The relationships of the Euparkeriidae and the rise of Archosauria

For the first time, a phylogenetic analysis including all putative euparkeriid taxa is conducted, using a large data matrix analysed with maximum parsimony and Bayesian analysis. Using parsimony, the putative euparkeriid Dorosuchus neoetus from Russia is the sister taxon to Archosauria + Phytosauria. Euparkeria capensis is placed one node further from the crown, and forms a euparkeriid clade with the Chinese taxa Halazhaisuchus qiaoensis and ‘Turfanosuchus shageduensis’ and the Polish taxon Osmolskina czatkowicensis. Using Bayesian methods, Osmolskina and Halazhaisuchus are sister taxa within Euparkeriidae, in turn sister to ‘Turfanosuchus shageduensis’ and then Euparkeria capensis. Dorosuchus is placed in a polytomy with Euparkeriidae and Archosauria + Phytosauria. Although conclusions remain tentative owing to low node support and incompleteness, a broad phylogenetic position close to the base of Archosauria is confirmed for all putative euparkeriids, and the ancestor of Archosauria +Phytosauria is optimized as similar to euparkeriids in its morphology. Ecomorphological characters and traits are optimized onto the maximum parsimony strict consensus phylogeny presented using squared change parsimony. This optimization indicates that the ancestral archosaur was probably similar in many respects to euparkeriids, being relatively small, terrestrial, carnivorous and showing relatively cursorial limb morphology; this Bauplan may have underlain the exceptional radiaton and success of crown Archosauria.


Placement of euparkeriids outside of crown Archosauria
The current analysis contrasts with the placement of Euparkeria capensis within the crown, as the sister taxon to Ornithosuchidae+Ornithodira (within "Ornithosuchia"), uniquely found by Gauthier (1986; Fig. 1A; see Previous phylogenetic work in main text) and with Broom (1913a,b) who suggested an affinity between Euparkeria capensis and ornithosuchids. The affinity proposed by Broom (1913a,b) was however based only on overall unspecified similarity in skull shape, and the encroachment of the distal end of the ventral ramus of the squamosal and the dorsal ramus of the quadratojugal into the lateral temporal fenestra (1913a). No apomorphic features of the skull were identified to unite Euparkeria capensis with Ornithosuchus longidens to the exclusion of other crown taxa however. Furthmore, the degree of encroachment by the squamosal and quadratojugal differs substantially between the taxa, with the encroachment in Ornithosuchus longidens far more substantial, and approaching the condition in other pseudosuchians (e.g. gracilisuchids - Butler et al. 2014) where the squamosal contacts the postorbital bar. The characters used by Gauthier (1986) to place Euparkeria capensis within Archosauria were: (1) squamosal reduced and descending ramus gracile; (2) centra steeply inclined at least in first four postatlantal cervicals; (3) modifications of hindlimb and girdle correlated with semierect gait; (4) ventral flange of astragalus absent; (5) crocodile-reversed ankle joint (peg on calcaneam, socket on astragalus); (6) pedal digit V with less than four phalanges.
The "reduced" and "gracile" nature of the squamosal in Euparkeria reflects the overall gracile morphology of the skull and its size (with a thin ventral ramus of the squamosal similarly found in Coelophysis bauri -Colbert 1989) contrasting with the wide ramus in the larger Herrerasaurus ischigualastenis (Sereno and Novas 1994; Fig. 8H), and does not represent a particular novel feature of the element, meaning that its phylogenetic usefulness is highly questionable. Regarding the "steeply inclined" nature of the centra of the first four postatlantal cervical vertebrae, delimitation and characterization of the variation referred to was found to be difficult, with no clear demarcation between the state seen in Euparkeria capensis and that in, for example, Erythrosuchus africanus (Gower 2003).
Modifications correlated with a semierect gait is, as made clear by Sereno and Arcucci (1990), an extremely loosely defined character. Even if it is broken down into several more specific characters, such modifications are also found in all pseudosuchian taxa, and the only taxa in or immediately outside the crown to lack these modifications are the aquatic phytosaurs; on the stem, proterochampsids (e.g. Chanaresuchus bonapartei -Romer 1972b) also show the derived state. These modifications are thus certainly not a feature which convincingly unites Euparkeria capensis with ornithodires to the exclusion of pseudosuchians. The ankle characters used by Gauthier (1986) to unite Euparkeria capensis with "ornithosuchians" were thoroughly discussed by Sereno and Arcucci (1990). After inspection of the specimens and based on the description of Cruickshank (1979), these authors came to the conclusion, which is followed here, that Euparkeria capensis (like Osmolskina czatkowicensis) does not in fact show an ornithosuchian-like "crocodile reversed" ankle joint, nor indeed any other features of the ankle which unite it with either major archosaur lineage (see Fig. 11A).
Whilst Euparkeria capensis does (as outlined in Sereno and Arcucci 1990), like ornithosuchids, possess three phalanges on digit V, this character appears to be very labile, with proterosuchids (Cruickshank 1972) and erythrosuchids (Young 1964) showing three phalanges, proterochampsids possessing a single phalanx (Nesbitt 2011), and some pseudosuchians showing fewer than four (e.g. Neoaetosauroides shows two - Bonaparte 1971). Furthermore, on inspection of the material (SAM-PK-8309), the condition in Euparkeria capensis is not absolutely clear due to damage.

Phylogenetic placement of Vancleavea campi
The This placement in the current analysis is however only supported by two unambiguous synapomorphies: (1) presence of irregular pit and ridge ornamentation on dermal skull bones; (2) absence of lateral mandibular fenestra. Synapomorphy (1) however probably relates to the aquatic habitus of the taxa, as it is also shared by phytosaurs, and node support is low (Fig. 2). Support for this node in the analysis of Ezcurra et al. (2010) rests on the following unambiguous synapomorphies: (1) absence of fourth trochanter; (2) presence of mid-dorsal neural spines situated at mid-length between the zygapophyses; (3) maximum length of iliac blade less than three times maximum height. Whether Vancleavea campi possesses a true fourth trochanter is however controversial (see Nesbitt 2009), and this taxon was thus scored "?" for this taxon in the current analysis. Doswelliids are scored as showing a fourth trochanter in the current analysis as, although not pronounced, a muscle attachment ridge is present in the same location (i.e. at midshaft) and has been considered to be likely to be homologous (Schoch and Sues 2013). The character yielding synapomorphy (2) was excluded from the current analysis as it was deemed to be difficult to score, and that yielding (3) was excluded as it was felt to overlap with character 276 of Nesbitt (2011), but these characters would further support the results presented here. Synapomorphy (2) of Ezcurra et al. 2010 was the same sole synapomorphy found to support this clade in the analysis of Dilkes and Arcucci (2012).
The node in Schoch and Sues (2013) uniting Vancleavea campi with dosweliids was supported by four unambiguous synapomorphies: (1) loss of antorbital fossa; (2) regain of prominent pubic tubercle; (3) maximum length of iliac blade less than three times maximum height; (4) midshaft diameter of metatarsal IV less than that of metatarsal III. The character yielding synapomorphy (3) is discussed above. That yielding (1) was scored as inapplicable in the current analysis as Vancleavea campi lacks an antorbital fenestra. That yielding (2) was excluded from this analysis due to potential overlap with presence/absence of the pectineal process (character 289) and difficulty in scoring taxa from the literature; furthermore, the erythrosuchid Garjainia prima (PIN 951/25) shows a well-developed pubic tubercle, meaning that conclusions drawn regarding the phylogenetic position of erythrosuchids in respect to Vancleavea campi based on this character may be doubtful. The character yielding (4) (2008) is loss of postaxial intercentra (see discussion above).
The relative position of Vancleavea campi in respect to Erythrosuchus africanus (the next taxon crownward) found by Desojo et al. (2011) is supported by the following unambiguous synapomorphies: (1) acquisition of antorbital fossa; (2) mid-cervical/mid-dorsal centrum length ratio ≤1; (3) presence of preacetabular process; (4) reduction of prominent pubic tubercle to rugosity; (5) caudoventral process of ischium large (making ischium longer than ilium); (6) presence of fourth trochanter; (7) lateral fossa on dorsal centra below neurocentral suture; (8) anterior process of jugal broad and dorsally expanded anteriorly. The characters yielding (1), (4) and (6) are discussed above. That yielding (2) was excluded from the current analysis as it was deemed difficult to score, somewhat arbitrary in its delimitation, and to be very labile with both primitive and derived states widespread within and outside the crown.
That yielding (5) was excluded as it overlapped with character 298 of Nesbitt (2011; ischium length); character 298 could not, following Nesbitt (2011), be scored for Vancleavea campi because no complete ilium and ischium are preserved from the same individual. Regarding (7), presence or absence of a lateral fossa on the dorsal centra was considered to be problematic to score, as several taxa scored as "absent" actually show a rudimentary fossa (e.g. Mesosuchus browni -SAM-PK-6046), and the feature varies greatly along the dorsal column (Mesosuchus browni -SAM-PK-6046; Euparkeria capensis -SAM-PK-6047B). Additionally, the character overlaps with presence/absence of a prezygodiapophyseal lamina, with this character deemed to better represent the variation seen. Regarding (8), slender versus broad anterior process of the jugal was deemed to be too difficult to delimit satisfactorily, and was thus not included in the current analysis. Figure S1. Strict consensus tree of most parsimonious trees found in analysis including character 93. Numbers at nodes are standard (before slash) and GC (after slash) bootstrap values.