Abstract
A new somphospondylan titanosauriform from the Lower Cretaceous of Spain is described from the remains found at the Sant Antoni de la Vespa site (upper Barremian Arcillas de Morella Formation) located in Morella. Garumbatitan morellensis gen. et sp. nov. is diagnosed by 11 autapomorphies and eight local autapomorphies; and our phylogenetic analyses suggest that Garumbatitan morellensis might correspond to an early-branching somphospondylan. The presence of several somphospondylan traits in Garumbatitan morellensis supports the somphospondylan hypothesis. The phylogenetic distribution of some titanosauriform and somphospondylan novelties in the femur (markedly developed lateral bulge, high shaft eccentricity, linea intermuscularis cranialis, and trochanteric shelf) is discussed. The tarsus and pes of Garumbatitan morellensis are distinctive, being characterized by the loss of the calcaneum, relative slenderness of the metatarsals II, III, and IV when compared to the retracted metatarsals I and V, three pedal phalanges in digit IV, and reduced ungual III. The sauropod fauna of the Iberian Peninsula during the Hauterivian–Aptian shows a complex phylogenetic mosaic, including forms with Laurasian affinities, mainly titanosauriforms (Soriatitan, Garumbatitan, and possibly Tastavinsaurus and Europatitan), and Gondwanan affinities, the rebbachisaurid Demandasaurus. Faunal exchange during the Early Cretaceous between the Europe, North America, East Asia, and Africa is plausible.
INTRODUCTION
Titanosauriformes are a diverse and geographically widespread group of sauropods during the Late Jurassic to Cretaceous that ranged from the middle Oxfordian to the Cretaceous–Palaeogene boundary (e.g. Upchurch et al. 2004, D’Emic 2012, Mannion et al. 2013, 2017, Carballido et al. 2015, Gorscak and O’Connor 2016). The early evolution of the clade, as well as the early diversification of Somphospondyli and Titanosauria, is still poorly known, and several non-titanosaurian titanosauriforms have been recovered in distinct phylogenetic positions, depending on the morphological dataset, resulting in different palaeobiogeographic scenarios. Some non-titanosaurian titanosauriform groups, such as Brachiosauridae and Euhelopodidae, only began to receive cladistic support in more recent works (e.g. Wilson 2002, Upchurch et al. 2004, 2015, Royo-Torres 2009, Wilson and Upchurch 2009, D’Emic 2012, 2013, Mannion et al. 2013, 2017, 2019a, Carballido and Sander 2014, Carballido et al. 2015, Poropat et al. 2015, 2016, D’Emic et al. 2016).
The first unequivocal osteological evidence of Titanosauriformes is dated from the Late Jurassic, specifically from the middle Oxfordian, with the brachiosaurid Vouivria damparisensis (Lapparent 1943, Mannion et al. 2017). The early evolution of this clade is marked by the diversification of Brachiosauridae and Somphospondyli (e.g. D’Emic 2012, Mannion et al. 2013, 2017, Mocho et al. 2019a). Brachiosauridae span up to the Early–Late Cretaceous boundary, i.e. late Albian/early Cenomanian (Chure et al. 2010, D’Emic et al. 2016, Mannion et al. 2017), being present in Europe, Africa, and North America during the Late Jurassic and apparently restricted to North America and Europe, and possibly South America, Africa, Asia, and Middle East, in the Early Cretaceous (e.g. Buffetaut et al. 2006, Carballido et al. 2015, McPhee et al. 2016, Mannion et al. 2017, Liao et al. 2021). Brachiosauridae include the Late Jurassic Giraffatitan brancai, Lusotitan atalaiensis, Brachiosaurus altithorax, Vouivria damparisensis (Mannion et al. 2013, 2017, 2019b, Mocho et al. 2017a) and possibly Galveosaurus herreroi (Pérez-Pueyo et al. 2019) and Europasaurus holgeri (considered a non-titanosauriform by: Sander et al. 2006, Carballido and Sander 2014, Carballido et al. 2015, 2020, a brachiosaurid by D’Emic 2012, Mannion et al. 2013, 2017, 2019b, D’Emic et al. 2016). This group is also represented by some Early Cretaceous forms, such as Cedarosaurus weiskopfae, Venenosaurus dicrocei, Sonorasaurus thompsoni, Abydosaurus mcintoshi, Soriatitan golmayensis (e.g. Chure et al. 2010, D’Emic 2012, Royo-Torres et al. 2012, 2017a, D’Emic et al. 2016, Mannion et al. 2017, 2019a), and possibly Padillasaurus leivaensis (Carballido et al. 2015) and Tastavinsaurus sanzi (see: Royo-Torres et al. 2012, 2017a).
On the other hand, the evolutionary history of the early somphospondylans and titanosaurs remains uncertain, but the discovery of new taxa, the re-description of previously described forms, and the establishment of new morphological datasets have greatly improved our knowledge about this sector of the macronarian phylogenetic tree (e.g. Wilson and Upchurch 2009, Mannion and Calvo 2011, Zaher et al. 2011, D’Emic 2012, Mannion et al. 2013, 2017, 2019a, b, Carballido and Sander 2014, Carballido et al. 2015, Poropat et al. 2015, 2016, 2021, 2023, Gorscak and O’Connor 2016, Martínez et al. 2016, Torcida Fernández-Baldor et al. 2017, González Riga et al. 2018, Mocho et al. 2019a, Gallina et al. 2021). Oceanotitan dantasi, identified in the upper Kimmeridgian levels of the Praia de Amoreira-Porto Novo Formation in Portugal (Mocho et al. 2019a), and Australodocus bohetii from the Upper Jurassic levels of the Tendaguru Formation in Tanzania (see: Mannion et al. 2013, 2019a) are possibly the older members of Somphospondyli. Unfortunately, only some non-titanosaurian somphospondylans are represented by relatively complete specimens [e.g. Euhelopus zdanskyi, Wilson and Upchurch (2009); Phuwiangosaurus sirindhornae, Suteethorn et al. (2009)], and several Early Cretaceous forms have shown distinct phylogenetic positions within, or even outside, Somphospondyli, such as Tastavinsaurus sanzi, Padillasaurus leivaensis, Chubutisaurus insignis, Sauroposeidon proteles, Wintonotitan wattsi, Ligabuesaurus leanzai, Angolatitan adamastor, and the members of Euhelopodidae, an endemic group of East Asia (e.g. Royo-Torres 2009, Carballido et al. 2011a, 2015, D’Emic 2012, 2013, Mannion et al. 2013, 2017, 2019a, Carballido and Sander 2014, Poropat et al. 2015, 2016, D’Emic et al. 2016, Moore et al. 2020, Bellardini et al. 2022), resulting in a lack of strong consensus on the topology of early branching somphospondylans. The Early Cretaceous is also marked by the origin and early diversification of titanosaurs, which reached a wide geographic distribution in the Barremian–Aptian and became the dominant component of sauropod faunas in the Late Cretaceous (e.g. D’Emic 2012, Gorscak and O’Connor 2016).
The Early Cretaceous is a key period to better understand the diversification of the Somphospondyli clade. The study of the sauropod fossil record from the Lower Cretaceous, including the description of new titanosauriform specimens, is necessary to understand the diversification of somphospondylans in this period and the composition of the Early Cretaceous sauropod fauna, which also include some non-macronarian lineages, such as turiasaurs, rebbachisaurids, and dicraeosaurids (e.g. Wilson and Allain 2015, Mannion et al. 2017, Royo-Torres et al. 2017a, b, Coria et al. 2019, Mannion 2019, Allain et al. 2022). The Iberian Peninsula has recently been an important source of sauropod discoveries in Hauterivian–Aptian (Lower Cretaceous) rocks, particularly material referred to Titanosauriformes (including members of Somphospondyli) (e.g. Canudo et al. 2008, Royo-Torres 2009, Santos-Cubedo et al. 2010,Royo-Torres et al. 2012, 2017a; Mocho et al. 2017b, Torcida Fernández-Baldor et al. 2017) and Rebbachisauridae (Torcida Fernández-Baldor et al. 2011). These new discoveries are shedding light on the palaeobiogeographic relationships of the European sauropod taxa and help to identify possible events of faunal isolation and dispersal (Torcida Fernández-Baldor et al. 2011, Royo-Torres et al. 2017a). Until now, four sauropod taxa have been described from the Lower Cretaceous levels of the Iberian Peninsula: the rebbachisaurid Demandasaurus darwini (Torcida Fernández-Baldor et al. 2011), and three titanosauriforms, the somphospondylans Tastavinsaurus sanzi (Canudo et al. 2008) and Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017), and the brachiosaurid Soriatitan golmayensis (Royo-Torres et al. 2017a). However, other important fossil sites with partial titanosauriforms have been found, such as El Canteret (Morella; Sanz et al. 1982, Yagüe et al. 2001), Sant Antoni de la Vespa (Morella; Ortega et al. 2006, Gasulla et al. 2011, 2012, Mocho et al. 2016, 2022), and Mas de Palau (Morella; Gasulla et al. 2021).
In this study, we present the first detailed systematic analyses of the titanosauriform remains found in the fossil site of Sant Antoni de la Vespa (Morella, Castelló) from the Arcillas de Morella Formation (upper Barremian). The Arcillas de Morella Fm. was deposited in the Maestrat Basin being rich in vertebrate fossil sites, many of which are notable for their dinosaur fauna (e.g. Sanz et al. 1982, Ortega et al. 2006, Santos-Cubedo et al. 2009, 2010, Gasulla et al. 2011, 2012, 2015, 2021). In this unit, three new dinosaurs have been established: the styracosternan ornithopod Morelladon beltrani (Gasulla et al. 2015), and the spinosaurid theropods Vallibonavenatrix cani (Malafaia et al. 2019) and Protathlitis cinctorrensis (Santos-Cubedo et al. 2023). Several sauropod specimens have been identified from this formation, including some found before the 1980s (e.g. Sanz et al. 1982, Sanz 1984, 1996, Ruiz-Omeñaca and Pereda-Suberbiola 1999, Ruiz-Omeñaca et al. 2003, Pérez-García et al. 2009, Mocho et al. 2017b) and several specimens found since then, some of which comprise partial skeletons (e.g. Sanz et al. 1982, Yagüe et al. 2001, Ortega et al. 2006, Gasulla et al. 2008, 2011, 2012, 2021, Santos-Cubedo et al. 2010, Mocho et al. 2016). Since the 1980s, not much attention was paid to these specimens until more recent studies, where they started to be described in detail (e.g. Mocho et al. 2017b).
The fossil site of Sant Antoni de la Vespa (Morella, Castelló) was discovered in 1998 and fieldwork was carried out in 2005 and 2008. Remains of at least four sauropod individuals were collected, three of which are titanosauriforms, making this locality one of the most productive fossil sites for titanosauriform remains in the Lower Cretaceous strata of Europe. The most complete titanosauriform individual comprises an articulated posterior cervical-to-anterior dorsal vertebrae sequence, dorsal ribs, anterior to posterior caudal vertebrae, chevrons, interclavicle, pelvis elements, and hindlimb elements, including two almost complete pedes. Near this specimen, ribs and hindlimb elements of at least, two smaller individuals were identified, including two almost complete hindlimbs. All comparable elements referable to this titanosauriform present a similar morphology, suggesting that they belong to members of the same taxon. The Sant Antoni de la Vespa titanosauriform was recently related with Somphospondyli, sharing several morphological features with Tastavinsaurus sanzi (late Barremian–early Aptian, Spain) (Mocho et al. 2016). Here, we provide a detailed description and systematic analysis of all material that is available for study, including part of the preserved elements allocated to the largest specimen and all the remains from the smallest one, which probably corresponds to a subadult individual. The systematic study of the Sant Antoni de la Vespa titanosauriform provides important data on the evolutionary history of the Iberian titanosauriforms and improves our knowledge of the faunal composition of the Barremian–Aptian terrestrial ecosystems established in the Maestrat Basin region (Spain). This study also includes the establishment of a new titanosauriform taxon and proposes an updated phylogenetic framework for the evolution of somphospondylans in the Iberian Peninsula.
Geological settings
The Sant Antoni de la Vespa fossil site is located in the eastern sector of the Iberian Range (Fig. 1A). The Iberian Range is an intraplate alpine fold belt, developed between the Late Cretaceous (Maastrichtian) and the early–middle Miocene (Langhian–Serravalian). This alpine fold belt is the result of the tectonic inversion of the Iberian Trough, which incorporates a Mesozoic marine basin placed in the eastern sector of the Iberian Plate. The formation of this basin was a consequence of a Mesozoic crustal thinning stage (Salas and Casas 1993). According to these authors the evolution of the tectonic subsidence in the Iberian Trough, without keeping in mind the sedimentary load, shows two rifting stages: (i) Late Permian–Late Triassic and (ii) Late Jurassic–Early Cretaceous. During the second rifting stage the Iberian Trough was divided into several basins (Cameros, Central-Iberian, Maestrat, and Iberian-Levantine). Important thicknesses of continental and shallow-marine siliciclastic deposits were formed in these basins.
Geographical and geological situation of the type locality of Garumbatitan morellensis gen. et sp. nov.. A, simplified geological map showing the location of Morella (Castellón, Spain) and of the Sant Antoni de la Vespa Quarry. B, simplified diagram of litho-stratigraphic framework of chronostratigraphic and lithostratigraphic units of the Maestrat Basin (based on: Bover-Arnal et al. 2016). C. NW–SE cross-section across in the region of Sant Antoni de la Vespa, detailing the lithostratigraphy of Arcilla de Morella Formation.
The Maestrat basin
The Maestrat and Iberian-Levantine basins were placed in the eastern part of Iberia in direct connection with the north margin of Tethys Sea. They were separated by the Valencian Massif (Vilas et al. 1982). The sedimentary record of the Maestrat Basin is 6.5 km in thickness (Salas 1987). These materials outcrop in the Linking Zone, a structural domain of the Iberian Range. The Linking Zone is structured in a set of E–W folds and thrusts verging towards the North. During the second rifting stage, a set of ENE–WSW listric faults that concern the upper crust determined the compartmentalization of the western and southern parts of the Maestrat Basin (Guimerà and Salas 1996, Salas et al. 2020) in a group of small sub-basins (Oliete, Las Parras, Galve, Cedramán, Penyagolosa, and Orpesa). The central sector of the Maestrat Basin was divided into two half-grabens (Morella and La Salzedella), by the listric fault of Montalbán–Herbers and by the Turmell fracture zone (Salas et al. 2003). In the half-graben of Morella were deposited the materials of the Arcillas de Morella Formation, whereas in the La Salzedella half-graben were deposited the materials of the Margas de Cervera Formation (Salas et al. 2003).
The Arcillas de Morella Formation at Sant Antoni de la Vespa
Grey marls, white sandstones, and red clays compose the Arcillas de Morella Formation. Locally, it contains conglomerates and polygenic breccias. This unit is Late Barremian in age (Garcia et al. 2014, Villanueva-Amadoz et al. 2014, Bover-Arnal et al. 2016; Fig. 1B). The Arcillas de Morella Fm. is at Sant Antoni de la Vespa 125 m in thickness. It is on top of sandy and bioclastic limestones of the Calizas y Margas de Artoles Fm. Towards the top, it passes gradually to bioclastic limestones of the Calizas y Margas de Xert Fm. The Artoles and Xert formations are shallow marine carbonate platform deposits. Thick beds of red clays of continental origin characterize the unit of Arcillas de Morella, formerly named ‘las capas rojas de Morella’ (red clays of Morella). These beds of red clays have been the object of mining works. Sant Antoni de la Vespa is one of the localities in which the Arcillas de Morella Fm. is thicker, and it presents different lithologies and stratigraphic features to those of other outcrops that have been studied in the area. In this way, the red clays are only 10% of the total thickness, and 50% of the section is formed by sandstones and marls as filling of big channels in which the remains of marine and continental fauna can be found mixed together. The presence of these channels, some of which are 30 m deep, implies the development of erosive events, during which an important part of the sedimentary record has been removed.
Stratigraphy
In the outcrops of Sant Antoni de la Vespa can be distinguished four units (Fig. 1B, C). Unit A has a variable thickness between 13 and 22 m. It is formed by intercalations of beds of red clays, gray marls, and white sandstones. Grey marls form the base of the unit. It contains a layer of bioclastic limestones with marine fauna. The sandstones are formed by tabular beds (up to 40 cm thick) and contain dinosaur footprints (Santos-Cubedo et al. 2014).
Unit B is limited between the top of the unit A, at the base, and an erosional surface at the top. Two associations of facies form this unit (B1 and B2). The deposits of facies B1 consist of three amalgamated sandstone palaeochannels. In outcrop, these palaeochannels can be observed in oblique section. They have a lateral continuity of 400 m and are 10 m thick. The deposits of facies B2 are 25 m thick. They come in a position both laterally and superior to those of the B1 association. They consist of intercalations of layers of grey marls, white sandstones, and red clays similar to those of unit A. They form a set of transgressive–regressive sedimentary sequences.
Unit C is limited to the base by an erosional surface. This surface has a terraced configuration and a vertical development of 43 m. This is the base of a big palaeochannel structure that can be followed along 700 m. The infilling of this palaeochannel is formed by an onlapping set of bodies of sandstones and grey marls. The sandstone layers are between 3 and 15 m thick, and they present a lateral asymmetric structure. They are leaning towards the SE in the erosion surface, whereas towards the NW they pass transitionally to grey marls. Internally they have large cross-planar or convex-concave stratification and cross and wave ripple lamination. The upper part (12 m of this infilling) is formed by a set of intercalations of tabular thin beds of sandstones and grey marls. The dinosaur fossil site of Sant Antoni de la Vespa is placed in these deposits of the upper part of unit C.
Unit D is 37 m thick. It is formed by grey silty marls with intercalations of thin beds of bioturbated sandstones. Locally the marls contain red clayely intervals that include carbonate palaeosols with nodular structure (Santisteban et al. 2016).
Sedimentology
The materials of the four units of the Arcillas de Morella Formation outcropping in the section of Sant Antoni de la Vespa are interpreted as deposits formed in a wave-dominated deltaic system (Poza et al. 2010, Santisteban et al. 2012, Santos-Cubedo et al. 2014). Most of the bodies of sandstone are beach deposits; however, many of them are part of the infilling of a big erosional structure that form a palaeochannel. These palaeochannels are incised valleys developed as a consequence of relative eustatic sea-level falls.
The whole of the sedimentary record of the Arcillas de Morella Formation is characterized by numerous transgressive–regressive sequences of fourth and fifth orders (Santisteban and Santos-Cubedo 2011, Santisteban et al. 2012, 2016). We consider these as parasequences. These are grouped in six depositional sequences, bounded by erosional surfaces. The incised valleys, as those that are in units B1 and C of Sant Antoni de la Vespa, are related with these erosional surfaces.
In the studied section, unit A is formed by four parasequences belonging to the same depositional sequence. Amalgamated sandstone channels of three different depositional sequences form unit B1. Unit C, in which is placed the dinosaur fossil site, is a complete depositional sequence. The terraced erosional surface is the base of an incised valley, and it represents the falling stage system tract. Most of their infilling is formed by four parasequences that contain the sandstone beach bodies. This part forms the transgressive system tract. The upper 12 m of the infilling, composed of an accretional set of tabular thin beds, are estuarine deposits that we identify as the high stand system tract (Santisteban et al. 2016).
The fossil site of Sant Antoni de la Vespa is placed in the middle part of the Arcillas de Morella Formation, 53 m above the base of the unit. The fossil remains were buried in an estuarine environment inside an incised valley, during a highstand period (Santos-Cubedo et al. 2014).
Sant Antoni de la Vespa fossil site
In 1998, local archaeologist Miquel Guardiola found the proximal end of a large sauropod femur during an archaeological survey near the hermitage of Sant Antoni de la Vespa, south-west of the mountain known as ‘Mola de la Garumba’ in Morella (Castelló Province). The first fieldwork in the locality was carried out in 2005 (Fig. 2C, D) and resulted in the finding of two carcasses with low disarticulation from two different individuals of titanosauriform sauropods (Fig. 3). The remains of the two specimens were deposited on top of each other, probably at two different times. The two complete hindlimbs, as well as the ribs and some elements of the forelimbs, were removed from the overlying specimen. From the largest specimen, stratigraphically below, a complete hindlimb and part of the other, as well as a set of articulated caudal vertebrae, were removed. In 2008, a second palaeontological campaign was conducted, extending the area excavated in 2005 to the east and west (Fig. 2A, B). Two main areas of intervention were identified during this campaign. Immediately east of the bones found in 2005, the remains of an incomplete and eroded ilium were recovered, as well as a series of articulated dorsal vertebrae and a possible cervical vertebra along with corresponding ribs. These remains are in continuity with, and compatible with, the largest specimen identified in 2005. Also in continuity with the 2005 operation, two series of caudal vertebrae were found to the west. The series slightly above corresponds to a series of middle-posterior caudal vertebrae, whereas the series below represents a series of articulated middle caudal vertebrae with their chevrons. There is no criterion to indicate that both series belong to one and the same individual, but their sizes are compatible with the largest specimen. After preparing the remains from this palaeontological site, it has been possible to verify the presence of a few elements of a third and a fourth individual that had not been detected during the fieldwork, and they are similar in size to the smallest individual identified in the first campaign.
Photographs from the fieldwork conducted in Sant Antoni de la Vespa fossil-site in the year of 2005 (C, D) and 2008 (A, B), including the holotype and paratype of Garumbatitan morellensis gen. et sp. nov.. A, B, holotype specimen with a partial articulated-to-associated dorsal vertebrae and ribs; C, right hindlimb from holotype specimen; D, left hindlimb from the paratype specimen.
Garumbatitan morellensis gen. et sp. nov., type site-map showing the approximate association of the bones (different mapping methodologies were used in the fieldwork of 2005 and 2008, resulting in a schematic map only showing the relative position of the main elements found in the field). The holotype specimen is in orange and the paratype one is in yellow. The elements collected in 2005 are indicated by dashed lines, and the elements collected in 2008 by continuous line.
Institutional abbreviations
AODF, Australian Age of Dinosaurs Fossil, Winton, Queensland, Australia; CM, Carnegie Museum of Natural History, Pittsburgh, Pennsylvania, USA; DMNH, Denver Museum of Natural History, Denver, Colorado, USA; FCPTD, Fundación Conjunto Palaeontológico de Teruel-Dinópolis (material with acronym ‘CT’ is deposited herein), Teruel, Spain (CPT for the Museo Aragonés de Palaeontología); FMNH, Field Museum of Natural History, Chicago, Illinois, USA; HMN, Humboldt Museum für Naturkunde, Berlin, Germany; IANIGLA, Instituto Argentino de Nivología, Glaciología y Ciencias Ambientales, Colección de Palaeovertebrados, Mendoza, Argentina; MCNA, Museo de Ciencias Naturales de Álava/Arabako Natur Zientzien Museoa, Vitoria-Gasteiz, Spain; MG LNEG, Museu Geológico do Laboratório Nacional de Energia e Geologia, Lisboa, Portugal; MLP, Museo de La Plata, La Plata, Argentina; MNCN, Museo Nacional de Ciencias Naturales, Madrid, Spain; MNHN, Muséum National d’Histoire Naturelle, Paris, France; MPG, Museo Palaeontológico de Galve (material with acronym ‘ZH’ is deposited herein), Spain; MPZ, Museo Palaeontológico of the Universidad de Zaragoza, Zaragoza, Spain; MUCPv, Museo de Geología y Palaeontología of the Universidad Nacional del Comahue, Neuquén, Argentina; MZSP, Museu de Zoologia of the Universidade de São Paulo, São Paulo, Brazil; NMB; Naturhistorisches Museum Braunschweig, Braunschweig, Germany; NMMNH, New Mexico Museum of Natural History and Science, Albuquerque, New Mexico, USA; PMU, Palaeontological Museum, University of Uppsala, Uppsala, Sweden; PVL, Fundación-Instituto Miguel Lillo, Tucumán, Argentina; SAV, Sant Antoni de la Vespa fossil site Collection of the Museu Temps de Dinosaures-Museus de Morella, Morella, Spain; UNCUYO-LD, Laboratorio de Dinosaurios of the Universidad Nacional de Cuyo (UNCUYO-LD), Mendoza, Argentina; YPM, Yale Peabody Museum of Natural History, New Haven, Connecticut, USA; ZPAL, Instytut Palaeobiologii PAN, Warszawa, Poland.
Anatomical and other abbreviations
2cc, second cnemial crest [see Bonaparte et al. (2000) and Mannion et al. (2013)]; acf, anterior to anteromedial crest of the fibular proximal end; acdl, anterior centrodiapophyseal lamina; acf, anterior chevron facet; acr, anterior crest; amcr, anteromedial crest; aEI, average elongation index [anteroposterior length of centrum (excluding articular ball) divided by the mean average value of the mediolateral width and the dorsoventral height of the posterior articular surface of the centrum; following Upchurch (1995, 1998) and Chure et al. (2010)]; amt, articulation for metatarsal; asp, ascending process; aspa, articular surface for the ascending process (lateral condyle in Mannion et al. 2017); cam, camella (sensuWedel et al. 2000); cc, cnemial crest; cdr, caudal rib; cmr, camera (sensuWedel et al. 2000); cr, crest; dcr, dorsal crest; dmcr, dorsomedial crest; dp, depression; ep, epicondyle; exp, expansion; f, fossa; fh, femoral head; fia, fibular articular surface; fic, fibular condyle; fr, foramen; ft, fourth trochanter; fct, facet; gr, groove; icg, intercondylar groove; icr, intercondylar ridge; lb, lateral bulge; lc, lateral condyle; lg, longitudinal groove; lic, linea intermuscularis cranialis; lr, lateral ridge; lt, lateral trochanter; mc, medial condyle; mco, medial concavity to the fourth trochanter; mli, medial lip; mt, metatarsal; nc, neural canal; pbu, posterior bulge; pcdl, posterior centrodiapophyseal lamina; pcf, posterior chevron facet; pcr, posterior crest; pf, posterior face; phx, phalanx; plf, posterolateral fossa; pnf, pneumatic fossa; podl, postzygodiapophyseal lamina; posl, postspinal lamina; poz, postzygapophyseal process; prj, proximal projection of the ventrolateral crest in the metatarsal II; prsl, prespinal lamina; prz, prezygapophyseal process; pvp, posteroventral process [medial condyle of Mannion et al. (2017)]; rdg, ridge; spof, spinopostzygapophyseal fossa; spol, spinopostzygapophyseal lamina; sprf, spinoprezygapophyseal fossa; sprl, spinoprezygapophyseal lamina; st, step; str, strias; tb, tuberosity; tia, tibial articular surface; tic, tibial condyle; tprl, intraprezygapophyseal lamina; ts, trochanteric shelf; vlcr, ventrolateral crest; vmcr, ventromedial crest; vlr, ventrolateral ridge.
Systematic palaeontology
We use ‘Romerian’ terms (Wilson 2006) for the anatomical structures (e.g. ‘centrum’) and their orientation (e.g. ‘anterior’). In this study, the landmark-based terminology of Wilson (1999, 2012) and Wilson et al. (2011a) is used for vertebral laminae and fossae. The position of caudal vertebrae is based on Díez Díaz et al. (2013a), Mannion et al. (2013), and Tschopp et al. (2015). The studied specimens of Garumbatitan morellensis gen. et sp. nov. described here are housed at the Museu Temps de Dinosaures-Museus de Morella (Morella, Castelló), which is part of Museums Survey of the Generalitat Valenciana (government of the Valencian Community). All the measurements are documented in the Supporting Information, File S1.
Dinosauria Owen, 1842
Sauropoda Marsh, 1878
Neosauropoda Bonaparte, 1986
Macronaria Wilson and Sereno, 1998
Titanosauriformes Salgado et al., 1997
Somphospondyli Wilson and Sereno, 1998
Garumbatitan gen. nov.
Zoobank: urn:lsid:zoobank.org:act:0E7654DD-1802-4D58-AF25-69E25446CBFA
Type and only included species
Garumbatitan morellensis sp. nov.
Etymology
The generic name is derived from Garumba, referring to the peak known as ‘Mola de la Garumba’ near the Sant Antoni de la Vespa fossil site, which is one of the highest points in the municipality of Morella; titan, giant in Greek mythology.
Type locality and horizon
As for type and only species.
Diagnosis
See diagnosis for type and only species below.
Garumbatitan morellensis sp nov.
(Figs 4–23, Supporting Information, File S1)
Zoobank: urn:lsid:zoobank.org:act:2D373FD9-6A97-453C-B2EB-8D5E259876FD
Holotype
The largest individual found in Sant Antoni de la Vespa, which comprises an articulated posterior cervical-to-anterior dorsal vertebrae sequence (SAV08-100) and a partial dorsal centrum (SAV08-040); dorsal ribs (SAV08-101, SAV08-102, SAV08-103), anterior-to-posterior caudal vertebrae (SAV05-027, SAV05-028, SAV05-029, SAV05-030, SAV05-060va, SAV05-061, SAV08-060-061-063-065-067-066-064-068-069-070-071, SAV08-047, SAV08-048, SAV08-049, and SAV08- 050), six chevrons (SAV05-060chb, SAV05-063, SAV05-060cha, and SAV05-060chc, two are included in SAV08-060-061-063-065-067-066-064-068-069-070-071), an interclavicle (SAV05-055), left (SAV05-023) and right (SAV05-024) femora, left (SAV05-025) and right (SAV05-065) tibiae, left (SAV05-026) and right fibulae (SAV05-064), right astragalus (SAV05-066), and an almost complete right pes (SAV05-068), and left metatarsal II (SAV05-021) and IV (SAV05-024). The articulated posterior cervical-to-anterior dorsal vertebrae sequence (SAV08-100) is still unprepared, as well as other jackets with elements from the holotype, including some dorsal ribs (SAV08-30-46), nine caudal vertebrae, and two chevrons (SAV08-060-061-063-065-067-066-064-068-069-070-071), a partial ilium (SAV08-104), left femur (SAV05-023), left tibia (SAV05-025), and left fibula (SAV05-026).
Etymology
morellensis refers both to the ‘Arcillas de Morella’ Formation and to the town of Morella, where some of the first dinosaur remains in Spain were found and where the Sant Antoni de la Vespa fossil site is located.
Type locality and horizon
Sant Antoni de la Vespa fossil site, Morella, Castelló (Spain), Arcillas de Morella Formation, late Barremian in age, Maestrat Basin.
Diagnosis
The somphospondylan titanosauriform Garumbatitan morellensis can be diagnosed by 11 autapomorphies (marked with an asterisk), as well as eight local autapomorphies: (i) lateral pneumatic fossae in the anterior caudal centra [also shared with the somphospondylan Savannasaurus elliottorum, Poropat et al. (2020)]; (ii) posterior articular surface is more deeply concave than anterior one in anterior-middle caudal centra; (iii) depression near the anterior edge of the centrum and above the lateral crest in the middle-posterior caudal vertebrae*; (iv) middle-posterior caudal vertebral pedicels covered by a complex of three anteroposteriorly elongated ridges*; (v) middle-posterior caudal neural spines are expanded posteriorly, which results in lateral rounded boss*; (vi), femur with a lateral developed lateral bulge (44% of the transverse minimal width of the femoral shaft)*; (vii) the trochanteric shelf is proximolaterally-to-distolaterally oriented*; (viii) presence of a linea intermuscularis cranialis in the femoral anterior face and interrupted at its midlength; (ix) ratio of the mediolateral breadth of the tibial condyle to the breadth of the fibular condyle of the femur is 0.8 or less; (x) medial face of fibular diaphysis is concave transversely along its length, resulting in a D-shaped cross-section with a concave medial border, being bordered by two ridges (shared with some titanosaurs)*; (xi) absence of calcaneum; (xii) the metatarsals II, III, and IV are more gracile and significantly longer than the metatarsals I and V with an abrupt transition between metatarsals I and II and between metatarsals IV and V, which is well visible in dorsal view, resulting in the retraction of the most medial and lateral toes*; (xiii) metatarsal I with a tubercle on medial surface (situated at approximately midlength and equidistant from the dorsal and the ventral margins); (xiv) the proximal tip of the ventrolateral crest of the metatarsal II is laterally deflected and projected*; (xv) pronounced tuberosity near the ventromedial edge of the metatarsal V distal end; (xvi) lateral depression near the dorsolateral ridge of the distal end of pedal phalanx I.1*; (xvii) proximal surface of the phalanx II.1 has an ‘heart’-shaped outline*; (xviii) digit III with a reduced ungual (33% of the metatarsal III length)*; and (xix) loss of the pedal phalanges in the digit V.
Paratype
A partial skeleton comprising dorsal ribs (SAV05-046, SAV05-047), left (SAV05-031a) and right (and SAV05-031b) pubes and two almost complete hindlimbs, which include the left (SAV05-031) and right (SAV05-013) femora, the left (SAV05-036) and right (SAV05-032) tibiae, the left (SAV05-037) and right (SAV05-033) fibulae, a right astragalus (SAV05-034), one almost complete right pes (SAV05-035.a-l and SAV05-038.b), and some elements from the left pes (SAV05-035.m, SAV05-038.a and SAV05-038.c).
Referred material
One left metatarsal I (SAV05-044), metatarsal III (SAV05-056), and metatarsal IV (SAV05-058), and one left pedal phalanx I.1 (SAV05-057.b) and I-V (SAV05-057.c).
Description
Presacral vertebra
An almost complete sequence of possible posterior cervical to anterior dorsal vertebrae from the holotype individual is preserved; however, these vertebrae still need to be prepared. Only a part of the posterior section of a dorsal centrum is available for study (SAV05-40, Fig. 4A–E), which is interpreted as belonging to a middle or posterior dorsal vertebra. This centrum preserves a concave posterior articular surface, which is markedly dorsoventrally compressed (partially owing to taphonomy). The ventral surface is smoothly transversely convex. The ventral margin of the pneumatic fossa is preserved. The centrum is fractured, making it possible to observe internal camellate tissue bone. This type of tissue is also present in the cervical-to-dorsal sequence found in this fossil site. The presence of a camellate tissue bone in the presacral vertebrae was considered as a synapomorphy of Galveosaurus + Titanosauriformes (e.g. Wilson 2002, Upchurch et al. 2004, Carballido et al. 2011b, Mannion et al. 2013).
Dorsal ribs
Our description of the dorsal ribs is based on five partial dorsal ribs (Fig. 4F–Q) from the holotype (three anterior left ribs: SAV08-101, 102, 103) and paratype (one right anterior rib SAV05-047, and one right middle rib SAV05-046). The anterior ribs are the larger elements of the set. The proximal edge is poorly preserved in all elements and only SAV05-046 preserves a complete distal end (Fig. 4N, O). The proximal end has a triradiate cross-section, which corresponds to three well-developed proximodistal crests: (i) the anteromedial to medial crest; (ii) anterior crest; and (iii) posterior crest. The anteromedial to medial crest extends from the capitulum, which is anteromedially located, and deflects to the midpoint of the medial surface in the anterior half of the rib, being the most robust of the forementioned crests. The posterior and anterior crests correspond to the edges of the lateral surface, and the posterior edge is generally thinner and more acute than the anterior one. The anterior ribs are marked by a flat lateral face in proximal-to-middle section of the shaft (probably for the reception of the scapula), which is slightly anteroposteriorly expanded (e.g. Fig. 4H, I). Between the posterior and the anteromedial crests, the rib is markedly concave, corresponding to the posteromedial fossa. This fossa is dorsoventrally elongated and bears a crenulated surface and some foramina (Fig. 4F, G). The posteromedial fossa is dorsally bordered by a thick and robust ridge, which departs from the anteromedial crest to the posterior crest probably connected to the tuberculum. The anteromedial surface of the proximal end, bordered by the anterior and anteromedial crests, is less concave than the posteromedial fossa, and referred to the anteromedial fossa (no crenulated surface and foramina are present). The proximal end of the rib preserves several internal camerae (following: Wedel et al. 2000, Wedel 2003). The presence of a pneumatic foramen and internal cavities in the dorsal ribs was considered as a synapomorphy of Titanosauriformes (Wilson and Sereno 1998, Wilson 2002, Mannion et al. 2013), which is present in the Iberian Early Cretaceous titanosauriforms Tastavinsaurus sanzi (Royo-Torres et al. 2012) and Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017). In the middle section of the rib, the anteromedial crest, located at midpoint of the anteroposterior width of the medial surface of the rib shaft, becomes smoother and disappears. Additionally, the antero- and posteromedial fossae converges into a medial flat surface of the rib shaft, in the middle section of the rib, which quickly becomes transversely concave to the distal end. In the middle section of the more distal anterior dorsal rib of the holotype (SAV08-101) and in the anterior one of the paratype (SAV05-047) specimens, the anterior crest bifurcates, resulting in a second crest, which extends distally to the lateral side of the rib. In the middle and distal sections of the rib, the shaft becomes extremely compressed mediolaterally (the anteroposterior width is more than three times the mediolateral one) with acute anterior and posterior edges, and a flat-to-slightly concave medial and lateral surfaces (Fig. 4J). The presence of anterior dorsal ribs with a plank-like cross-section (i.e. anteroposterior width more than three times the mediolateral one) was considered as characteristic of Titanosauriformes (Wilson and Sereno 1998, Wilson 2002).
In the middle dorsal rib (Fig. 4N, O), the proximal end is also triradiate. The anteromedial crest is anteriorly displaced and expands in the direction of the capitulum, which is anteroposteriorly expanded. The antero- and posteromedial fossae of the proximal end are both concave, and the surface is damaged, which reveals a highly pneumatized internal tissue bone, composed of small camerae and camellae (following: Wedel et al. 2000, Wedel 2003). The anteromedial crest departs from the proximal end and extends distally to the anterior edge of the rib. The posterior pneumatic fossa of the proximal end contains foramina, and extends to the medial surface of the shaft, which is anteroposteriorly concave up to the distal end. Proximally, the anterior crest deflects to the lateral surface of the proximal end, and at the distal end, it also deflects to the lateral surface of the rib, being smoother, resulting in an anteroposterior convex lateral surface. The posterior crest of the proximal end remains in the same position along the proximodistal width of the rib. The distal end is less transversely compressed than the distal end of the anterior dorsal ribs (anteroposterior width is two times the mediolateral one).
Garumbatitan morellensis, gen. et sp. nov., dorsal vertebra and dorsal ribs. Dorsal vertebra SAV05-40 (holotype) in posterior (A), dorsal (B), right lateral (C), left lateral (D), and ventral (E) views. Right middle dorsal rib SAV05-046 (paratype) in posteromedial (F), lateral (N), and medial (O) views; left anterior dorsal rib SAV08-101 (holotype) in lateral (H) and medial (I) views; left anterior dorsal rib SAV08-102 (holotype) in lateral (J) and medial (K) views; left anterior dorsal rib SAV08-103 (holotype) in lateral (L) and medial (M) views; right anterior dorsal rib SAV05-047 (paratype) in medial (P) and lateral (Q) views. Missing edges indicated by dashed lines. Abbreviations: acr, anterior crest; cam, camella; cmr, camera; cr, crest; exp, expansion; gr, groove; mcr, medial crest; pcr, posterior crest; pf, pneumatic fossa; rdg, ridge. Scale bar equals 100 mm.
Anterior caudal vertebra
Five anterior caudal vertebrae and two isolated neural spines are described herein (Figs 5, 6A–F), which were found near the hindlimbs of the holotype specimen. They belong to the distal half of the anterior series, and their positions are estimated based on other titanosauriform specimens with reasonably complete caudal series (e.g. Janensch 1950, Royo-Torres et al. 2012, Mocho et al. 2017a). The four more anterior ones were found in articulation above the femur of the holotype specimen (SAV05-027: ≈eighth caudal vertebra, Fig. 5A–F; SAV05-028: ≈9th caudal vertebra, Fig. 5G–L; SAV05-029: ≈10th caudal vertebra, Fig. 5M–R; SAV05-030: ≈11th caudal vertebra, Fig. 5S–X), plus a more distal anterior caudal vertebra (SAV05-060va ≈ 12th caudal vertebra, Fig. 6A–F). The chevrons described below were found near these vertebrae. The neural arches of the first four preserved caudal vertebrae are heavily damaged. SAV05-027 preserves a centrum with a flat posterior articular surface with a depression in the centre (Fig. 5D). The presence of a flat posterior articular surface is common in non-titanosaurian macronarians such as Camarasaurus supremus (Osborn and Mook 1921), Lourinhasaurus alenquerensis, Brachiosaurus altithorax (D’Emic 2012), Cedarosaurus weiskopfae (Tidwell et al. 1999), and Aragosaurus ischiaticus (Royo-Torres et al. 2014). The anterior caudal vertebrae of Tastavinsaurus sanzi have a flat surface that bears a central concavity (Royo-Torres 2009), as in Lourinhasaurus alenquerensis and SAV05-027. All the caudal vertebral centra of Europatitan eastwoodi are described as amphicoelous (Torcida Fernández-Baldor et al. 2017), differing from the condition of SAV05-027 and, also from the anterior-most caudal vertebrae of Tastavinsaurus sanzi (Royo-Torres 2009). The remaining anterior caudal vertebrae of Garumbatitan morellensis are amphicoelous, i.e. the anterior and posterior articular surfaces are concave, becoming smoothly concave in the last anterior centra (the anterior articular surface becomes less concave than the posterior one). The presence of a less concave anterior articular surface than the posterior one, is a common trend in rebbachisaurids (Carballido et al. 2012, Mannion et al. 2019b), but also present in some somphospondylans such as Huabeisaurus allocotus, Gobititan shenzhouensis, Huanghetitan ruyangensis, Jiangshanosaurus lixianensis, Phuwiangosaurus sirindhornae, Savannasaurus elliottorum, Tangvayosaurus hoffeti, and Wintonotitan wattsi (e.g. D’Emic et al. 2013, Poropat et al. 2016, Mannion et al. 2019a). The average elongation index value [aEI, the anteroposterior length of centrum (excluding articular ball) divided by the mean average value of the mediolateral width and the dorsoventral height of the posterior articular surface of the centrum (following: Upchurch 1995, 1998, Chure et al. 2010)] is around 0.80–1.12 in the preserved anterior caudal vertebrae of Garumbatitan morellensis, which fits with the range shown by the anterior caudal vertebrae of Tastavinsaurus sanzi (aEI: 0.52–1.21, Royo-Torres et al. 2006) but much larger than the range shown by Eeuropatitan eastwoodi (aEI: 0.56–0.68, Torcida Fernández-Baldor et al. 2017). A circular and small tuberosity is present in the centre of the posterior articular surface of SAV05-060va [Fig. 6D; no tuberosities are present in Eeuropatitan eastwoodi (Torcida Fernández-Baldor et al. 2017) and they can be present in some middle caudal vertebrae of Tastavinsaurus sanzi (Royo-Torres 2009)], and Savannasaurus elliottorum (Poropat et al. 2020). The anterior articular surface is mediolaterally wider than dorsoventrally tall, and the maximum mediolateral width is ventrally displaced, which markedly differs from the morphology of the anterior articular surface of the anterior caudal vertebrae of Tastavinsaurus sanzi (Royo-Torres 2009, maximum mediolateral width is dorsally displaced) or Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017, with a subcircular anterior articular surface). The posterior articular surface is also wider than tall.
Garumbatitan morellensis, gen. et sp. nov., anterior caudal vertebrae from the holotype specimen. Anterior caudal vertebra SAV05-027 (A–F), SAV08-028 (G–L), SAV08-029 (M–R), SAV08-030 (S–X) in right lateral (A, G, M, S), anterior (B, H, N, T), left lateral (C, I, O, U), posterior (D, J, P, V), dorsal (E, K, Q, W; anterior towards right side), and ventral (F, L, R, X; anterior towards left side) views. Isolated anterior caudal neural spines SAV05-029 (Y–CC) and SAV05-030 (DD-HH) from the holotype specimen in dorsal (Y, DD), right lateral (Z, EE), anterior (AA, FF), left lateral (BB, GG), and posterior (CC, HH) views. Missing edges indicated by dashed lines. Abbreviations: acdl, anterior centrodiapophyseal laminae; acf, anterior chevron facet; cdr, caudal rib; f, fossa; fr, foramen; lr, lateral ridge; nc, neural canal; pcdl, posterior centrodiapophyseal laminae; posl, postspinal lamina; prsl, prespinal lamina; spol, spinopostzygapophyseal lamina. Scale bar equals 50 mm.
The lateral surface is anteroposteriorly concave and dorsoventrally convex, converging ventrally, resulting in a transversely narrow ventral surface. In SAV05-028 appears a longitudinal ridge in the lateral surface below the caudal rib and the lateral fossa (Fig. 5I). This ridge displaces ventrally in the following centra, occupying a more ventrolateral position in the transition with the middle caudal vertebrae. In SAV05-060va, another lateral ridge appears in lateral surface of the centrum, below the caudal rib. Near the anterior and posterior edges of the centrum, the lateral surface is covered by some rugosities. Bellow the caudal rib there is a shallow fossa, interpreted as pneumatic. Small foramina can be present in some of the fossae. The presence of pneumatic fossa lacking sharply defined margins on the lateral surface of anterior caudal vertebrae was recovered as a synapomorphy of Brachiosauridae (D’Emic 2012, Mannion et al. 2013), which can be present in some vertebrae of Tastavinsaurus sanzi and other somphospondylans such as Savannasaurus elliottorum and Padillasaurus leivaensis, being absent in Europatitan eastwoodi (Carballido et al. 2015, Mannion et al. 2019a, Poropat et al. 2020). The presence of this foramina has been recorded in other non-titanosaurian somphospondylans such as Savannasaurus elliottorum, Chubutisaurus insignis, and Gobititan shenzhouensis (Mannion et al. 2019a, b, Poropat et al. 2023). Also, the presence of small, shallow vascular foramina in lateral and ventral surfaces of anterior-middle caudal centra was recovered as synapomorphy of Titanosauriformes (Mannion et al. 2013); however, their absence seems to be characteristic of some brachiosaurids such as Cedarosaurus weiskopfae (Mannion et al. 2013), Galveosaurus herreroi (Pérez-Pueyo et al. 2019), Soriatitan golmayensis (Royo-Torres et al. 2017a), and the possible brachiosaurid Europasaurus holgeri (Mannion et al. 2013; recovered as a basal macronarian by Carballido et al. 2020), but present in Giraffatitan brancai, Vouivria damparisensis, Lusotitan atalaiensis, and several other somphospondylans (Mannion et al. 2019a, b). These foramina are absent in Tastavinsaurus sanzi (Mannion et al. 2013). The ventral surface is transversely concave between the posterior chevron facets, becoming flat anteriorly (in the ventral surface of SAV05-027 and SAV05-029 there is an anteroposterioly short ridge departing from the posterior chevron facet). The anterior chevron facets are eroded, but they seem to be well-developed. Several foramina are visible on the ventral surface. The neural canal is dorsoventrally taller than it is wide transversely with a quadrangular outline (straight ventral and lateral edges, and concave dorsal one; taphonomic deformation might have played a role in the morphology of the neural canal). Behind the pedicels there are two small depressions in the dorsal surface of the centrum, as well as a tuberosity, which is present up to the middle caudal vertebrae. There is a longitudinal ridge in both sides of the ventral surface of the neural canal near the pedicels. The dorsal surface of the neural canal is excavated in SAV05-030.
The caudal ribs are still dorsoventrally tall in SAV05-027 and SAV05-028, and they extend to the lateral surface of the neural arch (Fig. 5D, J). In the posterior surface of the caudal rib there is a ridge that is interpreted as the postzygodiapophyseal lamina (podl). The caudal ribs are laterally projected in posterior view and posterolaterally projected in dorsal view, deflecting posteriorly in the tip as occur with Tastavinsaurus sanzi (Royo-Torres 2009). In the preserved anterior caudal vertebrae, the caudal ribs do not reach the posterior articulation. The posterior deflection of the caudal ribs is a common feature in titanosauriforms, extending beyond the posterior end of centrum in Abydosaurus mcintoshi, Andesaurus delgadoi, Cedarosaurus weiskopfae, Chubutisaurus insignis, Europatitan eastwoodi, Giraffatitan brancai, Sonorasaurus thompsoni, Soriatitan golmayensis, Tastavinsaurus sanzi, and Tangvayosaurus hoffeti (e.g. Mannion et al. 2013, 2019a, b, D’Emic et al. 2016, Royo-Torres et al. 2017a); this differs from the condition preserved in SAV05-027. The caudal rib is dorsoventrally compressed in the last anterior caudal vertebrae, and they change from a sub-horizontal orientation to an anterodorsal-posteroventral one, in lateral view. The caudal rib is supported anteroventrally by a developed anterior centrodiapophyseal lamina (acdl) and posteroventrally by a rudimentary posterior centrodiapophyseal lamina (pcdl). Both laminae are rudimentary in the remaining anterior caudal vertebrae. The presence of acdl in anterior caudal vertebrae were recovered as a synapomorphy of Flagellicaudata (Tschopp et al. 2015), but it was identified in other sauropods taxa, including the brachiosaurids Giraffatitan brancai, Vouivria dampariensis, and non-titanosaurian somphospondylans Europatitan eastwoodi, Phuwiangosaurus sirindhornae, Tastavinsaurus sanzi, and Jiangshanosaurus lixianensis (Mannion et al. 2017, 2019a, b).
The anterior neural arch is anteriorly displaced but does not reach the anterior edge of the anterior articular surface of the centrum. The lateral surface of the pedicels is covered by rugosities. The pedicels are transversely compressed. The prezygapophyseal processes are anteriorly projected, transversely compressed, and surpass the anterior edge of the anterior articular surface of the centrum. The medial surface is flat whereas the lateral one is transversely convex. The spinoprezygapophyseal lamina (sprl) is single, connected to the prezygapophyseal facets but restricted to the base of the neural spine. There is no lateral tuberosity in prezygapophyseal processes as occur in some titanosaurs (Díez Díaz et al. 2016, González Riga et al. 2016, 2018). The ‘spinoprezygapophyseal lamina-process’ observed in some titanosauriforms (e.g. Giraffatitan brancai and Mendozasaurus neguyelap) and Losillasaurus giganteus (D’Emic 2012, Mannion et al. 2019a) is absent in Garumbatitan morellensis. Between the sprl there is a ventrally deep spinoprezygapophseal fossa (sprf), which is ventrally bordered by the intraprezygapophyseal lamina (tprl). The anterior surface of the neural spines is completely covered by the rugosities of the prespinal lamina (prsl, not medially restricted). The postzygapophyseal processes are posteriorly projected from the neural spine and the pedicels, almost reaching the posterior articular surface of the centrum [more posteriorly projected than in Tastavinsaurus sanzi (Royo-Torres 2009) and Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017)]. A spinopostzygapophyseal lamina (spol) extends from the dorsal edge of the postzygapophysis up to the distal end of the neural spine, and borders the spinopostzygapophyseal fossa (spof), which is covered by the rugosities of the postspinal lamina (posl). The morphology of the neural spine can be described based one complete neural spine from the 12th caudal vertebra (SAV05-060va) and from two broken and isolated neural spines recovered near SAV05-029 and SAV05-030. In anterior view, the neural spines are only slightly expanded and have a round dorsal edge. The posterior surface is wider than the anterior one. No dorsal grooves or lateral depressions are present as in Aragosaurus ischiaticus (Royo-Torres et al. 2014) and Oceanotitan dantasi (Mocho et al. 2019a). The lateral surface is flat and covered by some smooth rugosities near the tip. In the 12th caudal vertebra, the neural spine is subvertical, similar to the holotype of Tastavinsaurus sanzi (Royo-Torres 2009); however, in the more anterior ones of Garumbatitan morellensis, probably from the 8th to 11th position, the neural spines are interpreted as posterodorsally oriented as in Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017) and in the anteriormost caudal vertebra of Soriatitan golmayensis (Royo-Torres et al. 2017a), differing from the anterodorsally oriented neural spines of Tastavinsaurus sanzi (Royo-Torres, 2009), Cedarosaurus weiskopfae (Tidwell et al. 1999, DMNH 39045), and Venenosaurus dicrocei (Tidwell et al. 2001, DMNH 40932). The recovered neural spines are not as anteroposteriorly expanded as the more distal anterior caudal vertebrae of Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017).
Middle caudal vertebra
SAV05-061 corresponds to one of the first middle caudal vertebrae (probably the 16th caudal vertebra, Fig. 5G–L); however, there is a series of nine middle caudal vertebrae that still need to be prepared (SAV08-060-061-063-065-067-066-064-068-069-070-071). This vertebra preserves a peculiar morphology, which is also present in the first middle caudal vertebrae of Tastavinsaurus sanzi. The centrum is amphicoelous, i.e. the anterior and posterior articular surfaces are concave (Fig. 6G–J). No important tuberosities of pits are present in the articular surfaces. The anterior and posterior articular surfaces of the centrum are dorsoventrally compressed [much more than in Tastavinsaurus sanzi (Royo-Torres 2009) and Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017) and differing from the transversely compressed articular surface of Soriatitan golmayensis (Royo-Torres et al. 2017a)], with a straight dorsal and smoothly pointed lateral edges unlike the other Iberian titanosauriform taxa (e.g. Royo-Torres 2009, Mocho et al. 2017a, Royo-Torres et al. 2017a, Torcida Fernández-Baldor et al. 2017). The aEI of SAV05-061 is 1.15. The lateral surface of the centrum is anteroposteriorly concave and marked by two longitudinal ridges (Fig. 6I): (i) a ventrolateral ridge, which originates from the lateral ridge described in the anterior caudal vertebra, which displaced to more ventral position (corresponds here to the ventrolateral edge of the ventral surface and is only connected with the anterior chevron facets); (ii) a lateral longitudinal ridge located at midpoint of the dorsoventral width of the centrum (apparently absent in the first middle caudal vertebrae of Tastavinsaurus sanzi). This later ridge is placed below an anteroposteriorly elongated ridge that corresponds to the position where the caudal rib was in the anterior caudal vertebrae. The region between this rudimentary caudal rib and the lateral ridge, and between the lateral ridge and the ventrolateral one, is dorsoventrally concave. The ventral surface of the centrum is mediolaterally wide and flat (Fig. 6L). The posterior chevron facets are more developed than the anterior ones and preserve a semicircular outline. Several small foramina are visible in the ventral surface as occur in several titanosauriforms (Mannion et al. 2013). The neural canal is dorsoventrally higher than wide with a semi-oval outline in anterior view, and wider than higher with a quadrangular outline in posterior view. The ventral surface of the neural canal is flat. There is a tuberosity in the posterior half of the centrum dorsal surface connected with the longitudinal ridges that are in the ventral surface of the neural canal near the pedicels, which are also observed in anterior caudal vertebrae.
The anterior neural arch is anteriorly displaced (placed in the anterior half of the centrum) but does not reach the anterior edge of the anterior articular surface of the centrum (Fig. 6G, I). This anterior displacement of neural arch in middle caudal vertebrae is characteristic of titanosauriforms (Salgado et al. 1997) and in the non-neosauropods Cetiosaurus oxoniensis (Upchurch and Martin 2003), Moabosaurus utahensis (Britt et al. 2017), and Mierasaurus bobyoungi (Royo-Torres et al. 2017b). The lateral surface of the pedicels is rugose. The pedicels are transversely compressed. No anteroposteriorly oriented ridge and fossa (‘shoulder’) between the prezygapophyses and the postzygapophyses is present in the anterior-middle caudal vertebrae of Garumbatitan morellensis unlike Andesaurus delgadoi (Mannion and Calvo 2011), Lusotitan atalaiensis, and Giraffatitan brancai (Mannion et al. 2013), Huabeisaurus allocotus (D’Emic et al. 2013), and Sonorasaurus thompsoni (D’Emic et al. 2016). The prezygapophyseal processes are transversely compressed and surpass the anterior edge of the anterior articular surface of the centrum. The first middle caudal vertebra seems to be characterized by an anteroventral deflection of the prezygapophyseal processes as occurs in the 16th caudal vertebra of Tastavinsaurus sanzi (Royo-Torres 2009). The medial surface of the prezygapophyseal process is flat whereas the lateral one is transversely convex. The prezygapophyseal facets are rudimentary and poorly defined, facing medially. The sprl is single and fades out before reaching the tip of the prezygapophyseal processes and the dorsal end of neural spine. The sprf is restricted to base of the neural spine, which is ventrally delimited by the tprl (posteriorly located to the anterior articular surface of the centrum) (Fig. 6H, K). The dorsal half of the anterior surface of the neural spine is completely covered by the rugosities of the prsl (not medially restricted). The postzygapophyseal processes are posteriorly projected from the dorsal edge of the neural spine unlike Tastavinsaurus sanzi (Royo-Torres 2009). The spol extends from the dorsal edge of the postzygapophyses up to the distal end of the neural spines, becoming less pronounced in the dorsal half of the spine. The spol borders the spof, which is ventrally deep, and the dorsal two-thirds of the spof are covered by the rugosities of the posl (Fig. 6J). The region between the postyzygapophyses is not depressed as in the anterior and middle caudal vertebrae of Tastavinsaurus sanzi (Royo-Torres 2009) and the last preserved anterior caudal vertebrae of Garumbatitan morellensis (SAV05-060va). No hyposphenic structure seems to be present. The postzygapophyseal facets are reduced and undefined, covered by some rugosities and face laterally. The neural spine is posterodorsally oriented. The anterior surface has a straight profile from the top of the spine up to the tip of the prezygapophyseal processes (this profile is slightly different from the 16th caudal vertebra of Tastavinsaurus sanzi, Royo-Torres 2009). The dorsoposterior edge of the spine does not reach the posterior articular surface as in Tastavinsaurus sanzi (Royo-Torres 2009). The neural spine is only slightly expanded and has a round dorsal edge. The posterior surface is wider than the anterior one. In lateral view the neural spine is subtriangular in shape, and the anteroposterior width decreases dorsally, against the rectangular profiles of the dorsal end of the neural spine in Tastavinsaurus sanzi, and the anteroposteriorly expanded one of Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017). The anteroposterior width of the neural spine of Soriatitan golmayensis also decreases dorsally, but it has a subvertical neural spine (Royo-Torres et al. 2017a).
Garumbatitan morellensis, gen. et sp. nov., anterior and middle caudal vertebrae from the holotype specimen. Anterior caudal vertebra SAV05-060va in right lateral (A), anterior (B), left lateral (C), posterior (D), dorsal (E; anterior towards right side), and ventral (F; anterior towards left side) views. Middle caudal vertebra SAV05-061 in right lateral (G), anterior (H), left lateral (I), posterior (J), dorsal (K; anterior towards right side), and ventral (L; anterior towards left side) views. Missing edges indicated by dashed lines. Abbreviations: acf, anterior chevron facet; cdr, caudal rib; f, fossa; fr, foramen; lr, lateral ridge; nc, neural canal; pcf, posterior chevron facet; posl, postspinal lamina; prsl, prespinal lamina; poz, postzygapophyseal process; prz, prezygapophyseal process; rdg, ridge; spol, spinopostzygapophyseal lamina; sprl, spinoprezygapophyseal lamina; tb, tuberosity; tprl, intraprezygapophyseal lamina; vlr, ventrolateral ridge. Scale bar equals 50 mm.
Middle-posterior caudal vertebrae
Four middle-posterior caudal vertebrae (SAV08-047, SAV08-048, SAV08-049, and SAV08- 050, Fig. 7) were found in partial articulation, 1 m from the distal end of the series of nine middle caudal vertebrae (SAV08-060-061-063-065-067-066-064-068-069-070-071). Considering the general morphology and the size, we believe that all these specimens belong to the same individual (the series of nine caudal vertebrae still need to be prepared). The centra are amphicoelous, i.e. the anterior and posterior articular surfaces are concave (Fig. 7C, D, G), differing from the slightly procoelous caudal vertebrae of Tastavinsaurus sanzi (Royo-Torres 2009). No important tuberosities or pits are present in the articular surfaces. The anterior and posterior articular surfaces of the centrum are slightly dorsoventrally compressed, with a straight-to-concave dorsal edge and convex lateral ones. The aEI spans ranges from 1.39 to 1.46. The lateral surface is anteroposteriorly concave, dorsoventrally convex at midpoint, and marked by the lateral ridge (also present in middle caudal vertebrae), but only developed near the anterior and posterior edges of the centrum (Fig. 7C, D). Near the posterior edge of the centrum, there is a small depression below this lateral ridge in SAV08-49 (Fig. 7D); and near the anterior edge of the centrum and above the lateral ridge there is another depression present in all preserved middle-posterior caudal vertebrae (autapomorphy of Garumbatitan morellensis). The ventrolateral ridges are rudimentary or absent, and associated with the chevron facets. The ventral surface is transversely convex and, generally, slightly concave between the chevron facets (Fig. 7A, F). These facets are semicircular. The neural canal is subcircular.
The anterior neural arch is anteriorly displaced but does not reach the anterior edge of the articular anterior surface of the centrum. The lateral surface of the pedicels is covered by a complex of three anteroposteriorly elongated ridges (Fig. 7C, D, G): (i) one ridge located in the position which corresponds to the position of the caudal rib in the anterior caudal vertebrae; (ii) one at midheight of the neural arch pedicel; and (iii) the dorsalmost one extending from the dorsal margin of prezygapophyseal process. The presence of this complex of three ridges is considered herein as autapomorphic of Garumbatitan morellensis, and is absent in other titanosauriforms. The pedicels of the neural arch are transversely compressed. The prezygapophyseal processes are anteriorly projected and surpass the anterior edge of the anterior articular surface of the centrum. The medial and lateral surfaces of these processes are transversely convex and the prezygapophyseal facet is absent. In dorsal view, the prezygapophyseal processes are medially curved (this curvature is not so pronounced in Tastavinsaurus sanzi,Royo-Torres 2009). The sprl is single and does not reach the tip of the prezygapophyseal processes. They are developed in the ventral half of the neural spine reaching its dorsal end. The distance that prezygapophyses extend beyond the anterior margin of the centrum is less than 20% of centrum length (excluding ball); which differs from the condition shown by most of somphospondylan titanosauriforms, including Tastavinsaurus sanzi (Mannion et al. 2013, 2019a, b). The sprf is present along the total height of the neural spines, but shallower in the dorsal half. This fossa is ventrally delimited by the tprl, which is posteriorly located to the anterior articular surface of the centrum. The sprf is covered by the rugosities of the prespinal lamina, which bears a medially constricted ridge (Fig. 7I), absent in Tastavinsaurus sanzi (Royo-Torres 2009). The spol and spof are absent, and the posterior surface of the neural spine is rough. The neural spine is anteroposteriorly elongated, and the dorsal margin expands anteriorly and posteriorly. The posterior expansion of the neural spine coincides in the posterior projection of the spine and results in a lateral bulge, also considered an autapomorphy of Garumbatitan morellensis. The posterodorsal edge is posteriorly projected, located behind the posterior articular surface of the centrum, condition shared with Venenosaurus dicrocei (Tidwell et al. 2001) and some non-titanosauriform sauropods such as Camarasaurus (Osborn and Mook 1921, Gilmore 1925, Ostrom and McIntosh 1966, McIntosh et al. 1996a, 1996b). The dorsal edge is straight-to-smoothly convex in lateral view, sloping posteriorly, unlike the markedly concave dorsal edge of the middle-posterior caudal vertebrae of Tastavinsaurus sanzi (Royo-Torres et al. 2006), Astrophocaudia slaughteri (D’Emic 2013), Aragosaurus ischiaticus (personal observation, P.M. 2014), Cedarosaurus weiskopfae (personal observation, P.M. 2018), and slightly developed in some caudal vertebrae of Giraffatitan brancai (Janensch 1950).
Garumbatitan morellensis, gen. et sp. nov., middle-posterior caudal vertebrae from the holotype specimen. Middle-posterior caudal vertebrae SAV08-047, SAV08-048 and SAV08-049 in ventral (A; anterior towards right side), dorsal (B; anterior towards right side) and right lateral (C, D), anterior (I), and left lateral (J) views. Middle-posterior caudal vertebra SAV08-050 in dorsal (E; anterior towards left side), ventral (F; anterior towards right side), left lateral (G), and posterior (H) views. A, B, D, E, F, and J correspond to a 3D digital model. Missing edges indicated by dashed lines. Abbreviations: acf, anterior chevron facet; dp, depression; lr, lateral ridge; pbu, posterior bulge; pcf, posterior chevron facet; prsl, prespinal lamina; prz, prezygapophyseal process; rdg, ridge; sprf, spinoprezygapophseal fossa; sprl, spinoprezygapophyseal lamina; tprl, intraprezygapophyseal lamina. Scale bar equals 50 mm.
Chevrons
Four anterior chevrons are available for study (from the anteriormost preserved chevron to the posteriormost one: SAV05-060chb, SAV05-063; SAV05-060cha, and SAV05-060chc; Fig. 8), which probably belong to the first half of the tail. In addition, two other chevrons were identified within the jacket SAV08-060-061-063-065-067-066-064-068-069-070-071. Based on a relatively complete series of chevrons (e.g. Janensch 1950, Royo-Torres 2009, D’Emic et al. 2013), the preserved chevrons are interpreted as the third to the sixth. The proximal end of the dorsal rami is badly preserved in the some of the chevrons (e.g. right ramus of SAV05-063 and SAV05-060cha is absent; and left and right ones are broken in SAV05-060chc). The haemal canal is more than 40% of the total height of the chevron (around 40–44%) (Fig. 8B, G, L, Q), which is common in titanosauriforms such as Lusotitan atalaiensis, Europatitan eastwoodi, Soriatitan golmayensis, and Tastavinsaurus sanzi (Royo-Torres 2009, Mocho et al. 2017a, Royo-Torres et al. 2017a, Torcida Fernández-Baldor et al. 2017). The dorsal rami are transversely compressed with an anteroposteriorly convex lateral surface, and an anteroposteriorly flat medial one. The anteromedial edge of the dorsal rami is acute, resulting in a proximodistal crest, which converge, ventrally to the anterior crest of the distal end of the chevron. The posterior edge of the dorsal rami is rounded. The articular facets for the vertebrae are badly preserved, but some information can be obtained. These facets are more transversely expanded than anteroposteriorly, and the lateral edge is laterally projected (some rugosities are present laterally, below the articular facets). A posterior rugosity is present right below these facets. The proximal end is interpreted as open, i.e. the chevrons are not dorsally bridged.
The distal end is transversely compressed being anteroposteriorly longer than mediolaterally (the transverse compression becomes more significant towards the posteriormost preserved element, SAV05-060chc). The distal end is posteriorly deflected (Fig. 8C, H, M, R). The anterior and posterior edges are acute, resulting in anterior and posterior crests. The anterior crest preserves a step (Fig. 8O, H) that is absent in the more anterior chevrons of Tastavinsaurus sanzi [a step appears from the seventh chevron, at a more distal position (Royo-Torres 2009) than the anterior step observed in Garumbatitan morellensis)] and of Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017). The posterior edge is convex to straight in lateral view (no step is present as in Europatitan eastwoodi,Torcida Fernández-Baldor et al. 2017). Above this anterior crest and below the end of the haemal canal there is a triangular depression that persists in the preserved series of chevrons. Similarly, in the posterior surface of the distal end, there is a subtriangular depression below the end of the haemal canal, but only in the first preserved chevron. The lateral surface of the distal end is striated, and no bulge is observed.
Garumbatitan morellensis, gen. et sp. nov., anterior chevrons from the holotype specimen. Anterior chevron SAV05-060chb (A–E), SAV05-063 (F–J), SAV05-060cha (K–O), and SAV05-060chc (P–T) in dorsal (A, F, K, P; anterior towards bottom side), anterior (B, G, L, Q), left lateral (C, H, M, R), posterior (D, I, N, S), and right lateral (E, J, O, T) views. Missing edges indicated by dashed lines. Abbreviations: acr, anterior crest; gr, groove; pcr, posterior crest; st, step. Scale bar equals 50 mm.
Interclavicle
SAV05-055 shares a similar morphology to elements that have been interpreted as interclavicles such as NMB-1698-R from Spinophorosaurus nigerensis [see also: (Tschopp and Mateus 2013)]. The proximal end of an interclavicle was found near the holotype specimen (SAV05-055, Fig. 9). The shaft of this interclavicle is anteroposteriorly compressed, with a transversely convex anterior surface and flat posterior one (Fig. 9A). The proximal end becomes mediolaterally compressed, and the surface is striated. It is possible to identify a facet in the left side of the proximal end (the right face is not well individualized) for the contact with coracoids (following: Tschopp and Mateus 2013) (Fig. 9D). The proximal tip of the interclavicle is pointed and not bifurcated.
Garumbatitan morellensis, gen. et sp. nov., interclavicle SAV05-060chb from the holotype specimen in proximal (A), anterior (B), left (C), posterior (D), and right (E) views; and cross-section (F). Missing edges indicated by dashed lines. Abbreviations: fct, facet; gr, groove; rdg, ridge; str, striation. Scale bar equals 50 mm.
Pubes
The distal end of two pubes are described (right pubis, SAV05-031a; left pubis, SAV05-31b; Fig. 10). Besides the difference in robustness between these two pubes (SAV05-031a and SAV05-31b), they are referred as belonging the same individual, in this case, to the paratype specimen, being recovered below its left femur. The pubic shaft is transversely compressed and slightly lateromedially and anteroposteriorly expanded in its distal end [differing from a planar distal blade common in several titanosaurs, Poropat et al. (2016)]. The anterodistal edge is not projected, as in Europatitan eastwoodi (Torcida Fernández-Baldor et al. 2017); in this regard, it differs from the hook-shaped profile present in the camarasaurid Lourinhasaurus alenquerensis (Mocho et al. 2014), and the titanosaurforms Giraffatitan brancai (Janensch 1961) and Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009, Royo-Torres et al. 2012). A medial triangular facet is present near the distal end, which is considered to be part of the symphysis. The distal surface is rough and concave transversely. No crests and tuberosities are observed on the lateral and medial surface of the preserved portion.
Garumbatitan morellensis, gen. et sp. nov., pubes from the paratype specimen. Left pubis SAV05-31b in posterior (A), lateral (B), anterior (C), and distal (E) views. Right pubis SAV05-31a in distal (D), posterior (F), and anterior (G) views. Missing edges indicated by dashed lines. Scale bar equals 50 mm.
Femora
Four femora have been identified, two from the holotype specimen (left, SAV05-023; right, SAV05-024; Fig. 11) and two from the paratype specimen (left, SAV05-031; and right, SAV05-013; Fig. 12). The most complete femur of the holotype specimen is the right one, which lacks a sector of the proximal end. Considering the two other remains from the proximal half, which are still unprepared (including the femora head), we tentatively estimate a proximodistal length of around 1.9–2.0 m. Only the distal end of the left femur was recovered. In the paratype, the left femur is complete but presents some deformation and the right femur lacks the proximal half. The femur is straight in lateral and anterior views and has an important medial deflection of the proximal one-third (Fig. 12B, D). The femoral proximolateral margin is medial to the lateral margin of the distal half of the shaft as in other macronarians (Royo-Torres et al. 2012, Mannion et al. 2013) but differing from Tastavinsaurus sanzi, which lacks a strong medial deflection. This results in a well-developed lateral bulge, representing 44% of the narrowest transverse width of the femoral shaft (following Salgado et al. 1997), which might have been affected by some deformation. Considering the whole femoral morphology, particularly the distolateral projection of the distal end of the trochanteric shelf, which is not expected to be affected by the anteroposterior deformation of the femur, we consider plausible the presence of one of the most developed lateral bulges in sauropods, which would be characteristic of Garumbatitan morellensis. The diaphyses of these femora are markedly anteroposteriorly compressed. The mediolateral width of the diaphysis is around 3.23 times the anteroposterior width in the holotype, and 2.67–2.99 times in the paratype specimen, which is markedly higher than in other titanosauriforms from the Early Cretaceous of Iberia [Tastavinsaurus sanzi is 1.67–1.5 times, Royo-Torres (2009); Soriatitan golmayensis is 2.18, Royo-Torres et al. (2017a)]. These values are possibly slightly affected by some crushing, but we believe that it is not to significant. The cross-section of the diaphysis is elliptical, slightly anteroposteriorly wide in the lateral half. In the posterior surface, a broad trochanteric shelf emerges from the reduced greater trochanter that fades away at level of bulge apex (Fig. 12D). The trochanteric shelf has a prominent proximomedial-to-distolateral orientation, which is considered as autapomorphic. The fourth trochanter is located at the posteromedial border of the shaft, is dorsoventrally elongated ridge, and its distal tip is placed above femur midheight. It is associated with a medial and proximodistally elongated depression (Fig. 12E). In the paratype, the fourth trochanter is medially projected, being observed in anterior view (Fig. 12B). This feature was recovered as synapomorphy of Brachiosauridae (Mannion et al. 2013) and is absent in Tastavinsaurus sanzi (Royo-Torres 2009). However, this medial projection of the fourth trochanter can be a consequence of deformation, and the full preparation of the femur of the holotype might provide some insights about the orientation of the fourth trochanter. The femoral anterior surface bears a longitudinal crest, distally and proximally marked (Figs 11A, 12B, H). This crest is interpreted as the linea intermuscularis cranialis present in some lithostrotians (Otero 2010, D’Emic 2012) and in the non-lithostrotian titanosaur Diamantinasaurus matildae (Poropat et al. 2015, 2023). The femoral head is elliptical in proximal view, anteroposteriorly compressed, probably accentuated due the deformation. The femoral head is dorsomedially projected, and the proximal surface is rough and convex. There is a step below the femoral head, absent in Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009). The region between the lateral bulge and greater trochanter is anteriorposterioly compressed, and the bulge is anterolaterally deflected. The anterior and posterior surfaces of the lateral bulge are striated.
The distal end of the femur is smoothly concave anteriorly and posteriorly. The tibial condyle is longer anteroposterioly than the fibular one, and both bear rough and convex surfaces (Figs 11H, 12F, G). The medial surface of the tibial condyle is flat as in other sauropods such as Lourinhasaurus alenquerensis (Mocho et al. 2014) and Giraffatitan brancai (Janensch 1961). In distal view, the condyles are anteroposteriorly oriented. The distal end is perpendicular and slightly laterally bevelled (less than 10° from the horizontal), differing from the laterally bevelled distal end of Cedarosaurus weiskopfae, Tastavinsaurus sanzi, Soriatitan golmayensis, Phuwiangosaurus sirindhornae, ‘Paluxysaurus jonesi’ [considered as Sauroposeidon proteles by D’Emic and Foreman (2012)], and Vouivria damparensis among others (e.g. Mannion et al. 2013, 2017, 2019a, b, Royo-Torres et al. 2017a). The anterior expansion of the distal condyles is not significant unlike some lithostrotians (Wilson 2002, Mannion et al. 2013, 2019a, b). The tibial to fibular condylar anteroposterior length ratio is less than 1.2 as in Soriatitan golmayensis (Royo-Torres et al. 2017a; Mannion et al. 2019b), differing from the condition, which seems to characterize the Somphospondyli clade [1.2 or greater (Mannion et al. 2013) based on the LSDM matrix), including Tastavinsaurus sanzi. The ratio of mediolateral breadth of tibial condyle to breadth of fibular condyle is less than 0.8 (0.69–0.79), as occur in the brachiosaurids Brachiosaurus altithorax and Giraffatitan brancai and Vouivria damparensis, and in many somphospondylans (Wilson 2002, Poropat et al. 2016). This ratio is greater than 0.8 in Tastavinsaurus sanzi and Soriatitan golmayensis (Royo-Torres 2009, Poropat et al. 2016, Mannion et al. 2017, 2019a, b, Royo-Torres et al. 2017a). The epicondyle is laterally projected, well developed, and separated from fibular condyle by a longitudinal groove (more developed in the largest individual and referred to here as the posterolateral fossa; Figs 11D, G, 12D, F–H). This epicondyle is developed up to the distal margin of the femur, being visible and individualized from the fibular condyle in distal view. The intercondylar region is anteroposteriorly compressed in both femora and, in anterior view, it is possible to observe a marked local concavity between the condyles in smallest individual, but unpronounced in the largest one. Small intercondylar ridges are present in the holotype specimen (Fig. 11F), as occur in Vouivria dampariensis (Mannion et al. 2017).
Garumbatitan morellensis, gen. et sp. nov., right femur SAV05-024 from the holotype specimen in anterior (A, B*), medial (C*), posterior (D, F*) lateral (G*), and distal (E, I*) views; and cross-section of diaphysis at minimal diameter (E; anterior towards top side). *3D digital model. Missing edges indicated by dashed lines. Abbreviations: ep, epicondyle; fic, fibular condyle; lg, longitudinal groove; lic, linea intermuscularis cranialis; icg, intercondylar groove; icr, intercondylar ridge; plf, posterolateral fossa; tic, tibial condyle. Scale bar equals 100 mm.
Garumbatitan morellensis, gen. et sp. nov., femora from the paratype specimen. Left femur SAV05-031 in proximal (A), anterior (B), lateral (C), posterior (D), lateral (E), and distal (G) views. Right femur SAV05-013 in distal (F), anterior (H), lateral (I), posterior (J) and medial (K) views. Missing edges indicated by dashed lines. Abbreviations: ep, epicondyle; dp, depression; fh, femoral head; fic, fibular condyle; ft, fourth trochanter; lg, longitudinal groove; lic, linea intermuscularis cranialis; icg, intercondylar groove; mco, medial concavity to the fourth trochanter; plf, posterolateral fossa; st, step; tic, tibial condyle; ts, trochanteric shelf. Scale bar equals 100 mm.
Tibiae
In the Sant Antoni de la Vespa fossil site, four tibiae were found: two from the holotype specimen (right, SAV05-065, Fig. 13A; left, SAV05-025, not prepared), and two from the paratype one (left, SAV05-036, Fig. 14G–L; and right, SAV05-032, Fig. 14A–F, the most proximal section of the cnemial crest is fractured). The left tibia has a pronounced mediolateral compression and torsion between proximal and distal ends, and, in consequence, the morphological description will be mainly focused on the elements from the right side. The tibia is straight in anterior and lateral views (not arched as in Lusotitan atalaiensis,Mocho et al. 2017a). The proximal section is D-shaped with a straight lateral edge (corresponding to the fibular articular facet) and a concave to straight posterior edge (Figs 13A, 14C). In the proximal surface of the right tibia of the paratype (SAV05-032), the perpendicular width to articular fibular facet (approximately the mediolateral width of the proximal end) has a similar width to its perpendicular (approximately the anteroposterior width of the proximal end). However, this mediolateral axis is slightly longer in the right tibia of the holotype (SAV05-065), i.e. in the less deformed elements, the proximal end is slightly anteroposteriorly compressed. In the case of the left tibia of the paratype, the proximal section has a mediolaterally compressed outline owing to deformation (Fig. 14I), but the axis perpendicular to the fibular articular facet is still the longest one, which possibly acquired a different orientation because of the torsion of the proximal end. The proximal surface is rough and flat with a circular depression in the medial two-thirds of the surface. The region near the articular facet of the fibula extends slightly distally. The lateral surface of the proximal section is smoothly concave bearing a subtriangular fibular articular facet. This surface is anteriorly bordered by the cnemial crest and posteriorly by a lateral crest rising from the proximal surface (Fig. 14D, E). The cnemial crest is round [Fig. 13B; unlike the triangular cnemial crest of Tastavinsaurus sanzi (Royo-Torres 2009), Europasaurus holgeri (Carballido et al. 2020) and Lusotitan atalaiensis (Mocho et al. 2017a)], asymmetrical (the ventral edge is longer) and laterally directed in the non-deformed elements. A laterally projected cnemial crest (Figs 13A, 14C) is common in sauropods (Wilson and Sereno 1998, Wilson 2002, Mannion et al. 2013), but it is anterolaterally projected in several titanosauriforms such as Tastavinsaurus sanzi (Royo-Torres 2009). In the holotype specimen of Garumbatitan morellensis, the cnemial crest expands becoming thicker proximally. The posterior surface of the cnemial crest preserves some rugosities, but the ‘tuberculum fibularis’ present in Giraffatitan brancai and Vouivria dampariensis (Mannion et al. 2017), in many diplodocids (Harris 2007, Tschopp et al. 2015), and in Janenschia robusta (Mannion et al. 2019b), is absent. From the proximal surface extends a proximodistal ridge, posterior to the cnemial crest (Fig. 14C) interpreted as a ‘second cnemial crest’ (sensuBonaparte et al. 2000; ; Mannion et al. 2013). This crest is present in many eusauropods but absent in several somphospondylans and diplodocoids (Mannion et al. 2013, 2017, 2019a, b). The anteromedial surface of the proximal end is smoothly concave in the holotype specimen.
Garumbatitan morellensis, gen. et sp. nov., right tibia (SAV05-065), fibula (SAV05-064) and astragalus (SAV05-066) from the holotype specimen in proximal (A*), anterior (B*), medial (C*), posterior (D*), lateral (E*), and distal (F*) views. *3D digital model. Abbreviations: acr, anterior crest; aspa, articular surface for the ascending process; cc, cnemial crest; lt, lateral trochanter; pvp, posteroventral process; tb, tuberosity; tia, tibial articular surface. Scale bar equals 100 mm.
Garumbatitan morellensis, gen. et sp. nov., tibiae from the paratype specimen. Right tibia (SAV05-032) in proximal (A), anterior (B), medial (C), posterior (D), lateral (E), and distal (F) views. Left tibia (SAV05-036) in proximal (I), anterior (G), medial (H), posterior (J), lateral (K), and distal (L) views. Missing edges indicated by dashed lines. Abbreviations: 2cc, second cnemial crest; aspa, articular surface for the ascending process; cc, cnemial crest; cr, crest; fia, fibular articular surface; pvp, posteroventral process. Scale bar equals 100 mm.
The tibial shaft bears a sub-elliptical to D-shaped cross-section, with a flat posterolateral surface. The lateral edge of the shaft is acute, resulting in a crest structure. The distal section bears a transverse expansion (not so pronounced in the left tibia of the paratype specimen owing to deformation; Figs 13B, 14A, F) and the distal end mediolateral width to the long axis of the cross-section horizontally through the midshaft ratio ranges from 1.56 to 2.0 (excluding the most deformed tibia, SAV05-036). The mediolateral width : the anteroposterior width ratio of the distal end ranges from 1.75 to 1.62. The anterior surface of the distal end is flat-to-slightly concave, and the anteromedial edge bears a rough proximal ridge in the paratype specimen, particularly thicker in the holotype one. The articular surface for the ascending process is transversely elongated and laterally projected, with a flat surface, occupying a more dorsal position than posteroventral process, but relatively lower when compared with other macronarians such as Lusotitan atalaiensis (Mannion et al. 2013, Mocho et al. 2017a), Lourinhasaurus alenquerensis (Mocho et al. 2014), or Giraffatitan brancai (Janensch 1961). The posteroventral process is oval and smaller than the articular surface for the ascending process and bears a convex and rough surface. The articular surface for the ascending process and the posteroventral process are separated posteriorly by a not well-marked longitudinal concavity. The tibia is 64% of femur length based on the paratype specimen (the femur is not complete in the holotype one), against 55% in Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres et al. 2012). Similar values are observed in Cedarosaurus weiskopfae (63%), Euhelopus zdanskyi (63%), and Phuwiangosaurus sirindhornae (62%) (Poropat et al. 2016).
Fibulae
Four fibulae were found and referred to Garumbatitan morellensis, two from the holotype (right, SAV05-064, Fig. 13; left, SAV05-026, not prepared) and two from the paratype (right, SAV05-033, Fig. 15A–F; left, SAV05-037, Fig. 15G–L). The description is mainly based on the right elements of the paratype and holotype specimens (the left one from the holotype still needs to be prepared, and the left one from the paratype has an important mediolateral deformation, as occurs with the left tibia). The fibula is straight in anterior and lateral view (Figs 13E, 15A, G) lacking a pronounced sigmoid profile as many somphospondylans such as Oceanotitan dantasi (Mocho et al. 2019a), ‘Paluxysaurus jonesi’ (Rose 2007), Huabeisaurus allocotus (D’Emic et al. 2013), Phuwiangosaurus sirindhornae (Martin et al. 1999), and Tastavinsaurus sanzi (Royo-Torres 2009). The presence of a sigmoidal fibula was recovered as synapomorphy of Somphospondyli/Titanosauria (Mannion et al. 2013; based on the LCDM matrix). The proximal one-third of the fibula is anteroposteriorly expanded, more posteriorly than anteriorly. In medial view, there is a short triangular tibial articulation surface (occupying one-fifth of the fibular total length), differing from the long tibial scars present in basally branching macronarians (e.g. Ostrom and McIntosh 1966, Mocho et al. 2014) and Oceanotitan dantasi (Mocho et al. 2019a). This surface faces medially-to-slightly proximomedially. The ventral apex of the tibia articular surface is located in the anteromedial edge of the shaft, and coincides with the presence of a boss structure, more pronounced in the left fibula, interpreted as the anterior trochanter of Wilson and Sereno (1998), which is absent to rudimentary in other titanosauriforms, such as Lusotitan atalaiensis (Mocho et al. 2017a), Tastavinsaurus sanzi (Royo-Torres 2009, Royo-Torres et al. 2012), and Giraffatitan brancai (Janensch 1961). The anteromedial crest departs from this anterior trochanter to the proximal surface and occupies one-fifth of the fibular total length (as the tibial articular surface; Fig. 14C, D, I, J). The anteromedial crest is visible in proximal view being embraced by the cnemial crest of the tibia. The presence of an anteromedially directed crest extending into a notch behind the cnemial crest of the tibia was recovered as a synapomorphy of the somphospondylan clade Sauroposeidon + (Tastavinsaurus + (Euhelopodidae + (Chubutisaurus + Titanosauria))) by D’Emic (2012) and Somphospondyli/Titanosauria by Mannion et al. (2013; based on the LCDM matrix). This crest is transversely constricted and bears a longitudinal sulcus on its lateral face. The medial face of the fibular diaphysis is concave transversely along its length, resulting in a D-shaped cross-section with a concave medial border, being bordered by two ridges, which corresponds to the anteromedial and posteromedial edges of the diaphysis, which is here considered as autapomorphic of Garumbatitan morellensis [in Tastavinsaurus sanzi, only the proximal half is transversely concave medially, Royo-Torres et al. (2006)]. Fibular diaphysis with a transversely concave medial face can be observed in some titanosaurs, such as Lohuecotitan pandafilandi (Díez Díaz et al. 2016). The lateral trochanter is a complex structure comprising an oval and shallow tuberosity inserted in a flat area bordered by two proximodistal ridges (the posterior one is more pronounced and located near the posterior face of the diaphysis; Figs 13E, 14A, G), as occurs in several somphospondylans (Upchurch 1998, Mannion et al. 2013, 2017), but also present in the brachiosaurid Giraffatitan brancai (Mannion et al. 2013). The lateral trochanter is only slightly laterally pronounced, differing from the lateral projected lateral trochanters common in some somphospondylans (e.g. Ksepka and Norell 2006, Salgado and Carvalho 2008, Otero 2010, Díez Díaz et al. 2013b, Lacovara et al. 2014). The tip of the lateral muscle scar is located approximately at midshaft.
Garumbatitan morellensis, gen. et sp. nov., fibulae from the paratype specimen. Right fibula (SAV05-033) in proximal (A), lateral (B), anterior (C), medial (D), posterior (E), and distal (F) views. Left fibula (SAV05-037) in proximal (I), lateral (G), anterior (H), medial (J), posterior (K), and distal (L) views. Abbreviations: acr, anterior crest; cr, crest; lt, lateral trochanter; mli, medial lip; tia, tibial articular surface. Scale bar equals 100 mm.
The proximal edge of the fibula is straight and horizontal in lateral view. The proximal surface is rough and flat with a subrectangular outline (Figs 13A, 14C, I), different from the autapomorphic crescentic morphology of Tastavinsaurus sanzi (Royo-Torres et al. 2012). Distally, the anteromedial ridge of the diaphysis has a round projection. The lateral face of the distal end is transversely convex, partially related with the lateral projection of the distal end. The medial edge of the distal section is projected, forming a medial lip which articulates with the astragalus. The distal surface is rough and flat-to-concave and has a semicircular-to-oval outline as in several other sauropods (Royo-Torres 2009), with a flat medial edge and round lateral one (the lateral margin of the distal surface extends to the diaphysis of the fibula). This morphology is markedly distinct from the autapomorphic quadrangular morphology of the fibular distal end shown by Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009, Royo-Torres et al. 2012). Excluding the left fibula of the paratype, which is deformed: (i) the mediolateral width to the anteroposterior width ratio of the distal end is greater than 0.8; and (ii) the mediolateral width of distal end to the mediolateral width at the midshaft ratio is greater than 2.0, differing from Tastavinsaurus sanzi, which is characterized by a ratio lower than 2.0 (Royo-Torres et al. 2012).
Tarsus
Two complete astragali have been described: one right astragalus from the holotype (SAV05-066, articulated with the fibula and tibia, Fig. 13) and other right one from the paratype (SAV05-034, Fig. 16). The astragalus is wedge-shaped, and in proximal view it becomes anteroposteriorly narrow in its medial half (Fig. 16B), characteristic of neosauropods (Wilson 2002). Also in proximal view, the anterior edge is straight and transversely oriented. The posterior edge of the astragalus is straight and transversely oriented behind the ascending process, but the medial sector of the posterior edge is mainly straight and posterolaterally-anteromedially oriented, culminating in the round and blunt medial apex of the astragalus (Fig. 16B). In anterior view, the apex of the astragalus is proximodistally constricted (Fig. 16E) as occurs in derived eusauropods (Upchurch 1995, 1998, Mannion et al. 2017). The ascending process almost reaches the posterior margin of astragalus (when the dorsal surface of the ascending process is in horizontal; Fig. 16C, F), a condition commonly shared by the members of Neosauropoda (Wilson and Sereno 1998, Wilson 2002). The proximal surface of this process is rough and flat. The posterior surface of the ascending process is smoothly concave in the holotype (this surface seems to be eroded) and flat in the paratype (with a small foramen). From the proximomedial corner of the dorsal surface of the ascending process, a rudimentary proximodistal ridge is present that does not reach the posterior edge of the astragalus (Fig. 16D). In this sector a well-developed crest can be observed in several non-somphospondylan titansoauriforms, such as Lusotitan atalaiensis (Mannion et al. 2013, Mocho et al. 2017a) and Giraffatitan brancai (Janensch 1961), being considered as absent in Garumbatitan morellensis. The posterior margin of the astragalus lacks a tongue-like projection posteromedial to the ascending process (Fig. 16D), as occurs in titanosauriforms (Mannion et al. 2013). Medial to this rudimentary crest there is a foramen. The proximal surface of the tibial articular surface is broadly concave and smooth, slopping posteriorly. The mediolateral width to maximum proximodistal height ratio is greater than 1.8 [1.84–1.87; it is less than 1.8 in Tastavinsaurus sanzi (Royo-Torres et al. 2012)]; and the mediolateral width to the maximum anteroposterior length ratio is 1.87–1.94. The rough ventral surface of the astragalus is transversely convex and transits continuously to the also rough anterior surface. The articular surface for the fibula (lateral surface of the astragalus) faces laterally, is well-limited, and occupies the entire lateral surface of the astragalus. This articular surface contains two foramina separated by a ridge (this part of the astragalus is covered by sediment in the holotype specimen). The astragalus caps most of the distal end of the tibia, as in the brachiosaurid Vouivria dampariensis (Mannion et al. 2017), but is distinct from the reduced astragalus that characterizes most titanosauriforms (Ksepka and Norell 2006, Wilson and Upchurch 2009), including Tastavinsaurus sanzi (Royo-Torres et al. 2012). No calcaneum was identified, the absence of this element being considered as a reliable feature of this taxon, which is present in most non-titanosaurian sauropopods (e.g. Poropat et al. 2023), as in Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009, Royo-Torres et al. 2012) and Gobititan shenzhouensis (You et al. 2003). The preservation of the calcaneum is unusual; however, we believe that the absence of this element in Garumbatitan morellensis is plausible. This is supported by the fact that two nearly complete hindlimbs, including their pedes, were found in close association, and preserve almost all their distal elements (i.e. phalanges).
Garumbatitan morellensis, gen. et sp. nov., right astragalus SAV05-034 from the paratype specimen in distal (A), proximal (B), lateral (C), posterior (D), anterior (E), and medial (F) views. Abbreviations: asp, ascending process; fia, fibular articular surface; fr, foramen; tia, tibial articular surface. Scale bar equals 50 mm.
Pedes
A partially articulated to associated right pes (SAV05-068) was found for the holotype specimen (Figs 17A, 18, 20, 22). Some elements of the left pes (SAV05-021 and SAV05-024) of the holotype were also recovered. An almost complete right pes (SAV05-35) and a left metatarsal III (SAV05-038.a) was found associated and partially articulated to the smallest individual, which corresponds to the paratype specimen (Figs 17B, 19, 21, 23). Another three left metatarsals, here considered as referred materials, were found associated with these two legs (SAV05-044, SAV05-056, and SAV05-058). The presence of two left metatarsals III, with similar morphology and size, indicates the presence of at least two individuals with similar size (including the paratype specimen) in addition to the holotype specimen.
All five metatarsals are preserved in the holotype and paratype specimens, and when articulated, they bear a slightly arched profile with convexity facing dorsally. The shortness and robustness of metatarsals I and V relative to metatarsals II, III, and IV is considered here as autapomorphic of Garumbatitan morellensis. Three metatarsals I are preserved [the right of the holotype specimen (SAV05-068.a, Fig. 18A–F), which is mediolaterally deformed; the right one of the paratype (SAV05-35.a, Fig. 19A–F); and a referred left one (SAV05-044)]. Metatarsal I (Figs 17, 18) is three-quarters the length of metatarsal II, expands proximally and distally, and is the most robust metatarsal. The proximal surface is subtrapezoidal to subtriangular, rough, flat, and sloping medially. In proximal view, the medial edge is straight-to-slightly convex and the lateral one is concave for the reception of metatarsal II (Figs 18E, 19E). The ventrolateral edge of the proximal surface deflects distally extending to the ventrolateral edge of metatarsal I (with a ridge like-morphology, i.e. the ventrolateral crest; Fig. 19C). The dorsolateral edge of the proximal face is proximally projected (Fig. 18B). The proximal surface is angled ventromedially approximately 15° relative to the axis of shaft (Figs 18A, 19A). The diaphysis of metatarsal I is elliptical in cross-section in the holotype (probably a consequence of deformation) and triradiate in the paratype (with three main surfaces, the continuous dorsomedial, lateral, and ventral surface). In the proximal half, the dorsomedial surface is continuous and, medially, it curves posteriorly where it meets with the ventral one, resulting in the ventromedial crest/ridge (which is present along its total proximodistal width of the metatarsal). In the proximal half of the ventromedial crest, there is a pronounced elliptical tuberosity (Figs 18A, 19B), not described in Tastavinsaurus sanzi (Royo-Torres 2009). The presence of this tuberosity in the medial face of the metatarsal I, seems to be unique for the brachiosaurids Giraffatitan brancai, Sonorasaurus thompsoni, and Venenosaurus dicrocei (D’Emic et al. 2016, Mannion et al. 2019b). In the distal half of the metatarsal, the dorsomedial surface is subdivided in dorsal and medial surfaces, separated by a short proximodistal crest coming from the medial distal condyle. The distal half of the dorsal surface is flat (holotype) or slightly concave (paratype) between the distal condyles. The lateral surface is bordered by a proximodistally elongated ventrolateral and dorsolateral crests. The proximal half of the lateral surface is deeply concave and striated becoming flat distally, with a proximodistally elongated tuberosity (ridge-like in the holotype specimen). No ventrolateral projection in the distal end is present [this projection in common in diplodocids (Tschopp et al. 2015) and in other taxa such as Ligabuesaurus leanzai (Bonaparte et al. 2006) and Gobititan shenzhouensis (You et al. 2003)]. The distal condyles are well-marked, they are convex transversely and dorsoventrally, and are separated by a ventral intercondylar concavity (excavated in the paratype specimen). The lateral and medial condyles are located at a perpendicular plan relative to the axis of the diaphysis. In distal view, the condyles diverge ventrally, and the medial condyle is dorsoventrally elongated. The metatarsal I to metatarsal V proximodistal length ratio is 1.05, differing from some titanosauriforms, which are a characterized by a ratio less than 1.0 as Tastavinsaurus sanzi, Cedarosaurus weiskopfae, Sonorasaurus thompsoni, Gobititan shenzhouensis, and Ligabuesaurus leanzai (Mannion et al. 2013).
Garumbatitan morellensis, gen. et sp. nov., right pes of the holotype (A, B) and paratype (C, D) specimens in proximal (A, C) and dorsal (B, D) views. Missing edges indicated by dashed lines. Scale bar equals 50 mm.
Three metatarsals II were recovered [the right (SAV05-068.b, Fig. 18G–L) and left (SAV05-021) from the holotype specimen; and the right one (SAV05-035.b, Fig. 19G–L) from the paratype specimen]. They have a subrectangular proximal surface (Fig. 18K). In proximal view, the dorsal and ventral edges are straight (the dorsal and ventral edges have similar mediolateral widths), and the medial and lateral ones are convex and concave, respectively. The proximal surface slopes distomedially. This surface is rough and there are two ‘condyle’-shaped convex structures in the paratype specimen, dorsally and ventrally located, separated by a wide and smooth groove (this structure in condyles is not individualized in the holotype specimen). The dorsolateral apex of the proximal surface is proximally projected as in metatarsal I. The proximal surface extends to the diaphysis of the metatarsal from the dorsolateral and dorsomedial edges. From the dorsolateral edge of the proximal surface emerges the dorsolateral crest of the diaphysis, which reaches the distal condyle (connecting with the ridge that departs from the lateral distal condyle). The diaphysis is D-shaped in cross-section at midlength. The dorsal surface is transversely concave in the dorsal two-thirds and, distally, becomes flat (concave in the paratype specimen) between the proximodistal crests that departs from the lateral and medial distal condyles. In the paratype specimens, in this concave area, there is a small foramen (Fig. 19G). The dorsal surface transits continuously to the medial surface in the dorsal two-thirds of the diaphysis and it is separated by the proximodistal crest that comes from the medial distal condyle. The medial surface is transversely flat to transversely convex (with a shallow proximodistal groove in the paratype specimen). The medial surface of the distal condyle is smoothly concave. The ventral surface of metatarsal II is bordered laterally by the ventrolateral crest that connect the ventrolateral edges of proximal and distal surfaces (in the paratype this crest is interrupted at midlenght, and the proximal portion extends to the lateral surface of the diaphysis). The proximal tip of this ventrolateral crest (laterally displaced) is laterally deflected and projected (Figs 18I, 19I), which is considered as a possible autapomorphy of Garumbatitan morellensis. From the ventromedial edge of the proximal end departs a proximodistal crest only present in the dorsal one-third of the diaphysis (the transition between the ventral and medial surfaces of the diaphysis in the ventral two-thirds is continuous). The proximal half of the ventral surface is transversely concave and striated. The lateral surface is transversely concave and striated in the proximal half (for the reception of the metatarsal III) and flat in the distal half (marked by the proximodistal crest that extends from the lateral distal condyle in the paratype). In distal view, the distal condyles are dorsoventrally elongated, and the lateral one, is bevelled more than 20° from the sagittal plane of the metatarsal. The condyles are transversely and dorsoventrally convex and separated by an intercondylar groove ventrally marked (especially marked on the right metatarsal II of the holotype) (Figs 18L, 19L). The distal end does not have the ventrolateral projection present in diplodocids (Tschopp et al. 2015).
The metatarsals III and IV are slender and higher than metatarsals I and II, and the minimum transverse diameter of metatarsal III relative to the transverse diameter of the metatarsal I, and the minimum transverse diameter of metatarsal IV relative to the transverse diameter of the metatarsal I are both less than 65%. Four metatarsals III were recovered from the Sant Antoni de la Vespa fossil site [one right (SAV05-068.c, Fig. 18M–R) from the holotype; the right (SAV05-035.c, Fig. 19M–R) and the left (SAV05-038.a) from paratype specimen; and a referred left one (SAV05-056)]. The length of metatarsal III is 27–24% of the tibial length, unlike Tastavinsaurus sanzi, with 30% (Royo-Torres et al. 2012). Similar values are shown by the brachiosaurids Vouivria dampariensis (26%) and Sonorasaurus thompsoni (27%) (Mannion et al. 2013). The proximal surface is rough, flat, and has a teardrop shape (in SAV05-038.a and SAV05-056, this teardrop shape is lateromedially compressed due to deformation) in proximal view with a convex and concave medial and lateral edge, respectively, and a straight dorsal edge (ventral edge is constricted and mediolaterally shorter than the dorsal one) (Figs 18Q, 19Q). The proximal surface becomes transversely shorter dorsolaterally, and the dorsolateral edge curves distally, extending to the dorsolateral crest of the diaphysis. The dorsal surface of the diaphysis is bordered by the dorsolateral and the dorsomedial crests, connecting the dorsolateral and the dorsomedial edges of the proximal end to the distal end. The dorsal surface is flat and becomes concave proximally and distally (between the distal condyles), bearing two small foramina (Fig. 19M). Near the distal end, the dorsal surface of the dorsolateral and dorsomedial crests of the diaphysis preserve two pronounced bosses absent in the largest specimen. The lateral and medial surfaces of the metatarsal III are flat, but the lateral one becomes transversely concave in the proximal (striated surface) and distal end (with a small foramen in SAV05-035.c), The lateral and medial surfaces converge in the sagittal plane of the metatarsal III resulting in a constricted ventral face, corresponding to the ventrolateral crest of the diaphysis. The ventrolateral crest expands proximally, resulting in a flat platform, well-developed in the holotype specimen. The distal condyles are individualized and are convex transversely and dorsoventrally (the medial one is longer dorsoventrally than the lateral one), separated by a wide intercondylar groove that slightly progresses to the ventral surface (Figs 18R, 19R). The ratio of metatarsal III to metatarsal I proximodistal length is 1.34–1.37 and the ratio of metatarsal III to metatarsal IV proximodistal length is 1.06–1.07.
Four metatarsals IV were recovered in the Sant Antoni de la Vespa fossil site [the right (SAV05-068.d, Fig. 18S–X) and the left one (SAV05-042) from the holotype; the right one (SAV05-035.d, Fig. 19S–X) from paratype specimen; and a referred left one (SAV05-058)]. The metatarsal IV is the slenderest metatarsal. The proximal surface is quadrangular to subtrapezoidal (Fig. 19W), rough, and concave in the middle (as occurs in metatarsal II), near the medial edge. The medial and lateral edges are concave. The presence of a concave medial surface in the proximal end of the metatarsal IV (Figs 18W, 19W) characterizes many titanosauriforms (D’Emic et al. 2011, D’Emic 2012, Mannion et al. 2013, 2019b). The ventral edge of the proximal surface is shorter than the dorsal one. The dorsal edge of the proximal surface is straight, and in the holotype specimen, the proximoventral sector of the metatarsal is broken and displaced. The dorsolateral edge of the proximal surface deflects distally. The dorsal surface of diaphysis is transversely convex and becomes flat proximally and distally, between the distal condyles (there is a small foramen in this sector of the metatarsal). The dorsal face is bordered by a dorsolateral and dorsomedial crests only marked near the proximal and distal end. The medial face of the metatarsal is flat along its length, and proximally is marked by two proximodistal ridges: (i) a ventralmost ridge that corresponds to the ventromedial crest of the diaphysis (not associated with the ventrolateral crest of the diaphysis), which is not fully developed in the paratype specimen (Figs 18T, 19T); and (ii) a dorsalmost ridge that is proximodistally shorter and located near the ventromedial crest of the diaphysis (Fig. 19T), corresponding to prominent rugosities in the holotype specimen. The lateral surface is concave in the proximal end (this concavity extends to the distal half in the paratype specimen) and flat in the distal end. The lateral surface of the diaphysis faces ventrolaterally and converge with the medial surface ventrally resulting in a constricted ventral surface (=ventrolateral crest), as in the metatarsal III. This ventral ridge is double in the holotype specimen. The medial surface of the medial condyle has a small foramen in the holotype specimen. The distal condyles are individualized, with an intercondylar groove that slightly extends to the ventral surface of the metatarsal. The distal condyles are slightly dorsoventrally longer in the paratype specimen, and subcircular in the holotype specimen. The distal surface is perpendicular to the long axis of bone, differing from the medially bevelled distal end present in brachiosaurids (Mannion et al. 2013, D’Emic et al. 2016)
Two metatarsals V were found [a right one (SAV05-068.e, Fig. 18Y–AC) from the holotype; and a right one (SAV05-035.e, Fig. 19Y–AD) from the paratype specimen]. The metatarsal V has approximately 72% and 77% of the length of metatarsals III and IV, respectively, and is lateromedially compressed. The proximal surface (not preserved in holotype specimen) has an elliptical outline (the medial edge is straight and lateral one is convex), compressed lateromedially, and slopes laterally. The proximal and distal ends of metatarsal V expand dorsoventrally, with a marked proximal expansion (almost two times of the distal dorsoventral expansion), but not as marked as in Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009). The medial surface of metatarsal V is flat, and a marked crest extends from a boss-shaped structure near the distal face. The boss is much more pronounced in the holotype specimen and the ridge seems to be absent. The presence of a pronounced tuberosity near the ventromedial edge of the distal end of the metatarsal V (Figs 18Z, 19Z), is considered as an autapomorphy of Garumbatitan morellensis. The lateral surface of the diaphysis is dorsoventrally convex. The distal surface is rough, convex, and has a subcircular outline. The proximal end to the distal end maximum mediolateral width ratio is 1.67, differing from the greater ratio shown by Tastavinsaurus sanzi (2.32; Royo-Torres et al. 2012).
Garumbatitan morellensis, gen. et sp. nov., metatarsals from the holotype specimen. Right metatarsals I (SAV05-068.a), II (SAV05-068.b, second row), III (SAV05-068.c, third row), IV (SAV05-068.d, fourth row), V (SAV05-068.e, fifth row) in dorsal (A, G, M, S, Y), medial (B, H, N, T, Z), ventral (C, I, O, U, AA), lateral (D, J, P, V, AB), proximal (E, K, Q, W; dorsal towards top side), and distal (F, L, R, X, AC; dorsal towards top side) views. Abbreviations: amt, articulation for metatarsal; dlcr, dorsolateral crest; dmcr, dorsomedial crest; icg, intercondylar groove; lc, lateral condyle; mc, medial condyle; mt, metatarsal; prj, proximal projection of the ventrolateral crest in the metatarsal II; vlcr, ventrolateral crest; vmcr, ventromedial crest. Scale bar equals 50 mm.
Garumbatitan morellensis, gen. et sp. nov., metatarsals from the paratype specimen. Right metatarsals I (SAV05-35.a, first row), II (SAV05-035.b, second row), III (SAV05-035.c, third row), IV (SAV05-035.d, fourth row), V (SAV05-035.e, fifth row) in dorsal (A, G, M, S, Y), medial (B, H, N, T, Z), ventral (C, I, O, U, AA), lateral (D, J, P, V, AB), proximal (E, K, Q, W, AC; dorsal towards top side), and distal (F, L, R, X, AD; dorsal towards top side) views. Abbreviations: amt, articulation for metatarsal; dmcr, dorsomedial crest; icg, intercondylar groove; lc, lateral condyle; mc, medial condyle; mt, metatarsal; prj, proximal projection of the ventrolateral crest in the metatarsal II; vlcr, ventrolateral crest; vmcr, ventromedial crest. Scale bar equals 30 mm.
The possible phalangeal (including unguals) formula based on the holotype and paratype specimens is 2-3-3-3-0. The presence of three phalanges in pedal digit IV seems to be unique among eusauropods and only shared with Cedarosaurus weiskopfae (D’Emic 2013). The non-ungual phalanges are broader transversely than longer proximodistally. Three phalanges I.1 [a right one (SAV05-068.f, Fig. 20A–F) from the holotype; a right one (SAV05-035.f, Fig. 21A–F) from the paratype; and a referred left one (SAV05-057.b)]; and two phalanges II.1 [a right one (SAV05-068.g, Fig. 20G–L) from the holotype; a right one (SAV05-035.g, Fig. 21G–L) from the paratype] have been found at the fossil site. The phalanges I.1 and II.1 have a similar morphology. The dorsal surface of phalanx I.1 (mediolaterally narrower than the ventral one) is flat and medially and laterally bordered by rudimentary proximodistal crests (the dorsomedial and the dorsolateral crests), which separates it from the medial and lateral surfaces of the phalanx. The proximal surface is concave, has a semicircular-to-subrectangular outline (with a straight to concave ventral edge, Figs 20E, 21E), and is perforated by a foramen in SAV05-38.f. The dorsal and the ventral edges of the proximal surface are proximally projected, covering part of the distal end of metatarsal I. The medial surface of this phalanx is flat and perforated by small foramina in the holotype specimen. The lateral surface is dorsoventrally and proximodistally shorter than the medial one and marked by a depression near the dorsolateral crest, especially developed in the smaller individuals (considered herein as an autapomorphy of Garumbatitan morellensis). The ventral surface is proximodistally and mediolaterally concave and bears a small concavity near the medial distal condyle in the holotype and paratype specimen (absent in SAV05-057b). The distal condyles are individualized, extended to the ventral surface of the phalanx, dorsally connected, and ventrally projected; the medial condyle is shorter mediolaterally than the lateral one. In distal view, the medial condyle is medially bevelled (15–20°) and converges dorsally with the lateral one. The medial condyle is more ventrally projected than the lateral one. Both condyles are dorsoventrally elongated with sub-elliptical outlines. Near the ventral border of the distal surface, there is a wide intercondylar depression between the condyles.
The pedal phalanx II.1 is proximodistally longer than the phalanx I.1. In proximal view, phalanx II.1 has a semicircular-to-‘heart’-shaped (concave ventral edge) outline with a smooth, and flat-to-concave surface (more concave in the paratype specimen than in the holotype one) (Figs 20K, 21K). A proximal surface of phalanx II.1 with a ‘heart’-shaped outline is considered as autapomorphic of Garumbatitan morellensis. Similarly, to pedal phalanx I.1, the dorsal surface of phalanx II.1 is flat and mediolaterally shorter than the ventral one (in the holotype, the dorsal surface is mediolaterally convex and poorly preserved). The dorsal surface is bordered by the proximodistal elongated dorsomedial and dorsolateral crests. The medial surface is also marked by a depression near the distal end, but shallower than in pedal phalanx I.1 (less pronounced in the holotype specimen). The lateral surface is flat, and the ventral surface is concave to smoothly concave in the paratype specimen (two small foramina are present in the paratype specimen). In the ventralmost area of the ventral surface, there is a small concavity near the distal end (absent in the holotype specimen). The distal condyles are individualized (dorsally connected), extending to the ventral face of the phalanx. An intercondylar depression is present between the distal condyles and near the ventral edge of the distal surface. This depression does not extend to the ventral surface of the phalanx. The condyles are dorsoventrally elongated (the medial one is lateromedially wider than the lateral one).
Only one right phalanx III.1 (SAV05-068.h, Fig. 20M–R) was recovered belonging to the holotype specimen. It has a flat dorsal surface and bears some smooth rugosities. The lateral surface of this phalanx is dorsoventrally shorter than the medial one and it extends continuously to the dorsal surface. The medial surface is flat, separated from the dorsal edge by a smooth dorsomedial crest. The outline of proximal surface is unknown (the ventral edge is broken), but the dorsal edge is convex. The proximal surface is concave. The dorsal edge is slightly proximally projected. The distal end is covered by pedal phalanx III.2, and so it is not possible to describe it. The distal condyles extend to the ventral surface of the phalanx.
Two pedal phalanges IV.1 [a right one (SAV05-068.i, Fig. 20S–X) from the holotype; and a right one (SAV05-035.h, Fig. 21Y–AD) from the paratype] are recognized (the distal condyle in phalanx IV.1 of the holotype is eroded). They are more dorsoventrally compressed than pedal phalanx III.1. The phalanx IV.1 of the paratype specimen seems to be mediolaterally more elongated than that of the holotype. The proximal surface is smoothly concave in the paratype and concave in the holotype, preserving an elliptical outline (mediolaterally elongated). The dorsal surface is concave proximodistally and convex mediolaterally. This surface is separated from the ventral surface by the ventromedial and ventrolateral crests (these crests are not pronounced). The ventral surface is concave in the paratype and flat in the holotype specimen (with two small foramina). The distal condyles are transversely expanded in the paratype specimen; however, this expansion is not visible in the holotype specimen due to the preservation. The condyles are not well-individualized and only slightly extended to the ventral surface.
There is one preserved pedal phalanx II.2 [the right one (SAV05-035.j, Fig. 21M–R) from the paratype]. The phalanx II.2 is wedge-shaped in dorsal view being proximodistally constricted on the lateral side, producing a tongue-shaped structure. This phalanx is suboval in proximal view and bears a concave proximal surface with small circular foramina and some dorsoventral struts. The ventromedial edge of the proximal surface is ventrally projected bordering dorsally the ventral surface of the phalanx. The ventral surface is small and concave and perforated by a small foramen. The dorsal surface is proximodistally concave and extends continuously to the medial side up to the ventromedial crest, which separates the ventral surface from the medial one. The distal surface is marked two convex medial and lateral condyles, not well individualized dorsally but separated by a wide intercondylar concavity in the ventral half of the distal surface. The condyles extend to the ventral surface of the phalanx.
Two right phalanges III.2 were found: one in articulation with pedal phalanx III.1 in the holotype specimen (Fig. 20M–R), and another one (SAV05-038.b, Fig. 21S–X) found near the left metatarsal III (SAV05-38.a), and it is considered as belonging to the right hindlimb of the paratype specimen. The pedal phalanx III.2 is larger than the phalanx II.2. The phalanx III.2 is wedge-shaped in dorsal view being proximodistally constricted on the lateral side (less proximodistally constricted than in the pedal phalanx II.2). This phalanx has a teardrop-shaped outline in proximal view (laterally constricted and with a ventral straight edge). The proximal surface is dorsoventrally concave and mediolaterally convex, rough, and bears some small foramina. The ventral edge of the proximal surface is ventrally projected, bordering dorsally the ventral surface, producing a tongue-shaped structure. The dorsal surface is flat, triangular (laterally constricted), and is separated from the medial face by the dorsomedial crest. The medial surface is flat to slightly concave. The ventral surface is sub-rectangular in ventral view but laterally constricted, being lateromedially and proximodistally concave (two foramina are present near the lateral edge). The distal surface is marked by two convex medial and lateral condyles, well individualized and dorsally connected but separated by a wide intercondylar concavity in the ventral half of the distal surface. The condyles extend do the ventral surface of phalanx and diverge ventrally (the medial condyle is dorsoventrally oriented and the lateral one is dorsomedially-ventrolaterally oriented).
Three pedal phalanx IV.2 were recognized [a right one (SAV05-068.j, Fig. 20Y–AD) from the holotype; a right one (SAV05-035.i, Fig. 21AE–AJ) from the paratype; and a referred left one (SAV05-057.c)]. The pedal phalanx IV.2 is markedly proximodistally shorter than phalanx IV.1, and extremely mediolaterally larger than proximodistally. The proximal surface is elliptical and dorsoventrally compressed in proximal view. The proximal surface is flat (straight or smoothly convex mediolaterally). In the holotype the ventral edge of the proximal surface is proximally projected. The dorsal surface is continuous to the lateral and medial surfaces, resulting in a mediolateral convex surface and proximodistally concave. This surface is separated from the ventral surface by the ventromedial and ventrolateral crests. The distal condyles are individualized with a smooth intercondylar depression in the middle. They are projected ventrally and extend significantly to the ventral surface of the phalanx. The ventral surface of the phalanx is smoothly concave with a tuberosity in its medial half.
Garumbatitan morellensis, gen. et sp. nov., pedal phalanges from the holotype specimen. Right phalanx I.1 (SAV05-068.f, first row), II.1 (SAV05-068.g, second row), III.1 and III.2 (SAV05-068.h, third row), IV.1 (SAV05-068.i, fourth row), IV.2 (SAV05-068.j fifth row) and IV.3 (SAV05-068.m, sixth row) in dorsal (A, G, M, S, Y, AE), medial (B, H, N, T, Z, AF), ventral (C, I, O, U, AA, AG), lateral (D, J, P, V, AB, AH), proximal (E, K, Q, W, AC, AI; dorsal towards top side), and distal (F, L, R, X, AD, AJ; dorsal towards top side) views. Abbreviations: dp, depression; fr, foramen; icg, intercondylar groove; lc, lateral condyle; mc, medial condyle. Scale bar equals 30 mm.
Garumbatitan morellensis, gen. et sp. nov., pedal phalanges from the paratype specimen. Right phalanx I.1 (SAV05-035.f, first row), II.1 (SAV05-035.g, second row), II.2 (SAV05-035.j, third row), III.2 (SAV05-038.b, fourth row), IV.1 (SAV05-035.h, fifth row), IV.2 (SAV05-035.i, sixth row) and IV.3 (SAV05-035.m, seventh row) in dorsal (A, G, M, S, Y, AE, AK), medial (B, H, N, T, Z, AF, AL), ventral (C, I, O, U, AA, AG, AM), lateral (D, J, P, V, AB, AH, AN), proximal (E, K, Q, W, AC, AI, AO; dorsal towards top side), and distal (F, L, R, X, AD, AJ, AP; dorsal towards top side) views. Abbreviations: dp, depression; icg, intercondylar groove; lc, lateral condyle; mc, medial condyle. Scale bar equals 30 mm.
Two right pedal unguals I.2 [almost complete, from both holotype (SAV05-068.k, Fig. 22A–E) and paratype (SAV05-035.k, Fig. 23A–E)]; one right pedal ungual II.3 [from the holotype specimen (SAV05-068.l, Fig. 22F–J), which is badly preserved and separated in two pieces; and fragments from the left one of the paratype (SAV05-038.c)]; one right pedal ungual III.3 [from the paratype specimen (SAV05-35.l, Fig. 23F–J)]; and two right IV.3 [a complete one from the holotype (SAV05-068.m, Fig. 20AE–AJ); and one from the paratype (SAV05-035.m, Fig. 21AK–AP), which is badly preserved] were identified. Except the fourth ungual, all the unguals are sickle-shaped, much deeper dorsoventrally than broader transversely, with a convex dorsal edge, and concave ventral edge in lateral view. The ungual I.2 is less curved than the one from Tastavinsaurus sanzi (Canudo et al. 2008). The ungual I.2 in the holotype specimen is dorsoventrally deeper than the ungual I.2 of the paratype specimen. They are mediolateraly compressed and bevelled laterally. The lateral and medial surfaces are dorsoventrally convex, and, in unguals I.2 and II.3, there is a longitudinal groove on both sides, the lateral one deeper than the medial one. The lateral groove does not curve and intersects the dorsal edge of the ungual at the distal end. The proximal surface is concave, dorsally constricted, and laterally deflected (this deflection is not present in Tastavinsaurus sanzi,Canudo et al. 2008). The phalanx III.3 lacks lateral and medial grooves, and it is markedly shorter than the ungual of the digits I and II, corresponding to 33% of the metatarsal III length, an autapomorphy of Garumbatitan morellensis. Near the proximal surface, there is a round tuberosity in the transition between the medial and ventral surfaces in the unguals I.2 and III.3. In the ungual I.2, there is another tuberosity in the distal half of the ventral edge, which appears in many titanosauriforms (Canudo et al. 2008, Mannion et al. 2013, 2019a, b). The right phalanx IV.2 is rudimentary and oval (dorsoventrally compressed). The medial side is dorsoventrally constricted. The proximal surface is pierced by two foramina and the dorsal surface is transversely concave. The phalanx of the fifth toe is considered absent, also characteristic of Garumbatitan morellensis.
Garumbatitan morellensis, gen. et sp. nov., right pedal unguals I.2 (SAV05-035.k, A–E) and II.3 (SAV05-35.l, F–J) from the holotype specimen in dorsal (A, F), lateral (B, G), medial (C, H), proximal (D, I), and ventral (E, J) views. Abbreviations: dcr, dorsal crest; fr, foramen; gr, groove; pf, posterior face; tb, tuberosity. Scale bar equals 30 mm.
Garumbatitan morellensis, gen. et sp. nov., right pedal unguals I.2 (SAV05-068.k, A–E) and III.3 (SAV05-068.l, F–J) from the paratype specimen in dorsal (A, F), lateral (B, G), medial (C, H), proximal (D, I), and ventral (E, J) views. Abbreviations: dcr, dorsal crest; gr, groove; pf, posterior face; tb, tuberosity. Scale bar equals 50 mm.
Phylogeny
To provide a phylogenetic analysis for Garumbatitan morellensis we use the dataset of Poropat et al. (2023), which results from the updated version of the ‘LSDM’ data matrix of Mannion et al. (2013), incorporating the augmented version of this matrix presented in Upchurch et al. (2015), Poropat et al. (2016, 2021), Mannion et al. (2017, 2019a, b), and González Riga et al. (2018). The scoring of Lusotitan atalaiensis was updated based on the recent information provided by Mocho et al. (2017a) and personal observations, which resulted in the following re-scoring: 48 (0→0&1), 59 (?→1), 145 (? →0), 162 (1→?), 188 (1→0), 189 (?→1), 191 (?→1), 206 (0→1), 210 (?→1), 211 (?→1), 235 (?→0), 334 (0→?), 372 (1→0), 419 (0→?), 481 (0→?), 483 (0→?), 500 (0→?), 524 (?→0), 525 (?→0). Some characters of Tastavinsaurus sanzi were re-scored based on personal observations: C59 (0→0&1), C162 (0→1), 163 (?→0), C168 (?→1), C169 (?→0), C175 (0→?), 256 (0→1). The data matrix (Mesquite and TNT files) is provided in the Supporting Information, File S3. We also incorporate the scoring of Oceanotitan dantasi based on Mocho et al. (2019a) and personal observations. The holotype of Diamantinasaurus matildae (AODF 0603) and the referred specimens, AODF 0836 and AODF 0906, were combined as a single operational taxonomic unit (OTU). First, and supported by our description and comparative analyses, the two most complete specimens are scored together in the same taxonomic unit for the first two analyses. However, we repeated the first two analyses with the holotype and paratype separately to check and confirm if the analysis also supports our a priori hypothesis of both belonging to the same taxon.
Following the recent iterations of this data matrix by Mannion et al. (2019a, b) and Poropat et al. (2021, 2023), the characters 11, 14, 15, 27, 40, 51, 104, 122, 147, 148, 195, 205, 259, 297, 426, 435, 472, and 510 were treated as ordered multistate characters, and eight unstable and highly incomplete taxa were excluded a priori (Astrophocaudia, Australodocus, Brontomerus, Fukuititan, Fusuisaurus, Liubangosaurus, Malarguesaurus, and Mongolosaurus). This pruned dataset was analysed using the ‘Stabilize Consensus’ option in the ‘New Technology Search’ in TNT v.1.5 (Goloboff et al. 2008, Goloboff and Catalano 2016), to find the most-parsimonious trees (MPTs), using sectorial searches, drift, and tree fusing, with the consensus stabilized five times. After, we used the resultant trees as the starting topologies to perform a ‘Traditional Search’, using tree bisection–reconnection. We ran two versions of this analysis: in the first one we used equal weighting of characters and in the second we applied extended implied weighting (Goloboff et al. 2006, 2018, Goloboff 2014, Mannion et al. 2017). We applied a k-value of nine, following the analyses performed by Mannion et al. (2019b) and Poropat et al. (2021, 2023), which are based on the recommendations of Goloboff (2014) and Tschopp and Upchurch (2019). However, some authors have suggested that the implied weighting might propagate errors and result in poorer topological accuracy (Congreve and Lamsdell 2016, O’Reilly et al. 2016). To test the robustness of the phylogenetic hypotheses, bootstrap values (absolute frequencies based on 5000 replicates) were obtained using TNT 1.5. (Goloboff et al. 2008). The list of synapomorphies for the analysis II is documented in the Supporting Information (Supporting Information, File S2).
RESULTS
Analysis I
The initial parsimony analysis considered equal weights. The ‘New Technology Search’ yielded 208 trees of length 2714 steps (CI: 0.216; RI: 0.596). These 208 trees were then used as the starting trees for a ‘Traditional Search’ using tree bisection–reconstruction (TBR). This analysis recovered more than 999 999 most parsimonious trees (MPTs) of length 2714 steps without finishing (CI: 0.206; RI: 0.572). A strict consensus tree from this interrupted search showed little resolution throughout most of the topology (Fig. 24A). Garumbatitan morellensis is included in a wide polytomy together with Tastavinsaurus, Europatitan, and Late Jurassic brachiosaurids. Within this polytomy, some groups have been recovered as monophyletic, such as Diplodocoidea, Turiasauria, or Wintonotitan + deeply nested somphospondylans.
Phylogenetic relationships of Garumbatitan morellensis within more deeply nested eusauropods obtained from the Poropat et al. (2023) data matrix: A, strict consensus cladogram from 999 999 most parsimonious trees (MPTs) of length 2714 steps using equal weights; B, strict consensus cladogram from 231 most parsimonious trees (MPTs) of length 140.80561 steps using extended implied weights with a k-value of 9. Note that these trees were produced following the a priori exclusion of eight unstable taxa: Astrophocaudia, Australodocus, Brontomerus, Fukuititan, Fusuisaurus, Liubangosaurus, Malarguesaurus, and Mongolosaurus.
Analysis II
The present dataset was analysed using implied weighting in TNT with a k-value of nine. The ‘New Technology Search’ yielded 82 trees of length 140.80561 steps (CI: 0.214; RI: 0.591). These 82 trees were then used as the starting trees for a ‘Traditional Search’ using TBR, which produced 231 MPTs of length 140.80561 steps (CI: 0.214; RI: 0.591), with an appreciable resolved strict consensus (Fig. 24B). The Pruned Trees’ option in TNT recognized Abydosaurus, Nopcsaspondylus, Padillasaurus, and Sarmientosaurus as the most unstable taxa. Bootstrap values within Macronaria are lower than 50%, except from Camarasaurus + Lourinhasaurus.
The general topology obtained in this analysis is better resolved than in the ones obtained in the first analysis and it shows some differences relative to the topology achieved in the implied weight analyses performed by Mannion et al. (2019a, b) and Poropat et al. (2021). In this topology, Tehuelchesaurus is placed outside of Macronaria, within a clade including Bellusaurus, Haestasaurus, and Janenschia in a more basally branching position than Camarasaurus + Lourinhasaurus. The topology at the base of Somphospondyli is slightly different, with the inclusion of Garumbatitan morellensis and allocation of Tastavinsaurus to a more basally branching position within Somphospondyli. An important polytomy is found at the base of the clade that includes more derived somphospondylans than Tastavinsaurus (see Fig. 24B), which includes Padillasaurus, Sauroposeidon, ‘Paluxysaurus’, Ligabuesaurus, Cloverly titanosauriform, and a clade with the remaining somphospondylans. Garumbatitan morellensis is recovered as a member of Somphospondyli, in a more deeply nested position than Dongbeititan and in a more basally branching position than Tastavinsaurus. The inclusion of Garumbatitan morellensis within Titanosauriformes is supported by the presence of (i) camellate internal tissue texture in middle-posterior dorsal vertebrae (C141); and (ii) small, shallow, vascular foramina piercing the lateral and/or ventral surfaces of the anterior-middle caudal centra (C180). Somphospondyli is supported by seven synapomorphies, but the presence/absence of none of them can be identified in the available material of Garumbatitan. Dongbeititan is recovered as the more basally branching somphospondyli, followed by Garumbatitan morellensis. The somphospondylan clade that includes Garumbatitan morellensis and more deeply nested somphospondylans is supported by the following six synapomorphies: (i) dorsoventral height of haemal canal divided by total chevron height is 0.40 or greater in anterior chevrons (C35); (ii) trochanteric shelf in the posterior surface of the femoral proximal end (C256); (iii) anteromedialy directed crest in the fibular proximal end (extending into a notch behind the cnemial crest of the tibia) (C262); (iv) lateral muscle scar of the fibula is formed from two vertically elongate, parallel ridges (C263); (v) absence of a tongue-like projection posteromedial to the ascending process in the astragalus (C269); (vi) undivided posterior fossa of the astragalus (C552). The present analysis recovered eight local autapomorphies for Garumbatitan morellensis: (i) lateral pneumatic fossae in the anterior caudal centra (C178); (ii) linea intermuscularis cranialis in the anterior surface of the femur (C257); (iii) fourth trochanter visible in anterior view (C258; not included in the diagnosis because this feature might be effected by deformation); (iv) calcaneum is absent (C270); (v) pedal digit IV having at least three phalanges (C277); (vi) posterior articular surface more deeply concave than the anterior one in anterior-middle caudal centra (C350); (vii) ratio of the mediolateral breadth of the tibial condyle to the breadth of the fibular condyle of the femur is 0.8 or less (C389); and (viii) metatarsal I with a tubercle on medial surface (situated at approximately midlength and equidistant from the dorsal and the ventral margins) (C540). Tastavinsaurus presents a close phylogenetic position to the Sant Antoni del la Vespa sauropod and appears as the following deeply nested somphospondylan member. The clade that includes Tastavinsaurus and the remaining somphospondylans is supported by five synapomorphies, which are absent in Garumbatitan morellensis. In addition, 18 local autapomorphies are recovered for Tastavinsaurus: (i) 11 of them are absent in Garumbatitan morellensis; (ii) five are not possible to confirm in the available of material, and, finally, (iii) two of them show polymorphism, and they share one of the conditions.
Soriatitan is recovered as a member of Brachiosauridae (this clade is supported by nine synapomorphies) within a clade that includes the Early Cretaceous brachiosaurids Venenosaurus, Cedarosaurus, and Abydosaurus from the North American continent. Europatitan is placed as more deeply somphospondylan, in a non-titanosaurian clade, which includes, Chubutisaurus, Angolatitan, Europatitan, and Huanghetitan.
A second run of our analysis was performed, where both holotype and paratype specimens of Garumbatitan morellensis were analysed separated, to test our hypothesis that both belongs to the same taxon. The ‘New Technology Search’ yielded 90 trees of length 140.88155 steps (CI: 0.214; RI: 0.591). These 90 trees were then used as the starting trees for a ‘Traditional Search’ using TBR, which produced 231 MPTs of length 140.88155 steps (CI: 0.214; RI: 0.591). The recovered topology is the same as the one recovered in our first version of the analyses (with implied weight), and the holotype and paratype specimens shared the same node, which is supported by five features: (i) the anterior surface of the femoral shaft preserves a linea intermuscularis cranialis (C257); (ii) the calcaneum is absent (C270); (iii) at least three phalanges in the pedal digit IV (C277); (iv) ratio of the mediolateral breadth of the tibial condyle to the breadth of the fibular condyle in the femur is 0.8 or less (C389); and (v) metatarsal I with a rugosity on the medial surface (at approximately midlength) (C540).
DISCUSSION
The equal weighting analysis is poorly resolved and, in our implied weighting analyses, Garumbatitan morellensis is recovered as an early branching somphospondylan and the strict consensus is well resolved. This phylogenetic position is also supported by the presence of several characters in Garumbatitan morellensis that are commonly shared by somphospondylan titanosauriforms.
The detailed study of this new titanosauriform allowed us to obtain new information to better understand the phylogenetic distribution of some anatomical features, and the appearance of some somphospondylan novelties. As mentioned earlier, in the description of Garumbatitan morellensis (see Systematic palaeontology), we identified a set of features that supports the placement of this taxon within Titanosauriformes (or within a more inclusive group but more derived than the Camarasauridae, depending on the considered phylogeny), such as: (i) camellate tissue bone in the presacral vertebrae (e.g. Wilson 2002, Upchurch et al. 2004, Carballido et al. 2011b, Mannion et al. 2013, 2019b); (ii) pneumatic foramen and internal cavities in dorsal ribs (Wilson and Sereno 1998, Wilson 2002, Mannion et al. 2013); (iii) anterior dorsal ribs with a plank-like cross-section in the distal end (Wilson and Sereno 1998,Wilson 2002; recovered as a synapomorphy of Macronaria by: Mannion et al. 2013); (iv) small, shallow vascular foramina in lateral and ventral surfaces of anterior-middle caudal centra (Mannion et al. 2013); (v) posterior deflection of the caudal ribs (Mannion et al. 2013, 2019b); (vi) anterior displacement of neural arch in middle caudal vertebrae (e.g. Salgado et al. 1997); (vii) haemal canal is more than 40% of the total height of the chevron (e.g. Tidwell et al. 1999, Wilson 2002, Mannion and Calvo 2011, Mannion et al. 2013, 2019b, Mocho et al. 2017a, Royo-Torres et al. 2017a); (viii) well-developed lateral bulge in the femur (Salgado et al. 1997, Wilson 2002, Carballido et al. 2011b, Mannion et al. 2013; see Femoral lateral bulge and femoral shaft eccentricity); (ix) transverse diameter of the femoral shaft is approximately three times the anteroposterior shaft length (Mannion et al. 2013; see Femoral lateral bulge and femoral shaft eccentricity); (x) posterior margin of astragalus lacks a tongue-like projection posteromedial to the ascending process (Mannion et al. 2013); (xi) a concave medial surface in the proximal end of the metatarsal IV (D’Emic et al. 2011, D’Emic 2012, Mannion et al. 2013, 2019b); (xii) ungual I.2 with a tuberosity in the distal half of the ventral edge (Canudo et al. 2008, Mannion et al. 2013, 2019b).
The presence of brachiosaurids in the upper Hauterivian–lower Aptian deposits of the Iberian Peninsula was recently confirmed with the establishment of Soriatitan golmayensis (Royo-Torres et al. 2012). The phylogenetic position of Tastavinsaurus sanzi was also controversial, and some phylogenetic analyses positioned it within this clade (Royo-Torres et al. 2012, 2014, 2017a, Mocho et al. 2014). Garumbatitan morellensis shares some features that have been recovered as diagnostic of Brachiosauridae. One of these features is the presence of pneumatic fossa lacking sharply defined margins on the lateral surface of anterior caudal vertebrae; however, this feature also appears in some non-titanosaurian somphospondylans, such as Tastavinsaurus sanzi and Padillasaurus leivaensis (Mannion et al. 2013, 2017, 2019b). Another feature that was considered a synapomorphy of Brachiosauridae is the presence of a tuberosity on the medial surface of the metatarsal I (D’Emic et al. 2016, Mannion et al. 2019b), which is also present in the metatarsal I of Garumbatitan morellensis. The topology of the strict consensus tree from the performed implied weight analysis, which is much better resolved, suggests that its presence is a convergence brachiosaurids. A femoral fourth trochanter, visible in anterior view, was also recovered as an ambiguous synapomorphy of Brachiosauridae by Mannion et al. (2013, based on the LCDM dataset), and it is also present in Garumbatitan morellensis. This feature has been identified in several other sauropod forms, including diplodocoids, non-neosauropod eusauropods, and somphospondylans (e.g. Gallina and Apesteguía 2005, Curry Rogers 2009, Mannion et al. 2019a, b).
On the other hand, Garumbatitan morellensis seems to lack some of the features identified as brachiosaurid synapomorphies (or synapomorphies of brachiosaurid subclades). One of these, is the presence of a metatarsal IV with a medially bevelled distal end (see: D’Emic 2012, Mannion et al. 2013), which is observed in several brachiosaurids (Janensch 1961, Tidwell et al. 2001, Chure et al. 2010, D’Emic et al. 2016). The clade containing Abydosaurus, Cedarosaurus, and Venenosaurus in the topology recovered by D’Emic (2012) is characterized by taxa with anteriorly oriented neural spines in middle caudal vertebrae and reduced femoral fourth trochanter, both of which are absent in the holotype and paratype specimens of Garumbatitan morellensis.
The analysis using extended implied weights recovered Garumbatitan morellensis as an early branching somphospondylan more deeply nested than Dongbeititan dongi, and more basally branched than Tastavinsaurus sanzi. Below, we will comment on some anatomical features that are present in Garumbatitan morellensis and their occurrence along the somphospondylan lineage. The presence of some synapomorphies of Somphospondyli has been identified in Garumbatitan morellensis (depending on the phylogenetic approach, they can characterize more exclusive somphospondylan clades). In the phylogenetic hypothesis proposed by D’Emic (2012) some features in the fibula seem to be diagnostic for Somphospondyli. One of these features is the presence of an anteromedially directed crest in the proximal end of the fibula extending into a notch behind the tibial cnemial crest that is present in most somphospondylans and absent in brachiosaurids (D’Emic 2012, Mannion et al. 2013). Its presence was recovered as a synapomorphy of Sauroposeidon + (Tastavinsaurus + (Euhelopodidae + (Chubutisaurus + Titanosauria))) by D’Emic (2012). Another feature observed in the fibula of Garumbatitan morellensis and common in somphospondylans is the presence of a complex lateral trochanter consisting of a tuberosity bounded by two proximodistal ridges (Upchurch 1998, Mannion et al. 2013, 2017). The absence of a posterior tuberculum in the posterior proximal surface of the astragalus is also common in somphospondylans (D’Emic 2012), as well as a metatarsal IV with an embayment in the medial surface of the proximal end [D’Emic 2012; but considered as synapomorphy of Titanosauriformes by Mannion et al. (2013), based on the LCDM dataset)]. All these features are also present in the possible somphospondylan Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009, Royo-Torres et al. 2012).
However, some diagnostic features of somphospondylans are absent in Garumbatitan morellensis, suggesting that they might be diagnostic of a more exclusive somphospondylan clade, being acquired later in the early evolution of this titanosauriform lineage. Somphospondylans are characterized by an elongated prezygapophyseal process that extends beyond the anterior margin of the centrum less than 20% of the centrum length (excluding articular condyle) (Mannion et al. 2013, 2019b). However, this feature is absent in the early branching Garumbatitan morellensis, suggesting that it is acquired by more deeply nested somphospondylans such as Tastavinsaurus sanzi. Another somphospondylan feature (also present in Tastavinsaurus sanzi) is the presence of a femur with a tibial to fibular condylar anteroposterior length ratio of 1.2 or greater (Mannion et al. 2013), which differs from Garumbatitan morellensis. The presence of a subcircular proximal section in the tibia is a synapomorphy of the neosauropods (e.g. Wilson 2002, Upchurch et al. 2004, Carballido and Sander 2014) that is present in Garumbatitan morellensis. However, the proximal section of the tibia is transversely compressed in several somphospondylans, as in Sauroposeidon proteles (Rose 2007), Euhelopus zdanskyi (Wiman 1929), Phuwiangosaurus sirindhornae (Martin et al. 1999), or Huabeisaurus allocotus (D’Emic et al. 2013). The absence of a ‘second cnemial crest’ characterizes several somphospondylans (Mannion et al. 2013, 2017, 2019b) but it seems to be present in some early branching somphospondylans such as Garumbatitan morellensis, suggesting that this feature appeared in more deeply nested somphospondylans, and may be reversed in Phuwiangosaurus sirindhornae and Ligabuesaurus leanzai (Mannion et al. 2013). The presence of sigmoidal fibular shaft characterizes many somphospondylans, including Tastavinsaurus sanzi (D’Emic 2012, Mannion et al. 2013), but it is absent in the recovered fibulae of the sauropod from Sant Antoni de la Vespa. The lateral trochanter of fibula is not laterally pronounced in some somphospondylans (e.g. Ksepka and Norell 2006, Salgado and Carvalho 2008, Otero 2010, Díez Díaz et al. 2013b, Lacovara et al. 2014). Following the phylogenetic hypothesis of D’Emic (2012), the members of Somphospondyli also share a medially unexpanded fibular distal end, lacking a medial lip, unlike Garumbatitan morellensis and Tastavinsaurus sanzi, which are featured by the plesiomorphic condition of the Titanosauriformes clade: a well-developed distal medial lip in the fibula.
In some phylogenetic approaches it has been suggested that the presence of a distal tibial end with a transverse width two times or greater than the anteroposterior width is a synapomorphy of Titanosauria (Wilson 2002) or Lithostrotia [Mannion et al. (2013), based on LSDM dataset]. This feature is present in the holotype of Garumbatitan morellensis but absent in the paratype one, suggesting that this ratio may vary during ontogeny, with possible allometry of the distal tibial end.
Femoral lateral bulge and femoral shaft eccentricity
The presence of a well-developed lateral bulge in the femur has been considered a common feature of Titanosauriformes and some stem forms such as Tehuelchesaurus benitezii (Carballido et al. 2011b) and Aragosaurus ischiaticus (Royo-Torres et al. 2014) and was considered as a synapomorphy of the group in several phylogenetic analyses (e.g. Wilson 2002, Upchurch et al. 2004, D’Emic 2012). Nevertheless, some authors have pointed out the difficulty of identifying intermediate situations as in Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009, Royo-Torres et al. 2012) or Tehuelchesaurus benitezii (Carballido et al. 2011b), describing more formally restricted conditions as in Salgado et al. (1997) and Mannion et al. (2013) (Fig. 25). Along sauropod evolution, some diplodocids such as Diplodocus carnegii (Hatcher 1901), Tornieria africana (Remes 2006), Apatosaurus louisae (Gilmore 1936, CM 3018), and an unnamed diplodocid from the Portuguese Upper Jurassic also acquired a pronounced lateral bulge (Mocho et al. 2017c). Salgado et al. (1997) tentatively quantified the presence of a pronounced lateral bulge as ‘the distance from a straight line that contains the greater trochanter and the lateral point of the femoral shaft where it reaches its minimum transverse width up to a parallel that comprises the outset point of the lateral bulges, is greater than 30 percent the minimum transverse width of the shaft’, which was recovered as a synapomorphy of Titanosauriformes (Salgado et al. 1997: 16). In this case, the bulge of Garumbatitan morellensis shares the apomorphic condition with a bulge equal to 44% of the minimum transverse width of the shaft. A lateral bulge being greater than 30% of the minimum transverse width of the shaft is shared by many titanosauriforms such as Giraffatitan brancai (37%), ‘Paluxysaurus jonesis’ (37%), Phuwiangosaurus sirindhornae (34%), Rapetosaurus krausei (33%), Saltasaurus loricatus (38%), Neuquensaurus australis (40%), Ampelosaurus atacis (33%), and Lirainosaurus astibiae (41%), among others (Fig. 25). However, some somphospondylans have less pronounced lateral bulges, such as Euhelopus zdanskyi (23%), Dongbeititan dongi (25%), Tastavinsaurus sanzi (17%), and Opisthocoelicaudia skarzynskii (27%). Apart from Giraffatitan brancai, other brachiosaurids are characterized by poorly developed lateral bulges (Vouivria dampariansis, 18%; Brachiosaurus altithorax, 22%), suggesting that the apomorphic condition of Giraffatitan brancai (i.e. the presence of a developed lateral bulge) is a convergence with most somphospondylans. In addition, and considering the topology obtained from the strict consensus of our implied weighting analyses, early branching somphospondylans are characterized by poorly developed lateral bulges (e.g. Tastavinsaurus sanzi, Dongbeititan dongi), suggesting that the condition of Garumbatitan morellensis was independently acquired and represents a convergence with more deeply nested somphospondylans and Giraffatitan brancai.
Schematic draws of eusauropod femora (not to scale) showing the development of the lateral bulge following Salgado et al. (1997, in blue) and the medial deflection of the proximal end following Mannion et al. (2013): A, Tehuelchesaurus benitezii in posterior view (Carballido et al. 2011b); B, Apatosaurus louisae in anterior view (Gilmore 1936, CM 3018); C, Camarasaurus grandis in anterior view (YPM 1901, Ostrom and McIntosh 1966); D, Lourinhasaurus alenquerensis in posterior view (Mocho et al. 2014, MG 4931); E, Aragosaurus ischiaticus in posterior view (ZH-2, Royo-Torres et al. 2014); F, Vouivria dampariensis in posterior view (MNHN.F.1934.6 DAM 36, Mannion et al. 2017); G, Brachiosaurus altithorax in anterior view (FMNH P 25107, Riggs 1903); H, Giraffatitan brancai in posterior view (HMN St 291, Janensch 1961); I, Dongbeititan dongi in posterior view (Wang et al. 2007); J, Garumbatitan morellensis in posterior view (SAV05-023); K, Phuwiangosaurus sirindhornae in posterior view (Martin et al. 1999); L, Euhelopus zdanskyi in posterior view (PMU 234, Wiman 1929); M, Tastavinsaurus sanzi in anterior view (MPZ 99/9, Royo-Torres 2009); N, ‘Paluxysaurus jonesi’ in anterior view (Rose 2007); O, Daxiatitan binglingi in posterior view (You et al. 2008); P, Opisthocoelicaudia skarzynskii in posterior view (Borsuk-Białynicka 1977); Q, Rapetosaurus krausei in anterior view (Curry Rogers 2009, FMNH PR 2209); R, Saltasaurus loricatus in posterior view (Powell 1992, PVL 4017); S, Neuquensaurus australis in anterior view (MLP-CS 9, Otero 2010); T, Lirainosaurus astibiae in posterior view (MCNA 7468, Díez Díaz et al. 2013); U, Ampelosaurus atacis in posterior view (Le Loeuff 2005, Vila et al. 2012). *Reversed.
Based on Royo-Torres (2009) and Royo-Torres et al. (2012), Mannion et al. (2013) proposed a different way to evaluate the prominence of the femur lateral bulge (C255) with the following formulation: ‘femur, proximolateral margin, above the lateral bulge: level with or lateral to the lateral margin of the distal half of the shaft (0); medial to the lateral margin of the distal half of the shaft (1)’. Garumbatitan morellensis bears the derived state shared by most of titanosauriforms, excluding, for example, Tastavinsaurus sanzi, and more derived forms such as Saltasaurus loricatus, Patagotitan mayorum, Rapetosaurus krausei, Opisthocoelicaudia skarzynskii, Alamosaurus sanjuanensis, and Ampelosaurus atacis (Borsuk-Białynicka 1977, Powell 1992, Curry Rogers 2009, Royo-Torres 2009, Vila et al. 2012, Mannion et al. 2013, Carballido et al. 2017) (see Fig. 25).
Several authors have used the eccentricity of the femoral shaft (i.e. mediolateral width to anteroposterior width ratio at midshaft) in their morphological datasets (e.g. Wilson 2002, Upchurch et al. 2004, Carballido et al. 2012, 2017, Mannion et al. 2013, 2017, 2019b, Poropat et al. 2016, 2021). The mediolateral width to anteroposterior width ratio at midshaft in Garumbatitan morellensis is 2.7–3.1. Following the definition proposed by Mannion et al. (2013: C65, based on: Wilson 2002): ‘Femur shaft eccentricity, mediolateral width to anteroposterior width ratio at midshaft: less than 1.85 (0); 1.85 or greater (1)’; Garumbatitan morellensis shares the apomorphic condition, common in members of Titanosauriformes (Mannion et al. 2013, 2017, 2019b, Poropat et al. 2016, 2021). However, some non-titanosaurian titanosauriforms seem to retain the plesiomorphic condition, with less eccentric femoral shafts, such as, Huabeisaurus allocotus, Tastavinsaurus sanzi, Cedarosaurus weiskopfae, Chubutisaurus insignis, Gobititan shenzhouensis, and Euhelopus zdanskyi. The plesiomorphic condition is also present in a few titanosaurs (Mannion et al. 2013, 2019b). For example, based on the values for femoral shaft eccentricity obtained by Mannion et al. (2013), we conclude that Garumbatitan morellensis exhibits an extreme development of the femoral shaft eccentricity, being higher than in many non-titanosaurian titanosauriforms that share the apomorphic condition [Brachiosaurus altithorax (2.1); Giraffatitan brancai (2.2); Fukuititan nipponensis (1.9); ‘Paluxysaurus jonesi’ (1.9)]. The values shown by Garumbatitan morellensis are closer to the those of some deeply nested titanosauriformes, including titanosaurs, such as Ligabuesaurus leanzai (2.4), Phuwiangosaurus sirindhornae (2.6), Opisthocoelicaudia skarzynskii (2.3) (Mannion et al. 2013); Lirainosaurus astibiae (1.6–2.6, Díez Díaz et al. 2013b); and Ampelosaurus atacis (1.9–5, Vila et al. 2012). The extreme development of the lateral bulge and the eccentricity of shaft that characterizes the femora of Garumbatitan morellensis may have been enhanced by taphonomic deformation. However, the pronounced medial deflection of the proximolateral edge of the femur, and the fact that the most eccentric femoral shaft corresponds to femora that present less evidence of anteroposterior deformation, supports that the morphology of these femora was close to that which characterized them before burial and fossilization. Eccentricity in Garumbatitan morellensis is variable, being less pronounced in the smallest individual (2.7–3.0) than in the largest specimen (3.1), suggesting that this variability may be related to ontogeny.
The presence of a pronounced lateral bulge and eccentric femoral shafts has been considered as ‘wide-gauge’ features (Wilson and Carrano 1999, Carrano 2005). Nowadays, it can be concluded that the development of the lateral bulge and medial deflection of the proximal end in the femur, as well as the presence of high values for femoral shaft eccentricity, are relatively homoplastic and have a wider distribution within Eusauropoda, but are commonly shared by most members of Titanosauriformes (e.g. Wilson and Carrano 1999, Mannion et al. 2013, 2019b, Britt et al. 2017). This suggests that wide-gauge tracks may be produced by different lineages of eusauropods (e.g. Henderson 2006, Santos et al. 2009, Mannion et al. 2013). The femoral morphology of Garumbatitan morellensis suggests that this possible somphospondylan was a putative trackmaker of wide-to-intermediate-gauge tracks. The presence of wide-to-intermediate gauge tracks are known from the Early Cretaceous of Spain and Portugal (e.g. Casanovas et al. 1995, 1997, Moratalla et al. 2003, Santos 2008, Moratalla 2009). Other sauropod taxa from the Hauterivian–Aptian of Spain are Soriatitan golmayensis, Demandasaurus darwini, Tastavinsaurus sanzi, and Europatitan eastwoodi. No complete femur is preserved from the brachiosaurid Soriatitan golmayensis (the proximal end is missing), so the development of the lateral bulge cannot be quantified. The mediolateral width to anteroposterior width ratio at midshaft is 2.1 (Royo-Torres et al. 2017a), a relatively high ratio that is commonly shared by titanosauriforms. According to measurements obtained by Torcida Fernández-Baldor (2012), the rebbachisaurid Demandasaurus darwini is also characterized by a high eccentricity of the femoral shaft (2.1), consistent with the ratios in titanosauriforms. However, this taxon lacks a developed lateral bulge and medial deflection. Tastavinsaurus sanzi lacks a developed lateral bulge and markedly eccentric femoral shaft (Royo-Torres 2009, Royo-Torres et al. 2012). The femoral morphology of the somphospondylan Europatitan eastwoodi is unknown (Torcida Fernández-Baldor et al. 2017). Based on femoral morphology, Garumbatitan morellensis seems to be the best adapted taxon known from the Early Cretaceous of the Iberian Peninsula to produce wide-gauge tracks, and possibly Soriatitan golmayensis. On other the hand, Demandasaurus darwini and Tastavinsaurus sanzi seem to be less well adapted to produce wide-gauge tracks based on its femoral morphology (they also lack a medially bevelled distal end). Titanosaurs are good candidates for these wide-gauge tracks from the Lower Cretaceous of the Iberian Peninsula, as this group of sauropods was present in Europe during the Early Cretaceous (e.g. D’Emic 2012, Dal Sasso et al. 2016, Mocho et al. 2019b); and some authors noted the presence of titanosaurian affinities in some specimens found in the Barremian–Albian of the Iberian Peninsula, such as an anterior caudal vertebra from Galve (Camarillas Formation, Sánchez-Hernández et al. 2007) and a femur from Cinctorres (Arcillas de Morella Fm., Santos-Cubedo et al. 2010). However, they were recently referred as Eusauropoda indet. and Titanosauriformes indet., respectively (Mocho et al. 2017b).
Linea intermuscularis cranialis
Garumbatitan morellensis also has an interesting feature on the anterior surface of the femur, namely the presence of a well-defined linea intermuscularis cranialis. The linea intermuscularis cranialis, or femorotibialis crest, delimits the insertion site of the femorotibialis internus muscle medially and that for femorotibialis externus muscle laterally (Otero and Vizcaíno 2008, Otero 2010). This feature is considered as plesiomorphic in archosaurs (Hutchinson 2001) and absent in sauropods, but present in deeply nested titanosauriforms (Otero 2010, D’Emic 2012, Poropat et al. 2015) such as Saltasaurus loricatus, Rocasaurus muniozi, Neuquensaurus australis, and Bonatitan reigi (Otero 2010). This feature was proposed as a putative synapomorphy of Saltasaurinae (Otero 2010) and Alamosaurus + ‘Saltasaurini’ (D’Emic 2012). The study of some titanosaurian specimens from the Upper Cretaceous of Spain revealed the presence of this feature in taxa such as Lohuecotitan pandafilandi (personal observation; Díez Díaz et al. 2016, PM 2020) and Lirainosaurus astibiae (personal observation, P.M. 2016). This crest is also present in Alamosaurus sanjuanensis (D’Emic 2012, Mannion et al. 2013, 2019b), Diamantinasaurus matildae (Poropat et al. 2015), Dreadnoughtus schrani [see supplementary material of Lacovara et al. (2014); Ullmann and Lacovara (2016) consider that its presence is not clear], Uberabatitan ribeiroi (Silva Junior et al. 2019), Rinconsaurus caudamirus (Pérez Moreno et al. 2023), and a specimen from Bellevue locality (Campagne-sur-Aude, France; Upper Cretaceous) (Vila et al. 2012). All these occurrences suggest that it may be a synapomorphy of a more inclusive group within Titanosauria.
The presence of a linea intermuscularis cranialis in Garumbatitan morellensis corresponds to one of the first occurrences of this feature in a non-titanosaurian sauropod; besides this ridge is much shallower in the Sant Antoni de la Vespa sauropod than in saltasaurines (personal observation, P.M. 2019). A similar ridge is also present in the femur of Brontosaurus excelsus from the Upper Jurassic of the Morrison Formation (Ostrom and McIntosh 1966; personal observation, P.M. 2017). The presence of this feature in these taxa should be considered a local autapomorphy. The linea intermuscularis cranialis is absent in the remaining non-lithostrotian titanosauriforms, including brachiosaurids [e.g. Giraffatitan brancai, Vouvria dampariensis, and Soriatitan golmayensis (personal observation, P.M. 2014, 2015)]. No other Early Cretaceous sauropod taxa from the Iberian Peninsula share this feature with Garumbatitan morellensis (Pereda Suberbiola et al. 2003, Royo-Torres 2009, Royo-Torres et al. 2012, 2017a, Torcida Fernández-Baldor 2012; personal observation, P.M. 2014, 2015).
Femoral trochanteric shelf
The presence of a trochanteric shelf in the posterior surface of the femoral proximal end was considered a possible synapomorphy of saltasaurines by Otero (2010) and reported in several rebbachisaurids and titanosaurs (Sereno et al. 2007, Mannion et al. 2013). This feature has been identified in several macronarian taxa, such as Aragosaurus ischiaticus (Royo-Torres et al. 2014), as well as the somphospondylans ‘Paluxysaurus jonesi’, Huabeisaurus allocotus, Ruyangosaurus giganteus, Diamantinasaurus matildae, Xianshanosaurus shijiagouensis, Malawisaurus dixeyi, Epachthosaurus sciuttoi, Muyelensaurus pecheni, Mendozasaurus neguyelap, Pitekunsaurus macayai, Rinconsaurus caudamirus, Alamosaurus sanjuanensis, Jainosaurus cf. septentrionalis, Saltasaurus loricatus, Opisthocoelicaudia skarzynskii, Uberabatitan ribeiroi, Lirainosaurus astibiae, and Lohuecotitan pandafilandi, a titanosaur femur from the Molí del Baró-2 locality (Tremp Formation, Spain) (e.g. Wilson et al. 2011b, D’Emic 2012, Vila et al. 2012, D’Emic et al. 2013, Díez Díaz et al. 2013b, 2016, Mannion et al. 2013, 2017, 2019a, b, Poropat et al. 2015, González Riga et al. 2018, Silva Junior et al. 2019, Páramo et al. 2020, 2022, Pérez Moreno et al. 2023; personal observation, P.M. 2014–20). In addition, our personal observations noted the presence of a trochanteric shelf in Tastavinsaurus sanzi and Giraffatitan brancai (Janensch 1961), which is proximodistally short. The trochanteric shelf is absent or rudimentary in non-macronarians sauropods (excluding some rebbachisaurids; Pereda Suberbiola et al. 2003, Sereno et al. 2007; Torcida Fernández-Baldor et al. 2011, Torcida Fernández-Baldor 2012, Mannion et al. 2013, 2019b), in early branching macronarians such as Lourinhasaurus alenquerensis (Mocho et al. 2014) and Camarasaurus (Osborn and Mook 1921, Ostrom and McIntosh 1966; personal observation, P.M. 2017–18), and in some titanosaurs such as Patagotitan mayorum (Otero et al. 2020). Within Titanosauria, the development of the trochanteric shelf has an important morphological variability, especially related with its proximodistal length (Vila et al. 2012, Páramo et al. 2020, 2022). Future work will be needed to address the characterization of the proximodistal length and transverse width of the trochanteric shelf, as this may be a feature with taxonomic significance for titanosaurs.
What Garumbatitan tells on the evolution of the tarsus and pes?
The Sant Antoni de la Vespa fossil site has yielded at least two nearly complete pedes, one of which is described in detail in this work. Complete or partially complete pedes are relatively rare in the fossil record, and only preserved in some taxa within Titanosauriformes: Cedarosaurus weiskopfae (FMNH PR 977, D’Emic 2013); Tangvayosaurus hoffeti (Allain et al. 1999, Jannel et al. 2022); Gobititan shenzhouensis (You et al. 2003); Sonorasaurus thompsoni (D’Emic et al. 2016); two partially complete pedes of Tastavinsaurus sanzi (Royo-Torres 2009, Royo-Torres et al. 2012); Diamantinasaurus matildae (Poropat et al. 2023); Epachthosaurus sciuttoi (Martínez et al. 2004); Opisthocoelicaudia skarzynskii (Borsuk-Białynicka 1977); Notocolossus gonzalezparejasi (UNCUYO-LD-302, González Riga et al. 2016); Mendozasaurus neguyelap (IANIGLA-PV 077-78, González Riga et al. 2018); Tapuiasaurus macedoi (MZSP-PV 807, Zaher et al. 2011); a pes referred to Antarctosaurus wichmannianus (Huene 1929); a titanosaur specimen from Agua del Padrillo (Argentina) (UNCUYO-LD-313, González Riga et al. 2015), an unnamed taxon from La Invernada (Argentina) (MUCPv-1533, González Riga et al. 2008), and material referred to Alamosaurus sanjuanensis (NMMNH P-49967, D’Emic et al. 2011) (see Fig. 26).
Schematic draws of eusauropod pedes in dorsal view (not to scale): A, Shunosaurus lii (Zhang 1988); B, Omeisaurus tianfuensis (He et al. 1988); C, Brontosaurus excelsus (CM 89, Hatcher 1901); D, Camarasaurus grandis (McIntosh et al. 1996b); E, Cedarosaurus weiskopfae (FMNH PR 977, D’Emic 2013); F, Tastavinsaurus sanzi (MPZ 99/9, Royo-Torres 2009); G, Garumbatitan morellensis (holotype specimen); H, Garumbatitan morellensis (paratype specimen); I, Gobititan shenzhouensis (You et al. 2003); J, Epachthosaurus sciuttoi (Martínez et al. 2004); K, titanosaurian specimen from Agua del Padrillo (Argentina, UNCUYO-LD-313, González Riga et al. 2015); L, titanosaurian specimen from La Invernada (Argentina, MUCPv-1533, González Riga et al. 2008); M, Notocolossus gonzalezparejasi (UNCUYO-LD 302, González Riga et al. 2016); N, Opisthocoelicaudia skarzynskii (Borsuk-Białynicka 1977); O, pes referred to Alamosaurus sanjuanensis (NMMNH P-49967, D’Emic et al. 2011). *Reversed.
One of the most important features of Garumbatitan morellensis is the absence of the calcaneum, which is shared with some titanosaurs, such as Opisthocoelicaudia skarzynskii (Borsuk-Białynicka 1977), Epachthosaurus sciuttoi (Martínez et al. 2004), and possibly Notocolossus gonzalezparejasi (González Riga et al. 2016). The calcaneum is present in the non-somphospondylan titanosauriform Giraffatitan brancai (Janensch 1961) and in the somphospondylans Tastavinsaurus sanzi (Royo-Torres et al. 2012), Dongbeititan dongi (Wang et al. 2007), Gobititan shenzhouensis (You et al. 2003), and the euhelopodids Erketu ellisoni (Ksepka and Norell 2006) and Euhelopus zdanskyi (Wilson and Upchurch 2009). Considering the topology of the cladograms recovered in this study, which considers Garumbatitan morellensis as an early branched somphospondylan, the absence of the calcaneum represents a local autapomorphy.
The pes of Garumbatitan morellensis is relatively slender (Fig. 26), especially the metatarsals II, III, and IV. In addition, the pes of Garumbatitan morellensis is characterized by the shortness of metatarsal V and metatarsal I (this trend is present in the holotype and the paratype). The metatarsal V is markedly reduced, with a metatarsal V to IV proximodistal width ratio of 0.74 and a proximodistal width of metatarsal V to III ratio of 0.7 (ratios are based on the paratype specimens). These values are comparable to titanosaurs with reduced metatarsal V in relation to the metatarsal III and IV, such as MUCPv-1533, Epachthosaurus sciuttoi, Opisthicoelicaudia skarzynskii, Rapetosaurus krausei (Borsuk-Białynicka 1977, González Riga et al. 2008, Curry Rogers 2009), and possibly Diamantinasaurus matildae [for the proximodistal width of metatarsal V to III ratio, Poropat et al. (2023)]. This short metatarsal V, in relation to metatarsals IV and III, results in a step between these metatarsals when we observe the pes of Garumbatitan morellensis in dorsal view (Fig. 17). This step is not evident in other non-titanosaurian somphospondylans, such as Tastavinsaurus sanzi, Gobititan shenzhouensis, Cedarosaurus weiskopfae, and Sonorasaurus thompsoni (Royo-Torres 2009, D’Emic 2013, D’Emic et al. 2016), and it is considered as a convergence with the titanosaurs Opisthicoelicaudia skarzynskii, Rapetosaurus krausei and MCUPv-1533. In Tastavinsaurus sanzi, the metatarsal V is proximodistal longer than the metatarsal I [for measures see Royo-Torres (2009)], unlike Garumbatitan morellensis.
The transition between the proximodistal widths of the metatarsal I to metatarsal II is also abrupt as in Sonorasaurus thompsoni, Cedarosaurus weiskopfae, Diamantinasaurus matildae, and Opisthocoelicaudia skarzynskii, in contrast to Tastavinsaurus sanzi, Gobititan shenzhouensis, Tangvayosaurus hoffeti, Epachthosaurus sciuttoi, and Notocolossus gonzalezparejasi (see Fig. 26). This mutual retraction of the metatarsals I and V, is considered an autapomorphy of Garumbatitan morellensis, which is not present in other sauropods.
The phalangeal formula for Garumbatitan morellensis (based on the pes of the two specimens available) is interpreted as 2-3-3-3-0(?), different from the phalangeal formula proposed by Royo-Torres et al. (2012) for Tastavinsaurus sanzi based on a referred specimen (CT-19) and holotype (MPZ99/9): 2-3-4-2-1. Comparison with the data of González Riga et al. (2016: table 3) shows that the pes of Garumbatitan morellensis preserves the same formula of basal macronarians for digits I to III. On the other hand, Garumbatitan morellensis preserves a similar phalangeal formula to Cedarosaurus weiskopfae, which is estimated to be 2-3-3-3-1. Garumbatitan morellensis preserves the plesiomorphic condition in sauropods with more than two phalanges in digit IV, as found in Mamenchisaurus, Omeisaurus, and Shunosaurus. Neosauropods are characterized by the presence of only two phalanges on digit IV (He et al. 1988, Mannion et al. 2013, 2019a, b; González Riga et al. 2016), but some exceptions have been reported: (i) D’Emic (2013) suggests the presence of three phalanges in Cedarosaurus weiskopfae, but Gallup (1989) noted only two; and (ii) a drawing provided by Jannel et al. (2022), suggests the presence of three pedal phalanges for digit IV of Tangvayosaurus hoffeti, but Allain et al. (1999) stated two phalanges in digit IV. In our analyses the presence of three phalanges is a local autapomorphy of Garumbatitan morellensis.
No phalanges were found for digit V, but its complete absence may be an artefact of post-mortem disarticulation and no preservation. The absence of the pedal phalanges in the digit V is shared by titanosaurs (see: Poropat et al. 2023, and references therein) but differs from the condition present in the non-titanosaurian somphospondylans Tastavinsaurus sanzi (Royo-Torres et al. 2012) and Gobititan shenzhouensis (You et al. 2003), and in the brachiosaurid Cedarosaurus weiskopfae (D’Emic 2013). This situation suggests that the loss of this phalanx might occur earlier during the evolution of Somphospondyli (i.e. synapomorphy of a more inclusive group than Titanosauria) or represents an autapomorphy of Garumbatitan morellensis that could result from the reduction of the metatarsal V. The complete pes of Garumbatitan shenzhouensis is important for understanding the evolution of this character. This taxon was found in the Albian of China and tentatively considered a basal titanosaur (You et al. 2003). However, recent phylogenetic approaches suggest that Gobititan shenzhouensis is a non-titanosaurian somphospondylan and a possible member of Euhelopodidae (Mannion et al. 2013, 2017). If this hypothesis is confirmed and based on the two the topologies of cladogram recovered in this study, the loss of the pedal phalanx in the digit V is indeed a local autapomorphy of Garumbatitan morellensis.
The ungual of the digit III in Garumbatitan morellensis has a proximodistal width smaller than those of metatarsal III (i.e. ungual to metatarsal length ratio in the digit III is 0.33). This ratio is higher than 0.4 in Tastavinsaurus sanzi (0.48, Royo-Torres 2009), Cedarosaurus weiskopfae (D’Emic 2013), Alamosaurus sanjuanensis (NMMNH P-49967; D’Emic et al. 2011), Gobititan shenzhouensis (You et al. 2003), Epachthosaurus sciuttoi (Martínez et al. 2004), MCUPv-1533 (González Riga et al. 2008), Opisthocoelicaudia skarzynskii (Borsuk-Białynicka 1977), and Notocolossus gonzalezparejasi (González Riga et al. 2016) that also have higher values for this ratio. A digit III with a reduced ungual is an autapomorphy of Garumbatitan morellensis.
In conclusion, Garumbatitan morellensis is characterized by the absence of calcaneum and by the presence of some important changes in the pes morphology: elongated metatarsals, reduced metatarsals I and V, three phalanges on the three central digits of the pes, and a reduced ungual of the digit III. According to the phylogenetic hypotheses proposed in this work and the tree topologies of other authors (e.g. Mannion et al. 2013, 2017, 2019a, b, Carballido and Sander 2014, Poropat et al. 2015, 2016, 2021, 2023, Upchurch et al. 2015, D’Emic et al. 2016, González Riga et al. 2016, 2018), this morphology seems to be the result of a unique adaptation occurring in this taxon. More derived titanosaurs acquired posteriorly some of these features through convergence, such as loss of the calcaneum, reduction of the metatarsal V, the loss of the pedal phalanx V.1, and reduction of the ungual phalanx of the digit III.
Garumbatitan and Hauterivian–Aptian sauropods of the Iberian Peninsula
The historically earliest discoveries that are referred to ‘colossal reptiles’ or sauropods from the Lower Cretaceous of the Iberian Peninsula were provided by Vilanova y Piera (1872) in Morella (Barremian, Spain) and Sauvage (1897–98) in Boca de Chapim (Barremian–Aptian, Portugal). After these first references, several works were published on the Iberian Early Cretaceous sauropods, especially in the Maestrat and Cameros basins, but also in the Iberian Ranges, Pre-Betic and Lusitanian Basins (e.g. Royo y Gómez 1926, Lapparent and Zbyszewski 1957, Lapparent 1966, Lapparent et al. 1969, Sanz et al. 1982) and, more recently, several new specimens have been found (e.g. Yagüe et al. 2001, Canudo et al. 2002, 2004a, b, 2008, Pereda Suberbiola et al. 2003, Fuentes Vidarte et al. 2005, Ortega et al. 2006, Sánchez-Hernández et al. 2007, Gasulla et al. 2008, 2011, 2012, 2021, Royo-Torres 2009, Santos-Cubedo et al. 2010, Torcida Fernández-Baldor et al. 2011, 2017, Royo-Torres et al. 2012, 2017a, Mocho et al. 2016, 2022). Some of this material is very incomplete and provides little information on the Iberian sauropod diversity from the Hauterevian to the Aptian. Herein, we will discuss the most complete sauropod fossil record, mainly from the Valanginian to the Aptian, including the most taxonomically informative specimens, for which we provide a direct comparison (if possible) with Garumbatitan morellensis.
An important fossil record has been found in the Arcillas de Morella Formation, which was deposited during the late Barremian in the Maestrat Basin (sensuBover-Arnal et al. 2016). All authors agree with the presence of titanosauriforms in this formation, with some specimens thought to be related to Brachiosauridae (Sanz et al. 1982, Yagüe et al. 2001, Ortega et al. 2006, Gasulla et al. 2008) and others to Titanosauria (Santos-Cubedo et al. 2010). Mocho et al. (2017b) provided a detailed study for several specimens collected in Arcillas de Morella Fm. around the Morella locality (Castelló) and belonging to the historical collection of the Museo Nacional de Ciencias Naturales in Madrid (Spain). In that study, a higher sauropod diversity was hypothesized in this formation, with three possible different titanosauriforms (recognized based on three different types of humeri and other material): a possible ‘laurasiform’ and two somphospondylans. Mocho et al. (2016, 2017b) pointed out the possibility that the Sant Antoni de la Vespa titanosauriform (=Garumbatitan morellensis) represents a member of ‘Laurasiformes’. The present phylogenetic approach suggests the inclusion of Garumbatitan morellensis within Somphospondyli, as an early branching form, more deeply nested than Dongbeititan dongi and more basally branching than Tastavinsaurus sanzi (Figs 24, 27). As a result of that, the monophyly of ‘Laurasiformes’ is herein not supported.
Time-calibrated phylogenetic tree, showing geographic distribution of camarasaurids and non-titanosaurian titanosauriforms. Topology corresponds to a consensus strictus from trees obtained with implied weight analysis (see Analyses II) and resolved after the a posteriori deletion of Abydosaurus, Nopcsaspondylus, Padillasaurus, and Sarmientosaurus in TNT. Global palaeogeographic reconstructions from Fossilworks (http://fossilworks.org/) showing the distribution of Late Jurassic (150 Ma) and Early Cretaceous (126 Ma) of brachiosaurids in (pink) and non-titanosaurian somphospondylans (light brown). An interrogation mark is placed on two occurrence points for two Late Jurassic titanosauriforms, Oceanotitan and Autrodolodocus, which present an uncertain phylogenetic position, but were referred as putative somphospodylans by some authors.
Another informative sauropod specimen recovered in this formation is the specimen collected at El Canteret (Sanz et al. 1982). This specimen was aligned with Brachiosauridae by Sanz et al. (1982) and Yagüe et al. (2001). Recentely, Mocho et al. (2017b) noted that the humerus of the sauropod from El Canteret shares some affinities with Somphospondyli. This specimen presents some distinct features from the Sant Antoni de la Vespa sauropod: (i) the femur has a medially projected femoral head, unlike Tastavinsaurus sanzi (Royo-Torres 2009) and Garumbatitan morellensis; (ii) the fibula has a higher robustness index (RI = 0.21) than the fibulae of Garumbatitan morellensis (RI = 0.17) and Tastavinsaurus sanzi (RI = 0.16–0.19; Canudo et al. 2008, Royo-Torres 2009, Royo-Torres et al. 2012). However, this specimen still needs an accurate systematic review. Mocho et al. (2017b) found no support for the presence of diplodocoids, brachiosaurids, and titanosaurs in the Arcillas de Morella Fm., as previously proposed by other authors (e.g. Sanz et al. 1982, Pereda Suberbiola and Ruiz-Omeñaca 2005, Santos-Cubedo et al. 2010). Mocho et al. (2017b) also noted the presence of material with ‘Laurasiform’ affinity, in particular, MNCN 59691 and MNCN 59697, which share the anterodorsal projection of the anterior and middle neural spines with Tastavinsaurus sanzi (Royo-Torres 2009), Cedarosaurus weiskopfae (Tidwell et al. 1999), and Venenosaurus dicrocei (Tidwell et al. 2001), which was interpreted as absent in Garumbatitan morellensis. A femoral proximal end (MNCN 9363) was collected in the same area and described by Mocho et al. (2017b). This femur shows a pronounced development of the lateral bulge, which enables it to be regarded as cf. Garumbatitan morellensis. The systematics of the sauropod material collected in the Arcillas de Morella Fm. is complex, and its incompleteness precludes direct comparison of all the identified specimens. For example, three different morphologies of humerus were identified, and the presence of three different titanosauriforms, including two somphospondylans, was hypothesized (Mocho et al. 2017b). No humerus is preserved from the Sant Antoni de la Vespa sauropods, and it cannot be directly compared with the previously identified ones. However, the sauropod from El Canteret, which also preserves hindlimb elements, can be confidently differentiated from Garumbatitan morellensis. For the moment, we agree with Mocho et al. (2017b) on the composition of the sauropod fauna, after the detailed description of the Sant Antoni de la Vespa sauropod, named here as Garumbatitan morellensis.Gasulla et al. (2021) noted the discovery of a new sauropod dinosaur in Mas de Palau (MP-1a), represented by a set of associated bones that includes cervical and dorsal vertebra, dorsal and cervical ribs, pelvic, forelimb and hindlimb bones (including humerus, ischium, and tibia). A preliminary assessment of this specimen allowed us to identify a combination of features shared with Titanosauriformes (e.g. presence of camellate bone tissue in the presacral vertebrae). Several of the elements preserved in MP-1a can be directly compared with those of the holotype specimens from the currently established titanosauriform taxa of the Early Cretaceous of Spain (i.e. Tastavinsaurus sanzi, Soriatitan golmayensis, Europatitan eastwoodi, and Garumbatitan morellensis) and El Canteret and Mas de Eroles sauropods. MP-1a becomes one of the key specimens from the Lower Cretaceous of the Iberian Peninsula, for understanding the phylogenetic relationships of those taxa. In conclusion, the sauropod fauna recovered from this formation possibly comprises three titanosauriforms, including the somphospondylans Garumbatitan morellensis and the El Canteret sauropod.
The Arcillas de Morella Formation directly underlies the Xert Fm., which in turn is overlain by the Forcall Fm. (Villanueva-Amadoz et al. 2014, Bover-Arnal et al. 2016). Important sauropod specimens have also been reported from these formations, in particular the somphospondylan Tastavinsaurus sanzi, represented by two different specimens: the holotype coming from the de Peñarroya de Tastavins (Canudo et al. 2008, Royo-Torres 2009) and a second specimen found at El Castellar (Teruel; Forcall Fm., lower Aptian; Royo-Torres et al. 2012). Royo-Torres et al. (2012) regarded it as a basal somphospondylan [following the definition of Wilson and Sereno (1998)], belonging to a new clade defined as Laurasiformes (Royo-Torres 2009, Royo-Torres et al. 2012). Different phylogenetic positions have been proposed for Tastavinsaurus sanzi and Laurasiformes (e.g. Royo-Torres 2009, Carballido et al. 2011b, 2015, 2020, D’Emic 2012, Royo-Torres et al. 2012, 2017a, Mannion et al. 2013, 2017, 2019a, b, Carballido and Sander 2014, Poropat et al. 2015, 2016, 2021, 2023, Upchurch et al. 2015, D’Emic et al. 2016, Torcida Fernández-Baldor et al. 2017, Mocho et al. 2019b). Laurasiformes were defined by Royo-Torres (2009) as a branch-based clade including all macronarian taxa more closely related to Tastavinsaurus sanzi than to Saltasaurus loricatus. This clade includes the Laurasian titanosauriforms Venenosaurus dicrocei, Tastavinsaurus sanzi, and Cedarosaurus weiskopfae, and is supported by nine synapomorphies (see: Royo-Torres et al. 2012). These authors also suggested the inclusion of Lourinhasaurus alenquerensis and Aragosaurus ischiaticus within this clade, but the recent full description of these taxa reveals that they represent non-titanosauriform macronarians (Mocho et al. 2014, Royo-Torres et al. 2014). ‘Laurasiformes’ has been recovered in some phylogenetic proposals as a brachiosaurid or somphospondylan subclade (Royo-Torres 2009, Royo-Torres et al. 2012, 2014, Mocho et al. 2014). More recently, Royo-Torres et al. (2017a) recovered the clade Laurasiformes within Brachiosauridae, and composed of Tastavinsaurus sanzi, Cedarosaurus weiskopfae, Venenosaurus dicrocei, Abydosaurus mcintoshi, and Soriatitan golmayensis. However, other recent phylogenetic analyses do not support the monophyly of this clade or propose a different configuration for Tastavinsaurus sanzi, recovering it as an early branching macronarian, a brachiosaurid, or, more frequently, a non-titanosaurian somphospondylan with possible affinities with other Laurasian forms (e.g. Europatitan eastwoodi, Sauroposeidon proteles, and ‘Cloverly titanosauriform’) or even Gondwanan forms such as Angolatitan adamastor, Ligabuesaurus leanzai, Chubutisaurus insignis, or Wintonotitan wattsi (e.g. D’Emic 2012, Mannion et al. 2013, 2017, 2019a, b, Carballido and Sanders 2014, Poropat et al. 2015, 2016, 2021, D’Emic et al. 2016, Torcida Fernández-Baldor et al. 2017, Carballido et al. 2020). In none of the analyses performed was Garumbatitan morellensis recovered as a sister-taxon of Tastavinsaurus sanzi or within ‘Laurasiformes’ [following the proposal of Royo-Torres et al. (2012)]. These two taxa were both found in the Maestrat Basin, but in different geological formations. Both are similar in age, being the Arcillas de Morella Fm. dated to the late Barremian, and Xert and Forcal formations to the latest Barremian and early Aptian, respectively (Bover-Arnal et al. 2016). Several morphological features allowed the distinction of the sauropod from Sant Antoni de la Vespa from Tastavinsaurus sanzi.
The revised diagnosis of Tastavinsaurus includes 19 autapomorphies (Royo-Torres et al. 2012; 18 local autapomorphies are recovered in our Analysis II). In the available material of Garumbatitan morellensis, it is possible to evaluate the presence or absence of nine of the diagnostic features proposed by Royo-Torres et al. (2012). Seven of them are absent in Garumbatitan morellensis: (i) acute angle of the cranioventral corner of the distal extremity of the pubis; (ii) short tibia (55% of the length of the femur); (iii) crescentic proximal end of the fibula; (iv) quadrangular distal end of the fibula; (v) robustness index value of 1.0 (0.97) for the metatarsal I; (vi) area of the proximal end of metatarsal II is 30% larger than that of metatarsal I; and (vii) a greatly expanded proximal end of metatarsal V (three times the width of the distal end). The autapomorphy (14) of Tastavinsaurus sanzi proposed by Royo-Torres et al. (2012), i.e. the presence of a well-developed anterior trochanter forming a crest that projects anteromedially (=anteromedial crest of the fibular proximal end), is present in Garumbatitan morellensis. However, this feature is also present many other somphospondylans, such as ‘Paluxysaurus jonesi’ (Rose 2007), Gobititan shenzhouensis (You et al. 2003), Phuwiangosaurus sirindhornae (Martin et al. 1999, Mannion et al. 2013), Malawisaurus gomani (Gomani 2005), Erketu ellisoni (personal observation, P.M. 2018), and Euhelopus zdanskyi (Wilson and Upchurch 2009; personal observation, P.M. 2018). The autapomorphy (18) of Tastavinsaurus of Royo-Torres et al. (2012) is the presence of pronounced proximolateral projections in metatarsal I and II. These projections are also present in Garumbatitan morellensis, Cedarosaurus weiskopfae (D’Emic 2013), Venenosaurus dicrocei (Tidwell et al. 2001), Sonorasaurus thompsoni (D’Emic et al. 2016), and the Late Jurassic macronarians Giraffatitan brancai (Janensch 1961) and Camarasaurus (e.g. McIntosh et al. 1996b; personal observation, P.M. 2017–18).
Numerous differences between Tastavinsaurus sanzi and Garumbatitan morellensis can be identified from the anatomical comparison of both specimens and from the obtained cladogram topologies. The tail morphology is similar, especially in the anterior section of the tail; however, some important anatomical differences can be noticed in Garumbatitan morellensis, such as: (i) subvertical-to-posterodorsally oriented anterior caudal neural spines; (ii) small and shallow vascular foramina piercing the lateral and/or ventral surfaces of the anterior and middle caudal centra; (iii) the distance that prezygapophyses extend beyond the anterior margin of the centrum is less than 20% of centrum length (excluding ball); (iv) amphicoelous middle and posterior caudal centra; and (v) middle-posterior caudal neural spine has a straight-to-slightly convex dorsal edge in lateral view, posteriorly oriented. The femur of Garumbatitan morellensis is markedly distinct from the femora of Tastavinsaurus sanzi. Some of the most important features distinguishing Garumbatitan morellensis are: (i) a step below the femoral head; (ii) linea intermuscularis cranialis (local autapomorphy of Garumbatitan morellensis); (iii) medial deflection of the proximal third; (iv) much more developed lateral bulge (Fig. 12, autapomorphy of Garumbatitan morellensis); (v) femoral shaft transverse width/anteriorposterior width ratio is higher in Garumbatitan morellensis (2.7–3.1) than in the holotype of Tastavinsaurus sanzi (1.5–1.6, Royo-Torres 2009) and in the referred specimen found at El Castellar (2.1, Royo-Torres et al. 2012); (vi) ratio of the mediolateral width of the tibial condyle to the width of the fibular condyle is less than 0.8; (vii) length of the tibia is 64% of the femur length (55% in Tastavinsaurus sanzi). The fibula of Garumbatitan morellensis is stout and straight in lateral view, unlike the marked sigmoidal fibula of Tastavinsaurus sanzi (not so pronounced in the referred specimen from El Castellar). Also, in lateral view the proximal end is markedly asymmetrical in Tastavinsaurus sanzi (Canudo et al. 2008, Royo-Torres 2009, Royo-Torres et al. 2012), unlike the condition of Garumbatitan morellensis, which resembles the more symmetrical end present in basal titanosauriforms, such as the Late Jurassic brachiosaurids Giraffatitan brancai (Janensch 1961), Lusotitan atalaiensis (Mocho et al. 2017a), and Vouivria dampariensis (Mannion et al. 2017). An important difference is the presence of a semicircular distal surface in Garumbatitan morellensis, a common condition in sauropods and different from the subquadrangular distal end of Tastavinsaurus sanzi, which was considered diagnostic for this taxon by Royo-Torres et al. (2012). The tibia of Garumbatitan morellensis is more slender than those of Tastavinsaurus sanzi and the cnemial crest in the former taxon is round and laterally projected, unlike the tibia of the latter taxon, which has a subtriangular and anterolaterally projected cnemial crest. Comparison between the astragali of Garumbatitan. morellensis and Tastavinsaurus sanzi is limited because of the poor preservation of these elements in the latter taxon. The mediolateral width to maximum proximodistal height ratio is greater in Garumbatitan morellensis (1.84–1.87) than Tastavinsaurus sanzi (1.32–1.79, Royo-Torres et al. 2012). Tastavinsaurus sanzi is characterized by the presence of calcaneum, which is absent in the recovered pedes from the Sant Antoni de la Vespa fossil site. The pes of Garumbatitan morellensis also differs from Tastavinsaurus sanzi: (i) there is a tuberosity in the medial surface of the metatarsal I, which is absent in Tastavinsaurus sanzi; (ii) the metatarsal I is longer than the metatarsal V, unlike Tastavinsaurus sanzi; (iii) the metatarsal III length to tibia length ratio is 0.24–0.27 compared to the plesiomorphic condition present in Tastavinsaurus, >0.30; (iv) the shortness of metatarsals I and V relative to the metatarsals II, III, and IV (autapomorphy of Garumbatitan morellensis); (v) the proximal end to the distal end maximum mediolateral width ratio is 1.67 in Garumbatitan morellensis and 2.32 in Tastavinsaurus sanzi; (vi) the lateral edge of phalanges I.1 and II.1 are proximodistally constricted. The proximal section of the pes also shows important differences, mainly in the proximal outline of metatarsals III and IV (see Figs 18, 19).
Important sauropod occurrences have also been reported from the upper Barremian–lower Aptian Castrillo de la Reina Fm., deposited in the Cameros Basin, west of the Maestrat Basin. This sauropod fauna comprises the rebbachisaurid Demandasaurus darwini (Castrillo de la Reina Fm., upper Barremian–lower Aptian) and the somphospondylan Europatitan eastwoodi (Castrillo de la Reina Fm. in Salas de los Infantes, Burgos). Demandasaurus darwini is a rebbachisaurid member of Nigersaurinae (Torcida Fernández-Baldor et al. 2011, Whitlock 2011, Carballido et al. 2012), or Rebbachisaurinae by Wilson and Allain (2015), and can be easily differentiated from titanosauriforms, especially considering the morphology of its axial skeleton (e.g. anterior caudal neural spines with a petal-shaped transverse section; the orientation of the transverse processes in the caudal vertebrae). Europatitan eastwoodi is titanosauriform with somphospondylan affinities and distinct from Tastavinsaurus sanzi (Torcida Fernández-Baldor et al. 2017). This titanosauriform comprises a tooth, several axial elements, and some appendicular bones. However, we can make a direct comparison with Garumbatitan morellensis only for some elements (dorsal ribs, anterior caudal vertebra, chevrons, and pubis). In addition, in our analysis, both taxa occupy different positions in the cladogram, with Europatitan eastwoodi being recovered as a more deeply nested somphospondylan within a subclade including Baotianmansaurus henanensis and Huanghetitan, which are sister-taxa of Angolatitan adamastor + Chubutisaurus insignis. The detailed comparison of the Sant Antoni de la Vespa sauropod with the available data allows us to recognize several features that are not present in Europatitan eastwoodi: (i) anterior caudal vertebrae with flat posterior articular surface; (ii) larger aEI for the anterior caudal vertebrae (aEI for Garumbatitan morelllensis is 0.80–1.12 and for Europatitan eastwoodi is 0.56–0.68); (iii) presence of pneumatic and shallow fossa in the lateral surface of the anterior caudal centra; (iv) caudal ribs do not extend beyond the margin of the posterior articular surface; and (v) no step is present in the anterior crest of the chevrons. The future preparation of the preserved section of cervical and dorsal vertebrae of Garumbatitan morellensis will be important to provide more data on the relationship between these taxa.
Soriatitan golmayensis is a brachiosaurid titanosauriform found in the sedimentary rocks of the Golmayo Formation, upper Hauterivian–lower Barremian (Cameros Basin), on the Zorralbo site (Soria) (Fuentes Vidarte et al. 2005, Royo-Torres et al. 2017a). The Zorralbo sauropod is not temporally correlative with the already established titanosauriforms from the Lower Cretaceous strata of Spain and is clearly different from Tastavinsaurus sanzi and Garumbatitan morellensis, especially in the morphology of the femur (extremely slender femur and distal end only slightly transversely expanded). Also, this brachiosaurid differs from the Sant Antoni de la Vespa sauropod by: (i) the absence of small and shallow vascular foramina in the anterior and middle caudal vertebrae; (ii) the caudal ribs extended beyond the posterior articular surface of the caudal centrum; (iii) the lesser eccentricity of the femoral shaft (2.18 in Soriatitan golmayensis and 2.7–3.1 in Garumbatitan morellensis); and (v) the distal end of the femur is laterally bevelled. In addition, none of the diagnostic features proposed for this brachiosaurid have been found in the material of Garumbatitan morellensis.
The sauropod diversity in the Iberian Peninsula during the Hauterivian–Aptian is richer than previously thought, and the sauropod fauna is composed of the rebbachisaurid Demandasaurus darwini, the brachiosaurid Soriatitan golmayensis, and at least three somphospondylans, Tastavinsaurus sanzi, Garumbatitan morellensis, and Europatitan eastwoodi. Some specimens yet undescribed, such as the sauropods from El Canteret, Mas de Eroles (Sanz et al. 1982, Yagüe et al. 2001), and Mas de Palau (Gasulla et al. 2021), also share some affinities with Somphospondyli. The rebbachisaurid Demandasaurus darwini was recovered as a closely related form with other Aptian–Cenomanian North African rebbachisaurids (e.g. Torcida Fernández-Baldor et al. 2011, Wilson and Allain 2015, Canudo et al. 2018, Xu et al. 2018). The titanosauriform Soriatitan golmayensis has been recovered within the Brachiosauridae clade, closely related to the Early Cretaceous forms of North America (Royo-Torres et al. 2017a,Mannion et al. 2019b; this study). The phylogenetic relationship of the Iberian somphospondylans to other Early Cretaceous somphospondylans in the European and non-European regions is still controversial. Different phylogenetic positions were proposed for Tastavinsaurus sanzi and Europatitan eastwoodi. The former taxon was recently recovered as a somphospondylan titanosauriform, in close relationship with some of the Early Cretaceous forms from East Asia and Gondwana (e.g. Mannion et al. 2013, 2017, 2019a, b, Upchurch et al. 2015, Poropat et al. 2016, 2021, Mocho et al. 2019a, Royo-Torres et al. 2021). However, some authors have recovered Tastavinsaurus sanzi as a non-titanosauriform macronarian (e.g. Carballido and Sander 2014, Carballido et al. 2020) or as a member of Brachiosauridae closely related to Early Cretaceous taxa from North American (Royo-Torres et al. 2012, 2014, 2017a, Mocho et al. 2014). The phylogenetic position of Europatitan eastwoodi is still uncertain, but some phylogenetic approaches recovered it within a clade with East Asian and Gondwanan forms in accordance with the results obtained in this work. In our implied weight analysis, Garumbatitan morellensis is placed as a basally branching somphospondylan with the Laurasian titanosauriforms Tastavinsaurus sanzi and the Early Cretaceous Dongbeititan dongi from East Asia. Oceanotitan dantasi, which was considered an early branching somphospondylan in one of the analyses performed by Mocho et al. (2019a), is recovered here as an early branching member of Macronaria.
These results reveal a complex phylogenetic mosaic for the sauropod fauna of the Iberian Peninsula during the Early Cretaceous, composed of forms with Laurasian affinities, mainly titanosauriforms (Soriatitan golmayensis, Garumbatitan morellensis, and possibly Tastavinsaurus sanzi and Europatitan eastwoodi) and Gondwanan affinities (the rebbachisaurid Demandasaurus darwini) (Fig. 27). However, the systematics of many sauropod occurrences from the Lower Cretaceous of Europe is uncertain, and some of them are represented by significantly incomplete specimens. A diverse faunal composition has been proposed, including turiasaurs, rebbachisaurids, brachiosaurids, and non-titanosaurian and titanosaurian somphospondylans (Poropat et al. 2022, and references therein). Faunal contacts during the Early Cretaceous have been suggested between the sauropod fauna of Europe, especially the Iberian Peninsula and that of North America, East Asia, and Africa. Rebbachisaurids possibly dispersed between Africa and Europe via the Apulian route during the Early Cretaceous, which is supported by the affinities of Demandasaurus and the Early Cretaceous rebbachisaurids of Gondwana (e.g. Canudo et al. 2009, 2018, Torcida Fernández-Baldor et al. 2011, 2017, Wilson and Allain 2015), and possibly some somphospondylan lineages as well, which may explain the possible relationship of some European and East Asian forms to Gondwanan ones. Brachiosaurids achieved a wide palaeobiogeographic distribution in the Late Jurassic, being already present in Europe, Africa, and North America. However, a possible dispersal across Europe and North America, with the establishment of a land connection between North America and Fennoscandia across the Barents Shelf in the earliest Barremian (~131–129 Mya) (Brikiatis 2016) have been suggested (Royo-Torres et al. 2017a). Finally, a dispersal route between Europe and East Asia during the Early Cretaceous [a Late Jurassic dispersal event is not ruled out, Liao et al. (2021)] seems to be supported by the close relationship of some Iberian forms with some non-titanosaurian somphospondylans of East Asia recovered in this work and some previous phylogenetic approaches (e.g. Poropat et al. 2016, Royo-Torres et al. 2017a, Mocho et al. 2019a), such as Dongbeititan dongi, Baotianmansaurus henanensis, and Huanghetitan (e.g. but also with the Gondwanan forms Angolatitan adamastor, Chubutisaurus insignis; and in a few of the recovered topologies with Wintonotitan wattsi, Ligabuesaurus leanzai, and the North American Sauroposeidon proteles and the ‘Cloverly titanosauriform’). The presence of a tooth like those of Euhelopus was also considered evidence for a close link between some sauropod forms of Europe and Asia (Canudo et al. 2002). This close relationship between Iberian and East Asian taxa has also been revealed for other taxonomic groups, suggesting a faunal exchange between Europe and Asia domains during the Early Cretaceous, possibly in the Barremian (e.g. Cuenca-Bescós and Canudo 2003, Sweetman 2006, 2008, Vullo et al. 2012, Allain et al. 2014, Martin et al. 2015, Buscalioni and Poyato-Ariza 2016). The complex palaeobiogeographic distribution of the somphospondylan titanosauriforms, reflected in the different phylogenetic approaches, could be a consequence of the cosmopolitan distribution that somphospondylans were able to achieve in the Early Cretaceous (or possibly in the Late Jurassic). However, most early branching members of Somphospondyli are represented by incomplete specimens, which adds some degree of uncertainty on the obtained phylogenetic approaches. The future discovery and reassessment of the already established non-titanosaurian somphospondylan taxa, in particular, the Early Cretaceous representatives of Asia, will be key to understand the early evolution of this clade.
New discoveries will be necessary to understand the evolutionary history of the sauropod faunas in the Barremian–Aptian sequence in the Maestrat Basin. The relationship between the sauropod faunas from different regions of the Iberian Peninsula needs to be clarified, especially between the Maestrat and Cameros Basins. Future systematic studies on the remaining material from Sant Antoni de la Vespa (Arcillas de Morella Fm., Maestrat Basin) and the specimen from El Cateret (Arcillas de Morella Fm., Maestrat Basin), as well as some new occurrences in the Arcillas de Morella Formation (Gasulla et al. 2021), will help to better understand these relationships. Additionally, a detailed systematic study of several sauropod specimens from the Lower Cretaceous of England and France (e.g. Upchurch et al. 2011, Allain et al. 2022) will be important to better understand this palaeobiogeographic scenario in Europe.
CONCLUSIONS
Remains of three sauropod individuals belonging to a unique taxon were found at Sant Antoni de la Vespa (Castelló, Morella), in the sediments of the Arcillas de Morella Formation (upper Barremian). These specimens have a set of exclusive characters (19 features) allowing the establishment of a new taxon, Garumbatitan morellensis. Garumbatitan morellensis is recovered as an early branching somphospondylan titanosauriform in the implied weights analyses, with a well-resolved strict consensus. These analyses also placed the somphospondylan titanosauriforms Tastavinsaurus sanzi and Dongbeititan dongi as early branching members of the clade. Our comparative analysis also identified several characters in Garumbatitan morellensis that are commonly shared with somphospondylan titanosauriforms.
Our results allow us to discuss the phylogenetic distribution of some titanosauriform and somphospondylan novelties, especially some related to the morphology of the femur, tarsus, and pes. The femur of Garumbatitan morellensis is characterized by a relatively apomorphic morphology exhibiting several synapomorphies of Somphospondyli or even more exclusive clades (e.g. Saltasaurinae): (i) pronounced lateral bulge; (ii) a strong eccentricity of the shaft; (iii) the presence of a linea intermuscularis cranialis at the anterior surface of the shaft; and (iv) the presence of a trochanteric shelf in the posterior face of the proximal end. These may indicate an earlier appearance of these features in the phylogeny of somphospondylans, or a convergence of Garumbatitan morellensis with more deeply nested somphospondylans, such as the members of Titanosauria. The morphology of the femur suggests that Garumbatitan morellensis is a putative producer of wide-gauge tracks, which are well-represented by some fossil track sites in the Lower Cretaceous of Spain and Portugal. The tarsus and pes of Garumbatitan morellensis are characterized by a unique combination of features, highlighting the absence of the calcaneum, the relative slenderness of the metatarsals II, III, and IV with the retraction of the metatarsals I and V, the presence of three pedal phalanges in the digit IV, and the reduction of the ungual of the digit III. The pes morphology of Garumbatitan morellensis is unique within Somphospondyli and differs markedly from the morphology of Tastavinsaurus sanzi, which is represented by a well-preserved pes.
Garumbatitan morellensis is clearly distinguishable from other Iberian titanosauriforms (Soriatitan golmayensis, Tastavinsaurus sanzi, and Europatitan eastwoodi). In the case of Tastavinsaurus sanzi, which is recovered in a closer phylogenetic position than the other Iberian forms, the comparative study between these taxa supports the validity of Garumbatitan morellensis, which can be distinguished on the basis of its diagnostic characters and it is shown to lack the autapomorphies of Tastavinsaurus sanzi. Garumbatitan morrellensis is characterized by a distinct morphology of the hindlimbs, especially the femur, fibula, and pes, possibly reflecting differences gauge patterns when compared to Tastavinsaurus sanzi and Soriatitan golmayensis.
The somphospondylan Garumbatitan morellensis is the first sauropod taxon named on the basis of remains from the Arcillas de Morella Formation. The sauropod fauna recovered from this formation might consist of three titanosauriforms (based on Mocho et al. 2017b), including the somphospondylans Garumbatitan morellensis and the El Canteret sauropod. According to the most recent phylogenetic approaches and the topology recovered here, the sauropod fauna of the Iberian Peninsula during the Hauterivian–Aptian seems to consists of a complex phylogenetic mosaic composed of forms with Laurasian affinities, mainly titanosauriforms (the brachiosaurid Soriatitan golmayensis, and the somphospondylans Garumbatitan morellensis and, possibly, Tastavinsaurus sanzi and Europatitan eastwoodi), and Gondwanan affinities (the rebbachisaurid Demandasaurus darwini). There were probably faunal contacts between the sauropod faunas of Europe, North America, East Asia, and Africa during the Early Cretaceous: (i) rebbachisaurids may have dispersed between Africa and Europe via the Apulian route, and possibly some somphospondylan lineages as well; (ii) brachiosaurids were present in Europe and North America since the Late Jurassic, but a dispersion across Europe and North America during the Early Cretaceous with the establishment of a land connection between North America and Fennoscandia is not ruled out; and (iii) the close relationship of some Iberian forms with some non-titanosaurian somphospondylans of East Asia is a possible result of an exchange between Europe and Asia (a Late Jurassic dispersal event is also not ruled out). Future research on the remaining material from Sant Antoni de la Vespa and other historical and new specimens from the Arcillas de Morella Fm. will be important to improve our understanding of the phylogenetic relationships of Garumbatitan morellensis and other Iberian titanosauriforms, and to improve our understanding of the early evolution of the Somphospondyli clade.
SUPPLEMENTARY DATA
Supplementary data are available at Zoological Journal of the Linnean Society online.
File S1. Measurements of the sudied bones of Garumbatitan morellensis.
File S2. Map of synapomorphies.
File S3. TNT file.
ACKNOWLEDGEMENTS
This work was funded by the Portuguese Fundação para a Ciência e a Tecnologia (FCT) I.P./MCTES through a CEEC Individual contract (CEECIND/00726/2017) and the national funds (PIDDAC)—UIDB/50019/2020. This research was also supported by the Synthesys Project (http://synthesys3.myspecies.info/) which is financed by the European Research Council under the FP7 (FR-TAF-5072) (PM) and (DE-TAF-6138) (PM). This research was also supported by the Ministerio de Ciencia e Innovación of Spain (PID2019-111488RB-I00). For allowing accessing specimens, we want to thank: R. Allain (MNHN, France); S. Chapman (NHMUK, UK); A. Cobos and L. Alcalá (FCPT-Dinópolis, Spain); C. Corral (MCNA, Spain); S. Devincenzi (IANIGLA, Argentina); J.M. Herrero (MPG, Spain); P. Ortiz (PVL, Argentina); X. Pereda Suberbiola (UPV, Spain); M. Ramalho and R. Silva (MG, LNEG, Portugal); M. Reguero (MLP-Pv, Argentina); D. Schwarz and O. Hampe (MB.R, Germany). We also appreciate the critical comments of Rafael Royo-Torres. In memoriam of the archaeologist Miquel Guardiola Fígols who, thanks to his fieldwork, was able to locate the Sant Antoni de la Vespa fossil site. We want to thank the companies Renomar and Arcillas Vega del Moll and the institutions Castelló Cultural and Morella City Council for their collaboration. The Willi Hennig Society sponsors TNT cladistics software. We also appreciate the critical comments P.D. Mannion, S.F. Poropat, and an anonymous reviewer, and the editor, J. Streicher, for comments and suggestions.
CONFLICT OF INTEREST
None declared.
DATA AVAILABILITY
The specimens in this study are in the collections of the Museu Temps de Dinosaures-Museus de Morella (Morella, Castelló), which is part of Museums Survey of the Generalitat Valenciana (government of the Valencian Community). All other data are available in the Supporting Information. This article and associated nomenclatural acts are registered in ZooBank.
