The taxonomy of a new parvicursorine alvarezsauroid specimen IVPP V20341 (Dinosauria: Theropoda) from the Upper Cretaceous Wulansuhai Formation of Bayan Mandahu, Inner Mongolia, China

Systematic Palaeontology

Locality and horizon

IVPP V20341 was found at Bayan Mandahu, Inner Mongolia, China, a locality of the Campanian-aged Wulansuhai Formation (Eberth, 1993). On July 2nd 2013, a team member (JBS) discovered isolated alvarezsauroid bones weathering out of a clifftop exposure of a fine-grained, red structureless aeolian quartz arenite, ∼3 km SE of ‘The Gate’ locality (41°43′15.3″N, 106°44′43.3″E; Fig. 1), close to the location of Eberth’s (1993) ‘7/12/90/2’ stratigraphic section, but not as far North as his ‘7/12/90/1’ section (Eberth, 1993). The locality lies within an larger area that Jerzykiewicz et al. (1993) called the ‘South Escarpment’ locality, but the specific locality of IVPP V20341 is referred to as the ‘Eagle’s Nest’ because a large predatory bird nest was found ∼3 m from the find. On July 3rd 2013, another team member (MP) further explored the sublocality and recovered several additional bones within 1 m of the original material. Following the depositional environments and facies transitions identified by Eberth (1993) at Bayan Mandahu, the specimen was deposited in the zone 1 palaeographic zone. The latter consists of alluvial, lacustrine and aeolian sediments deposited in a distal alluvial fan or braid-plain environment adjacent to an aeolian dune field. Given the sandy depositional environment of the specimen, IVPP V20341 was nicknamed ‘Xiaoshalong’, which is Mandarin Chinese for ‘little sand dragon’.

Description and comparison

IVPP V20341 appears to belong to an individual at or near skeletal maturity (e.g., subadult or adult) because the neurocentral sutures on the vertebrae preserved appear to be completely closed (Brochu, 1996; Irmis, 2007). However, this inference should be treated as a tentative one in the absence of other ontogenetically-informative fusion in the appendicular skeleton as well as relevant histological data. The skeletal elements that are common to both IVPP V20341 and Linhenykus are generally smaller in the former than in the latter. This suggests that IVPP V20341 was probably lighter in weight than Linhenykus, which itself weighed around 450 g (Xu et al., 2010b; Xu et al., 2013b). IVPP V20341 lacks a femur and skull so a more accurate proxy-based body mass estimate was not possible (Christiansen & Fariña, 2004; Therrien & Henderson, 2007). The incomplete vertebral column and the missing skull also prohibited a meaningful measurement of body length. However, given the relative size of IVPP V20341 to Linhenykus, the former was probably lighter than other Asian parvicursorines, with the exception of Shuvuuia and Parvicursor (see Table S2).

Axial skeleton

An articulated series of 4 partial anterior cervical vertebrae (A–D) (Fig. 2A), an isolated anterior portion of an anterior cervical vertebra (Fig. 2B) and 4 isolated fragmentary anterior caudal vertebrae (A–D) represent the axial skeleton (Fig. 3).

Caudal vertebrae

IVPP V20341 includes four isolated anterior caudal vertebrae: one that is well-preserved and three others that are poorly preserved (Fig. 3). These are referred to as caudals A–D in order of their relative anteroposterior position along the tail, as determined using the anteroposterior position of the neural arch pedicle and transverse processes and the relative development of the furrows and ridges along the ventral surfaces of the centra. The dimensions of the caudal vertebrae given in Table 1 do not unequivocally support the proposed ordering, nor any other ones. This probably reflects the large amount of missing data, particularly in caudals C and D (Table 1), and variability in vertebral geometry changes along the tail, as has been measured in a wide range of theropods (Pittman et al., 2013).

Caudals A–C appear to be anterior ones because the neural arch pedicle is anteriorly placed along the anteroposterior length of the centrum, as evident in other parvicursorine alvarezsauroids including Alvarezsaurus (MUCPv 54; Chiappe, Norell & Clark, 2002; Bonaparte, 1991), Achillesaurus (MACN-PV-RN 1116; Martinelli & Vera, 2007), Linhenykus (IVPP V17608; Xu et al., 2010b; Xu et al., 2013b), Parvicursor (PIN 4487/25; Karhu & Rautian, 1996), Shuvuuia (MPC 100/975; Chiappe, Norell & Clark, 2002) and Xixianykus (XMDFEC V0011; Xu et al., 2010c).

In lateral view the posterior margin of the narrowest portion of caudal A’s neural arch (the neck) is approximately two-thirds along the anteroposterior length of the centrum, whereas in the first free caudal of Parvicursor (PIN 4487/25; Karhu & Rautian, 1996) and the proximal caudals of Alvarezsaurus (MUCPv 54; Chiappe, Norell & Clark, 2002; Bonaparte, 1991) this is less than halfway along the same length and in the first caudal of Xixianykus (XMDFEC V0011; Xu et al., 2010c) this is about halfway along the same length. The condition in IVPP V20341 is comparable to the supposed first caudal of Patagonykus (MCF-PVPH 37; Novas, 1997) and the middle and distal caudals of Alvarezsaurus (Chiappe, Norell & Clark, 2002; Bonaparte, 1991). This suggests that the position of the posterior margin of the neural arch neck along the anteroposterior length of the caudal centrum of IVPP V20341 is unhelpful in determining the relative position of caudals A–C along the tail.

Caudals A to D show a posterior migration of the transverse processes (only the posterior ridge of the process is visible on the left lateral side of caudal D), but these are all still situated anteriorly on the centrum, which identifies IVPP V20341 as a parvicursorine alvarezsauroid (Xu et al., 2013b). The transverse processes form a broad subhorizontal ridge that originates from the anterodorsal corner of the centra (caudals A and B) rather than the posterior end of the prezygapophyses as in Linhenykus (best examples in caudals 3, 5, 7 and 8 of IVPP V17608; Xu et al., 2013b) and Shuvuuia (anterior caudals (caudals 3–8?) of MPC 100/975; Chiappe, Norell & Clark, 2002). It may be possible that caudals A and B of IVPP V20341 are caudals 1 and 2 of Linhenykus, but this is unlikely given that the first caudal of the latter has a sharp ventral keel (Xu et al., 2013b), a feature that neither caudals A or B possess. In both IVPP V20341 and Linhenykus the broad subhorizontal ridge does deflect posteriorly towards the dorsal edge of the posterior articular face (Xu et al., 2013b), although this appears to deflect more ventrally in the former specimen. However, the latter probably relates to subtle differences in caudal position between these specimens.

The caudals become anteroposteriorly shorter from caudal A to B (caudals C and D are not anteroposteriorly complete), but this pattern is not emphasised here because centrum length does not decrease continuously along the tail of Linhenykus—its middle caudals are actually anteroposteriorly longer than its anteriormost ones (IVPP V17608; Xu et al., 2013b). Caudals A and B possess a longitudinal furrow along the centrum’s entire ventral surface and this is bordered laterally by two ventral keels. This feature is also observed in Linhenykus (Xu et al., 2013b), Parvicursor (PIN 4487/25; Karhu & Rautian, 1996; Chiappe, Norell & Clark, 2002), Patagonykus (supposed 20th caudal of MCF-PVPH 37; Novas, 1997) and Shuvuuia (MPC 100/975; Chiappe, Norell & Clark, 2002). However, this feature is less developed in caudal B and is barely visible in caudal C, where the ridges are low and the furrow is broad and shallow. The ventral surface of caudal D has been eroded down to the cortical bone. The ventral surfaces of caudals A, B and C therefore support their proposed positional ordering: caudal A is the anteriormost one whilst caudal C is the posteriormost one.

Caudals A–D are laterally pinched. Caudals A and B are procoelous, but this cannot be confirmed in caudals C and D owing to a missing posterior articular face in the former and a missing anterior articular face in the latter. In caudals A and B, the concave anterior articular face is deep, whilst the posterior condyle is well-developed and hemi-spherical in shape. Procoely in caudal vertebrae is also observed in Haplocheirus (IVPP V15988; Choiniere et al., 2010), Shuvuuia (MPC 100/975; Chiappe, Norell & Clark, 2002), Mononykus (MPC N107/6; Perle et al., 1994), Xixianykus (XMDFEC V0011; Xu et al., 2010c), Alvarezsaurus (MUCPv 54; Chiappe, Norell & Clark, 2002; Bonaparte, 1991) and potentially in Patagonykus (MCF-PVPH 37; only the posterior articular surfaces are preserved (Novas, 1997)). However, the first caudal of Linhenykus is amphiplatyan (Xu et al., 2013b) whilst a proximal caudal of Achillesaurus—tentatively assigned as the fourth in the series—is amphicoelous (biconcave) (Martinelli & Vera, 2007).

Caudals A–D lack the sharp ventral keel that has been associated with the anteriormost caudals of many parvicursorines, so they probably represent caudals located further along the tail—at least posterior to the ventrally-keeled first caudal, as in Linhenykus and Xixianykus (Xu et al., 2013b; Xu et al., 2010c). Shuvuuia and Achillesaurus have a sharp ventral keel on their first two caudals (the first two caudals of Shuvuuia as identified by Chiappe, Norell & Clark (2002); a keel is not observed on the first caudal of Achillesaurus owing to specimen damage, but this feature is inferred because of the keels present on the last sacral and second caudal (Martinelli & Vera, 2007)) and this feature is also present on an anterior caudal of Alvarezsaurus (Bonaparte, 1991), Mononykus (Perle et al., 1994) and, supposedly, of Parvicursor as well (Karhu & Rautian, 1996). Patagonykus has a seemingly unique ventral surface amongst parvicursorines as the assumed first caudal has a ventral surface that is transversely narrow and slightly flat (Novas, 1996).

No evidence of chevron articulation facets was found on caudals A, B and D—the only ones that preserve the posterior ventral surface—as in caudals 1–3 of Linhenykus (IVPP V17608). However, these facets are well-developed on the posteroventral surface of a distal caudal—supposedly the 20th caudal—of P. puertai (MCF-PVPH 37; Novas, 1997). In Linhenykus (IVPP V17608) chevron articulation facets are weakly developed on the posteroventral surfaces of more posterior proximal caudals (caudals 4 and 5) and strongly developed on the anteroventral surface of a middle caudal (caudal 13) (Xu et al., 2013b). However, the absence of chevron articulation facets on caudals A, B and D of IVPP V20341 could become a distinguishing feature from Linhenykus if further finds can demonstrate that this morphology exists on or beyond caudal 4, in contrast to Linhenykus.

On the anterior portions of the right and left lateral surfaces of caudal centra A and B respectively, there is a very small weakly developed foramen, but this is absent on the opposing side of the centrum. However, the small size of these foramen and their differing positions suggests that they probably have little taxonomic significance. The lateral surfaces of caudals A–D lack both large, oval-shaped and small, subcircular fossa—unlike the first and second caudals of Patagonykus (MCF-PVPH 37; Novas, 1996) and Achillesaurus (MACN-PV-RN 1116; Martinelli & Vera, 2007) respectively—further supporting that caudals A–D are not the anteriormost ones. Foramen are absent from the caudals of Linhenykus but IVPP V20341 and Linhenykus (IVPP V17608) both have broad, shallow fossa on the lateral surfaces of their centra (e.g., caudal 5 and caudals A–D respectively).

The neural spine of caudal A is partially preserved and is missing its dorsoposterior portion. However, with what is present it is evident that the neural spine is rod-like, quite tall dorsoventrally, anteroposteriorly short and dorsoposteriorly directed. This suggests that caudal A is an anterior one as this type of neural spine morphology is found in the proximal caudals of Linhenykus (caudal two of IVPP V17608; Xu et al., 2013b), Parvicursor (the neural spine of the first caudal in PIN 4487/25 is dorsoventrally tall and dorsoposteriorly directed overall, but its rounded tip protrudes by a relatively small height beyond the dorsal margin of the postzygapophyseal facets and is dorsally directed (Karhu & Rautian, 1996)), Patagonykus (the first caudal of MCF-PVPH 37 is dorsoventrally tall and weakly dorsoposteriorly directed (Novas, 1997)) and Shuvuuia (MPC 100/975; Chiappe, Norell & Clark, 2002). In contrast, the proximal neural spines of Alvarezsaurus (MUCPv 54) are dorsally directed and have a subtriangular lateral profile (Chiappe, Norell & Clark, 2002). The anterior margin of the damaged neural spine on caudal A lies above the neural pedicle, whereas the whole neural spine is located posterior to the pedicle—at the level of the postzygapophyses —in the anterior caudals of Linhenykus (e.g., caudals 2–4 of IVPP V17608; Xu et al., 2013b). This difference is probably because of the differing caudal positions being compared, given that theropod neural spines generally extend further posteriorly as their inclination increases along the anterior portion of the tail. However, caudal A of IVPP V20341 does not appear to be an anteriormost caudal because of the absence of a sharp ventral keel and the presence of procoelous articular faces and fossae, as mentioned earlier in the caudal vertebrae section. This suggests that the anterior and posterior margins of caudal A’s neural spine are perhaps more damaged than they seem to be.

The shape of the neural canals in caudals A–C are laterally wider and more semi-circular compared to the laterally narrower and more oval-shaped ones of Linhenykus (observable in caudals 2, 7 and 13 of IVPP V17608; Xu et al., 2013b), Patagonykus (MCF-PVPH-37, the supposed first caudal (Novas, 1997)) and Parvicursor (PIN 4487/25, the supposed first caudal (Karhu & Rautian, 1996)). However, this feature cannot be confirmed as an autapomorphy of IVPP V20341 because it is plausible that laterally wider and more semi-circular neural canals might be present in as yet unknown portions of the vertebral columns of existing parvicursorines, since neural canal size and shape changes along the vertebral column of theropods (and other vertebrates). The ventral surface of the neural canal of caudal D bears a longitudinal ridge. However, the distribution of this characteristic amongst parvicursorines is unclear owing to a paucity of appropriate specimens.

Appendicular skeleton

The appendicular skeleton comprises of a potential left scapular shaft (Fig. 4) and a range of possible pedal phalanges (Fig. 5), including unknown digit II/III and IV phalanges, a right II-1, II-2 and IV, as well as a potential left III-2 and III-3. Linear measurements of the appendicular skeleton are given in Table 2.

Parvicursorines at Bayan Mandahu

The period of deposition represented at Bayan Mandahu is not known accurately, but some lithologies such as the structureless sandstones seem to have been rapidly deposited by sandstorm events, whilst others such as the carbonates (caliche) were probably deposited more slowly over thousands of years (Jerzykiewicz et al., 1993). This suggests that deposition probably happened over thousands of years at Bayan Mandahu, although this cannot be confirmed until radiometric dating can establish a chronology for the local depositional history. Given this estimate, the locations of IVPP V20341 and Linhenykus specimens (IVPP V17608, IVPP V18190) near the top and bottom of the local rock succession (Fig. 1) suggests that the deposition of both taxa was probably separated by a similar magnitude of time. This time interval perhaps makes it more likely that IVPP V20341 belongs to Linhenykus, but it is possible that there were two or more distinct genera in the locality that had separate and/or overlapping temporal ranges. IVPP V20341 and Linhenykus were preserved under broadly similar semi-arid conditions—the former is preserved in a red structureless sandstone layer whilst the latter is preserved in a more resistant nodule-rich red sandstone layer (Table S1). This indicates that IVPP V20341 and Linhenykus lived in a similar environment within or close to a dune field, according to Eberth’s (1993) depositional model for Bayan Mandahu (alluvial material washed off the nearby palaeo-Lang Shan mountain range was fringed by a dune environment). The persistence of environmental conditions potentially favours the longevity of an existing genus, but this likelihood cannot be used to justify IVPP V20341’s taxonomy. Thus, there are contextual arguments for the referral of IVPP V20341 to Linhenykus but these are seemingly weak ones. The study of the specimens themselves demonstrates that IVPP V20341 is distinguishable from Linhenykus monodactylus by a feature of the anterior caudal vertebrae: transverse processes originate from the dorsoanterior corner of the centra (anterior caudals A and B) rather than the posterior end of the prezygapophyses, as in Linhenykus (e.g., caudals 3, 5, 7 and 8 of IVPP V17608; Xu et al., 2013b). This is because the first caudal of Linhenykus is ventrally-keeled but neither caudals A or B are.

There are a number of suspected differences between the cervical and caudal vertebrae of IVPP V20341 and Linhenykus that have also been discussed. However, these must remain as such until the vertebral number of the relevant vertebrae can be clarified—particularly by further parvicursorine discoveries. Owing to the uncertainty in the identification of elements from the appendicular skeleton, their differences with Linhenykus are also excluded here.

IVPP V20341 compared to other Asian parvicursorines

∼350 km separates Bayan Mandahu and the closet Mongolian parvicursorine locality (Mononykus olecranus at Bayan Dzak (Andrews, 1932)). During the Upper Cretaceous the mountain ranges within the Gobi Basin (composed of Palaeozoic and Mesozoic rocks) were being subjected to extensional tectonism (Jerzykiewicz et al., 1993) that presumably promoted sediment deposition through the creation of accommodation space. This geological setting divided the Gobi Basin and created obstacles to faunal interaction, which probably promoted vicariance. The latter would help to explain why the Bayan Mandahu fauna seems to be distinct from Djadokhtan ones (Xu et al., 2015). If common Bayan Mandahu and Djadokhtan parvicursorines were found, this would suggest that at least some elements of the faunas are similar, which would advocate a complex scenario of selective isolation to explain the pattern of animals observed. Fortunately, for the skeletal elements that they share in common, IVPP V20341 lacks the autapomorphies of any Mongolian parvicursorine, so this material does not appear to contradict the distinctiveness of the Bayan Mandahu fauna (see Table 3). However, its status as a distinct taxon or specimen of Linhenykus remains unclear. Parvicursor and Ceratonykus are both known from the Upper Cretaceous (Lower Santonian, Alifanov & Barsbold, 2009; Middle Campanian, Karhu & Rautian, 1996) Barun Goyot Formation (Table S1) and appear to have shared their living environment. Niche partitioning by these taxa—if at all— probably relates to their body size differences as Parvicursor is smaller than Ceratonykus (75.6 mm long tibiotarsus in Parvicursor, PIN 4487/25 (Karhu & Rautian, 1996); 89 mm long right and left tibiotarsus in Ceratonykus, MPC 100/124 (Alifanov & Barsbold, 2009)). If IVPP V20341 is shown to be a valid taxon at a later date, the sharing of a relatively harsh semi-arid environment with Linhenykus might support niche partitioning too e.g., if IVPP V20341 has a more conventional parvicursorine hand morphology than Linhenykus (Xu et al., 2013b).

IVPP V20341 compared to other alvarezsauroids

IVPP V20341 is seemingly distinct amongst alvarezsauroids because of possible cervical procoely and the presence of caudal neural canals (caudals A–C) that are laterally wider and more semi-circular compared to the laterally narrower and more oval-shaped ones of Linhenykus (caudals 2, 7 and 13) and of the supposed first caudals of Patagonykus (MCF-PVPH 37; Novas, 1997) and Parvicursor (Karhu & Rautian, 1996). However, these potential autapomorphies have caveats that need to be considered. Cervical procoely is unknown in alvarezsauroids, but only one complete neck specimen is known and this belongs to the basalmost taxon Haplocheirus sollers (Choiniere et al., 2010). If the eleven cervical vertebrae of the latter is similar amongst all alvarezsauroids—an assumption that is speculative based on current fossil evidence—then the 8 and 9 cervical vertebrae preserved in Mononykus (MPC 107/6; Perle et al., 1994) and Shuvuuia (MPC 100/975; Chiappe, Norell & Clark, 2002) respectively may actual represent near complete series. These three neck specimens provide an indication of the neck morphology in basal alvarezsauroids and derived parvicursorines—the latter of which IVPP V20341 should most closely resemble—so the absence of procoely in all three specimens is significant. However, Haplocheirus lacks the strong opisthocoelous condition of Mononykus (MPC 107/6; Perle et al., 1994) and Shuvuuia (MPC 100/975; Chiappe, Norell & Clark, 2002), which shows that there is significant variation in articular surface morphology within the clade, although it is currently impossible to say if such variation might include multiple taxa with a procoelous condition. Alvarezsauroid tail morphology is better understood than neck morphology because the former is known from more fossil material belonging to a broader range of taxa. The most complete caudal series are found in Haplocheirus (IVPP V15988, 15? caudals (Choiniere et al., 2010)), Alvarezsaurus (MUCPv 54, 13 caudals (Chiappe, Norell & Clark, 2002; Bonaparte, 1991)), Linhenykus (IVPP 17608, 13 caudals (Xu et al., 2013b)) and Shuvuuia (MPC 100/975, 19 caudals (Chiappe, Norell & Clark, 2002); MPC 100/120; 22? caudals (Suzuki et al., 2002)). The most complete alvarezsauroid tail is represented by specimen MPC 100/120 of Shuvuuia, which preserves direct evidence of approximately 22 caudals (Suzuki et al., 2002). However, the gaps in the caudal series suggest a caudal count upwards of 35 caudals (Suzuki et al., 2002). The relatively large semi-circular caudal neural canals of IVPP V20341 are absent in Linhenykus, Patagonykus and Parvicursor—the only taxa that had specimens where the shape of the caudal neural canal could be determined from firsthand study or from the literature. This represents a small sample size and given that the neural canal varies in size and shape along the vertebral column of theropods (and other vertebrates), this potential autapomorphy cannot be supported unequivocally. In the cases of both of these tentative autapomorphies of IVPP V20341, future fossil specimens are needed to test their validity.

Potentially informative features for alvarezsauroid phylogeny

New parvicursorine anatomical information obtained through the study of IVPP V20341 warrants editing of five Longrich & Currie (2009) phylogenetic characters (characters 3, 7–9 and 21) as well as the erection of a new character.

The possibly procoelous cervicals of IVPP V20341 are currently unique amongst alvarezsauroids which requires character 3 of Longrich & Currie (2009) to be edited. Cervical procoely is probably a derived alvarezsauroid condition since the majority of alvarezsauroids have opisthocoelous cervicals and the basal condition seems to be amphicoelous, amphiplatyan or platycoelous (amphi- to platycoelous in Haplocheirus (IVPP V15988; Choiniere et al., 2010) and amphiplatyan in Alvarezsaurus (MUCPv 54; Chiappe, Norell & Clark, 2002; Bonaparte, 1991)). However, given the unknown combinations of these vertebral types in alvarezsauroid necks and their changes through time, this character is not ordered here:

Cervical centra: amphicoelous, amphiplatyan or platycoelous (0), opisthocoelous (1), procoelous (2), amphicoelous, amphiplatyan or platycoelous AND opisthocoelous or procoelous (3), amphicoelous, amphiplatyan or platycoelous AND opisthocoelous AND procoelous (4) (after Perle et al., 1994).

At a qualitative level, we observed noticeable changes in the ventral surface width of parvicursorine cervicals along their series. This suggests that further quantitative study is needed to maximize the phylogenetic utility of this feature and build upon the ordered character 7 of Longrich & Currie (2009).

As mentioned, the lateral sides of the cervicals of IVPP V20341 are less depressed than those of Linhenykus (IVPP 17608; Xu et al., 2013b), and this depression is limited to the ventrolateral portion of the centra. However, the position of these cervicals may not be analogous. To accommodate this difference and the expected variability in the degree of lateral surface depression along the neck of parvicursorines—pending more in-depth quantitative studies—character 8 of Longrich & Currie (2009) was edited:

Lateral surfaces of cervical centra: convex or flat (0), strongly to mildly depressed across part of or the entire surface (1), convex or flat AND strongly to mildly depressed across part of or the entire surface (2).

Comparisons made between the preserved cervical and caudal vertebrae of IVPP V20341 and all other parvicursorines has highlighted variation in the ventral surface along each series, including the relative development of furrows (partly or fully) and keels (absent, small or large in caudals) as well as the distribution of rounded and/or flat smooth ventral surfaces. To reflect these observations, character 9 of Longrich & Currie (2009) has been reworded:

Ventral surfaces of cervical centra: smooth and flat and/or smooth and rounded (0), longitudinal furrow partly or fully spanning the length of the centrum (1), both conditions are present (2) (after Novas, 1996).

Character state 2 is added because current evidence cannot rule out the possibility that states 0 and 1 are present in the same cervical series. However, the character remains unordered in the absence of evidence regarding how this trait evolved across Alvarezsauroidea.

In consideration of the amphiplatyan first caudal of Linhenykus (IVPP 17608; Xu et al., 2013b) and the amphicoelous (biconcave) proximal (4?) caudal of Achillesaurus (MACN-PV-RN 1116; Martinelli & Vera, 2007), character 21 of Longrich & Currie (2009) is expanded to:

Caudal vertebrae: amphiplatyan or amphicoelous (0), or procoelous (1) (after Novas, 1996).

To utilise the potential of the caudal ventral keel towards reconstructing alvarezsauroid phylogeny whilst considering their poorly known extent along the tail, a new character limited to the first caudal is proposed:

Ventral surface of the first caudal vertebrae: not transversely narrow (0), ‘pseudo-keel’ present—the ventral surface is transversely narrow and slightly flat (1), sharp keel present (2).

Bayan Mandahu as a distinct fauna within the Upper Cretaceous Gobi Basin

The Wulansuhai Formation rocks of Bayan Mandahu, Inner Mongolia comprise of lithologies that are similar to the Djadokhta Formation rocks of Bayan Dzak, Mongolia (Eberth, 1993; Jerzykiewicz et al., 1993). These lithologies indicate that both formations were deposited mostly under semi-arid conditions as alluvial and aeolian sediments, but the presence of some mudrocks shows that some deposition occurred under wetter climatic conditions (Eberth, 1993; Jerzykiewicz et al., 1993). Many Bayan Mandahu fossils have been referred to taxa known from the Djadokhta Formation (Jerzykiewicz et al., 1993; Currie & Peng, 1994; Dong & Currie, 1996), which both share a vertebrate fauna of dinosaurs, lizards, turtles, mammals and birds. The Wulansuhai Formation was assigned a Campanian age based on its lithological and faunal similarities with the Campanian-aged Djadokhta Formation (Jerzykiewicz et al., 1993), which itself was dated based on faunal and magnetostratigraphic data (see Xu et al. (2015) and references therein). The absolute age of the Wulansuhai Formation is still wanting so the stratigraphic correlation of these formations remains equivocal. However, an increasing body of evidence suggests that the two faunas represented in both formations are actually distinct (Makovicky, 2008): several previous referrals of Bayan Mandahu specimens to Djadokhta taxa have been rejected (Xu et al., 2012; Longrich, Currie & Dong, 2010) whilst several taxa unique to Bayan Mandahu have been described (Xu et al., 2010b; Godefroit et al., 2008; Longrich, Currie & Dong, 2010; Xu et al., 2010a; Xu et al., 2013a; Xu et al., 2013b). Unfortunately, the taxonomic status of IVPP V20341 does not contribute strong support towards the hypothesis that Bayan Mandahu is faunally distinct from the Djadokhta Formation, but it does not appear to contradict it.