Postcranial anatomy of Pissarrachampsa sera (Crocodyliformes, Baurusuchidae) from the Late Cretaceous of Brazil: insights on lifestyle and phylogenetic significance

The postcranial anatomy of Crocodyliformes has historically been neglected, as most descriptions are based solely on skulls. Yet, the significance of the postcranium in crocodyliforms evolution is reflected in the great lifestyle diversity exhibited by the group, with members ranging from terrestrial animals to semi-aquatic and fully marine forms. Recently, studies have emphasized the importance of the postcranium. Following this trend, here we present a detailed description of the postcranial elements of Pissarrachampsa sera (Mesoeucrocodylia, Baurusuchidae), from the Adamantina Formation (Bauru Group, Late Cretaceous of Brazil). The preserved elements include dorsal vertebrae, partial forelimb, pelvic girdle, and hindlimbs. Comparisons with the postcranial anatomy of baurusuchids and other crocodyliforms, together with body-size and mass estimates, lead to a better understanding of the paleobiology of Pissarrachampsa sera, including its terrestrial lifestyle and its role as a top predator. Furthermore, the complete absence of osteoderms in P. sera, a condition previously known only in marine crocodyliforms, suggests osteoderms very likely played a minor role in locomotion of baurusuchids, unlike other groups of terrestrial crocodyliforms. Finally, a phylogenetic analysis including the newly recognized postcranial features was carried out, and exploratory analyses were performed to investigate the influence of both cranial and postcranial characters in the phylogeny of Crocodyliformes. Our results suggest that crocodyliform relationships are mainly determined by cranial characters. However, this seems to be a consequence of the great number of missing entries in the data set with only postcranial characters and not of the lack of potential (or synapomorphies) for this kind of data to reflect the evolutionary history of Crocodyliformes.


INTRODUCTION
2010. The postcranial bones are referred to Pissarrachampsa sera, primarily due to the presence of features compatible with the postcranial morphology of other baurusuchids, but also because the relatively restricted locality ''Inhaúmas-Arantes Farm'' provided, so far, exclusively materials referred to P. sera. The material assigned to the holotype was not collected at the same time as the skull (Montefeltro, Larsson & Langer, 2011). However, this association is assumed as the postcranial elements were spacially identified during the first expedition placed only few centimeters from the back of the skull, in its natural anatomical position in the same stratigraphic level in the outcrop. Also, it is unlikely that the specimen assignations employed here is wrong, due to discrepant sizes, anatomical overlaps and different locations in the quarry.

Description
The postcranial remains of Pissarrachampsa sera were compared within the context of Crocodyliformes although special attention was given to the morphology of other baurusuchids with postcranium. The comparisons focused on first-hand examination of specimens (Table 1), however, published resources were also used.

Dorsal vertebrae
Seven dorsal vertebrae are partially preserved in the holotype of Pissarrachampsa sera (LPRP/USP 0019), all of which exhibit the typical amphicoelous morphology seen in Notosuchia (Pol, 2005;Nascimento & Zaher, 2010). Five partial vertebrae are articulated in a series, one of which lacks part of the neural arch, , and are recognized as mid-to caudal-dorsal vertebrae, whereas the other two are isolated and very likely belong to a more cranial position in the vertebral series (Fig. 1D). One of the features used to determine the axial position of the preserved vertebrae was the relative position of the parapophysis and diapophysis. In notosuchians, as in Baurusuchus albertoi (Nascimento & Zaher, 2010), Sebecus icaeorhinus (Simpson, 1937), and Notosuchus terrestris (Woodward, 1896), the diapophysis is located more dorsally in cranial dorsal vertebrae, but migrate to a more ventral position caudally along the series (Pol, 2005;Nascimento & Zaher, 2010;Pol et al., 2012). On the other hand, the parapophysis is located ventrally in cranial-dorsal vertebrae, and migrate to a more dorsal position in more caudal elements, until it reaches the same dorsoventral level of the diapophysis (Pol, 2005;Nascimento & Zaher, 2010;Pol et al., 2012). The vertebrae in the articulated series show no evidence of para-and diapophyses migration, with both structures located at the same dorsoventral level at the distal portion of the transverse process. In addition, the preserved prezygapophyses are fused with the transverse processes. In closely related taxa, such as Baurusuchus albertoi and Notosuchus terrestris, this fusion is present in vertebrae caudal to the seventh dorsal element (Pol, 2005;Nascimento & Zaher, 2010), also suggesting that this sequence does not belong to cranial-dorsal vertebrae.
The vertebrae of Pissarrachampsa sera have an elliptical centrum in cranial view and are constricted at the middle, as typical for notosuchians (Pol, 2005). The centrum is slightly craniocaudally longer than high (measured from the ventral margin to the level of the ventral limit of the neural channel), and the dimensions are approximately the same in all  (2006) Baurusuchus albertoi MZSP-PV 140;Nascimento (2008) and Nascimento & Zaher (2010) Baurusuchus salgadoensis UFRJ DG 285-R; Vasconcellos & Carvalho (2010) Caiman sp. LPRP/USP N 0008; MZSP 2137; Brochu (1992) and Nascimento (2008) Chimaerasuchus paradoxus IVPP V8274; Wu & Sues (1996) Crocodylus sp. Brochu (1992) Edentosuchus tienshanensis Pol et al. (2004) Lomasuchus palpebrosus Leardi et al. (2015) Mahajangasuchus insignis FMNH 2721 (research cast of UA8654); Buckley & Brochu (1999) Mariliasuchus amarali UFRJ-DG-50-R, UFRJ-DG-105-R; Nobre & Carvalho (2013) Melanosuchus niger Brochu (1992) and Nascimento (2008) Microsuchus schilleri Leardi, Fiorelli & Gasparini (2015) Notosuchus terrestris MACN-PV RN 1037; MACN-PV RN 1044, MACN-PV N 109;MUCPv-137;Pol (2005) and Fiorelli & Calvo (2008) Orthosuchus stormbergii SAM-PK 409; Nash (1975) Sues & Brinkman (1996) Uberabasuchus terrificus CPPLIP 0630; Vasconcellos (2006) Uruguaysuchus aznarezi Pol et al. (2012) Yacarerani boliviensis Leardi et al. (2015) preserved centra (28 mm long, and 19 mm high). The preserved portion of the neural spine in the third vertebra of the sequence suggests that this structure projects cranially, as in caudal dorsal vertebrae of Baurusuchus albertoi. However, the neural spine of caudal-dorsal vertebrae of Baurusuchus bends caudally on its distal end (Nascimento & Zaher, 2010); a condition not accessible in P sera. The transverse processes are caudally oriented, and project horizontally in cranial and caudal views. The base of the prezygapophyseal process is located ventral to the upper margin of the neural canal, and projects dorsally and laterally. There is also a slight caudal projection and the prezygapophyses do not extend beyond the cranial limit of the vertebral centrum. The articulation area between the pre-and postzygapophyses is slightly oblique in relation to the Cross-hatched areas represent broken surfaces. Black areas represent sedimentfilled areas. Abbreviations: dpon, depression between the postzygapophysis and the neural spine; ns, neural spine (base); ncs, neurocentral suture; pf, postspinal fossa; poz, postzygapophysis; prz, prezygapophysis; tp, transverse process; vc, vertebral centrum. Scale bar equals 5 cm. horizontal plane of the vertebral column. The postzygapophyses, in the second and third vertebrae of the articulated series, are dorsally curved and project from the caudalmost part of the transverse processes. There is a deep fossa cranial to the postzygapophysis, at the intersection of the neural spine with the transverse process. Pol et al. (2012) suggest that this fossa is exclusively found in notosuchians. The cranial limit of this fossa is marked by a ridge, which extends laterally from the base of the neural spine to half of the lateral length of the transverse process.
One of the isolated vertebrae ( Fig. 1D) provides additional information on the vertebral morphology of Pissarrachmpsa sera. The dimensions of this vertebral centrum are approximately the same as for those of the articulated series. However, the neural arch is slightly craniocaudally longer. Also, its neural canal exhibits a rounded opening in cranial view. In caudal view, the postzygapophyses are connected by the postspinal fossa (Pol et al., 2012). The U-shaped ventral margin of this fossa forms a groove located ventral to the dorsal margin of the neural canal (Fig. 1D), a feature that is also observed in cervical and dorsal vertebrae of Baurusuchus albertoi (Nascimento, 2008;Nascimento & Zaher, 2010). This groove becomes progressively wider dorsally, until it merges with the zygapophyses. Also, in dorsal view, the cranialmost part of the fossa is lateromedially narrower than the area between the postzygapophyses.
The suture line between the neural arch and the vertebral centrum is clearly distinguishable in the best preserved isolated vertebra, and it is very likely that the neurocentral suture was also not completely closed in the dorsal vertebrae of the articulated series. Brochu (1996) proposed a cranial to caudal closure pattern of this suture for the crown-group Crocodylia, so that juveniles retain the suture opened in caudal presacral vertebrae. Irmis (2007) observed a similar pattern in phytosaurs and tentatively suggested it is typical of members of the Pseudosuchia lineage, but not of the Avemetatarsalia lineage. However, after analyzing dorsal vertebrae of Notosuchus terrestris, Pol (2005) commented that this pattern described in Brochu (1996) might not be valid for Crocodyliformes outside the Crocodylia clade, such as Pissarrachampsa sera. As the vertebrae described here belong to the holotype, which is likely a mature specimen based on comparisons to smaller specimens from the type locality, our results reinforce the inference of Pol (2005). Finally, Ikejiri (2012) showed that sutures of presacral vertebrae remain opened even in some very mature extant alligators, and Bailleul et al. (2016) have demonstrated that addressing the stage of maturity of archosaurian specimens based on the level of sutural closure in the skull can be misleading. In this context, vertebral sutural closure should not be used as the single factor when inferring the stage of maturity in crocodyliforms.

Forelimb
Ulna. The right ulna of the holotype of Pissarrachampsa sera is preserved (LPRP/USP 0019), as well as a smaller referred right ulna (LPRP/USP 0740) that corresponds to a juvenile individual. The holotypic ulna is damaged at both ends (Fig. 2). Its maximum proximodistal length is 16.5 cm, and the midshaft mediolateral width is 1.8 cm. The general shape is similar to that of other crocodyliform ulnae, including baurusuchids and
The proximal end of the ulna is craniocaudally expanded compared to both the shaft and distal end, as in other crocodyliforms. Since the proximal end is damaged, the structures of the articular surface with the humerus are not preserved. The olecranon process is severely damaged, hampering the assessment of its morphology. Nevertheless, two expansions are preserved in the proximal end, a cranial process and a conspicuous lateral process. Prior to taphonomic damage, the proximal surface of the lateral process corresponded to the ulnar radiohumeral surface, but the radial facet is still preserved. In proximal view, the ulna-radius articulation forms a sinusoidal contact (Fig. 3A). In caudal view, distal to the olecranon processes, scars are seen for the insertion of the M. triceps brachii tendon (Meers, 2003).
& Kellner, 2011). Due to the fragmentary condition of the region, the flexor ridge that would mark the limit between M. pronator quadratus and M. flexor digitorum longus pars ulnaris (Meers, 2003) is not preserved. However, the latter muscle extends distally until the cranial oblique process of the distal condyle, as seen by the well-marked scars for its insertion proximal to the process, as seen in many fossil taxa (as Baurusuchus albertoi, Stratiotosuchus maxhechti, Simosuchus clarki) and also in living forms (Brochu, 1992;Riff, 2007;Nascimento, 2008;Sertich & Groenke, 2010).
The distal end of the ulna has a craniocaudal breadth 45% shorter than that of the proximal end. The distal condyle has both cranial and caudal oblique processes turned medially. These processes are about the same size, giving the bone a heart-shaped outline in distal view. The craniolateral process is not completely preserved, due to a damage that also affected the distal surface of the condyle, preventing a precise assessment of the ulnare and radiale articulations. Yet, preserved parts suggest the ulnar articulation with the carpal bones was similar to that of other mesoeucrocodylians, such as Stratiotosuchus maxhechti, in which the cranial oblique process articulates with the radiale and the caudal process articulates with the ulnare (Riff & Kellner, 2011).
Radius. The right radius is preserved in the holotype of Pissarrachampsa sera (LPRP/USP 0019). The straight proximodistal extension of its slender shaft gives the bone a rod-like shape; which seems to be exaggerated due to the badly preserved proximal and distal ends (Fig. 4). Its maximum proximodistal length is 16 cm, and the midshaft mediolateral width is 1.4 cm. This general shape resembles that of other baurusuchid radii (Nascimento & Zaher, 2010;Vasconcellos & Carvalho, 2010;Godoy et al., 2014), but it is less robust than in

Stratiotosuchus maxhechti
The lateral and medial processes of the proximal condyle are not complete but the lateromedial expansion of the proximal end is clear, as in most crocodyliforms (Pol, 2005). The proximal end of the radius is bent cranially at an angle of approximately 25 • . In cranial view (Figs. 4A-4B), the radiohumeral articular surface bears a concavity for the articulation of the radial condyle of the humerus. In caudal view (Figs. 4E-4F), part of a crest is seen, adjacent to the lateral process of the proximal condyle. This crest is described by Pol (2005) for Notosuchus terrestris as a thin proximodistal crest and is also present in Simosuchus clarki, as well as in the baurusuchids Stratiotosuchus maxhechti and Baurusuchus albertoi (Nascimento & Zaher, 2010;Sertich & Groenke, 2010;Riff & Kellner, 2011). The ulnar facet is poorly preserved, but it is represented in caudal view by a concavity between the lateral and medial processes. The medial process of the proximal condyle bears, on its medial surface, the scar for the tendon of M. humeroantebrachialis inferior (Figs. 4E-4H). This scar was described by Turner (2006) for Araripesuchus tsangatsangana, and is also present in Simosuchus clarki and Baurusuchus albertoi (Nascimento & Zaher, 2010;Sertich & Groenke, 2010). Caudodistal to this scar, the tubercle for the insertion of M. biceps brachii is seen (Meers, 2003).
The radial shaft is elliptical in cross-section, and marked by scars and ridges for muscle insertions. In cranial view (Figs. 4A-4B), distal to the proximal condyle, the scar for the M. abductor radialis insertion is present, lateral to the tuberosity for the insertion of M. humeroradialis. This scar extends distally to the midlenght of the shaft, as in other notosuchians and living crocodylians (Meers, 2003;Pol, 2005;Turner, 2006;Sertich & Groenke, 2010). More distally, in the midline of the cranial surface, a proximodistally elongated ridge separates the insertions of M. supinator laterally and M. pronator teres, medially, along most of the shaft (Meers, 2003). This ridge is also seen in Baurusuchus albertoi, but less marked than in Stratiotosuchus maxhechti (Nascimento & Zaher, 2010;Riff & Kellner, 2011). The proximodistally long insertions of M. extensor carpi radialis brevis and M. pronator quadratus are better seen, respectively, on the lateral and caudal surfaces (Figs. 4C-4F) (Meers, 2003). A well-developed, proximodistal elongated ridge marks the caudal limit of M. extensor carpi radialis brevis and the lateral limit of M. pronator quadratus (Meers, 2003) at the lateral surface of the distal half of the shaft (Figs. 4C-4D). This ridge extends from the first to the third quarters of the shaft, resembling that of Simosuchus clarki, Baurusuchus albertoi and Aplestosuchus sordidus (Godoy et al., 2014) (Sertich & Groenke, 2010Nascimento & Zaher, 2010), but is smoother than that of Stratiotosuchus maxhechti (Riff & Kellner, 2011). Also in lateral view, another ridge, in the proximal half of the shaft, separates the insertion extensions of M. extensor carpi radialis brevis and M. abductor radialis (Meers, 2003). This ridge almost reaches the cranial surface, as in other baurusuchids, differing from the pattern seen in Simosuchus clarki, in which the ridge is restricted to the lateral surface (Sertich & Groenke, 2010;Nascimento & Zaher, 2010;Riff & Kellner, 2011;Godoy et al., 2014).
The distal end of the radius is lateromedially expanded and strongly compressed craniocaudally. In distal view, the caudal surface is concave for the articulation with the ulna (Fig. 3B). On the caudal surface of the distal end (Figs. 4E-4F) a small vascular foramen is seen medial to the ulnar articulation concavity. The radiale articulates with the cranial convex surface of the radius. This articulation gives the radial distal end two separate condyles, a more distally extended medial condyle and a lateral one, as seen in Stratiotosuchus maxhechti and Simosuchus clarki (Sertich & Groenke;Riff & Kellner, 2011).
The proximal surface of the right radiale of Pissarrachampsa sera (holotype, LPRP/USP 0019) is not completely exposed. However, as the preserved medial two-thirds of the surface are concave, this appears to be also the condition of the lacking portion, whereas the lateral third is occupied by a proximally directed convex lateral process. The same pattern is found in Simosuchus clarki, Stratiotosuchus maxhechti, Notosuchus terrestris, Baurusuchus albertoi, Sebecus icaeorhinus, and Yacarerani boliviensis (Novas et al., 2009) (Pol, 2005Riff, 2007;Nascimento & Zaher, 2010;Sertich & Groenke, 2010;Pol et al., 2012;Leardi et al., 2015). The exposed portion of the proximal surface represents the articulation for the distal end of the radius, as described for Baurusuchus albertoi, Simosuchus clarki, Stratiotosuchus maxhechti and Araripesuchus tsangatsangana (Turner, 2006;Riff, 2007;Nascimento & Zaher, 2010;Sertich & Groenke, 2010). The presence of a marked longitudinal crest in the cranial surface of the radiale has been described for several notosuchians, such as Notosuchus terrestris, Baurusuchus albertoi, Sebecus icaeorhinus, Stratiotosuchus maxhechti, and Yacarerani boliviensis (Pol, 2005;Riff, 2007;Nascimento & Zaher, 2010;Sertich & Groenke, 2010;Pol et al., 2012;Leardi et al., 2015). On the other hand, Turner (2006) describes a ''median ridge'' in Araripesuchus tsangatsangana, which may correspond to the longitudinal crest. There is no sign of such a crest in the exposed surface of the radiale of Pissarrachampsa sera, but its absence cannot be confirmed as most of the cranial surface of the radiale is embedded in the rock matrix.
Sertich & Groenke (2010) described a prominent pit and a raised rugosity for Simosuchus clarki, which topologically corresponds to the proximal portion of the cranial longitudinal crest in Mahajangasuchus insignis, and represents the insertion of the M. extensor carpi radialis longus (Meers, 2003). The presence of raised scars medial and lateral to this pit is has also been described for Simosuchus clarki, consistent with the origin of the superficial extensor muscles for digits I, II and III (Brochu, 1992;Meers, 2003;Sertich & Groenke, 2010). In Pissarrachampsa sera, despite the lack of the pit, it is possible that the exposed surface of the radiale includes the insertion areas of those extensor muscles, or at least those lateral to the pit in Simosuchus clarki. Abbreviations: I mc, metacarpal I; II mc, metacarpal II; III mc, metacarpal III; IV mc, metacarpal IV; V mc, metacarpal V; dph, distal phalanx; mph, medial phalanx; pph, proximal phalanx; rdl, radiale; uln, ulnare. Scale bar equals 5 cm.
The distal end of the ulnare is more expanded than the proximal, as in Notosuchus terrestris, Sichuanosuchus shuhanensis (Wu, Sues & Dong, 1997), Baurusuchus albertoi, Araripesuchus tsangatsangana, Stratiotosuchus maxhechti, Simosuchus clarki, Yacarerani boliviensis, and most non-Crocodylia crocodyliforms (Pol, 2005;Turner, 2006;Riff, 2007;Nascimento & Zaher, 2010;Sertich & Groenke, 2010;Leardi et al., 2015). Yet, the bone is not exposed enough to see if this expansion is symmetrical, as in Simosuchus clarki and Yacarerani boliviensis, or more marked medially, as in Notosuchus terrestris, Stratiotosuchus maxhechti and Baurusuchus albertoi (Leardi et al., 2015) Manus. Two right manus are associated to Pissarrachampsa sera, one of the holotype (LPRP/USP 0019) and an isolated one (LPRP/USP 0745). The holotypic right manus ( Fig. 5) is composed of five digits: the first includes the metacarpal and the proximal phalanx; the second includes the metacarpal, a poorly preserved proximal phalanx, and the distal phalanx; the third includes the metacarpal and fragments of the medial portions of three phalanges; the last two digits include only the metacarpals. The right manus of LPRP/USP 0745 preserves (albeit partially) all five metacarpals, an incomplete proximal phalanx of the digit I, and a fragment that might represent the proximal phalanx of the digit III. The holotypic manus is better seen in ventral view (Fig. 5B), whereas LPRP/USP 0745 has only its dorsal surface exposed.

Pelvic girdle
Ilium. One left ilium is partially preserved for Pissarrachampsa sera (Fig. 6), from a referred specimen (LPRP/USP 0742). It lacks the distal part of the postacetabular process, most of the preacetabular process, and the ventral portion of the acetabular region. The acetabulum is deep, as in Baurusuchus albertoi and Sebecus icaeorhinus, as a result from the strictly lateral orientation of the supraacetabular crest (Nascimento & Zaher, 2010;Pol et al., 2012). On the other hand, the supraacetabular crest of Araripesuchus tsangatsangana projects not only laterally, but also dorsally, which gives a shallower aspect to the acetabulum (Turner, 2006). In some neosuchians and living taxa, the crest is strongly inclined dorsally, giving an accentuated shallow aspect to the acetabulum in lateral view (Leardi, Fiorelli & Gasparini, 2015).
In Pissarrachampsa sera, the morphology of the dorsal surface of the acetabular roof resembles that of Baurusuchus albertoi (Figs. 6A-6B) (Nascimento & Zaher, 2010). In both taxa, the dorsal component of the supraacetabular crest is confluent with the remaining dorsal portion of the bone, extending as a flat horizontal surface, giving the ilium a broad aspect. On the other hand, in Sebecus icaeorhinus, Microsuchus schilleri (Dolgopol de Sáez, 1928), and living forms, such as Caiman latirostris (Daudin, 1802) (MZSP 2137), the supraacetabular crest is not confluent with the rest of the dorsal margin, but has a medial boundary (Pol et al., 2012;Leardi, Fiorelli & Gasparini, 2015). In Sebecus icaeorhinus and Caiman yacare (Daudin, 1802), the dorsal margin is sloped, with the portion corresponding to the supraacetabular crest lying dorsal to the medial portion of the iliac dorsal surface (Nascimento, 2008;Pol et al., 2012). Given the great lateral projection of the supraacetabular crest, the maximum width of the dorsal margin of the ilium of Pissarrachampsa sera is located right above the caudal margin of the acetabular area. The rest of the dorsal surface becomes gradually narrower in the direction of both the pre-

Figure 6 Pissarrachampsa sera (LPRP/USP 0742), photographs and schematic drawing of the left ilium in dorsal (A and B), medial (C and D), and lateral views (E).
Cross-hatched areas represent broken surfaces. Abbreviations: ac, acetabulum; acr, acetabular roof; das, dorsal portion of the articular surface for the second sacral rib; dmar, dorsal margin of the acetabular roof; pap, postacetabular process; imr, ridge on the medial surface of the ilium; s 1r, articular surface for first sacral rib; s 2r, articular surface for second sacral rib. Scale bar equals 5 cm. and postacetabular processes. Rugosities on the dorsal surface of the supraacetabular crest indicate the area for the attachment of M. iliotibialis 1 and 2 (Romer, 1923;Leardi, Fiorelli & Gasparini, 2015). In Pissarrachampsa sera, most of this surface is rugose, indicating a greater area for the attachment of those muscles.
The proximal portion of the postacetabular process is at least four times dorsoventrally higher than lateromedially wide, and its dorsal margin is slightly caudoventrally oriented. In medial view, it is possible to see the medial expansion of the dorsal portion of the postacetabular process, forming a ridge that extends craniocaudally (Figs. 6C-6D). This ridge marks the dorsal limit of a concave surface on the medial portion of the ilium. Ventrally, this concavity is delimited by a curved ridge, which corresponds to the dorsal part of the articular surface for the second sacral rib (see Pol et al., 2012), and this same morphology is also seen in Baurusuchus albertoi and Sebecus icaeorhinus (Nascimento & Zaher, 2010;Pol et al., 2012). On the other hand, in Theriosuchus pusillus (Owen, 1879) and some extant taxa, such as Caiman yacare and Melanosuchus niger (Spix, 1825), there is no evidence of a supraacetabular process medial crest, which gives a more flattened aspect to the process above the articular surface for the second sacral rib (Wu, Sues & Brinkman, 1996).
Baurusuchus albertoi has a total of three sacral vertebrae, with the articulation surface for the third element located in the distal portion of the postacetabular process (Nascimento & Zaher, 2010). Three sacral vertebrae are also found in other baurusuchids, such as Baurusuchus salgadoensis (Carvalho, Campos & Nobre, 2005) (Vasconcellos & Carvalho, 2010) and Aplestosuchus sordidus (Godoy et al., 2014), and there is no evidence of a different condition in Pissarrachampsa sera, although this remains speculative due to the absence of more complete remains.
Ischium. Both left and right ischia of the holotype of Pissarrachampsa sera (LPRP/USP 0019) are partially preserved, lacking the distal portions of the ischial blade, and of the iliac and pubic peduncles. Despite the incompleteness, the typical crocodyliform ischial morphology is recognizable (Figs. 7A-7B), with a lateromedially constricted ischial blade, a caudal process which would probably contact the ilium, and a cranial process which likely contacted both ilium and pubis (Sertich & Groenke, 2010). The notch between both processes formed the ventral margin of the perforate acetabulum, similar to the condition seen in mesoeucrocodylians such as Chimaerasuchus paradoxus, Mahajangasuchus insignis, Stratiotosuchus maxhechti, and Sebecus icaeorhinus (Wu & Sues, 1996;Buckley & Brochu, 1999;Riff & Kellner, 2011;Pol et al., 2012). The proximal parts of both processes differ in thickness, with a more extended cranial process, as seen in Stratiotosuchus maxhechti and Sebecus icaeorhinus (Riff & Kellner, 2011;Pol et al., 2012). In these two taxa, however, the cranial process expands distally, becoming more robust, an unknown condition for Pissarrachampsa sera.
On the lateral surface of the ischial blade (Figs. 7A-7B), a ridge extends dorsoventrally along its proximal third marking the limits of muscles attached to the ischium. The ischium is very constricted lateromedially, cranial and caudal to this ridge, giving a sharp aspect to its margins. Caudal to the ridge is the area for attachment of both M. flexor tibialis internus pars 3 laterally and M. ischiotrochantericus medially (Hutchinson, 2001a). In the distal portion of the ischial blade, only the cranial margin is constricted, as the dorsoventral ridge becomes confluent with the caudal margin, which becomes more rounded. The constricted cranial margin corresponds to the attachment surface for M . puboischiofemoralis externus pars 3, on the medial surface of the bone (Hutchinson, 2001a;Riff, 2007). In cranial and lateral views, it is possible to see a tubercle on the dorsal portion of the ischial blade, ventral to the cranial process of the ischium. Stratiotosuchus maxhechti bears a similar tubercle, which is interpreted as the attachment point for M. pubioischiotibialis (Riff & Kellner, 2011).
Given the incompleteness of the pelvis of Pissarrachampsa sera, the isolation of the pubis from the acetabulum cannot be asserted. Yet, in all Crocodyliformes, except protosuchians, the pubis is excluded from the acetabulum by the cranial process of the ischium, which represents the articulation point for the proximal end of the pubis (Colbert & Mook, 1951). In Pissarrachampsa sera, the partially preserved proximal articulation is lateromedially constricted, and more constricted in its cranial third, giving it a pear-shaped aspect. This lateromedial constriction extends distally along the shaft, as also seen in Stratiotosuchus maxhechti (Riff, 2007). Pissarrachampsa sera and Stratiotosuchus maxhechti also share the proximal pubic shaft bent approximately 30 degrees in relation to the pubic blade. In other notosuchians, such as Araripesuchus tsangatsangana and Simosuchus clarki, and also in the living Crocodylia, such bending is unknown (Turner, 2006;Riff, 2007;Sertich & Groenke, 2010). The pubic blade is craniocaudally constricted in its medial third, which forms the pubic symphysis. Lateral to the laminar symphyseal region, the ischial blade does not show any evidence of the craniocaudal constriction. The attachment area for both M. puboischiofemoralis externus pars 1 and 2 is probably located in the proximal two thirds of the transitional area between the constricted and non-constricted regions of the pubic blade, in the caudal and cranial surfaces respectively (Romer, 1923).

Hindlimb
Femur. There are four preserved femora known for Pissarrachampsa sera. The femoral pair of the holotype (LPRP/USP 0019), as well as two smaller isolated and partially preserved left and right elements (LPRP/USP 0743 and LPRP/USP 0744). The smaller right femur is still in articulation with tibia and fibula, but the following description is based mostly on the holotypic material (Fig. 8), since these are better preserved. The femur is virtually straight in cranial and caudal views, and its proximodistal length is about 24 cm. It is longer than the tibia and or fibula, as seen in most other Mesoeucrocodylia (Leardi, Fiorelli & Gasparini, 2015). In medial and lateral views, the shaft is slightly bowed cranially, and the proximal and distal ends are cranially and caudally curved. The proximal articulation surface is medially inturned, as seen in Baurusuchus albertoi and Stratiotosuchus maxhechti, but not as displaced as in Araripesuchus tsangatsangana (Turner, 2006;Nascimento & Zaher, 2010;Riff & Kellner, 2011). In proximal view (Figs. 8I-8J), the robust articular surface is rounded and rugose at its distal portion, with scars for muscle insertion, whereas the caudolateral extension of the head is slender, as in other baurusuchids and Mariliasuchus amarali (Carvalho & Bertini, 1999) (Nascimento & Zaher, 2010;Riff & Kellner, 2011;Nobre & Carvalho, 2013). At this point, in caudal view (Figs. 8E-8F), there is a proximodistally extensive ''greater trochanter'' placed laterally, extending cranially and parallel to the ''medial proximal crest,'' at the caudal most extension of the head (Pol et al., 2012). The ''medial proximal crest'' turns caudally in Pissarrachampsa sera, and not medially as in Sebecus icaeorhinus (Pol et al., 2012).
Other muscle scars seen along the shaft, as well as a foramen mediodistal to the cranial flange. Laterodistal to the flange lies the insertion area for the M. iliofemoralis (Hutchinson, 2001b) and distal to the flange, there is an extensive intermuscular line that almost reaches the proximal limit of the intercondylar fossa (Romer, 1956). This corresponds to the M. femorotibialis internus (Hutchinson, 2001b) and its distal most extension forms a longitudinal ridge, named here ''femorotibialis ridge.'' This intermuscular line does not form a ridge in the juvenile specimen, and is interpreted as an ontogeny-related character. Caiman sp. (LPRP/USP N 0008) also has this intermuscular line, but it does not form a ridge. The presence of this ridge is not clear in other notosuchians, except for Stratiotosuchus maxhecthi and Aplestosuchus sordidus, in which it is smoother than in Pissarrachampsa sera (Riff & Kellner, 2011;Godoy et al., 2014). On the caudal face of the femoral shaft (Figs. 8E-8F), the linea intermuscularis caudalis extends obliquely, from the fourth trochanter to the proximal portion of the lateral condyle, and forms the lateral border of the popliteal fossa. This scar corresponds to the boundary between M. femorotibialis externus, craniomedially, and M. adductor femoris 1 & 2, caudolaterally (Hutchinson, 2001b).
The two distal condyles are well developed, forming the intercondylar fossa cranially and a deep popliteal fossa caudally. The latter is rugose, as in Stratiotosuchus maxhechti, whereas the intercondylar fossa has smoother scars for muscle insertions (Romer, 1956;Riff & Kellner, 2011). The lateral or fibular condyle has a laterodistal concavity, possibly related to the fibular articulation. It is about two times larger than the medial or tibial condyle, which is not as distally expanded as the lateral condyle, a general crocodyliform condition (Sertich & Groenke, 2010;Pol et al., 2012). In lateral view (Figs. 8G-8H), the rugose surface above the lateral condyle makes the insertion of M. gastrocnemius (Brochu, 1992;Sertich & Groenke, 2010). Cranially, the distal portion of the femur has a well developed medial supracondylar ridge, whereas the lateral supracondylar ridge is smoother. This differs from the condition in Sebecus icaeorhinus, which lacks a marked transition from the cranial to the lateral surfaces of the distal femur (Pol et al., 2012). The caudal surface (Figs. 8E-8F) of the distal femur bears both medial and lateral supracondylar ridges (the latter would be the distal extension of the linea intermuscularis caudalis), as well as a popliteal fossa between these (Hutchinson, 2001b;Pol et al., 2012). The medial supracondylar ridge forms a proximodistally oriented crest, above the medial condyle, separating the caudal and lateral surfaces of the distal portions of the femur. The medial facet of the distal portion of the femur is almost flat, cranially bound by the medial supracondylar ridge, whereas in Sebecus icaeorhinus this surface is slightly convex (Pol et al., 2012).
Tibia. Both tibiae of the holotype (LPRP/USP 0019) are nearly complete, and articulated with the fibulae in their original position (Fig. 9). Additionally, there is a smaller isolated right tibia (LPRP/USP 0741), as well as the additional right tibia in articulation with femur and fibula (LPRP/USP 0744). The shafts of the articulated tibia and fibula are very close to one another (Figs. 9A-9B), as are the radius and ulna. This condition is different from that of modern crocodylians (e.g., Caiman and Melanosuchus) in which this distance is larger. The tibia of Pissarrachampsa sera is similiar in robustness to the tibiae of most crocodyliforms, differing from the more gracile elements of Araripesuchus tsangatsangana and Microsuchus schilleri (Brochu, 1992;Turner, 2006;Leardi, Fiorelli & Gasparini, 2015). The tibia is 18.6 cm long, i.e., 77% the femur's length, the same ratio of Sebecus icaeorhinus. This differs from other notosuchians, such as the relatively short tibia of other baurusuchids, such as Baurusuchus albertoi and Stratiotosuchus maxhechti, (about 72%) and the elongated bone (82%) of Araripesuchus tsangatsangana (Pol et al., 2012).
The proximal and distal extremities of the tibia are mediolaterally well expanded. The proximal surface is divided into medial and lateral facets (Figs. 9A-9B), which respectively correspond to the articulation areas for the tibial and fibular condyles of the femur. In proximal view, the medial articulation (posteromedial proximal process of the tibia, according to Leardi et al., 2015) has a trapezoid-shape; a pattern also seen in other baurusuchids, such as Stratiotosuchus maxhechti and Baurusuchus albertoi (Nascimento & Zaher, 2010;Riff & Kellner, 2011). The medial articular facet is more protruded relative to the lateral one. The proximal surface of the medial facet forms a gentle concavity, corresponding to the ''proximal pit '' sensu Brochu (1992), and bears a pronounced deflection toward its caudomedial corner (Fig. 9). This condition is also observed in Sebecus icaeorhinus, which bears a gently protruded medial facet, but differs from Mariliasuchus amarali, Yacarerani boliviensis, and Stratiotosuchus maxhechti, in which that medial portion is weakly pronounced (Pol et al., 2012;Leardi et al., 2015). The latter condition is also present in modern crocodylians (e.g., Caiman, Melanosuchus and Alligator) resulting in equally projected facets. The lateral articular facet is semi-lunar in shape and slightly concave in proximal view. The cranial border is rounded and the caudal tip is somewhat deflected distally. It resembles the pattern of Sebecus icaeorhinus and Yacarerani boliviensis, differing from the weakly projected tip of Mariliasuchus amarali, Araripesuchus tsangatsangana and Stratiotosuchus maxhechti (Turner, 2006;Riff & Kellner, 2011;Pol et al., 2012;Nobre & Carvalho, 2013;Leardi et al., 2015).
Cranially, close to the medial margin of the distal expansion, there is a protuberance for insertion of M. interosseus cruris. This structure is placed more proximally in extant taxa, slightly developed in Caiman and Melanosuchus, but marked in Alligator (Brochu, 1992). Among Baurusuchidae, both Stratiotosuchus maxhechti and Baurusuchus albertoi bear the same protuberance, although less prominent in the latter (Nascimento & Zaher, 2010;Riff & Kellner, 2011). Craniolaterally, the distal end of the tibia is devoid of the circular depression for the attachment of the medial tibioastragalar ligament, which is clearly seen in Araripesuchus tsangatsangana (Turner, 2006).
Fibula. Both fibulae of the holotype of Pissarrachampsa sera (LPRP/USP 0019) are virtually complete (Fig. 9) and in articulation with the tibiae. This is also the case for the fibula of LPRP/USP 0744, preserved in articulation with femur and tibia. The fibula of the holotype is 17 cm long, slender and slightly shorter than the tibia. The fibular width corresponds to half that of the tibia, differing from Baurusuchus albertoi, the fibula of which is three times thinner than the tibia (Nascimento & Zaher, 2010). The proximal articular surface is gently concave, with the lateral border more developed than the medial. In proximal view, the fibula is crescentic in shape and the medial margin is slightly notched. In contrast, the proximal fibula of Stratiotosuchus maxhechti is caudally wedged (Riff & Kellner, 2011).
The fibular shaft is almost entirely compressed lateromedially, except in its middle portion, which is elliptical in cross-section. Laterally, the fibular shaft bears faintly developed ridges, as in Baurusuchus albertoi, corresponding to the origin of M. peroneus longus (sensu Brochu, 1992) or M. fibularis longus (sensu Hutchinson, 2002). A different condition is seen in Stratiotosuchus maxhechti, in which that ridge is well developed (Riff, 2007). Among extant crocodylians, both Caiman and Melanosuchus show weakly developed ridges on the lateral surface of the fibular shaft, whereas in Alligator the fibula bears well developed crests and a slightly rugose shaft lateral surface (Brochu, 1992). In medial view, the shaft is mostly smooth and lacks any distinctive muscle scar. However, the caudodistal surface is rugose, revealing scars possibly related to the attachment for M. interosseus cruris, as also observed in Araripesuchus tsangatsangana and Stratiotosuchus maxhechti (Turner, 2006;Riff, 2007). There is a small vascular foramen on the caudal surface near the midshaft. The tibial distal end is enlarged with a triangular distal outline, as in Araripesuchus tsangatsangana and Microsuchus schilleri (see Leardi, Fiorelli & Gasparini, 2015: character 425). As in Alligator, Caiman, and Melanosu''hus, a ''dis''al hook'' (sensu Brochu, 1992) contacts the tibia and tapers medially. This differs from the condition in Stratiotosuchus maxhechtiand Yacarerani boliviensis, in which the medial end of the distal margin of the tibia is rounded (Riff & Kellner, 2011;Leardi et al., 2015). The contact of the distal hook with the tibia is more proximal then the distal tibial articulation (Fig. 9), and differs from the pattern in Microsuchus schilleri, the distal hook of which contacts the tibia more distally. This hook is absent in Araripesuchus tsangatsangana and Yacarerani boliviensis (Turner, 2006;Leardi et al., 2015).
Tarsus. Both complete astragali and calcanea are preserved in articulation (Fig. 10) in the holotype of Pissarrachampsa sera (LPRP/USP 0019), although the more distal tarsal bones are not preserved. The best preserved left astragalus and calcaneum are slightly displaced from their original positions. The tarsal morphology of Pissarrachampsa sera is similar to that of other crocodylomorphs with the ''crocodile normal'' condition, in which the astragalar ''peg'' fits into the calcaneal ''socket'' (Chatterjee, 1978;Chatterjee, 1982). In this configuration, the astragalus is fixed in articulation with tibia and the ankle rotation occurs between astragalus and calcaneum (Brochu, 1992).
Proximally, the astragalus bears a concave and laterally elongate surface for articulation with the distal tibia (Figs. 10A-10B). The division of this surface for the reception of medial and lateral condyles of the tibia is weak and both facets are similar in lateromedial extension. These are bounded caudally by a ridge, but this structure is more developed
The cranial and proximal portions of the cranial body form a well-developed rounded articular surface (a roller) that articulates medially with the astragalus and proximally with the fibula. This morphology is widespread, also seen in living forms and other fossil crocodylians, as Baurusuchus albertoi, Stratiotosuchus maxhechti, Sebecus icaeorhinus, Simosuchus clarki, and Araripesuchus tsangatsangana (Brinkman, 1980;Turner, 2006;Sertich & Groenke, 2010;Nascimento & Zaher, 2010;Riff & Kellner, 2011;Pol et al., 2012). No ridge is present at the articular surface of the roller, which in Simosuchus clarki separates the medial articulation area for the astragalus and the lateral articulation area for the fibula (Sertich & Groenke, 2010). This rounded surface slopes abruptly cranioventrally, forming a distally directed surface, which probably contacted the fourth distal tarsal. In Pissarrachampsa sera, this surface is flat and elliptical in distal view, resembling the condition in Stratiotosuchus maxhechti (Riff & Kellner, 2011). The lateral portion of the cranial body forms a well-developed flat surface that lacks any articular facet. This surface is proximodistally restricted and does not overcome the proximodistal extension of the distal tuber. The medial face of the cranial body forms the calcaneal socket. Most of the morphology of this area is not accessible due the articulation with the astragalus, but a faint medial flange overhangs the calcaneal socket as in Simosuchus clarki (Sertich & Groenke, 2010).
Pes. Pissarrachampsa sera has three preserved pedes, the left pes of the holotype (LPRP/USP 0019) and two referred (a left and a right) pedes (LPRP/USP 0739 and LPRP/USP 0746). The holotype pes is represented by four articulated metatarsals (Fig. 11B), whereas LPRP/USP 0739 includes four isolated metatarsals, and LPRP/USP 0746 comprises four partially preserved articulated digits (Fig. 11A). Metatarsal V is not preserved in any of the specimens of Pissarrachampsa sera, following the trend of reduction of that metatarsal towards Crocodylomorpha (Parrish, 1987). Therefore, the four metatarsals preserved in Pissarrachampsa sera constitute the entire number of fully functional pedal digits, as in all living crocodylians and most fossil crocodyliforms (Riff, 2007).

Body size and mass estimates of Pissarrachampsa sera
The preserved elements of the holotype (LPRP/USP 0019), particularly the femora, allow estimating the body size and mass of Pissarrachampsa sera. Based on the protocol presented by Farlow et al. (2005), we estimated that Pissarrachampsa sera had a total length varying between 2.7 and 3.5 m, and a body mass between 81 and 163 kilograms (for detailed results see Supplemental Information). This significant variation is also observed in estimates for other terrestrial crocodyliforms, such as Protosuchus and Sebecus (Farlow et al., 2005;Pol et al., 2012). The regressions of Farlow et al. (2005) were built with data from Alligator mississippiensis, and might not be as accurate as desired for fossil taxa with different habits and body proportions, as already pointed out by other works (e.g., Young et al., 2011;Pol et al., 2012).
Indeed, the comparison with nearly complete baurusuchid specimens permits assessing the accuracy of these regressions for the group. Comparisons to more complete baurusuchids such as the 1.9 m long specimen referred to Baurusuchus salgadoensis (lacking only the skull and pectoral girdle), the 1.3 m long holotype of Baurusuchus albertoi (lacking the tip of tail and snout), and the 1.1 m long holotype of Aplestosuchus sordidus (lacking the tail) (Nascimento, 2008;Vasconcellos & Carvalho, 2010;Godoy et al., 2014) suggest that it is unlikely that any of these specimens reached the maximum length estimated for Pissarrachampsa sera (3.49 m) using the regressions. Further, after applying the formulas from Farlow et al. (2005) for Baurusuchus albertoi and B. salgadoensis (both with femora well preserved), we obtained a total length of approximately 3.8 m for both taxa (see Supplemental Information). Even though not completely preserved, this is evidence that, at least for baurusuchids, these regressions are overestimating the size of the specimens. Additionally, in order to test the validity of the mass estimates obtained with the formulas from Farlow et al. (2005), we also applied the equations presented by Campione & Evans (2012), which uses proximal (stylopodial) limb bone circumference to obtain total body mass, and seems to work well for many fossil taxa (e.g., Castanhinha et al., 2013;Benson et al., 2014;Reisz & Fröbisch, 2014). After applying the femur-based equation, the mass estimate obtained for Pissarrachampsa sera was approximately 71 kilograms, lower than the lowest value obtained using Farlow et al. (2005) formulas.
Regardless of the incompleteness of specimens and inaccuracy of size estimates, it is very likely that an adult individual of Pissarrachampsa sera reached at least 2 m (Fig. 12), placing the taxon amongst the largest terrestrial predators of Late Cretaceous environments in southeast Brazil, together with other baurusuchids and theropods (Riff & Kellner, 2011;Godoy et al., 2014). The Bauru Group rocks have provided numerous carnivorous crocodyliforms (e.g., Campos et al., 2001;Carvalho, Campos & Nobre, 2005;Godoy et al., 2014), particularly baurusuchids, and many titanosaur sauropods (e.g., Kellner & Azevedo, 1999;Salgado & Carvalho, 2008;Santucci & Arruda-Campos, 2011), but very few theropods (Novas et al., 2008;Bittencourt & Langer, 2011;Méndez, Novas & Iori, 2012;Azevedo et al., 2013). This has been used as evidence for the rearrangement of roles in this paleoecosystem, with baurusuchids occupying the typical ecological niche of theropods or at least competing for the same niche (Gasparini, Fernandez & Powell, 1993;Candeiro & Martinelli, 2006;Riff & Kellner, 2011). However, although the morphology of baurusuchids indicates a highly specialized predatory habit, similar to that of theropods, it seems unlikely that even larger baurusuchids could have preyed on adult sauropods (>8-meter length for some titanosaurs; Salgado & Carvalho, 2008), if assumed as solitary predators. Although young theropods could have had similar diets to baurusuchids, the morphological differences are also indicative of distinct feeding (Martinelli et al., 2013). Indeed, this hypothesis is supported by the single reliable and identifiable direct evidence of predation among baurusuchids, in which a small sphagesaurid (Mesoeucrocodylia, Notosuchia) was found in the abdominal cavity of the holotypic skeleton of Aplestosuchus sordidus (Godoy et al., 2014). As such, if adult sauropods had any predator in this Cretaceous ecosystem, theropods remain as the most likely ones, and the scarcity of theropods might reflect incomplete or biased sampling. Accordingly, some niche partitioning may have occurred, with baurusuchids preying on smaller animals, as well as young or hatchling sauropods, and adult theropods being able to prey on larger individuals.

Terrestriality in Pissarrachampsa sera
A series of anatomical features have been recognized as related to the terrestrial habits of Crocodyliformes, many of which are observed in the postcranial skeleton of Pissarrachampsa sera. Most of these concern an upright posture and gait, with the limbs held under the body rather than to the side as in extant crocodylians. A characteristic presumably linked to terrestriality is the reduced space between articulated ulna and radius in Pissarrachampsa sera. Although contrasting with the relatively large space in extant crocodylians, this pattern is also observed in other baurusuchids, such as Stratiotosuchus maxhechti and Baurusuchus albertoi, as well as in the terrestrial notosuchian Araripesuchus tsangatsangana (Brochu, 1992;Turner, 2006;Nascimento & Zaher, 2010;Riff & Kellner, 2011). Similarly, the space between tibia and fibula of Pissarrachampsa sera is also reduced. Further, the proximal portion of its tibia bears a well-protruded medial facet that corresponds to the articulation with the tibial condyle of the femur. The uneven proximal facets rotate the distal tibia laterally when in articulation with the femur. Accordingly, both propodium and epipodium were arranged on the same long axis (in caudal or cranial views), allowing a parasagittal movement of the leg during locomotion. This condition is also seen in the terrestrial notosuchians Sebecus icaeorhinus and Simosuchus clarki (Sertich & Groenke, 2010;Pol et al., 2012). The proximal articulation facets of the tibia are caudally separated by an excavated fossa flexoria, and cranially, by a large tuberosity for the insertion of M. flexor tibialis internus.This is evidence of a tight/stable knee joint in agreement with an erect posture. Also, the distal tibial articulation of Pissarrachampsa sera is obliquely disposed, with a more enlarged medial facet, as in Stratiotosuchus (Riff & Kellner, 2011). Extant crocodylians, on the other hand, bear equally developed distal ends (medial and lateral) of the tibia, allowing a range of sprawling to semi-erect high walk (Brinkman, 1980;Parrish, 1986;Parrish, 1987;Gatesy, 1991). This oblique articulation and the sharp distal end of the tibia fits tightly with the astragalus, and can reduce the range of movements. But it also indicates a stable articulation with the foot, allowing some lateral displacement, matching the medial displacement of the distal tibia, denoting an upright posture. This is similar to the ankle articulation morphology seen in the terrestrial sphenosuchians and protosuchians (Parrish, 1987), but it is also observed in more closely-related taxa, as Araripesuchus tsangatsangana and Sebecus icaeorhinus.
Additionally, the less curved femur of P. sera, in comparison to that of living crocodylians, is in accordance with a more erect posture. The faint curvature in this bone is similiar to that seen in Stratiotosuchus maxhechti, for which a parasagittal posture was also claimed (Riff & Kellner, 2011). Hutchinson (2001b argues that limb bones, such as the femur, with a less accentuate curvature are subjected to bending stresses rather than torsional stresses. That anatomical acquisition would then be related to a more erect posture and terrestrial habits in the archosaurian lineage, whereas bones under torsional stresses, such as sigmoid femora, are associated with forms with a sprawling posture. Still, some of features pointed out by Parrish (1987) as linked to a parasagittal posture in archosaurians are also observed in Pissarrachampsa sera, such as a well-developed and medially inturned femoral head, prominent caudally oriented femoral condyles, and a conspicuous fibular condyle (or lateral condyle). Further, the femur orientation is compatible with the morphology of the ilium of P. sera. The laterally projected and enlarged supraacetabular crest would make it impossible for the femur to be strictly laterally oriented (Riff & Kellner, 2011), but would be compatible with a vertical orientation of a parasagital posture. Still in the pelvic girdle, Pissarrachampsa sera possess a tubercle on the lateral surface of the ischium, located in the attachment area of M . pubioischiotibialis. Riff & Kellner (2011) pointed out that this tubercle is absent in extant forms, and its big size in Stratiotosuchus, similar to the morphology observed in P. sera, can indicate that this muscle was more developed in the baurusuchids. Indeed, Reilly & Blob (2003) show that, in Alligator, this muscle is activated during the ''high-walk'' locomotion mode, which is compatible with the interpretation of Riff & Kellner (2011) suggesting that a greater development of the M. pubioischiotibialis is compatible with a permanent parasagital posture, more related to a terrestrial lifestyle.

The lack of osteoderms in Pissarrachampsa sera
Pissarrachampsa sera is represented by a series of specimens all from the same locality. The specimens range from the relatively complete and fairly articulated holotype to isolated fragmentary cranial and postcranial elements. So far, no osteoderm was found associated with these specimens, neither elsewhere in the type locality. This raises the question whether the lack of osteoderms represents a taphonomic signature or a genuine anatomical feature of the taxon. In the latter case, Pissarrachampsa sera would be the first terrestrial crocodyliform to completely lack any body armor, with biomechanical implications to be explored.
The specimens of Pissarrachampsa sera were collected without rigorous taphonomic control, but there is geological and paleontological evidence that supports the absence of osteoderms as unrelated to taphonomy. The type locality of P. sera is assigned to the Adamantina Formation and the deposition of this geological unity is associated with arid to semi-arid conditions (Fernandes & Coimbra, 1996;Fernandes & Coimbra, 2000;Batezelli, 2015). In the same way, the local geology suggests a developed paleosol profile that is also indicative of arid to semi-arid conditions (JCA Marsola et al., unpublished data). In this scenario, the prolonged periods without sedimentation lead to erosion and pedogenesis. Furthermore, well-preserved and complete crocodyliform egg clutches are found in the same levels of the body fossils of Pissarrachampsa sera (Marsola, Montefeltro & Langer, 2011). Crocodyliform eggs are particularly fragile to long-range transport (Grellet-Tinner et al., 2006;Hayward et al., 2000), whereas the skeletal elements of P. sera do not show significant signs of abrasion caused by transport (Montefeltro, Larsson & Langer, 2011). Therefore, the decay and burial of the P. sera remains most likely occurred in a low-energy, probably sub-aerial environment.
Araújo -Júnior & Marinho (2013) analyzed the taphonomy of one specimen of Baurusuchus pachecoi from the same formation, collected in Jales (São Paulo, Brazil), which matches the putative pre-burial conditions experienced by Pissarrachampsa sera. In that study, osteoderms were found close to their in vivo position, even after being exposed to some degree of scavenging and sub-aerial decay. A similar pattern of osteoderm disarticulation was found by Beardmore et al. (2012) for the marine crocodyliform Steneosaurus (Geoffroy Saint-Hilaire, 1825), from the Posidonienschiefer Formation (Lower Jurassic, Germany), which decayed and were buried in a quiet-water, marine basin. In that case, osteoderms are placed close to the carcass even in specimens with greater degree of disarticulation. The same pattern is seen in actualistic taphonomic experiments in juvenile Crocodylus porosus (Schneider, 1801), in which the osteoderms remain at the vicinity of the carcass even with relatively prolonged subaerial and subaqueous decay (Syme & Salisbury, 2014, Fig. 6). In fact, a series of fossil crocodyliforms, both close and distantly related to Pissarrachampsa sera, are recovered with associated osteoderms, even showing a relatively advanced degree of disarticulation, as. Susisuchus anatoceps (Salisbury et al., 2003), Candidodon itapecuruense (Carvalho & Campos, 1988 (Iori, Carvalho & Marinho, 2016). We took into consideration the possibility that Pissarrachampsa sera had its osteoderms disarticulated earlier in the decay process. This is possible and is supported by specimens of closely-related notosuchians with fairly articulated postcrania but lacking osteoderms, such as Mariliasuchus amarali (UFRJ-DG-50-R), Notosuchus terrestris (MUCPv-137), Sebecus icaeorhinus (Pol et al., 2012). However, in the particular case of P. sera we regard this as unlikely, given the complete absence of these elements in the entire outcrop and the number of specimens recovered. Therefore, in light of all evidence we suggest the lack of osteoderms is an inherent and diagnostic feature of Pissarrachampsa sera.
The presence of osteoderms is considered plesiomorphic for Crocodyliformes (Scheyer & Desojo, 2011), as these structures are found in most pseudosuchians (Brown, 1933;Wu & Chatterjee, 1993;Clark & Sues, 2002;Sues et al., 2003;Pol & Norell, 2004;Clark, 2011;Nesbitt, 2011;Scheyer & Desojo, 2011). Likewise, this ancestral condition is inferred for most internal nodes of Crocodyliformes, which bear at least one pair of parasagittal rows forming the body armor (Salisbury & Frey, 2001;Frey & Salisbury, 2001;Hill, 2005;Pierce & Benton, 2006;Jouve et al., 2006;Marinho & Carvalho, 2009;Pol, Turner & Norell, 2009;Hill, 2010;Andrade et al., 2011;Pol et al., 2012;Nobre & Carvalho, 2013;Tennant & Mannion, 2014). The only exception known so far is the complete absence of osteoderms in the marine metriorhynchids, a feature probably associated with their aquatic lifestyle (Young et al., 2010;Young et al., 2013;Molnar et al., 2015). Similarly, metriorhynchids do not have palpebral bones roofing the orbits (Nesbitt, Turner & Weinbaum, 2012), and previous analyses of the crocodylian skeletogenesis show that postcranial osteoderms match the palpebral development (Vickaryous & Hall, 2008). In this case, it might have been a common cause underlying the successive loss of the palpebrals and postcranial osteoderms in Thalattosuchia and Metriorhynchidae. Molnar et al. (2015) presented evidence that the loss of osteoderms in Metriorhynchidae is related to an increasing aquatic adaptation in this group, whereas the rigid series of osteoderms of early crocodylomorphs would be related to terrestrial habits. In this scenario, the presence of non-imbricate osteoderms in teleosaurid thalattosuchians and the more flexible arrangement of these structures in the extant semi-aquatic forms would represent intermediate stages (Salisbury & Frey, 2001;Molnar et al., 2015). The presence of one pair of parasagittal rows of oval osteoderms is considered a plesiomorphic state for Baurusuchidae, as all specimens previously described with postcranial remains exhibit this pattern (Nascimento & Zaher, 2010;Vasconcellos & Carvalho, 2010;Araújo-Júnior & Marinho, 2013;Godoy et al., 2014). The osteoderms of these forms (e.g., Aplestosuchus sordidus) barely imbricate and are not sutured to their counterparts, which might represent an intermediate condition towards the total lack of osteoderms seen in P. sera. The phylogenetic position of P. sera among Pissarrachampsinae, as well as its smaller size when compared to Baurusuchinae, lead to two possible underlying factors for the absence of body armor in this taxon. It could be assigned as a synapomorphy of Pissarrachampsinae and interpreted as a historical factor, also implying the absence in other members of the clade, for which we still do not have information (Campinasuchus dinizi and Wargosuchus australis). Alternatively, if the absence of osteoderms is confirmed in the other smaller and early-diverging taxa, Cynodontosuchus rothi (Woodward, 1896) and Gondwanasuchus scabrosus (Marinho et al., 2013), this condition could be linked to the reduced size of the taxa.
Yet, in both scenarios, the complete absence of osteoderms in P. sera and the reduction of the body armor in other baurusuchids had biomechanical implications, with the osteoderms in other baurusuchids possibly playing a diminutive role in the bracing system and in the sustained terrestrial locomotion of these animals. This is different from what is inferred for other terrestrial Crocodylomorpha such as ''sphenosuchians'' and the peirosaurids, in which the osteoderms played an important role in the bracing system and sustained erect locomotion (Salisbury & Frey, 2001;Molnar et al., 2015;Tavares, Ricardi-Branco & Carvalho, 2015). One exception to the general pattern is the absence of osteoderms in the ''sphenosuchian'' Junggarsuchus sloani (Clark et al., 2004). This assertion is supported by the reduced transverse process and the verticalized zygapophyses which imply a bracing system not compatible to the extant forms (Salisbury & Frey, 2001). The preserved vertebrae in P. sera belong to caudal-dorsal postion therefore not overlapping the more cranial vertebrae preserved in Junggarsuchus sloani (Clark et al., 2004). However, the vertebrae of P. sera also have more verticalized zygapophyses suggesting reduced undulating lateral movements in both taxa. On the other hand, the transverse process preserved in P. sera is expanded and more similar to the extant forms than to Junggarsuchus sloani (Salisbury & Frey, 2001;Clark et al., 2004;Molnar et al., 2015). An expanded transverse process is also present in caudal-dorsal vertebrae of metriorhynchids (Young et al., 2013;Molnar et al., 2015). Accordingly, there is no perfect correlation between the occurrence of expanded transverse process and presence of osteoderms in crocodyliforms. In light of the evidence, we suggest that Baurusuchidae in general, and P. sera in particular, acquired a unique bracing system with little or no participation of the osteoderms in the sustained erect locomotion.

Phylogenetic analysis and the significance of postcranial characters in Crocodyliformes phylogeny
Here, for the first time, the postcranial data for Pissarrachampsa sera was included in a phylogenetic analysis. This resulted in scoring a total of 34 additional characters (see the Supplemental Information) for the taxon in the data matrix presented by Leardi, Fiorelli & Gasparini (2015), which is the most recent work including a substantial amount of postcranial characters. The resulting data matrix (439 characters and 111 taxa) was analysed in TNT (Goloboff, Farris & Nixon, 2008a;Goloboff, Farris & Nixon, 2008b) via heuristic searches under the following parameters: 10,000 replicates of Wagner Trees, hold 10, TBR (tree bi-section and reconnection) for branch swapping, and collapse of zero length branches according to ''rule 1'' of TNT (min.length = 0). The result of our analysis (Supplemental Information) was exactly that presented by Leardi, Fiorelli & Gasparini (2015), and all the clades are supported by the same set of synapomorphies as in the original study.
We also conducted exploratory analyses to investigate the significance of the postcranial anatomy for the phylogenetic relationships of crocodyliforms based on the data matrix used in this study. We created two subsets of the original matrix, one using only cranial characters (315 characters), and another solely with postcranial characters (124 characters). As some of the taxa in this dataset do not have either cranial or post-cranial data, we performed an extra ''control analysis'' including only taxa for which elements of both subsets of the skeleton are scored. This ''control analysis'' was performed to test whether simply removing taxa caused an impact on the overall relationships between taxa. A total of 39 taxa (all from the ingroup) were excluded following this criteria (Supplemental Information), and the 72 remaining taxa were used in the two exploratory analyses.  Montefeltro & Langer (2012) and Bronzati, Montefeltro & Langer (2015).

Figure 14
Strict consensus tree of the analysis based only on cranial characters. Name of clades between quotes indicates that their inclusivity differs from those of the ''control analysis.'' Clade with the node marked by a square (Sebecia) represents those not present in the ''control analysis.'' Silhouettes of representative crocodylomorphs from Bronzati, Montefeltro & Langer (2012) and Bronzati, Montefeltro & Langer (2015).
In the reduced strict consensus (Fig. 15), Notosuchia is recovered with a similar taxonomic content as in the original analysis (i.e., including peirosaurids, uruguaysuchids and ziphosuchians). However, the relationship between peirosaurids and uruguaysuchids, as well as among some other notosuchians, differ from the original results (Leardi, Fiorelli & Gasparini, 2015). Yet, the importance of postcranial morphology to support the affinities Figure 15 Reduced strict consensus tree of the analysis based only on postcranial characters after the exclusion of very unstable taxa. Clades identified with a white circle represent informal clades. Taxa marked with * have a seemingly anomalous position within each informal clade recovered. Silhouettes of representative crocodylomorphs from Bronzati, Montefeltro & Langer (2012) and Bronzati, Montefeltro & Langer (2015). of peirosaurids to notosuchians is strengthened, following previous evidences presented by Pol et al. (2012) andPol et al. (2014). Also, the presence of a monophyletic Notosuchia illustrates the peculiarity of the notosuchian postcranial anatomy, which could be related to the emergence of a new terrestrial lifestyle, different from other terrestrial crocodyliforms, such as the ''protosuchians.'' Further, the results of the analyses using only the postcranial information show that some ''protosuchians'' are found together with the notosuchians, in a clade with only terrestrial forms (the only exception being Leidyosuchus and the affinity of this taxa to the terrestrial forms is derived from characters based on osteoderm anatomy). The Thalattosuchia clade is also recovered in this analysis, illustrating the peculiar postcranial anatomy of these taxa linked to a fully aquatic lifestyle. Another clade recovered includes semi-aquatic crocodyliforms (the only exception being Shamosuchus), including goniopholidids and eusuchians, but their relations largely deviate from the ''control analysis.'' Overall, the results of these exploratory analyses indicate that crocodyliform relationships are strongly determined by skull characters. The postcranium has its importance in defining some relationships (i.e., those that appear in the control and original analyses but not in the analysis with cranial characters only), such as the affinity of peirosaurids and uruguaysuchids to Notosuchia, the monophyly of sebecosuchia (in the context of the original dataset used here). However, the general arrangement is still determined by characters related to the skull.
Finally, we do not consider that the results presented here reflect the inability of postcranial data to illustrate the evolutionary history of the group. Indeed, we consider that this is influenced by historical factors associated with the study of fossil crocodyliforms. Descriptions are usually based on skulls; postcranial elements are neglected, sometimes never described or mentioned in the descriptive works. However, the postcranium may play a bigger role in phylogenetic studies, as Crocodyliformes range from fully terrestrial animals to semi-aquatic and fully marine forms, and this diversity in lifestyle leads to different postcranial morphologies (e.g., Riff & Kellner, 2011;Molnar et al., 2015). Indeed, our exploratory analysis performed only with postcranial characters recovered three clades mainly representative of three different lifestyles (a ''terrestrial'' clade, a ''semi-aquatic'' clade, and a ''marine'' clade). However, the different homoplasy indexes show that this grouping is probably not a result of convergent events. The Rescaled Consistency Index (RCI-Farris, 1989) for the analysis with postcranial characters is 0.37, higher than those for the analyses with cranial characters (0.28), the control analysis (0.28), or the original analysis (0.22). A direct comparison of these values might be misleading, as different datasets exhibit particularities that could influence the results. For example, the higher RCI value for the postcranial dataset could result from the high percentage of missing data, as data of this nature cannot be homoplasious (71% in the postcranial dataset, against 37% in the cranial dataset, 47% in the control dataset, and 55% in the original dataset). On the other hand, this great number of missing data in the postcranial data set also suggests that there is still much to explore on the postcranial anatomy of Crocodyliformes, as the amount of missing data is not only related to the absence of preserved materials but also because studies describing postcranium are scarce. In this way, future work, describing more postcranial elements and proposing more characters based on this type of data will show if the phylogeny of Crocodyliformes is truly ''skull-based'' or merely ''skull-biased.''

CONCLUSIONS
The study of the postcranial skeleton of Pissarrachampsa sera allowed the recognition of some exclusive features of this taxon in the context of Baurusuchidae, such as the short and sharp crest at the craniolateral margin of the distal tibial expansion, the raised and proximodistally elongated iliofibularis trochanter of the fibula, and the more proximally placed contact between the fibular distal hook and the tibia. Also, some features related to a terrestrial lifestyle were identified, as the reduced interosseous space between both radio-ulna and tibia-fibula, the tubercle in the lateral surface of the ischium, as well as a well-protruded medial facet and a well-excavated fossa flexoria in the tibia.
An important feature is the complete absence of osteoderms in Pissarrachampsa sera, the first suggested for a terrestrial crocodyliform. This complete loss of body armor was previously known only for metriorhynchids, which have extreme adaptations for a fully marine habit. In this scenario, osteoderms probably played a minor role in locomotion of terrestrial baurusuchids, with their complete absence in Pissarrachampsa sera representing the endpoint of this trend in the group. Further, the body size and mass estimations indicate that P. sera was a large predator in the terrestrial ecosystems of the Bauru Group, but it is unlikely that it fed on adult sauropods also present at this stratigraphic unit.
Finally, our exploratory phylogenetic analyses indicate that, at least for the matrix used in this study, crocodyliform relationships are determined primarily by skull characters. However, this is more likely a consequence of the high percentage of missing data in the postcranial data set and not of the inability of this data to reflect the evolutionary history of Crocodyliformes.

AMNH
American Museum of Natural History, New York, USA.