Early crocodylomorph increases top tier predator diversity during rise of dinosaurs

Triassic predatory guild evolution reflects a period of ecological flux spurred by the catastrophic end-Permian mass extinction and terminating with the global ecological dominance of dinosaurs in the early Jurassic. In responding to this dynamic ecospace, terrestrial predator diversity attained new levels, prompting unique trophic webs with a seeming overabundance of carnivorous taxa and the evolution of entirely new predatory clades. Key among these was Crocodylomorpha, the largest living reptiles and only one of two archosaurian lineages that survive to the present day. In contrast to their existing role as top, semi-aquatic predators, the earliest crocodylomorphs were generally small-bodied, terrestrial faunivores, occupying subsidiary (meso) predator roles. Here we describe Carnufex carolinensis a new, unexpectedly large-bodied taxon with a slender and ornamented skull from the Carnian Pekin Formation (~231 Ma), representing one of the oldest and earliest diverging crocodylomorphs described to date. Carnufex bridges a problematic gap in the early evolution of pseudosuchians by spanning key transitions in bauplan evolution and body mass near the origin of Crocodylomorpha. With a skull length of >50 cm, the new taxon documents a rare instance of crocodylomorphs ascending to top-tier predator guilds in the equatorial regions of Pangea prior to the dominance of dinosaurs.

Diagnosis. Large-bodied (,3 m) crocodylomorph distinguished by the following features (loricatan autapomorphies denoted by an asterisk): six premaxillary teeth*; horizontally directed maxillary process of premaxilla; elongate, subtriangular antorbital fenestra (length to height ratio ,2.3); caudodorsally trending lateral ridge on maxilla terminates at margin of antorbital fenestra*; caudal process of maxilla rostrally pinched, minimum dorsoventral height at rostralmost corner of antorbital fenestra*; jugal with ornamented lateral boss*; caudally deep antorbital fossa, with anteriorly directed flange extending from rostral margin of lacrimal*; caudal margin of antorbital fossa vertically oriented (caudodorsal corner directly dorsal to caudoventral corner)*; antorbital fossa more than twice the estimated area of the orbit*; bifurcated caudal process of jugal bearing a small caudodorsally directed flange*; small, sub-conical, medial process of articular; pronounced crainocaudally oriented ridge on caudal aspect of lateral surface of angular; ectepicondylar crest proximal to the radial condyle of the humerus.

Description. All neurocentral sutures remain open in NCSM 21558
indicating a skeletally immature individual 13 . The skull is rostrally elongate (estimated minimum length . 50 cm) and lightly built (Fig. 1a). The maxillary process of the premaxilla projects horizontally; it is subequal in length and parallel with, the alveolar margin (Fig. 1b). The premaxilla bears six premaxillary teeth (Fig. S9b), and a subnarial notch along the caudoventral margin of the tooth row (Fig. 1b, S1) as in Dromicosuchus 14 NCSM 13733 (formerly UNC 15574), and the early crocodylomorph CM 29894. The palatal process exhibits a rostral palatal foramen (for the fourth dentary tooth 15 ).
A well-defined, rugose lateral ridge on the jugal process of the maxilla rises sharply to terminate at the ventral margin of the rostral antorbital fenestra (Fig. 1c), a condition otherwise undocumented in loricatans. A rugose, rostrally-oriented ridge on the rostral margin of the lacrimal bears a subtriangular prong, forming a keyhole shaped caudal margin of the antorbital fossa (Fig. 1d, S3). The descending process of the lacrimal widens rostrocaudally to contact the expanded rostral process of the jugal near the ventral orbital margin (Fig. 1a) as in crocodylomorphs. However, unlike crocodylomorphs, the ventral portion of the orbit in Carnufex is craniocaudally compressed, a condition that more closely resembles rauisuchids. The orbital area is markedly smaller than the antorbital fossa (,50%). The caudal process of the jugal is bifurcated (Fig. 1e, S4), a synapomorphy of Dinosauria, also present in Proterosuchus 18 .
The angular is slender and rims an elongate external mandibular fenestra (minimum 10 cm in length). In lateral view, the caudal aspect folds into a pronounced ridge ( Fig. 1f. S5) as in Junggarsuchus 19 , likely representing the insertion point for the m. pterygoideus ventralis 20 . The articular bears a saddle-shaped glenoid, as in crocodylomorphs generally. A ventromedial process of the articular is present, yet reduced, differing from the tongue-like condition of other loricatans 21 . The caudodorsal surface of the caudal process of the articular is concave as in Dromicosuchus 14 and Protosuchus richardsoni and bears a dorsomedial projection as in other basal crocodylomorphs 14,16 . In Carnufex, this projection is separated from the glenoid by a deep groove (Fig. 1g, S6) that is otherwise absent in crocodylomorphs more closely related to Crocodyliformes 18 .
Premaxillary tooth crowns are elongate, serrated, and slightly recurved, whereas the caudal maxillary tooth is serrated on mesial and distal carinae, stout and blade-like, with a weakly convex distal carina. All cranial elements except the articular are ornamented. Anastomosing pits and grooves are most pronounced on the jugal, where they form a rounded tuberosity (Fig. 1e), and on the lacrimal, where they coalesce into a rugose crest on the caudodorsal margin of the antorbital fossa (Fig. 1d). The large bodied, crocodylomorph Redondavenator also exhibits pronounced cranial ornamentation 11 , as do some large-bodied rauisuchids (e.g., Postosuchus 22 ). Ornamentation is weak to absent in small-bodied basal crocodylomorphs (e.g., Sphenosuchus 16 , Dromicosuchus 14 ) suggesting a possible correlation with body size.

Discussion
A comprehensive phylogenetic analysis of Archosauria including 79 taxa and 413 characters 7 posits Carnufex at the base of Crocodylomorpha in an unresolved polytomy with the skeletally immature postcranial skeleton CM 73372 (Fig. 2a, S9). We also provide the first phylogenetic placement of the Rhaetian pseudosuchian Redondavenator, substantiating this taxon as the largest Triassic crocodylomorph yet described 11 (Fig. S9). Carnufex and Redondavenator expand the diversity of top tier terrestrial predator guilds in the Late Triassic to at least five distinct archosaur clades and document vast overlap in body size between contemporary dinosaurs and crocodylomorphs (Fig. 2b). In contrast, the Triassic-Jurassic transition marks a shift to dichotomous body mass distributions between terrestrial members of these two clades, and a loss of top tier crocodylomorph diversity after the end-Triassic extinction (ETE) (Fig. 2b).
The Carnian-aged Pekin Formation preserves some of the oldest Triassic archosaur assemblages in North America and brings to bear unique biodiversity data on the composition of top predator guilds preceding the appearance of theropod dinosaurs on the continent 18,23 . To date, tetrapods of the Pekin Formation are well sampled, and capture a diverse assemblage comprised of dicynodontians 24 , traversodontid cynodontians 25 , aetosaurians 26 , and two species of crocodylomorphs (Carnufex, and a new small bodied taxon 27 with a femur length {FL} 5 133 mm) (Fig. 2c). With an estimated immature FL of 353-440 mm, Carnufex is the largest terrestrial predator in the Pekin Formation (Fig. 2d), vastly exceeding the body size of the earliest North American theropod dinosaurs (FL 174-265 mm) (Fig. 2b).
Early concepts of faunal homogenization across Pangea are unsupported by recent studies, which instead document latitudinally arrayed, paleoclimatic faunal provinces across the supercontinent 12,28 , although this pattern is likely restricted to assemblages preceding the ETE 29 . Predatory guild evolution in the Triassic was equally complex, with recent research supporting a diachronous replacement of the leading terrestrial predators-pseudosuchianswith theropod dinosaurs between proto-Laurasian and Gondwanan landmasses 23,30 . Our chronostratigraphic plots of body size (Fig. 2b) and paleogeographic region (Fig. 2d) support both hypotheses. Whereas Carnian terrestrial predator guilds in southern hemisphere faunas were exploited in part by large bodied theropod dinosaurs, northern and equatorial faunas of similar age have yet to yield definitive theropod remains 5,23,30 and appear instead to have been evolutionary centers for large bodied terrestrial crocodylomorphs 11 , as exemplified by Carnufex in the Carnian and subsequently by Redondavenator in the Rhaetian. The loss of large-bodied crocodylomorphs nearing the ETE may have spurred mesopredator release 31 or opportunistic invasion scenarios 30 , whereby smaller-bodied theropods subsequently assumed apex predator roles in paleoequatorial regions of proto-Laurasia.

Methods
Body size. We evaluated body size using the widely accepted proxy of femur length (FL) [32][33][34] . Measurements of FL were derived primarily from recent archosaurian datasets 32,35 . Select taxa relevant to our analyses lack a femur and required FL estimation. We derived scaling equation (1) for predicting FL by applying OLS regression analysis on bivariate plots of FL/humeral length (HL) (Fig. S10) and equation (2) using FL/skull length (SKL) (Fig. S11) for a variety of loricatans for which data were available (Table S1). We also performed OLS regressions on bivariate plots of log e HL against SKL (Fig. S12) to examine the reliability of our estimates against independent scaling relationships (equation [3]). Data was log transformed (log e). Scaling equations had high coefficients of determination (R 2 ) ranging from 0.94-0.97; however, these are influenced by low sample size (n 5 12) 36 We used scaling equations to estimate FL of Trialestes romeri based on a 220 mm measurement of SKL provided by Reig 37 ; Redondavenator quayensis based on a 600 mm SKL estimate provided by Nesbitt et al. 11 ; Pseudhesperosuchus jachaleri based on an estimated skull length of 130 mm 38 41 ), whereas, the estimated SKL of Z. rougieri closely approximates that of Cryolophosaurus ellioti (FL 769 mm 42 ). We did not estimate FL of the Early Jurassic theropods Dracovenator regenti and Lophostropheus airelensis because of the fragmentary nature of the remains (all lack appendicular elements and complete skulls) and because our conclusions revolve around Late Triassic fauna in the former instance. However, we note that published estimates of D. regenti size (5.5-6.5 m in length 42 ) would place this taxon with the range of body size already captured in Figure 2, adding no new data to our results. Estimates of FL generated for these taxa are listed in Table S2 and marked with an asterisk. We only provide a general estimate of FL for the ornithosuchid Venaticosuchus as approximating that of the ornithosuchid Riojasuchus, because the skull of this taxon is highly fragmentary and appears to have been close in size to Riojasuchus 9 .
We derived a FL range for Carnufex carolinensis of 354-441 mm, based on a measured HL of 207.7 mm and estimated minimum SKL of 500 mm, respectively. Given the large FL range produced from these two variables, we further explored FL for Carnufex by testing how accurately we could approximate known HL and estimated SKL values using the relevant scaling equation derived from our loricatan dataset. Our estimation of Carnufex HL using a minimum SKL length of 500 mm was 257 mm, 24% larger than our actual measurement on the preserved humerus of NCSM 21558. Conversely, our estimate of Carnufex SKL using the actual value for HL of CNSM 21558 was 385 mm, far shorter than the portion of the skull preserved (450 mm). These data indicate that the humeral to skull proportions of Carnufex are not a good fit to the regression, i.e., either the humerus of Carnufex is unusually short, or the skull unusually long, or both variations are compounded. Therefore, we present FL estimates derived both from the HL and SKL here. We note that NCSM 21558 is skeletally immature, having open neurocentral sutures across the cervical and dorsal series minimally. This immaturity plus our use of a minimum estimate for skull length yields a conservative range of 354-441 mm for FL. We expect the FL of a somatically mature Carnufex would fall within the upper values of our current estimates or perhaps well above. Our gross estimate of the body length of a skeletally immature Carnufex (,3 m) is based on comparative skeletal ratios in the closely related Dromicosuchus (NCSM 13731) and the nearly complete basal crocodylomorph NCSM 21722 27 .
Ecological inferences. Given that autecology cannot be observed for extinct taxa, paleontologists generally rely on the presence of ecomorphological traits to infer dietary inferences [43][44][45][46] and construct trophic networks 2,48,49 . We follow Mitchell & Makovicky 49 in assigning the extinct archosaurs to guilds (e.g., top-tier predator) based on body size, inferred diet, and habitat (e.g., terrestrial, semi-aquatic, aquatic). These ecological factors were taken from the published literature. We followed multiple authors in considering taxa of the following clades to represent the diversity of carnivorous, terrestrial, Triassic pseudosuchians: loricatans (Rauisuchidae 1 Crocodylomorpha) 6,10 , gracilisuchids 7 , poposauroids 50 , & ornithosuchids 9 ; and in assigning early theropods to this guild 5 . Dietary inferences for some Triassic pseudosuchians are ambiguous (e.g., Effigia 51 ); however, without quantitative analyses testing analogous ecomorphological traits in these taxa 45,47 , we include them as carnivores in keeping with the apparent dominant trophic habit of their clade, considering this a conservative approach. Currently known diversity places Carnufex and Redondavenator as the largest, terrestrial carnivores within their respective assemblages 11 , which generally denotes apex predator status 30 . However, given the potential of sampling biases (e.g., no rauishuchids recovered from the Pekin Formation) and the nuances of extant predator interaction 6,30 , we refrain from restricting these taxa to apex predator roles. Rather, we adopt a more conservative approach that allows for incomplete sampling of large-bodied carnivores, by considering Carnufex and Redondavenator to be minimally, components of top-tier predator guilds within Triassic faunas.
Phylogenetic protocol. We examined the evolutionary relationships of Carnufex and Redondavenator by inclusion in the recent, comprehensive analysis of archosaurs published by Butler et al. 7 , which is an expansion of Nesbitt 18 . The analysis includes 79 archosaurs and 413 characters. We followed Butler et al. 7 in a priori exclusion of the operational taxonomic units: Archosaurus rossicus, Prestosuchus chiniquensis, UFRGS 0156 T, UFRGS 0152 T, Lewisuchus admixtus, and Pseudolagosuchus major; and in designation of the following characters as additive: 32, 52, 75, 121, 137, 139, 156, 168, 188, 223, 247, 258, 269, 271, 291, 297, 328, 356, 399, and 413. Data coding, character tracing and tree manipulation/visualization were carried out using Mesquite ver. 2.75 52 . Phylogenetic analyses were executed in the program TNT 53 . We conducted heuristic searches on Wagner trees using TBR (tree bisectionreconnection) branch-swapping with 1,000 random addition sequences holding 10 trees per replicate, continuing subsequent TBR swapping on all stored minimum length trees (90 most parsimonious trees, TL 1,320). We assessed results using strict and reduced consensus methods and Bremer support values 54 . Ambiguous nodes were collapsed following Rule 1 of Coddington and Scharff 55 . Maximum agreement subtrees 56 were calculated in TNT and used to identify labile taxa and common topology among all MPTs.
In this analysis we recover three unambiguous synapomorphies of Crocodylomorpha 1 Carnufex: a sub narial gap (char. 11 state 1); an elongate lacrimal reaching the ventral aspect of the orbit (char. 39, state 1); and loss of fin-like hyposphen-hypantrum articulations in the vertebral series (char. 195, state 0). However, Carnufex clearly exhibits a mosaic bauplan that spans lightly built, cursorial crocodylomorphs and their large-bodied, robust sister taxa, rauisuchids. As a result, Carnufex also shares several skeletal features characteristic of rauisuchids including a bulbous longitudinal ridge on the maxilla (char. 26, state 2); non-tapering dorsal process of the maxilla (char. 29, state 1); as well as retaining some synapomorphies of Loricata, lost in crocodylomorphs more closely related to Alligator than Carnufex, including a distinct groove caudal to the glenoid fossa on the articular (char. 156, state 1); and a tall, narrow orbit (char. 142, state 1). Carnufex also possesses some traits convergent with theropod dinosaurs such as a bifurcated caudal process of the jugal (char. 71, state 3) and a dorsoventrally expanded caudal process of the jugal, also present in Revueltosaurus and some archosauromorphs (char. 27, state 2). We recover Carnufex as an unequivocal crocodylomorph in our analysis. Two steps are required to move Carnufex out of Crocodylomorpha (Fig. S9). The remaining nodes within Crocodylomorpha are supported by Bremer values of 1 (Fig. S9). The mosaic morphology exemplified by this taxon and its basal phylogenetic and stratigraphic position yields critical insight into the step-wise appearance of the crocodylomorph bauplan.
Redondavenator quayensis was described as a large bodied crocodylomorph 11 , yet has not been tested in a phylogenetic context. Although fragmentary (only 8% of characters can be coded) we included this taxon to substantiate this placement quantitatively. Our analysis posits R. quayensis as sister-taxon to Sphenosuchus acutus based on the shared presence of an elongated maxillary process of the premaxilla (char 2, state 1). Although the maxillary process of the premaxilla is incomplete in Redondavenator, we find that all taxa possessing five or more maxillary alveoli anterior to the antorbital fenestra also possess a maxilla in which the portion rostral to the antorbital fenestra is longer than the posterior process (char 2, state 1) (SD per obs.). This correlation may not prove exhaustive, given that Redondavenator only includes the anterior portion of the skull; however, we include coded Redondavenator for this trait as a testable hypothesis.
The skeletally immature postcranial skeleton CM 73372 has been variously interpreted as Postosuchus 28,57,58 and a Hesperosuchus-like basal crocodylomorph 18 . Our analysis is unable to resolve the relationship between Carnufex and CM 73372, recovering these taxa in a polytomy with a clade consisting of all remaining crocodylomorphs. Carnufex is represented predominantly by cranial elements and CM 73372 consists entirely of postcranial elements, therefore there is little overlapping data between these species to aid in phylogenetic resolution.
3D visualization and reconstruction. Elements of the skull and postcranial skeleton of Carnufex were scanned using a Creaform EXAscan TM high-resolution (0.050 mm) handheld surface scanner. Scans were captured in 1.0-0.02 mm resolution using VXelements 3D data acquisition software. Post processing and generation of 3D PDFs were accomplished in Geomagic StudioH. We used Autodesk Maya 2014 to produce a composite three-dimensional model of the skull using individual element scans of Carnufex (NCSM 21623), supplemented with scaled surface scans of cranial elements of the closely related crocodylomorph Dromicosuchus, and Junggarsuchus, performed using the same protocol as Carnufex. Skull width was modeled using the relatively undeformed skull roof of Dromicosuchus. The rostral dentary and braincase were modeled using scans of Junggarsuchus, whereas the rostral maxilla, quadrate, frontals, parietals, squamosals, nasals, and maxillary dentition were modeled from Dromicosuchus. The circumnarial region of the premaxilla, prefrontal, quadratojugal and remainder of the mandible were generated as de novo objects manipulated to reflect estimated proportions. Original three-dimensional scans of the skeletal elements of Carnufex are provided as individual 3D PDFs as Supplementary  Information (Figs. S1-8).
Methods summary. This published work and the nomenclatural acts it contains have been registered in ZooBank, the proposed online registration system for the International Code of Zoological Nomenclature (ICZN). The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix 'http://zoobank.org/'. The LSID for this publication is: urn:lsid:zoobank.org:pub:129E5681-0D2E-4A7D-BC04-A3E624151BF7.