Increasing the theropod record of Europe: a new basal spinosaurid from the Enciso Group of the Cameros Basin (La Rioja, Spain). Evolutionary implications and palaeobiodiversity

Abstract

A new member of Spinosauridae from the Enciso Group (uppermost Barremian–lower Aptian) from Igea (La Rioja, Spain) is here erected on the basis of axial, pelvic girdle, and hindlimb elements that exhibit a unique combination of characters. Riojavenatrix lacustris gen. et sp. nov. is one of the latest Iberian and European spinosaurid taxa. It retains a triangular pubic boot, like the megalosaurids, and a medial condyle of the femur that shows a transitional stage between the anteroposteriorly oriented long axis of non-spinosaurid theropods and the posteromedially oriented long axis of Spinosauridae. The spinosaurid record of Iberia ranges from the late Hauterivian–early Barremian to the latest Barremian–early Aptian so far, and both the oldest and the most recent evidence comes from the Cameros Basin, where spinosaurid remains are especially abundant in the Barremian deposits. A review of the spinosaurid record has allowed us to dismiss the presence of the genus Baryonyx from Iberia; hence, only Camarillasaurus, Iberospinus, Protathlitis, Riojavenatrix gen. nov., and Vallibonavenatrix are considered to be present in the Early Cretaceous of Iberia. According to this study, Riojavenatrix is one of the youngest baryonychines in the fossil record.

INTRODUCTION

Spinosauridae comprises a bizarre group of large-sized theropods, characterized by elongated snouts, conidonty, robust forelimbs with a large manual ungual phalanx I, and tall neural spines forming a dorsal sail in some taxa (e.g. Charig and Milner 1997, Sereno et al. 1998, Holtz et al. 2004, Bertin 2010, Hendrickx et al. 2015, Arden et al. 2019). First erected on the basis of disarticulated elements from the Cenomanian of Egypt and the description of the enormous Spinosaurus aegyptiacusStromer, 1915 (Stromer 1915, Bertin 2010), this tetanuran clade has been involved in great controversy. The morphology of the skull and teeth resembling those of crocodyliforms and the gut content with fish scales suggest a piscivorous diet, at least in part, but also feeding on dinosaurs and pterosaurs (e.g. Charig and Milner 1997, Buffetaut et al. 2004). Owing to their feeding ecology, spinosaurids were, at least in part, dependent on aquatic habitats. A predominantly aquatic lifestyle has also been proposed for some spinosaurid taxa, such as BaryonyxCharig & Milner, 1986 and Spinosaurus, even with a subaqueous foraging strategy (e.g. Ibrahim et al. 2014, 2020, Beevor et al. 2021, Fabbri et al. 2022). Nevertheless, this is still a highly debated topic (see e.g. Hone and Holtz 2017, 2021, Heckeberg and Rauhut 2020, Sereno et al. 2022, Isasmendi et al. 2023). Regardless, spinosaurids have usually been positively associated with coastal environments (Sales et al. 2016).

Traditionally, spinosaurids have been included within Megalosauroidea, as a sister group of megalosaurids (e.g. Sereno et al. 1998, Rauhut 2003, Allain et al. 2012, Hone and Holtz 2017, Malafaia et al. 2020a). Spinosauridae comprises the subfamilies Baryonychinae and Spinosaurinae (Sereno et al. 1998). Megalosauroidea was also considered paraphyletic by Rauhut and Pol (2019), and Spinosauridae would be early branching within Carnosauria as sister taxa of Megalosauridae + Allosauroidea (Rauhut and Pol 2019). Nevertheless, the more recent analysis carried out by Schade et al. (2023), based on a modified version of the matrix of Rauhut and Pol (2019), again recovered Megalosauroidea as a monophyletic clade, which included Spinosauridae within it. The division within Spinosauridae has also been questioned by Evers et al. (2015) and Sales and Schultz (2017). Indeed, Evers et al. (2015) did not find this traditional division to be strongly supported. According to the phylogenetic analyses carried out by Sales and Schultz (2017), baryonychines are recovered in a polytomy. Despite all the phylogenetic controversy, spinosaurids are always considered as a monophyletic group (e.g. Sereno et al. 1998, 2022, Allain et al. 2012, Carrano et al. 2012, Sales and Schultz 2017, Rauhut and Pol 2019, Malafaia et al. 2020a, Mateus and Estraviz-López 2022, Schade et al. 2023).

This clade of early-branching tetanurans has a wide geographical distribution, having been reported from Asian, European, North African, and South American deposits that range from the Early Cretaceous to the Cenomanian (Bertin 2010, Candeiro et al. 2017, Hone and Holtz 2017). Nevertheless, spinosaurid fossil remains are rather uncommon, mostly consisting of isolated teeth and bone fragments (Holtz et al. 2004, Bertin 2010, Hone and Holtz 2017). It is currently supposed that spinosaurids originated in Europe, with at least two subsequent Early Cretaceous migrations to Africa (Milner 2003, Barker et al. 2021), but recently, three baryonychine teeth have been reported from the Berriasian–Valanginian deposits of the Feliz Deserto Formation of Brazil, suggesting a more complex scenario (Lacerda et al. 2023). Hitherto, spinosaurids are particularly diverse in the Early Cretaceous European archipelago, where they are represented by the genera Baryonyx, CamarillasaurusSánchez-Hernández & Benton, 2014, CeratosuchopsBarker et al., 2021, IberospinusMateus & Estraviz-López, 2022, ProtathlitisSantos-Cubedo et al., 2023, RiparovenatorBarker et al., 2021, and VallibonavenatrixMalafaia et al., 2020a.

Besides Camarillasaurus, Iberospinus, Protathlitis, and Vallibonavenatrix, other spinosaurid remains from the Iberian Peninsula have been assigned to baryonychines, specifically to the genus Baryonyx (Torcida et al. 1997, Fuentes-Vidarte et al. 2001, Buffetaut 2007, Viera and Torres 2013, Figueiredo et al. 2015), but these attributions should be considered with caution owing to the fragmentary nature of the specimens. Most of these spinosaurid remains consist of scattered bones or isolated teeth of both baryonychine and spinosaurine theropods (e.g. Torcida Fernández-Baldor et al. 2003, Sánchez-Hernández et al. 2007, Alonso and Canudo 2016, Alonso et al. 2017, Gasca et al. 2018, Isasmendi et al. 2020, 2023). However, as new studies provide evidence for a more complex scenario in Iberia for spinosaurids, a revision of these fossils should be made based on the current record.

This paper is aimed to describe a partly articulated spinosaurid theropod from the Early Cretaceous Enciso Group of La Rioja (Spain). The specimen is here proposed as the holotype of a new theropod genus and species, Riojavenatrix lacustris, previously suggested to have baryonychine affinities (Malafaia et al. 2018) or even assigned to the genus Baryonyx (Viera and Torres 2013). The herein erected taxon contributes to a better understanding of the palaeobiodiversity of Spinosauridae in the Early Cretaceous of Europe and Gondwana. Furthermore, a review of the current spinosaurid record is presented, and the spinosaurid palaeobiodiversity of the Iberian Peninsula is analysed.

GEOGRAPHICAL AND GEOLOGICAL SETTING

The partly articulated specimen (CPI 1637–1648 and CPI 1675–1677) was found in the locality of Igea (southeast of La Rioja, Spain), specifically at the Virgen del Villar-1 site (Fig. 1). This site is located southwest of the main town of Igea, near the Virgen del Villar hermitage (Fig. 1A).

Geologically, the Virgen del Villar-1 site is located in the northeastern part of the Cameros Basin. The Cameros Basin formed in the second rifting stage that took place during the Late Jurassic and Early Cretaceous in the Iberian Mesozoic Rift (Mas et al. 2002). The deposits of the basin are mainly continental or coastal, reaching up to a thickness of 6500 m in its depocenter, dating back to the Tithonian–early Aptian (Martín-Chivelet et al. 2019 and references therein). Traditionally, these deposits were divided into five lithostratigraphic groups (Tera, Oncala, Urbión, Enciso, and Oliván groups) by Tischer (1966) or into eight depositional sequences (DS1–8) by Mas et al. (2002), where fluvial deposits derive laterally and upwards into lacustrine deposits (Mas et al. 1993, 2002, Hernán 2018).

Two sub-basins can be differentiated in the Cameros Basin: Eastern Cameros and Western Cameros sub-basins (Mas et al. 1993, 2002). The stratigraphic succession found at the Virgen del Villar-1 site is part of the DS7 sequence of the Eastern Cameros sub-basin. Specifically, the fossil locality is situated in the Enciso Group (Fig. 1A), which constitutes the central sector of the Cameros Basin (Tischer 1966, Hernán 2018). This group is composed of mixed siliciclastic–carbonate deposits that exceed 2000 m in thickness in the main depocenter of the basin (Clemente 2010, Suarez-Gonzalez et al. 2013), and its deposits have mainly been interpreted as a siliciclastic-influenced carbonate lacustrine and palustrine environment (Mas et al. 1993, 2002, 2011), also with fluvial and deltaic episodes (Hernán 2018).

Intense diagenesis and low-grade metamorphism (Casquet et al. 1992) affected the sediments of the Enciso Group. Initially, by deep diagenesis conditions reached at burial depths between 5000 and 6000 m, and posteriorly, by a later (Albian) low-grade hydrothermal metamorphism, reaching temperatures between 300 and 350°C and burial pressures under 1–2 kbar, respectively (Mata et al. 2001, Del Río et al. 2009, Omodeo-Sal et al. 2017 and references therein).

Regarding the age of the unit, several studies have proposed different chronological ranges (e.g. Alonso and Mas 1993, Mas et al. 1993, Martín-Closas and Alonso Millán 1998, Casas et al. 2009, Schudack and Schudack 2009). Nonetheless, recent studies date the Enciso Group from the latest Barremian to the early Aptian, with most of the unit being early Aptian in age, with likely lowermost upper Aptian deposits in the uppermost part of the group (Suarez-Gonzalez et al. 2013, 2015, Hernán 2018).

MATERIALS AND METHODS

The described material (CPI 1637–1648 and CPI 1675–1677) consists of a vertebral element (fragment of a dorsal neural arch), pelvic girdle remains (left pubis and right ischium), and hindlimb bones (proximal end of a right femur, shaft and distal end of a left femur, proximal right tibia, left tibia, an almost complete left fibula, left astragalus, left calcaneum, metatarsal III, a possible metatarsal II, three left non-ungual pedal phalanges, and left I-2 pedal phalanx) (Fig. 2). The left femur, tibia, astragalus, and the calcaneum were found articulated. All the material is likely to belong to the same individual, because it was recovered in association in 2 m² in the same part of the quarry, the fossils are of a consistent size, and there is a lack of duplicated elements. The fossil material is housed at the Centro de Interpretación Paleontológica de La Rioja-CPI in Igea (La Rioja, Spain).

The studied material was examined and measured first hand (all measurements are given in Supporting Information, Supplementary Material S1). The following spinosaurid specimens were also examined directly or via photographs by one or more of the authors: Baryonyx walkeriCharig & Milner, 1986 (NHMUK VP R9951); Camarillasaurus cirugedaeSánchez-Hernández & Benton, 2014 (MPG-KPC1–46); Iberospinus natarioiMateus & Estraviz-López, 2022 (ML 1190); Ichthyovenator laosensisAllain et al., 2012 (MDS-Savannakhet BK10-01–15); the proposed neotype of Spinosaurus aegyptiacus (FSAC-KK 118888; Ibrahim et al. 2014); Suchomimus tenerensisSereno et al., 1998 (MNN GDF500); and Vallibonavenatrix caniMalafaia et al., 2020a (MSMCa-1–6, 9–15, 18–20, 22–24, 27–28, 32–33 and 53–55). Besides the abovementioned taxa, two femora (CMP-3b/211 and CMP-MS-0/22) and a tibia (CMP-3c/188) published by Malafaia et al. (2018) were studied in person. Furthermore, comparisons with ‘Spinosaurus B’ Stromer, 1934 (Nr. 1922 X 45) were made using the drawings made by Stromer (1934). Only the specimen number is indicated when the observation was made first hand by any of the authors. When the study was based on figures from previous works, the figures are indicated, and when previous authors made the observations, the references are included.

Phylogenetic methodology

We analysed the phylogenetic position of the new genus and species described here using a modified version of the matrix of Rauhut and Pol (2021), which, in turn, was previously modified from Wang et al. (2017). This character data matrix is a large dataset of theropod dinosaurs that is updated currently and continuously by the Mesozoic Tetrapod work group (main institutions in LMU and BSPG, Germany). The coding of the new taxon was added to this matrix, together with a review and rescore of seven other taxa (Baryonyx, Camarillasaurus, Ichthyovenator, Sigilmassasaurus Russell, 1996, Spinosaurus, Suchomimus, and Vallibonavenatrix). For the analysis, Sigilmassasaurus and Spinosaurus genera have been distinguished as different operational taxonomic units (as proposed by Evers et al. 2015). The neotype specimen proposed for Spinosaurus by Ibrahim et al. (2014) and the other elements attributed to the same genus have been coded within Spinosaurus. The original set of taxa from both matrices used have not been modified (no other taxa have been excluded or included). The data matrix has 206 taxa and 774 characters, among which 118 are ordered multistate characters (see character list in Supporting Information, Supplementary Material S2; see NEXUS file of the matrix in Supporting Information, Supplementary File S1).

The matrix was managed using Mesquite v.3.01 (Maddison and Maddison 2008) and imported into TNT v.1.5 (Goloboff and Catalano 2016). We performed a heuristic tree search in order to find most parsimonious trees (MPTs), using the New Technology algorithms: sectorial searches and tree fusing, using the default settings for all of them. These algorithms were applied to new searched trees, using the driven search to find the minimum-length trees 100 times. Subsequently, we exposed the results to tree bisection–reconnection (TBR) as a branch-swapping algorithm, saving 100 trees per replicate. The resulting MPTs were summarized using a strict consensus. The consistency index (CI) and the retention index (RI) were calculated using the stats.run script. The branch support was tested using the methodology proposed by Goloboff et al. (2008) to calculate Bremer support values and resampling methods, such as bootstrap and jackknife (100 replicates, summarized as frequency differences).

Furthermore, to find unstable taxa, a reduced consensus tree was obtained using the iterPCR (iterative calculation of the Positional congruence reduced) method (Pol and Escapa 2009) with the iterpcr.run script. This methodology generated two outputs, (i) the reduced consensus, a summary tree showing all the unstable and pruned taxa and its alternative position in the tree, and (ii) the analysis of the characters that contradicts some of the alternative positions and potential critical characters (see Supporting Information, Supplementary Material S2). Two alternative approaches were carried out to evaluate unstable taxa: (i) the pruned trees method (Goloboff et al. 2008), using the command prunnelsen; and (ii) the finding of the maximum agreement subtree (MAST), the largest subtree shared by all the MPTs (Gordon 1979, 1980).

In order to evaluate the alternative placement of pruned taxa in previous analyses, some constrained trees were defined, forcing the unstable taxa in these alternative locations in the reduced consensus tree, using the command force=, and, posteriorly, making a search enforcing constraint (Goloboff et al. 2008).

The appendicular characters have been mapped using Mesquite v.3.01 (Maddison and Maddison 2008) in the reduced consensus trees obtained by the iterPCR analysis, and also in the strict consensus tree in TNT, using the common map of character analysis.

In addition to this phylogenetic analysis, the new taxon was also coded in a modified version of the matrix used by Mateus and Estraviz-López (2022) (NEXUS file in Supporting Information, Supplementary File S2), which, in turn, was modified from Arden et al. (2019), Evers et al. (2015), and Carrano et al. (2012) and with skull characters added from Hendrickx et al. (2020). This matrix has a total of 534 characters and 24 taxa that have not been modified in this study. The aim of the phylogenetic analysis conducted in this study was to test the hypothesis of the position of Riojavenatrix obtained in the previous analysis using a matrix focused mainly on early-branching tetanurans. The methodology followed to run this analysis in TNT v.1.5 was exactly the same as for the analysis executed with the modified matrix from Rauhut and Pol (2021).

Histology

Several fragmentary limb bones were selected for histological examination and prepared according to the usual methodology outlined by Chinsamy and Raath (1992), subsequently developed by Lamm (2013). Nomenclature and definitions of bone microstructures are derived from the work of Francillon-Vieillot et al. (1990).

Institutional abbreviations

AMNH, American Museum of Natural History, New York, NY, USA; BSPG, Bayerische Staatssammlung für Paläontologie und Geologie, Munich, Germany; CPGP, Centro Português de Geo-História e Pré-História, Lisbon, Portugal; CPI, Centro de Interpretación Paleontológica de La Rioja, Igea, Spain; FMNH, Field Museum of Natural History, Chicago, IL, USA; FPMN, Fukui Prefectural Museum, Fukui, Japan; FSAC, Faculté des Sciences Aïn Chock, Casablanca, Morocco; IVPP, Institute of Vertebrate Paleontology and Paleoanthropology, Beijing, China; IWCMS, Dinosaur Isle Museum, Sandown, UK; LMU, Ludwig Maximillian University, Munich, Germany; MDS-Savannakhet, Dinosaur Museum of Savannakhet, Savannakhet, Laos; MDS-Salas de los Infantes, Museo de Dinosaurios, Salas de los Infantes, Spain; MG, Museu Geológico, Lisbon, Portugal; ML, Museu da Lourinhã, Lourinhã, Portugal; MNBH, Musée National Boubou Hama, Niamey, Republic of Niger; MNHN, Muséum national d’Histoire naturelle, Paris, France; MNHN/UL, Museu Nacional de História Natural e da Ciência da Universidade de Lisboa, Lisbon, Portugal; MNN, Musée National du Niger, Niamey, Niger; MPG, Museo Paleontológico de Galve, Galve, Spain; MSM, Museo Paleontológico Juan Cano Forner, Sant Mateu, Spain; NHMUK, Natural History Museum, London, UK; OUMNH, Oxford University Museum of Natural History, Oxford, UK; SHN, Sociedade de História Natural, Torres Vedras, Portugal; SM, Sirindhorn Museum, Department of Mineral Resources, Kalasin, Thailand; USNM, National Museum of Natural History, Smithsonian Institution, Washington, USA; WMN, LWL-Museum für Naturkunde, Münster, Germany; YPM, Yale Peabody Museum, New Haven, CT, USA.

Systematic palaeontology

Dinosauria Owen, 1842
Theropoda Marsh, 1881,
Tetanurae Gauthier, 1986
Spinosauridae Stromer, 1915,
Baryonychinae Sereno et al., 1998
Riojavenatrix gen. nov.

lsid:urn:lsid:zoobank.org:act:C657EC04-18D2-45F0-8E4E-CB9D6C964626

Etymology:

Rioja (toponymy): in reference to La Rioja, the Spanish region where the holotype specimen was recovered, and venatrix: Latin for huntress.

Type and only species:

As for the type species (see below).

Diagnosis:

As for the type species (see below).

Riojavenatrix lacustris sp. nov.

lsid:urn:lsid:zoobank.org:act:F5E10E99-9BD8-4384-9FD1-36575E24B05D

Etymology:

Latin for ‘related to a lake’, which declines from the word ‘lacus’ (lake).

Holotype:

A partial skeleton of a single individual, including: CPI 1637, left femur; CPI 1638, left tibia and astragalus; CPI 1639A–B, left fibula; CPI 1640, left metatarsal III; CPI 1641A–B, right ischium; CPI 1642, right tibia; CPI 1643, right femur; CPI 1644, possible fragment of metatarsal II; CPI 1645, left phalanx III-1; CPI 1646, left phalanx III-3; CPI 1647, left phalanx I-2; CPI 1648, left phalanx IV-2 (or IV-3); CPI 1675A–B, left pubis; CPI 1676, left calcaneum; and CPI 1677, fragment of a dorsal neural arch.

Type locality and horizon:

Virgen del Villar-1 site, La Rioja, Spain; the Enciso Group is uppermost Barremian–lower Aptian (Suarez-Gonzalez et al. 2013, 2015), but the type locality is most likely to be lower Aptian within the Enciso Group.

Diagnosis:

A medium- to large-sized spinosaurid theropod with the following unique combination of characters within Spinosauridae: (i) a lateromedially thick and triangular pubic boot in distal view, with a straight posterolateral margin (similar to Ichthyovenator, although in Ichthyovenator this margin is concave, but it is absent in Baryonyx, Suchomimus, and FSAC-KK 11888); (ii) an anteroposteriorly expanded ischial boot with an anterodorsally oriented tip and an angular anterodistal surface (absent in Ichthyovenator and FSAC-KK 11888; similar to Megalosaurus bucklandii); (iii) a narrow, restricted and relatively deep articular groove on the proximal surface of the femur, which is anteromedially–posterolaterally inclined (distinct from those of Baryonyx and Suchomimus); (iv) a medial femoral condyle with the long axis exhibiting only slight posteromedial orientation (distinct from those in Baryonyx, ‘Spinosaurus B’ (Nr.1922 X 45), Suchomimus, and FSAC-KK 11888); (v) presence of a vertical ridge on the medial margin of the ascending process of the astragalus (potential autapomophy, but it could also be a character in Spinosauridae because no other spinosaurid astragalus has been described to date); (vi) height of ascending process of the astragalus more than twice the height of the astragalar body (potential autapomophy, but it could also be a character in Spinosauridae because no other spinosaurid astragalus has been described to date); (vii) presence of an anterior depression with a dorsally located foramen on the lateral surface of the calcaneum (autapomorphy; absent in other spinosaurids); and (viii) absence of a longitudinal groove on the medial surface of phalanx I-2 (potential autapomophy, but it could also be a character in Spinosauridae or even in Megalosauroidea; given that this element is not preserved in other spinosaurids or megalosauroids, a synapomorphy cannot be excluded).

Description
Axial skeleton

Dorsal vertebra (Fig. 3):

This piece consists of the mid- and posterior portions of the base of the neural arch, the bases of the transverse processes, and the neural spine of a mid- or posterior dorsal vertebra (CPI 1677). The neural arch does not preserve any of the zygapophyses. The thickness of the neural spine is constant across its anteroposterior length, but anteriorly, the interspinous ligament scar is present and slightly thickens the neural spine transversely (Fig. 3A). The right spinodiapophyseal fossa is smooth, whereas the left one is more rugose. The transverse processes are horizontal in lateral view and slightly inclined dorsolaterally (Fig. 3B).

Anteriorly, the spinoprezygapophyseal laminae would have bounded a deep and dorsoventrally elongate spinoprezygapophyseal fossa, which is filled with matrix. This was inferred considering the distance between both laminae, the posterior extension of the fossa, and the considerable missing anterior portion of the spine (Fig. 3A). On the posterior side, another fossa is present, interpreted here as the spinopostzygapophyseal fossa (Fig. 3C). This fossa is also dorsoventrally elongate and deep, delimited by the spinopostzygapophyseal laminae. Ventrally, these laminae merge to form the hyposphene (Fig. 3C). Underneath the left transverse process, a fragment of the posterior centrodiapophyseal lamina is centrally positioned, extending ventrally to the centrum (Fig. 3B). Posterior to this lamina, the postzygocentrodiapophyseal fossa is present on the right side of the neural arch (Fig. 3B). This fossa is also present on the left side, but it is less pronounced.

Appendicular skeleton
Pelvic girdle

Pubis (Fig. 4A–J):

Two fragments of the left pubis are preserved (CPI 1675A–B). The proximal fragment preserves the articular surface for the ischium, the surface of the acetabulum, and part of the iliac articulation (Fig. 4A–E). The iliac articulation is broad, strongly rugose, and somewhat convex in proximal view (Fig. 4E). Adjacent to the medial margin, an anteroposteriorly elongate concavity is present on this articular surface. However, this area is not well preserved and might be an artefact of abrasion. The iliac peduncle is not complete anteriorly, but it broadens towards the acetabular surface (Fig. 4E). The medial surface of the iliac peduncle is straight, becoming convex near the acetabular surface in proximal view. Posterior to the iliac articular facet, the acetabular surface is a small triangular concavity, which is located medially on the proximal surface and is posteromedially inclined (Fig. 4E). The ischial peduncle projects posteriorly and tapers posterodistally towards the ischial articulation (Fig. 4D). The ischial articulation is not completely preserved at the proximal end. It is triangular in posterior view and posteromedially directed (Fig. 4D). The obturator notch is distally opened, relatively small anteroposteriorly, and subcircular (Fig. 4B, C).

The distal fragment of the left pubis (Fig. 4F–J) lacks part of the symphysis and the medial portion of the distal expansion (Fig. 4J). The shaft of the distal fragment is straight and shows a teardrop-shaped cross-section at the level of the symphysis. The pubic apron projects as a blade from the shaft (Fig. 4J). The symphysis is almost entirely missing, but the preserved distal part is medially projected. Furthermore, it is sigmoidal in medial view, with a large distal extension. Based on the breakage surface, its development, and distalmost extension of the preserved apron, the pubic apron would have almost reached or slightly extended along the pubic boot (Fig. 4J). The pubis is slightly expanded anteriorly on its distal end, and it has a greater posterior projection (Fig. 4G, I). In distal view, the pubic boot is mediolaterally broad, especially at its centre (Fig. 4H). Its anterior and medial surfaces are convex, whereas the lateral surface is gently concave. The posterior process of the pubic boot tapers posteriorly. In distal view, the pubic boot is triangular in shape (Fig. 4H).

Ischium (Fig. 4K–T):

The right ischium preserves its proximal end (including the iliac peduncle), part of the shaft and the ischial boot (CPI 1641A–B). Both lateral and medial surfaces of the iliac peduncle are convex (Fig. 4Q). At the level of the surface of the acetabulum, the medial surface is flat and the lateral one is slightly concave, whereas near the pubic peduncle, the medial surface becomes concave and the lateral surface convex (Fig. 4Q). In the posterior margin of the iliac peduncle, there is a dorsoventrally directed crest, forming a bulge near the dorsal margin. This bulge can be observed in medial and lateral views (Fig. 4L–N). Anterior to this crest, the medial surface shows some proximodistally oriented furrows.

The iliac peduncle is mediolaterally expanded, with an oval contour in proximal view (Fig. 4Q). Although it is abraded, the articular surface of the iliac peduncle is concave posteromedially. A protuberance is present at the centre of the concavity, bounded anteriorly and medially by a groove. This concavity is anterolaterally delimited by a transversely oriented edge (Fig. 4Q). The surface of the acetabulum bears a pronounced concavity that becomes shallower anteriorly. In proximal view, the acetabular surface has a medial crest, partly broken. This crest extends anteriorly from the anteromedial edge of the iliac articulation (Fig. 4Q).

The preserved cross-section of the ischial diaphysis is somewhat oval, with a rounded crest, posterolaterally directed, running longitudinally on the lateral surface of the shaft (Fig. 4O, R, S). The posterior surface of the diaphysis is concave proximally and flattens towards the ischial boot, becoming a crest at the boot (Fig. 4S). Both medial and anterior surfaces of the shaft are flat (Fig. 4O, P). The ischial boot, triangular in medial or lateral view, is not strongly expanded anteroposteriorly and exhibits an angular anterodistal surface (Fig. 4P, R). The posterior half of the lateral surface is convex, whereas the anterior half shows a concavity caused by the anterior process of the ischial boot (Fig. 4R). This anterior process is connected to the diaphysis by a blade that becomes gradually less prominent proximally (Fig. 4R). The blade is laterally inclined distally and becomes vertical proximally. The medial surface of the ischial boot is flat and shows some longitudinal striations that run proximally to the shaft (Fig. 4P). In distal view, the ischial boot shows a triangular outline owing to the tapering of the anterior process (Fig. 4T).

Hindlimb

Femur (Fig. 5):

The right femur preserves only its proximal end (CPI 1643), and the left femur consists of most of the diaphysis and the distal articular end (CPI 1637). The femoral head is gently convex and subcircular in medial view (Fig. 5D). Its proximodistal axis is slightly longer than the anteroposterior one. The femoral head is medially oriented and anteromedially directed at an angle of ~20° (Fig. 5J). The proximal surface of the femur is posterolaterally inclined. There is an anteroproximally inclined groove (i.e. the oblique ligament groove) on the posterior surface (Fig. 5C). This groove separates the femoral head from the shaft; therefore, the latter is well offset medially. Medial to this groove, the posterior lip of the femoral head is well developed and extends slightly beyond the posterior surface of the head (Fig. 5C, J). The femoral head has a concave ventral surface. This surface becomes broader anteriorly, making the femoral head overhang the medial surface of the femoral diaphysis (Fig. 5A, C). On the anterior side of the femoral head, there is an oblique ridge that extends from the femoral head to the proximal end of the diaphysis (Fig. 5A). In proximal view, a deep, broad, and oblique articular groove can be noticed. This is more pronounced anteromedially and becomes shallower and narrower posterolaterally. In the same view, the femoral head is laterally confluent with the greater trochanter, which is incomplete and narrower than the head (Fig. 5J).

The left femur lacks the proximal end and the diaphysis proximal to the fourth trochanter. The preserved length of the left femur measures ~515 mm. The shaft of the femur is rather straight in anterior and posterior views, but it is anteriorly bowed in lateral and medial views (Fig. 5E–H). The diaphysis is oval in cross-section, the mediolateral axis being larger than the anteroposterior one, with a maximum circumference of 322 mm. The anterior surface of the shaft is convex and slightly pinched in the middle (Fig. 5E). The pinch extends 190 mm distal from the proximal fracture. Medial and lateral to this pinch, the surface becomes flat. The lateral surface is convex and becomes flat next to the distal expansion of the bone (Fig. 5F). The medial surface is also convex but leads to a shallow and large concavity, which is located posterior to the medial epicondyle and extends until half of the medial condyle (Fig. 5H). The posterior surface of the femoral shaft is flat and faces posterolaterally (Fig. 5G).

The fourth trochanter is located posteromedially in the proximal part of the preserved shaft. Only its most distal end is preserved, consisting of a well-developed and prominent longitudinal crest that becomes broader proximally (Fig. 5G). Medially, there is a shallow and smooth groove; and laterally, there is a shallow, smooth, and broad concavity, which is excavated in the posterior surface of the diaphysis in lateral view.

The distal end of the femur is slightly more medially than laterally expanded in posterior view (Fig. 5G). The medial epicondyle or medial distal crest is found on the mediodistal surface. This medial epicondyle is rounded, low, and not well developed (Fig. 5E). On the anterior surface of the medial epicondyle, the attachment for the muscle femorotibialis externus is a small rugose patch. The extensor groove is broad and slightly V-shaped in distal view (Fig. 5I) and becomes even shallower and broader proximally (Fig. 5E). The flexor groove is rather broad, deep, and U-shaped in distal view (Fig. 5I). The flexor groove is bounded by two ridges medially and laterally, proximal to the condyles (Fig. 5G). The medial ridge runs vertically from the proximal margin of the medial condyle, whereas the lateral ridge is more prominent and oblique, originating at the proximal end of the tibiofibular crest. The medial crest merges with the shaft proximally.

The anterior surface of the lateral condyle is rounded. The medial condyle is slightly more flattened and anterolaterally oriented in the anterior surface (Fig. 5E). At the distal end, the condyles are robust, being wide mediolaterally. Both medial and lateral condyles project distally to an almost equal extent in anterior view (Fig. 5E). In the same view, the lateral condyle projects distally and slightly laterally, whereas the medial condyle is solely distally projected. The medial condyle is teardrop-shaped in posterior view (Fig. 5G). In distal view, the lateral condyle is rounded and the medial one is more elliptical with the long axis slightly posteromedially oriented (Fig. 5I). The distal condyles are not separated by an anteroposterior pronounced groove in the distal surface, but there is a central shallow groove that connects to the tibiofibular crest (Fig. 5I). The tibiofibular crest is broad and positioned on the posterior surface of the lateral condyle (Fig. 5G, I). In posterior view, the crista tibiofibularis is teardrop-shaped and oblique with respect to the long axis of the lateral condyle (Fig. 5G). The tibiofibular crest is somewhat trapezoidal in distal view. This is laterally bounded by a broader groove, and medially, a narrow and deep groove separates the crest from the distal margin of the condyle or the flexor groove. Between both grooves, there is a posterolaterally elevated ridge in distal view (Fig. 5I).

Tibia (Figs 6A–J, 7A–E):

The right tibia (CPI 1642) preserves only its proximal portion (Fig. 6A–D, I). The left tibia (CPI 1638) is complete, including the articulated astragalus (Figs 6E–J, 7A–E). The right tibia lacks the cnemial crest and the medial condyle. However, the lateral condyle and the proximal end of the fibular crest are well preserved.

The left tibia shows many furrows on the medial surface close to the proximal end. Furthermore, these are also present on the anterior margin of the medial malleolus. The shaft of the left tibia is straight and fairly robust, with a length of ~810 mm (Fig. 6E–H). The cross-section of the tibial diaphysis is oval, where the mediolateral axis is the largest. At mid-shaft, the diaphysis narrows laterally and becomes wider medially. In lateral view, the anterior surface of the tibia is concave proximally, becoming convex distally. The posterior surface is rather straight close to the proximal end and concave distally. The anterior surface is flat and the posterior surface gently convex along the whole shaft. The lateral surface is convex, whilst the medial margin is flat proximally and becomes concave at mid-shaft.

The cnemial crest of the left tibia is rounded (Fig. 6E, F). Dorsally, it expands mediolaterally and flexes laterally in anterior view. The cnemial crest projects from the anterior surface of the diaphysis beyond the proximal articular surface of the tibia (Fig. 6E, F). The lateral surface of the cnemial crest bears a longitudinal tuberosity near its anteroposterior corner (Fig. 6F, J). In proximal view, the lateral condyles are offset from the cnemial crest by a poorly developed incisura tibialis (Fig. 6I, J). The lateral condyles are large, rounded, and broad. The preserved portions of both medial condyles suggest that they were also large. There is no process extending anterolaterally from the lateral condyle in either tibia, but there is a notch present between the medial and lateral condyles in the right tibia (Fig. 6I). In proximal view, the posterior outline of the left tibia is nearly straight owing to the lack of the medial condyle, whereas the medial surfaces are smoothly convex in both tibiae.

The fibular flange is a low longitudinal crest with a slightly broadened base (Fig. 6E–G). It is situated proximally on the lateral surface of the diaphysis, but it is distally located to the proximal end of the tibia. The outline of the base of the fibular crest is oval, but narrow, in lateral view (Fig. 6F). This is separated from the proximally located lateral process associated with the fibular crest. In the right tibia, the proximal end of the lateral process associated with the fibular crest also reaches the proximal end of the tibia (Fig. 6B, E–G). These processes are rounded and proximodistally oriented (Fig. 6B, F).

The distal end of the tibia is expanded mediolaterally, being anteroposteriorly narrower. Both the lateral and medial malleoli extend somewhat equally distally, with the lateral malleolus being larger and slightly more distally projected than the medial one (Figs 6E, G, 7A, C). This makes the distal surface almost horizontal, being ~15° with respect to the horizontal or ~75° with respect to the long axis of the tibial shaft. The lateral malleolus is rounded and the medial one more angled in posterior view (Figs 6G, 7C). In medial view, the medial malleolus shows a flat surface bounded posteriorly by a longitudinal ridge, which is mediodistally inclined (Figs 6G, H, 7C, D). The medial malleolus is transversely wider and more robust than the lateral one. The facet for the fibula shows longitudinal furrows and occupies a large portion of the anterior surface of the lateral malleolus (Figs 6E, F, 7A, B). The astragalar facet of the tibia is high (~150 mm), broad, and triangular and occupies more than half of the distal end of the anterior surface of the tibia (Figs 6E, 7A). This facet is delimited proximomedially by the supraastragalar buttress. The latter is developed as a well-developed and bluntly rounded ridge. It originates mediodistally and it is almost vertical basally, becoming a low oblique ridge proximally. The supraastragalar buttress is more pronounced distally and gradually fades proximolaterally (Figs 6E, 7A).

Fibula (Fig. 6K–T):

The left fibula (CPI 1639A–B) is almost complete but lacks part of the distal diaphysis. The fibula is a slender bone with a rather thin shaft and with an estimated length of 750 mm. The proximal end is strongly expanded anteroposteriorly and slightly expanded mediolaterally. Furthermore, the fibula is deformed, displaying an S-like shape at ~340 mm from its proximal end. Below the S-like deformation, the shaft of the fibula is somewhat twisted, a feature that might be associated with the deformation (Fig. 6K–N).

The proximal end of the fibula is more posteriorly than anteriorly expanded (Fig. 6L, N). The posterior margin of the proximal fibula becomes thinner posteriorly to form a thin, blade-shaped margin (Fig. 6O). This margin is more rounded than the anterior one in medial view, which is more acute in lateral view (Fig. 6L, N). The proximal margin is slightly concave at the level of the medial fossa and becomes convex near the anterior and posterior margins in lateral view (Fig. 6L, N). In proximal view, the proximal margin of the fibula is comma-shaped, with the anterior margin being wider than the posterior one (Fig. 6O).

The medial fossa reaches the proximal margin and occupies half the expansion of the proximal end of the tibia. It is shallow, inverted teardrop-shaped, and centrally positioned in medial view (Fig. 6N). Anteriorly, the medial fossa is bounded by an elevated edge, which creates the anterior margin of the proximal end and broadens proximally. The posterior margin of the medial fossa is posteriorly opened (Fig. 6N). Below the medial fossa, the medial surface is planar. In the shaft, a posteromedially located low ridge runs ventrally and becomes oblique below the iliofibularis tubercle, bounding an anterior groove on the medial surface. At the level of this tubercle, a shallow depression can be noticed on the medial surface of the shaft (Fig. 6N). Below the oblique ridge, a proximally well-developed groove is present, which extends parallel to the diaphysis distally. The posterior edge of this groove consists of a medially raised edge. Distally, the groove fades at the distal end of the preserved shaft when the diaphysis expands (Fig. 6N).

At the proximal expansion of the fibula, the lateral surface is convex anteriorly, becoming almost flat in the middle of the lateral side and again slightly convex posteriorly (Fig. 6L). Some longitudinal furrows are present on the anterolateral surface of the fibula. Posteriorly, the proximolateral surface of the fibula has a longitudinal and vertically oriented depression. Another broader, but slightly shallower depression is present on the anterolateral surface, at the same level as the posterior one (Fig. 6L). The fibular diaphysis has a strongly convex lateral surface, but it is more planar at the level of the iliofibularis tubercle.

The insertion for the iliofibularis muscle is located on the anterior margin of the fibular shaft, approximately at one-third from the proximal surface (Fig. 6K, L, N). It is low and rugose in anterior view and is formed by two protuberances, which are bounded laterally by a longitudinal, shallow, and rugose groove (Fig. 6K, L, N).

The distal end is expanded anteroposteriorly and transversely. At the beginning of the distal expansion, both anterior and posterior edges are sharp. Distally, the anterior part is rounded (Fig. 6P), whereas the posterior margin is sharp edged and slightly inclined laterally (Fig. 6R). The medial surface for the articulation with the tibia is flat proximally (Fig. 6S). However, in distal view, the anterior portion of the medial surface appears convex, becoming concave posteriorly (Fig. 6T). On the lateral surface of the distal expansion, an oblique ridge starts anteroproximally and runs posterodistally down to the ventral end (Fig. 6Q). Posteroproximally, overlying this ridge, there is a shallow and broad depression. The distal outline of the fibula is teardrop-shaped. This surface shows a depression anteromedially and becomes convex posteriorly (Fig. 6T).

Astragalus (Figs 6E–H, 7A–E):

The left astragalus is almost complete (CPI 1638), lacking only part of the anterodistal margin of the medial condyle of the astragalar body. The astragalus is L-shaped in lateral view, with a relatively tall ascending process (Figs 6E, F, 7A, B). The astragalus measures 155 mm in height and 122 mm mediolaterally. In anterior view, the astragalar body shows an hourglass shape, with concave distal and proximal surfaces (Figs 6E, 7A). However, the distal concavity is broader and less pronounced than the dorsal one. In distal view, the astragalar body is rectangular (Fig. 7E). The medial condyle of the astragalus is expanded anteroposteriorly. There is a shallow distal groove, becoming broader and slightly deeper laterally. The groove is bounded anteriorly by a parallel low crest that is slightly more prominent laterally and almost reaches the astragalus–calcaneum contact (Fig. 7E). In anterior view, there is another groove, horizontal and shallow, developed between the condyles, at the mid-height of the astragalar body (Fig. 7A). This groove has its maximum depth medially, gradually gets shallower laterally, and extends until the lateral rim. There is a bulge on the groove. Between the astragalar body and the ascending process, there is a depression; hence, the ascending process is offset from the astragalar body (Fig. 7B). This depression is broader and more marked laterally, showing a triangular shape in proximal view. The facet for the fibula is not preserved. The medial side of the posterior margin of the astragalar body is more proximally projected than the lateral side (Fig. 7C). This posterior margin is gently convex over most of its extension, but laterally it becomes shorter and slightly concave. The astragalar body is smooth on most of its anterior surface. Nonetheless, the area that articulates with the calcaneum is rugose.

The medial condyle is more anteriorly projected than the lateral one in ventral view; hence, the astragalus is narrower laterally (Fig. 7E). Besides the anteroposterior expansion of the astragalar body, the distal condyles are oriented anterodistally. The distal condyles of the astragalus are rounded anteroposteriorly (Fig. 7E).

The ascending process of the astragalus is laminar, triangular, and proximolaterally oriented (Fig. 7A). The height of the ascending process is twice the height of the astragalar body, with the former measuring ~103 mm and the latter 52 mm. The ascending process is located slightly laterally, and it does not reach the lateral margin of the astragalar body (Fig. 7A). It arises at around one-fifth from the medial margin of the astragalar body. The process is transversely broad, with its base extending mediolaterally ~65% of the transverse width of the astragalar body. The medial margin of the ascending process shows a well-defined ridge that is perpendicular to the mediolateral axis of the astragalar body and arises from the anterodorsal surface of the astragalar body (Fig. 7A). This ridge is parallel to the ventral part of the astragalar buttress of the tibia. The ridge fades where the ascending process tilts laterally at an angle of 40–45° with respect to the long axis of the tibia. Proximally, this margin gradually becomes oriented more transversely. The lateral margin of the ascending process is rather straight and vertical (Fig. 7A). The proximal end of the ascending process is placed at three-quarters of the transverse width of the astragalar body from the medial side. The ascending process does not contact the supraastragalar buttress (Fig. 7A).

Calcaneum (Fig. 7F–K):

The left calcaneum (CPI 1676) is complete, with a height of 63 mm. Its anteroposterior width measures 72 mm, whereas its mediolateral width is 29 mm. The contact of the calcaneum with the astragalus is slightly sinuous, and the tibia overlaps the calcaneum. The calcaneum is reniform anteriorly and laterally (Fig. 7H, I). In lateral view, the anteroproximal margin is more proximally elevated than the posteroproximal margin (Fig. 7I). The proximal rim is concave for the articulation with the fibula, and the proximal articular surface is proximally directed. The articulation for the distal end of the fibula is well developed (Fig. 7F). The distal profile is strongly convex in lateral view (Fig. 7I). In proximal view, the calcaneum is anteriorly broad, narrowing posteriorly, with a sharp edge that is oriented posterolaterally (Fig. 7F). This edge broadens distally and has a crescent shape in order to accommodate the lateral malleolus of the tibia (Fig. 7J). The lateral surface has a broad vertical groove and another anteriorly located depression. The latter is as deep as the groove, triangular, and bears a foramen at its proximalmost point (Fig. 7I). Between the groove and the depression, a triangular convexity is present. The medial surface of the calcaneum shows a concave facet for the lateral malleolus and an anteriorly located groove to articulate the lateral condyle of the astragalus. The groove is posteriorly inclined and parallel to the anterior margin of the calcaneum (Fig. 7K).

Metatarsal III (Fig. 8A–F):

The left metatarsal III is complete (CPI 1640). The diaphysis is straight, measuring 364 mm proximodistally, and with a subrectangular cross-section at mid-shaft (Fig. 8C–E). Its borders are rounded, except for the medial surface, which is planar. Its proximal end is anterodorsally and posteroventrally expanded, especially posteroventrally projected in lateral view (Fig. 8C, E).

In proximal view, metatarsal III has an hourglass-shaped outline, the anterodorsal part of which is more lateromedially expanded than the posteroventral one (Fig. 8F). The articular surface for metatarsal II is proximally located in the medial surface of metatarsal III (Fig. 8E). It is concave, densely striated, and reaches the middle of the diaphysis. The articular surface for metatarsal IV is posteroventrally concave and convex anterodorsally close to the proximal end of metatarsal III (Fig. 8C). This surface is also striated, as in the articulation for metatarsal II. The medial articular surface is bounded anterodorsally and posteroventrally by two elevated edges that fade distally. The posteroventral edge is more pronounced than the anterodorsal one. On the anterodorsal surface, there is a marked scar (Fig. 8B). Another scar is located on the posteroventral surface, below the proximal expansion (Fig. 8D).

The distal condyle is divided posteroventrally by a shallow groove (Fig. 8A). On the anterodorsal surface, there is no marked hyperextensor pit; however, a subtle depression is present adjacent to the condyle (Fig. 8A, B). The distal condyle is subrectangular in distal view, with the medial side being anterodorsally and posterooventrally larger than the lateral side (Fig. 8A). The medial collateral ligament pit is much deeper and larger than the lateral one. Both of them are suboval (Fig. 8C, E).

Phalanx I-2 (Fig. 8G–K):

CPI 1647 is interpreted as the ungual phalanx of digit I based on the small size of the element (it measures 63 mm proximodistally) and its significant curvature. This (left) phalanx is complete. It is long, narrow, distally pointed, and strongly arched in lateral and medial views (Fig. 8H, J). The phalanx I-2 is triangular in cross-section but has an oval articular surface in proximal view (Fig. 8K). The articular surface for the articulation of phalanx I-1 is concave. This surface is slightly pointed dorsally in proximal view, and the proximodistal lip is not well developed. Ventrally, this surface is rounded. The flexor tuberosity, placed on the ventral side of the articular surface, is not very pronounced, but it is more developed than the lip (Fig. 8K). Both medial and lateral surfaces are convex proximally, but the lateral surface is slightly flatter (Fig. 8H, J). The dorsal surface of phalanx I-2 is convex, and the ventral side is comparably flatter (Fig. 8G, I). There are no ventral or flexor fossae. Both medial and lateral margins are softly convex; hence, this phalanx is symmetrical. Only the lateral surface bears a longitudinal groove. This is well developed and faces at about two-thirds proximally, becoming progressively shallower (Fig. 8H).

Phalanx III-1 (Fig. 8L–Q):

CPI 1645 is interpreted as a proximal phalanx because it lacks a proximal keel. Furthermore, it can be assigned confidently to digit III based on the rather symmetrical distal condyles and not as asymmetrical proximal articular surface. The left phalanx III-1 is virtually complete, measuring 118 mm in length, with the flexor tubercle partly eroded and lacking the ventral part of the medial condyle. The height and width of the phalanx change proximodistally, being higher proximally. Between both articular surfaces, the neck constricts the phalanx, especially ventrally, such that the phalanx is wider than tall at mid-shaft (Fig. 8N, Q).

The dorsal margin is smooth and gently convex, becoming flat and inclined distally. Above the proximal margin of the collateral ligament pits is the extensor fossa (Fig. 8M). It is deep and oval, with the longest axis directed transversely. The ventral surface is also slightly mediolaterally convex, but proximodistally concave in lateral and medial views (Fig. 8M, P, Q). Proximally, there is a proximolaterally oriented oblique groove that separates the flexor tubercle into two processes. The proximal articular surface is devoid of any keel separating the articular facets. This surface is weakly concave and subcircular, with a pronounced, medially inclined proximodorsal lip. This surface is partly deformed (Fig. 8O).

Close to the proximal end, there is a proximoventral fossa on both medial and lateral sides (Fig. 8N, Q). These are rugose and roughly triangular. Distally, the lateral collateral ligament pit is subcircular and deeper than the medial one. The medial collateral ligament pit extends more proximally and is oval (Fig. 8N, Q).

Both condyles are equally developed (Fig. 8L). The long axis of the lateral condyle is inclined slightly laterally and the medial condyle is somewhat inclined medially. The lateral condyle is slightly more dorsally and ventrally projected, whereas the medial condyle is slightly more distally projected (Fig. 8M, P). The condyles are separated by a shallow intercondylar sulcus. In distal view, the condyles are mediolaterally expanded on the ventral surface. In ventral view, the lateral condyle ends more abruptly proximally or, at least, it is more pronounced than the medial one (Fig. 8P).

Phalanx III-3 (Fig. 8R–W):

CPI 1646 can be attributed confidently to digit III because of the symmetry of the distal condyles and the proximal articular surface. The specimen is not a proximal phalanx because it presents a keel on the proximal surface and is here identified as phalanx III-3 based on the position of the collateral ligament pits, which are dorsally displaced, and the lack of a dorsal extensor fossa. The left phalanx III-3 is complete and almost symmetrical axially. Compared with phalanx III-1, this is smaller, measuring 67 mm in length, but stouter and proportionally broader transversely. It is taller and wider proximally and distally. The neck of phalanx III-3 especially constricts the pedal element.

The dorsal surface is convex, but concave in lateral or medial view (Fig. 8T, W). There is no extensor fossa on the dorsal surface of this phalanx (Fig. 8S). The ventral surface is flat, but it is also concave in medial or lateral view (Fig. 8T, W). Two rounded and considerably broad proximoventral crests are present, and they are bounded medially and laterally by a rugose area (Fig. 8T, V, W). Proximally and centrally positioned on the ventral surface, there is a process. The proximal end of the process is missing, but the preserved part is broad, rounded, and proximodistally oriented. The proximal articular surface is triangular in outline in proximal view, with a well-developed and centred proximodorsal lip (Fig. 8U). A vertical median keel separates the lateral and medial articular facets. Both articular facets are equally developed and subsymmetrical.

In medial and lateral views, the proximoventrally located fossae are less developed in phalanx III-3 than in phalanx III-1 (Fig. 8T, W). On the medial side, the fossa is restricted to a triangular rugose patch. On the lateral margin, this fossa is oval and very small. Both collateral ligament pits are more dorsally placed, in comparison to phalanx III-1. These pits are deep, subcircular to oval, and almost equally developed.

The distal condyles are not symmetrical. The medial condyle is more dorsoventrally developed compared with the lateral one (Fig. 8R). The long axis of the lateral condyle is laterally oriented, and it is medially directed in the medial condyle. In ventral view, the lateral condyle ends more abruptly proximally (Fig. 8V).

Phalanx IV-2 (or IV-3) (Fig. 8X–AB):

There is a proximal fragment of a left pedal phalanx IV-2 or IV-3 (CPI 1648). This identification is based on high asymmetry of the proximal articular surface, which indicates that it would be from the digit IV. Furthermore, the presence of a keel on the proximal articular surface and the relatively large size of the element suggest that it would not be the distalmost non-ungual phalanx of digit IV. This fragment broadens and increases in height proximally. The neck of the phalanx would also contract the bone (Fig. 8X, Y, AA, AB). At this level, it is triangular in cross-section with rounded margins.

The dorsal surface of this phalanx is convex, but it is concave in lateral or medial view (Fig. 8Y, AB). The preserved ventral surface shows a smooth fossa, which is bounded by two proximoventral ridges (Fig. 8AA). The lateral ridge is broader than the medial one. Close to the proximal rim, there is a rugose surface. At the proximal rim, there is a centred proximal process. In proximal view, the proximal articular surface is triangular and bears a median keel that separates both articular surfaces for the condyles (Fig. 8Z). The proximodorsal lip is highly pronounced and medially directed. The lateral articular facet is smaller but deeper than the medial one. The lateral articular facet is triangular (almost right-angled triangular) in proximal view, and the medial one is oval. Dorsal to the proximoventral ridges, there are no proximoventral fossae. In this area, a triangular rugose surface is present on both sides (Fig. 8Y, AB).

Phylogenetic results

Matrix from Rauhut and Pol (2021)

The phylogenetic analysis resulted in 10 000 most parsimonious trees, with a length of 5386 steps (CI = 0.180; RI = 0.614). The strict consensus (Supporting Information, Supplementary Material S2, Fig. S1) shows a rather well-resolved topology of general non-tetanuran relationships, including the monophyletic Ceratosauria. Although Tetanurae is also solved as monophyletic, the non-Coelurosauria tetanuran clades are poorly solved, with more completely known taxa, such as Allosaurus Marsh, 1877, AsfaltovenatorRauhut & Pol, 2019, Concavenator Ortega et al., 2010, or Spinosaurus, placed within a large and complex polytomy with several taxa and clades of different degrees of instability within Tetanurae. Piatnitzkysauridae, megalosaurines (Megalosaurus Buckland, 1824, Torvosaurus gurneyi Hendrickx & Mateus, 2014, Torvosaurus tanneri Galton and Jensen, 1979, and WiehenvenatorRauhut et al., 2016); some megaraptorans; metriacanthosaurids; some carcharodontosaurids; tyrannosaurids, and Coelurosauria are found as monophyletic groups within this large polytomy in Tetanurae. However, Riojavenatrix is located in this high-grade polytomy, branching early with the majority of other members of Carnosauria (Allosauridae + Megalosauridae clade, currently redefined by Rauhut and Pol 2021) and with all spinosaurids.

Therefore, we focus on the outcomes of the iterative PCR and other methodologies of pruning in order to analyse the unstable taxa that affect this complex polytomy and to determine whether Riojavenatrix is one of these unstable taxa, to evaluate the causes of this uncertainty, and to use this information to find possible solutions.

The reduced strict consensus from iterPCR (Fig. 9) was finished after a total of seven iterations. A total of 27 terminal taxa and one clade (megaraptorans) were pruned as unstable taxa, where Riojavenatrix was included (pruned in iteration 4). In iteration 3, several clades within the tetanuran polytomy were established as monophyletic, such as Megalosauroidea. Riojavenatrix was branched in a polytomy with other spinosaurids within Megalosauroidea. In the final reduced consensus, for the interrelationships of spinosaurids, Camarillasaurus, IrritatorMartill et al., 1996, Riojavenatrix, Sigilmassasaurus, and Vallibonavenatrix (Fig. 9) were found as unstable taxa in the MPTs owing to the fragmentary nature of the majority of members of Spinosauridae, with many characters coded as missing data. With these taxa removed, Spinosauridae are split into two groups, Baryonychinae (Baryonyx + Suchomimus) and Spinosaurinae (Ichthyovenator + Spinosaurus). In order to evaluate the position of Riojavenatrix, a reduced consensus with manual pruning (command punnelsen) of the unstable spinosaurids, except Riojavenatrix, was carried out. The results (in Supporting Information, Supplementary Material S2, Fig. S2 and S3) show that the polytomy is still a high-degree one and that spinosaurids still remain located within this polytomy. The alternative positions of the pruned taxa (Camarillasaurus, Irritator, Sigilmassasaurus, and Vallibonavenatrix) are as follows: (i) all of them as early branching tetanurans; (ii) all of them in a dichotomy with Riojavenatrix; (iii) Irritator related to Suchomimus in a dichotomy; (iv) Irritator located in a dichotomy with Baryonyx; and (v) Vallibonavenatrix related to Leshansaurus Li et al., 2009 in a dichotomy.

Using prunnelsen, to improve the establishment of clades for the node of Tetanurae (node 226 in TNT, considered as a polytomy of degree 38) and calculating prunings until three taxa, we can see that the most common spinosaurids pruned are Irritator, Riojavenatrix, and Vallibonavenatrix. These are the same results as in the iterPCR methodology, where these taxa were also considered as unstable taxa. The manual pruning of these taxa shows a reduced strict consensus, with a monophyly of Baryonyx and Suchomimus, where these three pruned taxa have several alternative positions. Riojavenatrix could be located as an early branching baryonychine or in a dichotomy with Suchomimus within Baryonychinae (Supporting Information, Supplementary Material S2, Fig. S2 and S3). The agreement subtree was not obtained by TNT owing to the numerous alternative results when dichotomies are forced.

The alternative position of the unstable taxa in spinosaurids (Irritator, Riojavenatrix, and Vallibonavenatrix), based on the results in the pruning methodology of TNT, were forced in several constraints (see Supporting Information, Supplementary Material S2). The constraint obtained by forcing Riojavenatrix as a baryonychine is rather well supported, requiring only three additional steps (length of 5389, 200 trees retained). When forcing the position of Riojavenatrix within Baryonychinae, the high-degree polytomy is mostly resolved, and this recovered most ‘traditional’ monophyletic groups within Tetanurae. One of the groups recovered as monophyletic is Spinosauridae, whose monophyly is also recovered in the reduced consensus tree from the iterPCR (Fig. 9). However, apart from the forced monophyly of baryonychines with Riojavenatrix, other monophyletic groups within Spinosauridae are not established. In this result, the sister taxa of Spinosauridae are a dichotomic group of Lourinhanosaurus Mateus, 1998 and Monolophosaurus Zhao & Currie, 1993. When forcing Riojavenatrix and Suchomimus together, the support has a similar result, also requiring only three additional steps (length of 5389, 200 trees retained). However, the position of this forced group has numerous alternative positions within Spinosauridae, probably for the splitting up of Baryonychinae.

The Bremer support analysis (all taxa included) shows many clades with minimal support. Tetanurae, including the high-degree polytomy, has a Bremer support of one. However, monophyletic groups, such as Megalosauridae, have high values of Bremer support (three and five). The results of support for jackknife and bootstrap analyses show that most of the groups have values <50%. Spinosaurids are all grouped together here, but with a low jackknife support value of two. In bootstrap, Baryonyx, Irritator, and Suchomimus are grouped together, with a support value of four, but the other spinosaurids are branched in the polytomy of tetanurans, which has a bootstrap value of 27 (see Supporting Information, Supplementary Material S2).

As was found out by the iterPCR and other methods of pruning, there are numerous unstable taxa in the matrix. This taxon instability could be one of the reasons for these low measures of support. Therefore, the position of all the unstable spinosaurids pruned in the reduced consensus by the iterPCR (see above) was ignored during another round of the jackknife and bootstrap support analyses. The results reveal that Spinosauridae improved their support values (jackknife values: 20 for Spinosauridae, 70 for Baryonychinae, and 37 for Spinosaurinae; bootstrap values: 8 for Spinosauridae, 52 for Baryonychinae, and 23 for Spinosaurinae; see Supporting Information, Supplementary Material S2). The appendicular skeleton has rarely been described in spinosaurids so far. The placement of Riojavenatrix within Spinosauridae in the phylogenetic results allows us to discuss those appendicular characters that might diagnose Spinosauridae and support the placement of Riojavenatrix within the clade. For that, appendicular characters have been mapped on the reduced consensus trees using Mesquite and in the strict consensus in TNT.

Riojavenatrix shares the following characters with megalosauroids within Carnosauria: faint scar of the muscle femorotibialis externus [character (ch.) 692; shared with Cryolophosaurus and some early branching theropods]; weak medial epicondyle of the femur (ch. 694; plesiomorphic condition present in Herrerasaurus Reig, 1963, Dilophosaurus wetherilli Welles, 1970, some coelophysoids, and some coelurosaurs); metatarsals <50% of tibial length (ch. 746; present in Afrovenator Sereno et al., 1994, Eustreptospondylus Walker, 1964, Riojavenatrix, and Spinosaurus, and some other theropods outside megalosauroids).

After mapping appendicular characters of Riojavenatrix on the reduced consensus trees and the previous anatomical comparison, the following unique combination of characters supports its placement as spinosaurid: the pubic apron extended from the middle of the pubic shaft (ch. 646; present also in Ichthyovenator; but absent in Spinosaurus and unknown in Baryonyx and Suchomimus; also present in megalosaurids and Allosaurus); deep oblique ligament groove on posterior surface of femoral groove (ch. 678; shared with most theropods, but absent in Afrovenator, Megalosaurus, and Torvosaurus tanneri); femoral head strictly directed medially in the anteroposterior plane (ch. 680; potential synapomorphy of Spinosauridae within megalosauroids, although also present in Afrovenator; outside megalosauroids, it is also present in allosauroids and Coelurosauria).

Synapomorphies in Baryonychinae that also are present in Riojavenatrix are as follows: posteromedially oriented long axis of medial condyle of femur in distal view (ch. 693; absent in Spinosaurus and other theropods, but unknown in Ichthyovenator; also recovered in other taxa outside Spinosauridae); incomplete ossification in the fibular crest (ch. 710; present in Suchomimus, but unknown in Baryonyx; absent in other megalosauroids); bluntly rounded vertical ridge as a buttress for astragalus in the tibia (ch. 719; present in Suchomimus and Riojavenatrix, and distinct from the rest of the megalosauroids and carnosaurs).

Synapomorphies in Spinosaurinae that also are present in Riojavenatrix are as follows: ascending process of astragalus offset from astragalar body by a pronounced groove (ch. 739; present in Spinosaurus and Riojavenatrix, absent in Suchomimus, and unknown in other spinosaurids; also present in Wiehenvenator within megalosauroids, and in some coelurosaurs).

Within Spinosauridae, baryonychines have the following synapomorphies that are absent in Riojavenatrix: absence of an expanded pubic boot (ch. 639; shared with Afrovenator within megalosauroids, some coelophysoids, some ceratosaurians, and alvarezsauroids); lateral malleolus projects far distal to medial malleolus (ch. 717; present in Suchomimus; absent in Spinosaurus and Riojavenatrix, but is poorly coded in the matrix); anterior portion of fibular proximal end mediolaterally wider than the posterior portion (ch. 724; present in Baryonyx and other megalosauroids, but distinct from Spinosaurus and Riojavenatrix, some allosauroids, and some Coelurosauria).

Within Spinosauridae, spinosaurines have following synapomorphies that are absent in Riojavenatrix: large and ovoid pubic obturator foramen (ch. 652; shared with Ichthyovenator and Spinosaurus, also present in metricanthosaurids and Concavenator); low proximal projection of cnemial crest (ch. 702; present in Spinosaurus, but unknown in Ichthyovenator and Baryonyx; and absent in Suchomimus, other megalosauroids, allosauroids, coelophysoids, and some ceratosaurians); low and rounded fibular crest (ch. 713; present in Spinosaurus and other theropods, such as Asfaltovenator, some ceratosaurians, and coelurosaurs; but absent in Suchomimus and Riojavenatrix within spinosaurids); presence of a deep oval fossa on medial surface of fibula (ch. 730; potential synapomorphy of spinosaurines within megalosauroids, but it is unknown in Ichthyovenator; also present in allosauroids); fibular facet on astragalus reduced and facing laterally (ch. 734; present in Spinosaurus, but absent from Riojavenatrix and other megalosauroids and carnosaurs, except for Mapusaurus Coria & Currie, 2006, but unknown in other spinosaurids).

The potential autapomorphy of Riojavenatrix within spinosaurids based on the phylogenetic analysis is as follows: height of ascending process of the astragalus more than twice the height of astragalar body (ch. 735; present in Riojavenatrix but not in Suchomimus and other megalosauroids; also present outside spinosaurids in Allosaurus, Tyrannosaurus Osborn, 1905, Albertosaurus Osborn, 1905, and some derived coelurosaurs).

Matrix from Mateus and Estraviz-López (2022)

The alternative phylogenetic analysis, using the matrix proposed by Mateus and Estraviz-López (2022), resulted in 307 MPTs, with a length of 1123 steps (CI = 0.550; RI = 0.573). The strict consensus (Fig. 10A) shows a well-resolved topology in most of the monophyletic groups of theropods. Regarding results in non-spinosaurid groups, Piatnitzkysaurus Bonaparte, 1979 is branched earlier in a polytomy of Carnosauria, and it is not recovered as a megalosauroid, unlike the results of Carrano et al. (2012). In Megalosauroidea, there is only a degree 5 polytomy formed by the megalosaurids Afrovenator, Eustreptospondylus, DuriavenatorBenson, 2008, and Dubreuillosaurus Allain, 2005 (Afrovenatorinae in the study by Carrano et al. 2012) and the dichotomy of Megalosaurinae. Megalosaurinae (Torvosaurus + Megalosaurus) is a monophyletic group in these results. Within Spinosauridae, baryonychines (Baryonyx + Suchomimus) are recovered in a dichotomy. The rest of the spinosaurids (including the new taxon Riojavenatrix) form a degree 7 polytomy within the group, where the node of Baryonychinae is also included.

Only three taxa are pruned from the iterPCR of this consensus as unstable and collapsing nodes in the strict consensus: Eocarcharia Sereno & Brusatte, 2008, Monolophosaurus, and Afrovenator. Other taxa are considered equally unstable, always forming polytomies even if they are pruned. These equally unstable taxa are as follows: Baryonyx, Iberospinus, Ichthyovenator, Irritator, Riojavenatrix, Spinosaurus, Suchomimus, and Vallibonavenatrix. Most of these taxa have scored characters that support alternative positions in different trees; therefore, these characters should be re-evaluated to analyse the instability of these taxa. In addition, most spinosaurids show numerous missing data in their coding, and if they were to be scored, they might help to resolve the positions (Pol and Escapa 2009). The IterPCR script analyses the missing data of each unstable taxon and evaluates their ancestral condition. If the missing characters have different optimization in the ancestral node of the problematic taxon in different MPTs, it implies that their scoring (if possible) might help to solve the instability of the taxon (Pol and Escapa 2009). This is the case for Riojavenatrix, where if ≤113 of all missing data were scored, they might help to resolve its position (for the list of these characters obtained in the iterPCR script results, see Supporting Information, Supplementary Material S2). However, all these characters are associated with missing bones or missing anatomical elements in the material for Riojavenatrix. If new material for this taxon become known in future, all these characters should be re-evaluated and scored in subsequent phylogenetic approaches to define a stable position for Riojavenatrix.

The agreement subtree (Fig. 10B) shows a dichotomy between Suchomimus and Baryonyx, which is the only dichotomy that is consistent in all the MPTs, and they are the only taxa that have been not considered as unstable and pruned. All the other spinosaurids are equally considered as unstable, and there are six possible combinations of agreement subtrees (see list of pruned taxa in Supporting Information, Supplementary Material S2). This result is consistent, in part with the iterPCR one, but the latter analysis considers all the spinosaurids as equally unstable (including Suchomimus and Baryonyx). Following the methodology for pruning trees incorporated in TNT (prunnelsen), node 31 was improved if Afrovenator was pruned, node 30 was improved if Monolophosaurus was pruned, and finally, node 29 was improved if Eocarcharia was pruned. The results from iterPCR and prunnelsen are similar regarding the unstable taxa and pruning (Afrovenator, Monolophosaurus, and Eocarcharia), because both methodologies have similar rules for stopping (Pol and Escapa 2009). However, equally unstable taxa in the polytomy (spinosaurids) are not taken into consideration in the pruning methodology of TNT, whereas they are in the iterPCR. Regarding the agreement subtrees, its algorithm continues pruning taxa until the subtree is dichotomous, and for that, this method does not prune the same most unstable taxa, like the other methods mentioned, and keeps Suchomimus and Baryonyx as stable taxa because they form a dichotomy in all the subsets. Therefore, although Iberospinus and Riojavenatrix are left in the agreement subtree, they should still be considered as unstable taxa, as was inferred from the iterPCR.

The Bremer support analysis (all taxa included) shows many clades with minimal support. This is the case for Spinosauridae, which show a Bremer support of two. Jackknife and bootstrap analyses have values <50% in some groups, especially within Megalosauroidea. Spinosaurids are supported by 35 (jackknife) and 14 (bootstrap) values (Fig. 10A).

The unambiguous synapomorphies (see Supporting Information, Supplementary Material S2) from the strict consensus for spinosaurids are as follows: presence of webbing at base of neural spines in dorsal vertebrae (ch. 179); accessory centrodiapophyseal lamina in dorsal vertebrae (ch. 180); and expanded infraprezygapophyseal fossa in dorsal vertebrae (ch. 181). However, these characters are missing data in Riojavenatrix. The ambiguous synapomorphies that are shared by Riojavenatrix and others spinosaurids are as follows (for the complete list, see Supporting Information, Supplementary Material S2): large and oval obturator foramen of pubis (ch. 285; present in Ichthyovenator and Riojavenatrix); expanded and triangular morphology of distal end of ischium (ch. 297; present in Ichthyovenator, Riojavenatrix, and Vallibonavenatrix, but absent in Baryonyx or Suchomimus); posteromedial orientation of medial condyle of femur in distal view (ch. 312; present in Baryonyx, Riojavenatrix, and Suchomimus); bluntly rounded vertical ridge on medial side as buttress for astragalar in the tibia (ch. 320; present in Riojavenatrix and Suchomimus); fibular flange is not extended to proximal end of the tibia (ch. 322; present in Riojavenatrix and Suchomimus); and almost double height of the ascending process with respect to the height of astragalar body [ch. 331; present in (ch. state 2): Riojavenatrix and Suchomimus].

Comparisons with Spinosauridae and other theropods

Results from phylogenetic analysis support a placement of Riojavenatrix within the Spinosauridae. This new taxon has a unique combination of spinosaurid characters supporting this placement, according to the phylogenetic analysis performed with the matrix of Rauhut and Pol (2021): (i) a pubic apron that extends from the middle of the pubic shaft; (ii) a deep oblique ligament groove on the posterior surface of the femoral groove; and (iii) femoral head strictly directed medially in the anteroposterior plane.

Within Spinosauridae, Riojavenatrix shares three synapomorphies with members of Baryonychinae in the first phylogenetic analysis: (i) the postermedially oriented long axis of the medial condyle femur in distal view; (ii) the incomplete ossification in the fibular crest; and (iii) the bluntly rounded vertical ridge as buttress for astragalus in the tibia. However, Riojavenatrix shares only one synapomorphy with members of Spinosaurinae: the ascending process of astragalus offset from astragalar body by a pronounced groove.

Its placement within Spinosauridae is supported by six ambiguous synapomorphies according to the results from the phylogenetic analysis performed with the matrix of Mateus and Estraviz-López (2022): (i) a large and oval pubic obturator foramen; (ii) expanded and triangular morphology of the distal end of the ischium; (iii) the posteromedially oriented long axis of the medial condyle femur in distal view (Baryonychinae synapomorphy based on the first phylogenetic analysis using the matrix of Rauhut and Pol 2021); (iv) the bluntly rounded vertical ridge as buttress for astragalus in the tibia (Baryonychinae synapomorphy based on the first phylogenetic analysis using the matrix of Rauhut and Pol 2021); (v) fibular flange is not extended to the proximal end of the tibia owing to incomplete ossification (Baryonychinae synapomorphy based on the first phylogenetical analysis using the matrix of Rauhut and Pol 2021); and (vi) the height of the ascending process of the astragalus being double the height of the astragalar body. Therefore, this unique combination of spinosaurid synapomorphies based on two phylogenetic analysis implies that Riojavenatrix can be regarded confidently as a spinosaurid.

The fossil record of Spinosauridae is, in most cases, fragmentary, with only a few well-known taxa. Comparison with the spinosaurid fossil record is limited by the absence of material overlapping between Riojavenatrix and the European Ceratosuchops and Riparovenator, the African Cristatusaurus and Sigilmassasaurus, and the South American AngaturamaKellner and Campos, 1996, Irritator, and Oxalaia Kellner et al., 2011. Nevertheless, the fossil remains of Riojavenatrix lacustris overlap with enough material of other African, Asian, and European spinosaurids (Charig and Milner 1997, Sereno et al. 1998, Allain et al. 2012, Ibrahim et al. 2014, Sánchez-Hernández and Benton 2014, Malafaia et al. 2018, 2020a, Mateus and Estraviz-López 2022) to allow us to make comparisons and evaluate the possible synonymy between them, especially between the herein described taxon and the other Iberian spinosaurids.

Pelvic girdle

Pubis:

Riojavenatrix preserves an obturator notch in the pubis, like several tetanurans and more derived theropods (Hutchinson 2001). The preserved diaphysis of the pubis in Riojavenatrix is straight and similar to that of Baryonyx (NHMUK VP R9951), Iberospinus (Mateus and Estraviz-López 2022), Ichthyovenator (MDS-Savannakhet BK10-11), and Suchomimus (MNN GDF500), and unlike the curved shaft of FSAC-KK 11888. In Iberospinus, a longitudinal groove extends along the pubic shaft, a feature that is also present, albeit subtly, in Ichthyovenator (Mateus and Estraviz-López 2022) and in Baryonyx (NHMUK VP R9951). This groove is not present in the preserved shaft of Riojavenatrix, because it is present at the proximal end of the pubic shaft. The pubic apron of Riojavenatrix would have reached almost to the pubic boot and extends much further distally than in Baryonyx (NHMUK VP R9951), Ichthyovenator (MDS-Savannakhet BK10-11), and Suchomimus (MNN GDF500). The distal position of the pubic apron of Riojavenatrix also differs from that of FSAC-KK 11888, because in the latter it does not reach the pubic boot. In distal view, the triangular-shaped distal end of the pubis of Riojavenatrix resembles that of Ichthyovenator (MDS-Savannakhet BK10-11) and megalosaurids (Fig. 11A, B). However, in Ichthyovenator (MDS-Savannakhet BK10-11) the lateral surface is much more concave (Fig. 11B), giving it an L-shaped distal outline (Allain et al. 2012), in comparison to the more triangular pubic boot of Riojavenatrix in distal view (Fig. 11A). This triangular-shaped distal end of the pubis is related to a strong mediolateral expansion of the anterior part of the pubic boot, which was proposed as a feature of Baryonychinae (Sereno et al. 1998, Allain et al. 2012). However, the pubic boots of FSAC-KK 11888, Baryonyx (NHMUK VP R9951), and Suchomimus (MNN GDF500) are much narrower mediolaterally throughout their whole anteroposterior length in distal view than those of Ichthyovenator (MDS-Savannakhet BK10-11) and Riojavenatrix (Fig. 11A–E). In Suchomimus (MNN GDF500) and Baryonyx (NHMUK VP R9951), a subtle L-shape can be noticed in distal view, whereas it is more oval in FSAC-KK 11888, differing from Riojavenatrix (Fig. 11A, C–E).

Ischium:

The iliac peduncle of the ischium of Riojavenatrix shows a planoconcave articulation, a feature shared with Baryonyx (NHMUK VP R9951), FSAC-KK 11888, and Suchomimus (MNN GDF500). According to Allain et al. (2012) and Malafaia et al. (2020a), the articulation of the ischium with the ilium is peg-and-socket in Ichthyovenator and Vallibonavenatrix. Nevertheless, the morphology of these iliac peduncles in the latter taxa is similar to that of other spinosaurids (i.e. planoconcave); and they do not have the well-developed and deeply excavated fossa on the iliac peduncle, referred to as the peg-and-socket articulation, featured in abelisauroids and carcharodontosaurians (Sereno et al. 2004, Carrano et al. 2012). Therefore, all spinosaurids have a planoconcave articulation of the ischium to the ilium, a feature also seen in Riojavenatrix. In proximal view, the iliac articulation is less anteroposteriorly elongate in Riojavenatrix than in Baryonyx (NHMUK VP R9951) and FSAC-KK 11888, resembling that of Ichthyovenator (MDS-Savannakhet BK10-12–13), Suchomimus (MNN GDF500), and Vallibonavenatrix (MSMCa-1–3). The latter two also have a medially positioned and anteroposteriorly oriented crest with an adjacent groove on the medial surface of the acetabulum (Malafaia et al. 2020a). In Baryonyx (NHMUK VP R9951), a similar anteroposteriorly oriented groove is present. This is also lateral and medially bounded by two low, rounded, and parallel ridges that, together with the groove, fade anteriorly. Despite Riojavenatrix having a similar crest, it lacks the adjacent groove. In lateral view, the preserved portion of the ischial diaphysis and the distal portion of the Riojavenatrix ischium are more slender than those of FSAC-KK 11888, Ichthyovenator (MDS-Savannakhet BK10-12–13), Suchomimus (MNN GDF500), and Vallibonavenatrix (MSMCa-1–3). The anterior margin of the shaft and the distal portion of the Riojavenatrix ischium are transversely proportionately as thick as in Suchomimus (MNN GDF500) and Vallibonavenatrix (MSMCa-1–3). Nevertheless, the ischial boot expands further anteroposteriorly with respect to the shaft in Riojavenatrix than in FSAC-KK 11888, Ichthyovenator (MDS-Savannakhet BK10-12–13), Suchomimus (MNN GDF500), and Vallibonavenatrix (MSMCa-1–3) (Fig. 11F–J). In FSAC-KK 11888, the distal expansion is posteriorly directed, differing further from the Riojavenatrix ischium. The ischial boot is rounded in lateral view in FSAC-KK 11888 and Suchomimus (MNN GDF500), whereas it is more triangular in Ichthyovenator (MDS-Savannakhet BK10-12–13), Vallibonavenatrix (MSMCa-1–3), and Riojavenatrix (Fig. 11F–J). Nevertheless, the anterodistal surface of the ischial boot is only angular in Riojavenatrix. Furthermore, the anterior tip of the ischial boot in Riojavenatrix resembles that of Megalosaurus (OUMNH J.13565; Benson 2010, fig. 15G–L). This tip is not as marked in Ichthyovenator (MDS-Savannakhet BK10-12–13), and it is not present in FSAC-KK 11888 and not preserved in Suchomimus (MNN GDF500) and Vallibonavenatrix (MSMCa-1–3) (Fig. 11F–J).

Hindlimb

Femur:

The femur of Riojavenatrix shows several similarities with megalosauroid theropods (Fig. 12), such as the rounded medial epicondyle and the small rugose patch for the attachment of the muscle femorotibialis externus (Carrano et al. 2012). Furthermore, like other spinosaurids, the medial condyle of the left femur of Riojavenatrix has a posteromedial orientation (Benson 2010), and not posterolateral as previously suggested by some authors (e.g. Benson 2010, Carrano et al. 2012, Malafaia et al. 2018). Nevertheless, the posteromedial displacement in the new taxon is remarkably much less marked than in Baryonyx (NHMUK VP R9951), FSAC-KK 11888, ‘Spinosaurus B’ (Nr. 1922 X 45; Stromer 1934), and Suchomimus (MNN GDF500) and more similar to CMP-3b/211 (Fig. 12A–F). Indeed, this condition is somewhere between that of these spinosaurids and other megalosauroids, such as Megalosaurus (NHMUK PV 31806; Benson 2010: fig. 16I–J). The distal end of the Riojavenatrix femur also differs from Baryonyx (NHMUK VP R9951) and FSAC-KK 11888 in the distal extension of the lateral condyle. This distal end projects further distally than the medial condyle in Baryonyx (Charig and Milner 1997, Carrano et al. 2012) and the right femur of FSAC-KK 11888. In Riojavenatrix, instead, the femoral condyles are almost equally projected, a feature also seen in CMP-3b/211, ‘Spinosaurus B’ (Nr. 1922 X 45), and Suchomimus (MNN GDF500). Furthermore, the femoral condyles in FSAC-KK 11888 are notably narrower (Fig. 12A, F), and its fourth trochanter is hypertrophied (Ibrahim et al. 2014). The main axis of the lateral condyle of Baryonyx (NHMUK VP R9951) and FSAC-KK 11888 is more anteroposteriorly directed and narrower than in Riojavenatrix (Fig. 12A, D, F). In Riojavenatrix, the lateral condyle is almost as rounded and broad as that of Suchomimus (MNN GDF500), with a similar orientation, and it is virtually identical to CMP-3b/211 (Fig. 12A, B, E). The extensor groove is also more developed in Baryonyx (NHMUK VP R9951), ‘Spinosaurus B’ (Nr. 1922 X 45), and FSAC-KK 11888 than in Riojavenatrix, and this is more similar to that of Suchomimus (MNN GDF500) and CMP-3b/211 (Fig. 12A–F). The flexor groove in Baryonyx (NHMUK VP R9951) is also wider and deeper in comparison to Riojavenatrix.

The head of the femur of Riojavenatrix is subcircular, as in CMP-MS-0/22, CMP-3b/211, Baryonyx (NHMUK VP R9951), and Suchomimus (MNN GDF500), and it differs from that of FSAC-KK 11888, which is more oval in medial view. The articular groove of the proximal surface of the right femur resembles that of CMP-MS-0/22 and CMP-3b/211, owing to the latter ones being relatively narrow and deep (Malafaia et al. 2018) (Fig. 12G–J). However, in Baryonyx (NHMUK VP R9951) and Suchomimus (MNN GDF500), the proximal articular grooves are relatively broader and not as restricted, at least, posterolaterally (Fig. 12I–J). Furthermore, this articular groove is more anteriorly located and anteroposteriorly oriented in Riojavenatrix (Fig. 12G). Riojavenatrix also lacks the longitudinal V-shaped groove present on the medial surface of the shaft of Baryonyx (Charig and Milner 1997). In the FSAC-KK 11888 femur, the crista tibiofibularis is more posteriorly projected in distal view than in Riojavenatrix (Fig. 12F). The CMP-MS-0/22 and CMP-3b/211 femora differ almost exclusively from those of Riojavenatrix in the bowing of the femoral shaft, with CMP-3b/211 being straighter in lateral view, and with a slightly broader tibiofibular crest.

Tibia:

The supraastragalar buttress of Riojavenatrix is a vertical ridge located on the medial side, a feature also shared with Chilantaisaurus Hu, 1964 (Benson and Xing 2008), CMP-3c/188, FSAC-KK 11888, and Suchomimus (Rauhut 2003, Benson and Xing 2008, Carrano et al. 2012, Malafaia et al. 2018). The shaft of the left tibia of Riojavenatrix is straight, whereas in CMP-3c/188 it is medially curved distally owing to the medially projected medial malleolus (Malafaia et al. 2018). In Suchomimus (MNN GDF500), the tibial shaft is laterally bowed, a feature also observed, but to a lesser extent, in FSAC-KK 11888. Unlike the tibia in Riojavenatrix, the shaft of the Camarillasaurus tibia is ‘G-shaped’ owing to a deep groove present posteriorly on its lateral surface (Sánchez-Hernández and Benton 2014). This structure is also described in the left tibia of FSAC-KK 11888 (Samathi et al. 2021). However, this could be the result of preservation or pathology (Samathi et al. 2021). Riojavenatrix differs further from CMP-3c/188 in lacking the concavity situated posterior to the fibular crest that bears a tibial foramen (Malafaia et al. 2018). Suchomimus (MNN GDF500) has a proximodistally oriented groove that is parallel and anterior to the fibular flange. In Camarillasaurus, this is a subtle depression and has a distally located foramen (Sánchez-Hernández and Benton 2014). These are not present in Riojavenatrix. The cnemial crest follows the same pattern of Camarillasaurus, FSAC-KK 11888, and Suchomimus, where the tibia narrows towards the anteriormost part of this structure (Samathi et al. 2021). However, in Riojavenatrix, the tip of the cnemial crest is more laterally directed, differing from that of Camarillasaurus (MPG-KPC8), FSAC-KK 11888, ‘Spinosaurus B’ (Nr. 1922 X 45), and Suchomimus (MNN GDF500) (Fig. 13A, C–E), and being more similar to other non-spinosaurid theropods, such as Allosaurus (USNM 4734; Gilmore 1920: fig. 49), Condorraptor Rauhut, 2005, FukuiraptorAzuma & Currie, 2000 (FPMN 9712220; Azuma and Currie 2000: fig. 13C), Megalosaurus (NHMUK PV 31809; Benson 2010: fig. 17G), SinraptorCurrie & Zhao, 1993 (IVPP 10600; Currie and Zhao 1993: fig. 22I), and Tyrannosaurus (Brochu 2003). In addition, the cleft between both tibial condyles is narrower in Riojavenatrix than in Camarillasaurus (MPG-KPC8) (‘intercondylar groove’ of Sánchez-Hernández and Benton 2014), ‘Spinosaurus B’ (Nr. 1922 X 45), and FSAC-KK 11888, and it is more similar to that present, for instance, in Suchomimus (MNN GDF500) and Megalosaurus (Benson 2010) (Fig. 13B–E). Furthermore, in Riojavenatrix the lateral process of the lateral condyle of the tibia is proximodistally longer and narrower than the shorter and triangular one of Camarillasaurus (MPG-KPC8).

Distally, the supraastragalar buttress is more marked and thinner in Riojavenatrix than in CMP-3c/188. This crest smoothly inclines laterally in CMP-3c/188 and rises from the medial edge of the tibia, whereas it is more angular and with its distalmost end being located on the anterior side of the tibia in Riojavenatrix. This feature is similar in Suchomimus (MNN GDF500), but the tibia is slightly damaged at this point. The distal portion of the supraastragalar buttress is more inclined in FSAC-KK 11888 than in Riojavenatrix. The facet for the ascending process of the astragalus of the Riojavenatrix tibia is not as broad as in CMP-3c/188 and Suchomimus (MNN GDF500), occupying slightly more than half of the anterior surface of the tibia distally, but it extends almost as high as in the latter taxa. The mediolateral extension of this facet in Riojavenatrix is certainly more similar to FSAC-KK 11888, but in the latter the apicalmost point of the facet is more centrally placed. On the posterior side, the CMP-3c/188 tibia shows a marked concavity between the lateral malleolus and a vertical triangular crest (Malafaia et al. 2018), which is a slight depression in Riojavenatrix. The angle between the tibial malleoli is also much higher in CMP-3c/188 than in Riojavenatrix, being more angular, straighter, and almost equally directed distally in Riojavenatrix. In this feature, the tibia of Riojavenatrix resembles the specimen FSAC-KK 11888 and somewhat that of Suchomimus (MNN GDF500). However, the lateral malleoli of FSAC-KK 11888 and Suchomimus (MNN GDF500) are slightly more distally projected than in Riojavenatrix. In posterior view, the distal margin between the tibial malleoli of Riojavenatrix is straight, as in Suchomimus (MNN GDF500). In FSAC-KK 11888, there is a concavity, as in CMP-3c/188 (Malafaia et al. 2018), which is more pronounced in the latter specimen. The distal expansion of the Riojavenatrix tibia is not as pronounced as in Suchomimus, because the lateral malleolus is larger and projects further laterally in Suchomimus (MNN GDF500) than in Riojavenatrix. This projection in Riojavenatrix resembles that of CMP-3c/188 and FSAC-KK 11888.

Fibula:

In comparison to other spinosaurids, the fibula of Riojavenatrix is more slender. As in other megalosauroid theropods, the fibula of Riojavenatrix shows a shallow medial fossa (sensuCarrano et al. 2012) or lacks the medial depression (sensuBenson 2010 and Rauhut et al. 2016), differentiating them from most of other neotheropods. In proximal view, the Riojavenatrix fibula shows a comma-shaped outline. However, this is posteriorly straighter than in Wiehevenator (WMN P27479 and 27502) and it thins posteriorly more than in Baryonyx (NHMUK VP R9951; note that the better-preserved Baryonyx fibula is here regarded as a left one), Camarillasaurus (MPG-KPC unnumbered), and FSAC-KK 11888. Moreover, the fibula is more C-shaped in proximal view in Baryonyx (NHMUK VP R9951; crescent-shape of Charig and Milner 1997) and FSAC-KK 11888. In medial view, the posterior margin of the proximal end in Baryonyx is more proximally projected than the anterior edge in Riojavenatrix. In Baryonyx (NHMUK VP R9951), it is also more elevated but less pronounced, and distint from Camarillasaurus (MPG-KPC unnumbered) and FSAC-KK 11888, which have an almost even proximal end. The medial fossa is deeper in FSAC-KK 11888 and proportionally shallower in Baryonyx (NHMUK VP R9951) compared with Riojavenatrix. Instead, Riojavenatrix has a medial fossa on the fibula that resembles that of Suchomimus (MNN GDF500). Nevertheless, the distal end of the fossa is more acute in Riojavenatrix than in Suchomimus (MNN GDF500), but more rounded than that of Baryonyx (NHMUK VP R9951) and FSAC-KK 11888. This fossa extends more distally in Baryonyx (NHMUK VP R9951) and FSAC-KK 11888 than in Riojavenatrix, which is similar to Suchomimus (MNN GDF500). The proximal expansion of the fibula is more abrupt in Riojavenatrix than in FSAC-KK 11888, especially the posterior blade. This blade is blunter in Riojavenatrix than in Baryonyx (NHMUK VP R9951). Below the medial fossa, the iliofibularis tubercle is not especially marked in Riojavenatrix and thus similar to the condition observed in FSAC-KK 11888 and Suchomimus (MNN GDF500). In Baryonyx (NHMUK VP R9951), this is even less noticeable and lacks the groove present in the herein described taxon.

Astragalus:

The distal condyles of the astragalus are anterodistally projected in Riojavenatrix with an angle similar to other tetanurans (Sereno et al. 1996, Carrano et al. 2012). In addition to the herein described one, only one other, still undescribed, spinosaurid astragalus is known (MNN GDF, unnumbered and referred to Suchomimus according to Rauhut 2003). The ascending process of the latter astragalus is taller than that of Allosaurus (Sereno et al. 1998). In Suchomimus, the ascending process of the astragalus is higher than the astragalar body, a feature also proposed for Spinosauridae among early branching tetanurans (Carrano et al. 2012). Indeed, the ascending process for the astragalus is 1.6 times taller than its body in that referred to Suchomimus (Carrano et al. 2012). The ascending process of the astragalus of Riojavenatrix is taller than its body, like the one referred to Suchomimus. Nevertheless, this is even taller in Riojavenatrix, doubling the height of the astragalar body. Owing to the absence of other spinosaurid astragali, it is difficult to compare the Riojavenatrix astragalus with other taxa. Therefore, it is not possible to assess whether the vertical ridge on the medial margin of the ascending process of the astragalus is an autapomophy of Riojavenatrix or a synapomorphy among Spinosauridae.

Calcaneum:

The Riojavenatrix calcaneum is smaller in size than those of Baryonyx (NHMUK VP R9951) and Iberospinus (ML1190-31) (Fig. 14). In particular, this calcaneum is mediolaterally narrower and higher than that of Iberospinus (ML1190-31) (Fig. 14A, B). Moreover, the calcaneum of Riojavenatrix bears a foramen located in an anterolateral depression, which is so far present only in this taxon. The calcaneum of Baryonyx (NHMUK VP R9951) differs from the one of Riojavenatrix in having a centrally located large depression and several, but smaller, foramina, and from Iberospinus (ML1190-31) because the latter lacks any of them (Fig. 14A–C). Furthermore, the astragalar facet of Baryonyx bears two fossae (Charig and Milner 1997) that are not present in Riojavenatrix.

Pes:

Pedal elements are scarce and usually fragmentary among Spinosauridae. The hourglass-shaped proximal end of the third metatarsal of Riojavenatrix is also present in other tetanurans (Carrano et al. 2012). This metatarsal is robust in Riojavenatrix, resembling the shape of those of Chilantaisaurus (Benson and Xing 2008), Sinraptor (Currie and Zhao 1993) and Torvosaurus (Britt 1991). In addition, unlike avetheropods, the cross-section of the shaft of the metatarsal III of Riojavenatrix is rectangular, similar to early branching tetanuran theropods (Carrano et al. 2012). Charig and Milner (1997) mentioned the presence of two distal ends of metatarsal bones belonging to Baryonyx, indicating that they could belong to metatarsal II, III, or IV. Despite these being fragmentary, they have been identified here as the distal ends of the metatarsals II and III. The metatarsal that is not ginglymoid would correspond to the distal end of metatarsal III and the other to the distal end of metatarsal II, owing to the latter showing a ventral groove and being mediolaterally broad in distal view compared with the distal ends of metatarsals IV. Furthermore, the metatarsal assigned to digit II is too dorsoventrally low for it to be considered to be from digit IV. In distal view, the third metatarsal of Riojavenatrix is not ginglymoid, a condition also present in other megalosauroids, such as Baryonyx (NHMUK VP R9951), Megalosaurus (Benson 2010), and other theropods (Norell et al. 2001). In Riojavenatrix, the distal condyle is much more mediolaterally broad than in the metatarsal III of Baryonyx (NHMUK VP R9951). Furthermore, this is square-shaped and dorsoventrally larger than mediolaterally wide in Baryonyx (NHMUK VP R9951). Moreover, the ventral margin of the distal condyle of the metatarsal III of Riojavenatrix is concave, whereas it is flat in Baryonyx (NHMUK VP R9951).

Many spinosaurid pedal unguals from Gondwanan formations have been reported, and all of them show a flat ventral side (see Stromer 1934, Novas et al. 2005, Ibrahim et al. 2014, Maganuco and Dal Sasso 2018, de França et al. 2021). Ibrahim et al. (2014) reconstructed the pes of FSAC-KK 11888, also with a flat-bottomed I-2. This is not the case for phalanx I-2 of Riojavenatrix, which is remarkably curved. Furthermore, the pedal unguals of FSAC-KK 11888 are broader than deep, differing from the preserved phalanx I-2 of Riojavenatrix. The pedal ungual phalanx of Baryonyx (NHMUK VP R9951) does not show a flat ventral side and bears a longitudinal groove on each side. The ventral surface of Iberospinus (Mateus and Estraviz-López 2022) is flatter than that of other European spinosaurids, but its preservation does not permit a more precise comparison. A single longitudinal groove in pedal unguals is a plesiomorphy found among early branching theropods, with the exception of Abelisauroidea, in which two bifurcating grooves are present (Carrano and Sampson 2008). The I-2 pedal ungual of Riojavenatrix shows only one longitudinal groove on the lateral side, whereas it is devoid of grooves on the medial side. Therefore, it differs from most other theropods in this feature [e.g. Alectrosaurus Gilmore, 1993 (Mader and Bradley 1989); Bambiraptor Burnham et al., 2000; Deinonychus Ostrom, 1969; Tyrannosaurus (Brochu 2003): FMNH PR2081; and an indeterminate carcharodontosaurian from Portugal (Malafaia et al. 2019); Fig. 15]. But also, it differs from all the spinosaurid ungual phalanges described so far, which have a longitudinal groove on the medial side (see Stromer 1934, Charig and Milner 1997, Novas et al. 2005, Ibrahim et al. 2014, Maganuco and Dal Sasso 2018, de França et al. 2021, Mateus and Estraviz-López 2022). Therefore, the presence of a longitudinal groove on the lateral surface of pedal phalanx I-2 and the absence of it on the medial side might be an autapomorphy of Riojavenatrix. Nevertheless, the absence of pedal unguals that can be attributed confidently to digit I in other spinosaurids, or even in Megalosauroidea, prevents this assertion with certainty, because it can also be a synapomorphy of any of the latter groups.

Histological analysis

Bone sections show, in general, a 5- to 10-mm-thick, moderately thin cortex of primary compact bone, which grades into a large medullary region completely filled by bony trabeculae (Fig. 16A, B). The bone microstructure is too poorly preserved for a precise histological assessment, owing to the intense diagenesis and low-grade metamorphism experienced by the sediments of the Enciso Group (see ‘Geographical and geological setting’). Therefore, the bone matrix microstructure and the type of primary vascularity are not recognizable. However, some general patterns of the bone microstructure can be identified in the recovered material. In this sense, clusters of Haversian systems are clearly identifiable in the innermost region of the compacta (Fig. 16E), where at least two overlapping generations of secondary osteons are visible. Isolated secondary osteons and large resorption rooms incompletely filled by lamellar bone (i.e. incipient secondary osteons) are also present on the peripheral regions of the compacta. When present, they tend to be organized in circular rows, in the same manner as that seen in other groups of dinosaurs (Company 2011, Cerda et al. 2017).

Other histological structures recorded in the cortex consist of well-defined lines of arrested growth, which result in a cortical stratification. They are more evident in the external region of the compacta, where the secondary remodelling is less invasive (Fig. 16B–D). The presence of these growth lines indicates that the animal grew with a seasonal cyclicity, probably annual (D’Emic et al. 2023 and references therein). Given that the subperiosteal surface of the bones has been partly eroded away by abrasion, it is not possible to determine whether the external fundamental system, a proxy of somatic maturity composed of closely spaced resting lines, was once present in the samples.

The presence of large erosion spaces coated only by a thin layer of lamellar bone (i.e. immature Haversian osteons) and only two generations of secondary osteons, with scarce overlapping, indicates that the processes of osteonal remodelling was not complete and was still active when the animal perished. This fact, along with the presence of intense secondary reconstruction in the internal perimedullary region, but scarce towards the outer cortex, combined with the existence of cyclical growth rings, points to a somatically immature, still growing adult.

DISCUSSION

Phylogenetic discussion and implications in Spinosauridae

Spinosauridae, as a family, gathers members that are fragmentary, aside from them lacking plenty of overlapping material. However, some taxa are well represented, such as Baryonyx, Irritator, Suchomimus, and FSAC-KK 11888. This fragmentary nature of the spinosaurid record is one of the main causes of the instability of this clade in the phylogenetical analyses. Most of the representative taxa of Spinosauridae are considered unstable in the parsimonious analyses of the two studied matrices. Three of these unstable taxa are Iberian representatives (Camarillasaurus, Riojavenatrix, and Vallibonavenatrix) based on the matrix of Rauhut and Pol (2021); and Riojavenatrix, Iberospinus and Vallibonavenatrix in the case of the matrix of Mateus and Estraviz-López (2022). This scarce record leads to a lot of missing entries in these matrices. Some of these characters might help to determine the phylogenetic position of the unstable taxa if the scoring was established. In the case of Riojavenatrix, there are 113 characters (from a total of 534) that might help to resolve its position if they were scored based on the iterPCR method in the matrix of Mateus and Estraviz-López (2022). The scoring of these characters is not yet possible for Riojavenatrix because the elements are missing, but the analysis of which characters could be supporting this instability allows us to evaluate the instability of this taxon and the hypothetical reason for it, in order to improve it in future approachs, re-evaluating all the problematic characters. However, the only record of this new taxon comprises mainly bones from the pelvic girdle and the hindlimb, although a new individual is currently under study, which was recovered in a nearby locality. Despite the Riojavenatrix holotype not being a very complete skeleton, the recovered elements of it overlap significantly with other members of Spinosauridae, most importantly with all of those described in the Iberian Peninsula. Therefore, this allows us to make an in-depth comparison of this specimen with other members of Spinosauridae and to justify the establishment of a new taxon.

Although some of the features present in Riojavenatrix typically show up in Spinosauridae, some aspects are worth commenting. Spinosaurids show the ‘broadly triangular’ outline of the pubic boot in distal view (Carrano et al. 2012; ch. 290). However, among spinosaurids this shape varies considerably. In Baryonyx, Suchomimus, and FSAC-KK 11888, the pubes are lateromedially compressed and show a somewhat L-shape (sensuAllain et al. 2012). Ichthyovenator still shows the L-shaped pubic boot (Allain et al. 2012), but it is more triangular. If compared with these taxa, the Riojavenatrix pubic boot is virtually triangular in outline and fairly robust. These features are similar to those characters found in the megalosaurids Afrovenator (MNBH TIG 1), Eustreptospondylus (OUMNHJ.13558), and Streptospondylus Meyer, 1832 (MNHN 8605).

The orientation of the medial condyle of the Riojavenatrix holotype femur is also unusual. This is posteromedially oriented, which is considered to be a synapomorphy of Spinosauridae (Charig and Milner 1997, Benson 2010, Carrano et al. 2012). This synapomorphy (ch. 311:1) was also listed from the results here obtained from the phylogenetic analysis of Mateus and Estraviz-López (2022). Nevertheless, the extent of this shifting compared with other non-spinosaurid theropods is much more marked in Baryonyx (NHMUK VP R9951), Suchomimus (MNN GDF500), ‘Spinosaurus B’ (Nr. 1922 X 45), and, at least, the right femur of FSAC-KK 11888, and has an intermediate position in Riojavenatrix. The other Iberian spinosaurid femur from the Arcillas de Morella Formation (CMP-3b/211) shows this same trait. Given that the slightly posteromedially oriented medial condyle is present in two femora, each from a different individual, a possible palaeopathological influence on this character can be ruled out. Furthermore, this intermediate stage in the Riojavenatrix femora, together with the similar distal outline of its pubic boot to other megalosaurid megalosaurians, might point to the hypothesis that Riojavenatrix could be an earlier branching spinosaurid.

Riojavenatrix is younger than Baryonyx, Camarillasaurus, Ceratosuchops, Iberospinus, and Riparovenator, and most probably Vallibonavenatrix, coeval with Ichthyovenator and Suchomimus (note that the age of Ichthyovenator and Suchomimus is not very accurate), and older than the more derived mid-Cretaceous African and South American spinosaurids (Charig and Milner 1997, Sereno et al. 1998, Allain et al. 2012, Sánchez-Hernández and Benton 2014, Sales and Schultz 2017, Malafaia et al. 2020a, Barker et al. 2021, Mateus and Estraviz-López 2022). Therefore, in the European basins there are seemingly more derived baryonychine spinosaurids (i.e. Baryonyx, Ceratosuchops, and Riparovenator), with the latter two being closely related to Suchomimus (Barker et al. 2021), and Suchomimus is a baryonychine that already shows both of the more derived features (i.e. a more mediolaterally narrow pubic boot and a more posteromedially oriented medial condyle of the femur). If additional skeletal remains of more complete specimens support this hypothesis, this spinosaurid would be one of the youngest baryonychines described hitherto and still retaining some plesiomorphic features.

Spinosaurid palaeobiodiversity in the Iberian Peninsula

The Iberian Peninsula has yielded many spinosaurid remains that can shed light not only on the Iberian spinosaurid palaeobiodiversity, but also on the composition of Western Europe theropod faunas. The oldest Iberian putative spinosaurid remains come from the upper Hauterivian–lower Barremian Pinilla de los Moros Formation (Fuentes-Vidarte et al. 2001, Canudo and Ruiz-Omeñaca 2003). They consist of fragmentary elements regarded as Baryonyx sp. from the El Juguete site (Fuentes-Vidarte et al. 2001). Nevertheless, this specimen does not belong to the genus Baryonyx and, possibly, not even to Spinosauridae, owing to the combination of the presence of a ridge on the anterior surface of the shaft of the chevron, instead of a groove, and the absence of paired anterior processes on the proximal end of the chevron. The described chevron (MDS-Salas de los Infantes JBS,1) lacks an anterior longitudinal groove, which is present in Baryonyx, Camarillasaurus, SM-KK14 (Charig and Milner 1997, Samathi et al. 2021), and FSAC-KK 11888. In MDS-Salas de los Infantes JBS,1, instead, the anterior surface is convex, with a ridge appearing distally, similar to the chevrons assigned to Vallibonavenatrix (Malafaia et al. 2020a). Furthermore, MDS-Salas de los Infantes JBS,1 bears anterior processes, a feature apparently not present in Spinosauridae (Carrano et al. 2012). It has been reported that these processes are not present in the preserved haemal arches of Baryonyx, Riparovenator, Suchomimus, SM-KK14 (Charig and Milner 1997, Barker et al. 2021, Samathi et al. 2021), and apparently, in Vallibonavenatrix. In Camarillasaurus (MPG-KPC45, 60, and 63), most of the preserved chevrons lack the anterior process, but the left anterior process is present in MPG-KPC47. Likewise, some of the Baryonyx (NHMUK VP R9951) and FSAC KK 11888 (O.W.M. Rauhut, pers. comm.) chevrons seem to show anterior processes, but they are not as developed as in the El Juguete theropod. An isolated tooth studied by Torcida et al. (1997) from the Tenada de Costalomo site (MDS-Salas de los Infantes CLST,2), which is contemporaneous to the El Juguete specimen, would be the oldest definitive Iberian spinosaurid skeletal remains described to date, if the El Juguete specimen is not regarded as a spinosaurid.

In the lower Barremian deposits of the Iberian Peninsula, spinosaurid remains become rather common, specifically in the Lusitanian and Maestrazgo basins. These remains mainly comprise isolated spinosaurid teeth (e.g. Infante et al. 2005, Ruiz-Omeñaca et al. 2005, Ruiz-Omeñaca 2006, Buffetaut 2007, Sánchez-Hernández et al. 2007, Gasca et al. 2008, 2011, 2018, Mateus et al. 2011, Figuereido et al. 2015, Alonso and Canudo 2016, Alonso et al. 2018), but also postcranial remains (e.g. Gasca et al. 2018) from the Blesa, Camarillas, El Castellar, Mirambel, and Papo Seco formations (for further information, see Isasmendi et al. 2020, Malafaia et al. 2020b). Late Barremian skeletal remains regarded as Spinosauridae are also abundant, represented by isolated teeth (e.g. Ruiz-Omeñaca et al. 1998, Canudo et al. 2004, 2008) and, less commonly, postcranial remains (Malafaia et al. 2018) from the Arcillas de Morella and Artoles formations in the Maestrazgo Basin (for further information, see Isasmendi et al. 2020, Malafaia et al. 2020b). In the Cameros Basin, late Barremian spinosaurid remains have also been described in different formations, but these might also be early Aptian in age (see e.g. Pereda-Suberbiola et al. 2003, Torcida Fernández-Baldor et al. 2003, Alonso et al. 2017). The Enciso Group has also been dated as latest Barremian–early Aptian (Suarez-Gonzalez et al. 2013). However, given that the baryonychine remains described by Isasmendi et al. (2020, 2023) were recovered in the middle part of the group, they are highly likely to be Aptian in age, probably being the youngest spinosaurid remains described in Iberia to date.

Despite most of the remains being fragmentary, several specimens have allowed five different Iberian spinosaurid genera to be erected, aside from Baryonyx. Camarillasaurus cirugedae was initially regarded as a basal ceratosaur (Sánchez-Hernández and Benton 2014), but later identified as a member of Spinosauridae (Rauhut et al. 2019, Barker et al. 2021, Samathi et al. 2021). Its phylogenetic position within Spinosauridae is still unresolved, being regarded as a spinosaurine by Barker et al. (2021), or an early member of Spinosauridae by Samathi et al. (2021). The results after the iterPCR of the matrix of Rauhut and Pol (2021) locate Camarillasaurus as a member of Spinosaurinae, with two alternative positions in this clade: (i) as an early spinosaurine; or (ii) as more closely related to Ichthyovenator than to Spinosaurus. The results of the reduced consensus by manual pruning are not enough to evaluate the position within the clade, because the Spinosaurinae were not recovered. This specimen comes from the Fuente Arnar outcrop of the lower Barremian Camarillas Formation (Maestrazgo Basin) (Sánchez-Hernández and Benton 2014).

Iberospinus natarioi is another early Barremian Iberian spinosaurid (Mateus and Estraviz-López 2022). This specimen, recovered from the Papo Seco Formation (Lusitanian Basin), was initially assigned to Baryonyx walkeri, but its unusual features were also highlighted by Mateus et al. (2011). However, in a later revision and with the study of more material, a new taxon was erected (Mateus and Estraviz-López 2022). These authors proposed that Iberospinus is closely related to baryonychines rather than to spinosaurines. This taxon is not included in the matrix of Rauhut and Pol (2021), but in our second analysis, which uses the matrix of Mateus and Estraviz-López (2022), Iberospinus and Riojavenatrix are closely related in the agreement subtree, supporting that the Portuguese taxon is more likely to be a baryonychine than a spinosaurine.

The third Iberian spinosaurid, Vallibonavenatrix cani, comes from the younger Arcillas de Morella Formation, which is late Barremian in age (Malafaia et al. 2020a). This genus was initially positioned within Spinosaurinae (Malafaia et al. 2020a). This position is also supported by Mateus and Estraviz-López (2022). However, in the phylogenetic analysis performed by Barker et al. (2021), the genus is recovered as a baryonychine or as an early spinosaurid, sister taxon of the node of Baryonychinae and Spinosaurinae. In our results, based on the matrix of Rauhut and Pol (2021), the reduced consensus by iterPCR recovers Vallibonavenatrix in three different alternative positions: (i) as an early member of Spinosaurinae; (ii) as a spinosaurine more closely related to Ichthyovenator than to Spinosaurus (both results similar to those of Malafaia et al. 2020a); but also (iii) as a megalosaurid forming a clade with Dubreuillosaurus and Leshansaurus (the position related to Leshansaurus is also observed in the consensus reduced by manual pruning). Also from the same formation, the taxon Protathlitis cinctorrensis was erected (Santos-Cubedo et al. 2023) and was recovered as the earliest baryonychine in the phylogenetic analysis carried out by Santos-Cubedo et al. (2023).

The last erected genus is the herein described one, Riojavenatrix lacustris. The holotype individual was recovered from the uppermost Barremian–lower Aptian Enciso Group. However, the stratigraphic position of the Virgen del Villar-1 locality within the group suggests an early Aptian age for Riojavenatrix. This spinosaurid was previously regarded as Baryonyx (Viera and Torres 2013), but the osteological differences allow this specimen to differ from the British taxon. Here, in the reduced consensus by iterPCR, Riojavenatrix is recovered in three alternative positions: (i) as a baryonychine related more closely to Baryonyx than to Suchomimus; (ii) as a baryonychine related more closely to Suchomimus than to Baryonyx; and (iii) as a spinosaurine related more closely to Ichthyovenator than to Spinosaurus (probably owing to the morphology of the ischial boot that is shared with Vallibonavenatrix and Ichthyovenator and supports this alternative position). Despite the fragmentary nature of these genera in comparison to other well-known theropods, the existing overlapping material supports the presence of five different spinosaurid taxa in Iberia (see also Malafaia et al. 2020a, b, Mateus and Estraviz-López 2022).

Furthermore, the current and better-known spinosaurid record does not back up the presence of Baryonyx in Iberia. Even the isolated teeth show differences that prevent their attribution to the genus. The teeth from the Castrillo de la Reina and Pinilla de los Moros formations of the Western Cameros Basin (MDS-Salas de los Infantes C-15,30; C-15,32; CLST,2; TBMV,13) were assigned to cf. Baryonyx by Torcida et al. (1997). These teeth have flutes on both lingual and labial surfaces, which is a feature not present in most of the teeth of the holotype of Baryonyx (Charig and Milner 1997, Hendrickx et al. 2019). Indeed, only one tooth shows both fluted sides, with a single flute on the labial surface (Hendrickx et al. 2019). These features are present (i.e. the presence of both fluted lingual and labial surfaces) in the baryonychine isolated tooth from the Boca do Chapim site (CPGP.1.06.2), which was previously attributed to Baryonyx sp. by Figuereido et al. (2015) and, at least, the tooth of a right dentary fragment (MG 29A; Buffetaut. 2007), all from the Papo Seco Formation. Furthermore, another baryonychine isolated tooth (MNHN/UL.I.F2.176), also from the Boca do Chapim site, has at least five flutes on the labial surface (Malafaia et al. 2013, 2020b), also differing from the teeth of the Baryonyx holotype. In the light of these considerations, the specimens differ significantly from the Baryonyx teeth; hence, they can be identified assuredly as baryonychine teeth, but not from the genus Baryonyx. Gasca et al. (2018) also described an isolated manual ungual phalanx from the El Castellar Formation (CSC1-4), regarding it as aff. Baryonyx sp. and already suggested that it would probably belong to a spinosaurid different from Baryonyx. As the Iberian and European spinosaurid framework is becoming more complex, we consider here that CSC1-4 might be better regarded as Spinosauridae indet. Therefore, all the fossil material previously attributed to Baryonyx does not belong to this genus, but to indeterminate baryonychines or spinosaurids.

The presence of different spinosaurid taxa in a restricted area and time interval has also been reported, for instance, in south-eastern England, with Ceratosuchops and Riparovenator from the Barremian Wessex Formation, and the large-sized indeterminate spinosaurid from the overlying upper Barremian Vectis Formation of the Wessex sub-basin, and Baryonyx from the Barremian Upper Weald Clay Formation of the Weald sub-basin (Charig and Milner 1986, 1997, Barker et al. 2021, 2022). The same has been suggested in other Gondwanan localities, such as the Araripe Basin, which has yielded holotype specimens of the spinosaurines Angaturama and Irritator (Kellner and Campos 1996, Martill et al. 1996, Sales and Schultz 2017; but for the possible synonymy, see e.g. Charig and Milner 1997, Sereno et al. 1998, 2022, Buffetaut and Oujada 2002, Dal Sasso et al. 2005), and also in the Kem Kem beds (Evers et al. 2015), the Baharyia Oasis (Stromer 1934), and the Elrhaz Formation (Taquet and Russell 1998).

In Iberia, different baryonychine tooth morphotypes have been distinguished depending on the extension of the mesial carina and/or the presence or absence of mesial denticles (e.g. Ruiz-Omeñaca et al. 1996, Canudo and Ruiz-Omeñaca 2003, Ruiz-Omeñaca 2006, Gasca et al. 2008, Isasmendi et al. 2020). This could suggest the presence of different baryonychine taxa or could be attributable to pseudoheterodonty (Canudo and Ruiz-Omeñaca 2003, Ruiz-Omeñaca 2006, Isasmendi et al. 2020). Regardless, the presence of baryonychines and spinosaurines in the Lower Cretaceous deposits is well documented via dental elements (e.g. Infante et al. 2005, Sánchez-Hernández et al. 2007, Alonso et al. 2017, 2018), even in the same site, such as La Cantalera-1 (Barremian, lower Blesa sequence; Aurell et al. 2018), where both isolated baryonychine and spinosaurine teeth have been described (Alonso and Canudo 2016). This supports the presence of at least two coeval spinosaurid taxa in Iberia. Concerning the Iberian spinosaurid taxa, Camarillasaurus and Iberospinus come from lower Barremian deposits from different basins (Sánchez-Hernández and Benton 2014, Mateus and Estraviz-López 2022), whereas Vallibonavenatrix and Protathlitis are late Barremian (Malafaia et al. 2020a) and Riojavenatrix early Aptian in age. Protathlitis and Vallibonavenatrix have only been reported in the Maestrazgo Basin, whereas Riojavenatrix is present in the Cameros Basin.

CONCLUSIONS

The description of a partial skeleton from the uppermost Barremian–lower Aptian deposits of the Enciso Group (DS7) of Igea (La Rioja, Spain) has led to the erection of a new spinosaurid theropod genus and species, Riojavenatrix lacustris. This is the first theropod erected from the Cameros Basin and provides new knowledge regarding the barely known spinosaurid palaeodiversity of the Iberian Peninsula. Despite a poor preservation of Riojavenatrix bones, given that the finest histological structures have been obliterated by diagenetical alteration, certain poorly preserved growth marks and osteonal remodelling processes remain visible in the bone. This suggests a somatically immature, still growing subadult individual. The phylogenetic results obtained propose that Riojavenatrix is tentatively positioned as a baryonychine, although an alternative position within Spinosaurinae is also recovered. Moreover, Riojavenatrix and most other spinosaurids are considered unstable taxa in the phylogenies. Riojavenatrix corresponds to a medium- to large-sized spinosaurid dinosaur that differs from the other members of the clade owing to the combination of the following: a lateromedially thick and triangular pubic boot in distal view; an anteroposteriorly expanded and angular ischial boot with an anterodorsally oriented tip; a narrow, restricted and relatively deep proximal articular groove of the femur, which is anteromedially–posterolaterally inclined; a slightly posteromedially oriented medial femoral condyle; a longitudinal ridge on the medial margin of the ascending process of the astragalus (potential autapomorphy, but it could also be a character of Spinosauridae); height of ascending process of the astragalus more than twice the height of the astragalar body (potential autapomorphy, but it could also be a character of Spinosauridae); an anterior depression with a dorsally located foramen on the lateral surface of the calcaneum; and the absence of any longitudinal groove on the medial surface of phalanx I-2 (potential autapomorphy, but it could also be a character of Spinosauridae or even Megalosauroidea). Furthermore, the Riojavenatrix pubis and femur show features not as derived as other members of Spinosauridae. Spinosaurids were common in the Barremian faunas of Iberia, with the oldest definitive spinosaurid material (i.e. isolated baryonychine teeth) coming from the upper Hauterivian–lower Barremian deposits of the Western Cameros Basin, and the youngest (i.e. Riojavenatrix lacustris) from the uppermost Barremian–lower Aptian of Eastern Cameros Basin. Furthermore, after the revision of other Iberian fossils regarded until now as from the genus Baryonyx, the presence of this genus in Portugal and Spain has been dismissed. Therefore, only five spinosaurid genera are here considered to be present in the Iberian Peninsula: Camarillasaurus, Iberospinus, Protathlitis, Riojavenatrix, and Vallibonavenatrix.

[Version of Record, published online 19 February 2024; https://zoobank.org/urn:lsid:zoobank.org:pub:214C8AFF-46EF-4BEC-814D-417F0AE66659]

ACKNOWLEDGEMENTS

The studied specimen was recovered during an emergency excavation in 2005 with permission from the Consejería de Educación, Cultura, Deporte y Juventud (Dirección General de Cultura y Turismo, Servicio de Conservación y Promoción del Patrimonio Histórico Artístico) of La Rioja Government. We thank the Dirección General de Cultura (Servicio de Conservación y Promoción del Patrimonio Histórico Artístico) for issuing the excavation permit and financing the fieldwork, and the Reserva de la Biosfera de los valles Leza, Jubera, Cidacos y Alhama and Consejería de Sostenibilidad, Transición Ecológica y Portavocía del Gobierno de La Rioja for their collaboration. We would also like to acknowledge the technical staff of the Centro de Interpretación Paleontológica de La Rioja Adrián Páramo, Alba Marco and Adrián Blázquez for preparation of the specimen and for sharing photographs of it. We also thank the late José Angel Torres for all this arduous work and his commitment to the project. We further acknowledge José Ignacio Canudo (Aragosaurus, Unizar), Phil J. Currie (University of Alberta), Elisabete Malafaia (UL, UNED), Oliver Rauhut (SNSB-BSPG), and Ronan Allain (MNHN) for their helpful advice and for providing photographs of the comparative material. Special thanks are owed to Susannah Maidment (NHM), Nour-Eddine Jalil (MNHN), José Miguel Gasulla (UNED), Miguel Angel Herrero (MPG), and Juan Cano (MSM) for providing access to specimens under their care, and to the Mesozoic Tetrapod work group (main institutions in LMU and BSPG, Germany) for allowing us to use their character data matrix. We are very grateful to Scott Hartman for letting us use his skeletal diagram for Riojavenatrix. Finally, we would like to thank the field crew ‘Garras’ for their hard work during field seasons in the Virgen del Villar-1 site and the referees Elisabete Malafaia, Oliver W. M. Rauhut, and one anonymous reviewer.

AUTHOR CONTRIBUTIONS

Erik Isasmendi (Conceptualization, Data Curation, Formal analysis, Investigation, Project Administration, Writing-Original Draft Preparation, Writing-Review and Editing), Elena Cuesta (Conceptualization, Data Curation, Formal analysis, Investigation, Project Administration, Writing-Original Draft Preparation, Writing-Review and Editing), Ignacio Diaz-Martinez (Conceptualization, Data Curation, Formal analysis, Investigation, Project Administration, Writing-Original Draft Preparation, Writing-Review and Editing), Julio Company (Conceptualization, Data Curation, Formal analysis, Investigation, Writing-Original Draft Preparation, Writing-Review and Editing), Patxi Saez-Benito (Conceptualization, Investigation, Writing-Original Draft Preparation, Writing-Review and Editing), Luis I. Viera (Writing-Review and Editing), Angelica Torices (Supervision, Writing-Review and Editing), Xabier Pereda-Suberbiola (Conceptualization, Formal analysis, Investigation, Funding Acquisition, Supervision, Writing-Original Draft Preparation, Writing-Review and Editing).

CONFLICT OF INTEREST

None declared.

FUNDING

This work was supported by the Spanish Ministry of Science, Innovation and Universities and the European Regional Development Fund (projects CGL2017-85038-P and PID2021-122612OB-I00, MINECO/FEDER, UE); the Basque Government/Eusko Jaurlaritza (research groups IT418-19 and IT1485-22), and the University of the Basque Country (UPV/EHU, group PPG17/05). Erik Isasmendi is supported by a PhD fellowship of the Basque Government/Eusko Jaurlaritza (PRE_2019_1_0215). Ignacio Díaz-Martínez is funded by Unión Europea-NextGenerationEU ‘Requalification of Spanish university system for 2021–2022’ in the UPV-EHU, and supported by a Ramón y Cajal fellowship (RYC-2022), and by the Ministry of Science and Innovation of Spain.

DATA AVAILABILITY

All data is incorporated into the article and its online supplementary material. The measurements of the holotype of Riojavenatrix lacustris are gathered in Supporting Information, Supplementary Material S1 and the phylogenetic data are included in Supporting Information, Supplementary Material S2. The data matrices are available in Supporting Information, Supplementary Files S1 and S2.

REFERENCES