The Late Triassic Ischigualasto Formation at Cerro Las Lajas (La Rioja, Argentina): fossil tetrapods, high-resolution chronostratigraphy, and faunal correlations

Present knowledge of Late Triassic tetrapod evolution, including the rise of dinosaurs, relies heavily on the fossil-rich continental deposits of South America, their precise depositional histories and correlations. We report on an extended succession of the Ischigualasto Formation exposed in the Hoyada del Cerro Las Lajas (La Rioja, Argentina), where more than 100 tetrapod fossils were newly collected, augmented by historical finds such as the ornithosuchid Venaticosuchus rusconii and the putative ornithischian Pisanosaurus mertii. Detailed lithostratigraphy combined with high-precision U–Pb geochronology from three intercalated tuffs are used to construct a robust Bayesian age model for the formation, constraining its deposition between 230.2 ± 1.9 Ma and 221.4 ± 1.2 Ma, and its fossil-bearing interval to 229.20 + 0.11/− 0.15–226.85 + 1.45/− 2.01 Ma. The latter is divided into a lower Hyperodapedon and an upper Teyumbaita biozones, based on the ranges of the eponymous rhynchosaurs, allowing biostratigraphic correlations to elsewhere in the Ischigualasto-Villa Unión Basin, as well as to the Paraná Basin in Brazil. The temporally calibrated Ischigualasto biostratigraphy suggests the persistence of rhynchosaur-dominated faunas into the earliest Norian. Our ca. 229 Ma age assignment to Pi. mertii partially fills the ghost lineage between younger ornithischian records and the oldest known saurischians at ca. 233 Ma.

www.nature.com/scientificreports/ With one of the richest land biotas recorded worldwide, the Ischigualasto Formation of north-western Argentina represents a unique "window" into Late Triassic biodiversity and evolution. This stratigraphic unit is well known from the Ischigualasto Provincial Park (IPP), San Juan Province, with a fossil record composed of plants, fishes, and most of the known tetrapod groups of the time, i.e., temnospondyls, rhynchosaurs, archosauriforms (including dinosaurs), dicynodonts, and cynodonts [1][2][3] . Radioisotopic dates of various vintages have given the Ischigualasto fauna a temporal context, elevating its global significance in understanding the Triassic land ecosystems, as well as the early evolution of dinosaurs 3 . Nevertheless, exposures of the Ischigualasto Formation outside the IPP have only been briefly explored, delivering only subordinate fossil records 3 . One exception is the site known as Hoyada del Cerro Las Lajas [4][5][6] in La Rioja Province, where the northernmost known outcrops of the formation are exposed ( Fig. 1; see also fig. 1 in Baczko et al. 7 ). Explored by several expeditions starting in the early sixties (see Historical background and motivation in the Supplementary Information), the fossil record of the area appears meagre compared to that of the IPP and it has been described as "a poorly fossiliferous outcrop" (p. 20 in Martínez et al. 3 ), but includes key specimens, such as the holotypes of the ornithosuchid Venaticosuchus rusconii and the probable ornithischian Pisanosaurus mertii.
Aiming to expand the fossil collections of the Hoyada del Cerro Las Lajas and to investigate the chronostratigraphic context of previous fossil collections, our team explored the Cerro Las Lajas area in the course of four expeditions from 2013 to 2019. Here, we report on more than 100 new tetrapod fossil specimens collected form the Ischigualasto Formation at Cerro Las Lajas. Detail stratigraphy of its over 1,000 m-thick succession, integrated with high-precision U-Pb zircon geochronology of three interlayered tuffs, provide a high-resolution chronostratigraphic framework for the Ischigualasto Formation in the the Hoyada del Cerro Las Lajas. In this context, we discuss palaeobiologic aspects of the Ischigualasto fauna and their implications for Late Triassic tetrapod evolution.  Black bars are individual zircon analyses used in weighted mean age calculation. Horizontal shaded band represents the weighted mean 206 Pb/ 238 U date and its 95% confidence level internal uncertainty. Arrows point to older (detrital) analyses that fall outside the plot area. See Table 1 and Table S2 for complete U-Pb data and for details of calculated dates and their uncertianites. Table 1. Summary of calculated U-Pb ages and their uncertainties. Notes: Latitude/Longitude relative to WGS84 datum. X internal (analytical) uncertainty in the absence of all external or systematic errors. Y incorporates the U-Pb tracer calibration error. Z includes X and Y, as well as the uranium decay constant errors. MSWD mean square of weighted deviates. n number of analyses included in the calculated weighted mean date, out of a total number of # analyses.    Abbreviations: adf, adductor fossa; aoc, anguli oris crest; ap, acromial process; ar, articular; as, astragalus; ca, calcaneum; ce, centrale; cen, centra; co, coronoid; ct, calcaneal tuber; dcb, dentary cutting blade; dp, diapophysis; dt, dentary; dtc, deltopectoral crest; ect, ectepicondyle; ent, entepicondyle; ep, epipophysis; f.fi, facet for fibula; f.ti, facet for tibia; gf, glenoid fossa; itf, infratemporal fenestra; ju, jugal; lg, longitudinal groove; ltb, lower temporal bar; LTBA, lateral tooth bearing area; Mg, Meckelian groove; mx, maxilla; MTBA, medial tooth bearing area; nag, non-articular gap; ns, neural spine; ob, border of orbit; po, postorbital; poz, poszygapophysis; pra, prearticular; prz, prezygapophysis; pt, pterygoid; qj, quadratojugal; sa, surangular; sp, splenial; sq, squamosal; sy, symphysis; tc, tibial crest; tp, transverse process; tu, tuberosity. Scale bar: 2 cm.
Scientific RepoRtS | (2020) 10:12782 | https://doi.org/10.1038/s41598-020-67854-1 www.nature.com/scientificreports/ Scotia 37 , and other specimens of H. sanjuanensis 38 . By contrast, the single longitudinal groove is approximately centred on the tooth plate in H. gordoni 39 and an indeterminate hyperodapedontine from the Ischigualasto Formation 40 , whereas there are two longitudinal grooves in H. huenei 29 and Teyumbaita sulcognathus 41 . The lateral tooth bearing area (LTBA) has five longitudinal tooth rows and the crowns in the two medialmost rows (L1 and L2) are worn to the root, resembling the condition of most hyperodapedontines (e.g. H. mariensis, UFRGS-PV 0149T, 0408T; H. huxleyi 32 ). By contrast, only one or two longitudinal tooth rows are present in the LTBA of North American hyperodapedontines (Wyoming form 42 ; Nova Scotia form 37 ), Te. sulcognathus 41 , and I. genovefae 35 . The number of rows in the medial tooth bearing area (MTBA) and presence of lingual teeth cannot be determined in CRILAR-Pv 584 because of damage.
The dentary forms more than half of the hemimandible, has a tapering anterior end, and does not form part of the mandibular symphysis, which are character-states retained by all hyperodapedontines [43][44][45] (Fig. 7a, e). The dentary has a single, transversely thin cutting blade with one row of mesiodistally compressed teeth. There is no lingual teeth and no medially bulged area in the dentaries of CRILAR-Pv 583, 584, 646, 650, as typical of H. sanjuanensis 38 and also reported for an unnamed hyperodapedontine from Nova Scotia 37 . By contrast, all other hyperodapedontines have lingual teeth on the dentary 34,38 . A well-developed coronoid prominence is formed by the dentary, surangular, and coronoid bones. A deep and lateroventrally opened posterior surangular foramen is located at level with the glenoid fossa. The retroarticular process is short and its dorsal surface is damaged in CRILAR-Pv 583.
The anterior dorsal vertebra possesses a spool-shaped, taller than long centrum that lacks a ventral keel (Fig. 7b, c). The neural spine is restricted to the posterior two-thirds of the neural arch and does not extend between the bases of the prezygapophyses. The tibia has transversely expanded proximal and distal ends (Fig. 7d). The shaft possesses a well-developed, proximolaterally to distomedially oriented tibial crest on its anterior surface, as occurs in other rhynchosaurids 43,46,47 . The distal articular surface of the bone is transversely convex and slants proximolaterally. The proximal row of tarsals is composed of a centrale, astragalus, and calcaneum, as occurs in other rhynchosaurs 48 (Fig. 7f). The centrale is not fused to the astragalus and its proximal surface extensively contributes to the tibial facet. The proximal surface of the astragalus has tibial and fibular facets separated from one another by a non-articulating gap. The posterior surface lacks a posterior groove and the autapomorphic transverse boss present in Te. sulcognathus 46 . The medial half of the proximal surface of the calcaneum is occupied by the fibular facet and the lateral half of the bone is developed as a laterally projected calcaneal tuber. The calcaneal tuber is anteroposteriorly narrower than proximodistally tall and the latter axis is rotated approximately 45° from the proximodistal plane of the proximal tarsus.
Comments. These specimens can be referred to Hyperdapedontinae because of the presence of the following synapomorphies 34 : mandible dorsoventral depth > 0.25 times its total length (CRILAR-Pv 583); dentary with mesiodistally compressed teeth (all specimens); posteriormost dentary teeth on the posterior half of the lower jaw (CRILAR-Pv 583); and astragalus with centrale facet greater than the tibial facet (CRILAR-Pv 584, although this condition is unknown in the immediate sister-taxa to Hyperodapedontinae). In addition, CRILAR-Pv 584 can be included in the Hyperodapedon clade 44 because of the presence of a maxillary tooth plate with more than two tooth rows in the LTBA and four or more tooth rows on its anterior half. Within Hyperodapedontinae, these specimens can be referred to H. sanjuanensis because the absence of lingual teeth in the dentary has been considered an autapomorphy of this species 38 . Nevertheless, the unnamed Nova Scotia hyperodapedontine apparently also lacks lingual dentary teeth 37 and this feature may be an apomorphy of a more inclusive clade of hyperodapedontines (i.e. H. sanjuanensis + North American forms 34 ). We preferred here to maintain this character-state as diagnostic of H. sanjuanensis until more information of the Nova Scotia hyperodapedontine is published. In any case, the maxillary tooth plate of CRILAR-Pv 584 differs from those of the Nova Scotia hyperodapedontine in the presence of a higher number of tooth rows in the LTBA and, at least for this specimen, such combination of character-states still supports its referral to H. sanjuanensis.
Material. CRILAR-Pv 585, articulated partial left side of cranium missing lacrimal, prefrontal, anterior region of palate and almost entirely the skull table; fragment of right maxilla; partial braincase; at least 14 postaxial cervical and anterior-middle dorsal vertebrae; several ribs and gastralia; both scapulae; left humerus; distal end of right humerus; and multiple indeterminate bone fragments (Fig. 7g-n).
Description. The overall morphology of the skull of CRILAR-Pv 585 resembles that of other hyperodapedontine rhynchosaurs in the presence of a ventral border of the orbit positioned dorsal to the mid-height of the infratemporal fenestra, a massive and anterodorsally-to-posteroventrally sloping jugal, and a closed lower temporal bar (e.g. I. genovefae 45 (Fig. 7g,h). The infratemporal fenestra is kidney-shaped, with a notched posterior border. This outline is a result of the strongly concave anterior margin of the ascending process of the quadratojugal, as occurs in H. huenei 29  www.nature.com/scientificreports/ The occlusal surface of the maxillary tooth plate of CRILAR-Pv 585 is divided into equally broad LTBA and MTBA by a longitudinal groove (Fig. 7h), as also occurs in H. gordoni 39 and an indeterminate hyperodapedontine from the Ischigualasto Formation 40 . By contrast, the maxillary tooth plate has a broader LTBA in all other hyperodapedontines with a single groove 34 . The maxillary tooth plate of CRILAR-Pv 585 also differs from those of Te. sulcognathus, H. huenei, and the morphotype 2 of H. tikiensis, which possess two longitudinal grooves that define a third, central tooth bearing area 29,33,41 . CRILAR-Pv 585 has four longitudinal tooth rows at the posterior end of the LTBA and three rows in the MTBA. In addition, there is a row of six lingual teeth, well-spaced from one another and located on the medial surface of the maxilla, dorsally to the MTBA (Fig. 7k), resembling the condition in H. huenei 29 and a Zimbabwean hyperodapedontine 36 . However, CRILAR-Pv 585 differs from these two forms in the presence of lingual tooth crowns that are mainly oriented ventrally rather than perpendicular to the occlusal surface, and from all rhynchosaurs in the presence of lingual teeth restricted to the anterior half of the tooth plate.
The jugal forms the ventral border of the orbit and bears an anterodorsally-to-posteroventrally oriented anguli oris crest that overhangs laterally the maxilla (Fig. 7g). The lateral surface of the jugal is coarsely ornamented by low ridges and bulges on its main body and striations adjacent to the orbital edge. No secondary anguli oris crest is present on the main body of the jugal, contrasting with Te. sulcognathus 41 and I. genovefae 45 . The lateral surface of the posterior process of the jugal of CRILAR-Pv 585 lacks the deep and posterodorsally well-rimmed depression located on the ventral half of the base of this process in H. huxleyi (ISIR 01), H. huenei (UFRGS-PV 0132T), and referred specimens of H. mariensis (UFRGS-PV 0149T). The posterior process of the jugal forms the entire ventral border of the infratemporal fenestra in CRILAR-Pv 585, as occurs in H. huenei 29 and Te. sulcognathus 41 . By contrast, the anterior process of the quadratojugal contributes to the ventral border of the opening in I. genovefae 45  The palatine of CRILAR-Pv 585 contacts the ectopterygoid posterolaterally and, as a result, excludes the maxilla from the border of the infraorbital foramen. The pterygoid possesses a cup-shaped, dorsomedially projected process that received the basipterygoid process of the parabasisphenoid. This facet indicates the presence of a basal articulation two times dorsoventrally taller than transversely broad, as occurs in Te. sulcognathus 41 and other species of Hyperodapedon 45 .
The basioccipital possesses a long occipital neck and basal tubera broadly separated from one another. The exoccipital contacted its counterpart on the floor of the endocranial cavity, as occurs in several other hyperodapedontines (e.g. H. huenei, UFRGS-PV 0132T; H. mariensis, UFRGS-PV 0149T; H. sanjuanensis, MACN-Pv 18185), but contrasting with the absence of such contact in Te. sulcognathus 41 . The occipital surface of the base of the paroccipital process possesses a ventrally well-defined depression on its dorsal half, resembling a condition previously reported as autapomorphic of Te. sulcognathus 41 .
The postaxial cervical vertebrae have a spool-shaped centrum that lack a ventral keel and possess a shallow depression on its dorsolateral surface (Fig. 7l). By contrast, the anterior-middle cervical vertebrae of Te. sulcognathus have a median ventral keel 46 . There is a tall, crest-shaped (i.e. conical) epipophysis on the dorsal surface of the postzygapophysis, perhaps absent only in the posteriormost cervical vertebrae (Fig. 7m). The neural spine is restricted to the posterior half of the neural arch. The centra of the anterior and middle dorsal vertebrae are generally longer and slightly more transversely compressed than the cervical centra (Fig. 7n). The anteriormiddle dorsal neural arches possess comma-shaped transverse processes in cross-section and lack laminae. The postzygapophyses lack an epipophysis.
The scapula is anteroposteriorly expanded at both the proximal and distal ends (Fig. 7i). The very base of the acromial process is thick, ridge-like and distinctly laterally raised, resembling the condition in most hyperodapedontines (e.g., H. sanjuanensis, MACN-Pv 18185; H. huxleyi, ISIR 01; H. tikiensis 33 ). By contrast, this process is sub-circular and blunt in Te. sulcognathus 46 . The scapular blade has distinctly divergent anterior and posterior margins, as occurs in most hyperodapedontines (e.g., H. huxleyi 32 ; H. mariensis, MCN 1867-PV), but the scapular blade possesses a tab-like, poorly developed posterior expansion in Te. sulcognathus (UFRGS-PV 0232T). The proximal and distal ends of the humerus are distinctly transversely expanded and their main axes rotated approximately 40° from one another (Fig. 7j). The deltopectoral crest is mainly anteriorly oriented. The distal end has a very deep, subtriangular, and concave anterior fossa and a shallower and more proximally extended posterior fossa. The lateral surface of the distal end possesses a deep longitudinal ligament groove (= ectepicondylar groove) that is anteriorly delimited by a supinator ridge, resembling the condition in other hyperodapedontines (e.g., H. tikiensis 33 ; H. huxleyi 32 ; H. gordoni 39 ; H. sanjuanensis, MACN-Pv 18185; Te. sulcognathus, UFRGS-PV 0232 T).
Comments. CRILAR-Pv 585 is identified as a hyperodapedontine rhynchosaur because of the presence of the following synapomorphies of the clade 34 : jugal without an elevated orbital rim; fully closed lower temporal bar; anguli oris crest extended onto the anterior process of the jugal, but not the maxilla; maxilla well laterally overlapped by the jugal; maxillary tooth plate with cushion-shaped LTBA; and maxillary teeth with conical and 'pyramidal' crowns. In addition, CRILAR-Pv 585 shares with other members of the Hyperodapedon clade the following synapomorphies 34 : maxillary tooth plate with more than two tooth rows in the MTBA; maxillary tooth plate with four or more tooth rows of occlusal teeth on its anterior half; parabasisphenoid with a basipterygoid process wider than long (inferred from the shape of the basal articulation on the pterygoid); and postaxial cervical postzygapophyses with crest-shaped epipophysis. Among hyperodapedontines, CRILAR-Pv 585 differs from other taxa in the presence of an autapomorphic row of ventrally oriented lingual teeth restricted to the anterior half of the maxillary tooth plate. This new species will be formally named and described in detail in a future contribution.  32,34,38,49 . By contrast, the longitudinal groove is centred on the tooth plate in H. gordoni, the above Hyperodapedon sp. nov., and an indeterminate hyperodapedontine from the Ischigualasto Formation 40 . The longitudinal groove narrows anteriorly, resembling the condition in some other hyperodapedontines (e.g. UFRGS-PV 0149T, 0408T). The longitudinal groove bows slightly laterally and is very deep, with a V-shaped cross-section. The LTBA possesses four longitudinal tooth rows and the MTBA has three rows. The presence of more lateral longitudinal tooth rows than medial ones is consistent with the condition in H. sanjuanensis (MACN-Pv 18185, MCP-PV 1693), H. mariensis (UFRGS-PV 0149T, 0408T), H. huxleyi (ISIR 01), and the holotype of H. tikiensis 33 . By contrast, the MTBA has more longitudinal tooth rows than the lateral one in H. gordoni 39 , Su. stockleyi (SAM-PK-11705), and the unnamed hyperodapedontine from Nova Scotia 37 . The preserved L1 and M1 tooth crowns are strongly worn on the walls of the longitudinal groove, exposing the root in coronal section. The teeth of both toothbearing areas are relatively small and closely packed, as occurs in most hyperodapedontines with the exception of I. genovefae 35,45 . The preserved tooth crowns of both tooth-bearing areas have a circular cross-section, but it is not possible to determine the presence of pyramidal teeth because the posterior region of the tooth plate is not preserved. The preserved portion of the medial surface of the bone lacks lingual teeth, but it not possible to determine if they were present more posteriorly or anteriorly in the tooth plate. The fragment of dentary of CRILAR-Pv 582 possesses a V-shaped cross-section as a result of the presence of a transversely thin and sharp occlusal cutting blade. It is not possible to observe teeth in the dentary fragment.
The presence of a single longitudinal groove and more than two tooth rows in the MTBA of the maxillary tooth plate of CRILAR-Pv 582 allows referring this specimen to the Hyperodapedon clade 34 . Its maxilla differs from those of hyperodapedontines with a centrally located single longitudinal groove (i.e. H. gordoni, an hyperodapedontine from the Ischigualasto Formation, and the above Hyperodapedon sp. nov.), more tooth rows in the MTBA than in the LTBA (Su. stockleyi), less than two tooth rows in the LTBA (I. genovefae, the unnamed hyperodapedontines from Nova Scotia and Wyoming) or with two longitudinal grooves (H. huenei and Te. sulcognathus). Instead, the morphology of CRILAR-Pv 582 is congruent with that of H. sanjuanensis, H. mariensis, H. huxleyi, the holotype of H. tikiensis, and the unnamed Zimbabwean hyperodapedontine. Material. CRILAR-Pv 586 (Fig. 8), partial cranium lacking most of the skull roof and right side, partial left hemimandible, a median segment of the right hemimandible, atlas, axis and third cervical vertebra, four middleposterior cervical vertebrae, a fragment of humeral shaft, a probable metacarpal lacking the distal end, eight non-ungual phalanges, and indeterminate bone fragments (Fig. 8h). CRILAR-Pv 587, partial left premaxilla, right maxilla, left nasal, basicranium, right atlantal neural arch, five postaxial cervical vertebrae, a posterior dorsal vertebra, right scapula, coracoid, clavicle, humerus, ulna and femur, proximal and distal ends of fibula, a right metacarpal probably from digit II, two non-ungual and two ungual phalanges, and several ribs (Fig. 8c,e-g,i). CRILAR-Pv 588, partial left maxilla lacking its lateral edge and anterior tip, fragment of the medial dentary crest and indeterminate bone fragments. CRILAR-Pv 595, partial skull with almost complete right side and missing the left orbital and temporal regions, braincase and post-dentary bones with exception of the angulars (Fig. 8a, b). CRILAR-Pv 642, partial left maxilla lacking most of the ascending process, anterior tip, and occlusal surface of the LTBA, and-still in the field-partial postcranium. CRILAR-Pv 643, middle third of right maxilla, six postaxial centra, ventral end of clavicle, proximal and distal ends of tibia and indeterminate bone fragments of bones of a very small-sized individual. CRILAR-Pv 645, articulated right quadrate, squamosal and paroccipital process, probable partial left postfrontal, partial right parietal and pterygoid, partial parabasisphenoid, posterior two-thirds of left dentary, articulated posterior half of right surangular, prearticular and fragment of articular, six postaxial cervical vertebrae, two articulated probable anterior dorsal vertebrae, two dorsal or anterior caudal vertebrae, right scapula and humerus, partial scapular blade of another right scapula, multiple fragments of ribs and gastralia and several indeterminate bone fragments. This specimen consists of at least two individuals found in association with CRILAR-Pv 595. CRILAR-Pv 651, partial left premaxilla and dentary, posterolateral corner of right maxilla, partial right jugal, postorbital, quadrate, pterygoid and ectopterygoid, axial centrum, at least three postaxial centra, right scapula, humerus, ulna and radius, fragment of left scapula, distal third of metacarpal or metatarsal, three non-ungual phalanges, several rib fragments, and several probable cranial and postcranial indeterminate bone fragments (Fig. 8d,j-m).
Description. The skull of Teyumbaita sp. nov. is broader than long, with a jugal representing the main component of its lateral surface, a fully closed lower temporal bar, and a dorsolaterally facing orbit (Fig. 8a,b), as occurs in other hyperodapedontines 41,44,45,50 . The nasals form a straight posterodorsal border in the single external naris, contrasting with the diagnostic notched border at the median line present in Te. sulcognathus 41 . The lateral surface of the main body of the jugal is deeply concave and lacks a secondary anguli oris crest (CRILAR-Pv 595, 651), contrasting with I. genovefae 45  www.nature.com/scientificreports/  www.nature.com/scientificreports/ base of the posterior process of one specimen (CRILAR-Pv 651) has a thick, rugose ridge that does not extend further anteriorly on the bone ( Fig. 8d: saoc). A very similar structure was interpreted as a posteriorly restricted secondary anguli oris crest in another referred specimen of Te. sulcognathus 41 . The presence of this latter ridge and the absence of a lateral depression at the base of the posterior process differs from the condition present in species of Hyperodapedon (e.g. H. sanjuanensis, MACN-Pv 18185; H. huenei 29 ; H. huxleyi, ISIR 01; H. gordoni 39 ; H. mariensis, UFRGS-PV 0149T). The anterior margin of the ascending process of the quadratojugal is slightly convex in lateral view, as occurs in Te. sulcognathus 41 , but contrasting with the concave margin of H. huenei 29 and Hyperodapedon sp. nov. (CRILAR-Pv 585).
The maxillary tooth plate of the currently largest specimen of Teyumbaita sp. nov. has a transverse width of 47.6 mm (CRILAR-Pv 587), which is ca. 84% smaller than the largest specimen of Te. sulcognathus (transverse width = 56.7 mm, UFRGS-PV 290T). The maxillary tooth plate possesses two longitudinal grooves that define three tooth bearing areas (Fig. 8g), as occurs in Te. sulcognathus 41 , H. huenei 29 , the morphotype 2 of H. tikiensis 33 , and several non-hyperodapedontine rhynchosaurids 43,51 . The lateral groove is considered homologous to the single longitudinal sulcus of most hyperodapedontines (Chatterjee 52 and subsequent authors). As a result, the MTBA is subdivided into a central medial tooth bearing area (cMTBA) and an inner medial tooth bearing area (iMTBA). The entire MTBA is broader than the LTBA, as is the case in Te. sulcognathus and H. huenei 29,41 , but it contrasts with the distinctly broader LTBA present in both morphotypes of H. tikiensis 33 . Both longitudinal grooves converge anteriorly at the anterior third of the maxilla of Teyumbaita sp. nov., resembling the condition in Te. sulcognathus and the morphotype 2 of H. tikiensis 33,41 , but the medial groove is restricted to the posterior third of the tooth plate in H. huenei 29 .
The number of tooth rows in the LTBA, cMTBA, and iMTBA show minor variation among preserved specimens. The LTBA and cMTBA have two longitudinal tooth rows posteriorly, and the iMTBA has two (CRILAR-Pv 587, 642, 643) or three to four (CRILAR-Pv 588) tooth rows on the posterior half of the tooth plate. The arrangement of these occlusal tooth rows resembles that of Te. sulcognathus 41 . In addition, there is a row of well-spaced lingual teeth immediately medial to the cushion-shaped iMTBA ( In the braincase, the ventral ends of the exoccipitals contact their counterparts on the floor of the endocranial cavity ( Fig. 8e,f), as occurs in most other hyperodapedontines (see above), but contrasting with the absence of such contact in Te. sulcognathus 41 . The parabasisphenoid has an oblique, posterodorsally to anteroventrally oriented, main axis in lateral view. The basipterygoid processes are dorsoventrally taller than transversely broad, as in other hyperodapedontines with the exception of I. genovefae 45 .
The dentary has a lateral cutting blade and a lower and transversely thicker medial cutting blade, as occurs in Te. sulcognathus and non-hyperodapedontine rhynchosaurids 41 . Multiple lingual teeth are located on the top of the medial blade, immediately medial to it, and on a medially bulged border, being disposed in a crowded pattern, resembling the condition in Te. sulcognathus 41 , but differing from the well-spaced lingual teeth of H. huenei 29 and several other hyperodapedontines (e.g. H. mariensis, MCN 1867-PV).
The morphology of the atlas (Fig. 8h) closely resembles that of Te. sulcognathus 46 and other hyperodapedontines (e.g. H. huxleyi 32 ). The dorsal margin of the neural spine of the axis possesses a strongly convex central portion in lateral view that becomes concave at the level of the postzygapophyses, resembling the condition in H. gordoni 39 . By contrast, the posterior portion of the dorsal margin of the axial neural spine of Te. sulcognathus 46 (UFRGS-PV 0232T, 0298T) is convex in lateral view. The postaxial cervical vertebrae have a relatively short centrum with a thick, ventral keel. The postzygapophyses have a stout, crest-like epipophysis that vary in the series from short structures that do not extend posteriorly beyond the postzygapophyseal facet to substantially longer epipophyses, as occurs in Te. sulcognathus 46 . The best preserved dorsal vertebra has a spool-shaped centrum without a ventral keel (Fig. 8i). The neural arch possesses short and thick anterior and posterior centrodiapophyseal and postzygodiapophyseal laminae. There is no epipophysis, nor a hyposphene or hypantrum.
The scapula has a broad and fan-shaped blade, more anteriorly than posteriorly expanded (Fig. 8j). By contrast, the posterior margin of the scapular blade is nearly straight in Te. sulcognathus 46 . The acromial process is well-raised from the rest of the bone and mainly laterally projected, resembling the condition in some other hyperodapedontines (e.g. H. huxleyi, ISIR 01), whereas in Te. sulcognathus this process is shorter and blunt 46 . The humeral entepicondyle lacks the autapomorphic well-developed longitudinal groove of Te. sulcognathus 46 . The ectepicondyle has a tall supinator ridge and shallow ligament groove, which are mainly proximodistally oriented (Fig. 8k), as is the case in Te. sulcognathus 46 , H. gordoni (Benton 1983), and Hyperodapedon sp. nov. (CRILAR-Pv 585). By contrast, these ridge and groove are oblique, posteroproximally to anterodistally oriented, in H. sanjuanensis (MACN-Pv 18185) and H. huxleyi 32 . The ulna lacks an olecranon process and has a subtriangular lateral tuber in proximal view (Fig. 8l). The femur has a well-developed internal trochanter and a tibial condyle with a posteromedially oriented apex, as in some other hyperodapedontines (e.g. Te. sulcognathus and I. genovefae 45,46 ).

Description and comments.
Most of the specimens (CRILAR-Pv 589-593, 597, 598) here referred to Teyumbaita sp. are represented by partial, generally isolated, maxillae with two longitudinal grooves that extend anteriorly beyond the posterior third of the tooth plate and, thus, can be referred to this genus 41 . However, the presence of the row of lingual teeth that is diagnostic of Teyumbaita sp. nov. cannot be determined in these specimens because of damage or lack of preservation. Similarly, other specimens (CRILAR-Pv 594, 596) that do not preserve a maxilla are referred to the genus Teyumbaita because of the presence of lateral and medial cutting blades in the dentary and more than two longitudinal rows of dentary lingual teeth that are disposed in a crowded pattern 41 . By contrast, the dentary lingual teeth of H. huenei are less numerous and well-spaced from one another 29 . The morphology of these specimens is consistent with those of both T. sucolgnathus and Teyumbaita sp. nov., but there is no character-state that allow determining them at an alpha-taxonomy level. www.nature.com/scientificreports/ the maxilla and small pits and ridges of variable size on the dentary. In contrast, the maxillae of Chanaresuchus bonapartei, Gualosuchus reigi, Tropidosuchus romeri, and Pseudochampsa ischigualastensis are not ornamented 57 .
The right maxilla of CRILAR-Pv 579 is fairly complete, with a damaged anterior end, and preserved in articulation with the anterior process of the jugal. The anterior process of the maxilla extends a long way anteriorly to the antorbital fenestra, has a subhorizontal dorsal slope, and is strongly ornamented on its external surface (Fig. 9a). The premaxillary contact, at the anterior margin of the process, is not preserved. The ascending process forms the anterodorsal margin of the antorbital fenestra. The transition between the anterior and ascending processes is gradual, forming an obtuse angle, similar to the condition of Ch. bonapartei and Ps. ischigualastensis. As in Proterochampsa nodosa and Pro. barrionuevoi, the antorbital fossa excavates only the ascending process of the maxilla, where it forms a broad, anteriorly rounded depression. The palatal process is dorsoventrally low, forming the lateral and anterior borders of the internal nares (Fig. 9b). The ventral surface of this process bears some, possibly neurovascular, small foramina, as well as some striated areas for probably soft tissue attachment. Its dorsal surface is smooth, and slightly convex. The posterior process of the maxilla is laterally overlapped and excluded from most of the external ventral border of the antorbital fenestra by the jugal (Fig. 9a). In contrast, the maxilla forms most of the ventral border of the antorbital fenestra in other proterochampsids. The antorbital fenestra is mainly dorsally oriented, unlike the laterally facing fenestra of Tro. romeri, Cerritosaurus binsfeldi and rhadinosuchines. Ten tooth positions are preserved in the maxilla, but the total number would have been slightly higher, resembling the tooth counts of other specimens of Proterochampsa (e.g., at least 11 tooth positions in PVSJ 77 58 ). The preserved maxillary tooth crowns are relatively short and distally recurved (Fig. 9a,b). They have denticles on the apical three-quarters of the distal margin of the crown, whereas the mesial margin is not serrated. The denticles have a quadrangular edge and the interdenticular spaces do not extend onto the central region of the crown as blood grooves. There are 5 denticles per mm and they are apically oriented (Fig. 9c), contrasting with the orthogonal orientation of the maxillary denticles of other proterochampsids and the widespread condition in other archosauriforms. The apex of an erupting, mesial replacement tooth is visible between the bases of two interdental plates. The medial wall of the alveoli is formed by pentagonal interdental plates that do not contact one another (Fig. 9b). This exposes a wide groove extending along the bases of the interdental plates, with a replacement pit for each tooth 27,59 (Fig. 9b).
The preserved anterior tip of jugal possesses an ornamented external surface similar to that on the maxilla (Fig. 9a). The bone reaches anteriorly the level of the anterior border of the antorbital fenestra, as occurs in other specimens of Pro. barrionuevoi 58 .
The anterior end of dentary (Fig. 9d,e) lacks most of its ventral region. In medial view, the preserved ventromedial region includes a small part of the roof of the Meckelian canal, showing that the latter was ventrally placed. In lateral view, the dentary surface is strongly ornamented by several pits and small ridges, unlike the smooth surface of that bone present in Ps. ischigualastensis, Tro. romeri, and Ch. bonapartei. The preserved portion of dentary bears five alveoli, which are oval and anteroposteriorly longer than broad, and three teeth in situ with different degrees of eruption. The dentary tooth crowns have a serrated distal carina, whereas the mesial one lacks serrations, as occurs in the maxillary teeth. There are 6 denticles per mm in the dentary teeth. Their morphology is similar to those of the maxillary teeth, but they are orthogonal to the main axis of the crown (Fig. 9f).
Four vertebral centra were recovered, the best preserved of which belongs to an anterior or mid-cervical vertebra because its ventral surface bears a well developed longitudinal keel and the parapophyses are located slightly above the mid-height and adjacent to the anterior margin of the centrum (Fig. 9g, h). The ventral keel extends along the entire length of the centrum, as also seen in cervical vertebrae of Ps. ischigualastensis and Ch. bonapartei. The centrum is approximately as tall as long in lateral view (Fig. 9g), differing from the longer than tall cervical vertebrae of Ps. ischigualastensis. The anterior and posterior articular facets are shallowly concave and suboval, being lateromedially broader than tall. The base of the parapophysis is sub-oval, longer than tall.
The lateral surface of all the centra lacks a lateral fossa. The disarticulated cervical neural arch (Fig. 9i) lacks laminae and the neural spine is very tall, slightly anteroposteriorly expanded at its distal end, and vertical in lateral view. The distal end of the neural spine has longitudinal striations and lacks a spine table, as is the case in other proterochampsids 58 . The spine has a rounded anterodorsal corner, whereas the posterodorsal one is broken. The right prezygapophysis is complete, with a transversely broad and dorsomedially facing articular facet. The postzygapophyseal articular surfaces are ventrolaterally facing. The preserved right diapophysis is entirely located on the neural arch and at level with the roof of the neural canal. The diapophysis is long and slightly posteriorly oriented in dorsal view.
A small section of long-bone diaphysis is preserved. It has a smooth surface, an oval cross section, and may correspond to a femur because of the presence of a large, drop-shaped nutrient foramen. CRILAR-Pv 579 is assigned to Proterochampsa based on the following synapomorphies of the genus 58 : dermal sculpturing consisting of nodular protuberances and prominent ridges with smaller periodic nodular growths along their length, antorbital fossa restricted to an elongate depression on the maxilla anteriorly to the antorbital fenestra; dorsoventrally flattened skull with dorsally facing antorbital fenestrae. The specimen from the Hoyada del Cerro Las Lajas can be referred to Pro. barrionuevoi because it possesses a low, but coarse ornamentation Material. CRILAR-Pv 581, articulated basioccipital, ventral end of exoccipitals and parabasisphenoid, left paraoccipital process, five fragmentary vertebral centra, two neural spine fragments, proximal end of right femur, and proximal end of right tibia (Fig. 10).
Description. The occipital condyle is wider than high, slightly bilobed, and has a subcircular notochordal pit on the dorsal margin of its occipital surface (Fig. 10a-c). The latter condition is also present in rhynchosaurs     www.nature.com/scientificreports/ Arc. arborensis), in which the exoccipitals contact at the midline. The exoccipitals barely contribute to the occipital condyle and where it is broken off, we can recognize at least one exit for cranial nerve XII (Fig. 10b). The posteriormost region of the parabasisphenoid is preserved, contributing to the basal tubera. In its anterolateral contact with the right exoccipital, at the base of the braincase cavity, a small foramen can be recognized and putatively referred to part of the wall of the metotic foramen. This condition is seen in other pseudosuchians such as Ari. babbitti 69 , Ef. okeeffeae 68 , and B. kupferzellensis 70 . The base of the right paroccipital process of the opisthotic is partially preserved (Fig. 10f,g). It is subtriangular in cross-section at the base and oval towards its distal end, being anteroventrally to posterodorsally flattened. The paroccipital process has a well excavated stapedial groove on its posteroventral surface that opens into the brain cavity through the metotic foramen (Fig. 10g).
The vertebral remains of CRILAR-Pv 581 are very fragmentary (Fig. 10d, e,h-k), mainly represented by articular surfaces of centra, which have circular profiles and are markedly concave . The most complete ones are hourglass-shaped in ventral view, strongly constricted towards the body of the vertebra, and the parapophysis can be identified on the ventrolateral margin of one of them. Strongly constricted centra can be seen in cervical vertebrae of Sillosuchus longicervix (PVSJ 85, PVL 2267), Sh. inexpectatus (TTUP 09001), and Ef. okeeffeae 71 among poposauroids and an unnamed early crocodylomorph from the Ischigualasto Formation in the Hoyada de Ischigualasto (PVSJ 846, 890 72 ). Two fragmentary neural spines are preserved (Fig. 10d,e). They are laterally compressed and do not expand distally; the dorsal margin is rounded, straight in lateral view, and has transverse striations.
The proximal end of the right femur ( Fig. 10l-n) is anterolaterally to posteromedially compressed resembling that of Sh. inexpectatus (TTUP 3870), Ef. okeefeae 68 , Ari. babbitti 73 , and Poposaurus gracilis (TTUP 11613). Its proximal surface is well preserved, showing a longitudinal straight groove (Fig. 10l) as seen in several pseudosuchians (e.g. Pre. chiniquensis, SNSB-BSPG AS XXV 10; Pop. gracilis, TMM 31100-408, UCMP 28359; Sil. longicervix, PVSJ 85; Aetosauroides scagliai, PVL 2073). As typical of pseudosuchians, the proximal head is rounded and not clearly differentiated from the shaft, unlike those of non-aphanosaurian avemetatarsalians, in which the head is separated from the shaft by a notch or a concave depression ventral to the femoral head 74 . A moderately developed greater trochanter can be recognized on the posterolateral region of the proximal end of the femur, granting it a quadrangular shape in posterior view (Fig. 10m,n). The posteromedial tuber 71 is the largest of the proximal tubers and is subtriangular in proximal view, contrasting with the anteromedial tuber, which is smaller and more rounded, a condition similar to that of Poposaurus gracilis (TMM 31100-408, UCMP 28359). The anterolateral tuber is rounded and wide, occupying the medial third of the anterior margin of the femoral head. The intertrochanteric fossa is seen between the greater trochanter and the posteromedial tuber, it is shallow and level with the greater trochanter. The shaft is strongly anterolaterally to posteromedially compressed with the cortical bone collapsed where the bone is broken off.
The proximal end of the right tibia is partially preserved and the proximal surface is convex (Fig. 10o,p). The anterior margin of the tibia is rounded whereas the posterior one is sharper. The lateral condyle is well-developed and offsets anteriorly from the medial condyle as in several other archosauriforms (e.g., Ch. bonapartei, PVL 4575). The medial surface of the proximal end of the tibia is slightly concave anteriorly and convex posteriorly.
Comments. CRILAR-Pv 599 can be referred to Paracrocodylomorpha, in particular the clade composed of Saurosuchus galilei + Crocodylomorpha, based on the following synapomorphy 81 : anteroposteriorly shortened (blade-like) basal tubera of the basioccipital. Within this group it resembles Crocodylomorpha in the absence of contact between the ventral margin of the exoccipitals. Nevertheless, this condition is also present in shuvosaurid poposauroids (Ef. okkeeffeae and Sh. inexpectatus) 71 . In this regard, there are other character states of CRILAR-Pv 599 that resembles poposauroids: vertebrae with very constricted (hourglass-shaped) centra; unexpanded distal end of neural spines; anterolateral to posteromedially compressed proximal end of femur; straight longitudinal furrow on the proximal surface of femur. By contrast, it differs from shuvosaurid poposauroids in the presence of a rounded posteromedial tuber, lower than the anteromedial one, on the proximal end of the femur. As a result, the morphology of CRILAR-Pv 599 does not completely match that of poposauroids but we refrain to unambiguously refer it to Crocodylomorpha because of its fragmentary condition and similarities with Po. gracilis. Another paracrocodylomorph is known from the Hoyada del Cerro Las Lajas, the crocodylomorph Trialestes romeri (PVL 3889) 75 , and although it has a congruent overlapping morphology with the vertebrae of CRILAR-Pv 599, there are no preserved diagnostic features to determine if this specimen belongs to the same species. Material. CRILAR-Pv 580, several fragmentary teeth, two fragmentary vertebral centra, numerous fragments and natural casts of paramedian, ventral, and appendicular osteoderms, fragmentary ribs, right fragmentary coracoid, proximal end of right humerus, proximal end of both ulnae, distal end of left tibia, and distal end of metatarsal (Fig. 11).
Description. Several teeth are preserved as fragments and natural casts. Preserved tooth crowns lack their apices and are circular in cross-section (Fig. 11a), being clear that they were conical, resembling those of A. scagliai (PVL 2052(PVL , 2059  www.nature.com/scientificreports/ AbIII/1995) and De. spurensis (TTUP 09420). Because of their poor preservation, it cannot be determined if they had wear facets or serrations.
Both recovered vertebrae are represented by poorly preserved partial centra, which are spool-shaped (Fig. 11b,c), as typical of aetosaurs. The articular surfaces are circular, but it cannot be determined whether they represent anterior or posterior facets. The several incomplete ribs are elliptic in cross-section and some have a sharp edge on the anterolateral or posterolateral margin.
The partial right coracoid is represented by the glenoid fossa (with a medially expanded coracoid lip) and the posteroventral portion of the bone (Fig. 11f-h). The glenoid fossa is teardrop-shaped, with the tapering portion towards the scapula. The subglenoid lip is damaged and it cannot be determined whether it had a postglenoid process or not. The proximal fragment of the right humerus preserves a globose head and the proximal part of the deltopectoral crest (Fig. 11d,e). The proximal end of the humerus is anteroposteriorly compressed, corresponding to a gracile element that resembles that of A. scagliai (PVL 2073) and contrasts with the thick humeral head of N. engaeus (PVL 3525). Both preserved ulnae have a clearly discernible tapering, but relatively short olecranon process at the proximal end (Fig. 11i-k). The olecranon process is shorter than the long axis of the proximal end of the ulna as in other suchians (e.g., Riojasuchus tenuisceps: PVL 3828; A. scagliai: PVL 2073; Fasolasuchus tenax: PVL 3850; P. kirkpatricki: TTUP 9000). Particularly, the olecranon of CRILAR-Pv 580 is lateromedially compressed and sharp, resembling that of A. scagliai (PVL 2073) and Typothorax coccinarum (NMMNH-P 56299), and contrasting with that of most aetosaurs, which are wider and more rounded (St. olenkae: ZPAL AbIII/1179; Aetosaurus ferratus: SMNS 5770 S16; N. engaeus: PVL 3525). A lateral tuber for the articulation of the radius is present on the proximal end of the ulna of CRILAR-Pv 580, as in other suchians 71 .
The distal end of the left tibia is elliptic in cross-section with the anterior margin slightly tapering. The distal articular surface would have contacted the astragalus and its posterolateral corner projects further ventrally (Fig. 11l,m). This asymmetric distal end with a ventral projection is also present in  (Fig. 11n). The distal articular surface is convex and circular in cross-section. A distal fragment of metatarsal is preserved and possesses a ginglymoid articulation (Fig. 11o-q). Two condyles occupy the ventral surface, with the lateral one more ventrally projected than the medial one, resembling the right metatarsal I of A. scagliai (PVL 2052). The cross-section of the shaft is elliptic, being dorsoventrally compressed.
Numerous osteoderms are preserved in association, mainly as natural moulds. Most represent paramedian osteoderms (wider than long), there are a few ventral elements (quadrangular), and two are appendicular osteoderms (rhomboidal) (Fig. 11r-u). Their external ornamentation pattern is composed of long ridges and deep grooves that radiate from the dorsal eminence. There are some pits near this dorsal eminence as well. The characteristic anterior bar without ornamentation can be identified in some casts of paramedian osteoderms. Their internal surface is flat and unornamented. The appendicular osteoderms present a longitudinal eminence that projects from the anterior to the posterior margin and lack an anterior bar.
Comments. CRILAR-Pv 580 is here assigned to A. scagliai (by monotypy) based on the morphology of the teeth, the presence of a tapering olecranon on the ulna, a well-developed ventral projection on the distal end of tibia, and the general morphology of the osteoderms. Discussion ischigualasto chronostratigraphic framework. Our measured succession of the Ischigualasto Formation at the Hoyada del Cerro Las Lajas is 1059 m-thick (92% exposed), compared to a maximum formation thickness of 691 m in the central IPP 79 . Our Bayesian age-stratigraphic model based on three new high-precision U-Pb CA-ID-TIMS tuff dates from the Hoyada del Cerro Las Lajas can be extended to the Ischigualasto Formation and its rich fossil record at IPP based on the following considerations: (a) the dated 'Toba-2' marker tuff bed, 107 mab at Hoyada del Cerro Las Lajas, is a direct correlative of the Herr Toba bentonite, ~ 20 m above base of the formation at IPP and, (b) the relatively sharp, but conformable contact of the Ischigualasto Formation with the overlying Los Colorados Formation is expected to be the same age at both location (see stratigraphy in Supplementary Information). Therefore, we place conservative age constraints of 230.2 ± 1.9 Ma and 221.4 ± 1.   83 , which could be younger based on our results. Based on our new U-Pb geochronology, the 1059 m-thick succession of the Ischigualasto Formation at the Hoyada del Cerro Las Lajas was deposited in 8.8 (± 1.9) Myr, which translates into an average sediment accumulation rate of ~ 120 m/Myr. Compared to the most expanded part of the formation at IPP (691 m) with an average accumulation rate of ~ 74 m/Myr, the Cerro Las Lajas succession is ~ 1.6 times more expanded. taphonomical model. The Ischigualasto Formation was deposited under a semi-arid palaeoclimate, and the presence of floodplains along with the development of palaeosoils at the mid-section, indicates periodic rainfall 84 . The attributes observed in the vertebrate remains (large concentration of pristine, articulated, and non-weathered fossils, autochthonous assemblages, general absence of scavenging, little or no reworking) at the base of the Ischigualasto Formation indicates a specific taphonomic model; i.e., short transport and exposure (= "census assemblage" according to the Johnson's Model I), with rapid burial and low time-averaging, resulting in three-dimensionally arranged skeletal remains, with no sorting or orientation, and a polytypic taxonomic content 85 . Rapid burial, allied with various biological processes (e.g., faunistic turnovers, local extinctions), directly and positively influence fossil preservation. These biological and physical factors would haved a direct effect on bone input rates and time-averaging. Accordingly, the lower levels (first 300 m) of the Ischigualasto Formation have the highest concentration and the best preservartion of fossil vertebrates 3,86 .
Based on the above model, the palaeoecosystem of the Ischigualasto Formation as recorded in the Hoyada del Cerro Las Lajas was initially characterized by a high biocenosic (life assemblages) load, represented by an abundant vertebrate association, similar to that seen in outcrops of the IPP, in San Juan Province 3 . Due to several death factors (e.g., torrential storms, increased volcanism, and biotic factors), a poor thanatocenosis phase (pre-burial death assemblage) occurred, and rapid burial factors quickly introduced the remains into the taphocenosis phase.
The middle and top levels of the Ischigualasto Formation in the Hoyada del Cerro Las Lajas have a meagre fossil record or are almost devoid of fossils, as also observed in the IPP 3 . The reasons behind this pattern are not clear. Pyroclastic deposits, such as bentonites, ignimbrites, and welded tuffs, indicate a high volcanic influence in the area and elsewhere [87][88][89][90] . These materials are much more abundant in the upper third of the studied succession, reaching significant thicknesses in certain intervals. These volcanogenic deposits suggest proximity to volcanic centers that may have resulted in more hostile palaeoenvironments, with a reduced biomass. Yet, this is a highly conjectural inference, given that the biocenosis and thanatocenosis phases are unknown. On the other hand, the poor fossil record of these upper beds could in part be the result of lower sedimentation rates that would directly affect the taphocenosis phase, with skeletal remains being more intensely weathered, reworked, or scavenged, leading to reduced fossil preservation. Although equally extensive fossil prospecting has been carried out in the upper levels of the Ischigualasto Formation in the Hoyada del Cerro Las Lajas, sampling biases cannot be completely ruled out as an explanation for the lack of fossils. Indeed, a combination of the above factors may explain the taphonomy of the Ischigualasto Formation and the scarcity of fossils from its upper levels. faunal correlations. The fossil collection effort undertaken in the Ischigualasto Formation at the Hoyada del Cerro Las Lajas clearly did not produce a sampling as complete as that available for the IPP. Thirty-five specimens were identified to the genus level, compared to the nearly 1,000 specimens identified over the last 25 years in the IPP 3 . Yet, some significant faunal patterns have emerged at the Hoyada del Cerro Las Lajas that deserve further scrutiny. One of these is the stratigraphic separation between the sampled rhynchosaur genera, with Hyperodapedon occurring up to 260 mab and Teyumbaita occurring immediately above that. In contrast, the range of the traversodontid cynodont Exaeretodon spans nearly the entire fossil-bearing strata (120-400 mab), as is the case in the IPP 3 , whereas the archosauriforms Aetosauro. scagliai and Pro. barrionuevoi have single records together with Hyperodapedon and Teyumbaita, respectively. This is also respectively the case for the inferred provenances of Pi. mertii and V. rusconii plus Tri. romeri (see above).
A two-fold subdivision of the studied assemblages is, therefore, conceivable (Fig. 5), with an older fauna including Hyperodapedon, Exaeretodon, A. scagliai, and possibly Pi. mertii, occurring between 115 and 260 mab, succeeded by a younger fauna with Teyumbaita, Exaeretodon, Pro. barrionuevoi, and possibly V. rusconii and Tri. romeri, between 260 and 350 mab. Based on the dominant rhynchosaurs, these assemblages are herein referred to as Hyperodapedon and Teyumbaita biozones, respectively. The former was recorded above the 'Toba-2' tuff dated at 229.25 ± 0.10 Ma, with the 228.97 ± 0.22 Ma tuff positioned within the beds with Hyperodapedon. As for the Teyumbaita biozone, our age-depth model (see above) constrains it to between ca. 227. 94  www.nature.com/scientificreports/ with the record of Exaeretodon sp. in the eponymous biozone. Yet, the most noticeable biostratigraphic pattern seen in IPP is the high abundance of the rhynchosaur "Scaphonyx" sanjuanensis in the lower 100 m of the section, with the taxon decreasing in abundance in the next 200 m, until it disappears at 300 m 3 . A key discussion involves the identification of the IPP rhynchosaurs. "Scaphonyx" is a nomen dubium that may refer to any hyperodapedontine rhynchosaur, and the main taxon identified in IPP is better referred to as Hyperodapedon sanjuanensis 29,40 . Indeed, this species represents the totality of the rhynchosaurs recorded in the Hoyada de Ischigualasto 3 . The biostratigraphic patterns recognized in the Hoyada del Cerro Las Lajas have some resemblances to those of IPP. The cynodont Exaeretodon is the taxon with the broadest range in both areas. However, the significantly lower collection effort in the Hoyada del Cerro Las Lajas hampers a confident estimate of the Exaeretodon range. As such, the upper ca. 700 m of the Ischigualasto Formation at the Hoyada del Cerro Las Lajas may be devoid of fossils in part due to insufficient sampling, although a less abundant fossil record in the upper portions of the Ischigualasto Formation is also seen in IPP and fits the taphonomic model proposed above.
We speculate that the replacement of Hyperodapedon by Teyumbaita at about 260 mab may not be simply a preservation artefact, in part based on the high abundance of fossils at this level. Based on our age model, the Hyperodapedon and Teyumbaita biozones in the Hoyada del Cerro Las Lajas may respectively correlate (Fig. 12) to the lower and upper parts of the Scaphonyx-Exaeretodon-Herrerasaurus biozone at IPP 3 . Another interesting biostratigraphic pattern found in the Hoyada del Cerro Las Lajas is the abundance of Teyumbaita around 260-300 mab. Although in much lower numbers because of a poorer sampling, this compares with the likely older Hyperodapedon proportional richness at the base of IPP sections. Preliminary reports mentioned the presence of Te. sulcognathus (= "Scaphonyx" sulcognathus) in the "upper" levels of the Ischigualasto Formation in the Hoyada de Ischigualasto 91,92 . Yet, these reports did not provide collection numbers for the putative Teyumbaita specimens, which could not be located in the collections or restudied here, neither there is precise information of how "upper" were these specimens collected in the section of IPP. In any case, these putative Teyumbaita records in IPP were reported as stratigraphically above those of "Scaphonyx" sanjuanensis, matching the biostratigraphic pattern described here for the Hoyada del Cerro Las Lajas. Only a broad alpha-taxonomy revision of the Hyperodapedon-clade 93 specimens of IPP would shed light on the biostratigraphic distribution of the rhynchosaurs in that area 38,40 . It would be important to see if Teyumbaita occurs in that area and, if present, whether or not it is stratigraphically separated from Hyperodapedon as in the Hoyada del Cerro Las Lajas. This will be a test of the rhynchosaur turnover as a useful biostratigraphic marker.
Previously, the rhynchosaur genus Teyumbaita was only conclusively recognised in the Late Triassic beds of the Santa Maria Supersequence 94 in southern Brazil (Fig. 12). Montefeltro et al. 41 revised the three betterknown records of the taxon, which was found above those of Hyperodapedon in all sites it occurs, matching the pattern seen in the Hoyada del Cerro Las Lajas. All the records of Teyumbaita in Brazil are isolated, with no other associated index-fossils in beds of the same site. Accordingly, such Teyumbaita-bearing beds cannot be directly correlated with putatively coeval strata from other Triassic sites in southern Brazil, such as those where Exaeretodon abound in the absence of confirmed records of Hyperodapedon 95,96 . In fact, one previous record of Hyperodapedon (based on specimen MCN 3509PV), along with Exaeretodon in the Janner site [96][97][98] , is yet to be confirmed and, until further scrutiny, MCN 3509PV is safely assigned only to the Hyperodapedon-clade 34 . In any case, the Argentinean record of Teyumbaita supports the correlation of the strata where Exaeretodon is more abundant than Hyperodapedon in that country (i.e. upper Scaphonyx-Exaeretodon-Herrerasaurus and Exaeretodon biozones) 3 with the southern Brazilian beds where Teyumbaita and/or Exaeretodon occurs/abounds in the absence of Hyperodapedon 95 , i.e. Exaeretodon sub-assemblage zone 99 . The inferred ca. 228-227 Ma age for the Teyumbaita-rich beds in Argentina also matches the recently reported radioisotopic ages for the Hyperodapedon and Riograndia Assemblage Zones in Brazil 93 . These have been dated at ca. 233 and 226 Ma, respectively, and are consistently positioned below and above (Fig. 12) the Teyumbaita beds in the proposed stratigraphic schemes of the Santa Maria Supersequence 94 .
The Late Triassic tetrapod faunal compilation of South America 93 revealed a gap in the tetrapod fossil record of western Pangaea near the Carnian-Norian boundary, bounded from below by the lower Ischigualasto Formation fauna, and from above by the faunas of the upper Ischigualasto Formation and Riograndia Assemblage Zone. The revised age model for the Ischigualasto Formation presented here essentially fills that purported gap, showing the continuity of a faunal structure recognized in strata such as those of the Hyperodapedon Assemblage-Zone in Brazil (dated at ca. 233 Ma) 93 across that stage boundary, at least in palaeolatitudes close to 40°-50° South. For example, rhynchosaurs still abound and proterochampsids still occur in such strata, differing from younger beds that lack such taxa 93 .
Therefore, it seems that the major Late Triassic turnover seen in the terrestrial tetrapod biotas of western Gondwana post-dates the assemblages currently known for the Hoyada del Cerro Las Lajas (and most probably the Carnian-Norian boundary), which is followed by Norian non-fossiliferous deposits that evidence an increase in humidity (see above). This has a significant impact on the first occurrence of tetrapod groups such as saurischian dinosaurs and crocodylomorphs in the fossil record.
In Brazil, the ca. 226 Ma (early Norian) dated beds in which the Riograndia Assemblage Zone was recorded preserve sedimentary environments that drastically depart from those yielding the Hyperodapedon Assemblage Zone (including the Teyumbaita-bearings beds), the lower part of which was dated as ca. 233 Ma. This transition represents the replacement of an ephemeral anastomosed fluvial-lacustrine system (Alemoa Member of the Santa Maria Formation) by a perennial braided fluvial system (Caturrita Formation) 100 , which indicates a pluviosity increase, as also suggested by Th/U geochemical data 101 . As such, it seems that the Norian onset in southwestern Pangaea was marked by a humidity increase, as seen in the Caturrita Formation in Brazil and the upper levels of the Ischigualasto Formation in Argentina, after a more arid period that itself post-dated the Carnian Pluvial Event 102,103 . This coincides with a major biotic turnover 93 , when faunas with rhynchosaurs, proterochampsids, and herrerasaurid dinosaurs were replaced by faunas with the oldest plateosaurian sauropodomorphs 104  Pisanosaurus mertii and dinosaur origins. Pisanosaurus mertii Casamiquela 4 has been for a long time regarded as the oldest known ornithischian, but its dinosaur affinity was recently challenged by Agnolín and Rozadilla 107 , as well as in briefer accounts by Baron et al. 108 and Baron 109 . A comprehensive historical account of the taxon relationships has been provided 8 and there is no need to be duplicated here. Suffice to say, apart from broadly expressed scepticism [110][111][112][113][114] , the ornithischian affinity of Pi. mertii was only questioned on numerical phylogenetic grounds by the three papers mentioned above, and the matrices in Baron et al. 108 and Baron 109 are not independent from one another. In those hypotheses, the proposed alternative was to nest Pi. mertii among silesaurids, a dinosauromorph group usually positioned immediately outside the Dinosauria (but see Langer and Ferigolo 115 ). Given the potential evolutionary importance of this Hoyada del Cerro Las Lajas taxon, here we review the anatomical evidence brought forward by Agnolín and Rozadilla 107 in support of the silesaurid affinity of Pi. mertii, as well as features that may instead suggest its ornithischian affinity. Agnolín and Rozadilla 107 provided a compelling review of the anatomy of Pi. mertii and supported previous arguments claiming that the elements of its holotype belong to a single individual. Yet, a bone fragment that seems to represent a partial right femoral shaft was found among the Pi. mertii material 116 . It has an asymmetric fourth trochanter, as seen in non-neotheropod saurischian dinosaurs, in contrast to the pendant trochanter of ornithischians. This possible femur fragment has a preserved length of ca. 1.5 cm and a transverse width slightly below 1 cm, thus belonging to an individual considerably smaller than the holotype of Pi. mertii. As a result, we agree with Sereno 116 that bones of a smaller reptile-possibly an early saurischian dinosaur-may be stored together with and probably were associated to the holotype of Pi. mertii. Nevertheless, we still adhere to previous claims that the bones historically associated to Pi. mertii-which do not include this probable femur-belong to a single individual based on the matching size of the bones and field data describing the degree of articulation of the specimen when it was collected 107 . Here, we do not aim to review the anatomy of Pi. mertii, but regard this as a much-needed future enterprise, especially if assisted by non-destructive, tomographic techniques. Instead, we focus on revising anatomical traits that might help resolving the contentious placement of the taxon as either an ornithischian or a silesaurid.
We agree with Agnolín and Rozadilla 107 that the specimen preservation does not allow the positive identification of an external mandibular fenestra in Pi. mertii, but neither allows to confirm its absence. Yet, Pi. mertii has a depressed area on the lateral surface of the post-dentary portion of the hemimandible (Fig. 13a) that is recognized only in heterodontosaurids, i.e. "external mandibular fossa" of Sereno 116 , among early dinosauromorphs [117][118][119] . Its more deeply excavated anteroventral corner is in a position similar to that occupied by the external mandibular fenestra of ornithischians 116 . Indeed, a reduced fenestra could be an evidence of the ornithischian affinity of Pi. mertii, whereas its putative absence would be autapomorphic for Pi. mertii.
For about two centimetres posterior to the broken tip of the bone, the medial surface of the dentary of Pi. mertii is medially expanded at its ventral margin (Fig. 13b, c). A similar condition is seen in some ornithischians 116,119 , but not in silesaurids 115,120 . Also, the dentary of Pi. mertii possesses a gradual torsion towards its anterior end, where the cross-section of the bone gets dorsolateral to ventromedially oriented (Fig. 13c). This condition is present in at least some ornithischians (e.g. Eocursor parvus, SAM-PK-K8025), but absent in silesaurids and other early dinosauromorphs. The presence of a tall coronoid process on a relatively short dentary, producing a strongly concave dorsal margin of the bone in lateral-medial views, also resembles more the condition in heterodontosaurids 17 than in silesaurids 121 , saurischians 93,117,122 , and even other ornithischians 119,123 ; but a better quantification of these differences must be provided before they can be used to infer the affinities of Pi. mertii. Also, the apparent lack of replacement foramina in the Pi. mertii lower jaw may support its heterodontosaurid affinities within ornithischians 116 , as those are present in other early members of that group 123,124 , as well as in silesaurids 115,125 .
The dentary of Pi. mertii has a strong, bulbous ridge extending below the tooth row, especially at its posterior half, forming a "buccal emargination" 116 , between the ridge and the teeth (Fig. 13a,b). A similar structure also appears to extend above the maxillary tooth row, but this is harder to confirm given the incomplete preservation of that bone. Such a strong ridge and emargination is absent in silesaurids 115 and dinosaurs in general, including early sauropodomorphs 93,122 . On the contrary, this is seen in several 116,119,[126][127][128] , although not all, early ornithischians 123,124,129 . Accordingly, the presence of a buccal emargination in Pi. mertii better supports an ornithischian, rather than silesaurid affinity.
Inferring the mode of tooth attachment in Pi. mertii is exceedingly hard because of poor preservation (and conjectural) without the support of CT-Scan or histology techniques. Thus (contra Agnolín and Rozadilla 107 ), until more detailed data is available, we consider that tooth attachment cannot be used to infer either affinity (i.e. silesaurid or ornithischian) for Pi. mertii. Nevertheless, repreparation of the base of one of the maxillary teeth revealed the absence of fusion to the surrounding bone. Most preserved tooth crowns of Pi. mertii are broader labiolingually than mesiodistally (Fig. 13b), or at least equally broad in those two axes 107 . This is a very unusual condition among early dinosauromorphs and is approached only by the highly modified molariform teeth of some heterodontosaurids 116 . We agree with Agnolín and Rozadilla 107 that tooth crowns of Pi. mertii lack signs of carinae or denticles, but we disagree with those authors (p. 22) 107 in that the lack of denticles support a silesaurid affinity for Pi. mertii, as these elements are clearly seen in most representatives of the group 115,120,130 . Unlike Agnolín and Rozadilla 107 , we found it difficult to define enamel thickness or the presence of longitudinal ridges in the Pi. mertii teeth. Yet, we concur with those authors that a blunt primary ridge, like that of ornithischians 116,123,124,127 is seen in some teeth. This produces a morphology that most closely resembles that of the constricted, cup-shaped bases of Heterodontosaurus tucki and Lycorhinus angustidens molariform Scientific RepoRtS | (2020) 10:12782 | https://doi.org/10.1038/s41598-020-67854-1 www.nature.com/scientificreports/ crowns 116,131 . Similarly, the maxillary tooth crowns of Pi. mertii are medially inclined (Fig. 13d), resembling the condition in He. tucki (SAM-PK-K1332), but contrasting with the vertical teeth of silesaurids. Finally, tooth crowns of Pi. mertii are only similarly short as those of some silesaurids if these are compared to the "blade-like" crowns of plesiomorphic putative members of the group such as Lewisuchus admixtus 132,133 . Accordingly, contra Agnolín and Rozadilla 107 , short tooth crowns cannot be used prima facie (i.e. in the absence of a phylogenetic framework) to infer a silesaurid affinity for P. mertii.    4 ) that a pair of isolated vertebrae of Pi. mertii represent cervical elements. The centra are parallelogram-shaped in lateral view, a condition unrecognized in early dinosauromorph tail vertebrae. As such, they more closely resemble the short cervical elements of early ornithischians 116,135 than the more elongated vertebrae of silesaurids 136 . The trunk vertebrae of Pi. mertii are generally compressed lateromedially, with tall neural arches (about as deep as the centra) and well-developed prezygadiapophyseal, postzygadiapophyseal, prezygaparapophyseal, and anterior and posterior centrodiapophyseal laminae (see fig. 6 in Agnolín and Rozadilla 107 ). This condition is very similar to that of Silesaurus opolensis 136 , markedly departing from the morphology of early ornithischian trunk vertebrae, which are not as lateromedially compressed and have dorsoventrally short neural arches with poorly defined or no lamination 116,123,135,137 . As for the impressions of the sacral vertebrae, we agree with most authors that there is evidence of at least four elements and that the sacral ribs are shared between two vertebrae. Silesaurids have either two or three sacral vertebrae 120,121 and sacral ribs shared between two vertebrae are seen in both silesaurids 120,130 and early ornithischians 132,138 (Scelidosaurus harrisonii, NHMUK PV R1111). Hence (contra Agnolín and Rozadilla 107 ), the latter trait cannot be employed prima facie to infer a silesaurid affinity for Pi. mertii.
Pelvic features are very hard to identify, including if the acetabulum was open or closed, even if partially. Yet, we concur with Sereno 124 that none of the modifications seen in the opisthopubic pelvis of ornithischians can be recognized in Pi. mertii. On the contrary, the puboischial articulation is dorsoventrally extended and the ischial symphysis is not restricted to the distal end of the bone, suggesting a plesiomorphic propubic pattern. The popliteal fossa of the femur of Pi. mertii seems to be overprepared along its distal two centimeters, but we agree with Agnolín and Rozadilla 107 that ridges surrounding the fossa can be traced along the five distal centimeters of the bone (see fig. 11 in Agnolín and Rozadilla 107 ). The associated tibia is about 16 cm long, so we could infer a minimal femoral length of 15 cm, suggesting that such ridges extended over the distal third of the bone. This condition matches that of silesaurids, in which the popliteal fossa extends for more than one fourth of the femoral length 132 , but conditions similar to that of Pi. mertii, with rather subtle proximally extending ridges, are seen in various early dinosauromorphs, including ornithischians 127 (Scutelosaurus lawleri, MNA V175; Laquintasaura venezuelae). Also (contra Agnolín and Rozadilla 107 ), the cranial intermuscular line of early dinosaurs does not usually reach the distal third of the bone 139,140 , so that its absence in the preserved portion of femur of Pi. mertii is not evidence for its non-dinosaurian affinity.
As stated by Agnolín and Rozadilla 107 , the tibia of Pi. mertii is indeed devoid of a cnemial crest that expands proximally relative to the femoral facet. This configuration is very similar to that of Sacisaurus agudoensis 115 , departing from the typical dinosaur condition, including that of most ornithischians 116,127,138 (Sc. lawleri, MNA V175). Therefore, the reduced cnemial crest of Pi. mertii seems to better fit a silesaurid affinity. Yet, this character has a more complex distribution, with proximally unexpanded crests seen in undescribed specimens of La. venezuelae and the holotype of Lesothosaurus diagnosticus (NHMUK PV RU B17), and a more projected crest is seen in Asilisaurus kongwe 120 . A fibular crest as that of most theropods and silesaurids is indeed seen in the tibia of Pi. mertii and lacking in most early ornithischians 127,135,138 (Sc. lawleri, MNA V175), but we see no reason (contra Agnolín and Rozadilla 107 ) to disregard its homology to the crest present in heterodontosaurids 141 . In addition, the posterior hemicondyles of the proximal end of the tibia are separated from one another by a deep and very transversely broad notch (see Fig. 6l in Irmis et al. 111 ) that closely resembles the condition in several early ornithischians (e.g. He. tucki, SAM-PK-K1332; Eo. parvus, SAM-PK-K8025; Sc. lawleri, UCMP 130580). By contrast, the posterior hemicondyles of the tibia are separated by a distinct change in slope or a narrow groove in other early dinosauromorphs.
The articulation with the astragalus hampers a proper assessment of the distal outline of the tibia, but it is possible to infer that it is at least as broad anteroposteriorly as lateromedially, as occurs in most silesaurids 120,130 and early dinosaurs, such as herrerasaurids and early sauropodomorphs 122,142 . Instead, neotheropods, ornithischians 142 , and Sa. agudoensis 115 , have a much more lateromedially expanded distal end of the bone. We agree with Agnolín and Rozadilla 107 that the posterolateral margin of the tibia of Pi. mertii is not concave as in early ornithischians (L. diagnosticus, NHMUK PV RU B17; Sc. lawleri, MNA V1752), but we disagree that this feature supports its silesaurid affinity, as a similar plesiomorphic condition is also seen in most early dinosauromorphs, including dinosaurs 142 . As for the descending process of the tibia (= outer malleolus or posterolateral process) of Pi. mertii, it expands only slightly lateral to the anterolateral corner of the distal end of the bone (see fig. 12c in Agnolín and Rozadilla 107 ), resembling the condition of Asili. kongwe 120 , early sauropodomorphs, and herrerasaurids 122,142 , but markedly differing from neotheropods, ornithischians 142 , and Sa. agudoensis 115 , which bear an extensive descending process. Also, the descending process is well developed in some specimens of Si. opolensis (ZPAL Ab III 413, 415), although not as much as in ornithischians, whereas other specimens of that silesaurid (ZPAL Ab III 403/1, 460/3) bear a short process as in Pi. mertii. Finally, as mentioned by Agnolín and Rozadilla 107 the distal part of the fibula in Pi. mertii is not as slender as that of ornithischians 116,123,135 , retaining instead the transversely broader condition that is plesiomorphic among dinosauromorphs.
The anteroposterior breadth of the lateral margin of the astragalus of Pi. mertii is about three-fourths of its lateromedial width (Fig. 13e). Thus, that bone is proportionally less lateromedially expanded than in silesaurids 120,130 and early saurischians 122,139,143,144 , but has similar proportions to that of ornithischians (Sc. lawleri, MNA V175) 141 . Similarly, the proximodistal height of the astragalar ascending process with respect to the height of the astragalar body (see fig. 12d,e in Agnolín and Rozadilla 8 ) resembles more the condition in ornithischians (Sc. lawleri, MNA V175) than the shallower process of silesaurids 120,130 . However, the lateromedially narrow and medially sloping astragalar ascending process of Pi. mertii, as seen in anterior view (see fig.12c in Agnolín and Rozadilla 107 ), resembles the condition in silesaurids more than that of early ornithischians.
The calcaneum of Pi. mertii (Fig. 13e) retains a well recognizable calcaneal tuber and an expanded posteromedial corner of the bone, forming the "ventromedial projection" 143 . This shape is a modification of the general plesiomorphic subtriangular calcaneum seen in most early dinosauriforms 120 116,129,141 . The anterior and lateral margins of the calcaneum-fibula articulation reveal a convex calcaneal facet, with a slightly straighter posterior part, whereas the medial view reveals a concave calcaneum margin (see fig. 13a,b in Agnolín and Rozadilla 107 ). As discussed by Agnolín and Rozadilla (p. 11,19) 107 , but with the sides reversed, such a complex articulation is more typical of non-dinosaurian archosaurs, although lacking in lagerpetids and Si. opolensis 143,145 . On the other hand, the fibula-calcaneum articulation of Pi. mertii clearly differs from that of ornithischians, in which the main basin that occupies the posterior two-thirds of the proximal articulation of the calcaneum receives the outer malleolus of the tibia, whereas the fibula articulated only to the top of the raised anterior third of the bone 129,141 (Sc. lawleri, MNA V1752). Agnolín and Rozadilla 107 mentioned that metatarsal IV of Pi. mertii resembles those of silesaurids and saurischians because it is compressed at the proximal end. Indeed, the available data for early ornithischians (Le. diagnosticus, NHMUK PV RU B17; He. tucki, SAM-PK-K1332) reveal a more robust proximal articulation of metatarsal IV. On the contrary, contra Agnolín and Rozadilla (p. 22) 107 , the ungual phalanx of the fourth pedal digit of Pi. mertii, and the only available for the taxon (see Fig. 15 in Agnolín and Rozadilla 107 ), lacks the marked dorsoventral flattening present in silesaurids, contradicting a possible silesaurid affinity.
The features discussed above (Table 2) show that most character-states previously used to support a silesaurid affinity for Pi. mertii are likely plesiomorphic for Dinosauriformes, being also present in other dinosauromorphs and early dinosaurs. Some other characters are variable among silesaurids and ornithischians. Conversely, we found over ten character-states that are shared only by Pi. mertii and heterodontosaurids and/or other early ornithischians among early dinosauriforms. As a result, we consider that the ornithischian affinity of Pi. mertii rests on much stronger grounds than the silesaurid hypothesis. A quantitative analysis of the phylogenetic relationships of Pi. mertii goes beyond the scope of this paper and it will be conducted in the near future, integrating the new information provided here.
As discussed by previous authors 107,109,111,116,124 , considered as an ornithischian, Pi. mertii fills the long ghost lineage between other members of the group (Early Jurassic) 109 and the oldest known saurischians (ca. 233 Ma) 93 . Here, we constrain the age of Pi. mertii as ca. 229 Ma, showing that this species is latest Carnian. As a result, the long ghost lineage is transferred into the ornithischian clade, with the group absent in the fossil record for more than 30 My. Pisanosaurus mertii provides key clues about the early evolutionary history of Ornithischia and it is thus one of the most important components of the Hoyada del Cerro Las Lajas fauna. However, at the same time, it shows how deficient is our current knowledge of the first million years of evolution of this main dinosaur lineage.

conclusions
• The most complete succession of the Late Triassic Ischigualasto Formation is exposed in the Hoyada del Cerro Las Lajas, consisting of more than 1,000 m of fluvial-channel and flood overbank deposits with high volcanic input, which have produced historical tetrapod fossils. www.nature.com/scientificreports/ petrography. Bone and sedimentary rock thin-sections were prepared at the CRILAR Petrographic Laboratory using the protocol described by Fiorelli et al. 147 ; specimens were washed with distilled water and cut using a Buehler PetroThin™ device, dried at 40 °C in an oven for 24 h, and subsequently attached with epoxy resin to glass slides of 28 × 48 × 1.8 mm dimensions. Thin-section analyses were made with a Leica DM2500P petrographic microscope. Images were captured with a Leica DFC295 digital camera attached to the microscope and connected to a computer for data processing, editing and measurements. The images and figures were designed and edited with CorelDRAW X5. Petrographic results are described in detail in the Supplementary Information.
U-pb geochronology. Several samples of tuff from the Ischigualasto Formation at Hoyada del Cerro Las Lajas were collected for U-Pb zircon geochronology by the CA-ID-TIMS method. Table S2 shows the extended U-Pb zircon geochronology results. Sample processing, isotopic analyses and data reduction was carried out at the Massachusetts Institute of Technology Isotope Laboratory. Details of analytical methods and procedures are fully explained in the Supplementary Information.
Bayesian age-stratigraphic model. A Bayesian age-depth model has been employed to extrapolate statistically robust ages for the stratigraphic levels of interest (e.g., fossiliferous intervals) constrained by dated tuff beds. For this model, the Bchron software package 148,149 written for R 150 was used. Detailed information on age modelling and interpretation can be found in section iii of the Supplementary Information.