (1975). The specimen was first considered to be a partial maxilla with affinities to the diprotodontid genus Ngapakaldia, though considerably smaller (Tedford et al. 1975). It was subsequently reinterpreted as the dentary of a large petauroid by Tedford & Kemp (1998) based on an anteroposteriorly decreasing molar length gradient (considered a shared primitive state), increasing crown height anteriorly, a linear and peripheral cristid obliqua, and an entoconid as the largest talonid cusp (considered shared derived states). Crosby et al. (2001) subsequently suggested that isolated teeth from Geilston Bay, attributed by Tedford & Kemp (1998), to Petauroidea, were more congruent with phalangerid affinities given their well-defined lophids, though they did not specifically make reference to the dentary in question. We agree with Tedford & Kemp (1998) that the markedly peripheral cristid obliqua and postprotocristid preclude the taxon from being a member of Phalangeroidea, sensu Aplin & Archer (1987).

We propose instead that NHM UK PV OR 40157 may belong to an early vombatoid. The morphological attributes used to refer the specimen to Petauroidea by Tedford & Kemp (1998) are all shared by early vombatoids. The relative height of the entoconid as the tallest talonid cusp on NHM UK PV OR 40157 appears to be a product of greater wear to the buccal than lingual cusps, as occurs in early vombatoids. Further attributes of molar morphology that are shared with vombatoids include: a bilobed outline from occlusal view; a generally simple and bulbous profile of the primary cusps; transverse lophs that are constricted at roughly mid-width; and enamel that is thicker on the buccal rather than lingual face of the lower molars; as well as being roughly intermediate in size between those of Nimbavombatus boodjamullensis and species of Rhizophascolonus (Table 3). The juncture between the postprotocristid and cristid obliqua is also very close to the buccal margin, indicating that, in their unworn condition, these cristids were likely quite anteroposteriorly straight, rather than descending posterolingually and anterolingually, respectively. As is characteristic of early vombatids, but not species of Mukupirna and Marada, the protoconid and hypoconid are positioned very close to the buccal margin, and the molars lack a distinct precingulid. The taxon also shares with Rhizophascolonus ngangaba Brewer, Archer, Hand & Price, 2018, a cuspate lingual cingulid forming a metastylid on m1 (also present in Ma. arcanum). The molars on NHM UK PV OR 40157 nonetheless differ notably from all referred vombatids in that enamel does not extend down the roots on the buccal face of m2–m4. Taken together, a vombatoid affinity for the specimen would seem plausible, though this cannot be confirmed owing to the high degree of wear to the molars.

Table 3. Measurements (in mm) of the molar specimens from the Tarkarooloo LF vombatoid, the Geilston Bay LF ?vombatoid, as well as the basal vombatids Nimbavombatus boodjamullensis, Rhizophascolonus ngangaba and R. crowcrofti.

Specimen	Tooth position	L	AW	Pw
Tarkarooloo LF, Vombatidae gen. et. sp. indet.
MV P157575 L	m2 or m3	>5.7	7.1
Tarkarooloo LF, Vombatomorphia fam., gen. et sp. indet.
MN P157576 R	m4?	>10	>7.2	>6.7
MV P157563	Upper molar?	>6.3		>7.3
Geilston Bay LF ?vombatoid
NHM UK PV OR 40157 R	m1	9.9*	6.3*	6.9
	m2	8.4	7.0	6.7
	m3	8.0*	6.9	6.3
	m4	7.9	5.8	5.6
Nimbavombatus boodjamullensis
QM F51404	m1	7.1	4.2	5.5
QM F23773	m2 or m3	8.0	5.8	6.0
Rhizophascolonus ngangaba
QM F23764	m1	11.5	4.9	6.6
QM F57967	m1	10.8	4.8	6.2
QM F57968	m2	13.0	6.7	7.9
QM F23768	mx	11.9	7.2	7.3
Rhizophascolonus crowcrofti
QM F52745	mx	13.3	10.4	10.4
QM F52746	mx	15.5	10.3	10.1
QM F57960	mx?	14.5	9.5	9.1
QM F57961	mx	14.0	10.7	10.6
QM F57962	mx	13.4	10.0	10.1

Measurements of NHM UK PV OR 40157 taken were from a cast (SAMA P.21377) and those of the vombatids are from Appendix 3 of Brewer et al. (2018). Abbreviations: L, length; AW, anterior width; PW, posterior width.

Comments on vombatiform interrelationships

Within Vombatiformes, the family Phascolarctidae (koalas), often placed within the infraorder Phascolarctomorphia following Aplin & Archer (1987), have long been considered sister to Vombatomorphia + Thylacoleonidae (e.g., Marshall et al. 1990, Myers & Archer 1997, Myers et al. 1999, Black 2008, Black et al. 2012b, Black et al. 2014, Brewer et al. 2015). More recently, Thylacoleonidae has instead been recovered as the most basal diverging clade in Vombatiformes (Gillespie et al. 2016, Beck et al. 2020), or outside Vombatiformes altogether and classified as Diprotodontia incertae sedis (Beck et al. 2022). We find strong support for monophyly of Vombatomorphia and Phascolarctidae (BPP = 1.00, BS = 90%: Fig. 11) to the exclusion of Thylacoleonidae. Following the phylogenetic definition of Vombatiformes proposed by Beck et al. (2020), Thylacoleonidae was weakly supported within the suborder under Bayesian inference (BBP = 0.56), while unresolved under parsimony. We tentatively uphold the subordinal nomination of Thylacoleonidae as Diprotodontia incertae sedis by Beck et al. (2022), because the analyses in the present study included only a single representative diprotodontian from outside of these clades; namely, the burramyid Cercartetus lepidus.

Among vombatomorphians, Ilariidae was strongly supported under Bayesian inferences (BPP = 0.94; BS = 78%) as basally diverging to a paraphyletic Wynyardiidae, with species of Muramura sister to a clade comprising Namilamadeta snideri + Vombatoidea + Diprotodontoidea (BPP = 1.0). Paraphyly of ?Wynyardiidae, as represented by species of Muramura and Namilamadeta, was also found by Beck et al. (2022). Although long tentatively referred to Wynyardiidae, the monophyly of the one known species of Wynyardia, W. bassiana Spencer, 1901, with those of Muramura, Namilamadeta and Ayekaye has never been robustly demonstrated. In part, this reflects the overwhelming systematic focus that mammalian palaeontologists have historically put on the cheek dentition: W. bassiana was described from a partial skeleton lacking the dentition (Spencer 1901). Referral of further species to Wynyardiidae was initially based on similarities in post-cranial material and non-dental cranial characters (Rich & Archer 1979), and later on the basis of a conceptual ‘?wynyardiid’ dentition after Tedford et al. (1977), Rich & Archer (1979), and Pledge (1987). To more robustly assesses whether the referred ?wynyardiid taxa form a monophyletic group would require a phylogenetic analysis that more comprehensively samples non-dental characters, enabling inclusion of Wynyardia bassiana.

Diprotodontoid interrelationships also remain equivocal owing to their highly autapomorphic cheek tooth morphology, making identification of key dental homologies problematic. In this study, strong support was found under Bayesian inference for a monophyletic vombatoid–diprotodontoid clade (BPP = 0.88: Fig. 11). This clade received low support (0.16 BPP) in Beck et al. (2020), while Beck et al. (2022) instead recovered diprotodontoids forming a polytomy with ilariids and a wynyardiid–vombatid clade. Within Diprotodontoidea, earlier interpretations of palorchestid molar homologies, in particular those of the stylar cusps, have been used to polarize them as the sister group to diprotodontids (Murray 1990, Black 2006). Here, interpretations of stylar cusp homologies in diprotodontids (see char. 40, Supplementary data 2), as polarized by the upper molars of Raemeotherium yatkolai Rich, Archer and Tedford, 1978 (SAMA P43060), have contributed to a nested placement for palorchestids within Diprotodontoidea, sister to Ngapakaldia tedfordi. A sister relationship between palorchestids and species of Ngapakaldia is consistent with the original hypothesis of Stirton (1967), which was based largely by P3 morphology coupled with several basicranial attributes, though the latter have since been held as symplesiomorphies within Diprotodontoidea, after Archer (1984), and Murray (1986). In this study, only a representative sample of taxa from this clade were included in the analyses, and very few were scored for their postcranial skeletons. Comprehensive phylogenetic assessment targeted at diprotodontoids, including their cranial and postcranial skeleton as well as review of their dental homologies, is needed to more robustly resolve the interrelationships therein.

Comments on the diet and palaeoecology of Mukupirna fortidentata

The species of Mukupirna possess a unique suite of craniodental adaptations compared to other vombatiforms. It has been suggested that both Mu. nambensis and species of early vombatids may have eaten subterranean food items, such as roots and tubers, obtained via scratch-digging (Brewer et al. 2008, Brewer et al. 2018, Beck et al. 2020). Additional insights into the functional morphology of the group are provided by the complete upper and lower dentition of Mukupirna fortidentata sp. nov., as well as referred postcranial specimens.

Craniodental

One of the striking features of the upper and lower first incisors of Mu. fortidentata is their considerable labiolingual width. In anterior mammalian teeth (incisors and canines), their cross-sectional area is correlated to resistance to bending (Biknevicius et al. 1996, Bacigalupe et al. 2002, Freeman & Lemen 2008). With respect to Mu. fortidentata, the considerable labiolingual width of the first incisors suggests that they might have been able to resist high bending stresses associated with the exertion of high bite forces. Conversely, the mediolateral width of the first incisors is markedly narrower, tapering anteriorly towards a slightly pointed apex. This likely focused the bite force on a smaller area, improving their effectiveness at puncturing or fracturing food items (Samuels 2009, Martin et al. 2016).

Angulation of the incisors is also often related to the amount of pressure required for initial food processing (e.g., Martin et al. 2016). Steeply inclined lower incisors, such as those of Mu. fortidentata, have a bite point that is closer to the jaw adductor muscles, which effectively increases bite strength for a fixed gape by reducing the length of the out-lever (Samuels 2009, Druzinsky 2010). Steeply inclined lower incisors also provide the structural strength to receive high stress loads by distributing force down the longitudinal axis of the tooth, rather than oblique to it.

The lower incisor morphology of Mu. fortidentata bears some general similarity to that of species in the extinct eutherian suborder Tillodontia (e.g., see Gazin 1953), such as Tillodon fodiens (Marsh 1875). The diet of tillodonts has been interpreted to comprise a high proportion of tough vegetation, including roots and tubers (e.g., Schoch 1986, MacFadden 2000, Rose 2006). Compared to rodents, the lower incisor morphology of Mu. fortidentata bears notable resemblance to those with nut- and fruit-dominated diets, such as tree squirrels and the tree rat Mesembriomys gouldii (Gray 1843): e.g., in terms of labiolingual length, mesolateral width, inclination, and tapering of width towards the apex. In taxa with first incisors that are mesolaterally broad, a more transversely oriented blade is often formed, which functions as a wider cutting surface, representing a better structure for acquiring vegetation (Samuels 2009). This morphology is exemplified by specialist herbivores that engage habitually in gnawing/cropping woody/fibrous plant material, such as beavers, tuco-tucos and the lesser bamboo rat.

In Mu. fortidentata, the considerable depth of the dentary below p3 and m1 is expected to represent a further adaptation for resisting high stress loads associated with the incisors. Bite force potential is also likely increased by the relatively short and broad rostrum, reducing the out-lever and increasing the temporalis moment arm (Greaves 2000, Santana et al. 2012), with its generally robust morphology also providing the structural strength to receive high stress loads. Despite their markedly different diet, thylacoleonids have a broadly similar jaw morphology for much the same reason, with steeply inclined lower incisors, greatest depth of the dentary below p3, as well as a short and broad rostrum.

Several aspects of the cheek tooth morphology of Mu. fortidentata are also congruent with the processing of relatively hard food items. The large, splayed roots extending ventrally far beyond the alveoli most likely represent an adaptation to accommodate relatively high stress loads during mastication. The enamel on the cheek teeth is fairly thick (0.7 mm at the buccal margin of the upper molars), which can be an adaptation for resistance to tooth fracture induced by hard object feeding (e.g., Lucas et al. 2008, Vogel et al. 2008, Barani et al. 2012). The high crown height on the anterior molars (e.g., the juncture between the postprotocristid and cristid obliqua occurs at two-thirds of the crown-height on m1) may be an adaptation for the processing of food items high in abrasives (e.g., Janis 1988, Jardine et al. 2012, Damuth & Janis 2014), though less so than their hypselodont relatives.

The steep gradient in molar dimensions suggests that the neutral point, where the maximum bite force can be achieved without accompanying secondary rotational forces (Bramble 1978), is near the front of the cheek tooth row. Larger bite forces could be generated with more posterior biting, but the risk of condyle disarticulation is higher. This steep anteroposteriorly decreasing gradient in molar dimensions is not particularly suitable for thorough masticatory grinding by the posterior molars.

Among diprotodontians, the cheek tooth morphology of Mukupirna species have some similarity to those of the musky rat-kangaroo, Hypsiprymnodon moschatus Ramsay, 1876, and species of bettong, particularly Bettongia penicillata Gray, 1837. These taxa share molars with bulbous cusps linked by weak crests in a bunolophodont configuration, and third premolars with prominent ridges that descend the buccal and lingual faces. Hypsiprymnodon moschatus is frugivorous, primarily consuming the flesh of fruits and less commonly the seeds (Dennis 2002). By comparison, the diet of Bettongia penicillata is dominated by underground fungal fruiting bodies (truffles), though also includes invertebrates, seeds and other plant material (Zosky et al. 2017). The function of the more distinctly plagiaulacoid premolars in these macropodoids is to initiate and propagate cracks in hard food materials via a scissor-like puncture crushing or shearing action (e.g., McNamara 2014). In species of Mukupirna, the apices of the primary cusps appear to occlude more directly, as might be suitable for crushing or shearing by via transverse movement. As noted by Beck et al. (2020), the ridges on the premolar in Mukupirna may function to strengthen the enamel.

The molar morphology in Mukupirna species is also strongly reminiscent of that in species of the Old World monkey subfamily Cercopithecinae, such as macaques, guenons, and vervet monkeys. In cercopithecines, fruit is the most common dietary component, though many are quite opportunistic and also consume seeds, nuts, leaves, rhizomes, grasses, bark, insects and small vertebrates (e.g., Kay 1978, Harding 1981, Ungar 2019). The proportion of each food group taken varies between species, with these differences generally manifested in the prominence of molar structures; wherein more folivorous cercopithecines tend to have relatively higher cusps, longer shearing blades and larger crushing basins for a given tooth length than frugivorous relatives (Kay 1978). In primates, dietary composition is also influenced by the physical properties of food items in the context of seasonal fluctuations in their availability (van Schaik et al. 1993). Foods that are nutritious and relatively easy to process, such as fruits, are prioritized. During periods when these preferred foods are scarce, primates must switch to foods that are harder to process, of lower nutritional quality, or both (Marshall & Wrangham 2007, Rosenberger 2013, Berkovitz & Shellis 2018). One might expect dietary composition in species of Mukupirna to follow to similar patterns.

Postcranial skeleton

The postcranial specimens attributed to Mu. fortidentata are generally similar to the equivalent elements of Mu. nambensis, attesting to comparable functional morphology. In particular, with respect to the distal right humerus (NTM P13348), there are large attachment areas for the flexors and extensors of the elbow, wrist and manus, suggesting significant strength in these movements. Moreover, the depth of the radial and olecranon fossae, coupled with the range of movement indicated by the extent of the trochlea, suggest that the elbow was well braced against lateral stresses whilst still capable of being fully extended or fully flexed.

The shape and development of the styloid facet on the pisiform (NTM P13347) suggests moderate stress was placed on the pisiform–ulna joint. This differs from extant arboreal taxa, and reflects that comparatively more weight rested on the back of the manus. The enlarged attachment for the m. carpi ulnaris suggests significant strength in flexion of the wrist, perhaps associated with scratch-digging, consistent with suggestions by Beck et al. (2020).

With respect to the distal left tibia (NTM P13346), the orientation of the medial talar facet suggests that a greater degree of inversion of the pes was possible than in the vombatids and diprotodontids studied. This is supported by the deep groove for the tendon of the m. tibialis cranialis, which is one of the main muscles responsible for inversion of the pes.

The relatively shallow trochlea and near-horizontal malleolar facets of the talus indicate a mobile tibio-talar joint with little bracing against lateral stresses. The extensive navicular facet suggests a prominent, functional first digit. The narrow flexor groove and the caudoplantar edge of the talus suggests a proportionately smaller flexor tendon mass than in most other taxa studied.

The CLAJP of the calcaneus indicates a significant degree of movement/rotatory capability in the talocalcaneal joint. The orientation of the tuber is more like that seen in climbing and digging marsupials, and the tuber does not display the thickening typical of terrestrial diprotodontids (or Ngapakaldia). The cubonavicular facet also displays a degree of interlocking to help brace against lateral stresses that is more common amongst the climbing taxa studied.

Body mass

Mukupirna nambensis is among the largest marsupials known from the Etadunnan land mammal age (after Megirian et al. 2010), second only to species in the ilariid genus Ilaria (Beck et al. 2020). The largest craniodental specimens (e.g., NTM P13262: Table 1) referred to Mu. fortidentata are almost 20% larger than their equivalents on the Mu. nambensis holotype (AMNH FM 102646), while the smallest specimens (e.g., NTM P11999 Table 1) have overlapping dimensions. The referred postcranial specimens are, conversely, smaller than those of Mu. nambensis by up to 20%, with exception of the right pisiform, NTM P13347.

The body mass estimates for Mu. fortidentata are quite varied (Table 4), and as such, should be treated tentatively. That based on minimum humeral circumference (NTM P13348) yielded an estimate of 48 kg, whereas the dental-based estimates from the holotype (NTM P11997) ranged from 26 to 84 kg, reflecting the unequal cheek tooth proportions in the taxon. The largest dental specimen referred to the taxon (left P3, NTM P13262) yielded a body mass estimate 31% higher than the corresponding value derived from the holotype (Table 4). These body mass estimates for Mu. fortidentata overlap with that for Mu. nambensis derived from total skull length (46 kg) in Beck et al. (2020), but differ markedly from those based on femoral (143 kg) and humeral (171 kg) circumference. We regard the most parsimonious explanations for this discrepancy to be that either: some of the postcranial specimens are incorrectly referred to Mu. fortidentata; or the postcranial skeleton of Mu. fortidentata was proportionately smaller than that of Mu. nambensis.

Table 4. Body mass estimates (in kg) for Mukupirna fortidentata, based on measurements taken from craniodental (following Myers 2001) and postcranial (following Richards et al. 2019) material.

Specimen	Variable x (mm)	Regression equation	Smearing estimate (%)	Body mass (kg)
NTM P13348 L	MHC (63)	logBM = 2.671(log ×)−0.128	–	48
NTM P11997 R	UMRL (44.3)	logBM = –0.418 + 3.011(log x)	4.4	36
NTM P11997 R	UMORL (41.3)	logBM = –0.567 + 3.072(log x)	4.7	26
NTM P11997 R + L/2	3UPW (11.25)	logBM = 1.775 + 2.991(log x)	1.3	84
NTM P13262 L	3UPW (12.3)	logBM = 1.775 + 2.991(log x)	1.3	110

Abbreviations: logBM, log body mass; MHC, minimum humeral circumference; 3UPW, upper third premolar maximum width; UMORL, upper molar row length at the crown; UMRL, upper molar row length at the alveoli.

Palaeoecological conclusions

Mukupirna fortidentata shows several craniodental adaptations for the processing of hard plant material under relatively high stress loads. In particular, the robust and steeply inclined lower incisors with somewhat pointed apices likely played a role in piercing, fracturing and/or gnawing solid food items prior to mastication. We expect that Mu. fortidentata was something of a generalist, with a diet that may have included fruits, seeds, nuts, bulbs, tubers, rhizomes, and, to a lesser degree, shoots and leaves. The postcranial material of Mu. fortidentata shows several characteristics, broadly consistent with those of Mu. nambensis, that suggest the elbow, proximal carpus and tarsus had a high range and strength of movement, as well as being generally well braced against lateral stress.

Later vombatoids, in comparison, have craniodental adaptations (including hypselodonty) for the mastication of more abrasive plant material. Therefore, it is possible that in Vombatoidea the specialization for processing hard food items preceded that for the consumption of more abrasive plant material. Alternatively, and perhaps more likely, by the late Oligocene, prior to the evolution of hypselodonty, early vombatoids may have already radiated to fill a variety of niches for the processing of hard, tough and/or abrasive plant material.

Biochronological significance

By resolving that Vombatomorphia? fam., gen. et sp. nov. of Murray and Megirian (2006) represents a species of Mukupirna, we provide additional support for the relative temporal proximity of the Pwerte Marnte Marnte Local Fauna and the basal local faunas of the Namba Formation as well as, by implication, their Etadunna correlates (see Murray & Megirian 2006, Crichton et al. 2023), which together form the Etadunnan land mammal age (Megirian et al. 2010). With respect to stage-of-evolution, Mu. fortidentata has several craniodental attributes that appear more derived than in Mu. nambensis, as polarized by the morphologies presented by species of Muramura and Namilamadeta, which could be interpreted as support that the Pinpa LF is in fact older than the Pwerte Marnte Marnte LF. These attributes include: a proportionally steeper anteroposteriorly decreasing molar width gradient; splaying of the molar roots outwards from the crown towards the maxilla/dentary; and lacking a diastema between the I3 and C1. That being said, it could also be interpreted that Mu. fortidentata presents the more plesiomorphic condition with respect to the alignment pf P3 relative to M1 and, possibly, the relative size of the postcranial skeleton. Comprehensive sampling of the fauna is needed to more precisely constrain the relative age of the Pwerte Marnte Marnte Local Fauna, ideally coupled with additional age constraints.

Conclusion

Mukupirna fortidentata sp. nov. from the late Oligocene Pwerte Marnte Marnte Local Fauna, Northern Territory, is an early vombatoid, closely related to Mu. nambensis from the late Oligocene Pinpa Local Fauna of the Namba Formation. The phylogenetic placement of the maradid Marada arcanum is resolved to lie within Vombatoidea, as sister to species of Mukupirna + Vombatidae. Mukupirna fortidentata and Ma. arcanum appear to lack any synapomorphies in the dentary or lower dentition that would support their forming a clade to the exclusion of Vombatidae. Craniodental attributes of Mu. fortidentata suggest that this species was well adapted to processing hard plant material under relatively high stress loads. The dental material from Mu. fortidentata also assists in the identification of two further allied taxa: a vombatid from the younger late Oligocene Tarkarooloo Local Fauna, South Australia; and a ?vombatoid from the earliest Miocene Geilston Bay Local Fauna, Tasmania. The Lake Tarkarooloo taxon, in particular, provides support that early vombatids were present by at least 24.1 Ma in the late Oligocene. This would imply that representatives from all three referred vombatoid families had originated by the late Oligocene.

Acknowledgements

We thank M.-A. Binnie and D. Stemmer, respectively, for providing access to South Australian Museum Palaeontology and Mammal Collections, as well as T. Ziegler and R. Schmidt for providing access to the Museums Victoria Palaeontology Collections. We are grateful to W. Klein from Orange Creek Station for allowing access to the Pwerte Marnte Marnte fossil site. We thank C. Burke, S. Arman, W. Handley and G. Gully for organizing and/or assistance with the 2014, 2020 and 2022 field trips to the site. We also thank C. Burke for preparatory guidance, and J. Blokland for support relating to the phylogenetic component. We acknowledge the Southern Arrernte People, known as the Pwerte Marnte Marnte Aboriginal Corporation, for their custodianship of the lands on which the fossil locality is situated. We thank reviewers K. Black, P. Brewer and J. Louys, and editor Robin Beck, for their thoughtful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplementary research materials for this article can be accessed at https://doi.org/10.1080/03115518.2023.2181397.

Additional information

Funding

AIC was supported by The Australian Government Research Training Program Scholarship. Support for the 2014 field trip to the site was provided by a Patterson Memorial Grant from the Society of Vertebrate Palaeontology to AMCC.