Cranial osteology of the ankylosaurian dinosaur formerly known as Minmi sp. (Ornithischia: Thyreophora) from the Lower Cretaceous Allaru Mudstone of Richmond, Queensland, Australia

Ankylosauria

Ankylosauria is a clade of quadrupedal, herbivorous ornithischian dinosaurs, whose members are characterized, in part, by extensive parasagittal rows of dermal ossifications on the dorsal and lateral surfaces of the neck, trunk and tail, and an unusual cranial architecture that incorporates ornamentation (also the result of dermal ossifications) (Vickaryous, Maryańska & Weishampel, 2004). Ankylosauria and its sister taxon Stegosauria are united to form the clade Eurypoda. Eurypodans and a number of closely related ornithischians that are similarly characterized by extensive dermal ossifications (e.g., Scutellosaurus lawleri, Emausaurus ernsti and Scelidosaurus harrisonii) collectively form the clade Thyreophora (Sereno, 1986; Sereno, 1999; Parish, 2005; Thompson et al., 2012). Ankylosauria generally includes two clades: Nodosauridae and Ankylosauridae (Sereno, 1986; Sereno, 1999; Coombs & Maryańska, 1990; Lee, 1996; Kirkland, 1998; Xu, Wang & You, 2001; Vickaryous et al., 2001; Vickaryous, Maryańska & Weishampel, 2004; Hill, Witmer & Norell, 2003; Thompson et al., 2012). However, a number of analyses also weakly support the inclusion of a third clade: Polacanthidae/Polacanthinae (see Kirkland, 1998; Carpenter, 2001; Parish, 2005).

Phylogenetic analyses that include Minmi (excluding those conducted by Leahey, Molnar & Salisbury, 2008; Leahey, Molnar & Salisbury, 2010; Arbour & Currie, in press) have treated the holotype Minmi paravertebra (QM F10329) and the Marathon specimen (QM F18101) as the same genus, despite the fact that the latter was only tentatively assigned to Minmi based on one defining feature (discussed above) and was not fully described or taxonomically assessed. Despite this, the specific phylogenetic position of Minmi has varied between analyses. This is most likely due to the limited published information on QM F18101’s osteology, both for the skull and the postcranium, as well as the fragmentary condition of Minmi paravertebra. Generally, Minmi has been assigned to two relatively basal positions within Ankylosauria. The taxon has been regarded as a stem ankylosaurian—that is positioned basal to the two major ankylosaurian clades (Kirkland, 1998; Carpenter, 2001; Leahey, Molnar & Salisbury, 2008; Leahey, Molnar & Salisbury, 2010; Arbour & Currie, in press). Other phylogenetic analyses have found Minmi to be a basal member of Ankylosauridae (e.g., Sereno, 1999; Xu, Wang & You, 2001; Hill, Witmer & Norell, 2003; Vickaryous, Maryańska & Weishampel, 2004; Thompson et al., 2012). Herein, we provide further information for the skull that may enable future clarification of the taxonomic relationships of QM F18101.

Regardless of the conflicting placements of QM F18101, its basal position in relation to Ankylosauria fits well with its ‘mid’ Cretaceous age. The early evolution of ankylosaurian thyreophorans has been difficult to decipher due to the rarity and fragmentary nature of specimens from the Jurassic and Lower Cretaceous, particularly in the landmasses that once comprised Gondwana. Due to its completeness and age, QM F18101 thus represents an ideal specimen with which to investigate aspects of the early evolution of Ankylosauria.

Ontogenetic stage of the specimen

Our understanding of ontogeny in ankylosaurian dinosaurs is dominated by changes that occur in the postcranial skeleton (e.g., Maryańska, 1977; Coombs, 1986; Coombs, 1995; Coombs & Maryańska, 1990; Jacobs et al., 1994; Pereda Suberbiola, 1994; Godefroit et al., 1999; Pereda Suberbiola & Barrett, 1999; Xu, Wang & You, 2001; Hill, Witmer & Norell, 2003; Vickaryous, Maryańska & Weishampel, 2004), with more recent studies having utilised the microstructure of bone and various dermal ossifications (Hayashi et al., 2010; Stein, Hayashi & Sander, 2013; Ősi et al., 2014). Far-fewer ontogenetic changes have been recognised in the cranium. The most widely accepted change is the progressive fusion of the cranial sutures on the dorsal surface of the skull (e.g., Maryańska, 1977; Jacobs et al., 1994; Godefroit et al., 1999; Xu, Wang & You, 2001; Hill, Witmer & Norell, 2003; Vickaryous, Maryańska & Weishampel, 2004; Burns et al., 2011). Beyond this, most of the discussion on the identification of ontogenetic stages in ankylosaurians based on cranial morphology has been sourced from the numerous juvenile specimens of Pinacosaurus (Maryańska, 1971; Maryańska, 1977; Godefroit et al., 1999; Hill, Witmer & Norell, 2003; Burns et al., 2011) and other genera where numerous specimens are known (e.g., Euoplocephalus Penkalski, 2001; Adontosaurus Arbour & Currie, 2013). In all cases, however, it is unclear whether the traits discussed are truly ontogenetic or if they are the result of taxonomic differences or varying states of preservation Hill, Witmer & Norell, 2003; Penkalski, 2001; Arbour & Currie, 2013; Burns et al., 2011. For example, Burns et al. (2011) recently stated that larger specimens of Pinacosaurus have a greater number of marginal denticles on their teeth than smaller specimens of the same genus, thus this may represent an ontogenetic trait. However a reduced number of marginal denticles is also a feature of some nodosaurids (Coombs & Maryańska, 1990), thus it is possible that the feature Burns et al. (2011) suggested may be genus-specific ontogenetic trait, but clearly more specimens of other ankylosaurians and further investigation is necessary. Histological studies alongside gross morphological studies are likely to help assess whether these traits discussed by the aforementioned authors are ontogenetic or otherwise.

QM F18101 could be considered a sub-adult ankylosaur based on its small size along with the absence of fusion of the cranial elements, scapula-coracoid and pelvic elements (Molnar, 1996a), however Molnar argued that other factors suggest that it was near-mature to mature at death.

The lack of fusion of various cranial and postcranial elements seen in QM F18101 may not necessarily be related to immaturity, as these features are also plesiomorphic for the clade and cranial fusion may have developed only in more derived taxa (Molnar, 1996a). Two of the four known ankylosaurian skulls that exhibit cranial sutures are agreed to be juvenile specimens of the derived genus Pinacosaurus, a derived genus of Late Cretaceous ankylosaurian. However the other two, QM F18101 and Cedarpelta (Carpenter et al., 2001), are thought to be sub-adult–adult(?) and adult respectively. These taxa are positioned basally in relation to Ankylosauria, either as stem ankylosaurians (Leahey, Molnar & Salisbury, 2008; Leahey, Molnar & Salisbury, 2010; QM F18101 only, Arbour & Currie, in press) or basal members of clades within Ankylosauria (Hill, Witmer & Norell, 2003; Vickaryous, Maryańska & Weishampel, 2004; Thompson et al., 2012; Cedarpelta, only Arbour & Currie, in press). Furthermore, it is also noteworthy that a lack of fusion between the scapula and coracoid is also seen in some adult ankylosaurs (Coombs & Maryańska, 1990; Carpenter et al., 1999; Pereda Suberbiola, 1994), such that the ontogenetic utility of this feature may be more complex than previously thought.

Molnar (1996a) considered it noteworthy that all the specimens assigned to ‘Minmi’ (QM F10329, QM F33286, QM F33565, QM F33566, AM F35259, AM F119849) are approximately the same size (approximately 2.5–3 m total length) despite the fact that they originate from different locales and time periods, with the implication being that it would be highly unlikely that only juveniles were preserved. Furthermore the majority of early forms of ankylosaurs are small (less than 3 m e.g., Mymoorapelta maysi Kirkland & Carpenter, 1994, Gargoyleosaurus parkpinorum Kilbourne & Carpenter, 2005), compared with the more derived taxa from the Late Cretaceous of Asia and North America (approximately 6 m e.g., Ankylosaurus Carpenter, 2004). This would also suggest that the features that are typically considered ontogenetic in the more derived ankylosaurians may actually be the plesiomorphic condition.

A bone microstructure analysis was planned for this study in an effort to further assess the maturity of the specimen. Unfortunately preliminary testing of both Minmi paravertebra and QM 18101 revealed that the bone has undergone considerable remineralisation and oxidisation (a common process in surface rocks of these locales), such that histological investigations were not possible. A more detailed study of the ontogenetic status of QM F18101 is currently being undertaken with the description of its postcranial elements. For the present study, we concur with Molnar (1996a) that, in the absence of any unambiguous ontogenetic characters, it is most prudent to consider QM F18101 an almost mature or newly mature individual.

Preservation

The skull of QM F18101 has an unusual morphology for an ankylosaurian, and its state of preservation makes the interpretation of some aspects of its osteology problematic, particularly on the palate. For these reasons we have included a separate preservation section (a more detailed version will be available in Leahey, PhD thesis).

The skull of QM F18101 is approximately 90 per cent complete (Figs. 2 and 3). The majority of sutures are evident; only the suture between the parietals is completely fused, while the suture between the frontals is partially fused but still discernable. Matrix and dermal ossifications (mainly ossicles) occupy the nasal vestibules, orbits and adductor chambers. Matrix, ossicles and isolated teeth obscure parts of the rostral opening of the nasal cavity and the choanae. Matrix covers the dorsal two-thirds of the braincase and is also present in the interptyergoid vacuity, cranioquadrate canal and the foramen magnum. Numerous fractures transect the skull, and some of these have additionally resulted in the displacement of some elements (see Figs. 2 and 3).

The apex of the maxillary rostrum is not preserved, such that the nares as well as the rostral-most parts of the premaxillae, maxillae and nasals are missing. The left premaxillary ‘beak’ was removed from the nasal vestibule during preparation of the skull (see Molnar, 2001b: Fig. 16.5, N.B. it is not labelled) (Fig. 4). The right premaxillary beak was not recovered.

The left lateral side of the skull is well preserved (Figs. 3A and 3B), but some slight dorsoventral crushing has occurred to the ventral edge of the maxilla and the lacrimal. The orbital process of the jugal has broken and has been displaced slightly dorsally, its rostral tip is lost. The ventral margin of the jugal and quadratojugal are not preserved. Osteoderms, ossicles and matrix obscure (from left lateral aspect) the caudal-most edge of the left lateral side of the skull (parts of the squamosal, postorbital, jugal, quadratojugal, quadrate, and the quadrate foramen) (Figs. 3A and 3B).

The right lateral side of the skull has been crushed ventromedially (Figs. 3C and 3D), affecting the preservation of all elements from the right lateral side. Surface bone is missing in parts, some elements are highly fragmented and all bones have moved from their original positions: the right nasal vestibule is laterally flattened; the ventral part of the right maxilla, including the maxillary tooth row, has twisted medially and now sits within the oral cavity obscuring parts of the palate; the ventral parts of the right jugal and quadratojugal have broken and now lie lateral to the cheek region; the orbital process of the jugal has also moved caudolaterally; and, the ventral-most part of the squamosal is missing. A long piece of bone of unknown origin extends across the orbital cavity (Figs. 3C and 3D).

In ventral aspect (Figs. 2C and 2D), a considerable amount of crushing is apparent, with the majority of the ventrally exposed bones exhibiting varying degrees of fragmentation. The right maxilla has rotated medially into the oral cavity, disarticulated caudally, and now overlies the rostral edge of the right pterygoid and the caudal end of its rostral process. The caudal-most part of the right maxilla and right ectopterygoid are missing. Although highly fragmented the left maxillary tooth row remains in its in vivo position. Extensive fracturing obscures the sutural boundaries of the left ectopterygoid. The roots of fourteen teeth remain in their alveoli in the right maxilla, and twenty in the left maxilla. Some disarticulated and broken crowns and roots are scattered on the maxillae. The maxillary secondary palate and the vomer exhibit fragmentation as a consequence of the medially turned right maxillae. The caudal edge of the maxillary secondary palate and the vomer are not preserved (Figs. 2C and 2D). The location, or presence, of the palatines is not clear. Additionally the resolution of the CT imagery is poor in this area and thus they cannot be located via this method (Figs. 2C and 2D). In ventral aspect, the parasphenoid has disarticulated from the basisphenoid, and the adjacent areas are slightly damaged. The rostral processes of the pterygoids have moved laterally slightly relative to the parasphenoid. Collectively, the palatal surfaces of the parasphenoid and the rostral processes of the pterygoids are directed slightly dorsal and to the left lateral (Figs. 2C and 2D). The rostral process of the left pterygoid has disarticulated and rotated medially, such that its lateral surface now lies in the horizontal plane. An irregular-shaped portion of the lateral surface of the left pterygoid is missing. The caudal ends of the rostral processes of both pterygoids are obscured by matrix. The left pterygoid remains in natural articulation, apart from its contact with the basisphenoid. Lateral to this articulation, the medial half of the ventral surface of the left pterygoid is missing. A fragment of the bone, possibly from the mandible, overlies most of the ventral surface of the left pterygoid. The right pterygoid-basisphenoid contact is also broken, and the right pterygoid now lies slightly dorsolateral to the basisphenoid, such that it is unclear if the pterygoid–basisphenoid suture was fused. A small portion of the mandibular process of the right pterygoid is broken (Figs. 2C and 2D). The ventral edges of the jugals, quadrates (including the articular condyles) and quadratojugals are not preserved. The right quadrate is displaced, a fracture across its shaft, so that it now lies in the horizontal plane, across the adductor chamber. The mandibular condyle of the right quadrate is now angled rostrally and overlies the caudal end of the right maxilla, lateral to the caudal-most alveolus (Fig. 2D ‘q_i’), whereas the squamosal-articular end of the right quadrate underlies the right pterygoid but overlies the lateral end of the right otoccipital (Fig. 2D ‘q_ii’).

Although the lateral walls of the braincase (Figs. 2C and 2D) have some minor fractures, the general regions that comprise it (e.g., the prootic, laterosphenoid, orbitosphenoid) are discernible in the CT scan data, as are some but not all of the sutures between them.

In caudal aspect, all elements are preserved with some minor fragmentation (see Figs. 3G and 3H), however much of the supraoccipital is missing, thus exposing the parietal–otoccipital–supraoccipital contact.

The dentary and the coronoid process of the left mandibular ramus are partially preserved, whereas only a portion of the dentary from the right mandibular ramus remains.

Osteology

Overall, the skull is approximately 30 per cent longer than it is wide. In dorsal aspect, the skull has a pentagonal outline, with the greatest width occurring immediately caudal to the orbits. In caudal aspect the occiput is transversely wider than it is high dorsoventrally (Figs. 3G and 3H), much of it can also be observed in dorsal aspect (Figs. 2A and 2B). In ventral aspect the skull is triangular in outline, with the greatest width towards the caudal edge, and the tooth rows are medially inset (Figs. 2C and 2D). In lateral aspect, the cranial roof is flat, rising slightly dorsally immediately caudal to the orbits (i.e., at the postorbital). There is also a slight arching of the maxillary rostrum (within the rostral half of the nasals) (Figs. 3A–3D). The dorsal surfaces of the prefrontals, supraorbitals and postorbitals form an approximately 90° angle with the lateral surfaces of the skull. Dermal ossifications are present on the skull (see below and Molnar, 2001b). The majority of cranial sutures on QM F18101 are unfused, with the exception of some of those involved in the braincase. Other sutures may be obscured from view by dermal ossifications (Figs. 2 and 3). The antorbital, supratemporal and lateral temporal fenestrae are closed but a quadrate foramen is retained (Figs. 2A–2B and 3A–3B).

Rostral region

Premaxilla

The premaxilla comprises a main lateral body and a rostral beak. The main body of the premaxilla forms approximately 25% of the total length of the preserved portion of the skull in lateral aspect (Figs. 3A–3D). The premaxilla is a vertically orientated plate. The caudal-most third is rectangular and completely bordered by the maxilla. Immediately rostral to this, the dorsal margin sharply rises dorsally at approximately 70°, to a point where the sutures of the nasal, premaxilla and maxilla intersect. Rostral to this point the dorsal margin of the premaxilla extends rostroventrally towards its ventral margin at approximately 45°. The dorsal margin of the premaxilla is bordered dorsally and medially by the nasal. The dorsal edge of the premaxilla and the lateral-most edge of the nasal contact each other at an oblique angle. This contact occurs lateral to the nasal vestibule. It is unclear if the premaxilla contributes to the border of each naris. The premaxilla is bordered ventrally by the maxilla along its entire preserved length. The ventral margin, rostral to the caudal rectangular portion of the premaxilla (described above), dips sharply ventrally and then grades more gently towards the apex of the maxillary rostrum. Thus the rostral two-thirds of the premaxilla tapers towards the apex of the maxillary rostrum. The internal surface of the premaxilla is concave and contributes to the rostral ventrolateral walls of the nasal cavity. The interior surface and rostral-half of the exterior surface of the premaxilla consist of smooth bone (Fig. 5C). Along the ventral margin of the left premaxillae is an area that is roughened with scattered foramina (Fig. 3A). Such foramina are often cited as evidence for the possible presence of a rhamphotheca, particularly amongst ornithischian dinosaurs (Barrett, 2001), although Cuff & Rayfield (2015) noted that the presence of premaxillary foramina alone is not sufficient evidence. Nevertheless, many workers have regarded a diversity of ankylosaurs as having beaks that likely were covered with a rhampotheca (Coombs, 1971; Coombs & Maryańska, 1990; Rybczynski & Vickaryous, 2001), an interpretation that we follow here for QM F18101.

The preserved portion of the left premaxillary beak (Fig. 4) has a triangular outline approximately 50 mm dorsoventrally long and 40 mm at its greatest width. The medial and lateral edges are straight. The rostral edge curves at its junctions with the medial and lateral edges, thus suggesting the presence of a premaxillary notch. The element is concave internally and very slightly convex externally, the rostral edge curves at a greater angle. It is at the most 3 mm thick on the medial edge, and thins towards the lateral. The external surface is dotted by foramina, some considerably deep, and the internal surface is smooth.

Temporal region

Palatal region

Pterygoid

The pterygoid comprises a main body and a rostral process. It is visible mainly in ventral aspect, but the caudal edge of the main body is visible in caudal aspect. The pterygoids do not contact one another caudally near the braincase; they are completely separated by the basisphenoid. In ventral aspect, the main body of the pterygoid forms a broad concave (very slightly convex dorsally) plate that is divided medially by the basisphenoid (Figs. 2C–2D and 3G–3H). The rostrolateral corner thickens where it contacts the ectopterygoid, but due to poor preservation, this contact is unclear. The lateral edges of the pterygoid arch ventrally, contacting the ventromedial portion of the pterygoid processes of the quadrate. The caudolateral corner of the pterygoid contacts the pterygoid ramus of the quadrate via a scarf joint. The caudal margin thins caudally. The quadrate process is dorsoventrally thin and caudolaterally directed. A small, circular foramen is present on the right pterygoid, close to the rostral edge and approximately a third of the length of the element from its medial edge. The corresponding area on the left pterygoid is obscured.

The rostral (vomerine) process of the pterygoid arises from the rostral edge of the main body of the pterygoid, most likely near the junction with the ectopterygoid (Figs. 2C and 2D). The processes are approximately 30% longer rostrocaudally than at its widest point transversely. The rostral processes arch medially to meet at the midline approximately level with the orbit, thus forming an interpterygoidal vacuity over the parasphenoid. The rostral processes of the pterygoids are closely associated with the parasphenoid, and appear to have closely neighboured it laterally. This is evident along the medial edge and surface, which parallel the fusiform shape of the parasphenoid. Due to poor preservation, it is unclear where the pterygoid-palatine contact is situated.

The rostral (vomerine) processes of the pterygoid were originally thought to be the palatines, since in ankylosaurians the palatine and pteryoid form a broad, rostrally extensive shelf (‘caudoventral’ secondary palate) that obscure the parasphenoid from view. However compared with other dinosaurs, particularly stegosaurians and other ornithischians, the structures seen in QM F18101 are more similar to the rostral processes of the pterygoids and will thus be interpreted as such herein (Fig. 6E). Unfortunately CT resolution is poor in this area such that it cannot be utilised to verify this interpretation.

Braincase region

Mandible

The preserved portions of the left mandibular ramus (dentary + coronoid process) are described in detail by Molnar (1996a) and will not be commented on further here. Figure 6A presents a restoration of the mandible in lateral aspect.

Dermal ossifications

The majority of the dermal ossifications of QM F18101 were described in Molnar (1996a) and Molnar (2001b). Of note, the majority of the dermal ossifications on the skull are generally flat. QM F18101 does not possess any spines, horns or ‘boss-like’ dermal ossifications on its skull. The dermal ossifications with the most topography are those of squamosal-postorbital region, which have shallow keels. The largest of the keels occurs on the most ventral of these. This dermal ossification is missing a ventral portion, however the end of the keel is complete suggesting that it is only missing the ventral edge. Thus it appears that QM F18101 does not possess a quadratojugal or squamosal horn or ‘boss’.

The dermal ossifications of the orbit and nasal regions were not discussed by Molnar (1996a), thus they are described herein. Three dermal ossifications are associated with the dorsal half of the orbit (Figs. 3A–3D). These are most clearly seen in left lateral aspect, as all three are not fused to the supraorbital and postorbital bones. The most rostral dermal ossification (D1) is a thin, quadrilateral-shaped element with its long axis aligned at approximately 45° to the rostrolateral edge of the supraorbital. The middle dermal ossification (D2) is very slightly crescentic in shape. It lies directly ventral to D1, dorsoventrally along the caudal edge of the lacrimal. The most caudal dermal ossification (D3) has a crescentic outline and lies along the ventral edge of the supraorbital and the rostral edge of the postorbital. A similar series of ossifications on the right lateral side of the skull differs slightly from those on the left. D1 and D3 on the right are fused to supraorbital and the postorbital, respectively, whereas D2 remains detached. These elements may not be analogous to the ‘anterior and posterior supraorbital caputegulum’ (Arbour & Currie, 2013) as they occur within the orbit, ventral to the supraorbital and rostroventral to the postorbital. The anterior and posterior supraorbital caputegulum of all other ankylosaurians occur on the lateral edge of the skull roof, dorsolateral to the supraorbital.

The oblong dermal ossification that occurs in the right nasal vestibule (Figs. 3C and 3D) may resemble the condition seen in Pinacosaurus (Maryańska, 1977), particularly the dermal ossification that occurs more laterally to the ‘lateral osteodermal mass’ that caps the premaxillary-nasal region in Pinacosaurus grangeri (IVPP V16853) (Burns et al., 2011: Fig. 2). The exact in vivo position of this element, however, is unclear.

Previously described by Molnar (1996a) are irregular ‘grooves’ or sulci on the dorsal surface of the skull (Fig. 7), which he presumed to be impressions caused by epidermal plates/scutes, most likely the ‘caputegulae’ described in other ankylosaurians (Blows, 2001; Arbour & Currie, 2013). If QM F18101 did possess these dermal ossifications, their lack of preservation may be explained by them not being ossified (i.e., keratinous) or not being co-ossified to the skull (Vickaryous, Russell & Currie, 2001; Arbour et al., 2013). These sulci are generally deeper on the right side of the skull than they are on left (Fig. 7). Additionally, the central part of the skull table, the frontals and the area immediately surrounding it, are slightly depressed but not to the extent of the sulci. Surrounding these sulci, parts of the supraorbitals, prefrontals, nasals and postorbitals have a rugose texture. Patches of this rugose texture are also found within the depressed area but not within the sulci (Fig. 7). This most likely represents reworking of the bone (‘periosteal osteogenesis’ Carpenter et al., 2001).

Nasal cavity

The nasal cavity of QM F18101 presents some challenges to interpret given that (1) ankylosaurians in general have unusual and highly divergent nasal cavities that loop through the snout (Witmer & Ridgely, 2008: 1379 & 1383; Miyashita et al., 2011), (2) QM F18101 likely has a fairly basal phylogenetic position within Ankylosauria such that it is not clear to what extent the nasal morphology of other ankylosaurians applies to QM F18101, and (3) QM F18101 is incomplete rostrally and the CT scan data are often equivocal. These caveats aside, we here attempt to describe the major features of the nasal cavity based on digital segmentation of the CT scan data (Fig. 8, see also Supplemental Information 1).

As in other vertebrates, three main parts of the nasal cavity can be identified: the vestibule, the nasal cavity proper, and the olfactory region. The rostral portion of the skull is missing, and thus the full extent of the nasal vestibule cannot be restored. The vestibule is largely bounded by the premaxilla and nasal bones, both of which are better preserved on the left side. Not enough of the premaxillae are preserved to determine whether the pattern of premaxillary air sinuses and narial apertures described for Pinacosaurus (Hill, Witmer & Norell, 2003) were present. Likewise, much of the looping of the airway that was described as passing partly through the nasal vestibule in Panoplosaurus and especially Euoplocephalus (Witmer & Ridgely, 2008) cannot be fully assessed in QM F18101 due to incomplete preservation. Nevertheless, enough is preserved to suggest that QM F18101 had a more complicated airway than did typical non-ankylosaurian dinosaurs (Witmer, 1997; Witmer, 2001; Sampson & Witmer, 2007; Witmer & Ridgely, 2008; Bourke et al., 2014) (Fig. 9). The position of the nostril cannot be known precisely due to absence of the rostral-most portion of the snout, but it was presumably located rostroventrally as in other ankylosaurians (Hill, Witmer & Norell, 2003; Witmer & Ridgely, 2008), and indeed most other dinosaurs and amniotes in general (Witmer, 2001). The choanae opens into the oral cavity caudal to the maxillary secondary palate (Fig. 8C).

The course of the airway between the nostril and choanae is not entirely clear but some options are presented here (Fig. 9). As shown in Figs. 8 and 9, the main airway passed from the nostril region to run dorsomedially adjacent to the nasal septum (which is discernible in QM F18101, even if incompletely mineralized and preserved, as noted above). The airway passed dorsally over the maxillary secondary palate to reach a large open chamber, comprising the nasal cavity proper, bounded by the nasal and frontal bones dorsally, the maxilla, premaxilla, and lacrimal laterally, and maxilla ventrally. The medial portion of the chamber is presumably the olfactory recess (coloured red in Figs. 8 and 9), given its location adjacent to the olfactory bulb of the brain in the typical cul-de-sac caudodorsomedial to the choana (Witmer & Ridgely, 2008). There are no olfactory turbinates preserved in QM F18101 to confirm this functional segregation of the nasal cavity proper, but this relationship is highly conserved evolutionarily, and other ankylosaurians have been described with olfactory turbinates in this region (Maryańska, 1977; Witmer & Ridgely, 2008). The main respiratory (i.e., non-olfactory) airway passed more laterally into the nasal cavity proper (Fig. 9).

From this chamber, the airway may simply have passed ventromedially to reach the choana. In this non-looping-airway hypothesis (Figs. 9B and 9E), the open space extending rostrally above the maxillary secondary palate could be regarded as part of the nasal cavity proper and the space within the maxilla above the tooth row would be a paranasal sinus homologous with that of the antorbital sinus of other archosaurians, which would be consistent with earlier interpretations of ankylosaurians nasal cavities (e.g., Maryańska, 1977; Witmer, 1997; Vickaryous, Maryańska & Weishampel, 2004), prior to the discovery of the looping airways of both nodosaurids and ankylosaurids (Witmer & Ridgely, 2008). Again, given the potentially very basal position of QM F18101, a more plesiomorphic condition of the nasal cavity cannot be ruled out on phylogenetic grounds. However, a looping-airway hypothesis is also tenable (Figs. 9C and 9F), in which case the rostral space above the secondary palate could have transmitted a loop of the airway. Supporting the looping-airway hypothesis is the rostrally-facing, basin-shaped structure in the nasal bone described above (Figs. 5B–5C and 8). The nasal basin is hard to interpret as anything other than as a structure housing some portion of the nasal cavity. The space within the basin is walled off from the dorsomedial tract of the main airway by the nasal bone (Figs. 5B and 5C), suggesting that the nasal basin housed either a novel paranasal air sinus within the nasal bone or a loop of the airway within the nasal vestibule (Figs. 9C and 9F). Unfortunately, the front portion of the snout that would have housed these nasal loops or sinuses is not preserved, and thus we cannot know how the space within the nasal basin connected with other components of the nasal air spaces. Likewise, in the looping-airway hypothesis, it is not clear how the airway would loop back to the choana, and it is necessary to invoke that the airway looped both fore and aft through the space above the secondary palate (Figs. 8C and 8F), which is reasonable but for which there is no osteological evidence at this time. Thus, although the precise course of the airway in QM F18101 is not clear at present, there is at least some evidence (e.g., the nasal basin, partial partitioning of the nasal cavity proper) that some of the complexity of the nasal cavity that characterized more derived ankylosaurians (Witmer & Ridgely, 2008) had evolved already in QM F18101.

Cranial (brain) endocast

A cranial endocast of QM F18101 was generated from the CT scan data (Figs. 10A–10C, see also Supplemental Information 1) and provides information on brain structure in QM F18101. As is typically the case for reptiles (Hopson, 1979; Witmer et al., 2008), the neural tissue of the brain apparently did not fill the endocranial cavity and thus the endocast is a reflection of the dural envelope more so than the brain itself. The cranial endocast has been presented previously for a range of ankylosaurians, including Euoplocephalus (Figs. 10D and 10E, herein) (Coombs, 1978b; Hopson, 1979; Witmer & Ridgely, 2008; Miyashita et al., 2011), Polacanthus (Norman & Faiers, 1996), Struthiosaurus (Pereda Suberbiola & Galton, 1994), Panoplosaurus (Witmer & Ridgely, 2008), and Hungarosaurus (Ősi, Pereda Suberbiola & Földes, 2014). The endocast of QM F18101 is generally well preserved with little to no evidence of distortion. It is similar to those of other ankylosaurians but has some important general differences. For example, the length of endocast represents approximately 40% of QM F18101’s total skull length, whereas the endocast of Euoplocephalus represents approximately 25% of skull length. Also, the endocast of QM F18101 has some unusual attributes that probably represent incomplete ossification of the bones walling in the endocranial cavity, whereas those of other ankylosaurians are extremely well ossified (Vickaryous, Maryańska & Weishampel, 2004; Witmer & Ridgely, 2008; Miyashita et al., 2011; Ősi, Pereda Suberbiola & Földes, 2014).

The paired olfactory bulbs in QM F18101 are modestly developed and apparently contact each other at the midline. This latter point represents the primitive condition (as characterized by Stegosaurus; Figs. 10F and 10G; see also Galton, 2001) and contrasts with that of other ankylosaurians in which the olfactory bulbs are strongly divergent from the midline, as in Euoplocephalus (Figs. 10D and 10E; see also Hopson, 1979; Miyashita et al., 2011), Panoplosaurus (Witmer & Ridgely, 2008), and, to a lesser extent Hungarosaurus (Ősi, Pereda Suberbiola & Földes, 2014). As in other ankylosaurians just noted (see also Pereda Suberbiola & Galton (1994) on Struthiosaurus), the olfactory tract is relatively short. The cerebral hemispheres are clearly indicated on the endocast as lateral swellings in the rostral portion of the endocast, but distinct optic lobes and cerebellum are not grossly observable, which is typical of other ankylosaurians (Hopson, 1979; Witmer & Ridgely, 2008; Miyashita et al., 2011; Ősi, Pereda Suberbiola & Földes, 2014). The endocast of QM F18101 does not have a floccular lobe of the cerebellum, which is the condition in most ankylosaurians, although two specimens of Euoplocephalus display a flocculus (Fig. 10D) (Hopson, 1979; Miyashita et al., 2011). The endocast shows a median dural peak that is in the approximate position of the pineal gland (epiphysis cerebri). The pineal has been identified in endocasts of Euoplocephalus (Coombs, 1978b; Miyashita et al., 2011), and so its presence in QM F18101 is perhaps not unexpected, but it is worth pointing out that a diversity of dinosaurs have a variety of median dural peaks, not all of which are likely to be epiphyseal in origin (Witmer et al., 2008; Witmer & Ridgely, 2009). As noted above, the pituitary fossa within the basisphenoid preserves the general disposition of the gland. Unlike any other ankylosaurian endocasts, in which the pituitary projects more or less directly vertically below the cerebral region, in QM F18101 the pituitary angles strongly caudally from the infundibular region such that it underlies the brainstem. The outgroup condition (e.g., Stegosaurus; Fig. 10F) is to have a more vertical pituitary, and thus QM F18101 would have a derived condition. Another derived condition of the pituitary is that the cerebral carotid arteries enter the pituitary fossa not at its distal terminus, as in Stegosaurus and other ankylosaurians, but rather enter at about one-third the length of the pituitary from the distal terminus.

Many of the cranial nerves and vascular structures can be identified in the CT data and are included in the digital endocast (Fig. 10). The optic nerves (CN II) apparently emerge together near the midline rather than being widely separated from each and directed more or less laterally, as in other ankylosaurians (Hopson, 1979; Miyashita et al., 2011; unpublished endocast of Edmontonia, AMNH 5831, done at Ohio University). QM F18101 displays the primitive condition in this regard, resembling Stegosaurus and most other dinosaurs (Hopson, 1979). In more advanced ankylosaurians such as Euoplocephalus, Edmontonia, and Panoplosaurus (L Witmer, pers. obs., 2015), the orbits and eyeballs are more caudally positioned within the skull, resulting in the lateral course of not just the optic nerves, but also the nerves supplying the extraocular muscles (CN III, IV, VI). The canals for these nerves in QM F18101 are not as discrete as in other ankylosaurians but what is preserved is consistent with QM F18101 retaining a more plesiomorphic condition, suggesting that, as with the optic nerves, the extraocular muscle nerves were not directed strongly laterally. As noted above, the side wall of the braincase is not as ossified as in other ankylosaurians (Vickaryous, Maryańska & Weishampel, 2004), and there are large gaps on either side of the laterosphenoid (Fig. 5E). Typically, the oculomotor (CN III) and trochlear (CN IV) nerves pass through foramina in the suture located between the laterosphenoid and orbitosphenoid, but in QM F18101 the corresponding region is a large open space without discrete canals. Cranial nerves III and IV presumably travelled through the ventral portion of the aperture, whereas the dorsal portion of the aperture probably transmitted the orbitocerebral vein, which is found in ankylosaurians, as well as other dinosaur clades (Witmer et al., 2008; Miyashita et al., 2011). The abducens nerve (CN VI) reaches the orbit by traversing a rough canal arising from the region of the brainstem, which is typical. It may merge with the laterosphenoid-orbitosphenoid aperture, but the CT scan data are not entirely clear on this point.

The trigeminal nerve (CN V) in archosaurians emerges through a foramen between the laterosphenoid and prootic (Romer, 1956) but in QM F18101, again, there is a large unossified gap rather than a discrete foramen (N.B. this is not an artefact of preservation). As a result, no distinctions can be made between the three branches of the trigeminal nerve. On the left side, a small canal traverses the braincase wall to pass from the endocranium to reach the laterosphenoid-prootic aperture; this canal almost certainly transmitted the transversotrigeminal vein (= rostral middle cerebral vein), in that in many archosaurians this vein opens into or just above the trigeminal foramen (Sampson & Witmer, 2007; Witmer et al., 2008; Witmer & Ridgely, 2009). The facial nerve (CN VII) is visible on both sides of QM F18101 passing through the prootic just rostral to the labyrinth of the inner ear. The vestibulocochlear nerve (CN VIII) is visible on the right side of the endocast, before disappearing into the inner ear. Better preserved on the left side, the vagal (= metotic, jugular) canal is present, and presumably transmitted the glossopharyngeal (CN IX), vagus (CN X), and accessory (CN XI) nerves, much as in other ankylosaurians and indeed other sauropsids (Romer, 1956; Sampson & Witmer, 2007). The left side of the endocast displays three canals, all of which we interpret to have transmitted branches of the hypoglossal nerve (CN XII). The number of hypoglossal foramina is variable in dinosaurs, but some specimens of Euoplocephalus and Amtosaurus have three (Averianov, 2002; Miyashita et al., 2011). Finally, the endocast of QM F18101 shows the course of the dorsal head vein, passing through an aperture between the prootic, laterosphenoid, and parietal to reach the adductor chamber in the temporal region. The dorsal head vein was present in Euoplocephalus (Figs. 10D and 10E; N.B. unlabelled in Fig. 7 of Miyashita et al., 2011) and, as noted above, a wide diversity of other dinosaurs.

Endosseous labyrinth of the inner ear

The endosseous labyrinth of the inner ear of QM F18101 (Figs. 10A–10C and 11A–11B) is proportionally enormous relative to both the size of the skull and the relative sizes of the labyrinths in other ankylosaurians (Witmer & Ridgely, 2008; Miyashita et al., 2011) (Figs. 10D–10E and 11C–11D). Despite the skull of QM F18101 (approx. 25 cm, rostrocaudally) being smaller than that of Euoplocephalus (AMNH 5337, 5405 and NHMUK R4947 approx. 40–50 cm, rostrocaudally) or Panoplosaurus (NMC 2759, ROM 1215 approx. 35 cm, rostrocaudally), the inner ear is much larger in absolute size (Fig. 11, see also Supplemental Information 1). Moreover, the inner ear is highly divergent in morphology, as well. For example, rather than being separated by bone from the endocranial cavity (as in basically all other archosaurians groups; L Witmer, pers. obs., 2015), in QM F18101 the prootic and otoccipital bones fail to enclose the inner ear medially such that the vestibular portion of the inner ear is broadly open to the endocranial cavity (Fig. 5E). Likewise, the ventral limit of the inner ear, which in other archosaurians would correspond to the cochlear (or lagenar) region, is also incomplete or poorly ossified, such that it, too, is largely open (Fig. 5F). These attributes are previously unreported in dinosaurs (if not all of Sauropsida). The morphology is bilaterally symmetrical and shows no overt signs of pathology. The functional consequences of these modifications are obscure. In dorsal aspect, the semicircular canals appear to form a triangle, with the rostral and caudal semicircular canals being angled at approximately 45° to the sagittal plane, rostrolaterally and caudolaterally, respectively, and together form slightly more than a 90° angle. The lateral semicircular canal runs parallel to the median plane and forms acute angles with each of the vertical canals (Fig. 11B). The crus communis is extremely short, probably due to the large size of the vestibular region, which also contributes to the semicircular canals all seeming very short. The vestibule is spherical, which is unlike any known dinosaur. This condition is more similar to that in turtles and Sphenodon (Walsh et al., 2009), although even in these clades the inner ear is more discretely enclosed by bone. The cochlear region of the inner ear is hard to interpret given that its ventral terminus is so poorly defined. The fenestra vestibuli (= ovalis) of QM F18101 is large, relatively round, and laterally orientated. No columella (stapes) is preserved, and, combined with the curious structure of the inner ear, the auditory apparatus thus remains obscure.

Taxonomic status of QM F18101

Based on the clear number of osteological differences, and thus diagnostic characters, between the cranium of QM F18101 and other ankylosaurians (see ‘Diagnosis’ and discussed above) we consider the specimen known as Minmi sp. (QM F18101) sufficiently distinct from other ankylosaurian taxa to warrant referral to a new genus and species: Kunbarrasaurus ieversi gen. et sp. nov. Furthermore preliminary investigations into the postcranial elements of QM F18101 reveal that a number of significant differences exist between it and the holotype of Minmi paravertebra (QM F10329), but this is part of an ongoing study that will be presented elsewhere. Accompanying the description of the postcranium of K. ieversi will be an in-depth analysis into its phylogenetic position.

Of note, Minmi paravertebra (QM F10329) has recently being assigned to nomen dubium by Arbour & Currie (in press). As Minmi paravertebra comprises postcranial material only, an in-depth comparison and reappraisal of the specimen will be conducted in conjunction with the redescription of the postcranium of QM F1810; something that is beyond the scope of this study. Pending the forthcoming publication, we still consider Minmi paravertebra sufficiently different to other ankylosaurians, particularly within Australia and other Gondwanan landmasses, to represent a distinct genus. Consequently Kunbarrasaurus represents the second named genus of ankylosaurian from Australia.