ArticlePDF Available

Revision of Romanian sauropod dinosaurs reveals high titanosaur diversity and body-size disparity on the latest Cretaceous Haţeg Island, with implications for titanosaurian biogeography

Journal of Systematic Palaeontology

February 2025
23(1)

DOI:10.1080/14772019.2024.2441516

License
CC BY 4.0

Authors:

Verónica Díez Díaz

Museum für Naturkunde - Leibniz Institute for Research on Evolution and Biodiversity

Philip David Mannion

University College London

Zoltan Csiki-Sava

University of Bucharest

Paul Upchurch

University College London

The Hat , eg Basin and surrounding areas in Transylvania, western Romania, have been a hotspot for research on vertebrate faunas of the Late Cretaceous European Archipelago. One of the historically earliest titanosaurian sauropod dinosaurs to be discovered, the 'dwarfed' species Magyarosaurus dacus comes from lower Maastrichtian deposits in this basin; however, this species has been neglected, with no modern treatment of its anatomy, taxonomy or phylogenetic affinities. Via detailed anatomical study of historical and undescribed remains, combined with archival data, we identify shared autapomorphies that link multiple partial skeletons. Our analysis of hundreds of specimens (including >20 monospecific assemblages) enables the stabilization of the type species Magyarosaurus dacus. We propose the presence of three additional, but only partly contemporaneous taxa-Paludititan nalatzensis, Petrustitan ('Magyarosaurus') hungaricus n. gen. and Uriash kadici n. gen. n. sp. (the latter being amongst the largest known sauropods of the Late Cretaceous European Archipelago). We present a new phylogenetic analysis (152 taxa scored for 570 characters), with implications for broader titanosaurian evolutionary relationships and biogeography: Magyarosaurus is recovered either as a member or a close relative of Saltasauridae; Paludititan has affinities with Lognkosauria, along with the contemporaneous Lohuecotitan; Petrustitan is most closely related to South American early diverging eutitanosaurian taxa; and Uriash also shares affinities with Gondwanan taxa. Our findings strengthen the hypothesis that latest Cretaceous European titanosaurs belonged to Gondwanan lineages that invaded the former area during the Barremian-Albian. We interpret the presence of body-size disparity as either evidence that large-bodied taxa were ecologically excluded from body-size reduction by competition with small-bodied titanosaurs, or that dwarfing occurred stratigraphically earlier among several lineages and the small-bodied titanosaurs on Hat , eg Island are the descendants of existing dwarfed ancestors. Finally, we find no indication of a body size-related titanosaurian turnover in the uppermost Cretaceous of the Transylvanian area.

Content uploaded by Verónica Díez Díaz

Content may be subject to copyright.

Journal of Systematic Palaeontology

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/tjsp20

Revision of Romanian sauropod dinosaurs reveals high

titanosaur diversity and body-size disparity on the

lateﬆ Cretaceous Haţeg Island, with implications for

titanosaurian biogeography

Verónica Díez Díaz, Philip D. Mannion, Zoltán Csiki-Sava & Paul Upchurch

To cite this article: Verónica Díez Díaz, Philip D. Mannion, Zoltán Csiki-Sava & Paul

Upchurch (2025) Revision of Romanian sauropod dinosaurs reveals high titanosaur

diversity and body-size disparity on the latest Cretaceous Haţeg Island, with implications

for titanosaurian biogeography, Journal of Systematic Palaeontology, 23:1, 2441516, DOI:

10.1080/14772019.2024.2441516

To link to this article: https://doi.org/10.1080/14772019.2024.2441516

UK Limited, trading as Taylor & Francis

Group.

View supplementary material

Published online: 20 Feb 2025.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tjsp20

Revision of Romanian sauropod dinosaurs reveals high titanosaur diversity and

body-size disparity on the latest Cretaceous Hat¸eg Island, with implications for

titanosaurian biogeography

Ver

onica D

ıez D



ıaz

, Philip D. Mannion

, Zolt

an Csiki-Sava

and Paul Upchurch

Museum f€

ur Naturkunde, Leibniz-Institut f€

ur Evolutions- und Biodiversit€

atsforschung, Invalidenstraße 43, 10115 Berlin, Germany;

Department of Earth Sciences, University College London, Gower Street, London, WC1E 6BT, UK;

Laboratory of Paleontology,

Faculty of Geology and Geophysics, University of Bucharest, 010041 Bucharest, Romania

(Received 23 May 2024; accepted 14 November 2024)

The Hat

,eg Basin and surrounding areas in Transylvania, western Romania, have been a hotspot for research on

vertebrate faunas of the Late Cretaceous European Archipelago. One of the historically earliest titanosaurian sauropod

dinosaurs to be discovered, the ‘dwarfed’species Magyarosaurus dacus comes from lower Maastrichtian deposits in this

basin; however, this species has been neglected, with no modern treatment of its anatomy, taxonomy or phylogenetic

affinities. Via detailed anatomical study of historical and undescribed remains, combined with archival data, we identify

shared autapomorphies that link multiple partial skeletons. Our analysis of hundreds of specimens (including >20

monospecific assemblages) enables the stabilization of the type species Magyarosaurus dacus. We propose the presence

of three additional, but only partly contemporaneous taxa –Paludititan nalatzensis,Petrustitan (‘Magyarosaurus’)

hungaricus n. gen. and Uriash kadici n. gen. n. sp. (the latter being amongst the largest known sauropods of the Late

Cretaceous European Archipelago). We present a new phylogenetic analysis (152 taxa scored for 570 characters), with

implications for broader titanosaurian evolutionary relationships and biogeography: Magyarosaurus is recovered either

as a member or a close relative of Saltasauridae; Paludititan has affinities with Lognkosauria, along with the

contemporaneous Lohuecotitan;Petrustitan is most closely related to South American early diverging eutitanosaurian

taxa; and Uriash also shares affinities with Gondwanan taxa. Our findings strengthen the hypothesis that latest

Cretaceous European titanosaurs belonged to Gondwanan lineages that invaded the former area during the Barremian–

Albian. We interpret the presence of body-size disparity as either evidence that large-bodied taxa were ecologically

excluded from body-size reduction by competition with small-bodied titanosaurs, or that dwarfing occurred

stratigraphically earlier among several lineages and the small-bodied titanosaurs on Hat

,eg Island are the descendants of

existing dwarfed ancestors. Finally, we find no indication of a body size-related titanosaurian turnover in the uppermost

Cretaceous of the Transylvanian area.

Keywords: Cretaceous; Titanosauria; Hat

,eg Basin; Romania; island dwarfism; palaeobiogeography

Introduction

Titanosaurian sauropod dinosaurs include the largest

animals to ever walk on land, with gigantic species that

exceeded 60 tonnes (Benton et al., 2014; Bonaparte &

Coria, 1993; Carballido et al., 2017; Sander et al.,

2011). By the late Early Cretaceous, titanosaurs had

achieved a near-global distribution (Gallina et al., 2021;

Gorscak & O’Connor, 2016; Mannion, Upchurch, Jin,

et al., 2019), with their remains known from every land-

mass in the Late Cretaceous (Cerda et al., 2011; Curry

Rogers, 2005; Hocknull et al., 2009; Upchurch et al.,

2004). Despite a rich and global fossil record, titanosau-

rian evolutionary relationships are poorly known

(Carballido et al., 2022; Curry Rogers, 2005; Gorscak

et al., 2023; Mannion, Upchurch, Schwarz, et al., 2019),

limiting our understanding of this diverse clade of

megaherbivores (the only group of sauropods to survive

into the latest Cretaceous; Salgado et al., 1997). In par-

ticular, European titanosaurs have largely been neglected

in phylogenetic analyses (Mannion, Upchurch, Schwarz,

et al., 2019). This neglect mainly stems from the histor-

ical predominance of Gondwanan species, as well as the

scarcity and incompleteness of Laurasian remains, espe-

cially from Europe (Le Loeuff, 1993; Mannion,

Upchurch, Schwarz, et al., 2019; Wilson & Upchurch,

2003). However, this has begun to change through a

combination of reassessments of existing taxa and speci-

mens (D



ıez D

ıaz et al., 2011,2015;D



ıez D

ıaz, Pereda-

Suberbiola, et al., 2012,2013a,2013b;D



ıez D



ıaz,

Tortosa, et al., 2013), as well as the discovery of new

remains, including articulated, partial skeletons (Csiki,

Corresponding author. Email: diezdiaz.veronica@gmail.com

#2025 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits

unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The terms on which this article has been published allow the

posting of the Accepted Manuscript in a repository by the author(s) or with their consent.

Journal of Systematic Palaeontology, 2025

Vol. 23, No. 1, 2441516

http://dx.doi.org/10.1080/14772019.2024.2441516

Published online 20 Feb 2025

Codrea, et al., 2010;D



ıez D



ıaz, Garcia, et al., 2012;

D

ıez D

ıaz et al., 2014,2016,2018,2021; Garcia et al.,

2010; Le Loeuff et al., 2013; Vila et al., 2022). As

such, the latest Cretaceous (late Campanian–

Maastrichtian) European sauropod fossil record, in par-

ticular, is starting to reveal a rich evolutionary history,

with increasingly recognized importance for biogeo-

graphical scenarios (e.g. Holwerda et al., 2018;

Mannion, Upchurch, Schwarz, et al., 2019; Sallam

et al., 2018; Vila et al., 2022) and growing incorporation

into phylogenetic analyses (D

ıez D

ıaz et al., 2018,

2021; Gorscak et al., 2023; Gorscak & O’Connor, 2019;

Sallam et al., 2018; Vila et al., 2022).

Despite these developments, one of the historically

earliest titanosaurs to be discovered from Europe,

Magyarosaurus dacus, from the uppermost Cretaceous

(Maastrichtian) beds of Hat

,eg Basin in Transylvania,

western Romania (Fig. 1A), remains neglected, with no

modern treatment of its anatomy, taxonomy or phylo-

genetic affinities. This is despite the referral of numer-

ous remains to this taxon, representing most parts of the

skeleton (e.g. Huene, 1932), including cranial elements

(Weishampel et al., 1991)(Fig. 2). First described as a

species of Titanosaurus (T. dacus) by Nopcsa (1915),

this taxon was subsequently placed in its own genus,

Magyarosaurus, by Huene (1932), who erected two add-

itional species (M. transsylvanicus and ?M. hungaricus).

Subsequent to Huene’s(1932) taxonomic revision, with

some notable exceptions (see below), most titanosaur

remains recovered from the Maastrichtian of the

Transylvanian area were simply referred to

Magyarosaurus (or Titanosaurus)dacus, usually without

supporting evidence (e.g. Codrea et al., 2010; Jianu

et al., 1997; Jianu & Weishampel, 1999; Mamulea,

1953a; Vremir, 2010; Vremir et al., 2015). This practice

resulted in the view that a monospecific (at most mono-

generic) titanosaur taxon was present in the

Transylvanian area of Romania during the latest

Cretaceous (e.g. McIntosh, 1990; Upchurch et al.,

2004). The diminutive size of these fossil remains has

also interested researchers since the initial discovery of

the Transylvanian titanosaurs (e.g. Grellet-Tinner et al.,

2012; Jianu & Weishampel, 1999; Nopcsa, 1915; Stein

et al., 2010; Weishampel et al., 2010), particularly given

the gigantic nature of most members of the group

(Bonaparte & Coria, 1993; Carballido et al., 2017;

Lacovara et al., 2014). The consensus is that this repre-

sents an instance of island dwarfism during heightened

Late Cretaceous sea levels, when Europe formed an

archipelago (e.g. Csiki-Sava et al., 2015; Nopcsa,

1923b; Stein et al., 2010; Weishampel et al., 1991,

2010).

The taxonomy of Magyarosaurus is problematic.

Firstly, there is a great deal of uncertainty regarding the

historical association of specimens, making it unclear

whether it represents a chimaera. Secondly, although

most authors have considered only M. dacus to be valid

(e.g. Le Loeuff, 1993; Upchurch et al., 2004), it is

unclear whether the other two named species are syno-

nyms or represent greater diversity. Thirdly, an add-

itional taxon –Paludititan nalatzensis –was recognized

in 2010 (Csiki, Codrea, et al., 2010), demonstrating the

presence of at least a second approximately contempor-

aneous Romanian titanosaur. This emphasizes the dan-

ger that past referrals of fragmentary and non-

overlapping specimens to Magyarosaurus might have

created a chimaeric taxon. Finally, osteohistological

research targeting the Romanian titanosaurs has also

hinted at the presence of at least one other, large-sized

taxon in the faunal assemblage besides the small-bodied

M. dacus and Paludititan (Stein et al., 2010). As such,

it remains unclear what actually comprises

Magyarosaurus, limiting its utility in evolutionary and

palaeobiogeographical studies of Titanosauria.

In this study, we provide a comprehensive revision of

Magyarosaurus. We combine detailed anatomical study

of historical and undescribed remains, coming mainly

from the Hat

,eg Basin, with archival data to identify

shared autapomorphies that link multiple partial skele-

tons. This forms the basis for a revision of the systematics

of M. dacus,?M. hungaricus and M. transsylvanicus, and

an evaluation of their phylogenetic and biogeographical

affinities with respect to Paludititan and other titanosaurs

from Europe and elsewhere.

Institutional abbreviations

LPB (FGGUB), Laboratory of Palaeontology, Faculty of

Geology and Geophysics, University of Bucharest,

Romania; MCDRD, Muzeul Civilizatiei Dacice si

Romane (Museum of Dacian and Roman Civilisation),

Deva, Romania; NHMUK, Natural History Museum,

London, UK; SZTFH, Collection of the Supervisory

Authority for Regulatory Affairs (formerly the Geological

Institute of Hungary [MAFI], Budapest, Hungary); UBB,

Babes

,-Bolyai University, Cluj-Napoca, Romania.

Axial anatomical abbreviations

ACDL, anterior centrodiapophyseal lamina; ACPL,

anterior centroparapophyseal lamina; CDF, centrodiapo-

physeal fossa; CPOL, centropostygapophyseal lamina;

D, diapophysis; lSPRL, lateral spinoprezygapophyseal

lamina; mSPRL, medial spinoprezygapophyseal lamina;

NC, neural canal; PA, parapophysis; PACDF, parapo-

centrodiapophyseal fossa; PCDL, posterior

2V.D



ıez D

ıaz et al.

Revision of Romanian sauropod dinosaurs 3

centrodiapophyseal lamina; PCPL, posterior centropara-

pophyseal lamina; PO, postzygapophysis; POCDF, pos-

terior centrodiapophyseal fossa; PODL,

postzygodiapophyseal lamina; PPDL, paradiapophyseal

lamina; PRCDF, prezygocentrodiapophsyeal fossa;

PRDL, prezygodiapophyseal lamina; PRE, prezyga-

pophysis; PRSL, prespinal lamina; SDF, spinodiapophy-

seal fossa; SPDL, spinodiapophsyeal lamina; SPOF,

spinopostzygapophyseal fossa; SPOL, spinopostzygapo-

physeal lamina; SPRF, spinoprezygapophyseal fossa;

TPOL, interpostzygapophyseal lamina.

Other abbreviations

aEI (sensu Chure et al., 2010), average elongation

index, which scales the non-condylar centrum length

divided by the mean average of the mediolateral width

and dorsoventral height of the non-condylar centrum

articular surface; DWI (sensu Poropat et al., 2016), dis-

tal width index, defined as the ratio of the mediolateral

width of the distal end to the total proximodistal length

of the limb bone; ECC (sensu Wilson & Carrano,

1999), eccentricity index, calculated as the mid-shaft

mediolateral width divided by the anteroposterior diam-

eter of long bones; Hafd (sensu Upchurch et al., 2015),

humeral anconeal fossa depth divided by the anteropos-

terior width of the distal end; PWI (sensu Gonz

alez

Riga, 2003), proximal width index, the ratio of the

mediolateral width of the proximal end to the total prox-

imodistal length of the limb bone; RI (sensu Wilson &

Upchurch, 2003), robustness index, the ratio of the

mean average of the mediolateral widths of the proximal

and distal ends, as well as the minimum mediolateral

diameter of the shaft, to the total proximodistal length

of the limb bone.

Geological setting

Uppermost Cretaceous fossiliferous continental beds are

distributed patchily across several sedimentary basins in

western Romania; the most productive and concentrated

localities are in the Hat

,eg Basin and the south-western

Figure 2. Record of documented anatomy based on all Transylvanian titanosaur remains described in this work, summarized in one

individual. Not to scale. Skeletal drawing after Scott Hartman, used with permission.

Figure 1. Distribution of titanosaur fossil occurrences in Romania. A, position of the major areas with fossiliferous uppermost

Cretaceous continental deposits (labelled boxes); and B, areal distribution of the uppermost Cretaceous continental beds from these

major areas, highlighted in light green (HB –Hat

,eg Basin, RMB –Rusca Montan

a Basin, TB –south-western Transylvanian Basin).

C–E, main titanosaur-bearing fossiliferous localities in C, Rusca Montan

a Basin; D, south-western Transylvanian Basin (SbF –Sebes

Formation); and E, Hat

,eg Basin (DCF –Densus

,-Ciula Formation, Pb –‘Pui beds’,RMb–‘R^

aul Mare beds’, SpF –S

ınpentru

Formation). For details on key occurrences (coded A to V in Fig. 1E) see text (‘Key localities and skeletal associations’section).

4V.D



ıez D

ıaz et al.

part of the neighbouring and larger Transylvanian Basin

(Codrea et al., 2010; Csiki-Sava et al., 2016; Nopcsa,

1905; Vremir et al., 2015)(Fig. 1). Spatially more

restricted and relatively fossil-poor deposits are also

known from the Rusca Montan

a Basin (Fig. 1), as well

as from the western and north-western marginal areas of

the Transylvanian Basin (e.g. Codrea et al., 2010,2012;

Codrea & Godefroit, 2008; Csiki-Sava et al., 2016).

These deposits were laid down at the feet of mountain

chains uplifted by latest Cretaceous orogenic events that

shaped the nappe structure of the inner Carpathian area

(e.g. Willingshofer et al., 2001). They are dominated by

fluvial (coarser channel and finer-grained floodplain)

siliciclastic sediments, associated in the more western

areas (central-western Hat

,eg Basin, Rusca Montan

Basin) with volcanoclastic material, volcanic tuffs, and

minor lava flows, as well as coal bed intercalations

(Csiki-Sava et al., 2016). Deposition took place within

the confines of a subtropical island (‘Hat

,eg Island’or

Transylvanian Landmass), part of the Late Cretaceous

European Archipelago that fringed the northern margin

of the Mesozoic Neotethys-Alpine Tethys oceanic realm

(e.g. Benton et al., 2010; Csiki-Sava et al., 2015). Such

a palaeogeographical position, as well as the presence of

a dominantly semi-arid climate (e.g. Therrien, 2005), is

supported by palaeomagnetic studies that place Hat

,eg

Island at an approximate palaeolatitude of 29N

(Panaiotu & Panaiotu, 2010). The age of these continen-

tal beds is loosely constrained to the Maastrichtian by

their position atop biostratigraphically dated Campanian

marine beds (Melinte-Dobrinescu, 2010;T¸ab

ar

a et al.,

2022; Vremir et al., 2014), corroborated by magnetostra-

tigraphy (Panaiotu & Panaiotu, 2010), palynostratigra-

phy (Antonescu et al., 1983; Botfalvai et al., 2021; van

Itterbeeck et al., 2005), and a few radiometric ages

(Bojar et al., 2011). Locally, the transitional-to-continen-

tal beds also include part of the uppermost Campanian,

marking the onset of terminal Cretaceous continental

deposition well before 72 Ma (B

alc et al., 2024; Vremir

et al., 2014).

Several approximately synchronous lithostratigraphi-

cal units have been defined within the continental upper-

most Cretaceous of Transylvania (for a recent review

see Csiki-Sava et al., 2016), although direct correlation

between these is largely tentative because of their

patchy spatial exposure and lack (even locally) of exten-

sive marker beds. In the Hat

,eg Basin, these units

include: the S^

ınpetru Formation (Maastrichtian) in the

central part of the basin; the partly volcanoclastic

Densus

,-Ciula Formation (Maastrichtian) in its north-

western marginal areas (Grigorescu, 1992); and as yet

not formally defined sedimentary successions in the

west-central (‘R^

aul Mare Beds’; probably ’middle’to

upper Maastrichtian) and eastern-central (‘Pui Beds’,

upper part of lower to lower part of upper

Maastrichtian) areas of the depression. Fossils, including

vertebrate remains, are common throughout these units

(Csiki-Sava et al., 2016; Grigorescu, 1983; Nopcsa,

1915). In the Transylvanian Basin, the major fossil-bear-

ing unit is the Sebes

,Formation (uppermost Campanian–

upper Maastrichtian; (Vremir, 2010; Vremir et al.,

2015); ¼S

,ard Formation (Codrea et al., 2010), distrib-

uted in its south-western part, whereas the

Maastrichtian–Paleocene Jibou Formation (Codrea et al.,

2010) that outcrops in its western-north-western areas

has yielded only rare latest Cretaceous fossils (Codrea

& Godefroit, 2008). Finally, the uppermost Cretaceous

(Maastrichtian) continental beds of the Rusca Montan

Basin have yet to receive a formal lithostratigraphical

designation; these have only recently yielded vertebrate

fossils (Codrea et al., 2012; Csiki-Sava et al., 2016;

Vasile & Csiki, 2011).

For more than a century (e.g. Nopcsa, 1897), a rich

continental assemblage has been recovered from these

Transylvanian uppermost Cretaceous deposits. Plants are

represented by diverse palynological assemblages (e.g.

Antonescu et al., 1983; Botfalvai et al., 2021; Csiki

et al., 2008;T¸ab

ar

a et al., 2022;T¸ab

ar

a & Csiki-Sava,

2024; van Itterbeeck et al., 2005), mesofloral elements

(seeds, fructifications; May Lindfors et al., 2010), as

well as macroplant remains (e.g. Popa et al., 2014,

2016). Invertebrates are less well studied, but ostracods

(Silye et al., 2014), gastropods (P

all-Gergely et al.,

2023; Pan

a et al., 2002), and insects (Augustin et al.,

2019; Csiki, 2006; Vremir, 2009) have all been

reported. Vertebrates make up the best-known compo-

nent of the local palaeoecosystem, and these include

diverse taxa of fish, anurans, albanerpetontids, turtles,

lizards, snakes, crocodyliforms, pterosaurs, non-avian

dinosaurs (various theropods, titanosaur sauropods, rhab-

dodontid and hadrosauroid ornithopods, nodosaurid

ankylosaurs), birds, and multituberculate mammals (e.g.

Benton et al., 2010; Csiki-Sava et al., 2015,2016;

Nopcsa, 1923a; Weishampel et al., 2010).

Among the dinosaurs, titanosaur sauropods represent

one of the most commonly occurring groups (e.g. Csiki,

Grigorescu, et al., 2010, fig. 3; Grigorescu, 1983).

Titanosaur remains are widespread, having been discov-

ered in most major vertebrate localities in the Hat

,eg,

south-western Transylvanian, and Rusca Montan

basins, as both associated and isolated skeletal elements

(e.g. Botfalvai et al., 2021; Codrea et al., 2010,2012;

Csiki, Codrea, et al., 2010; Grigorescu, 1983; Huene,

1932; Nopcsa, 1915; Vremir, 2010; Vremir et al., 2015).

Caudal vertebrae, pectoral girdle, and appendicular ele-

ments represent the most commonly occurring titanosaur

Revision of Romanian sauropod dinosaurs 5

skeletal remains recovered, whereas presacral vertebrae,

elements of the pelvic girdle, and skull remains are

uncommon to very rare (Fig. 2; see below and

Supplemental Material). Contrary to their wide geo-

graphical distribution in Romania, their stratigraphical

distribution seems to be more skewed, with the few lat-

est Campanian–earliest Maastrichtian vertebrate-bearing

localities apparently lacking titanosaur remains (e.g.

B

alc et al., 2024; Vremir et al., 2014).

History of titanosaur discoveries and

research in Transylvania

Specimen discoveries

Despite their common presence in the continental upper-

most Cretaceous of Romania, and the relatively large

number of studies devoted to them in the past, the tax-

onomy, systematics and phylogenetic affinities of the

Transylvanian titanosaurs remain problematic.

Confusion and disagreement have arisen mainly because

of the often isolated and fragmentary nature of their

remains, as well as the quasi-absence of reliable locality

and association data for the material that was discovered

and excavated before 1990.

The first reports to mention the presence of latest

Cretaceous vertebrate remains in the Hat

,eg Basin were

by Halav

ats (1897) and Nopcsa (1897). Although nei-

ther of these publications specifically recorded the pres-

ence of sauropod remains in the area, inspection of the

SZTFH collections and registry records in Budapest

shows that the first sauropod remains currently on

record were discovered in 1896 in the Sibis

,el Valley,

south of S^

anpetru (stratotype area of the S^

ınpetru

Formation; Grigorescu, 1992, figs 1, 3). These speci-

mens, represented by two caudal vertebrae inventoried

as SZTFH Ob.1945 and Ob.1947 (Fig. 3), were col-

lected by Gyula Halav

ats, a geologist of the Hungarian

Royal Geological Survey, who was mapping in the cen-

tral-eastern parts of the Hat

,eg Basin (Halav

ats, 1897).

The Halav

ats specimens remained unidentified until

1904, when they were officially accessioned in the

SZTFH collections (L. Mak

adi, pers. comm. February

2019). The first explicit mention of sauropod remains in

the Hat

,eg Basin was by Ferenc (¼Franz) Nopcsa

(1902a). Slightly later, Nopcsa (1904,1905) noted the

presence of as many as three different sauropod taxa in

the Hat

,eg faunal assemblage, including Titanosaurus,

which he identified based on its caudal vertebral morph-

ology (and to which he also referred a humerus, an

ischium, and two femora), and two indeterminate taxa

whose presence was suggested by distinctive and diver-

gent vertebral and humeral morphologies. However,

none of these sauropod specimens were described or fig-

ured, and no information was offered in support of these

taxonomic identifications, except for a personal commu-

nication from the French geologist and palaeontologist

Charles Deper

et, who noted the similarity of some

Transylvanian caudal vertebrae to those of titanosaurs.

The presence of titanosaurs in the uppermost Cretaceous

Figure 3. First Late Cretaceous vertebrate remains reported in the Hat

,eg Basin (Halav

ats, 1897; Nopcsa, 1897), two titanosaur

anterior caudal vertebrae: SZTFH Ob.1945 in A, posterior; B, right lateral; C, anterior; and D, ventral views. SZTFH Ob.1947 1945

in E, anterior; F, left lateral; G, posterior; and H, ventral views. Scale bar equals 50 mm.

6V.D



ıez D

ıaz et al.

continental beds of the Transylvanian region was thus

first identified by Nopcsa based mainly on his personal

collection from S^

anpetru, excavated along the Sibis

,el

Valley (Hat

,eg Basin), and which was subsequently pur-

chased by the NHMUK (London) in 1906.

The aforementioned titanosaur specimens from the

Hat

,eg Basin were by no means the first ones discovered

in the wider Transylvanian region. In his geological

overview of south-western Transylvania, including the

Hat

,eg Basin, Nopcsa (1905) also mentioned the previ-

ous discovery of sauropod remains (a humerus and

femur) from R^

apa Ros

,ie, north-east of Sebes

,, in the

south-western Transylvanian Basin (Fig. 1D), that had

been misidentified as large mammal bones by Koch

(1894). To these, Nopcsa added a large-sized sauropod

ulna he had recovered from the same locality. Together

with a very incomplete and poorly preserved large limb

bone from B

ar

abant

,(also south-western Transylvanian

Basin; Fig. 1D), discovered in 1860 but only recognized

as a possible sauropod tibia more than a century later

(Codrea & M

arginean, 2007), these fossils from R^

apa

Ros

,ie represent not only the first sauropod remains, but

also the first dinosaur remains discovered from the

entire Transylvanian area. Unfortunately, the R^

apa

Ros

,ie specimens appear to be lost now, and the

B

ar

abant

,specimen is very poorly preserved; thus, the

precise nature and affinities of these early sauropod dis-

coveries from Transylvania remain unknown.

Another important early collection of vertebrate

(including titanosaur) fossils was amassed by Ottok

Kadi

c, a geologist of the Hungarian Royal Geological

Survey. He was assigned to map in the north-western

Hat

,eg Basin, where he discovered dinosaur bones

around V

alioara (Fig. 1E) in 1909 (Kadi

c, 1911). In

subsequent years, he returned to this area to conduct

several successful excavation campaigns, the results of

which today form the core of the SZTFH collection of

latest Cretaceous continental vertebrates in Budapest

(Botfalvai et al., 2021; Kadi

c, 1916). Together, the

NHMUK material collected by Nopcsa at S^

anpetru and

the SZTFH collection of Kadi

c from V

alioara allowed

Nopcsa (1915) to write his first synthesis on the com-

position and palaeobiology of the Transylvanian dino-

saur fauna, a study that also includes the first

description and taxonomic identification of a sauropod

from this fauna. Despite his previous comments con-

cerning the presence of several sauropod taxa in the

Hat

,eg Basin, and his earlier interpretation that titano-

saur-like vertebrae belonged instead to the hadrosauroid

Telmatosaurus (Nopcsa, 1910), Nopcsa (1915) recog-

nized the presence of a single taxon of titanosaur sauro-

pod as a new species of the (now obsolete; Wilson &

Upchurch, 2003) genus Titanosaurus, namely T. dacus.

He briefly characterized this new species and figured a

few of the referred skeletal elements from the SZTFH

collection (Nopcsa, 1915, pl. III, figs 4, 5, 8), repre-

sented by isolated middle and posterior caudal vertebrae

from V

alioara, and an ungual (now lost, but presumably

also from V

alioara). Later, in his second synthesis of

the Transylvanian vertebrate fauna, Nopcsa (1923a)

commented further on the anatomy of T. dacus, as well

as on the evolutionary and palaeobiogeographical impli-

cations of its presence in uppermost Cretaceous rocks of

eastern Europe.

Nopcsa had planned to describe this material himself

as part of his ‘Dinosaurierreste aus Siebenb€

urgen …’

series of monographs, according to letters to Friedrich

Huene in 1926–1927 (translated in Weishampel &

Kerscher, 2013), but these plans never came to fruition.

Thus, unlike the significantly larger amount of work

(and number of publications) he devoted to the study of

ornithischians and turtles of this fauna (e.g. Nopcsa,

1900,1902a,1923b,1929), these two short contribu-

tions (i.e. Nopcsa, 1915,1923a) represent the extent of

Nopcsa’s published research efforts towards the study of

the Transylvanian sauropod material. In order to com-

pensate for this shortcoming (partly caused by health

problems; see Weishampel & Jianu, 2011; Weishampel

& Kerscher, 2013), Nopcsa invited Huene to study and

describe the Transylvanian titanosaur material, sending

him his notes, drawings, and even shipping several

specimens from the Budapest collection to T€

ubingen,

south-west Germany, where Huene was based. This

study was first envisaged as a collaboration between the

two workers (Weishampel & Jianu, 2011, p. 203), but

Nopcsa subsequently entrusted it entirely to Huene

(Weishampel & Kerscher, 2013), and thus Huene

included an extensive chapter on the Transylvanian tita-

nosaurs in his monograph on saurischian dinosaurs

(Huene, 1932).

Huene’s detailed survey of the Transylvanian sauro-

pod material available at that time (i.e. the entirety of

the London and Budapest sauropod collections), sup-

ported by his previous work on South American titano-

saurs and overall sauropod diversity (e.g. Huene, 1929),

led him to refer this material without exception to

Titanosauria. Furthermore, he recognized morphological

differences between the Transylvanian titanosaur mater-

ial and the holotype of Titanosaurus (T. indicus) from

the uppermost Cretaceous of India (Lydekker, 1877),

which led him to erect a new genus, Magyarosaurus,

for the Hat

,eg titanosaur remains. In order to account for

perceived morphological, proportional, and dimensional

differences within the studied material, Huene also

erected two further species within Magyarosaurus, rec-

ognizing the small-to-medium sized M. dacus and M.

Revision of Romanian sauropod dinosaurs 7

transsylvanicus, and the larger ?M. hungaricus, while

assigning other remains as either Magyarosaurus sp. or

as indeterminate titanosaurs.

Following the Nopcsa and Kadi

c excavations and the

survey of Huene, there was a significant reduction in

the pace of the collection and study of Transylvanian

titanosaurs that extended until the last decades of the

twentieth century. Nevertheless, even though subsequent

collecting efforts were serendipitous and not systematic,

important titanosaur specimens were discovered in the

Hat

,eg region by prospecting geologists and university

staff carrying out fieldwork between 1932 and 1977.

Foremost of these are the fossils excavated by field

geologist Mihai Alfred Mamulea in the neighbourhood

of Ciula Mic

a(Fig. 1E). Although at first he only

reported the discovery of unspecified dinosaur bones in

bluish-reddish sandstones (Mamulea, 1953a), he speci-

fied that he had recovered from Ciula Mic

a‘a few

nicely preserved fossil remains belonging to T. dacus

NOPCSA’in a second, more detailed geological report

dealing with the central part of Hat

,eg Basin (Mamulea,

1953b, p. 250, in Romanian in the original). According

to personal information retrieved from faculty staff

members at the FGGUB (T. Neagu, D. Grigorescu, pers.

comm., 2004), Mamulea donated specimens he collected

in the Hat

,eg area to the LPB (FGGUB) collection after

retiring. Some of these specimens were figured in an

unpublished BSc thesis (Ungureanu, 1979), including

one well-preserved titanosaur humerus (LPB [FGGUB]

R.1047) displaying a paper label attached which indi-

cates that it was collected by Mamulea (Fig. 4);

unfortunately, this label was removed and probably lost

during subsequent preparation and consolidation of the

specimen. Based on this circumstantial evidence, as well

as preservational similarities and commensurate size, a

series of titanosaur skeletal elements were identified in

the LPB (FGGUB) collections that correspond to the

fossils excavated by Mamulea. Furthermore, the absence

of overlapping skeletal elements within this assemblage

(see ‘Key localities and skeletal associations’and

‘Systematic palaeontology’sections) and their commen-

surate size suggest that these specimens were not only

excavated together from the same locality, but they

most probably represent parts of one incomplete skel-

eton that includes associated and relatively well-pre-

served dorsal and caudal vertebrae (some of them still

in articulation), as well as limb elements. As such, this

individual (‘Individual E’: see ‘Key localities and skel-

etal associations’section) represents one of the few

cases of historical titanosaur discoveries from

Transylvania where skeletal association can be reason-

ably assumed, with important consequences for our

understanding of the anatomy and systematics of

Magyarosaurus.

Before 1980, further isolated titanosaur remains –

besides the ‘Mamulea specimen’–were also recovered

from S^

anpetru and Pui (Fig. 1E). Several titanosaur fos-

sils from S^

anpetru (caudal vertebrae and limb bones)

were excavated from a fossiliferous pocket by Nicolae

M

esz

aros and H. Krauss in 1966. These specimens,

together with isolated titanosaur bones found by

M

esz

aros and T

egl

as in 1979, are now registered in the

Figure 4. Left humerus LPB (FGGUB) R.1047 from Individual E, collected by M. A. Mamulea (indicated in the paper label

attached to it). A, anterior view; and B, posterior view, from Ungureanu (1979). C, paper label, from Ungureanu (1979). D, anterior

view. Scale bar equals 100 mm.

8V.D



ıez D

ıaz et al.

UBB collections (Jianu et al., 1997). All of these sauro-

pod remains, mostly fragmentary and rather poorly pre-

served, were later referred, without further detail or

argumentation, to M. dacus (Jianu et al., 1997). Dinc

et al. (1972) reported the discovery of an isolated caudal

vertebra from S^

anpetru (LPB [FGGUB] R.1062; Fig.

5A–G), and tentatively referred it to Titanosaurus cf.

indicus. These authors suggested in passim –and with-

out supporting evidence –that ‘T’.dacus might be syn-

onymous with T. indicus (see the Supplemental

Material). At Pui, dinosaur remains were collected by

Mamulea (1953b) and Stilla (1985); of these, a few iso-

lated titanosaur specimens donated by Alexandru Stilla

(scapula fragment, LPB [FGGUB] R.1153, and humerus

shaft, LPB [FGGUB] R.0555) are now curated in the

collections of the FGGUB.

Despite the occasional discovery of mainly isolated

and some more complete titanosaur specimens, such as

those noted above, it was only with the renewal of

palaeontologically focused exploration and excavation

efforts after 1977 that more significant collections of lat-

est Cretaceous continental vertebrates were amassed

(see Grigorescu [2010] for a review). These post-1977

efforts led to the assembly of important collections at

the University of Bucharest (LPB [FGGUB]), the

Babes

,-Bolyai University in Cluj-Napoca (UBB), and the

Muzeul Civilizat

,iei Dacice s

,i Romane in Deva,

Hunedoara County (MCDRD). Significantly, details on

locality and skeletal association were recorded for many

of these more recent titanosaur discoveries, a source of

information that is largely lacking for older collections

and whose absence has led to the confusion and

Figure 5. Titanosaur middle caudal vertebra LPB (FGGUB) R.1062 from S^

anpetru previously referred to Titanosaurus cf. indicus in

A, right lateral; B, ventral views; C, anterior; D, left lateral; E, posterior; F, ventral; and G, dorsal views. A and B are taken from

Dinc

a et al. (1972). NHMUK R.40867 (holotype of Titanosaurus indicus from India) in H, anterior; I, left lateral; J, posterior; K,

ventral; and L, dorsal views, for comparison. Scale bars equal 50mm.

Revision of Romanian sauropod dinosaurs 9

misinterpretation that has plagued the taxonomy of

Transylvanian titanosaurs. Starting in 1979, systematic

excavation campaigns focused primarily on the Hat

,eg

Basin, but also targeted the south-western Transylvanian

and Rusca Montan

a basins. These efforts were initiated

through collaboration between the FGGUB (coordinated

by Dan Grigorescu, more recently also including ZCs-S)

and the MCDRD (first conducted by Ion Groza, then

by Coralia-Maria Jianu). From the mid-1990s, these

prospecting and excavation activities were joined by the

UBB (coordinated by Vlad Codrea), the Transylvanian

Museum Society (M

aty

as Vremir), and the Ioan Raica

Museum, in Sebes

,-Alba (Radu Totoianu), and were

often made in collaboration with colleagues from

abroad. This work resulted in the discovery and excava-

tion of a large number of vertebrate localities that have

yielded a considerable amount of titanosaur material

either in isolation or mixed with other vertebrates.

These titanosaur specimens occasionally include associ-

ated skeletal remains (e.g. Botfalvai et al., 2021; Csiki-

Sava et al., 2012; Groza, 1983; van Itterbeeck et al.,

2004), and even partial skeletons (Csiki, Codrea, et al.,

2010). Because of these new discoveries, the areal

coverage of known titanosaur localities expanded to

include sites in the Hat

,eg Basin such as Pui (e.g.

Codrea & Solomon, 2012; Grigorescu et al., 1985;

,tiuc

a, 1983; van Itterbeeck et al., 2004), N

alat

,-Vad

(Csiki, Codrea, et al., 2010; 2016; Smith et al., 2002),

Livezi (Grigorescu & Csiki, 2008), Tus

,tea (Csiki-Sava

et al., 2012), and Boit

,a (Csiki-Sava et al., 2018), in add-

ition to further occurrences around S^

anpetru (Csiki,

Grigorescu, et al., 2010; Groza, 1983;S

,tiuc

a et al.,

1982), V

alioara and Ciula Mic

a (Botfalvai et al., 2021).

Outside of the Hat

,eg Basin, in the south-western

Transylvanian Basin, the R^

apa Ros

,ie locality continued

to yield titanosaur remains (e.g. Grigorescu, 1987, misi-

dentified as ankylosaur elements; Codrea et al., 2008;

Csiki & Vremir, 2011; Jianu et al., 1997), and new tita-

nosaur localities (S

,ard, Oarda de Jos, Lancr

am, Sebes

Glod, Teleac, MI6) were also identified (e.g. Codrea

et al., 2010; Vremir, 2010; Vremir et al., 2015). Finally,

rare titanosaur fossils were also discovered in the Rusca

Montan

a Basin (Codrea et al., 2012).

Recent decades have brought about some outstanding

titanosaur discoveries from the Transylvanian area.

These include the excavation and identification of a

well-preserved and fairly complete braincase at Pui

,tiuc

a, 1983; Weishampel et al., 1991), the only cur-

rently known cranial material of a Transylvanian titano-

saur, as well as isolated teeth (Botfalvai et al., 2021;

Grigorescu et al., 1985;}

Osi et al., 2017a). The La

C

arare locality from S^

anpetru has yielded the only

known titanosaur osteoderm from Romania, associated

with caudal vertebrae and limb bones, showing that at

least some Transylvanian titanosaurs were armoured

(Csiki, 1999). Although at the time of their first identifi-

cation, megaloolithid eggs from the Hat

,eg Basin were

preliminarily attributed to titanosaurs (Grigorescu et al.,

1990), most of these are now considered to belong to

hadrosauroid dinosaurs (e.g. Grigorescu, 2010).

Nevertheless, Grellet-Tinner et al. (2012) suggested that

certain megaloolithid eggs from Totes

,ti-Baraj (Fig. 1E)

were laid by titanosaurs, thus further expanding the list

of sauropod fossil localities. However, these titanosaur

remains will not be covered in this work, especially

given that their taxonomic affinities are practically

impossible to establish because of the lack of clear asso-

ciation with postcrania.

Taxonomic and phylogenetic work

The persistence of substantial systematic uncertainties

that surround Transylvanian titanosaurs is perhaps sur-

prising given the large volume of contributions that

have addressed this topic in the past. After the identifi-

cation of the titanosaur species T. dacus by Nopcsa

(1915) and its referral, together with two other new spe-

cies (M. transsylvanicus,?M. hungaricus), to the newly

erected genus Magyarosaurus by Huene (1932), the

issue of the taxonomic identity of the Transylvanian

sauropod material entered a long period of neglect, fol-

lowed by dissenting opinions. Steel (1970) and

McIntosh (1990) recognized all three previously formal-

ized taxa as valid species, but these authors assigned

them to Titanosaurus and Magyarosaurus, respectively.

Despite this, McIntosh (1990) noted that these species

could barely be diagnosed by autapomorphies, although

he recognized that it was conceivable that more than

one taxon was present within the known fossil material.

After first-hand study of the London and Budapest col-

lections of Transylvanian titanosaurs, Le Loeuff (1993)

suggested that the three species separated by Huene

within Magyarosaurus could not be differentially diag-

nosed, and that they were in fact synonymous, falling

under the name M. dacus (see also Csiki, 1999;Le

Loeuff, 2005a). Le Loeuff (1993) stated that the first

titanosaur specimens described by Nopcsa (1915), while

naming T. dacus, correspond to two articulated dorsal

vertebrae (NHMUK R.4896). Le Loeuff accordingly

designated these vertebrae as the lectotype of the species

(a designation also accepted, as ‘type’, by Jianu &

Weishampel, 1999). However, this is incorrect, as we

explain below in the ‘Systematic palaeontology’section.

Synonymy of the three Magyarosaurus species was also

accepted in the reviews of Wilson and Upchurch (2003),

Upchurch et al. (2004), and Curry Rogers (2005).

10 V. D



ıez D

ıaz et al.

The contrasting viewpoint that there was a higher

diversity of titanosaurs on the Transylvanian landmass

was re-stated by Csiki and Grigorescu (2004). More

specifically, the occurrence of large-sized titanosaur

specimens with a distinctive morphology was noted by

Csiki and Grigorescu (2006), Csiki et al. (2007), and

Csiki and Vremir (2011). These authors preliminarily

interpreted these large specimens as potentially repre-

senting taxa distinct from Nopcsa’s and Huene’s small-

sized M. dacus, but possibly attributable to Huene’s

larger ?M. hungaricus. The presence of a larger titano-

saur taxon was also supported by the osteohistological

study of Stein et al. (2010). More recently, the existence

of a higher taxic diversity was firmly established with

the description of a second small-bodied titanosaur

genus, Paludititan nalatzensis, from the Transylvanian

area (Csiki, Codrea, et al., 2010), and was also hinted at

in a preliminary overview of the Transylvanian titano-

saur material by Csiki et al. (2011). This high diversity

hypothesis was also supported by Mocho et al. (2023),

based on their interpretation of four distinct caudal ver-

tebral morphotypes.

An unusual outlier in the history of Transylvanian

titanosaur taxonomy is the proposed presence of

Titanosaurus cf. indicus in the Hat

,eg Basin, as sug-

gested by Dinc

a et al. (1972). However, our recent re-

location of this specimen (Fig. 5A–G) in the collections

of the LPB (FGGUB) demonstrates that it cannot be

referred to Titanosaurus indicus, and thus the proposed

presence of this taxon in the Transylvanian area cannot

be supported (see ‘Additional titanosaur remains’sec-

tion in the Supplemental Material).

Magyarosaurus, considered to be a ‘titanosaurid’by

McIntosh (1990), a non-saltasaurid lithostrotian by

Upchurch et al. (2004), and a saltasaurid by Le Loeuff

(2005a), has only been included previously in one

phylogenetic data matrix (Curry Rogers, 2005), where it

was recovered as a lithostrotian titanosaur of uncertain

placement (see also Mannion & Upchurch, 2011). In

light of the more recently recognized diversity of

Transylvanian titanosaurs, it is likely that the operational

taxonomic unit (OTU) used as Magyarosaurus by Curry

Rogers (2005) represents a ‘chimera’based on skeletal

elements derived from more than one taxon. More

recently, Paludititan has been included in several phylo-

genetic analyses, in which it was also recovered as a

lithostrotian titanosaur, although its position varied

between studies (Csiki, Codrea, et al., 2010;D



ıez D

ıaz

et al., 2018,2021; Gorscak & O’Connor, 2016,2019;

Navarro et al., 2022; Sallam et al., 2018; Vila et al.,

2022).

In summary, the taxonomic status of the

Transylvanian titanosaurs is far from resolved, as

emphasized by several authors (e.g. Csiki, 1999;Le

Loeuff, 1993; Upchurch et al., 2004; Weishampel et al.,

1991), but it appears that they formed a more diverse

assemblage than often acknowledged in the past. In light

of these taxonomic uncertainties, the phylogenetic affin-

ities of these titanosaurs also remain unclear.

Key localities and skeletal associations

In this section, we draw together multiple lines of evi-

dence (including field notes, museum labels, registry

notes, published data, and specimen preservation) to

establish a set of titanosaur individuals, or in some cases

assemblages from a given locality, in order to provide

the basis for recognizing diagnosable specimens and jus-

tifiable taxonomic referrals. We do not attempt to iden-

tify every specimen that has ever been referred to

Magyarosaurus: instead, we focus on key name-bearing

historic specimens and ‘Rosetta Stone’specimens that

are crucial in terms of their potential diagnostic charac-

ters and/or overlapping anatomy that enable compari-

sons and referrals.

Background

The complicated and ultimately poorly understood taxo-

nomic framework concerning the Transylvanian latest

Cretaceous titanosaurs stems from several different sour-

ces. One factor pertains to preservational biases, such as

the predominance of disarticulated and often isolated

specimens, and the severe underrepresentation or even

absence of certain body parts (especially the skull and

the pelvic girdle; Fig. 2). Disarticulated, isolated, and/or

fragmentary specimens, that were discovered in different

and sometimes distant localities, are often difficult to

compare meaningfully, let alone refer with certainty to

the same taxon. Despite this, aggregation of non-associ-

ated and anatomically non-overlapping specimens into

named taxa (primarily Magyarosaurus dacus) has been

the dominant taxonomic practice since the original dis-

covery of this fauna (see above). A second confounding

factor is the lack of detail in the available field docu-

mentation concerning the discovery and excavation of

many specimens. This is particularly true for the histor-

ical, pre-1978 collections, which include key material

such as the remains from S^

anpetru discovered by

Nopcsa (now registered in the NHMUK collections),

and those from V

alioara excavated by Kadi

c (now part

of the SZTFH collections): these two collections repre-

sent the foundation of the original taxonomic opinions

proposed by Nopcsa (1915) and Huene (1932).

Documentation for most of these early specimens simply

mentions S^

anpetru or V

alioara as their place of origin,

Revision of Romanian sauropod dinosaurs 11

without reference to a particular location in the sur-

roundings of these villages. Although general informa-

tion (approximate location, lithology, and/or

palaeontological content) for particular excavation sites

is recorded in some early publications (e.g. Kadi

c, 1916;

Nopcsa, 1902a,1902b,1914), it tends to lack explicit

details on both the precise identity of the titanosaur

specimens that were excavated from these sites and the

potential skeletal association of these specimens. Other

sources of provenance information (i.e. field notes, geo-

graphical and quarry maps, photographs, collection

registry entries, or labels accompanying the specimens)

are equally rare, albeit with some notable exceptions

(see below). Even more frustratingly, it appears that the

omission of the existing excavation or registry details

was common during this period, as most saliently dem-

onstrated by Nopcsa’s failure to convey information on

skeletal association for the titanosaur material he sent to

Huene for study in T€

ubingen (see below). In short, most

of the original information concerning the provenance

and field association of the historical titanosaur speci-

mens should be considered lost and nearly impossible to

reconstruct. The situation is somewhat improved for the

titanosaur material collected after 1977, for which more

precise locality and/or field association data are avail-

able (e.g. Csiki, Codrea, et al., 2010; Csiki-Sava et al.,

2012; Groza, 1983), or at least such information can be

retrieved from personal or orally transmitted accounts of

their discoveries, unpublished field notes, and/or

museum collection records (see below).

In this section, we review localities that have yielded

historically important (e.g. type specimens) and/or asso-

ciated skeletal material, in order to document the main

sources of information that are instrumental in our revi-

sion of the systematics of the latest Cretaceous

Transylvanian titanosaurs. Here, we gather, collate, and

synthesize all available data for the different key titano-

saur localities and individuals (Fig. 1E) to provide sup-

port for our inferences of association and provenance.

This is especially important given that we are dealing

with material collected over a period of more than a

century by researchers of different backgrounds and

aims, and covering a relatively large geographical area.

A necessary part of this process is to identify and clarify

confusions and misinterpretations promulgated by previ-

ous publications. For ease of discussion, we designate

each key set of associated specimens as ‘individuals’

(considered to represent one incomplete skeleton), or as

‘assemblages’when there is element duplication or other

reasons for questioning the presence of just one individ-

ual (though in some cases these assemblages appear to

largely comprise a single individual with just a small

number of extraneous elements from others). These

individuals and assemblages are labelled A, B, C, etc.

We also present a graphical overview of the skeletal

parts represented for the most important key individuals

to aid comparisons and highlight overlapping skeletal

parts among them.

Kadi

c locality I (5Magyarosaurus dacus type

locality), Ciula Mic

a–Assemblage A

When erecting the new taxon Titanosaurus dacus,

Nopcsa (1915, pl. III) figured only three specimens (two

procoelous caudal vertebrae and an ungual; Fig. 6). All

of these specimens belong to the Kadi

c collection that

was assembled in the neighbourhood of V

alioara village

(Kadi

c, 1916) and is housed in the SZTFH. Of the three

specimens figured by Nopcsa, and referred explicitly to

T. dacus, the ungual (Nopcsa, 1915, pl. III, fig. 8;

SZTFH Ob.3098; Fig. 6M, N) is now lost. The other

two referred specimens represent an incomplete distal

anterior caudal vertebra (Nopcsa, 1915, pl. III, fig. 4;

SZTFH Ob.3091; Fig. 6A–F) and a posterior caudal

centrum (Nopcsa, 1915: pl. III, fig. 5; SZTFH Ob.4215;

Fig. 6G–L). Only in the case of the latter specimen does

the plate caption make a direct reference to the species

epithet (i.e. dacus), although Nopcsa’s intention to refer

the other two figured specimens to the same species can

be reasonably inferred for two reasons. First, he recog-

nized only one titanosaur species, to which he referred

all the sauropod material collected throughout the Hat

,eg

Basin. Second, in the overview list of the figured mater-

ial that precedes the plates and their captions, the distal

anterior caudal (SZTFH Ob.3091) is also listed expli-

citly as T. dacus. In addition to these three specimens,

Nopcsa (1915, pl. III, figs 6, 7) referred two platycoe-

lous posterior caudal vertebrae to an indeterminate spe-

cies of Megalosaurus, a theropod dinosaur, but these are

clearly referable to a titanosauriform sauropod based on

first-hand inspection (SZTFH Ob.3105 in partim;

Fig. 7).

In his review of the Transylvanian titanosaurs, Huene

(1932, p. 263) emphasized that he regarded SZTFH

Ob.3091 as the specimen that typifies Magyarosaurus

(‘Titanosaurus’)dacus (“Diesen Wirbel nehme ich als

den Typus der Art”). Given that Nopcsa (1915) did not

formally designate a type for T. dacus, we follow

Huene (1932) in regarding SZTFH Ob.3091 as the lec-

totype for the species (see also Csiki, Codrea, et al.,

2010). This decision is also appropriate because SZTFH

Ob.3091 is the most complete and diagnostic element of

the three originally figured specimens.

Huene (1932) mistakenly mentioned that SZTFH

Ob.3091 came from S^

anpetru, but the type locality for

this taxon must be located somewhere in the vicinity of

V

alioara, the locus of Kadi

c’s collecting efforts.

12 V. D



ıez D

ıaz et al.

However, for a long time, the precise position of the

locality yielding SZTFH Ob.3091 remained unknown

because the only known account of the discovery and

excavation of these specimens was Kadi

c’s(1916) brief

and very general activity report. Recently, the hand-col-

oured map Kadi

c used in the field was found in the

archive of the SZTFH. This map marks the locations of

his excavation sites in the neighbourhood of V

alioara,

and has allowed their re-identification. Using a combin-

ation of approaches, including field mapping around the

original fossil-bearing sites, sedimentological and tapho-

nomic investigations, and rare earth element (REE) geo-

chemistry, coupled with the brief details on the sites

presented in Kadi

c(1916), Botfalvai et al. (2021) were

able to confidently identify the locality that yielded

SZTFH Ob.3091 as Locality I of Kadi

c(1916). This

locality occurs in the terminal part of the P^

ar^

aul

artopilor ravine, a left-side tributary of the R

achitova

Valley, south-west of V

alioara and near Ciula Mic

(Fig. 1E). Based on data from Csiki-Sava et al. (2016)

and Botfalvai et al. (2021), this locality can be placed in

the lowermost part of the ‘middle member’of the

Densus

,-Ciula Formation, and, as such, is most probably

of earliest Maastrichtian age (Fig. 8).

Other lines of evidence corroborate and extend the

above conclusions. Kadi

c(1916, p. 575) wrote with

respect to his Locality I: “…the matrix yielding the

bones was a red-coloured, sericite-rich mudstone rich in

carbonate concretions. The bones we found here are

thus reddish in colour, with a perfect preservation state-

…” (translated from Hungarian by ZCs-S). Based on

the taphonomic conditions, state of preservation, and the

REE spectra of the specimens in the SZTFH collections,

Botfalvai et al. (2021) suggested that the locality

Figure 6. Titanosaur remains used to erect Titanosaurus dacus by Nopcsa (1915). Anterior caudal vertebra SZTFH Ob.3091 in A,

right lateral; B, ventral; C, posterior; D, right lateral; E, ventral; and F, posterior views. Posterior caudal centrum SZTFH Ob.4215 in

G, left lateral; H, left lateral; I, anterior; K, dorsal; J, posterior; and L, ventral views. Ungual SZTFH Ob.3098 (now lost) in M,

medial; and N, lateral views. A–C, G, M and N are taken from Nopcsa (1915). Scale bars equal 100 mm for A–F, and 50 mm for

G–N.

Revision of Romanian sauropod dinosaurs 13

yielding SZTFH Ob.3091 also produced two other tita-

nosaur specimens from those they sampled geochemi-

cally: SZTFH Ob.3088, a left femur, and SZTFH

Ob.3089, a left humerus. The similar preservation style

and red-speckled yellowish-cream colour of the three

elements imply that they probably share a common

taphonomic history, and their commensurate size and

the spatially restricted nature of the excavations (called

‘nests’by Kadi

c, suggesting the presence of relatively

small lenticular bonebeds [Csiki, Grigorescu, et al.,

2010]) support the idea that they belong to the same

taxon or even the same individual.

There is another set of titanosaur remains in the

SZTFH collections, not included in the set sampled for

geochemistry by Botfalvai et al. (2021), that shares with

the specimens discussed above the same characteristics

in terms of preservation style, colour and size. This set

includes: SZTFH Ob.3100 (left ulna), SZTFH Ob.3101

(left radius), SZTFH Ob. 3096 (metacarpal), SZTFH

Ob.3086 (two left fibulae), SZTFH Ob.3102 (left fib-

ula), and one of the posterior caudal vertebrae included

under SZTFH Ob.3105, corresponding to the specimen

that Nopcsa (1915, pl. III, fig. 7) referred to

Megalosaurus sp. (see Fig. 7A–F). The similar preserva-

tion style of these specimens with that of SZTFH

Ob.3088, 3089 and 3091 strongly indicates that they

were excavated from the same locality as SZTFH

Ob.3091 (i.e. Locality I of Kadi

c, 1916), and –as sug-

gested by their largely commensurate size –that several

of them might belong to the same individual as the lec-

totypic anterior caudal vertebra. That some of these

bones are commensurate in size with the first, geo-

chemically sampled set, is independently supported by

osteohistology: Stein et al. (2010) sampled the humerus

(SZTFH Ob.3089) and the two fibulae accessioned

under SZTFH Ob.3086, and concluded that all these

remains exhibit the same stage of (advanced) ontogen-

etic development (Histological Ontogenetic Stage [HOS]

14). The presence of three left fibulae, representing the

only subset of anatomically overlapping remains, clearly

attests to the presence of at least three individuals.

However, their largely matching size and morphology

(see the ‘Systematic palaeontology’section) is consist-

ent with the assessment that these probably represent

individuals of approximately the same size that belong

to the same titanosaur species. In short, taking all of the

archival, morphological, taphonomic, osteohistological

and geochemical evidence together, we regard it as

highly probable that this entire set of titanosaur remains

came from a single spatiotemporally restricted fossil

site, and that most of the listed specimens belong to a

single disarticulated skeleton. In the following discus-

sions, we designate this set of elements, to which many

(maybe even most) of the titanosaur remains from

Figure 7. Titanosaur posterior caudal vertebrae SZTFH Ob.3105 referred to Megalosaurus sp. by Nopcsa (1915). First specimen in

A, right lateral; B, right lateral; C, ventral; D, posterior; E, anterior; and F, dorsal views. Second specimen in G, ventral; H, ventral;

I, anterior; J, left lateral; K, dorsal; and L, posterior views. A and G are taken from Nopcsa (1915). Scale bars equal 50mm.

14 V. D



ıez D

ıaz et al.

Locality I of Kadi

c belong to, as ‘Assemblage A’

(Fig. 9).

As a final historical note, it is worth pointing out that

some of the Assemblage A specimens were examined

first-hand by Huene during his study of Transylvanian

titanosaurs. These were sent to him at Nopcsa’s personal

instructions, as documented by handwritten notes

present in the SZTFH registry book (Fig. 10). Of these,

one of the fibulae registered as Ob.3086 was referred by

Huene (1932)toM. dacus, together with the lectotypic

caudal vertebra (Ob.3091), whereas two other fibulae

(the second specimen registered under Ob.3086, and

Ob.3102), together with the ulna (Ob.3100), were attri-

buted to the second species of the genus erected by him,

Figure 8. Synthetic stratigraphical distribution of key titanosaur individuals/assemblages (coded A to V; red stars –stratigraphical

location well constrained, pale red stars –stratigraphical position approximate; see text and Fig. 1E for more details), and other

significant titanosaur occurrences (grey stars). Abbreviations for stratigraphical columns: A–D,Hat

,eg Basin; A,S

ınpetru

Formation, Sibis

,el Valley section; B,‘R^

aul Mare Beds’;C,‘Pui Beds’;D, Densus

,-Ciula Formation; E, Rusca Montan

a Basin; F, G,

south-western Transylvanian Basin; F, left side of Mures

,Valley; G, right side of Mures

,Valley. Abbreviations for titanosaur

localities, other than key individuals/assemblages: Ba,B



ar

abant

,;Bo, Boit

,a; Cd, Ciugud; D, Densus

,;FD,F



arc

adeana; Li, Livezi;

Lc, Lancr

am; NV,N



alat

,-Vad, several sites; Od, Oarda de Jos; Pu, Pui, B

arbat River, several sites; RR, Sebes

,,R

apa Ros

,ie; RL,

apa Lancr

amului; SbG, Sebes

,-Glod; SC,S

anpetru, C

arare; Sd,S

,ard; SFH, Secas

,, Fetilor Hill; Sp,S

anpetru, several sites; SS,

anpetru, Scoab

a; Tc, Teleac; To, Totes

,ti; VF,V



alioara, F^

ant^

anele 1. Updated from Csiki-Sava et al. (2016, fig. 14) and Botfalvai

et al. (2021, fig. 12).

Revision of Romanian sauropod dinosaurs 15

M. transsylvanicus. It seems that Huene was unaware

that: (1) all these specimens most probably came from

the same fossiliferous site, and that some were possibly

from the same individual; and (2) further titanosaur

remains existed in the SZTFH collections, and that

some of these originated from the same Kadi

c Locality

I as the lectotype and thus potentially also belonged to

M. dacus. The selective way Nopcsa sent specimens

from Budapest to T€

ubingen is difficult to explain (see

also below), but it almost certainly affected Huene’s

opinions concerning the taxonomy and anatomy of the

Transylvanian titanosaurs.

The ‘Magyarosaurus’hungaricus type locality,

anpetru –Individual B

Among the material excavated by Nopcsa before 1906

in the Sibis

,el Valley, south of S^

anpetru (Fig. 1E), and

subsequently purchased by the NHMUK, there is an

associated and relatively large left tibia and fibula regis-

tered under the same accession number, NHMUK

R.3853. The fibula was described by Huene (1932,p.

269, pl. 47, fig. 1) as the holotype (and only specimen)

of his new species ?Magyarosaurus hungaricus, which

he differentiated from M. dacus and M. transsylvanicus

primarily based on the larger size and distinctive morph-

ology of the fibula. Furthermore, Huene (1932, p. 274,

pl. 48, fig. 3) suggested that the tibia bearing the same

specimen number as the holotype fibula could also

belong to the same species, based on its size, although

he referred to this tibia as Magyarosaurus sp. in the

text. First-hand examination of the fibula and tibia con-

firms that they are commensurate in size, have a similar

preservation style, and share an unusual interlocking

mechanism at their distal ends (see the ‘Systematic

palaeontology’section). These observations support

Huene’s assessment of potential conspecificity of the

two specimens. Furthermore, in our view, these features

also strongly suggest that these two specimens belonged

not only to the same species, but were excavated from

the same locality, and represent associated distal hind

limb elements of the same individual. Such an associ-

ation would also be consistent with these elements being

registered under the same NHMUK specimen number at

the time of their purchase in 1906. We therefore desig-

nate this material as Individual B (Fig. 11).

Unfortunately, almost nothing is known about the

locations, stratigraphy or fossil content of the fossilifer-

ous sites excavated by Nopcsa in the Sibis

,el Valley, and

field details are unavailable for most. This includes the

locality that yielded NHMUK R.3853, so it can only be

constrained as belonging to the type section of the

ınpetru Formation near S^

anpetru (Fig. 1E), and is thus

most probably of early to early late Maastrichtian age

(Csiki-Sava et al., 2016)(Fig. 8).

Kadi

c Locality VI (P^

ar^

aul Budurone), V

alioara

–Individual C

Kadi

c(1916, pp. 575–576) reported his Locality VI as a

‘nest’(Csiki, Grigorescu, et al., 2010) found in “fine-

grained bluish loose sandstones from where we have

excavated thick limb-bone fragments and exquisitely

Figure 9. Fossil remains belonging to Assemblage A, Kadi

locality I, the Magyarosaurus dacus type locality, Ciula Mic

Not to scale. Skeletal drawing after Scott Hartman, used

with permission.

Figure 10. Handwritten notes in the registry book of the SZTFH. A, indication that SZTFH Ob.3090 comprises eight ‘matching (¼

associated) caudal vertebrae’(‘€

osszeill}

o farokcsigoly

ak’in Hungarian). B, indication that caudal vertebrae SZTFH Ob.3090D and G

were sent to Huene in T€

ubingen (‘Elk€

uldve Huenenek, 2 drb.’in Hungarian).

16 V. D



ıez D

ıaz et al.

preserved large vertebrae. The colour of the bones is

here, too, black”(translated from Hungarian by ZCs-S).

As already noted by Csiki and Grigorescu (2006; see

also Csiki et al., 2011), and subsequently by Botfalvai

et al. (2021), the only set of bones in the SZTFH collec-

tions that matches this description is represented by a

series of large titanosaur remains comprising several

caudal centra (SZTFH Ob.3090), an incomplete right

humerus (SZTFH Ob.3104), fragmentary left and right

femora (SZTFH Ob.3103), and a metatarsal (SZTFH

Ob.3095). These specimens are all much larger than

most of the titanosaur remains from Transylvania, and

share a consistent preservation style with one another,

including their black colour with a shiny, smooth perios-

teum (except where it is originally fibrous in aspect).

Furthermore, handwritten notes in the registry book

of the SZTFH state that SZTFH Ob.3090 comprises

eight “matching [¼associated] caudal vertebrae”

(“€

osszeill}

o farokcsigoly

ak”in Hungarian in original; Fig.

10A), whereas SZTFH Ob.3103 consists of the “right

and left femur from a gigantic individual”(“Jobb 

es bal

combcsont 

ori

as p

eld

anyt

ol”in Hungarian in original;

Fig. 12A), with SZTFH Ob.3104 representing the

“humerus of the same (gigantic) individual”(“Humerus

ugyanatt

ol a p

eld.t

ol”in Hungarian in original; Fig.

12A). These notes were added when the specimens were

registered in 1914, i.e. approximately when they were

collected, and clearly indicate that this set of specimens

was identified as originating from the same skeleton.

The series of eight associated (and possibly articulated)

caudal vertebrae was initially included under the speci-

men number SZTFH Ob.3090. However, when one of

us (ZCs-S) first visited the SZTFH collections in 2003,

only four of these vertebrae could be located. Besides

the specimen numbers, the four remaining vertebrae are

also marked with poorly preserved letters written in ink:

B, D, G and H. Based on the morphology of the indi-

vidual elements (see description in the ‘Systematic

palaeontology’section), the vertebrae marked with B

and D occupied more anterior positions in the tail than

vertebrae G and H, and B appears to be somewhat more

anterior in position than D. Taken together, these obser-

vations suggest that the original eight vertebrae were let-

tered sequentially according to their position in the tail,

with A, C, E and F now missing. In 1927, two of these

vertebrae (D and G) were sent to Huene in T€

ubingen: a

handwritten note in the SZTFH register (Fig. 10B –

second line) reads “Sent to Huene, 2 pieces”(“Elk€

uldve

Huenenek, 2 drb.”in Hungarian in original), following

up a similar note (Fig. 10B –first line) for SZTFH

Ob.3089 dated 12 January 1927 which states that “By

Figure 11. Fossil remains belonging to Individual B, the

Petrustitan (‘Magyarosaurus’)hungaricus n. gen. type locality,

anpetru. Skeletal drawing after Scott Hartman, used

with permission.

Figure 12. A, handwritten notes in the registry book of the SZTFH on the titanosaurian remains referred to Individual C. B, fossil

remains referred to Individual C, the Uriash kadici n. gen. n. sp. type locality, Kadi

c Locality VI (P^

ar^

aul Budurone), V

alioara.

Skeletal drawing after Scott Hartman, used with permission.

Revision of Romanian sauropod dinosaurs 17

instruction of the Director [i.e. Nopcsa at that time], 1

piece sent to Huene to T€

ubingen”(“Igazgat

oi rendeletre

1 drb. elk€

uldve Huenenek T€

ubingenbe”, in Hungarian in

original). Huene figured and briefly described these two

vertebrae (Huene, 1932, p. 271–272, 274, pl. 47, figs 4,

5), referring at least one of them tentatively to

?Magyarosaurus hungaricus. However, it appears that

Huene was not aware that these two vertebrae belong to

a more complete vertebral series, nor that there were

appendicular elements also referable to this individual.

The right humerus (SZTFH Ob.3104) has a compli-

cated curatorial history. After being registered as a

humerus belonging to the same individual as the femora

SZTFH Ob.3103 (Fig. 12A), SZTFH Ob.3104 is men-

tioned as lost in the SZTFH registry book in 1938. In

2003, however, one of us (ZCs-S) located a small frag-

ment of the humeral midshaft that still bore the original

specimen number, in the same drawer as the SZTFH

Ob.3103 femora. A much larger portion of this humerus,

including the region of the deltopectoral crest, had been

re-registered as v.13491 in the second half of the twenti-

eth century. When examined in 2003, the latter element

was found to be of commensurate size, as well as hav-

ing a matching breakage surface and an identical preser-

vation style, with the smaller fragment still bearing the

original specimen number (SZTFH Ob.3104). The two

large and black-coloured humeral fragments thus defini-

tively belong to the same element, and were recon-

structed as such in 2003. Le Loeuff, (2005a, fig. 1A)

figured this specimen already reconstructed, missing

only a small fragment of the proximolateral end (which

was also relocated and restored in 2004), although he

still referred to it using the new registry number borne

by the larger fragment (v.13491) and not the original

specimen number (Ob.3014). However, as with Huene

(1932), Le Loeuff (2005a) was apparently not aware

that this humerus belongs to a large titanosaur individual

that is represented by further skeletal elements (most

probably because he used the new registry

number). Botfalvai et al. (2021) geochemically sampled

three specimens from this set of bones (two caudal ver-

tebrae from SZTFH Ob.3090, and the right femur,

SZTFH Ob.3103). Their analysis showed a remarkable

similarity between the three elements in their REE spec-

tra, lending further support to the idea that they origi-

nated from the same fossiliferous site.

Combining the above evidence, most of which comes

from notes and observations recorded around the time

of their discovery, it is highly probable that the set of

large black titanosaur bones documented above represent

elements of a single partial skeleton, herein identified as

Individual C (Fig. 12B). These specimens were exca-

vated from Locality VI of Kadi

c(1916), which was

recently reidentified by Botfalvai et al. (2021)ina

ravine joining the left side of V

alioara Valley, just

south-west of V

alioara village (Fig. 1E). Based on its

geographical position, the site can be placed in the

lower part of the ‘middle member’of the Densus

,-Ciula

Formation, although stratigraphically somewhat higher

than Locality I discussed above (Fig. 8). Csiki-Sava

et al. (2016) and Botfalvai et al. (2021) suggested that

the age of the locality is early Maastrichtian.

The remains belonging to Individual C have been

evaluated in several previous studies (e.g. Le Loeuff,

2005a; Stein et al., 2010; see ‘Dwarfism’section),

although mostly without noting either their skeletal asso-

ciation or site of origin. As noted above, two elements

of the caudal vertebral series SZTFH Ob.3090 were sent

to Huene in T€

ubingen by Nopcsa. Huene (1932, pp.

271–274, pl. 47, figs. 4, 5) identified these vertebrae as

Magyarosaurus sp., although he noted that they might

belong to ?M. hungaricus (a referral also emphasized in

the corresponding plate captions). The associated nature

of Individual C was first recognized and discussed by

Csiki and Grigorescu (2006). Following Huene’s argu-

ments and accepting his view of one large titanosaur

taxon present in the Transylvanian area, these authors

maintained the referral of these remains to ?M.

hungaricus.

Locality NV10, N

alat

,-Vad (5Paludititan

nalatzensis type locality) –Individual D

Unlike the historical titanosaur specimens from S^

anpetru

and V

alioara discussed above, the partial skeleton UBB

NVM 1 has well-documented provenance and skeletal

association data, and was established by Csiki, Codrea,

et al. (2010) as the holotype of the second titanosaur

genus described from the Transylvanian area –

Paludititan nalatzensis. It preserves parts of the dorsal

and caudal vertebral series, as well as most of the pel-

vis, together with a large number of rib fragments, sev-

eral chevrons, a small fragment of the femur, and two

pedal unguals (see the ‘Systematic palaeontology’sec-

tion). It was excavated in 2002 during a joint

Romanian–Belgian field season. One of the current

authors (ZCs-S) was actively involved in the excavation

process, and recorded the position of the different skel-

etal elements recovered. Numerous aspects support the

interpretation of one single disarticulated partial titano-

saur skeleton at this locality, which we designate as

Individual D (Fig. 13). These include: the progress of

the excavation itself; the resultant quarry map showing

the different skeletal parts in their disarticulated, but

quasi-in vivo position (see Csiki, Codrea, et al., 2010,

fig. 1B); the presence of in situ skeletal articulation in

the case of the pelvic girdle and a series of caudal

18 V. D



ıez D

ıaz et al.

vertebrae; the commensurate size and non-duplicated

nature of the elements; and the consistent preservation

style.

Besides specimen association data, there are also high-

resolution geographical and stratigraphical data available

for Locality NV 10 (N

alat

,-Vad) and thus for Individual

D. This locality belongs to the ’R^

aul Mare Beds’, a not

yet formally defined lithostratigraphical unit that either

represents the uppermost part of the S^

ınpetru Formation

(being slightly younger than most of the type section

located along the Sibis

,el Valley), or is a distinct unit (see

review in Csiki-Sava et al., 2016). The relatively exten-

sive outcrop at N

alat

,-Vad, lying in the lower part of the

aul Mare Beds, was mapped carefully in 2001–2002

(see Smith et al., 2002; van Itterbeeck et al., 2004), and

the position of the site yielding UBB NVM 1 is also

recorded with precision, in the upper part of the locally

outcropping section (Fig. 1E). Furthermore, preliminary

palynostratigraphical (van Itterbeeck et al., 2005) and

magnetostratigraphical (Panaiotu et al., 2011) data are

also available for the N

alat

,-Vad succession, suggesting

that the age of Locality NV 10 falls most probably around

the early/late Maastrichtian boundary or in the earliest

late Maastrichtian (Fig. 8). Based on the overall distribu-

tion pattern of the vertebrate fossil record of the Hat

,eg

Basin (e.g. Csiki-Sava et al., 2016; Therrien et al., 2009),

Locality NV 10 is most probably somewhat stratigraphic-

ally younger in age than the ?Magyarosaurus hungaricus

type locality, and definitely younger than localities I and

VI of Kadi

c, which makes Individual D stratigraphically

younger than Individuals/Assemblages A, B and C

(Fig. 8).

Mamulea locality, Ciula Mic

a–Individual E

As already noted, after Nopcsa ceased his collecting

activity at S^

anpetru in the 1920s, Mamulea was the first

to report the discovery of dinosaur remains from the

uppermost Cretaceous of the Hat

,eg Basin (Mamulea,

1953a, p. 250). No further details were given concerning

this discovery; however, in another report, Mamulea,

(1953b) added that the dinosaur remains he had discov-

ered at Ciula Mic

a were found in purplish sandstones

within the coarse, red, conglomeratic continental succes-

sion that, in his view, overlay a lower, dominantly grey-

ish ‘fluvio-lacustrine’succession. Because he considered

that this upper red unit (which he reported to be other-

wise palaeontologically barren) represented the

Paleocene, contrary to previous opinions expressed by

Nopcsa (1905), Schafarzik (1909,1910), Kadi

c(1916),

and Laufer (1925), Mamulea (1953b) also regarded the

dinosaur bones he found there as being reworked.

However, subsequent work in the Ciula Mic

a-V

alioara

area, from which these remains were collected, firmly

established that the red siliciclastic deposits are indeed

fossiliferous and are from the lower part of the ‘middle

member’of the uppermost Cretaceous Densus

,-Ciula

Formation (e.g. Antonescu et al., 1983; Bojar et al.,

2011; Botfalvai et al., 2021; Csiki-Sava et al., 2016;

Grigorescu, 1992). Thus, the titanosaur remains reported

by Mamulea must have originated from these lowermost

Maastrichtian deposits (Csiki-Sava et al., 2016)(Fig. 8).

More precise pinpointing of the geographical location of

this site is not currently possible, beyond it being in the

vicinity of Ciula Mic

a(Fig. 1E).

The titanosaur remains discovered by Mamulea at

Ciula Mic

a, and reported without any other details,

description, or illustrations, were re-identified in the col-

lections of the FGGUB. These consist of a set of post-

cranial remains that include incomplete dorsal and

caudal vertebrae, a chevron, the left humerus and radius,

several incomplete metacarpals, the left femur together

with corresponding proximal tibia-fibula, and the right

femur (see complete list of elements in the ‘Systematic

palaeontology’section).

The listed elements are commensurate in size, and

there is no duplication within the set. Their preservation

style is consistent, pointing to a common taphonomic

history, and all display a greyish-green-spotted yellow-

ish-reddish colour, suggesting that they come from a

somewhat reductive, variegated and relatively fine-

grained sediment (probably a fine sandstone). At least

some of the specimens are virtually complete and well

preserved (e.g. Fig. 4; see also the ‘Systematic palaeon-

tology’section), and others preserve their original,

in vivo skeletal articulation (i.e. dorsal centra, caudal

centra, and proximal tibia-fibula). These features argue

strongly against potential reworking or long-distance

transport being involved in the formation of the tapho-

coenosis corresponding to the set, and thus an alloch-

thonous origin, as advocated by Mamulea (1953b), can

be safely discarded. Finally, Stein et al. (2010) sampled

Figure 13. Fossil remains belonging to Individual D, the

Paludititan nalatzensis type locality, locality NV10, N

alat

,-Vad.

Skeletal drawing after Scott Hartman, used with permission.

Revision of Romanian sauropod dinosaurs 19

the histology of two specimens from this set (the left

humerus LPB [FGGUB] R.1047, and the left femur

R.1046), and found that their HOS values were very

similar (13 and 14, respectively; see the ‘Dwarfism’sec-

tion). Although circumstantial, this result is consistent

with these two elements belonging to a single

individual.

Based on all of this information, the set of titanosaur

elements from the FGGUB collections listed above can

be regarded as originating from one fossiliferous site,

and almost certainly represents the associated remains

of a single partial skeleton. We identify this as

Individual E (also termed the ‘Mamulea specimen’;

Fig. 14).

MD locality, Pui –Assemblage F

This locality was discovered in 2015, when a partial

titanosaur cervical vertebra (LPB [FGGUB] R.2505)

was found and excavated. Subsequent prospecting in

2016 yielded several other titanosaur remains, including

a well-preserved humerus (LPB [FGGUB] R.2506), a

metacarpal (LPB [FGGUB] R.2509), two femora (one

complete –LPB [FGGUB] R.2507; one partly complete

–R.2508), and a chevron (LPB [FGGUB] R.2510). At

that time, no other vertebrate remains had been found at

this site, although in 2019 it also yielded remains refer-

able to ornithischian dinosaurs. All the titanosaur speci-

mens were recovered from a small area and from the

same fossiliferous level, in a dominantly red-coloured,

fine to medium-grained silty sandstone, with local

greenish and sandier intercalations (ZCs-S pers. obs.).

Most of these remains are approximately commensurate

in size, and they share a similar whitish-reddish yellow

colour and preservation style, suggesting that they prob-

ably originate from one disarticulated partial skeleton.

One exception is represented by the less complete

femur: in addition to being less well preserved, it is

slightly larger than the other femur and comes from the

same side, indicating that a second, slightly larger tita-

nosaur individual is also represented at the MD site.

Nevertheless, this larger specimen is morphologically

similar to the other femur, and most probably they

belong to the same species. Thus, although we consider

it probable that all of these specimens, apart from the

larger femur, belong to the same individual, we conser-

vatively designate the titanosaur remains from the MD

locality as Assemblage F (Fig. 15).

The MD locality is situated in the lower part of the

uppermost Cretaceous continental succession cropping

out along the B

arbat River (Fig. 1E), a succession con-

sidered to be either part of the S

ınpetru Formation (e.g.

Grigorescu, 1992; Mamulea, 1953a), or a different, but

approximately time-correlative, unit (Csiki-Sava et al.,

2016; Therrien, 2005). A palynological sample collected

from the middle part of the succession, from a strati-

graphically higher level than that hosting site MD, was

dated as close to the early/late Maastrichtian boundary

(van Itterbeeck et al., 2005); as such, the MD locality

(and thus, Assemblage F) should be considered to be of

early Maastrichtian age (Fig. 8).

apa Mocioconilor locality, S^

anpetru –

Individual G

Renewed fossil vertebrate collecting in the Hat

,eg Basin

in the late 1970s led to the discovery of several impor-

tant vertebrate accumulations (‘fossiliferous pockets’or

small-sized lenticular macrovertebrate bonebeds; Csiki,

Grigorescu, et al., 2010) along the Sibis

,el Valley, near

anpetru (e.g. Grigorescu, 1983; Groza, 1983). One of

these accumulations is represented by the R^

apa

Mocioconilor (¼Mocioconilor ravine) site, situated on

the right side of the valley, between the villages of

anpetru and S

acel, and excavated by teams of the

Muzeul Civilizat

,iei Dacice s

,i Romane, Deva, between

1979 and 1982. The location of this site is provided in

Groza (1983, fig. 1; see also Csiki, Grigorescu, et al.,

Figure 14. Fossil remains belonging to Individual E, referred

here to Magyarosaurus dacus, Mamulea locality, Ciula Mic

Skeletal drawing after Scott Hartman, used with permission.

Figure 15. Fossil remains belonging to Assemblage F, referred

here to Magyarosaurus dacus, MD locality, Pui. Skeletal

drawing after Scott Hartman, used with permission.

20 V. D



ıez D

ıaz et al.

2010, fig. 1C), and, according to its stratigraphical posi-

tion, it should be placed in the lower part of the domin-

antly reddish-brown informal lower subunit of the

ınpetru Formation type section, which crops out along

the Sibis

,el Valley (Grigorescu, 1992; Therrien et al.,

2009)(Fig. 1E). This constrains its age most probably

to the early, or, at most, earliest late, Maastrichtian

(Csiki-Sava et al., 2016)(Fig. 8).

According to the brief, preliminary account of the

excavations by Groza (1983, p. 52, pl. 2, figs 1–6), this

site yielded several specimens in 1980; these included

six titanosaur posterior caudal vertebrae (MCDRD 255,

266, 267, 268, 269 [one of these six vertebrae could not

be located in the collections]), four of which were found

in articulation. Groza (1983) referred these specimens to

T. dacus. He also reported the recovery of an associated

pair of limb elements from the same site, collected in

1979. These elements (MCDRD 149 and 150) were

identified as a tibia and fibula of the rhabdodontid orni-

thopod Zalmoxes robustus (referred to as Rhabdodon

priscum by Groza, 1983, pp. 53–54, pl. 4, figs 4, 5).

However, examination of the published illustrations and

first-hand study of MCDRD 149 and 150 demonstrates

their sauropod affinities, and that they represent a

matching pair of radius and ulna instead.

Although not recognized as such by Groza (1983)at

the time of his report’s publication, further titanosaur

remains were also discovered at the R^

apa Mocioconilor

locality in the same excavations, including a metacarpal,

a partial ilium, and a fibula (see the complete list of

specimens in the ‘Systematic palaeontology’section).

These fossils match the previously listed elements well

in terms of their relative size and preservation style.

Moreover, I. Groza identified these bones as being

found together with the specimens discussed in his

report (personal communication in the early 1990s to

C.-M. Jianu, at that time curator of the palaeontology

collections of the MCDRD, and through her to one of

the authors [ZCs-S]), and were doubtlessly considered

by him to be associated with the latter. Based on these

data, we regard these elements from the MCDRD col-

lections originating from the R^

apa Mocioconilor site as

being conspecific, and most probably parts of the same

partial skeleton, which we designate as Individual G

(Fig. 16).

Green Four locality, S^

anpetru –Individual H

In the NHMUK collections there is a set of skeletal ele-

ments (inventoried under NHMUK R.4891) that is

accompanied by a handwritten label from the time of its

registration; this states that it contains “associated bones

of Titanosaurus, Szentp

eterfalva (all bones of this indi-

vidual marked ‘green 4’)”(Fig. 17A). This represents

the best and most detailed record of an original skeletal

association among the sauropod material excavated by

Nopcsa at S^

anpetru. This is also corroborated by notes

in the Nopcsa repository in the NHMUK archives, dated

1923, that lists the Transylvanian specimens Nopcsa

offered for purchase to the museum. These notes record

the “associated remains of one individual of

Titanosaurus, ribs, vertebrae, parts of pelvis, including

ilium, ?scapulae, 3 metacarpals, found in contact [ …],

Szentp

eterfalva”. As in the case of most of the other

assemblages discussed above, elements of this set

(including dorsal vertebrae, ribs, sternal plate, scapula,

coracoid, and three articulated metacarpals) are com-

mensurate in size, lack duplicated elements, and show a

similar preservation style, including the presence of a

light grey, fine-grained sandy-silty matrix that is still

adhered to some of the specimens. Based on these lines

of evidence, the remains registered under NHMUK

R.4891 can be considered as originating from one single

Figure 16. Fossil remains belonging to Individual G, referred

here to Petrustitan hungaricus n. gen., R^

apa Mocioconilor

locality, S^

anpetru. Skeletal drawing after Scott Hartman, used

with permission.

Figure 17. A, handwritten label from the NHMUK

collections, indicating that it contains ‘associated bones of

Titanosaurus, Szentp

eterfalva (elements under NHMUK

R.4891, all bones of this individual marked “green 4”)’.B,

fossil remains belonging to Individual H, Lithostrotia incertae

sedis, Green Four locality, S^

anpetru. Skeletal drawing after

Scott Hartman, used with permission.

Revision of Romanian sauropod dinosaurs 21

partial skeleton, identified here as Individual H (or the

‘green 4’individual; Fig. 17B).

Unfortunately, little is known about the provenance of

this material, except that it was excavated from the

ınpetru Formation, outcropping in the S^

anpetru area

(Fig. 1E). Because no more precise locality data are

available, its age can be constrained only as

Maastrichtian, following the review by Csiki-Sava et al.

(2016)(Fig. 8).

Several of the elements belonging to Individual H

were described and figured by Huene (1932). It is strik-

ing, however, that he decided to split up this material,

referring the articulated metacarpus to Magyarosaurus

transsylvanicus (Huene, 1932, pp. 268–269, pl. 46, fig.

10). By contrast, he regarded the dorsal ribs and pec-

toral girdle elements as belonging to an indeterminate

titanosaur (Huene, 1932, pp. 270–272), although he

explicitly emphasized that they probably originated from

one and the same large individual: “The left 2nd dorsal

rib …obviously of the same individual”and “… the

piece carries the same number as the first two large dor-

sal ribs of the same site, it is probably found together

with them and perhaps belongs to the same individual”

(“Die linke 2. Dorsalrippe …offenbar des gleichen

Individuums“and “… das St€

uck die gleiche Nummer

tr€

agt wie die beiden ersten großen Dorsalrippen glei-

chen Fundort, ist es wahrscheinlich mit ihnen zusammen

gefunden und geh€

ort vielleicht zum gleichen

Individuum”in German in the original). Given his usage

of the name ?Magyarosaurus hungaricus for the large-

sized taxon (and thus, any large-sized titanosaur

remains) from Transylvania, Huene (1932, p. 274) sug-

gested that the dorsal ribs and the pectoral girdle ele-

ments could conceivably belong to this species: “It is

also probably that these remains belong to the same spe-

cies as the left fibula R.3853 …, which would be

Magyarosaurus hungaricus”(“Es d€

urfte auch wahr-

scheinlich sein, daß diese Reste zur gleiche Art geh€

oren

wie die linke Fibula R.3853 …,das w€

are

Magyarosaurus hungaricus”in German in the original).

This taxonomic opinion was reiterated in the plate cap-

tions of the respective figures (Huene, 1932, pl. 47, figs

9–11). Huene’s(1932) decision to consider the remains

accessioned under NHMUK R.4891 as belonging to

more than one species is surprising, especially given

that the evidence of association was available to him,

and that he apparently studied these materials first-hand.

Tus

,tea titanosaur, Tus

,tea-Oltoane nesting site –

Individual I

The Tus

,tea-Oltoane dinosaur nesting site is the first of

its kind to be discovered in the uppermost Cretaceous of

the Transylvanian area (e.g. Grigorescu et al., 1990,

2010)(Fig. 1E). It has yielded eggs and nests assigned

to Megaloolithidae, an ootaxon that is usually consid-

ered to have been laid by titanosaurs (e.g. Horner, 2000;

Sander, 2008; Vila et al., 2010), as supported by the dis-

covery of embryonic remains associated with such eggs

in Argentina (Chiappe et al., 1998,2001). Accordingly,

the eggs from Tus

,tea were also referred initially to

Magyarosaurus (Grigorescu et al., 1990). However, they

were subsequently found to be associated with hatchling

remains of hadrosauroids (Grigorescu et al., 1994), and

have therefore been considered to belong to this group

instead (Botfalvai et al., 2017; Grigorescu, 2010,2017;

Grigorescu et al., 1994,2010). Further support for this

interpretation came from the total absence of sauropod

remains from this site, despite years of careful excava-

tions, a pattern that strikingly departs from the wide-

spread and abundant occurrences of titanosaur remains

in the Hat

,eg Basin (Csiki, Grigorescu, et al., 2010).

However, this situation changed in 2008, when caudal

vertebrae, chevrons, and an almost complete sacrum

with a co-ossified partial ilium of a titanosaur were dis-

covered at this locality (Csiki-Sava et al., 2012). All

these elements were found at approximately the same

stratigraphical level within the site, and were closely

associated. The tight clustering of these remains, com-

bined with the absence of other titanosaur specimens

from elsewhere within the nesting site, suggests that

they probably represent elements from one individual,

herein identified as Individual I (Fig. 18). Such a con-

clusion is also consistent with the matching size and

similar overall preservation of the specimens (Botfalvai

et al., 2017). This locality is situated in the middle–

upper part of the ‘middle member’of the Densus

,-Ciula

Formation (Fig. 1E), and thus the stratigraphical age of

Individual I is probably ‘middle’(late early to early

late) Maastrichtian (Csiki-Sava et al., 2016)(Fig. 8).

Figure 18. Fossil remains referred to Individual I, Lithostrotia

incertae sedis, Tus

,tea titanosaur, Tus

,tea-Oltoane nesting site.

Skeletal drawing (reversed) after Scott Hartman, used

with permission.

22 V. D



ıez D

ıaz et al.

The Groap

a titanosaur assemblage, S^

anpetru –

Assemblage J

The Groap

a locality is one of the major sites excavated

in the Sibis

,el Valley after 1977 (Csiki, 2005; Csiki,

Grigorescu, et al., 2010; Groza, 1983, fig. 1C).

Researchers from both the FGGUB and MCDRD have

worked at this site intermittently over the years, but

unfortunately there is no record of the exact position of

the skeletal elements recovered from here, except for

the period between 1992 and 1994, when more system-

atic excavations took place. Since 1995, when the

quarry was considered largely depleted and was aban-

doned, it has been revisited occasionally to uncover fos-

sils that were still eroding out, but this activity has had

little success. Accordingly, taphonomic information such

as position in the site, orientation, and skeletal associ-

ation, is scant or completely absent for most of the

specimens excavated from here. Even the complete list

of specimens originating from the Groap

a locality is dif-

ficult to reconstruct, because in many instances these

were labelled under different locality names in the two

participating institutions, when they were labelled at all.

Nevertheless, it is clear that the locality was relatively

richly fossiliferous and that more than 50% of the iden-

tifiable macrovertebrate remains discovered here belong

to titanosaurs (Csiki, 2005; Csiki, Grigorescu, et al.,

2010). Based on the number of overlapping skeletal ele-

ments and variation in size, at least seven titanosaur

individuals are represented at this site. These specimens

appear to share a common taphonomic history, although

there is some variation; for example, most of the titano-

saur fossils are isolated and often fragmentary, whereas

a few remains are articulated, including several series of

dorsal and (mainly) caudal vertebrae, as well as an

articulated tibia-fibula and associated partial pes. The

duplicated skeletal elements (humeri, femora, fibulae)

do not appear to represent distinct morphotypes, and do

not differ widely in overall size, suggesting that all these

individuals probably represent the same taxon and a

similar ontogenetic stage. The latter received support

from the osteohistological study of Stein et al. (2010),

who demonstrated that, despite their relatively small

size, all elements sampled from this site show an

advanced HOS (12–14) and thus were attributed to M.

dacus. Given that it is not possible to assign the recov-

ered elements to distinct individuals, we designate the

Groap

a locality titanosaur remains as Assemblage J

(Fig. 19).

The Groap

a locality is probably situated in the middle

part of the informal lower subunit of the S^

ınpetru

Formation type section (Fig. 1E; see also Csiki,

Grigorescu, et al., 2010, fig. 1C); as such, Assemblage J

is most probably early to ‘middle’Maastrichtian in age

(Csiki-Sava et al., 2016). This also means that

Assemblage J is stratigraphically somewhat younger

than Individual G from the R^

apa Mocioconilor site

(Fig. 8).

Unfortunately, although several titanosaur specimens

from the Groap

a Quarry are fairly complete, their over-

all state of preservation is often poor. Both pre-burial

and post-burial fracturing is common, and the perios-

teum of the bones is often missing entirely, exposing

the cancellous inner bone texture. These observations

are indicative of a relatively long sub-aerial exposure

period in their taphonomic history, during which, weath-

ering and environmental agents that promote disarticula-

tion, dispersal, breakage and splintering were active.

Furthermore, most of the titanosaur remains are pre-

served in a very hard, silica-cemented, fine, silty grey

matrix that covers the bones in a concretionary manner

and invades their incomplete and porous epiphyses.

These factors make preparation very difficult and time-

consuming to the extent of being almost prohibitive.

Because of these unfortunate circumstances, the morph-

ology of Assemblage J is less well known and docu-

mented than its skeletal representation would suggest.

Other less inclusive and/or less reliably supported

skeletal associations. In addition to the localities and

skeletal associations discussed above, there are several

other instances where skeletal association of a number

of titanosaur fossils from Transylvania can be ascer-

tained with some degree of confidence. In many of these

cases, however, skeletal association is restricted mainly

to one region of the body, thus providing a more limited

basis for comparison with the existing nominal taxa. In

other cases, inferences of skeletal association rely on

circumstantial evidence to a much greater degree.

Arguments to support these associations are sometimes

based on excavation records and/or registry or other

written information, but more commonly rest on similar

taphonomic features and commensurate size. Below, we

Figure 19. Fossil remains referred to Assemblage J,

Lithostrotia incertae sedis, Groap

a titanosaur assemblage,

anpetru. Skeletal drawing after Scott Hartman, used

with permission.

Revision of Romanian sauropod dinosaurs 23

discuss such assemblages, and briefly review the data

supporting their potential skeletal association (for a

more detailed description of these remains see the

Supplemental Material).

ar^

aul Neagului titanosaur, V

alioara –

Assemblage K

As noted below (see Individual N), a left humerus

accessioned under NHMUK R.4882 is part of a set of

“four imperfect limb bones”presented by Lady

Woodward in 1923 to the NHMUK, and was sup-

posedly collected from S^

anpetru (Nopcsa archive list

22e, dated May 1923). However, it seems unlikely that

this humerus came from the same locality as the other

three specimens. Firstly, the humerus shows a clearly

different preservation style compared to the other three

specimens, suggesting that they probably came from dif-

ferent localities. Secondly, one of the identifying fea-

tures of the humerus is a small quadrangular paper

sticker glued to it, with the number “26”written in ink.

As far as we are aware, there are only two other speci-

mens from the entire Transylvanian titanosaur collection

in the NHMUK that show this particular type of label-

ling: a partial left femur, NHMUK R.11126 (bearing

“28”), and a large scapular blade (NHMUK R.11144,

bearing “13”). No collecting data are available for

NHMUK R.11126, but the scapular blade is registered

as being part of the Nopcsa collection purchased

through Count Szapary in 1924. Moreover, besides the

small sticker, the scapula also has a larger paper label

attached to it that identifies its place of origin as

Valiora (¼V

alioara), and more specifically as the

“Pareu Niagului, right-side branch, second collecting

point”(“Pareu Niagului jobb oldali mell

ek

ag, 2

gy€

ujt}

opont.”, in Hungarian in original), which corre-

sponds perfectly with the details of one of the main

excavation sites (II) of Kadi

c in the V

alioara area

(Kadi

c, 1916). It is the only titanosaur material from the

entire NHMUK collection that is known to come from

this locality, and also one of only a few for which such

written confirmation of its provenance is available. It is

also worth noting that a large sample of vertebrate

remains, showing the same type of preservation and

dark brown to black colour, as well as the same type of

small quadrangular paper labels with numerals written

in ink, exists in the collections of the SZTFH. The prov-

enance of the latter collection can be definitively identi-

fied as the Kadi

c excavations from V

alioara.

Based on the above observations, it seems reasonable

to assume that the three NHMUK elements with these

quadrangular paper labels also originated from the fos-

siliferous localities excavated by Kadi

c along the P^

ar^

aul

Neagului (¼Pareu Niagului; Kadi

c, 1916), near

V

alioara, which, according to that report, produced

dominantly dark, blackish remains. Moreover, based on

their commensurate size, similar preservation style, and

dark colour, the humerus (NHMUK R.4882) and the

femur (NHMUK R.11126) could potentially have come

from the same locality, and possibly even from the

same individual (herein designated as Assemblage K).

Although their exact provenance remains unknown,

these specimens were most probably recovered from the

continental successions cropping out along the P^

ar^

aul

Neagului ravine (Fig. 1E), from the lower part of the

‘middle member’of the Densus

,-Ciula Formation, of

early Maastrichtian age (Botfalvai et al., 2021; Csiki-

Sava et al., 2016)(Fig. 8).

NHMUK R.3849, S^

anpetru –Individual L

Several titanosaur specimens are registered under

NHMUK R.3489 and are part of the Hat

,eg vertebrate

collection purchased from Nopcsa in 1906.

Unfortunately, as far as we are aware, no details are

available concerning the reasons for this grouping of

specimens. They include several caudal vertebrae, as

well as more or less complete elements of the right fore-

limb (humerus, ulna, radius) and hind limb (femur and

tibia). The humerus was described and figured by

Huene (1932, p. 265, pl. 45, fig. 5), who referred it to

M. dacus. All of the limb elements accessioned as

NHMUK R.3489 are of commensurate size and have a

similar preservation state (i.e. colour and preservation

style of the specimens, as well as colour and grain-size

of the matrix still adhering to the elements), which

might indicate that these specimens were excavated

from the same locality and might belong to the same

individual, herein designated as Individual L.

As in the other cases of material excavated by

Nopcsa at S^

anpetru, there are no precise locality data

associated with Individual L. Consequently, its proven-

ance can only be identified as the S^

ınpetru Formation

stratotype section along the Sibis

,el Valley (Fig. 1E) (see

also, e.g. Individuals B and O, both from S^

anpetru).

Dark Red titanosaur, V

alioara –Assemblage M

Huene (1932, pp. 265–266, pl. 45, fig. 6) described a

dark-red-coloured, partial right ulna from V

alioara,

which is part of the Kadi

c collection. He referred this

specimen to Magyarosaurus dacus on the basis of its

robustness. Based on the corresponding figure, this spe-

cimen (for which no other identification was given) can

be identified as SZTFH Ob.3099. Although incomplete,

it is a well-preserved element that shows a distinct

24 V. D



ıez D

ıaz et al.

bowing of the shaft. Huene (1932) noticed this bowing

and interpreted it as a genuine feature of the ulna; how-

ever, close inspection of the specimen demonstrates that

it is most probably the result of post-mortem tapho-

nomic deformation. Recognition of the taphonomic

nature of this feature has notable consequences because

Huene (1932) regarded the bowing as a key morpho-

logical difference between the otherwise similar ulnae

of the species M. dacus and M. transsylvanicus. Later in

the same monograph, Huene (1932) paired the ulna with

a fibula of Assemblage A (SZTFH Ob.3086a), when

attributing them to M. dacus. Meanwhile, Huene (1932,

p. 268, pl. 46, fig. 9) referred a well-preserved right

humerus to M. transsylvanicus, indicating that it had a

similar size and shape as the one he referred to M.

dacus (NHMUK R. 3849), but being more gracile,

together with an ulna and a second fibula from

Assemblage A (see above). Although no other identifi-

cation data is given for this humerus, either, except that

it comes from V

alioara, it can be clearly identified as

specimen SZTFH v.13492, belonging to the Kadi

c col-

lection. However, overall, the ulna SZTFH Ob.3099 that

Huene referred to M. dacus displays preservational char-

acteristics (including colour) similar to those of the right

humerus SZTFH v.13492 of M. transsylvanicus, sug-

gesting a potentially common taphonomic history.

Furthermore, given that no other specimen from the

Kadi

c collection shows the same combination of preser-

vation state and dark-red–purple colour with the ulna

SZTFH Ob.3099 and the humerus SZTFH v.13492, it is

conceivable that they indeed originated from the same

site, but one which is distinct from those yielding

Assemblage A, or individuals C and K, all of which are

characterized by differing sets of preservational features.

Nevertheless, these specimens most probably did not

come from the same individual, given that the ulna is

only slightly shorter than the humerus, and are thus des-

ignated here as Assemblage M. The humerus SZTFH

v.13492 was subsequently figured by Jianu and

Weishampel (1999, figs 5, 6) and by Le Loeuff (2005a,

fig. 1b), with both of these studies referring it to M.

dacus.

Unfortunately, the limited information available

regarding the provenance of the specimens in the Kadi

collection means that it is not currently possible to re-

identify the locality that yielded the humerus and ulna.

As noted above, it seems highly probable that they did

not originate from the main collecting sites excavated

by Kadi

c(1916) near V

alioara, because they fit neither

of the briefly described sets of preservational and tapho-

nomic attributes that characterize these sites.

Nevertheless, all of the fossils from the Kadi

c collection

originate either from the neighbourhood of V

alioara or,

in far fewer cases, that of nearby Boit

,a, situated east of

V

alioara (Fig. 1E). Uppermost Cretaceous continental

sediments crop out near both villages, belonging to the

lower part of the ‘middle member’of the Densus

,-Ciula

Formation. Given this setting, the stratigraphical age of

Assemblage M is likely to be early Maastrichtian

(Csiki-Sava et al., 2016)(Fig. 8).

NHMUK R.4882, S^

anpetru –Individual N

NHMUK R.4882 comprises four titanosaur limb ele-

ments. According to the NHMUK registry, they were

collected from S^

anpetru and then donated by Lady A. S.

Woodward in 1923. This information is also confirmed

by a list (no. 22e, dated May 1923) preserved in the

Nopcsa archive of the NHMUK, that enumerates several

fossils presented by Lady Woodward, and which men-

tions “four imperfect limb bones”under number

R.4882. Three of the four specimens (right ulna, femur

and tibia) are commensurate in size and show a similar

preservation style, suggesting that they might have origi-

nated from the same locality, and possibly from the

same individual. By contrast, the fourth specimen, a left

humerus, is notably different in preservation, and it

appears unlikely that it came from the same locality as

the other three specimens (see above, P^

ar^

aul Neagului

titanosaur, V

alioara –Assemblage K). As such, the

ulna, femur, and tibia are tentatively considered to rep-

resent a single titanosaur individual, designated herein

as Individual N.

As with other historical S^

anpetru collections, the pre-

cise locality data are not known for Individual N. It can

only be constrained as probably coming from the strato-

type S^

ınpetru Formation along the Sibis

,el Valley (Fig.

1E) (see e.g. Individuals B and O, also from S^

anpetru).

NHMUK unnumbered (VIII H 916) –

Individual O

Among the incompletely prepared specimens housed in

the NHMUK collections that were purchased from

Nopcsa in 1923, there is an unregistered left tibia and

fibula that both have “VIII H 916”written in ink on

their surfaces. This type of coded identification was

used by Nopcsa to mark specimens he considered to

represent parts of one individual; for example, in his let-

ter to the NHMUK dated 16 July 1920, preserved in the

Nopcsa repository in the NHMUK archives, he listed

several specimens of turtles, dinosaurs and crocodyli-

forms represented by associated material, offered to

the NHMUK for purchase, and in which such individu-

als are identified as “A VIII 1916”(Kallokibotion), “B

VIII 916”(Zalmoxes), etc. Assessment of skeletal

Revision of Romanian sauropod dinosaurs 25

associations in these cases was most probably based

on his own (otherwise specifically unrecorded) observa-

tions made during excavation of the specimens.

The unnumbered NHMUK tibia and fibula display a

similar preservation style to one another and preserve

the same type of sedimentary matrix (a light-coloured,

greenish-grey siltstone). They are commensurate in size,

and, most importantly, have the same unusual distal

articulation as that already noted in Individual B. Based

on all these observations, this tibia-fibula pair is consid-

ered as probably originating from one locality and repre-

senting the same individual, designated here as

Individual O.

As in the case of the other vertebrate remains that

Nopcsa excavated at S^

anpetru, there is no precise local-

ity data associated with Individual O. As such, its origin

can only be identified as the S^

ınpetru Formation strato-

type section along the Sibis

,el Valley (Fig. 1E), and

loosely constrained as early to early late Maastrichtian

in age (Csiki-Sava et al., 2016)(Fig. 8).

NHMUK R.3852, S^

anpetru –Individual P

Two associated partial pubes are registered under

NHMUK R.3852. These are of comparable size, repre-

sent the left and right sides, and have a very similar

preservation style and morphology. This suggests that

they probably belong to the same individual, which we

herein designate as Individual P. The right pubis was

briefly described and figured by Huene (1932, p. 273,

pl. 48, fig. 1), with him seemingly being unaware of its

counterpart. Huene (1932) tentatively assigned this

element to the large-sized Transylvanian titanosaur

taxon he identified as ?M. hungaricus. The provenance

of Individual P can only be identified as S^

anpetru (Fig.

1E), and it most probably originates from the stratotype

ınpetru Formation (see Individuals B and O, both also

from S^

anpetru).

NHMUK R.4892, S^

anpetru –Individual Q

NHMUK R.4892 includes an associated pair of scapulo-

coracoids, from S^

anpetru, that most probably belong to

the same individual (herein designated as Individual Q),

based on their identical size, morphology and preserva-

tion style. These specimens were acquired from Nopcsa

by the NHMUK in 1923. As with other historical

anpetru collections, the location and stratigraphical age

of this individual can only be loosely constrained as the

early to early late Maastrichtian stratotype section of the

ınpetru Formation (Csiki-Sava et al., 2016)(Figs

1E,8).

La Hum

a titanosaur, S^

anpetru –Individual R

One of the first fossiliferous localities identified at

anpetru, after the re-initiation of palaeontological

explorations in the last few decades of the twentieth

century, is the La Hum

a (Dup

au) site (Csiki,

Grigorescu, et al., 2010, fig. 1C). This is a ‘fossiliferous

pocket’that was excavated mainly in 1979 (Grigorescu,

1983;S

,tiuc

a et al., 1982). It has yielded numerous ver-

tebrate remains, including an incomplete left scapula

(LPB [FGGUB] R.0003) and coracoid (LPB [FGGUB]

R.0024) of a titanosaur. The two specimens are com-

mensurate in size and have a very similar preservation

style, suggesting that they could represent elements

from the same individual (herein designated Individual

R). A large titanosaur posterior caudal vertebra (LPB

[FGGUB] R.0017) was also recovered from the same

locality, but it is less clear whether it belongs to

Individual R. The only titanosaur skeletal element previ-

ously figured from this site (the scapula) was referred to

T. dacus by S

,tiuc

a et al. (1982, pl. I, figs 3, 4).

Although age constraints are poor on the right-side

succession of the Sibis

,el Valley, the La Hum

a site is

located in the lower part of the informal lower subunit

of the stratotype S^

ınpetru Formation (Fig. 1E).

Accordingly, the age of the site is most probably early

Maastrichtian (Csiki-Sava et al., 2016)(Fig. 8).

V

alioara purple titanosaur –Individual S

Two titanosaur skeletal elements registered in the

FGGUB collections, an incomplete right scapula (LPB

[FGGUB] R.1038) and an almost complete right corac-

oid (LPB [FGGUB] R.1051), possess a very similar

style of preservation. This includes the dark reddish-pur-

ple hue of the bones, the absence of the periosteum

exposing the underlying porous bone texture, and a dark

grey, fine silty matrix adhering to both bones, as well as

filling some of the pores. Based on these taphonomic

characteristics, these two elements can clearly be differ-

entiated from most other titanosaur bones in the

FGGUB collections. Coupled with their commensurate

size, it is possible that they came from the same site,

and potentially even from the same individual (herein

designated as Individual S). Locality data is only

recorded for the coracoid, and this is imprecisely identi-

fied as coming from V

alioara, in the north-western part

of the Hat

,eg Basin (Fig. 1E). Both elements were prob-

ably collected between 1978 and 1989, but no other

information is available concerning their provenance.

Based on the known distribution of the uppermost

Cretaceous continental beds of the Hat

,eg Basin, these

elements were probably recovered from the lower part

of the ‘middle member’of the Densus

,-Ciula Formation,

26 V. D



ıez D

ıaz et al.

and, as such, are early Maastrichtian in age (Csiki-Sava

et al., 2016)(Fig. 8).

NHMUK R.4880, S^

anpetru –Individual T

According to the list of specimens presented by Lady

Woodward to the NHMUK in 1923 (list 22e, dated May

1923, from the NHMUK Nopcsa Archive), NHMUK

R.4880 includes a set of “associated caudal vertebrae”

belonging to “Titanosaurus”. Indeed, there are several

titanosaur caudal vertebrae registered under this number

in the NHMUK collections, but based on their tapho-

nomic features some of these might not have even come

from the same site, let alone the same individual.

However, among those specimens registered under

NHMUK R.4880 there is a set of three procoelous anter-

ior–middle caudal vertebrae that can be arranged into a

continuous series, and share a similar preservation style,

suggesting that these indeed might have been originally

associated. This set is designated as Individual T.

As in the case of the other specimens presented to the

NHMUK by Lady Woodward, no more detailed proven-

ance data are associated with these specimens, except

that they came from S^

anpetru (i.e. from the stratotype

section of the S^

ınpetru Formation; Fig. 1E) (see also

Individuals B and O from S^

anpetru).

LPB (FGGUB) R.1358, S^

anpetru –Individual U

Specimen LPB (FGGUB) R.1358 consists of six partial

titanosaurian posterior caudal vertebrae discovered in

1994. This material was recovered from a discontinuous

series of outcrops along the left side of the Sibis

,el

Valley, south of S^

anpetru, identified as ‘Maluri’by the

local inhabitants (see Groza, 1983). The vertebrae were

discovered in semi-articulation, lying parallel with the

bedding plane, within a mottled, grey-greenish and

brownish fine silty sandstone. No other vertebrate

remains were identified in their proximity. Their degree

of articulation, good state of preservation, and position

within the fossiliferous level suggest that these vertebrae

were not transported. These observations are consistent

with this material coming from one skeleton (see Csiki,

Grigorescu, et al., 2010), and thus are regarded as the

associated remains of one individual, herein designated

as Individual U.

The locality yielding Individual U (Fig. 1E) is located

slightly upstream (i.e. up-section) of a large fossiliferous

outcrop (‘Scoab

a’; see Csiki-Sava et al., 2016) in which

the deposits were magnetostratigraphically dated as early

Maastrichtian (Panaiotu & Panaiotu, 2010). Based on

the position of the site from which Individual U was

discovered, it also belongs to the lower part of the infor-

mal lower subunit of the S^

ınpetru Formation, and is

thus most probably early Maastrichtian in age (Csiki-

Sava et al., 2016)(Fig. 8).

K2 titanosaur, V

alioara –Individual V

In an attempt to relocate the original collecting sites of

Kadi

c(1916), Botfalvai et al. (2021) identified several

new localities yielding titanosaur remains. Of these, the

most notable discovery consists of 10 caudal vertebrae

(LPB [FGGUB] R.2715), three of which are preserved

in articulation, and one chevron. The vertebrae, recov-

ered from Botfalvai et al.’s(2021) locality K2 from the

ar^

aul Neagului ravine west of V

alioara village (Fig.

1E), can be arranged into a discontinuous series along

the tail. The specimens were excavated from the same

bed, came from a small area (less than 1 m

), and are

taphonomically similar, which suggest that they most

probably represent elements of the same tail; this con-

clusion is also supported by the report of other associ-

ated/partially articulated dinosaur skeletons present in

the same locality (Magyar et al., 2024). Thus, only one

individual is likely to be represented by this material,

herein designated as Individual V. Deposits exposed

along the P^

ar^

aul Neagului ravine belong to the lower

part of the ‘middle member’of the Densus

,-Ciula

Formation, and, as such, are of early Maastrichtian age

(Fig. 8). Moreover, this locality is only very slightly

younger than locality I of Kadi

c(1916), which yielded

the type material of M. dacus (see above). The titano-

saur material from K2 is not yet completely prepared,

and is still under study; however, the preliminary report

of Botfalvai et al. (2021) suggested that it presents fea-

tures known solely in the titanosaur Paludititan (other-

wise represented only by Individual D). The detailed

interpretation of this specimen will have to await its full

preparation and description.

Methods

This section summarizes all methods used in this study,

with the exception of those pertaining to the phylogen-

etic analyses. The latter is presented in a separate sec-

tion within the Results given that the Romanian

titanosaur OTUs in our phylogenetic data set depend on

the results of our descriptions, comparisons and taxo-

nomic decisions set out below. All measurements are

detailed in Supplemental Material Table S1.

Anatomical descriptions

We use ‘Romerian’terms (Wilson, 2006) for anatomical

structures (e.g. ‘centrum’, not ‘corpus’) and their

Revision of Romanian sauropod dinosaurs 27

orientation (e.g. ‘anterior’, not ‘cranial’). The landmark-

based terminology for vertebral laminae (Wilson, 1999)

and fossae (Wilson et al., 2011) is also employed. Serial

variation in caudal vertebrae has been determined using

the criteria outlined by Mannion et al. (2013), wherein:

(1) anterior caudal vertebrae possess ribs, even reduced

ones; (2) middle caudal vertebrae lack ribs, but have

distinct neural spines and postzygapophyses; (3) poster-

ior caudal vertebrae lack ribs, as well as distinct neural

spines and postzygapophyses; and (4) distal caudal ver-

tebrae lack ribs and neural arches. To facilitate its

description, the scapulocoracoid is described with the

long axis of the scapular blade oriented horizontally.

Similarly, metacarpals are described with their internal

(palmar) surfaces facing ventrally, as if they were held

horizontally, as a standard orientation for description

(Mannion & Otero, 2012; Poropat, Upchurch, et al.,

2015), although in life they would have been held in a

vertical position, forming a ‘U’-shaped manus.

Estimation of body size and mass

For body size and mass estimates, we use the equations

proposed by Seebacher (2001) [M(kg) ¼

214.44L(m)1.46] for calculating size, and Packard

et al. (2009) [M(g) ¼3.352PerH þF2.125] and Benson

et al. (2018) [M(kg) ¼(10(2.749 log(PerH þF)–

1.104))/1000] for mass calculations for quadrupeds, in

which M ¼body mass, PerH þF¼sum of the perime-

ters of the humerus and femur in mm, and L ¼body

length. The latter equation is based on the scaling equa-

tions provided by Campione and Evans (2012) and

Campione et al. (2014).

Systematic palaeontology

Sauropoda Marsh, 1878

Titanosauriformes Salgado, Coria & Calvo, 1997

Titanosauria Bonaparte & Coria, 1993

Lithostrotia Upchurch, Barrett & Dodson, 2004

Eutitanosauria Sanz, Powell, Le Loeuff, Mart

ınez &

Pereda Suberbiola, 1999

Magyarosaurus Huene, 1932

Magyarosaurus dacus Nopcsa, 1915

1915 Titanosaurus dacus Nopcsa: 14–15

1915 Megalosaurus Nopcsa, partim:15

(Figs. 20–34)

Lectotype. Anterior caudal vertebra (SZTFH Ob.3091,

Assemblage A).

Paralectotype (Assemblage A). posterior caudal cen-

trum (previously referred to Megalosaurus, SZTFH

Ob.3105), left humerus (SZTFH Ob.3089), left ulna

(SZTFH Ob.3100), left radius (SZTFH Ob.3101), meta-

carpal (SZTFH Ob.3096), left femur (SZTFH Ob.3088),

three left fibulae (SZTFH Ob.3086 [two with the same

specimen number], SZTFH Ob. 3102).

Referred specimens. Individual E. three dorsal verte-

brae (LPB [FGGUB] R.1061 [two in articulation],

R.1063), three fragmentary dorsal ribs (LPB [FGGUB]

R.1861, R.1863, R.1865), four incomplete caudal verte-

brae (LPB [FGGUB] R.1069, R.1070 [two in articula-

tion], R.1071), one chevron (LPB [FGGUB] R.0076),

left humerus (LPB [FGGUB] R.1047), two fragments of

the left radius (LPB [FGGUB] R.1049 þR.1060), five

metacarpal fragments (LPB [FGGUB] R.0074, R.0077,

R.1053, R.1862, R.1864), left (LPB [FGGUB] R.1046)

and right (LPB [FGGUB] R.1992) femora, proximal end

of left tibia and fibula (LPB [FGGUB] R.2299).

Assemblage F. fragmentary cervical vertebra (LPB

[FGGUB] R.2505), chevron (LPB [FGGUB] R.2510),

right humerus (LPB [FGGUB] R.2506), metacarpal

(LPB [FGGUB] R.2509), right femur (LPB [FGGUB]

R.2507), right femur of a larger individual (LPB

[FGGUB] R.2508).

Locality and distribution. The type locality is Locality

I (Kadi

c, 1916), between V

alioara and Ciula Mic

a vil-

lages. This site occurs in the lowermost part of the

unnamed middle member of the Densus

,-Ciula

Formation (Fig. 1E), which is earliest Maastrichtian in

age (Fig. 8). Individual E was recovered from the same

stratigraphical unit, in the vicinity of Ciula Mic

a(Figs

1E,8). Assemblage F is also early Maastrichtian in age,

but comes from deposits along the B

arbat River, at Pui,

belonging either to the S^

ınpetru Formation (e.g.

Grigorescu, 1992; Mamulea, 1953b), or a different

(though penecontemporaneous) unit (Csiki-Sava et al.,

2016; Therrien, 2005)(Figs 1E,8).

Revised diagnosis. Magyarosaurus dacus can be diag-

nosed on the basis of five autapomorphies (marked with

an asterisk), as well as six ‘local’autapomorphies

(defined in Beeston et al., 2024 as an apomorphy that is

uniquely present in one taxon within a region of the

tree, but that is also convergently present in a phylogen-

etically distant taxon (or taxa) within the same higher

level clade): (1) anterior caudal centra with ventrolateral

ridges but without a midline furrow (Individual E and

Assemblage A); (2) posterior margin of the proximal

articulation of the chevron forms a small ‘hook’-like

process that overhangs the posterior surface of the

ramus (Individual E); (3) chevrons with a proximodistal

ridge located on the lateral surface of each ramus

(Individual E and Assemblage F); (4) marked promin-

ence extending distally from the humeral head along the

28 V. D



ıez D

ıaz et al.

posterior surface, giving the proximal end a posteriorly

deflected aspect in lateral/medial views (Individual E

and Assemblage F); (5) coracobrachialis fossa divided

into a small, shallow dorsomedial region and a large,

deep ventrolateral region by an obliquely curved ridge

(Individual E and Assemblage F); (6) position of the

well-developed bulge (presumably site for M. latissimus

dorsi) approximately equidistant from medial and lateral

margins of posterior surface (Individual E and

Assemblages A and F); (7) strong distolateral orienta-

tion of the interosseous ridge of the ulna, with the prox-

imal end of this ridge intersecting the anteromedial

margin of the shaft (Assemblage A); (8) shaft of radius

displays torsion of approximately 45(Assemblage A

and Individual E); (9) distal end of radius, broader ante-

roposteriorly on the lateral margin than on the medial

one (preservation means that this feature is only visible

in the radius of Individual E); (10) low eccentricity of

femoral midshaft (1.1–1.4) (Individual E, and

Assemblages A and F); (11) subtle ridge on the anter-

ior margin of the lateral surface of the fibula, with its

proximal tip situated at approximately the level of the

distal end of the lateral trochanter (Assemblage A).

Justification for referrals. The type material

(Assemblage A) can be diagnosed by autapomorphies

present in the anterior caudal vertebra (no. 1), humerus

(no. 6), ulna (no. 7), femur (no. 10), and fibula (no. 11).

Poor preservation and lack of anatomical overlap means

that Individual E and Assemblage F can only be com-

pared to Assemblage A via the humerus and the femur,

but here the referrals are supported by the presence of

autapomorphies 6 and 10. Moreover, Assemblage A and

Figure 20. Magyarosaurus dacus, cervical vertebra LPB (FGGUB) R.2505 (Assemblage F) in A, anterior; B, right dorsolateral; C,

posterior; D, ventral; and E, dorsal views. Scale bar equals 50mm.

Revision of Romanian sauropod dinosaurs 29

Individual E share an unusual combination of character

states in the anterior caudal centra, whereby they lack a

ventral midline furrow, but possess distinct ventrolateral

ridges (autapomorphy no. 1). Individual E and

Assemblage F can be compared directly via the humeri,

metacarpals and femora. Some of these elements pro-

vide strong evidence that E and F belong to the same

taxon, possessing autapomorphy nos. 3 to 6, 8 and 10.

Description and comparisons of Magyarosaurus

dacus

Cervical vertebra. LPB (FGGUB) R.2505 (Assemblage

F) is a poorly preserved cervical vertebra (Figs 20,21),

probably from the posterior region of the neck, based on

comparisons with titanosaurs with near-completely

known cervical series (e.g. Calvo, Porfiri, et al., 2007;

Carballido et al., 2017; Curry Rogers, 2009). Most of

Figure 21. Magyarosaurus dacus, interpretive drawing of the cervical vertebra LPB (FGGUB) R.2505 (Assemblage F) in A,

anterior; B, dorsolateral; and C, posterior views. Abbreviations: ACDL, anterior centrodiapophyseal lamina; CPOL,

centropostygapophyseal lamina; D, diapophysis; lSPRL, lateral spinoprezygapophyseal lamina; mSPRL, medial

spinoprezygapophyseal lamina; NC, neural canal; PA, parapophysis; PCDL, posterior centrodiapophyseal lamina; PO,

postzygapophysis; POCDF, posterior centrodiapophyseal fossa; PODL, postzygodiapophyseal lamina; PRE, prezygapophysis; PRSL,

prespinal lamina; SDF, spinodiapophyseal fossa; SPDL, spinodiapophsyeal lamina; SPOF, spinopostzygapophyseal fossa; SPOL,

spinopostzygapophyseal lamina; SPRF, spinoprezygapophyseal fossa. Scale bar equals 50mm.

Figure 22. Magyarosaurus dacus, dorsal vertebrae. Anterior dorsal vertebra LPB (FGGUB) R.1063 (Individual E) in A, left lateral;

B, posterior; C, right lateral; D, anterior; E, dorsal; and F, ventral views. Middle–posterior dorsal vertebrae LPB (FGGUB) R.1061

(Individual E) in G, left lateral; H, posterior; I, right lateral; J, anterior; and K, ventral views. Scale bar equals 100mm.

30 V. D



ıez D

ıaz et al.

the left half of the vertebra has been eroded away, and

it is missing the condyle, both prezygapophyses, and the

dorsal portion of the neural spine. The right side is also

incomplete in places, and some areas of the vertebra are

partly obscured by matrix. As a consequence of the

eroded left side, the internal tissue structure is clearly

visible, demonstrating the presence of camellae in both

the centrum and neural arch, as in most somphospondy-

lans (Wilson & Sereno, 1998).

Based on the nearly complete parapophysis and a

lamina that probably represents part of the anterior cen-

trodiapophyseal lamina (ACDL), little of the non-con-

dylar centrum is missing. The centrum is

anteroposteriorly short, with an aEI of approximately

1.74. This value varies between titanosaur taxa and

along the posterior cervical series, but an aEI of 1.74 is

comparable to that of some of the posterior cervical ver-

tebrae of taxa such as Alamosaurus (Tykoski & Fiorillo,

2017), Patagotitan (Carballido et al., 2017), and

Savannasaurus (Poropat et al., 2020). By contrast, the

posterior cervical vertebrae of some titanosaurs have a

lower aEI, including Abditosaurus (aEI ¼1.1; Vila

et al., 2022), and Saltasaurus (aEI ¼1.03–1.52; Powell,

2003).

The posterior cotyle is deeply concave and, although

incomplete, it is mediolaterally wider than dorsoven-

trally tall (Table S1). The morphology of the lateral

pneumatic excavation is difficult to discern because of

matrix and preparation: a sharp-lipped excavation is pre-

sent, but its depth, and whether there were dividing

ridges, cannot be determined (Figs 20B,21B). It is not

anteriorly restricted, and it does not extend to near the

posterior-most margin of the centrum. There are two

pneumatic foramina close to the dorsal surface of the

base of the parapophysis, inside the lateral pneumatic

excavation. The parapophysis projects ventrolaterally,

but it does not extend far beyond the ventral margin of

the centrum. In this regard, it differs from Overosaurus

(Coria et al., 2013) and some members of Lognkosauria

(Gonz

alez Riga et al., 2018). In Magyarosaurus, the

parapophysis appears to occupy more than half of the

length of the non-condylar centrum: such parapophyseal

elongation is a derived state that also occurs in the mid-

dle–posterior cervical vertebrae of the Iberian taxon

Abditosaurus, as well as Alamosaurus,Bonatitan,

Mansourasaurus, saltasaurines and Yongjinglong

(D’Emic, 2012; Li et al., 2014; Salgado et al., 2015;

Sallam et al., 2018; Vila et al., 2022). However, the

articular surface of the parapophysis in Magyarosaurus

is relatively small compared to these taxa, such that

much of its anteroposterior extent occurs where it

merges with the centrum and it is more akin to a poster-

ior centroparapophyseal-like lamina (PCPL). This lam-

ina fades out posteriorly, such that there are no

Figure 23. Magyarosaurus dacus, dorsal rib fragments. LPB (FGGUB) R.1861 (Individual E) in A, lateral/medial? view; and B,

transverse cross-section. LPB (FGGUB) R.1863 1861 (individual E) in C, lateral/medial? view; and D, transverse cross-section. Scale

bar equals 50 mm.

Revision of Romanian sauropod dinosaurs 31

ventrolateral ridges along the posterior third of the cen-

trum. As is also the case in most other titanosaurs

(Poropat et al., 2016; Upchurch, 1998), the dorsal sur-

face of the parapophysis is unexcavated. The preserved

area of the ventral surface of the centrum is transversely

flat anteriorly and gently convex posteriorly (Fig. 20D).

It is too incomplete to determine if a midline ridge was

present.

There is a prominent, posterior centrodiapophyseal

lamina (PCDL) that extends posteriorly and slightly ven-

trally (Fig. 21B). The ACDL is incomplete, but the dor-

sal-most preserved portion is steeply oriented. These

two laminae delimit a deep triangular centrodiapophy-

seal fossa (CDF), floored by a lamina that extends paral-

lel to the PCDL and roofs the lateral pneumatic

excavation on the centrum. The neural arch does not

extend to the posterior margin of the centrum. Each cen-

tropostzygapophyseal lamina (CPOL) is robust and

nearly vertical (Fig. 21C). There is no interpostzygapo-

physeal lamina (TPOL), but we cannot be certain

whether its absence is a preservational artefact. Despite

this, there is clearly no vertical midline lamina between

the roof of the posterior neural canal opening and the

postzygapophyses. The latter processes terminate

approximately level with the posterior margin of the

cotyle. This morphology characterizes the posterior cer-

vical vertebrae of most titanosaurs, but contrasts with

many other sauropods in which the postzygapophyses

terminate anterior to the posterior margin of the centrum

(Poropat et al., 2016; Tschopp et al., 2015). The postzy-

gapophyses are proportionally large, with articular facets

that face ventrally, partly laterally, and slightly

Figure 24. Magyarosaurus dacus, anterior caudal vertebrae. Lectotype SZTFH Ob.3091 (Assemblage A) in A, right lateral; B,

anterior; C, left lateral; D, posterior; E, dorsal; and F, ventral views. LPB (FGGUB) R.1069 (Individual E) in G, right lateral; H,

anterior; I, posterior; and J, ventral views. LPB (FGGUB) R.1070 (Individual E) in K, right lateral; L, left lateral; M, posterior; and

N, ventral views. LPB (FGGUB) R.1071 (Individual E) in O, right lateral, P, anterior; Q, left lateral; and R, posterior views.

Abbreviations: PO, postzygapophysis; PRE, prezygapophysis. The number 1 indicates the autapomorphy described in the text:

anterior caudal centra with ventrolateral ridges but without a midline furrow. Scale bar equals 100 mm.

32 V. D



ıez D

ıaz et al.

posteriorly. There appears to be a small epipophysis, at

least on the left postzygapophysis. The medial surfaces

of the postzygapophyses are not excavated.

The preserved proximal part of the diapophysis proj-

ects laterally. Its broad dorsal surface is predominantly

flat. A prominent postzygodiapophyseal lamina (PODL),

in conjunction with the PCDL and CPOL, define a deep

triangular postzygapophyseal centrodiapophyseal fossa

(POCDF) (Fig. 21B). Close to the base of the anterior

region of the neural spine, there is an elliptical, well-

defined foramen within the spinodiapophyseal fossa

(SDF). Although the presence of a SDF is common in

sauropod cervical vertebrae (Wilson et al., 2011), a

comparably deep opening has otherwise only been

reported in the posterior cervical vertebrae of

Alamosaurus (Tykoski & Fiorillo, 2017), Abditosaurus

(Vila et al., 2022), and some members of Lognkosauria

(Gonz

alez Riga et al., 2018) and Aeolosaurini (Coria

et al., 2013; Gorscak et al., 2017).

The lower sections of the SPRLs are preserved (Fig.

21A). There is both a medial and lateral SPRL (mSPRL

and lSPRL, respectively), although only their bases are

preserved (note that the left lSPRL is not preserved).

Their distal ends (i.e. the portions of SPRLs that would

have ascended to the spine summit) are not preserved, so

whether the mSPRLs and lSPRLs continued as separate

laminae or merged distally cannot be ascertained. There

is a sharp-lipped, oval pneumatic foramen within the

resultant SPRL fossa (SPRL-F). The presence of medial

and lateral branches of the SPRL is unusual in sauropods,

but it has been documented in posterior cervical vertebrae

of Mendozasaurus (Gonz

alez Riga et al., 2018)and

Neuquensaurus (Zurriaguz, 2016). However, even in

these taxa, a double SPRL is not consistently present in

Figure 25. Magyarosaurus dacus, posterior caudal vertebra SZTFH Ob.3105 (Assemblage A) in A, right lateral view, from Nopcsa

(1915); B, right lateral; C, ventral; D, posterior; E, anterior; and F, dorsal views. Scale bar equals 50 mm.

Revision of Romanian sauropod dinosaurs 33

posterior cervical vertebrae; as such, we do not consider

it a diagnostic feature of Magyarosaurus,pendingthe

discovery of additional specimens with this feature.

Although too poorly preserved to be certain, there is

some evidence for a midline prespinal lamina (PRSL)

within the spinoprezygapophyseal fossa (SPRF), as is the

case in the posterior-most cervical vertebrae of most som-

phospondylans (D’Emic, 2012; Salgado et al., 1997).

The broad spinopostzygapophyseal laminae (SPOLs)

bound a deep, but mediolaterally narrow

spinopostzygapophyseal fossa (SPOF), with no evidence

for a postspinal lamina (POSL) (Fig. 21C). The narrow

SPOF resembles those in the posterior cervical vertebrae

of some titanosaurs (e.g. Atsinganosaurus [D



ıez D

ıaz

et al., 2018], Overosaurus [Coria et al., 2013, fig. 2],

Trigonosaurus [Campos et al., 2005, fig. 4], and salta-

saurines [e.g. Zurriaguz, 2016, fig. 4]), but differs from

the wide fossa that characterizes others, including

Ampelosaurus (Le Loeuff, 2005b), Garrigatitan (D

ıez

D

ıaz et al., 2021), Kaijutitan (Filippi et al., 2019), and

Figure 26. Magyarosaurus dacus, chevrons. Chevron LPB (FGGUB) R.2510 (Assemblage F) in A, anterior; B, right lateral; C,

proximal (posterior towards top); D, posterior; and E, left lateral views. Chevron ramus LPB (FGGUB) R.0076 (Individual E) in F,

left lateral; G, proximal (anterior towards left); H, posterior; and I, right lateral views. The numbers 2 and 3 indicate the

autapomorphies described in the text: posterior margin of the proximal articulation of the chevron forms a small ‘hook’-like process

that overhangs the posterior surface of the ramus; and chevrons with a proximodistal ridge located on the lateral surface of each

ramus. Scale bar equals 50 mm.

34 V. D



ıez D

ıaz et al.

Mendozasaurus (Gonz

alez Riga et al., 2018). Given the

incomplete nature of the specimen, it is not possible to

obtain further information on the morphology of the cer-

vical neural spine of Magyarosaurus.

Anterior dorsal vertebrae. LPB (FGGUB) R.1063

(Individual E) preserves the centrum and base of the

neural arch of an anterior dorsal vertebra, missing

material from its posterior surface (Fig. 22A–F). Its

internal bone structure is camellate, a derived state

that characterizes the presacral vertebrae of sompho-

spondylans (Wilson & Sereno, 1998). The opisthocoe-

lous centrum is mediolaterally wide and

dorsoventrally low (Table S1), similar to the morph-

ology of most titanosaurs (Mannion, Upchurch,

Schwarz, et al., 2019). Its ventral surface is trans-

versely convex, lacking fossae or ridges (Fig. 22F).

In this regard, it differs from the anterior dorsal

Figure 27. Magyarosaurus dacus, humeri. Left humerus LPB (FGGUB) R.1047 (Individual E) in A, anterior; B, distal (posterior

towards bottom); C, medial; D, posterior; E, proximal (anterior towards top); and F, lateral views. Right humerus LPB (FGGUB)

R.2506 (Assemblage F) in G, anterior; H, distal (posterior towards bottom); I, medial; J, posterior; K, proximal (anterior towards

top); and L, lateral views. Left humerus SZTFH Ob.3089 (Assemblage A) in M, anterior; N, medial; O, posterior; and P, lateral

views. Abbreviations: dpc, deltopectoral crest; lat, lateral; post, posterior. The numbers 4,5

and 6indicate the autapomorphies

described in the text: marked prominence extending distally from the humeral head along the posterior surface, giving the proximal

end a posteriorly deflected aspect in lateral/medial; coracobrachialis fossa divided into a small, shallow dorsomedial region and a

large, deep ventrolateral region by an obliquely curved ridge; and position of the well-developed bulge (presumably site for M.

latissimus dorsi), which is situated approximately equidistant from medial and lateral margins of posterior surface, level with the

distal tip of the deltopectoral crest. Scale bar equals 100 mm.

Revision of Romanian sauropod dinosaurs 35

centra of Lirainosaurus (D

ıez D

ıaz et al., 2013b, fig.

3a), Lohuecotitan (D



ıez D

ıaz et al., 2016), and

Opisthocoelicaudia (Borsuk-Białynicka, 1977), which

are characterized by a longitudinal ventral ridge.

The parapophyses are situated mainly on the centrum,

extending onto the base of the neural arch (Fig. 22A,

C), supporting the inference that this specimen probably

represents the second or third dorsal vertebra, based on

comparisons with titanosaurs preserving complete dorsal

series (e.g. Rapetosaurus [Curry Rogers, 2009]) and

Trigonosaurus [Campos et al., 2005]). In this

Magyarosaurus specimen, the parapophyses are rela-

tively prominent processes that project quite far lat-

erally. Each lateral surface of the centrum is excavated

by a deep pneumatic foramen, located at approximately

midheight, lying slightly closer to the anterior margin

than the posterior one. This opening is oval, with an

acute posterior terminus, as is typical for the anterior

dorsal centra of most macronarians (Upchurch, 1998;

Upchurch et al., 2004). Unlike the lateral pneumatic

openings of the dorsal centra of many titanosaurs

(Bonaparte & Coria, 1993; Upchurch et al., 2004), that

of LPB (FGGUB) R.1063 is not set within a fossa. The

neural arch is anteriorly biased, and the floor of the

neural canal is flat and unexcavated.

Middle–posterior dorsal vertebrae. LPB (FGGUB)

R.1061 (Individual E) consists of two articulated mid-

dle–posterior dorsal centra, with the base of the neural

arch preserved in the more posterior element (Fig. 22G-

K). The posterior end of the more posterior centrum is

slightly incomplete. No parapophyses are present on

either centrum or the preserved portion of neural arch,

supporting the view that these vertebrae are not from

the anterior region of the dorsal series. The aEI of the

more anterior of these centra is 1.36, and that of the

Figure 28. Magyarosaurus dacus, left ulna SZTFH Ob.3100 (Assemblage A) in A, medial; B, proximal (anterior towards top); C,

posterior; D, anterolateral; and E, anterior views. The number 7indicates the autapomorphy described in the text: distolateral

orientation of the interosseous ridge of the ulna: proximal end of interosseous ridge, on anterior surface, intersects the anteromedial

margin of the shaft and slants distolaterally. Scale bar equals 100 mm.

36 V. D



ıez D

ıaz et al.

more posterior one is 1.34, higher than the values calcu-

lated for the middle–posterior dorsal vertebrae of

Lirainosaurus (D



ıez D

ıaz et al., 2013b). The centra

have a camellate internal structure. Both centra are

strongly opisthocoelous, and there is a slight central

depression on each condyle. In lateral view, the ventral

margins of the centra are strongly arched upwards. The

ventral surface of each centrum is gently convex trans-

versely, lacking ridges or excavations (Fig. 22F, K).

This is similar to the condition in most sauropods, but

differs from the midline ventral ridge that characterizes

the middle–posterior dorsal centra of some titanosaurs,

including Diamantinasaurus (Poropat, Upchurch, et al.,

2015), Futalognkosaurus (Calvo, Porfiri, et al., 2007),

and Opisthocoelicaudia (Borsuk-Białynicka, 1977), and

that is variably present in taxa such as Overosaurus

(Coria et al., 2013) and possibly Atsinganosaurus (D

ıez

D

ıaz et al., 2018). ‘Eye’-shaped lateral pneumatic open-

ings are present, extending across approximately the

middle third of the anteroposterior length of the non-

condylar centrum, at approximately midheight (Fig.

22A, C, G, I). Although filled with matrix, these lateral

pneumatic openings appear to be deep, and they are not

set in a fossa. The neural arch does not extend to the

posterior margin of the centrum.

Thoracic ribs. LPB (FGGUB) R.1861 (in two pieces),

R.1863, and R.1865, are small fragments of thoracic rib

shafts from Individual E (Fig. 23). There is no clear evi-

dence for pneumaticity, but this is potentially because

they are almost certainly quite distal fragments of the

shaft. LPB (FGGUB) R.1861 is ‘plank’-like, with the

anteroposterior width approximately three times that of

the mediolateral width (Fig. 23B). Although we cannot

be certain of its position in the dorsal series, such

‘plank’-like ribs are a derived state of Titanosauriformes

when present in the anterior thoracic region (Wilson &

Sereno, 1998). By contrast, the anteroposterior to medio-

lateral width ratio is closer to 2.0 in LPB (FGGUB)

R.1863, which has a narrow, ‘D’-shaped cross-section

(Fig. 23D). The absence of the plank-like structure in

LPB (FGGUB) R.1863 potentially indicates that it comes

from a more posterior rib within the thoracic region.

Anterior caudal vertebrae. Parts of five anterior caudal

vertebrae can currently be referred to Magyarosaurus

(Fig. 24). Specimens LPB (FGGUB) R.1069 (Fig. 24G–

J) and R.1071 (Fig. 24O–R) (Individual E) belong to

the proximal section of the tail. However, none of them

can be considered as being the first caudal vertebra

because of the presence of chevron facets. Their aEI

ranges from 0.68 to 0.81, a low value when compared

with the anterior-most caudal vertebrae of some early-

branching titanosaurs, such as Andesaurus (aEI ¼0.92,

Mannion & Calvo, 2011) and Malawisaurus (aEI ¼

0.96, Gomani, 2005), but closer to the aEI of

Lohuecotitan (where the aEI of the first six caudal cen-

tra ranges from 0.57 to 0.74; D



ıez D

ıaz et al., 2016),

Bonitasaura (aEI ¼0.77; Gallina & Apestegu

ıa, 2015),

and Trigonosaurus (aEI ¼0.81; Campos et al., 2005).

Figure 29. Magyarosaurus dacus, radii. Left radius LPB (FGGUB) R.1049 and R.1060 (Individual E) in A, anterior; B, distal

(anterior towards top); C, medial; D, proximal (posterior towards bottom); E, posterior; and F, lateral views. Left radius SZTFH

Ob.3101 (Assemblage A) in G, proximal (posterior towards bottom); H, posterior; I, medial; J, anterior; K, distal (anterior towards

top); and L, lateral views. The numbers 8 and 9 indicate the autapomorphies described in the text: shaft of radius displays torsion of

approximately 45; and distal end of radius, broader anteroposteriorly on the lateral margin than on the medial one. Scale bar equals

100 mm.

Revision of Romanian sauropod dinosaurs 37

Figure 30. Magyarosaurus dacus, Individual E metacarpals. Metacarpal LPB (FGGUB) R.0074 in A, dorsal; B, lateral/medial; C,

ventral; D, medial/lateral; and E, distal (ventral towards bottom) views. Metacarpal III? LPB (FGGUB) R.1864 in F, ventral; G,

lateral/medial; H, dorsal; I, medial/lateral; and J, proximal (ventral towards bottom) views; K, transverse cross-section at mid-shaft.

Left metacarpal IV? LPB (FGGUB) R.1053 in L, dorsal; M, lateral/medial; N, ventral; O, medial/lateral; and P, distal (ventral

towards bottom) views. Metacarpal IV? LPB (FGGUB) R.1862 in Q, dorsal; R, lateral/medial; S, ventral; T, medial/lateral; U,

transverse cross-section of the distal third of the shaft; and V, distal (ventral towards bottom) views. Right metacarpal V LPB

(FGGUB) R.0077 in W, lateral; X, ventral; Y, medial; Z, dorsal; and A’,distal (ventral towards left) views. Scale bar equals 10mm.

Figure 31. Magyarosaurus dacus, Assemblages A and F metacarpals. Metacarpal SZTFH Ob.3096 (Assemblage A) in A, lateral/

medial; B, dorsal?; C, ventral?; and D, distal (dorsal? towards bottom) views. Metacarpal LPB (FGGUB) R.2509 (Assemblage F) in

E, ventral; and F, dorsal views. Scale bar equals 50mm.

38 V. D



ıez D

ıaz et al.

LPB (FGGUB) R.1070 (Fig. 24K–N) (Individual E) is

an anterior caudal vertebra, in articulation with the pos-

terior part of the preceding vertebra, that comes from

further along the caudal sequence than LPB (FGGUB)

R.1069 and R.1071. The lectotype of Magyarosaurus,

SZTFH Ob.3091 (Fig. 24A–F) (Assemblage A), is prob-

ably from the distal part of the anterior series and has

an aEI of 1.01.

None of these caudal vertebrae present a camellate

internal tissue structure, contrasting with some

Figure 32. Magyarosaurus dacus, femora. Left femur SZTFH Ob.3088 (Assemblage A) in A, anterior; B, posterior views; and C,

transverse cross-section (at approximate midshaft). Left femur LPB (FGGUB) R.1046 (Individual E) in D, posterior; E, lateral, F,

proximal (posterior towards top); G, anterior; H, distal (anterior towards top); and I, medial views. Right femur LPB (FGGUB)

R.1992 (Individual E) in J, anterior; K, proximal (posterior towards bottom); L, posterior; and M, distal (posterior towards top)

views. Right femur LPB (FGGUB) R.2507 (Assemblage F) in N, anterior; O, lateral; P, proximal (posterior towards bottom); Q,

posterior; R, distal (posterior towards top); and S, medial views. Right femur LPB (FGGUB) R.2508 (Assemblage F) in T, anterior;

and U, posterior views. The number 10indicates the autapomorphy described in the text: low eccentricity of femoral midshaft (1.1–

1.4). Scale bar equals 100 mm.

Revision of Romanian sauropod dinosaurs 39

titanosaurs, in which the neural arch (and sometimes

the centrum too) of anterior caudal vertebrae is intern-

ally pneumatized (e.g. see Poropat et al., 2020). The

concave anterior articular surface of each centrum has

a slightly dorsoventrally compressed subcircular out-

line. The centra are strongly procoelous, and demon-

strate that a posterior condyle was present throughout

the anterior caudal vertebral series. This feature charac-

terizes nearly all lithostrotians (Mannion et al., 2013).

The posterior condyle is prominent in Magyarosaurus,

but is also constricted because of the presence of a

marked ridge on the outer margin of the posterior

articular face of the centrum, as in most somphospon-

dylans (Mannion et al., 2013). The condyle is located

in the centre of the articulation surface in SZTFH

Ob.3091 and LPB (FGGUB) R.1071, whereas it is

more dorsally shifted in LPB (FGGUB) R.1069 and

1070. This demonstrates that some of the purported

differences in morphotypes amongst Romanian titano-

saurian caudal vertebrae identified by Mocho et al.

(2023) might instead represent serial variation. In

SZTFH Ob.3091 there is a shallow midline groove on

the condyle (Fig. 24D).

The ventral surfaces of LPB (FGGUB) R.1069 and

R.1071 are transversely wide, whereas they are narrower in

SZTFH Ob.3091 and LPB (FGGUB) R.1070, as a result of

the ventral half of the centrum being more transversely

compressed than in the two more anterior caudal vertebrae.

The ventral surface is also flat in all of the anterior caudal

vertebrae, lacking a midline furrow. The presence of a

ventral midline furrow characterizes the anterior caudal

centra of many somphospondylans, including several tita-

nosaurs (e.g. Alamosaurus,Andesaurus,Epachthosaurus,

Malawisaurus,Opisthocoelicaudia,Saltasaurus;Mannion

& Calvo, 2011; Upchurch, 1998; Wilson, 2002), but many

taxa lack this feature. Amongst Titanosauria, a flat ventral

surface, without a midline groove, also characterizes the

anterior caudal centra of other western Eurasian species

(i.e. Atsinganosaurus,Lirainosaurus,Lohuecotitan,

Normanniasaurus,Paludititan,andVolgatitan;Averianov

& Efimov, 2018; Csiki, Codrea, et al., 2010;D



ıez D

ıaz

et al., 2013b,2016,2018; Mannion, Upchurch, Jin, et al.,

2019), as well as several South American taxa, e.g.

Figure 33. Magyarosaurus dacus, left tibia and fibula LPB (FGGUB) R.2299 (Individual E) in A, proximal (posterior towards

bottom); B, anterior; C, posterior (fibula barely visible on the left); D, lateral (only fibula visible); and E, medial (fibula not visible)

views. F, tibia and fibula articulated with femur LPB (FGGUB) R.1046 in anterior view. Scale bar equals 100mm.

40 V. D



ıez D

ıaz et al.

Baurutitan (Kellner et al., 2005)andMendozasaurus

(Gonz

alez Riga et al., 2018). All known anterior caudal

centra of Magyarosaurus bear anterior and, more promin-

ent, posterior chevron facets, although these are less well-

developed in LPB (FGGUB) R.1071. In SZTFH Ob.3091

and LPB (FGGUB) R.1070, the ventral surface is delimited

laterally by two longitudinal ridges that connect the anterior

and posterior chevron facets (Fig. 24F, N). Taxa that have a

midline ventral furrow in their anterior caudal centra typic-

ally also have ventrolateral ridges (including most of the

taxa mentioned above), though Lirainosaurus (D



ıez D

ıaz

et al., 2013b) and some specimens of Atsinganosaurus



ıez D

ıaz et al., 2018) are also characterized by the com-

bination of ventrolateral ridges without a midline furrow

observed in Magyarosaurus.

The lateral surfaces of the anterior caudal centra of

Magyarosaurus lack excavations or ridges. Transverse

processes are partially preserved on LPB (FGGUB)

R.1070 and 1071, where they are situated at the arch-

centrum junction. In SZTFH Ob.3091, they are reduced

to low, rugose prominences at the arch-centrum junc-

tion, supporting our inference that this specimen comes

from the distal part of the anterior tail series. These

transverse processes are gently excavated at midheight,

forming a shallow, anteroposteriorly oriented groove

(Fig. 24E).

The base of the neural arch is anteriorly placed on

the centrum. Only SZTFH Ob.3091 preserves enough of

the neural arch and spine to provide detailed anatomical

information. Although incomplete, the prezygapophyses

are slender processes that project anterodorsally beyond

the anterior margin of the centrum, as in most titanosau-

rian anterior caudal vertebrae, including those of other

European taxa (Csiki, Codrea, et al., 2010;D



ıez D

ıaz

et al., 2013b,2016,2018; Le Loeuff, 2005b; Le Loeuff

et al., 2013). The prezygapophyses are supported by

short, stout CPRLs, and are linked to each other by a

wide, horizontal TPRL. Each postzygapophysis has a

flat oval facet that faces mainly laterally and does not

project beyond the posterior margin of the neural spine.

Ventrally, the postzygapophyses are each supported by a

CPOL. No hyposphene-hypantrum complex is present,

an absence that is typical for the caudal vertebrae of

most titanosaur taxa (Upchurch, 1998; Upchurch et al.,

2011). The neural spine is almost complete and is dorso-

ventrally short, similar to that of taxa such as

Atsinganosaurus,Lirainosaurus,Paludititan,

Rapetosaurus and Saltasaurus (Csiki, Codrea, et al.,

2010; Curry Rogers, 2009;D



ıez D

ıaz et al., 2013b,

2018; Powell, 1992), but unlike the elongated spines of

Andesaurus,Muyelensaurus and Lohuecotitan (Calvo,

Gonz

alez Riga, et al., 2007;D



ıez D

ıaz et al., 2016;

Figure 34. Magyarosaurus dacus, fibulae. Left fibula SZTFH Ob.3086a (Assemblage A) in A, lateral; B, posterior; C, proximal

(medial towards bottom); D, medial; and E, anterior views. Left fibula SZTFH Ob.3086b (Assemblage A) in F, lateral; G, posterior;

H, proximal (medial towards bottom); I, medial; J, distal (medial towards top); and K, anterior views. Left fibula SZTFH Ob.3102

(Assemblage A) in L, lateral; M, posterior; N, medial; and O, anterior views. The number 11 indicates the autapomorphy described

in the text: subtle ridge on the anterior margin of the lateral surface of the fibula, with its proximal tip situated at approximately the

level of the distal end of the lateral trochanter. Scale bar equals 100 mm.

Revision of Romanian sauropod dinosaurs 41

Mannion & Calvo, 2011). The neural spine of SZTFH

Ob.3091 is transversely compressed and posterodorsally

oriented, as in many titanosaurs, including Andesaurus,

Atsinganosaurus,Lirainosaurus and Normanniasaurus



ıez D



ıaz et al., 2013b,2018; Le Loeuff et al., 2013;

Mannion & Calvo, 2011). It does not extend as far as

the posterior margin of the non-condylar centrum.

Weakly developed SPRLs define the anterolateral mar-

gins of the neural spine. The presence of weakly devel-

oped SPOLs can also be inferred, because there is a

central depressed area on the posterior surface of the

neural spine, between and above the postzygapophyseal

surfaces, with raised margins laterally.

Posterior caudal vertebra. SZTFH Ob.3105

(Assemblage A) is an almost complete posterior caudal

vertebra (Fig. 25). It has an aEI of 1.99. It is not pos-

sible to observe the internal tissue structure anatomy of

this specimen. The centrum is spool-like, slightly opis-

thocoelous, with a poorly developed anterior condyle. A

change in the type of articulation within the tail series is

not uncommon within Titanosauria, especially in the

posterior caudal vertebrae (e.g. Opisthocoelicaudia,

Bonitasaura and Rinconsaurus; Borsuk-Białynicka,

1977; Calvo & Gonz

alez Riga, 2003; Gallina &

Apestegu

ıa, 2015;P



erez Moreno et al., 2022). The out-

line of both articular surfaces is slightly sub-quadrangu-

lar. No ridges or rugosities are present on the lateral

surfaces of the centrum. The ventral surface is flat, and

only the posterior articular facets for the chevrons can

be seen. The neural arch is short, and is anteriorly

placed on the centrum. Only the bases of the prezyga-

pophyses are preserved, and the postzygapophyses are

absent. Although not complete, the neural spine was ori-

ginally anteroposteriorly long and dorsoventrally low.

No laminae or fossae are present in this specimen.

Chevrons. LPB (FGGUB) R.2510 (Assemblage F)

includes three fragments that come from a single chev-

ron, missing its distal end (Fig. 26). The proximal

articular surfaces cannot be assessed because of their

poor preservation (Fig. 26C), but the chevron is clearly

‘open’proximally (i.e. no bridge of bone links the left

and right proximal articulations above the haemal

canal), such that it is characterized by the typical open

‘Y’-shaped morphology described by Otero et al. (2012,

fig. 2), as is also the case in most macronarians (Calvo

& Salgado, 1995; Upchurch, 1998). Although the chev-

ron is slightly incomplete, the haemal canal was prob-

ably similar in length to the distal blade. Such a

relatively deep haemal canal (i.e. at least 40% of total

chevron length) is a derived state that occurs in most

titanosauriforms (Curry Rogers, 2005; Wilson, 2002). A

prominent ridge is present along most of the length of

the lateral surface of each ramus (Fig. 26A, D): this

starts close to the anterodorsal-most corner of the ramus

and extends posteroventrally, forming the posterior mar-

gin of the ramus at its distal tip. A similar ridge has

been identified on chevrons from the anterior–middle

caudal vertebrae of several titanosaurs, including

Alamosaurus,Arrudatitan,Baurutitan,Epachthosaurus

and Saltasaurus (Poropat et al., 2016; Santucci &

Arruda-Campos, 2011). It is clearly absent from the

chevrons of some titanosaurs, including those of other

European taxa (i.e. Abditosaurus,Atsinganosaurus,

Lirainosaurus,Lohuecotitan and Paludititan), although

it is not always present along all of the anterior–middle

caudal series in taxa that possess this feature. Therefore,

we consider this feature a ‘local’autapomorphy of

Magyarosaurus. It is possible that this ridge provides an

attachment site for M. caudofemoralis longus (D

ıez

D

ıaz et al., 2020; Otero & Vizca



ıno, 2008; Wilhite,

2003). Distal to this ridge, each proximal ramus is trans-

versely compressed. The preserved part of the distal

blade has a subtriangular horizontal cross-section, with a

prominent posterior midline ridge that forms the apex of

this triangle, as well as a less well-developed, broader

and flatter anterior counterpart. There are no ridges on

the lateral surface of the distal blade.

LPB (FGGUB) R.0076 (Fig. 26G–I) (Individual E)

preserves the proximal part of a ramus of a chevron,

probably also from the anterior region of the tail. The

proximal articular surface is too poorly preserved to

fully determine its morphology, but it is clearly antero-

posteriorly convex and lacks a transverse groove. The

absence of this groove distinguishes the chevrons of

Magyarosaurus from those of some aeolosaurines

(Powell, 2003; Santucci & Arruda-Campos, 2011),

Epachthosaurus (Poropat et al., 2016), members of

Lognkosauria (Gonz

alez Riga et al., 2018), as well as

both Lohuecotitan (D



ıez D

ıaz et al., 2016) and

Paludititan (see below). The posterior margin of the

proximal articulation forms a small ‘hook’-like process

that overhangs the posterior surface of the ramus. A

broadly comparable feature has otherwise only been

observed in the non-titanosaurian titanosauriforms

Cedarosaurus (DMNH 39045: PDM pers. obs. 2008)

and Gobititan (IVPP 12579: PDM and PU pers. obs.

2007) and, as such, is regarded as a potential local auta-

pomorphy of Magyarosaurus. Despite its incomplete

state, it seems unlikely that the chevron was ‘closed’

proximally by a bridge of bone above the haemal canal.

There is no posteroventrally oriented ridge on the lateral

surface of LPB (FGGUB) R.0076, such as that seen in

LPB (FGGUB) R.2510, but this could be because only

the proximal-most portion of the ramus is preserved, or

might reflect serial variation given that this feature is

42 V. D



ıez D

ıaz et al.

often absent in the anterior-most chevrons of those tita-

nosaurs that otherwise display it.

Humerus. LPB (FGGUB) R.1047 (Fig. 27A–F) (left

element, Individual E) and R.2506 (Fig. 27G–L) (right

element, Assemblage F) are virtually complete and well-

preserved humeri, with the former approximately 25%

longer than the latter. LPB (FGGUB) R.1047 is abraded

at its articular ends, and the tip of the deltopectoral crest

is damaged. SZTFH Ob.3089 (Fig. 27M–P)

(Assemblage A) is a left humerus that is incomplete

proximally and distally.

The Magyarosaurus humerus has a mediolaterally

constricted shaft and articular ends that are markedly

expanded relative to the shaft. The robusticity index

(RI) is approximately 0.28–0.30. There is a small degree

of torsion between the proximal and distal ends. The

proximal end is approximately 22% wider transversely

than the distal one, and is more than twice as wide as

the midshaft. As is the case in nearly all titanosauri-

forms (Poropat et al., 2016), most of this proximal

expansion occurs along the medial margin, with little

expansion of the lateral margin relative to the shaft. The

expansion of the proximal end gives the humerus a

proximal width index (PWI) of 0.33–0.44. As such, the

proximal end of the humerus does not conform to those

of titanosaur taxa typically regarded as having more gra-

cile humeri, such as Isisaurus (PWI ¼0.31; Jain &

Bandyopadhyay, 1997), nor to those that have extremely

robust humeri, e.g. Neuquensaurus (PWI ¼0.48; Otero,

2010) and Opisthocoelicaudia (PWI ¼0.51; Borsuk-

Białynicka, 1977). Instead, in this respect, the proximal

end is more similar to those found in taxa such as

Futalognkosaurus (PWI ¼0.38; Gonz

alez Riga et al.,

2016) and Lirainosaurus (PWI ¼0.39; D

ıez D

ıaz et al.,

2013a).

In anterior view, the profile of the proximal end is

slightly sinusoidal, with this feature being most notice-

able in LPB (FGGUB) R.2506 (Fig. 27G). This condi-

tion is similar to that in titanosaur taxa such as

Mansourasaurus (Sallam et al., 2018),

Opisthocoelicaudia (Borsuk-Białynicka, 1977), and

Saltasaurus (Powell, 1992), but contrasts with the con-

vex or straight margin that characterizes most sauropods

(Gonz

alez Riga, 2003; Upchurch, 1998), including

Ampelosaurus (Le Loeuff, 2005b) and Lirainosaurus



ıez D

ıaz et al., 2013a). The proximolateral corner of

the humerus has a ‘squared’morphology in anterior/pos-

terior view, as is the case in nearly all somphospondy-

lans (Carballido et al., 2012; Wilson, 2002). By

contrast, the proximal and medial surfaces converge to

form an acute, triangular projection, similar to those

observed in Angolatitan,Mendozasaurus, and

Paralatitan (Gonz

alez Riga et al., 2018; Mateus et al.,

2011; Smith et al., 2001).

In Magyarosaurus, the humeral head is situated

closer to the medial margin than the lateral one, and it

is strongly expanded posteriorly. A marked posterior

prominence characterizes the humeral head of most

neosauropods, although it is absent in many titano-

saurs and close relatives, including Alamosaurus,

Diamantinasaurus,Jainosaurus,Rapetosaurus and

saltasaurines (Bonaparte et al., 2006; Curry Rogers,

2009; Mannion, Upchurch, Jin, et al., 2019; Poropat

et al., 2016; Smith et al., 2001; Upchurch et al.,

2015). A steeply oriented ridge extends distally and

slightly laterally from the humeral head along the pos-

terior surface, fading out at approximately the level of

the deltopectoral crest. The posterior surface of the

proximal end of the humerus is transversely convex med-

ial to this ridge, and concave lateral to it. This prominent

ridge is not present in other titanosaurs or close relatives,

e.g. Diamantinasaurus,Mansourasaurus,Rapetosaurus

and Saltasaurus (Curry Rogers, 2009; Poropat, Upchurch,

et al., 2015;Powell,2003;Sallametal.,2018), including

European taxa that have well-preserved humeri, i.e.

Ampelosaurus,Atsinganosaurus,Garrigatitan and

Lirainosaurus (D



ıez D



ıaz et al., 2013a,2018,2021;Le

Loeuff, 2005b). Therefore, this ridge is regarded as an

autapomorphy of Magyarosaurus.

The anterior surface of the proximal third of the

humerus is concave, with the coracobrachialis fossa

bounded laterally by the deltopectoral crest. The

humerus of Magyarosaurus displays an autapomorphic

morphology in this region, with the coracobrachialis

fossa divided into a small, shallow dorsomedial region

and a larger, deeper ventrolateral region by an obliquely

curved, stepped ridge. Such division of this fossa is not

present in other titanosaurs, including European taxa

with well-preserved humeri (D

ıez D



ıaz et al., 2013a,

2018,2021; Le Loeuff, 2005b). Given that no other

tuberosities or distinct rugosities are present in this area,

it is possible that this ridge served as the insertion site

for M. coracobrachialis. Distally, the coracobrachialis

fossa is demarcated by a curved ridge.

The deltopectoral crest is anteromedially deflected,

slightly overhanging the anterior concavity. It is simi-

larly oriented in most titanosauriforms, including

Abditosaurus,Atsinganosaurus and Lirainosaurus



ıez D

ıaz et al., 2013a,2018; Vila et al., 2022),

although it projects mainly anteriorly in saltasaurines,

Diamantinasaurus,Mansourasaurus and Rinconsauria

(Mannion et al., 2013; Poropat et al., 2016; Sallam

et al., 2018). The most prominent part of the deltopec-

toral crest is missing in most Magyarosaurus specimens,

with the exception of LPB (FGGUB) R.2506 (Fig. 27I).

Revision of Romanian sauropod dinosaurs 43

In this specimen, the crest widens transversely towards

its distal end. This approximate doubling in width is a

feature of most saltasaurids, as well as some members

of Lognkosauria and Rinconsauria (Gonz

alez Riga

et al., 2018; Poropat et al., 2016; Wilson, 2002), and it

also characterizes Ampelosaurus,Atsinganosaurus and

Lirainosaurus (Le Loeuff, 2005b;D



ıez D

ıaz et al.,

2013a,2018). The deltopectoral crest terminates well

above the midlength of the humerus in Magyarosaurus.

The lateral surface of the deltopectoral crest is rugose,

becoming heavily striated distally, at least in LPB

(FGGUB) R.2506, which could indicate the insertion

area for M. deltoideus clavicularis and/or M. pectoralis

(Otero, 2018; Voegele et al., 2020). A short vertical

ridge extends from the proximal end of the humerus

along the lateral margin of the posterior surface. Such a

ridge is present in some other titanosaur taxa (e.g.

Diamantinasaurus,Epachthosaurus,Rapetosaurus and

Saltasaurus; Poropat et al., 2016). At the approximate

level of the most prominent portion of the deltopectoral

crest, there is a well-developed bulge on the lateral mar-

gin of the posterior surface that is visible in anterior

view. This characterizes nearly all titanosaurs (Mannion,

Upchurch, Schwarz, et al., 2019) and probably repre-

sents the insertion site for M. scapulohumeralis anterior

(Borsuk-Białynicka, 1977; Upchurch et al., 2015;

Voegele et al., 2020) or possibly M. deltoideus clavicu-

laris (Otero, 2018). There is a further prominence

approximately equidistant from the medial and lateral

margins of the posterior surface, level with the distal tip

of the deltopectoral crest. This presumably represents

the insertion site for M. latissimus dorsi, which charac-

terizes several other titanosaurs, namely Alamosaurus,

Opisthocoelicaudia,Patagotitan,Rukwatitan and salta-

saurines (Borsuk-Białynicka, 1977; Carballido et al.,

2017;D’Emic, 2012; Gorscak et al., 2014; Otero, 2010;

Voegele et al., 2020), as well as the European taxa

Atsinganosaurus,Garrigatitan and Lirainosaurus (D

ıez

D

ıaz et al., 2013a,2018,2021). However, only in

Garrigatitan and Lirainosaurus is it also situated closer

to the midline than the lateral margin (VDD pers. obs.),

and in those two it is a hypertrophied longitudinal bulge

and a subtly expressed ridge, respectively: thus, the

shape, development, and central position of this feature

is regarded as an autapomorphy of Magyarosaurus.In

the humeri referred to Magyarosaurus, this prominence

is confluent with the distal-most part of the subvertical

posterior ridge that extends from the humeral head.

Distal to the base of the deltopectoral crest, the

humerus constricts markedly into an anteroposteriorly

compressed oval shaft (the eccentricity index [ECC]

ranges from 1.26 to 1.44). The anterior surface of the

shaft is relatively flat, whereas the posterior one is

transversely convex, and these surfaces slightly con-

verge medially to form a rounded but angular medial

margin. In anterior/posterior view, the medial margin of

the shaft is strongly concave. By contrast, the lateral

margin is only slightly concave, or nearly straight, giv-

ing the humerus a canted aspect, resembling the morph-

ology in most titanosaurs, including Ampelosaurus,

Atsinganosaurus and Lirainosaurus (Le Loeuff, 2005b;

D

ıez D



ıaz et al., 2013a,2018).

The shaft expands gradually transversely as it

approaches the distal end of the humerus. Its degree of

expansion is moderate, as expressed by a distal width

index (DWI) of 0.31–0.36, similar to that reported in

Ampelosaurus (DWI ¼0.33), Mendozasaurus (DWI ¼

0.31), Isisaurus (DWI ¼0.30) and Lirainosaurus (DWI

¼c. 0.30) (Gonz

alez Riga, 2003; Le Loeuff, 2005b;



ıez D

ıaz et al., 2013a; Wilson et al., 2009). The distal

articular condyles extend to some degree onto the anter-

ior surface of the humerus, which is a feature primarily

restricted to taxa close to or within Saltasauridae,

including Alamosaurus,Opisthocoelicaudia and salta-

saurines (Wilson, 2002). As is also the case in nearly all

titanosaurs (D’Emic, 2012), the anterior surface of the

distal lateral condyle of the Magyarosaurus humerus is

undivided. On the anterior surface, the distal condyles

are separated by an intercondylar groove that deepens

laterally. The medial part of this intercondylar groove is

rugose. The lateral surface of the lateral condyle is exca-

vated by a longitudinal depression that faces anterolater-

ally, such that this region is divided into an anteriorly

projecting lateral distal condyle, and a laterally projec-

ting ectepicondyle. The ectepicondyle is wide and

rounded, with a flattened, scarred, anterolaterally facing

surface. In most preserved Magyarosaurus humeri, the

lateral surface of the distal end is slightly abraded;

nevertheless, a distinct, elongated and rugose anterolat-

eral surface can be discerned.

The distal condyles are shallowly but clearly divided

from one another along the distal surface, at least

in LPB (FGGUB) R.1047, such that the distal

margin of the humerus is concave in anterior/posterior

view. This is a feature that is otherwise known only

in Abditosaurus,Alamosaurus,Lirainosaurus,

Opisthocoelicaudia and saltasaurines (D



ıez D

ıaz et al.,

2013a; Vila et al., 2022; Wilson, 2002), as well as pos-

sibly Paralititan (Smith et al., 2001, fig. 2). In distal

end view, these condyles are oval to subquadrangular in

outline, with anteromedially to posterolaterally oriented

long axes. The medial condyle is somewhat smaller, but

more distally projected, compared to its lateral counter-

part. Such bevelling of the distal end is also present in

several other titanosaurs, including saltasaurines and

some members of Lognkosauria (Mannion, Upchurch,

44 V. D



ıez D

ıaz et al.

Jin, et al., 2019), as well as Lirainosaurus (D



ıez D

ıaz

et al., 2013a, fig. 3b), Mansourasaurus (Sallam et al.,

2018, fig. 2), and Rukwatitan (Gorscak et al., 2014, fig.

10). On the posterior surface of the distal end of the

Magyarosaurus humerus, the olecranon (¼supracondy-

lar or anconeal) fossa is wide and deep, bordered by

well-developed supracondylar crests, as is the case in

most titanosaurs (Upchurch et al., 2004,2015). The

medial supracondylar crest is more acute, and has a

slightly depressed posteromedial flattened surface. The

proximal portion of this shallowly depressed area, as

well as the medial surface of the medial supracondylar

crest, is rugose.

Ulna. The left ulna belonging to Assemblage A

(SZTFH Ob.3100), the only such element known for

Magyarosaurus, is essentially complete in terms of its

length, but there is material missing from the proximal

end, as well as a small anterior portion of the distal one

(Fig. 28). Despite its incomplete nature, the ulna is

clearly gracile, with an estimated maximum proximal

mediolateral width to ulna length ratio of less than

0.3. As such, it differs notably from the robust

ulnae of many titanosaurs (e.g. Diamantinasaurus,

Epachthosaurus,Lirainosaurus,Opisthocoelicaudia and

Saltasaurus), in which this ratio equals or exceeds 0.4

(Curry Rogers, 2005; Mannion et al., 2013; Wilson,

2002). In the Magyarosaurus ulna, the near-complete

anterolateral proximal process is short, whereas the

incomplete anteromedial process was clearly much lon-

ger. The posterior process is prominent and of compar-

able size to the anterolateral one, as is the case in most

titanosaurs (Upchurch et al., 2015; Poropat et al., 2016),

including Lirainosaurus and Lohuecotitan (D



ıez D

ıaz

et al., 2013a,2016). Relative to the long axis of the

anterolateral process, the posterior process is deflected

posterolaterally. The proximal surface is mostly incom-

plete, except for the anterolateral process, and this miss-

ing material might have exaggerated the apparent

prominence of the olecranon process.

The anterior surface (radial fossa) is the most strongly

concave of the surfaces of the proximal third of the

ulna, but the anterolateral and posterior surfaces are also

transversely concave. The anteromedial process is sup-

ported by a ridge that extends distally for most of the

length of the ulna. This anteromedial ridge appears to

bifurcate at approximately midlength of the ulna, or this

could be interpreted as the development of a second

ridge at this point. This results in a more posteriorly

placed ridge forming the anteromedial margin of the

shaft, and an anterior ridge that is deflected laterally and

terminates a short distance from the distal end.

Following Otero (2018) and Voegele et al. (2020), we

interpret this anterior lateral (interosseous) ridge to be

the origin of M. pronator quadratus that in life would

be connected by ligaments to a similar structure on the

radius. The distal end of this interosseous ridge lies

almost equidistant from the medial and lateral margins

of the ulna. Although the presence of an interosseous

ridge on the ulna is typical for virtually all sauropods,

its slanting orientation in Magyarosaurus appears to be

unique, given that this ridge typically follows the long

axis of the shaft in other taxa; consequently, we regard

this as an autapomorphy of Magyarosaurus.

The distal end of the ulna expands posteriorly. In

Malawisaurus and some saltasaurids (and closely related

taxa), the ulna lacks such a distal expansion (D’Emic,

2012), but Magyarosaurus shares a distally expanded

ulna with the majority of titanosaurs, including taxa

such as Lirainosaurus (D



ıez D

ıaz et al., 2013a),

Opisthocoelicaudia (Borsuk-Białynicka, 1977), and

Mendozasaurus (Poropat et al., 2016; Gonz

alez Riga

et al., 2018). In Magyarosaurus, the distal end of the

ulna has an oval outline, but it is possible that it was

‘comma’-shaped, with an anterior process that has not

been preserved. The morphology of the distal ulna gen-

erally resembles those of Diamantinasaurus,

Neuquensaurus,Rapetosaurus, and Lirainosaurus

(Curry Rogers, 2009, fig. 38E; D



ıez D

ıaz et al., 2013a,

fig. 3.5-6; Otero, 2010, fig. 4; Poropat, Upchurch, et al.,

2015, fig. 11H), but it is less transversely compressed

than those of some specimens referred to Ampelosaurus

(e.g. MDE C3-1296; Le Loeuff, 2005b, fig. 4.15a-b),

Atsinganosaurus (D



ıez D

ıaz et al., 2018, fig. 10L–P),

Garrigatitan (D

ıez D

ıaz et al., 2021, fig. 8), and

Lohuecotitan (D



ıez D

ıaz et al., 2016, fig. 5B).

Radius. The best-preserved radii are LPB (FGGUB)

R.1049 and R.1060 (Fig. 29A–F), which are portions of

a single left radius from Individual E, and SZTFH

Ob.3101, which is a left radius from Assemblage A

(Fig. 29G–L). LPB (FGGUB) R.1049 comprises the

proximal two-thirds (in two pieces), whereas LPB

(FGGUB) R.1060 preserves the distal end. There is a

small amount of material missing in between these seg-

ments, and the posterior surface is poorly preserved

throughout much of the element’s length. However,

SZTFH Ob.3101 is fairly complete, only missing small

fragments close to the distal edge, and the lateral border

of the proximal end.

The radius to humerus length ratio varies from

c. 0.65 (Individual E) to c. 0.72 (Assemblage A). The

ratio of Individual E lies within the typical titanosaur

range (Mannion et al., 2013; Mannion, Upchurch,

Schwarz, et al., 2019), whereas that of Assemblage A is

slightly higher. However, it is important to note that the

elements from Assemblage A might not be from the

same individual. There is a strong degree of torsion

Revision of Romanian sauropod dinosaurs 45

between the proximal and distal ends of the radius,

accentuated by the prominent interosseus ridge (see

below), such that their long axes are approximately 45

to one another. A similarly twisted radius also character-

izes several other titanosaurs, including Epachthosaurus,

Malawisaurus,Rapetosaurus,Savannasaurus and

Uberabatitan (Silva Junior et al., 2019; Mannion et al.,

2013; Poropat et al., 2016), whereas the shaft is much

straighter in most other taxa, including the only other

European taxon with a sufficiently well-preserved

radius, Ampelosaurus (Le Loeuff, 2005b). The proximal

articular surface is slightly deformed in LPB (FGGUB)

R.1049 (Fig. 29D), but has a ‘comma’-shaped outline,

with a prominent medial process that is anteroposteriorly

narrower than the remainder of the proximal end, as is

also the case in nearly all titanosauriforms (Poropat

et al., 2016; Upchurch et al., 2015). This also appears to

be the case in SZTFH Ob.3101, although the lateral por-

tion is incomplete (Fig. 29G). It is not possible to deter-

mine whether a medial ridge for attachment of the

combined M. biceps brachii and M. brachialis inferior

(see Otero, 2018; Upchurch et al., 2015; Voegele et al.,

2020) is present along the proximal quarter of the

shaft.

A proximodistally oriented interosseus ridge is pre-

sent on the posterolateral surface and extends for most

of the radius length (though note that preservation is

poor at its proximal part and along the distal third of

the radius in LPB [FGGUB] R.1049), as is the case in

all titanosaurs (Curry Rogers, 2005), as well as in sev-

eral non-titanosaurian titanosauriform taxa (Mannion

et al., 2013). As preserved, this ridge remains largely on

the posterolateral margin throughout the length of the

radius, despite the torsion between the proximal and dis-

tal ends. This ridge marks the probable insertion of M.

pronator quadratus (Otero, 2018; Voegele et al., 2020).

In the radius of Individual E, a subtle, vertical ridge is

also present on the posterior surface, close to the medial

margin, extending along the middle third of the shaft.

This ridge appears to be absent in SZTFH Ob.3101. The

posterolateral margin of the midshaft also forms a ridge

in both radii.

The lateral two-thirds of the distal end surface are

strongly bevelled (approximately 30) relative to a plane

perpendicular to the proximodistal axis of the radius.

Comparable bevelling characterizes the radius of nearly

all titanosaurs, although in saltasaurines and

Argyrosaurus this extends across the entire distal surface

(Mannion & Otero, 2012; Poropat et al., 2016; Wilson,

2002). The posterior surface of the distal end is gently

concave, forming a fossa between two subtle condyles.

This morphology characterizes most neosauropods, but

is lost in some titanosaurs, including Malawisaurus,

Opisthocoelicaudia, and saltasaurines (Upchurch et al.,

2015). In distal end view, the Magyarosaurus radius has

a flat anterior margin and the lateral half is notably

wider anteroposteriorly than the medial half (Fig. 29B),

especially when compared to the radius of Abditosaurus

(Vila et al., 2022). This feature is a local autapomorphy

of Magyarosaurus that also occurs in Neuquensaurus

(Upchurch et al., 2015, fig. 9P).

Metacarpals. Several fragments of metacarpals are pre-

served as part of Individual E (Fig. 30), and there is one

metacarpal from each of Assemblage A and F (Fig. 31),

although their incomplete nature makes their identifica-

tion difficult.

LPB (FGGUB) R.0074 (Individual E) is the distal

end of a metacarpal (Fig. 30A–E). In distal view, it has

a transversely elongated elliptical or ‘D’-shaped outline,

with a mildly concave ventral margin. The distal articu-

lar surface does not extend onto the dorsal surface of

the metacarpal.

LPB (FGGUB) R.1864 (Individual E) is the proximal

end of a possible metacarpal III (Fig. 30F–K). It has a

subtriangular outline in proximal view, although the

cross-section of the shaft becomes elliptical a short dis-

tance from this end.

LPB (FGGUB) R.1053 (Individual E) is another distal

end of a metacarpal (Fig. 30L–P). It is poorly preserved,

with some incomplete margins, but probably represents

a left metacarpal IV. It has a transversely elongate trap-

ezoidal outline in distal view, with a dorsal margin that

is wider than the ventral one. The latter margin is mildly

concave along its midline. One of the distal condyles of

LPB (FGGUB) R.1053 is incomplete, but its lateral

margin appears to be straight, whereas the other margin

clearly slopes to face partly ventrally, forming a ventro-

medial ridge.

LPB (FGGUB) R.1862 (Individual E) is the distal

end of a metacarpal (Fig. 30Q–V). It has a trapezoidal

outline in distal view, with a dorsal margin that is

slightly wider transversely than the ventral one. Such a

morphology characterizes the distal end of metacarpal

IV in many titanosaurs, e.g. Diamantinasaurus,

Epachthosaurus,Rapetosaurus and Saltasaurus (Poropat

et al., 2016), and thus we tentatively identify LPB

(FGGUB) R.1862 as such. One of this specimen’s distal

condyles (probably the medial) is slightly taller dorso-

ventrally than the other. This taller margin is deflected

such that it partly faces ventrally. The distal articular

surface is convex dorsoventrally, but it does not extend

onto the dorsal surface of the metacarpal. This restric-

tion of the distal articulation is consistent with the

derived condition in nearly all titanosauriforms

(D’Emic, 2012; Salgado et al., 1997).

46 V. D



ıez D

ıaz et al.

LPB (FGGUB) R.0077 (Individual E) consists of the

proximal end of a right metacarpal V (Fig. 30W-A’). In

proximal view, it has a dorsoventrally tall, subtriangular

outline that is more widely rounded ventrally, with only

one long and flat articular surface (the latter presumably

for metacarpal IV).

SZTFH Ob.3096 (Assemblage A) consists of the dis-

tal half of a metacarpal (Fig. 31A–D). The bevelling of

the distal articular end is similar to that present in LPB

(FGGUB) R.0074, although the outline of the distal

articular surface is different. The cross-section of the

shaft of SZTFH Ob.3096 is sub-circular. We are unable

to identify where LPB (FGGUB) R.0074 and SZTFH

Ob.3096 belong in the metacarpus.

LPB (FGGUB) R.2509 (Assemblage F) lacks both its

proximal and distal ends, limiting its identification and

interpretation (Fig. 31E–F). The preserved proximal and

distal parts are twisted relative to one another. A ridge

is present on the ventral surface, close to the midline of

the distal (?) portion. The medial and lateral margins of

the shaft are mostly straight in dorsal view.

Femur. Assemblage A preserves a partial left femur

(SZTFH Ob.3088; Fig. 32A–C). The proximal end is

missing, and the distal articulation is damaged.

Individual E preserves a pair of complete femora (left,

LPB [FGGUB] R.1046 [Fig. 32D–I] and right, LPB

[FGGUB] R.1992 [Fig. 32J–M]), and Assemblage F

includes two right femora (LPB [FGGUB] R.2507 [Fig.

32O–T] and R.2508 [Fig. 32U–V]). The proximal end

of LPB (FGGUB) R.2508 is missing, and the distal con-

dyles are eroded, whereas LPB (FGGUB) R.2507 is

largely complete.

Based on Individual E, the humerus to femur length

ratio for Magyarosaurus is 0.8. This ratio is similar to

that of the titanosaurs Diamantinasaurus (0.79; Poropat,

Upchurch, et al., 2015), Futalognkosaurus (0.79; Calvo,

2014), and Rapetosaurus (0.80; Curry Rogers, 2009),

whereas other titanosaurs for which this can be assessed

have notably lower (Opisthocoelicaudia ¼0.72 [Borsuk-

Białynicka, 1977]; Jainosaurus ¼0.74 [Wilson et al.,

2011]) or higher values (Dreadnoughtus ¼0.84

[Lacovara et al., 2014]; Epachthosaurus ¼0.88

[UNPSJB-PV 920; PDM and PU pers. obs., 2013]).

The femoral head projects primarily medially, such

that it is approximately perpendicular to the long axis of

the shaft. This is the plesiomorphic eusauropod condi-

tion, that also characterizes some titanosaurs (Curry

Rogers, 2005; Upchurch et al., 2004), including

Ampelosaurus (Le Loeuff, 2005b), Diamantinasaurus

(Poropat, Upchurch, et al., 2015), Lohuecotitan (D

ıez

D

ıaz et al., 2016), and Mendozasaurus (Gonz

alez Riga

et al., 2018). By contrast, most titanosaurs, including

Lirainosaurus (D

ıez D



ıaz et al., 2013a),

Opisthocoelicaudia (Borsuk-Białynicka, 1977),

Patagotitan (Carballido et al., 2017), and saltasaurines

(Curry Rogers, 2005), have a more dorsally directed

femoral head that is separated from the poorly devel-

oped greater trochanter by a shallow sulcus on the prox-

imal surface. Proximally, the lateral margin of the

Magyarosaurus femur is medially deflected, and there is

a well-developed and robust lateral bulge (although

these two features are less developed in the Assemblage

F specimens [Fig. 31O–V]). The lateral bulge of

Magyarosaurus is similar to that of the femoral frag-

ment referred to Abditosaurus (Vila et al., 2022), and is

more prominent than in Ampelosaurus,Lirainosaurus,

and Lohuecotitan (D



ıez D

ıaz et al., 2013a,2016;Le

Loeuff, 2005b). As in all titanosaurs (D

ıez D

ıaz et al.,

2013a; Mannion et al., 2013; Otero, 2010), the femur of

Magyarosaurus is characterized by a trochanteric shelf

located on the proximal third of the lateral portion of

the posterior surface, commencing below the greater tro-

chanter. This mediolaterally wide, but shallow shelf dis-

appears at the same level as the proximal tip of the

fourth trochanter. The latter process is a well-developed,

mediolaterally expanded, longitudinal ridge located on

the posterior surface of the shaft, close to the medial

margin. Its prominence distinguishes it from the barely

discernible fourth trochanters of saltasaurines (Mannion,

Upchurch, Jin, et al., 2019; Otero, 2010) and

Atsinganosaurus (D

ıez D

ıaz et al., 2018). The distal end

of the fourth trochanter of the Magyarosaurus femur

does not extend as far as the femoral midlength. The

medial surface of the femur, level with the fourth tro-

chanter, is rugose, probably for the insertion of M. cau-

dofemoralis longus (D

ıez D

ıaz et al., 2020; Ibiricu

et al., 2013).

The femoral shaft is mostly straight, with little curva-

ture. The anterior surface of the shaft is mildly convex

transversely, whereas the posterior one is flat. This

anterior convexity is primarily generated by a faint, lon-

gitudinal linea intermuscularis cranialis, which fades out

towards the distal end of the shaft. This feature is most

marked in the middle section of the shaft of LPB

(FGGUB) R.2507. Such a crest is absent in most sauro-

pods, but is present in several titanosaurs, including

Alamosaurus (D’Emic, 2012), diamantinasaurians

(Hocknull et al., 2021; Poropat, Upchurch, et al., 2015),

Lohuecotitan (D

ıez D

ıaz et al., 2016), Pellegrinisaurus

(Cerda et al., 2021), saltasaurines (Otero, 2010), and

Uberabatitan (Silva Junior et al., 2019). Although often

a subtle feature, it is clearly absent in most other titano-

saurs (Mannion, Upchurch, Schwarz, et al., 2019;

Poropat et al., 2016), including Atsinganosaurus and

Lirainosaurus (D



ıez D

ıaz et al., 2013a,2018). The

Magyarosaurus femoral shaft is only slightly

Revision of Romanian sauropod dinosaurs 47

anteroposteriorly compressed at midlength, with a sub-

circular cross-section (the ECC value ranges from 1.14

to 1.4). This midshaft eccentricity value is notably lower

than that in all other macronarians (Mannion et al.,

2013; Wilson & Sereno, 1998), including penecontem-

poraneous European taxa that preserve femora, i.e.

Garrigatitan (1.5), Lohuecotitan (1.6), Paludititan

(>1.7, although the femoral fragment is close to the dis-

tal end), Abditosaurus (>1.85), and especially

Ampelosaurus (>2.0), Lirainosaurus (mean average of

2.2), and Atsinganosaurus (2.6) (D



ıez D

ıaz et al.,

2013a,2016,2018,2021; Le Loeuff, 2005b; Vila et al.,

2022). As such, we regard a subcircular femoral mid-

shaft as an autapomorphy of Magyarosaurus.

The posterior surface of the distal condyles is dam-

aged in the available specimens. The distal condyles are

proximomedially bevelled, such that the fibular condyle

extends further distally than the tibial one (Fig. 32E, I,

P, T), a derived state that characterizes most titanosaurs

(Wilson, 2002). Intercondylar fossae separate the tibial

and fibular condyles on both the anterior and posterior

surfaces, continuous with the deep and wide intercondy-

lar groove on the distal articular surface itself. A shal-

lower sulcus is present between the main fibular

condyle and the posterolaterally oriented fibular epicon-

dyle, as is the case in nearly all eusauropods, with the

exception of saltasaurines (Beeston et al., 2024;

Carballido et al., 2017). The tibial condyle is trans-

versely narrower than the fibular condyle (tibial to fibu-

lar condyle width ratio ranges from 0.78 to 0.96),

similar to other titanosaurs, including Lirainosaurus,

Lohuecotitan, and some femora ascribed to

Ampelosaurus, in which this ratio is 0.8 (Poropat

et al., 2016; this study). In Magyarosaurus, the long

axis of each distal condyle is anteroposteriorly oriented,

and the articulations do not curve noticeably onto

the anterior surface. The latter feature contrasts

with the condition in saltasaurines, Alamosaurus,

Opisthocoelicaudia and Lirainosaurus (D

ıez D

ıaz et al.,

2013a; Wilson, 2002).

Tibia. The only specimen known to document this skel-

etal element in Magyarosaurus, LPB (FGGUB) R.2299

(belonging to Individual E), comprises the poorly pre-

served proximal ends of a left tibia and fibula in semi-

articulation, which are obscured by matrix in places

(Fig. 33). However, this specimen is important in that it

demonstrates clear anatomical differences with NHMUK

R.3853, i.e. the paralectotype of ‘Magyarosaurus’hun-

garicus (see below). By contrast with the latter species,

the long axis of the proximal articular surface of the

Magyarosaurus dacus tibia is oriented mediolaterally

(Fig. 33A), and the cnemial crest projects laterally,

roughly aligned with this long axis (Fig. 33C). The

proximal end of the tibia has an approximately ‘D’-

shaped outline, with a flat and transversely wider poster-

ior margin (Fig. 33A), similar to that of many titano-

saurs, including Abditosaurus (Vila et al., 2022). It

differs from the transversely compressed proximal tibiae

of Atsinganosaurus and Lirainosaurus (D

ıez D

ıaz et al.,

2013a,2018), as well as the subcircular morphology of

the proximal end that characterizes Ampelosaurus and

Lohuecotitan (D

ıez D

ıaz et al., 2016; Le Loeuff,

2005b).

Fibula. Three left fibulae are represented within

Assemblage A (two registered under SZTFH Ob.3086,

and Ob.3102). SZTFH Ob.3086a (Fig. 34A–E) lacks the

distal end, but otherwise is nearly complete. SZTFH

Ob.3086b (Fig. 34F–K) is largely complete, although

the proximal and distal ends are poorly preserved.

SZTFH Ob.3102 (Fig. 34L–O) lacks the proximal end

and its distal-most tip. As noted above, Individual E

includes the poorly preserved proximal end of a left fib-

ula (LPB [FGGUB] R.2299) (Fig. 33D, E).

In lateral view, the shaft of SZTFH Ob.3086a is sig-

moidal (Fig. 34A), whereas it is straighter in SZTFH

Ob.3086b and Ob.3102 (Fig. 34F, L). Otherwise, the

fibulae are broadly comparable in shape. There is clear

torsion of approximately 20between the long axes of

the proximal and distal ends. In general, the lateral sur-

face is anteroposteriorly convex, whereas the medial

surface is flat, giving the fibula a D-shaped cross-sec-

tion throughout most of its length. In lateral view, the

proximal end is anteroposteriorly expanded relative to

the shaft, with this expansion primarily occurring along

the posterior margin.

At the proximal end, the fibula has an anteroposteriorly

elongate comma-shaped profile, narrowing anteriorly (Fig.

34C, H). The proximal articular surface is mildly bevelled

along its anterior half, such that it faces slightly anteriorly.

The posterior margin of the proximal half forms a ridge

that broadens towards its distal end. Although slightly

incomplete, an anteromedial proximal process is clearly

present in SZTFH Ob.3086a and b (and probably also LPB

[FGGUB] R.2299). Such a process characterizes nearly all

somphospondylans (D’Emic, 2012; Mannion et al., 2013;

Wilson & Upchurch, 2003). There is no anterolateral tro-

chanter, such as that present in Jainosaurus (Wilson et al.,

2009), Laplatasaurus (Gallina & Otero, 2015),

Lohuecotitan (D

ıez D

ıaz et al., 2016), Mendozasaurus

(Gonz

alez Riga, 2003), and Uberabatitan (Salgado &

Carvalho, 2008). There is a break of slope on the lateral

surface of SZTFH Ob.3086a and Ob.3102, close to the

anterior margin, just posterior to the distal portion of the

anteromedial projection. The medial surface of the prox-

imal end of the fibula is only subtly striated and is not

48 V. D



ıez D

ıaz et al.

raised, as is also the case in most titanosauriforms

(D’Emic, 2012).

Although none of the fibulae preserve their entire length,

the lateral trochanter almost certainly did not extend to

midlength of the shaft. This trochanter is composed of a

prominent, proximodistally elongate, posterior ridge (less

well developed in SZTFH Ob.3102), and a shorter, less

prominent anterior ridge. There is a striated, gentle concav-

ity in between these two ridges. A similar morphology

characterizes most titanosaurs (Upchurch, 1998), includ-

ing Epachthosaurus (Mart

ınez et al., 2004), Lohuecotitan



ıez D



ıaz et al., 2016), Opisthocoelicaudia (Borsuk-

Białynicka, 1977), and Saltasaurus (Powell, 1992). By

contrast, there is a single oval trochanter in a small number

of titanosaurs, including Dreadnoughtus (Ullmann &

Lacovara, 2016) and Jainosaurus (Wilson et al., 2009). A

subtle ridge is present on the anterior margin of the lateral

surface of the fibula of Magyarosaurus, with its proximal

tip situated at approximately the level of the distal end of

the lateral trochanter. This ridge is directed steeply poster-

odistally, and it fades out before it reaches the distal-most

preserved portions of the fibula. It is absent in other titano-

saurs, including European taxa with well-preserved fibu-

lae, i.e. Abditosaurus (Vila et al., 2022), Ampelosaurus

(VDD pers. obs., 2024), Lirainosaurus (D



ıez D

ıaz et al.,

2013a), and Lohuecotitan (VDD pers. obs., 2014). As

such, we regard this ridge as an autapomorphy of

Magyarosaurus. The anterior margin of the distal third of

the shaft, as preserved, forms a ridge. Distally, the fibula

expands anteroposteriorly. In distal view, the fibula has a

tear-drop-shaped outline, with a flat medial surface and an

acute anterior edge (Fig. 34J).

Titanosauria Bonaparte & Coria, 1993

Lithostrotia Upchurch, Barrett & Dodson, 2004

Eutitanosauria Sanz, Powell, Le Loeuff, Mart

ınez &

Pereda Suberbiola, 1999

Petrustitan n. gen.

(Figs. 35–40)

Type species. Petrustitan hungaricus (Huene, 1932)n.

comb. (see below).

Etymology. The genus name, which translates roughly

as ‘stone titan’, derives from the Greek words p

etra

(pesqa), meaning ‘stone, rock’, alluding both to the

place of its type locality, S^

anpetru (a shorthand for

Saint Peter, in Romanian) village in the central Hat

,eg

Basin (Fig. 3), as well as to the origin of the holotype,

coming from the rocky outcrops of the S^

ınpetru

Formation exposed in this area, and the Greek word

‘titan’, a term often used to name titanosaurian

sauropods.

Generic diagnosis. As for the type (and only known)

species.

Petrustitan hungaricus (Huene, 1932) n. comb.

1932 Magyarosaurus hungaricus; Huene: 269

1970 Titanosaurus hungaricus: Steel: 77

Lectotype and paralectotype. NHMUK R.3853,

Individual B: associated left fibula (holotype) and tibia

(paratype).

Comment. Although not formally designating a holo-

type, Huene (1932, p. 269) referred only one specimen,

an almost completely preserved fibula (NHMUK

R.3853), to his new taxon ‘Magyarosaurus’hungaricus.

Moreover, he briefly described this specimen (albeit

incorrectly citing it as Brit. Mus. N. H. No. 3833 in the

text as well as in the caption to his plate 47), and briefly

compared it to other titanosaur fibulae: thus, this speci-

men is here chosen as the lectotype of Petrustitan hun-

garicus (Huene, 1932) n. comb. The left fibula

NHMUK R.3853 described by Huene can be positively

identified based on its figure (Huene, 1932, pl. 47, fig.

1a–c) and his comment that it has the same number as

one of the other titanosaur specimens from

Transylvania: the paralectotype left tibia NHMUK

R.3853 (also briefly described and figured as Brit. Mus.

N. H. R. 3853 by Huene, 1932, pp. 273–274, pl. 48, fig.

3a–c). As we argued above, these two elements were

almost certainly found associated and belong to the

same individual, identified here as Individual B.

Type locality. Sibis

,el River Valley, south of S^

anpetru

locality, central Hat

,eg Basin, Hunedoara County,

Romania (Fig. 1E). More precise locality information is

unavailable.

Type stratum and age. The holotype comes from the

stratotype section of the uppermost Cretaceous continen-

tal S

ınpetru Formation. The age of this succession was

estimated to the early to late Maastrichtian (Csiki-Sava

et al., 2016; Panaiotu & Panaiotu, 2010). As more pre-

cise locality data are not known, neither is a more pre-

cise age assessment possible for the type locality,

although it is most probably located within the informal

lower subunit of the formation (dated as early to early

late Maastrichtian), because the upper subunit is almost

completely barren as far as vertebrate remains are con-

cerned (Csiki-Sava et al., 2016).

Referred specimens. Individual G: six caudal vertebrae,

although only five appear to be accessioned (MCDRD

255, MCDRD 266, MCDRD 267, MCDRD 268,

MCDRD 269) and only the latter four specimens can

now be located, right ulna (MCDRD 149), right radius

Revision of Romanian sauropod dinosaurs 49

(MCDRD 150), right metacarpal I (MCDRD 152), right

ilium (MCDRD 148), left fibula (MCDRD 153).

Locality and distribution. The geographical and strati-

graphical locus of Individual G is broadly consistent

with those of the type specimen, but is known more pre-

cisely: R^

apa Mocioconilor (¼Mocioconilor ravine),

informal lower subunit of the S^

anpetru Formation (early,

or earliest late, Maastrichtian), Sibis

,el Valley, between

anpetru and S

acel villages (Fig. 1E; Csiki, Grigorescu,

et al., 2010, fig. 1C; Csiki-Sava et al., 2016; Grigorescu,

1992; Therrien et al., 2009).

Revised species diagnosis. Petrustitan hungaricus can

be diagnosed on the basis of one autapomorphy

(markedwithanasterisk),aswellastwo‘local’auta-

pomorphies: (1) tibia drawn out into slender laterodis-

tal process that inserts into a socket on the medial

surface of the distal fibula; (2) proximal end of the fib-

ula with a subtriangular outline; (3) posterodistally

oriented accessory ridge on lateral surface of fibula,

extending from midlength to one-quarter of fibula

length from distal end.

Justification for referrals. Individuals B and G can be

compared directly via the fibula. These fibulae share

autapomorphies 2 and 3 above, although the posterior

expansion of the proximal end is less well marked in

Individual G.

It is worth noting that Huene (1932, p. 274) tenta-

tively referred a number of other Transylvanian titano-

saur specimens to his new species ‘Magyarosaurus’

hungaricus besides the NHMUK R.3853 fibula, mainly

based on their assessed larger size and Huene’s opinion

that ‘M.’hungaricus was ‘THE large’Transylvanian

Figure 35. Petrustitan hungaricus n. gen., posterior caudal vertebrae. A, posterior caudal vertebrae reported by Groza (1983).

MCDRD 266 (Individual G) in B, right lateral; C, anterior; D, posterior; and E, ventral views. MCDRD 269 (Individual G) in F,

right lateral; G, anterior; H, posterior; and I, ventral views. MCDRD 267 (Individual G) in J, right lateral; K, anterior; L, posterior;

and M, ventral views. MCDRD 268 (Individual G) in N, right lateral; O, anterior; P, posterior; and Q, ventral views. R, MCDRD

266, 269, 267 and 268 (from left to right) in articulation, left view. Abbreviations: PO, postzygapophysis, PRE, prezygapophysis.

Scale bar equals 100 mm.

50 V. D



ıez D

ıaz et al.

titanosaur. However, from these specimens, only the

NHMUK R.3853 tibia can be referred with any certainty

to this taxon based on its highly probable association

with the lectotype fibula. The other specimens (dorsal

ribs, caudal centra, sternum, coracoid, ulna, metacarpal,

pubis, femur, tibiae; housed in the London and

Budapest collections) lack information on their potential

field association with the lectotype, and either do not

overlap with the lectotype material or else do not share

its autapomorphies. Indeed, some of these specimens are

here identified as parts of other, more complete skeletal

associations, designated as individual C (caudal centra;

see below), Individual H (dorsal ribs, sternum), and

Individual O (pubis), whereas others appear to be iso-

lated elements whose precise taxonomic affinities

remain currently undecided (see the ‘Additional individ-

uals and assemblages’section, and the Supplemental

Material). Thus, these specimens are here excluded from

Petrustitan hungaricus n. comb. pending further, more

complete discoveries.

Figure 36. Petrustitan hungaricus n. gen., right ulna MCDRD 149 (Individual G) in A, posterior; B, lateral; C, distal (lateral

towards top); D, proximal (posterior towards top); E, anterior; and F, medial views. Scale bar equals 100 mm.

Revision of Romanian sauropod dinosaurs 51

Description and comparisons of Petrustitan

hungaricus

Posterior caudal vertebrae. Groza (1983) reported six

caudal vertebrae found in articulation as part of

Individual G (Fig. 35A), but only four could be located

in the MCDRD collections (Fig. 35R). The first in the

series (MCDRD 266; Fig. 35B–E) is procoelous, with

MCDRD 269 (the second in the series) amphicoelous

(Fig. 35F–I), and the last two (MCDRD 267 [Fig. 35J–

M] and 268 [Fig. 35N–Q]) mildly opisthocoelous. The

centra are spool-like, with a subcircular profile in

transverse cross-section, and a slightly constricted cen-

tral region compared to the articular ends. This morph-

ology differs from the dorsoventrally compressed

posterior caudal centra of Lirainosaurus,

Muyelensaurus,Rinconsaurus and Saltasaurus (Calvo,

Porfiri, et al., 2007; Calvo, Gonz

alez Riga, et al., 2007;

Calvo & Gonz

alez Riga, 2003;D



ıez D

ıaz et al., 2013b).

The centra of Petrustitan are elongate, with aEI values

of 1.76–1.87, similar to the posterior caudal vertebra of

Magyarosaurus and most titanosaurs, with the exception

of Isisaurus and Opisthocoelicaudia, which have

Figure 37. Petrustitan hungaricus n. gen., right radius MCDRD 150 (Individual G) in A, proximal (posterior towards to); B,

anterior; C, lateral; D, posterior; E, distal (posterior towards top); and F, medial views. G, proximal ends of the right ulna MCDRD

149 and radius MCDRD 150 in articulation. Scale bar equals 100 mm.

52 V. D



ıez D

ıaz et al.

anteroposteriorly shorter centra (Poropat et al., 2016). In

Petrustitan, the ventral surfaces of the centra are flat

(Fig. 35E, I, M, Q). Chevron facets are present, but

poorly developed, especially the anterior ones. No ridges

or excavations occur on either the ventral or lateral sur-

faces of the centra.

Figure 38. Petrustitan hungaricus n. gen., right metacarpal I MCDRD 152 (Individual G) in A, proximal (dorsal towards top); B,

ventral, C, lateral; D, distal (lateral towards top); E, dorsal; and F, medial views. Scale bar equals 50 mm.

Figure 39. Petrustitan hungaricus n. gen., right ilium MCDRD 148 (Individual G) in A, posterior; B, lateral; and C, medial views.

Scale bar equals 100 mm.

Revision of Romanian sauropod dinosaurs 53

Figure 40. Petrustitan hungaricus n. gen., tibia and fibulae. Paralectotype left tibia NHMUK R.3853 (Individual B) in A, proximal

(lateral towards bottom); B, lateral; C, anterior; D, distal (anterior towards top); E, medial; and F, posterior views. Paralectotype left

tibia and lectotype left fibula NHMUK R.3853 (Individual B) in articulation in G, proximal (anterior towards bottom); H, anterior; I,

posterior; and J, distal (posterior towards top) views. Lectotype left fibula NHMUK R.3853 (Individual B) in K, lateral; L, anterior;

M, distal (anterior towards top); N, proximal (medial towards bottom); O, medial; and P, posterior views. Left fibula MCDRD 153

(Individual G) in Q, lateral; R, posterior; S, proximal (medial towards bottom); T, medial; U, distal (medial towards top); and V,

anterior views. The numbers 1, 2, and 3 indicate the autapomorphies described in the text: tibia drawn out into slender laterodistal

process that inserts into a socket on the medial surface of the distal fibula; proximal end of the fibula with a subtriangular outline;

and posterodistally oriented accessory ridge on lateral surface of fibula, extending from midlength to one-quarter of fibula length

from distal end. Scale bar equals 200 mm.

54 V. D



ıez D

ıaz et al.

Longitudinal ridges are present at the junction

between the centrum and neural arch (Fig. 35B, F, J, N,

R). The neural arch is situated on the anterior half of

the centrum. The prezygapophyses are slender and

elongate, projecting primarily anteriorly, well beyond

the anterior margins of their respective centra.

Postzygapophyses do not project as distinct processes:

rather, they form laterally facing oval articular surfaces.

The neural spines are incomplete and abraded at their

apices. Nevertheless, the spines are well enough pre-

served to determine that they are simple, dorsoventrally

low projections.

Ulna. A right ulna (MCDRD 149) is preserved as part

of Individual G (Fig. 36). The maximum proximal

mediolateral width to proximodistal length ratio is 0.39.

Thus, this is a relatively robust element, similar to the

ulnae of titanosaurs such as Diamantinasaurus (Poropat,

Upchurch, et al., 2015), Isisaurus (Jain &

Bandyopadhyay, 1997), Lirainosaurus (D

ıez D



ıaz et al.,

2013a), and saltasaurines (Otero, 2010; Powell, 1992),

and differing substantially from that of Magyarosaurus

(<0.3). In Petrustitan, the proximal articulation surface

is gently convex, with only a weakly developed olecra-

non process. In this regard, Petrustitan differs from

most titanosaurs, in which the olecranon is a prominent

process (Wilson & Sereno, 1998), although the ulnae of

Lirainosaurus (D



ıez D



ıaz et al., 2013a) and Patagotitan

(Otero et al., 2020) also lack a well-developed olecra-

non. Although the anterolateral process is incomplete

in this Petrustitan specimen, it is clear that the antero-

medial process of the proximal end was considerably

longer (Fig. 36D). This differs from Magyarosaurus,

in which the proximal processes are more similar in

length to one another (Fig. 28B). The lateral surface

of the anteromedial process is concave, and its distal

tip is gently deflected laterally. The proximal articular

surface of this process has a convex outline in lateral/

medial views. The posterior process is rounded and only

slightly developed when compared to the other two

processes. A ridge extends distally from the anteromedial

process along the entirety of the shaft. Although less well

developed, another ridge extends distally from the antero-

lateral process and merges into the shaft at approximately

midlength. These anteromedial and anterolateral ridges

are less conspicuous than in the ulnae referred to

Atsinganosaurus and Lirainosaurus (D

ıez D

ıaz et al.,

2013a,2018). The posterior process is buttressed by a

thickened edge that develops distally into a thinner ridge.

The anteromedial surface of the shaft, for articulation

with the radius, is relatively shallowly concave

(Fig. 36D, E) and thus contrasts with the deeper radial

articulations of Magyarosaurus (Fig. 28B, D),

Atsinganosaurus,Lirainosaurus and Lohuecotitan (D

ıez

D

ıaz et al., 2013a,2016,2018). In Petrustitan,asin

most sauropods, this surface bears an interosseous ridge

along its distal half. This ridge extends subparallel to

the long axis of the shaft, as is typical for other sauro-

pods, rather than apomorphically slanting across the

anterior surface, as in Magyarosaurus (see above). The

overall shape of the shaft differs from the more slender

ulna of Atsinganosaurus (D



ıez D



ıaz et al., 2018), but

also from the transversely compressed ulnae of

Ampelosaurus (C3-1296; Le Loeuff, 2005b) and

Garrigatitan (D

ıez D

ıaz et al., 2021).

In Petrustitan, the distal articular surface of the ulna

is convex, with a rounded subtriangular outline, and lit-

tle posterior expansion (Fig. 36C), differing from the

probable ‘comma’-shaped distal end in Magyarosaurus.

This distal morphology in Petrustitan is similar to that

of Diamantinasaurus (Poropat, Upchurch, et al., 2015),

but differs from the more angular distal ends seen in

Atsinganosaurus (D

ıez D

ıaz et al., 2018, fig. 10O),

Lirainosaurus (D



ıez D



ıaz et al., 2013a, fig. 3.8), and

Ampelosaurus (C3-1000; Le Loeuff, 2005b, fig. 4.19D).

Radius. A right radius (MCDRD 150) (Fig. 37), asso-

ciated with the ulna MCDRD 149 (Figs 36,37G), has

been recovered as part of Individual G. The proximal

articular surface is slightly convex, with an oval out-

line, as a result of a prominent medial process (Fig.

37A, B, D). The shaft is slender, almost straight, with

a subcircular cross-section at midshaft. A slight torsion

is present in the shaft; however, this is not as marked

as in the radius of Magyarosaurus. There is a medially

placed ridge, located slightly above midlength, that

probably represents the M. biceps brachii scar. On the

posterolateral surface, the interosseous ridge extends

from the proximal end. However, it is not possible to

determine the distal extent of these medial and postero-

lateral ridges, because part of the shaft has been recon-

structed in plaster. The posterolateral surface of the

distal end is shallowly concave for articulation with

the ulna. In distal view (Fig. 37E), the radius has a

rounded/oval outline, with a lateral prominence. The

lateral two-thirds of the distal end surface are bevelled

at an angle of c.33

to a plane perpendicular to the

proximodistal axis of the radius. The distal articular

surface is gently convex, contrasting with the relatively

flat one in Magyarosaurus.

Metacarpal. The right metacarpal I (MCDRD 152) of

Individual G is preserved (Fig. 38). This element is

incomplete and slightly eroded. The shaft is straight

and has a circular cross-section, with the proximal and

distal thirds twisted relative to each other. It lacks

the lateral bowing that characterizes the first metacarpal

of the titanosaurs Andesaurus and Argyrosaurus

Revision of Romanian sauropod dinosaurs 55

(Apestegu

ıa, 2005; Mannion & Calvo, 2011). In

Petrustitan, the proximal and distal regions of metacar-

pal I are transversely expanded in comparison to the

shaft. Both the proximal and distal articular surfaces are

convex and have rounded subrectangular outlines. The

distal articular surface does not extend onto the dorsal

surface of the metacarpal. As is the case in most titano-

sauriforms, the distal end is approximately perpendicular

to the long axis of the metacarpal (Mannion et al.,

2013; Wilson, 2002).

Ilium. A partial right ilium of Individual G (MCDRD

148) preserves most of the ventral portion of the preace-

tabular process, as well as the pubic peduncle (Fig. 39).

The preacetabular process clearly projects anterolaterally,

but the ilium is too incomplete to determine whether this

formed a sub-horizontal platform, such as that seen in

many titanosaurs (Powell, 2003). There is some indica-

tion of a ‘kink’on the ventral margin of the preacetabular

process, similar to that seen in some titanosaurs, includ-

ing saltasaurines, Alamosaurus and Dreadnoughtus

(D’Emic, 2012; Ullmann & Lacovara, 2016), but its pres-

ence in Petrustitan is not certain given the incomplete

nature of the element. The pubic peduncle is short and

robust, contrasting with the relatively more slender

peduncles of many titanosaurs, including

Diamantinasaurus,Overosaurus,Rapetosaurus,

Atsinganosaurus,Garrigatitan,Lirainosaurus,

Normanniasaurus and Paludititan (Coria et al., 2013;

Csiki, Codrea, et al., 2010; Curry Rogers, 2009;D



ıez



ıaz et al., 2013a,2018,2021; Le Loeuff et al., 2013;

Otero, 2010; Poropat, Upchurch, et al., 2015). In

Petrustitan, this process projects primarily perpendicular

to the long axis of the preacetabular process. Its posterior

margin is dorsoventrally concave in lateral view. The dis-

tal articular surface of the pubic peduncle is anteroposter-

iorly convex, such that the posterior half of the distal

surface faces ventrally and posteriorly. There is no evi-

dence for a triangular excavation on the lateral surface of

the upper part of the pubic peduncle, such as that seen in

Lirainosaurus (D



ıez D

ıaz et al., 2013a), Dreadnoughtus

(Ullmann & Lacovara, 2016), and Rocasaurus (Salgado

& Azpilicueta, 2000). It is not possible to ascertain the

internal tissue structure of the Petrustitan ilium.

Tibia. NHMUK R.3853 is a left tibia (Fig. 40A–F)

associated with the holotype fibula, together represent-

ing Individual B (Fig. 40G–J). The tibia is described

with the flat subtriangular anterior surface at the distal

end of the shaft facing directly anteriorly. The proximal

end is equidimensional (Table S1), forming a rounded,

quadrangular outline (Fig. 40A). This morphology

differs from that seen in the proximal tibia of

Magyarosaurus, which presents a relatively narrower,

sub-rectangular to ‘D’-shaped profile (Fig. 33A). In

Petrustitan, the cnemial crest expands anterolaterally

and is curved only very mildly along its length. In prox-

imal view, the cnemial crest projects less markedly out-

ward relative to the dimensions of the proximal end,

differing from the condition in Magyarosaurus.

Furthermore, whereas the latter taxon has a cnemial

crest that is approximately aligned with the long medio-

lateral axis of the proximal tibia, in Petrustitan the crest

extends perpendicular to the long axis of the proximal

tibial surface (the latter being oriented anteromedially–

posterolaterally). The embayment enclosed by the cne-

mial crest on the anterolateral surface, that receives the

head of the fibula, is shallow and relatively small

because of the transverse widening of the proximal

end. A similar condition is present in some other

titanosaurs, including Diamantinasaurus,Saltasaurus,

Atsinganosaurus and Lirainosaurus (Powell, 1992;D



ıez

D

ıaz et al., 2013a,2018; Poropat, Upchurch, et al.,

2015). The cnemial crest of Petrustitan fades out before

reaching the midlength of the tibia.

At midlength, the shaft is constricted, especially trans-

versely. At this point, the shaft has an anteroposteriorly

elongated oval cross-section. Although incompletely pre-

served, it is nevertheless clear that the distal end

expands strongly transversely, such that its distal-most

preserved mediolateral width is more than twice that of

the shaft at midlength. The anteroposterior expansion of

the distal end is slight, and the anterior face is wide and

flat. As a result, the distal articular end has a laterally

elongated and pointed subtriangular profile. Essentially,

the lateral malleolus projects as a prominent distolateral

process. The tip of this triangular process inserts into a

medial fossa at the distal end of the fibula (see below;

Fig. 40H–J, O), forming an unusual interlocking contact

between the tibia and fibula distally. This morphology

also occurs in the saltasaurines Neuquensaurus (e.g.

MCS −5 [PDM pers. obs., 2009] and MLP-CS 1264

[PDM & PU pers. obs., 2013]; see also Otero, 2010,

figs 11, 12) and Saltasaurus (PVL 4017: PDM & PU

pers. obs., 2013; see also Powell, 2003, pls 45, 46), but

is clearly absent in other sauropods for which this fea-

ture can be assessed, including the penecontemporan-

eous Iberoarmorican taxa Abditosaurus,Ampelosaurus,

Atsinganosaurus,Lirainosaurus, and Lohuecotitan



ıez D

ıaz et al., 2013a,2016,2018; Le Loeuff,

2005b; Vila et al., 2022). As such, we regard this

unusual morphology of the distal tibia and fibula as a

local autapomorphy of Petrustitan. As already noted,

the distal-most articular portion of the tibia is not pre-

served in NHMUK R.3853. There is no evidence for a

prominent ridge on the anteromedial margin of the pre-

served distal tibia; such a ridge characterizes the distal

56 V. D



ıez D

ıaz et al.

quarter of the tibiae of Atsinganosaurus,Lirainosaurus,

Abditosaurus and Uberabatitan (D

ıez D

ıaz et al.,

2013a,2018; Silva Junior et al., 2019, fig. 23; Vila

et al., 2022), although it remains possible that the tibia

of Petrustitan is too incomplete to preserve this

feature.

Fibula. The fibula of Petrustitan is represented by two

left elements, NHMUK R.3853 (the holotype, Individual

B; Fig. 40K–P) and MCDRD 153 (belonging to

Individual G; Fig. 40Q–V). NHMUK R.3853 is essen-

tially complete, whereas MCDRD 153, though almost

complete, has missing parts of the shaft and a distal

third restored with plaster. In lateral view, the proximo-

distal axis of the fibula follows a gentle sigmoidal curve

(Fig. 40L, R), as is the case in most somphospondylans

(Canudo et al., 2008). The proximal and distal ends are

slightly twisted relative to one another, such that the

medial surfaces at each end are not quite in the same

plane. Although the proximal end is slightly worn in the

available specimens, this region is characterized by the

presence of an anteromedially directed crest that extends

into a notch behind the cnemial crest of the tibia, as

also occurs in Magyarosaurus. Excluding the anterome-

dial process, the proximal articular surface of the fibula

has an anteroposteriorly elongate subtriangular outline

in Petrustitan (Fig. 40K, Q). This proximal profile is

formed by a nearly straight medial margin, and a trans-

verse narrowing towards its posterior end. Such a prox-

imal end outline differs from the ‘D’-shaped one in

Magyarosaurus (Fig. 40N, S) and the D-shaped to ellip-

tical outline in other titanosaurs, including

Ampelosaurus (VDD pers. obs., 2024) and Lohuecotitan



ıez D

ıaz et al., 2016). It is therefore regarded as auta-

pomorphic for Petrustitan. In the fibula of Petrustitan,

the proximal end is strongly expanded posteriorly rela-

tive to the shaft, such that the proximal half of the pos-

terior margin of the fibula is concave in lateral view.

This is an unusual condition that we regard as a possible

autapomorphy of Petrustitan. However, this feature is

not as hooked or as prominent in (the more poorly pre-

served) MCDRD 153 specimen (Fig. 40S), and thus we

have provisionally excluded this feature from our diag-

nosis. The proximal articular surface of the fibula is

irregular, and is bevelled such that it faces proximally

and also slightly anteriorly. The medial surface of the

proximal end appears to be striated, although not espe-

cially raised, as is also the case in other titanosauriforms

(D’Emic, 2012; Wilson & Sereno, 1998).

The fibula of Petrustitan lacks an anterolateral tro-

chanter along its proximal third. The lateral trochanter

extends distally to at least midlength of the fibula in

NHMUK R.3853 (Fig. 40K), but does not reach so far

distally in MCDRD 153 (Fig. 40Q, R, V). Most

eusauropods are characterized by the former condition

(Whitlock, 2011), although the lateral trochanter is more

proximally restricted in some titanosaurs, including

Epachthosaurus,Lohuecotitan and Mendozasaurus



ıez D



ıaz et al., 2016; Mannion, Upchurch, Jin, et al.,

2019). The lateral trochanter is composed of two sub-

parallel ridges: a prominent one (more laterally devel-

oped in MCDRD 153) that extends along the posterior

margin of the lateral surface, and a less developed one

that is situated more anteriorly (this ridge is less obvious

in MCDRD 153 because of its poorer preservation and

partial restoration in plaster). Although prominent, the

lateral trochanter of Petrustitan is not hypertrophied to

the same extent as those of the fibulae of Laplatasaurus

and Uberabatitan (Gallina & Otero, 2015). An add-

itional ridge is present on the anterior margin of the lat-

eral surface in NHMUK R.3853 (this feature cannot be

ascertained in MCDRD 153). Its proximal tip is situated

at approximately the midlength of the fibula. It is

directed steeply posterodistally, terminating distally at

about three-quarters of fibula length. This ridge is simi-

lar to that described in the fibula of Magyarosaurus but

appears to represent a convergently developed feature

(supported by our phylogenetic analyses –see below)

that we herein regard as a local autapomorphy of both

Petrustitan and Magyarosaurus. Just proximal to the

distal tip of this ridge in Petrustitan, the anterior margin

of the distal third of the fibula forms a well-developed

ridge, as in most eusauropods (Upchurch et al., 2004),

although this is poorly preserved in NHMUK R.3853.

The medial surface of the distal end is strongly con-

cave for reception of the laterodistal expansion of the

tibia (see above; Fig. 40O). As noted above, a similar

concavity on the distal fibula is otherwise only known

in Neuquensaurus and Saltasaurus. In distal view, the

fibula has a subtriangular outline, with flattened medial,

anterolateral and posterolateral margins (although the

latter two margins meet at a rounded corner) (Fig. 40M,

U). The anterior apex of this triangle is incomplete in

NHMUK R.3853. The distal end has a more rounded

outline in MCDRD 153, but this is likely to be the

result of poor preservation. This outline differs from the

tear-drop-shaped distal end present in Magyarosaurus

(Fig. 34J). A subtriangular distal fibular outline is a

derived state that occurs in several other titanosaurs,

such as Lirainosaurus (D



ıez D



ıaz et al., 2013a) and

Diamantinasaurus (Poropat, Upchurch, et al., 2015).

Titanosauria Bonaparte & Coria, 1993

Lithostrotia Upchurch, Barrett & Dodson, 2004

Uriash n. gen.

Figs 41–45

Revision of Romanian sauropod dinosaurs 57

Type species. Uriash kadici sp. nov.

Etymology. The genus name comes from the Romanian

word ‘urias

’(pronounced ‘uriash’) that refers to gigantic

humanoid fairy-tale characters in Romanian folk-lore

(see also the Hungarian term ‘

ori

as’), but also, gener-

ally, to something or someone colossal, gigantic. It

alludes to the large body size of this titanosaur taxon,

significantly larger than the other Transylvanian taxa.

Generic diagnosis. As for the type (and only known)

species.

Uriash kadici n. gen. n. sp.

1932 Magyarosaurus hungaricus Huene: 269

1970 Titanosaurus hungaricus Steel: 77

2005a Magyarosaurus dacus Le Loeuff: 15–16

2010 ‘Magyarosaurus’hungaricus Stein et al.: 9259

Holotype. Individual C: four posterior–distal caudal verte-

brae (SZTFH Ob.3090B, D, G, H; originally eight vertebrae

collected, but four are missing); incomplete right humerus

(SZTFH Ob.3104); incomplete right femur and one small

fragment (region of the fourth trochanter) of the left femur

(SZTFH Ob.3103); left metatarsal I (SZTFH Ob.3095).

Etymology. The species name honours Ottok

ar Kadi

(1876–1957), geologist and palaeontologist of the Royal

Geological Survey of Hungary and discoverer of several

continental vertebrate-bearing fossil localities in the

north-western Hat

,eg Basin, including the type locality

of Uriash, between 1909 and 1915.

Type locality. Locality VI of Kadi

c (see Kadi

c, 1916)

in the Budurone ravine (P^

ar^

aul Budurone), south-east of

V

alioara locality, north-western Hat

,eg Basin, Hunedoara

County, Romania (Fig. 71). The approximate position of

the original locality was recently relocated in the field

by Botfalvai et al. (2021).

Type stratum and age. The holotype comes from the

lower part of the unnamed middle member of the upper-

most Cretaceous continental Densus

,-Ciula Formation

(Csiki-Sava et al., 2016; Grigorescu, 1992). The age of

the local succession, once considered to be late

Maastrichtian (e.g. Antonescu et al., 1983), was recently

Figure 41. Uriash kadici n. gen. n. sp. holotype, posterior caudal vertebrae SZTFH Ob.3090 (Individual C). SZTFH Ob.3090B in A,

anterior; and B, ventral views. SZTFH Ob.3090D in C, left lateral; D; anterior; E, posterior; and F, ventral views. SZTFH Ob.3090G

in G, right lateral; H, posterior; I, left lateral; J, anterior; and K, dorsal views. SZTFH Ob.3090H in L, right lateral; and M, anterior

views. The numbers 1 and 2indicate the autapomorphies described in the text: posterior caudal vertebrae with a markedly

procoelous centrum and a distinctive hour-glass shaped contour in ventral view; and posterior caudal vertebrae with anteroposteriorly

elongated neural arch pedicels. Scale bar equals 50 mm.

58 V. D



ıez D

ıaz et al.

reassessed as early Maastrichtian (Fig. 10; Botfalvai

et al., 2021; Csiki et al., 2007; Csiki-Sava et al., 2016).

Diagnosis. Uriash kadici n. gen. sp. can be diagnosed

on the basis of four autapomorphies (marked with an

asterisk), as well as four ‘local’autapomorphies: (1)

posterior caudal vertebrae with a markedly procoelous

centrum and a distinctive hour-glass shaped contour in

ventral view; (2) posterior caudal vertebrae with antero-

posteriorly elongated neural arch pedicels; (3) presence

of a midline longitudinal convexity along the posterior

humeral shaft curving distally towards the medial edge;

(4) proximal half of humerus with an ‘anteromedial

arm’; (5) large and proximodistally elongated oval for-

amen in the posterolateral depression of the proximal

humerus; (6) hypertrophied bulge for insertion site of

M. latissimus dorsi; (7) femoral shaft starts to trans-

versely expand distally at a point close to its mid-

length; (8) hypertrophied femoral fourth trochanter.

Comment. As noted above, skeletal elements belonging

to Individual C had been briefly discussed by Huene

(1932) and Le Loeuff (2005a), without these authors

recognizing that these may represent parts of the same

incomplete skeleton. Huene (1932, p. 271, pl. 47, figs 4,

5) briefly described and figured two of the SZTFH

Ob.3090 posterior caudal centra and, again based on

their large size, tentatively referred them to his new spe-

cies ‘M.’hungaricus (here reassigned as Petrustitan

hungaricus nov. comb). Le Loeuff (2005a) discussed

the large size of the humerus specimen SZTFH Ob.3104

(referred by him to Magyarosaurus dacus, under speci-

men number MAFI V13491) as evidence of the non-

dwarfed status of this taxon. Our detailed comparisons

(see below) show that Individual C does not share any

of the autapomorphies used to diagnose Magyarosaurus,

Petrustitan and/or Paludititan, instead presenting a set

of autapomorphies of skeletal elements that can be dir-

ectly compared with those of the former three taxa, thus

precluding its referral to any of these other

Transylvanian titanosaurs.

Description and comparisons of Uriash kadici

Caudal vertebrae. As noted above, the lettering

sequence of the individual vertebrae appears to reflect

their relative position within the tail, with vertebrae

SZTFH Ob.3090B and Ob.3090D being more anteriorly

situated than Ob.3090G and Ob.3090H (Fig. 41). The

former pair have an approximately comparable morph-

ology, but one that is clearly distinct from that of the

latter pair, suggesting a substantial morphological

change along the tail within its posterior segment. None

Figure 42. Uriash kadici n. gen. n. sp. holotype, right humerus SZTFH Ob.3104 (Individual C) in A, anterior; B, medial; C,

posterior; and D, lateral views. Abbreviations: ama, anteromedial arm; dpc, deltopectoral crest. The dashed line represents the area

of the posterior longitudinal convexity. The numbers 3,4,5

and 6 indicate the autapomorphies described in the text: presence of a

midline longitudinal convexity along the posterior humeral shaft curving distally towards the medial edge; proximal half of humerus

with an ‘anteromedial arm’; large and proximodistally elongated oval foramen in the posterolateral depression of the proximal

humerus; and respectively hypertrophied bulge for insertion site of M. latissimus dorsi. Scale bar equals 100 mm.

Revision of Romanian sauropod dinosaurs 59

of the four remaining vertebrae of the series is complete,

preserving only the centra and bases of the neural

arches; furthermore, the articular ends of the centra are

often heavily damaged (especially those of G and H),

making the assessment of the type of articular surface

difficult in some of them. The vertebrae are all from the

posterior to distal region of the tail, as suggested by fea-

tures such as the absence of transverse processes, rela-

tive elongation of the centra, and absence of articular

facets for the chevrons (Mannion et al., 2013).

The first two vertebrae in this series (B and D) are

both strongly procoelous (Fig. 41A–F), with a deeply

concave anterior articular cotyle, differing from the pos-

terior caudal vertebrae of Magyarosaurus and

Petrustitan. The posterior articular surfaces of SZTFH

Ob.3090B and D are markedly convex and somewhat

wider than high. The ‘apex’of the posterior condyle is

located dorsally relative to the mid-height of the cen-

trum. In vertebra D, the centre of the condyle is exca-

vated by a circular pit (Fig. 41E). Vertebrae B and D

are both characterized by an hourglass-shaped morph-

ology in dorsal/ventral view, which results from an

elongated centrum that is strongly laterally constricted

at midlength (Fig. 41B, F). The lateral sides of the cen-

trum are consequently deeply concave anteroposteriorly,

but mildly convex dorsoventrally. A similar hourglass

morphology can be observed in the posterior caudal cen-

tra of Lirainosaurus and Atsinganosaurus (D

ıez D

ıaz

et al., 2013b,2018), but the specific anatomy of the

caudal vertebrae of Uriash –a combination of an

Figure 43. Uriash kadici n. gen. n. sp. holotype, femora. Large fragment of right femur SZRFH Ob.3103 (individual C) in A,

anterior; B, medial; C, posterior; and D, lateral views. Small fragment of left femur SZRFH Ob.3103 (individual C) in E, posterior;

and F, medial views. The numbers 7and 8 indicate the autapomorphies described in the text: femoral shaft starts to transversely

expand distally at a point close to its midlength; and hypertrophied femoral fourth trochanter. Scale bar equals 200 mm.

60 V. D



ıez D

ıaz et al.

hourglass-shaped cross-section and strong procoely –

has not been found in any other titanosaurs, including

the European taxa (D



ıez D



ıaz et al., 2013b,2016,2018;

Le Loeuff, 2005b). In Uriash, the dorsal and ventral sur-

faces of the centrum are approximately equal in trans-

verse width, although the maximum transverse width

occurs at the level of the aforementioned longitudinal

lateral crest. The ventral surface of the centrum is flat to

slightly convex transversely, and lacks a midline furrow

and ventrolateral ridges connecting chevron facets; only

the remnants of very short and low crests can be seen

near the posterior articular margin. The position of the

neural arch is clearly biased towards the anterior end of

the centrum; however, the neural arch pedicels are elon-

gated, extending for over half the length of the centrum

(Fig. 41C). Such arch elongation is not present, as far as

we are aware, in the posterior caudal vertebrae of other

Transylvanian specimens, or other Late Cretaceous

European titanosaurian taxa either, and might ultimately

represent an autapomorphy. The neural canals are trans-

versely narrow.

The morphology of the other two vertebrae of the ser-

ies (G and H, Fig. 41G–N) differs from that described

in vertebrae B and D. Vertebrae G and H have more

elongated, spool-like centra, that are not so markedly

constricted at midlength. Furthermore, these two distal

posterior or distal caudal vertebrae are not procoelous,

and are either amphicoelous or opisthocoelous (N.B.

given the damaged nature of the articular ends, espe-

cially the anterior ones, this is difficult to ascertain).

The posterior articular surfaces are slightly concave

(deeper in G than in H), surrounded by a stout, rounded

rim. This morphology is shared with Petrustitan, but

clearly differs from the generally procoelous-to-platy-

coelous posterior caudal vertebrae of most titanosaurs,

including Muyelensaurus,Rapetosaurus and

Saltasaurus, as well as the European titanosaurs

Magyarosaurus,Ampelosaurus,Atsinganosaurus and

Lirainosaurus (Calvo, Porfiri, et al., 2007; Calvo,

Gonz

alez Riga, et al., 2007; Curry Rogers, 2009;D



ıez

D

ıaz et al., 2013b,2018; Le Loeuff, 2005b). The ventral

surface is transversely flat and slightly dorsally arched

along its length, lacking ventrolateral longitudinal

ridges. The lateral sides of the centra are subparallel,

dorsoventrally flat, and slightly concave anteroposter-

iorly. A rounded protuberance lies at the level of the

junction between the neural arch and the centrum.

Neural arch pedicels are restricted to the anterior half of

the centrum, extending almost to the anterior margin of

the latter (Fig. 41G, M). The neural canal is reduced,

especially in vertebra H.

Humerus. The right humerus (SZTFH Ob.3104) pre-

serves the major part of the diaphysis, extending from

the region of the deltopectoral crest to close to the distal

epiphysis; thus, both the proximal and distal articula-

tions are missing (Fig. 42). Despite its incomplete pres-

ervation, this humerus displays several unusual features,

aside from its very large size, compared to most other

titanosaur remains from Transylvania, as also noted by

Le Loeuff (2005a).

Figure 44. Uriash kadici n. gen. n. sp. holotype, left metatarsal I SZTFH Ob.3095 (individual C) in A, proximal (ventral towards

bottom); B, ventral; C, distal (ventral towards top); D, medial; E, dorsal; and F, lateral views. Scale bar equals 50 mm.

Revision of Romanian sauropod dinosaurs 61

The preserved proximal segment of the humerus is

anteroposteriorly flattened, with a rounded and narrow

lateral margin and a wider, largely convex medial mar-

gin. Immediately distal to the broken proximal region, a

longitudinal angular crest appears on the medial edge of

the anterior surface, merging with the shaft distally

(Fig. 42A). This crest is not present in Magyarosaurus,

but is similar to the ‘anteromedial arm’that character-

izes the titanosaurs Argyrosaurus,Gondwanatitan,

Muyelensaurus,Paralititan and Rukwatitan (Gorscak

et al., 2017; Mannion & Otero, 2012). The anterior face

of the proximal end is generally concave. Only the distal

half of the deltopectoral crest (approximately from the

most prominent anteromedial part [¼tip] of the crest to

its distal base) is preserved. This crest projects mainly

medially, rather than anteromedially, overhanging the

anterior face, as in the humeri of Magyarosaurus and

Lirainosaurus (D



ıez D

ıaz et al., 2013a). However, it is

not as medially deflected as in the humerus of

Abditosaurus (Vila et al., 2022). The lateral side of the

distal part of the deltopectoral crest has a coarse texture.

Distal to the tip of the deltopectoral crest, there is a

weak prominence on the anterolateral edge of the

humerus: this potentially represents the insertion of M.

brachialis inferior. Distal to the most prominent part of

the deltopectoral crest, the shaft of the humerus narrows

gradually. This constriction of the shaft is relatively

mild compared to that seen in Magyarosaurus: the ratio

between the minimum shaft width and the mediolateral

width at the level of the deltopectoral crest tip is

approximately 0.73 in SZTFH Ob.3104, whereas it is as

low as 0.63 in the humerus LPB (FGGUB) R.1047 of

Magyarosaurus.

The posterior face of the proximal end is broadly flat

in SZTFH Ob.3104, but has a complex morphology

(Fig. 42C). In particular, this surface bears a proximo-

distally elongated depressed area in its lateral part, a

similarly elongated midline prominence, and another

shallowly concave area medially. The midline longitu-

dinal convexity continues distally along the shaft, curv-

ing towards the medial edge and gradually attenuating.

Within the lateral depression, just below the broken

proximal edge, there is a large and proximodistally elon-

gated oval foramen, which might represent an autapo-

morphy, as it has not been described in any other

titanosaur. Lateral and distal to this foramen, there is a

rugose, prominent, proximodistally elongated bulge

(insertion site of M. latissimus dorsi) on the posterior

face of the proximal end. This bulge starts slightly prox-

imal to the most prominent part of the deltopectoral

crest, and extends distally, approximately parallel to the

lateral margin, to a point level with where the distal part

of the deltopectoral crest merges into the humeral shaft.

The strongest projection of this bulge, which is also

where it reaches its greatest mediolateral width, is

approximately level with the most prominent part of the

deltopectoral crest. This bulge is so well developed that

it is clearly visible in both medial and lateral views

(Fig. 42B, D). A similarly well-developed posterior lon-

gitudinal bulge is present in the large humerus ascribed

to Garrigatitan, although it is more centrally located in

the latter taxon (D



ıez D

ıaz et al., 2021, fig. 5F). Given

that the corresponding feature in Magyarosaurus is less

well developed and centrally placed, we regard the

extreme prominence of this laterally positioned bulge in

the humerus of Uriash as a potential autapomorphy.

Medial to the proximal part of the bulge, the posterior

surface of the proximal end is excavated by a rugose,

elongated oval depression, bordered proximally by a

small sharp lip. A narrow angular crest originates lateral

and distal to the prominent bulge; this extends obliquely

distally for approximately 140 mm, merging with the lat-

eral margin of the shaft at a point level with the latter’s

minimum transverse width. The proximal part of this

crest is unevenly rugose, whereas its distal half is cov-

ered by coarse longitudinal striae, suggesting it was

probably a muscle insertion site. Level with this angular

crest, but on the midline of the posterior face, there is a

proximodistally elongated flat region marked by coarse

vertical striae.

The humeral shaft is relatively robust, with a moder-

ately flattened, transversely elongate oval cross-section

and an ECC of 1.65, higher than in Magyarosaurus (in

which ECC values range between 1.26 and 1.44). In

SZTFH Ob.3104, the lateral and medial surfaces of the

shaft are broadly rounded, the posterior face is mildly

convex transversely, and the anterior one is slightly con-

cave on the midline proximally and then flattens

distally.

Distal to the point of minimum shaft circumference,

the humerus expands again, but this is rather mild in the

preserved segment. In the distal-most preserved part of

the humerus, its posterior face becomes somewhat con-

cave because of the presence of the bases of the supra-

condylar ridges. The medial surface remains largely

rounded anteroposteriorly, whereas the lateral one bears

an angular crest covered by longitudinal striae (possibly

the area of origin for the M. flexor carpi ulnaris and

radialis; Voegele et al., 2020).

Femur. The femur is known from a large fragment

(most of the shaft, including the fourth trochanter) of a

right element (Fig. 43A–D), and a small fragment pre-

serving the region of the fourth trochanter from the left

element (Fig. 43E–F), both registered under the same

specimen number (SZTFH Ob.3103). Although the

proximal and distal ends are absent, the preserved length

62 V. D



ıez D

ıaz et al.

of the right femur suggests an estimated original length

of more than one metre, which is about twice the

size of the next largest femur known from the

Transylvanian area (LPB [FGGUB] R.1046, a specimen

referred to Magyarosaurus); this suggests an individual

comparable in size with the Argentinean early-branching

titanosaur Epachthosaurus (Mart

ınez et al., 2004). The

right femur is badly preserved and severely crushed

anteroposteriorly. The femur was probably robust, with

a highly anteroposteriorly compressed shaft, even taking

into account the crushing of the bone that might have

exaggerated this feature. It has an ECC of 2.48, much

higher than the values calculated for the femora of

Magyarosaurus (ECC range ¼1.14–1.17). The better

preserved, less crushed anterior surface of the shaft is

transversely rounded, being more convex in its medial

half and slightly depressed in its lateral half. As a result,

the greatest anteroposterior width of the shaft lies med-

ial to the midline. The anterior surface of the shaft

changes from transversely flat proximally to slightly

convex distally. A longitudinal rugose crest arises distal

to the midlength of the anterior surface and extends dis-

tally, which might be the linea intermuscularis cranialis.

The posterior face of the shaft is heavily crushed

inward. A weak angular crest is present near the medial

margin of the posterior surface, but it is unclear whether

or not this is a taphonomic artefact. The shaft starts to

expand transversely at a point close to its midlength,

contrasting with the condition present in most titano-

saurs, including Ampelosaurus,Atsinganosaurus,

Garrigatitan,Lirainosaurus and Lohuecotitan (D

ıez

D

ıaz et al., 2013a,2016,2018,2021), as well as

Magyarosaurus, in which the femur starts to expand

transversely closer to the distal extremity.

A notable feature of the femur, seen in both the left

and right element, is its extremely well developed,

prominent fourth trochanter, clearly differing from those

of most titanosaurs. This is located on the posteromedial

margin, as in most sauropods, but its great development

means that it is clearly visible in anterior, posterior, lat-

eral and medial views (Fig. 43). As such, we regard this

hypertrophied fourth trochanter as a potential autapo-

morphy. Two distinct rugose areas are present on the

fourth trochanter of both femoral fragments, which we

interpret as the insertion areas for M. caudofemoralis

longus and brevis. One of these is oval and located on

the medial tip of the fourth trochanter, covering its distal

two-thirds. The second muscle insertion is positioned in

the proximal third of the fourth trochanter, and runs dis-

tally posteriorly. The elongated rugosity on the medial

edge, proximal to the fourth trochanter, probably forms

the insertion area for M. puboischiofemoralis externus

(Voegele et al., 2021).

The flaring of the shaft in the distal-most preserved

part of the right femoral fragment suggests that the

breakage occurred near the base of the distal articular

end. At the broken distal end, the beginning of a longi-

tudinally elongated crest can be observed near the lateral

edge of the posterior surface. This represents the prox-

imal-most termination of the lateral (fibular) supracon-

dylar crest (Fig. 43C). The surface of this crest is

marked by coarse longitudinal striations. The posterior

face of the shaft is flattened, but the proximal portion of

the medial (tibial) supracondylar crest is also visible,

close to the medial margin.

Metatarsal. SZTFH Ob.3095 is a left metatarsal I (Fig.

44). This element is nearly complete, but lacks some of

the medial margin and small portions of the lateral mar-

gin of the proximal end. It has been broken into two

pieces and glued back together. The proximal end is

bevelled relative to the long axis of the element, such

that it faces slightly medially. The proximal articular sur-

face is flat to mildly concave. Although incomplete, the

metatarsal has a ‘D’-shaped outline in proximal view,

with the dorsoventrally tall, flat margin facing steeply

ventrolaterally. The dorsal surface of the shaft lacks the

foramina and the dorsolateral rugosity seen in several dip-

lodocoids (Tschopp et al., 2015; Upchurch, 1998). In dor-

sal view, the lateral margin is concave, but the medial

margin is too incomplete to assess its profile. The meta-

tarsal decreases in dorsoventral height distally, with the

lateral surface of the proximal half being gently concave

dorsoventrally.

The distal end is bevelled relative to the long axis of

the metatarsal, with the lateral condyle extending further

distally than the medial one. It also bears a ventrolateral

projection that results in the laterodistal condyle extend-

ing further laterally than the proximal end of the elem-

ent in dorsal view. In distal view, the metatarsal has a

mediolaterally elongated oval profile, decreasing in

dorsoventral height towards its lateral margin. The distal

articular surface is mediolaterally concave over most of

its extent, becoming convex towards its medial and lat-

eral margins. This surface is dorsoventrally convex,

although it flattens centrally.

Additional anatomical information on

Paludititan nalatzensis

Although Paludititan was described only just over a

decade ago (Csiki, Codrea, et al., 2010), additional

insights into its anatomy arise from both the new infor-

mation provided in the current work and also the numer-

ous other studies of titanosaur taxa published recently.

Therefore, below we provide an update on the

Revision of Romanian sauropod dinosaurs 63

diagnostic characters of Paludititan, together with a

more detailed anatomical description and comparison.

Titanosauria Bonaparte & Coria, 1993

Lithostrotia Upchurch, Barrett & Dodson, 2004

Eutitanosauria Sanz, Powell, Le Loeuff, Mart

ınez &

Pereda Suberbiola, 1999

Paludititan Csiki, Codrea, Jipa-Murzea & Godefroit,

2010

Paludititan nalatzensis Csiki, Codrea, Jipa-Murzea &

Godefroit, 2010

(Figs. 45–50)

Holotype. Individual D: three dorsal vertebrae (UBB

NVM 1-43, UBB NVM 1-44, UBB NVM 1-45), several

ribs and rib fragments (UBB NVM 1-12, UBB NVM 1-

20, UBB NVM 1-22 to UBB NVM 1-25, UBB NVM 1-

27 to UBB NVM 1-42), three anterior caudal vertebrae

(UBB NVM 1-3, UBB NVM 1-50, UBB NVM 1-58),

three anterior mid-caudal vertebrae (UBB NVM 1-1 –

two vertebrae in connection, UBB NVM 1-2), six mid-

posterior caudal vertebrae (UBB NVM 1-21, UBB

NVM 1-26, UBB NVM 1-46, UBB NVM 1-47, UBB

NVM 1-48, UBB NVM 1-49), several fragments of cau-

dal vertebrae (UBB NVM 1-19, UBB NVM 1-57), 12

incomplete chevrons (UBB NVM 1-4 to UBB NVM 1-

9, UBB NVM 1-15 to UBB NVM 1-18, UBB NVM 1-

24), partial right pelvis including almost complete pubis,

ischium and acetabular part of the ilium, in articulation

(UBB NVM 1-13), left ischium (UBB NVM 1-45), frag-

ment of right femur (UBB NVM 1-53), and two ungual

phalanges (UBB NVM 1-10, UBB NVM 1-11). Several

other fragments are either incompletely prepared (e.g.

UBB NVM 1-59) or unidentified in the original publica-

tion (UBB NVM 1-51, UBB NVM 1-52, here we iden-

tify them as clavicles).

Locality and distribution. S^

anpetru Formation, N

alat

Vad, Hat

,eg Basin, Romania. The UBB NVM 1

Figure 45. Paludititan nalatzensis holotype, dorsal vertebrae (Individual D). Anterior dorsal vertebra UBB NVM 1-43 in A, left

lateral; B, anterodorsal; C, right lateral; D, right lateral (interpretive drawing); E, posterior; and F, ventral views. Middle–posterior

dorsal vertebra UBB NVM 1-44 in G, left lateral; H, anterior; I, right lateral; and J, posterior views. Middle–posterior dorsal

vertebra UBB NVM 1-45 in K, left lateral; L, anterior; M, right lateral; and N, right lateral (interpretive drawing) views.

Abbreviations: ACDL, anterior centrodiapophyseal lamina; ACPL, anterior centroparapophyseal lamina; CDF, centrodiapophyseal

fossa; CPOL, centropostygapophyseal lamina; D, diapophysis; PA, parapophysis; PACDF, parapocentrodiapophyseal fossa; PCDL,

posterior centrodiapophyseal lamina; PCPL, posterior centroparapophyseal lamina; PPDL, paradiapophyseal lamina; PO,

postzygapophysis; POCDF, posterior centrodiapophyseal fossa; PRCDF, prezygocentrodiapophsyeal fossa; PRDL,

prezygodiapophyseal lamina; PRE, prezygapophysis; PRSL, prespinal lamina; SPDL, spinodiapophsyeal lamina; SPOL,

spinopostzygapophyseal lamina. The number 1indicates the autapomorphy described in the text: dorsal centra with weakly

developed anterior condyle. Scale bar equals 100 mm.

64 V. D



ıez D

ıaz et al.

specimen comes from site 10 of Smith et al. (2002);

according to van Itterbeeck et al. (2004), the position of

site 10 corresponds to the dark palaeosoil horizon at

56–58 metres.

Revised diagnosis. Paludititan nalatzensis can be diag-

nosed on the basis of two autapomorphies (marked with

an asterisk), as well as two ‘local’autapomorphies: (1)

dorsal centra with weakly developed anterior condyle;

(2) ischiadic peduncle of the ilium unusually well devel-

oped dorsoventrally, with an anteroposteriorly elongate

distal end; (3) iliac peduncle of ischium bearing a post-

erolateral buttress (i.e. a raised area on the posterolateral

margin of its proximal part); (4) non-coplanar distal

ischia.

Comment. Csiki, Codrea, et al. (2010) proposed a list

of four autapomorphies (comprising three vertebral fea-

tures and one corresponding to the ischium) and two

putative ones (pertaining to the lamination of the poster-

ior dorsal vertebrae and the pubis). These authors also

noted that Paludititan possesses a unique combination

of three character states pertaining to the axial skeleton.

All of these character states have been reviewed in the

current work: although some are no longer recognized

as autapomorphies, additional autapomorphies have been

newly identified.

Individual V (LPB [FGGUB] R.2715), comprising 10

caudal vertebrae and one chevron, recovered from the

lower part of the unnamed middle member of the

Densus

,-Ciula Formation (lower part of the lower

Maastrichtian) at locality K2, was preliminarily sug-

gested to be closely comparable to Paludititan

(Botfalvai et al., 2021). This assessment was based on a

caudal vertebral feature (“presence of amphiplatyan and

platycoelous caudal vertebrae intercalated between pro-

coelous caudal vertebrae in the mid-section of the tail”;

Csiki, Codrea, et al., 2010, p. 302) that is no longer

regarded as an autapomorphy of Paludititan in our

revised diagnosis, and thus cannot be used as a unique

feature to unite these two individuals. Furthermore, the

K2 titanosaur material requires further preparation and

detailed study before such a referral can be supported.

Description and comparisons of Paludititan

nalatzensis

Dorsal vertebrae. One anterior (UBB NVM 1-43) and

two middle–posterior (UBB NVM 1-44 and 1-45) dorsal

vertebrae of Paludititan are known (Fig. 45). UBB

NVM 1-43 is the most complete specimen (Fig. 45A–

F), preserving the centrum and much of the neural arch

and spine (though lacking most of the processes), with

details visible mostly only on the right side (Fig. 45C,

D). However, it is poorly preserved in many places, and

somewhat deformed (e.g. the neural spine has been dis-

torted to project almost entirely posteriorly). UBB NVM

1-44 (Fig. 45G–J) preserves the centrum and base of the

neural arch; the right side has been partly sheared, such

that it is dorsally displaced relative to the left side. UBB

NVM 1-45 preserves the anterior half of the centrum

and most of the neural arch and spine, although the lat-

ter is now separated from the rest (Fig. 45K–N).

The internal camellate structure of the vertebra is

clearly visible in UBB NVM 1-43. The elongated nature

of each centrum can be confirmed, mainly because of

the better preservation of the two more anterior dorsal

vertebrae. All of the centra are opisthocoelous, although

the anterior surfaces are only gently convex. This distin-

guishes them from the dorsal vertebrae of

Magyarosaurus, as well as most somphospondylans,

which are characterized by a prominent anterior condyle

throughout the dorsal series (Salgado et al., 1997). As

such, the presence of a mildly convex anterior condyle

is regarded as a local autapomorphy of Paludititan. The

ventral surface of each centrum is transversely convex

and lacks ridges or fossae (Fig. 45F). In lateral view,

the ventral margin of UBB NVM 1-43 is concave.

Generally, the single lateral pneumatic openings are

eye-shaped and occupy most of the length of the non-

condylar centrum; they extend over approximately the

dorsal two-thirds in UBB NVM 1-43 (Fig. 45A, C, D),

and the dorsal half of the UBB NVM 1-44 centrum

(Fig. 45G, I). These pneumatic openings are deep but

do not seem to ramify dorsally or ventrally. There is no

clear evidence that these pneumatic spaces were divided

by a ridge internally, but they are not well enough

Figure 46. Paludititan nalatzensis, thoracic ribs (Individual D) in A, lateral?; and B, medial views. Scale bar equals 100mm.

Revision of Romanian sauropod dinosaurs 65

preserved for this to be determined with certainty. The

lateral pneumatic opening of UBB NVM 1-43 does not

seem to be set within a wider fossa; that in UBB NVM

1-44 probably was set within such a fossa, although it is

difficult to discern.

No hyposphene-hypantrum complex is present, as is

also typical of the dorsal vertebrae of nearly all titano-

saurs, with the exception of Andesaurus and

Epachthosaurus (Mart

ınez et al., 2004; Salgado et al.,

1997; Upchurch, 1998), as well as some specimens of

Patagotitan (Carballido et al., 2017). In UBB NVM 1-

43, the neural arch does not extend to the posterior mar-

gin of the centrum. Neither the anterior or posterior

neural canal openings are well enough preserved in this

vertebra to be entirely certain of their outlines, but they

were probably dorsoventrally tall and elliptical. Poor

preservation in most specimens means that it is not pos-

sible to unequivocally determine whether the anterior

Figure 47. Paludititan nalatzensis, chevrons (Individual D). UBB NVM 1-4 in A, anterior; B, proximal (posterior towards bottom);

C, posterior; and D, right lateral views. UBB NVM 1-5 in E, anterior; F, proximal (posterior towards bottom); G, posterior; and H,

left lateral views. UBB NVM 1-6 in I, anterior; J, proximal (posterior towards bottom); K, posterior; and L, right lateral views. UBB

NVM 1-7 in M, anterior; N, proximal (posterior towards bottom); O, posterior; and P, right lateral views. UBB NVM 1-8 in Q,

anterior; R, proximal (posterior towards bottom); S, posterior; and T, left lateral views. UBB NVM 1-15 in U, anterior; V, posterior;

and W, left lateral views. UBB NVM 1-9 in X, posterior view. UBB NVM 1-16 in Y, anterior; and Z, posterior views. UBB NVM

1-17 in A0, anterior; and B0, posterior views. Scale bar equals 50mm.

66 V. D



ıez D

ıaz et al.

Figure 48. Paludititan nalatzensis, clavicles (Individual D). UBB NVM 1-51 in A, internal; C, external; and E, anterior? views.

UBB NVM 1-52 in B, internal; D, external; and F, anterior? views. Scale bar equals 50 mm.

Figure 49. Paludititan nalatzensis, partial right pelvis (Individual D) including an almost complete pubis, ischium and acetabular

part of the ilium, in articulation (UBB NVM 1-13) in A, medial; and B, lateral views. Right femur UBB NVM 1-53 (Individual D)

in C, posterior view. The numbers 2, 3 and 4 indicate the autapomorphies described in the text: ischiadic peduncle of the ilium

unusually well developed dorsoventrally, with an anteroposteriorly elongate distal end; iliac peduncle of ischium bears a

posterolateral buttress (i.e. a raised area on the posterolateral margin of its proximal part; and respectively non-coplanar distal ischia.

Scale bar equals 100 mm.

Revision of Romanian sauropod dinosaurs 67

neural canal opening was set within a fossa, but this

does seem to be the case in UBB NVM 1-45, in which

the anterior neural canal opening is clearly dorsoven-

trally tall and elliptical. The right prezygapophysis is

preserved in UBB NVM 1-43 (Fig. 45C, D); although

almost certainly accentuated by crushing, its flat articu-

lar surface is strongly tilted to face dorsomedially. A

PRDL is preserved, but little information can be ascer-

tained regarding its morphology (Fig. 45D). The prezy-

gapophyseal articular surfaces in UBB NVM 1-45 are

flat and much more gently inclined (less than 30to the

horizontal) than in UVV NVM 1-43. It is not possible

to fully determine the morphology of the CPRL.

Although each CPRL appears to be single in UBB

NVM 1-45, poor preservation of their anterior surfaces

makes this uncertain. A small, poorly preserved TPRL,

which is ‘V’-shaped in anterior view, is present in UBB

NVM 1-43 (Fig. 45B).

Parapophyses are only preserved in specimens UBB

NVM 1-43 (the right parapophysis) and NVM 1-45. The

parapophysis of UBB NVM 1-43 is situated on the

lower part of the arch, extending ventrally to the cen-

trum-arch junction (Fig. 45D), suggesting that this was

an anterior dorsal vertebra, in approximately position

D3. This parapophysis has a dorsoventrally tall, ellip-

tical shape, and does not seem to extend beyond the

anterior margin of the non-condylar centrum. By con-

trast, the parapophyses are situated at the level of the

ventral margin of the prezygapophyseal articular surfa-

ces in UBB NVM 1-45 (Fig. 45N), suggesting that this

specimen is a middle or posterior dorsal vertebra. The

parapophyses in the latter specimen are much shorter

dorsoventrally than in UBB NVM 1-43. Each parapoph-

ysis is supported from below by a steep, anterodorsally

oriented ACPL. PCPLs are present on both sides of

UBB NVM 1-45, where they appear to be singular,

Figure 50. Paludititan nalatzensis, pedal unguals (Individual D). Right pedal ungual from digit I or II UBB NVM 1-10 in A, lateral;

B, ventral; C, medial; D, distal (medial towards top); and E, dorsal views. Left pedal ungual from digit III UBB NVM 1-10 in F,

lateral; G, dorsal; H, medial; and I, ventral views. Scale bar equals 50 mm.

68 V. D



ıez D

ıaz et al.

although neither side is well enough preserved to con-

firm the latter. A PPDL is present on both sides in UBB

NVM 1-43, extending posterodorsally at approximately

45to the horizontal. The PCDL, PPDL and parapophy-

sis bound a parapophyseal centrodiapophyseal fossa

(PACDF). The diapophyses are preserved in UBB NVM

1-43, and project dorsolaterally (Fig. 45B, E); although

this seems to be slightly more lateral than normal, this

is almost certainly the result of dorsoventral crushing of

the neural spine, and the diapophysis has also been very

slightly displaced. The distal end of the diapophysis of

UBB NVM 1-43 is incomplete, and therefore we cannot

determine the morphology of its dorsal surface. ACDLs

are present in UBB NVM 1-44 and 45. These laminae

are posterodorsally directed (most clearly seen on the

right side) in UBB NVM 1-44, and are distinct from the

ventral expansion of the PCDL. The ACDL curves post-

erodorsally, rather than being straight, in UBB NVM 1-

45. Csiki, Codrea, et al. (2010, p. 302, fig. 2C1)

regarded the morphology of the ACDL as an autapo-

morphy of Paludititan:‘the dorsal segment of the anter-

ior centrodiapophyseal lamina curves anterodorsally and

extends parallel to the dorsal segment of the posterior

centrodiapophyseal lamina’, which led those authors to

interpret the presence of an accessory ACDL. However,

we re-interpret this as a taphonomic artefact, resulting in

the posteroventral displacement of the ACDL, with no

evidence for an accessory lamina (Fig. 45N); as such,

we exclude this feature from our revised diagnosis of

Paludititan. The ACDL and PCDL define a CDF. A

PACDF or PRCDF (it is not certain which fossa is pre-

sent because of the poor preservation of the roofing

lamina, i.e. PPDL or PRDL) is present in UBB NVM 1-

45, anterior to the ACDL, and bounded anteriorly by

the ACPL and parapophysis/prezygapophysis. The

PCDL is near-vertical in UBB NVM 1-45, whereas it is

steeply oriented anterodorsally in the middle–posterior

dorsal vertebrae. It is slightly expanded at its ventral

end, but it is not bifurcated.

The left postzygapophysis is partially preserved in

UBB NVM 1-43, though distorted, whereas only the

very base of the right postzygapophysis remains (Fig.

45C–E). Because of the distortion to the spine, the

articular surface of the right postzygapophysis faces pri-

marily medially. The postzygapophyseal articular surfa-

ces are flat and face posteroventrally, as well as partly

laterally, in UBB NVM 1-45 (Fig. 45M, N). Neither

CPOL is clearly preserved in UBB NVM 1-43, although

it is possible to trace their approximate positions. The

bases of the CPOLs are preserved in UBB NVM 1-44,

where they project posterodorsally. In UBB NVM 1-45,

each CPOL appears to be single (i.e. unbifurcated) and

vertical (Fig. 45M, N). No PODLs are present in any of

these vertebrae, but this absence could reflect poor pres-

ervation. A POCDF is defined by the PCDL, CPOL and

SPDL in UBB NVM 1-43 and 45 (Fig. 45D, N).

The neural spine of UBB NVM 1-45 is the best-pre-

served of the three dorsal vertebrae, and it projects

strongly posterodorsally (Fig. 45K, M, N). In lateral

view, it has approximately parallel anterior and posterior

margins, with a slight degree of anteroposterior shorten-

ing dorsally. Its dorsal tip is not preserved, but probably

only a small amount of bone is missing, and clearly it

was not bifurcated. In anterior view, the neural spine of

UBB NVM 1-45 has a subtriangular outline, narrowing

in transverse width dorsally (Fig. 45L). There is evi-

dence for weakly developed aliform processes where the

SPOLs and SPDLs probably converge, a short distance

from the spine summit. The distal neural spine morph-

ology is very unclear because of its poor preservation in

all three dorsal vertebrae. There is no clear evidence for

SPRLs as a result of the poor preservation of much of

the anterior surface of the neural spine of UBB NVM 1-

43. These laminae are also not preserved in UBB NVM

1-45, but it is unlikely that they were prominent struc-

tures in life. There are remnants of what could be por-

tions of a possible midline PRSL at the base of the

spine, and again just above midheight of the latter, but

both areas might just be broken surfaces. A prominent

midline PRSL is present along the length of the neural

spine of UBB NVM 1-45 (Fig. 45N), although this flat-

tens close to the spine summit, with the anterior surface

of the spine becoming strongly convex transversely in

this region. A midline POSL is present in UBB NVM

1-43 and 45. The SPOLs of UBB NVM 1-43 are mainly

vertically oriented (with a slight medial deflection), and

are clearly restricted to the posterolateral margins of the

neural spine (Fig. 45D, N). In UBB NVM 1-45, these

laminae are poorly preserved, and it is not possible to

determine if they were bifid. One prominent, unbifur-

cated SPDL is present on each side in UBB NVM 1-45.

As preserved, the long axis of the SDF is essentially

horizontal, but this is clearly the result of taphonomic

damage.

Thoracic ribs. Csiki, Codrea, et al. (2010) reported that

a number of plank-like ribs are preserved in Paludititan

(Fig. 46). No pneumatic features are visible, although

no proximal ends are preserved. There is a hollow filled

with matrix visible in the cross-section in some of these

elements, which was interpreted as evidence for

pneumaticity by Csiki, Codrea, et al. (2010), but this is

present in many sauropod ribs and appears to have been

non-pneumatic (Waskow & Sander, 2014).

Caudal vertebrae. Seven anterior caudal vertebrae are

preserved in Paludititan, and can be ordered into a

Revision of Romanian sauropod dinosaurs 69

series: UBB NVM 1-58, 1-50, 1-3, 1-2, 1-1 (the latter

specimen comprises two articulated caudal vertebrae),

and 1-21 (Csiki, Codrea, et al., 2010, fig. 3). Of these,

UBB NVM 1-58 and 50 are more poorly preserved than

the others. Although from the anterior region of the tail,

none of these specimens represent the anterior-most cau-

dal vertebrae. The aEI of these seven vertebrae ranges

from 0.94 to 1.33, encompassing the typical range of

aEI values of anterior caudal centra observed in several

early-branching titanosaur taxa (see above). All of the

centra are procoelous, with a deeply concave anterior

articular surface that has a circular profile (Csiki,

Codrea, et al., 2010, fig. 3E), with the exception of

UBB NVM 1-21, which is elliptical. The posterior con-

dyle is prominent but has a constricted base, shifted

slightly dorsally (Csiki, Codrea, et al., 2010, fig. 3D, F),

and those of UBB NVM 1-1 and 21 are less well devel-

oped than the others. The ventral surface of each cen-

trum is flat (Csiki, Codrea, et al., 2010, fig. 3G, I),

rather than possessing a ventral hollow as described by

Mocho et al. (2023), and becomes more transversely

compressed passing distally along the series.

Ventrolateral longitudinal ridges are only present on the

centrum of some anterior caudal vertebrae (i.e. UBB

NMV 1-1 and 3). The posterior chevron facets are more

developed than the anterior ones. The articular surfaces

of the posterior chevron facets face posteroventrally,

and project far below the condyle, especially in UBB

NVM 1-3. In the more anteriorly placed UBB NVM 1-1

vertebra, there is a low longitudinal ridge on the lateral

surface of the centrum at approximately midheight. This

ridge demarcates where the lateral surface meets the

ventrolaterally facing surface that extends toward the

ventral midline (see also Salgado & Garc

ıa, 2002).

However, there is no equivalent morphology in the

more posterior caudal vertebra UBB NVM 1-3.

The transverse processes are located below the junc-

tion between the neural arch and the centrum (Csiki,

Codrea, et al., 2010, fig. 3C, D, H). These processes are

large rounded bulges in UBB NVM 1-3 and 1-1,

whereas they are smaller and posterolaterally oriented in

UBB NVM 1-2. In the latter vertebra, and in UBB

NVM 1-50, the transverse processes are dorsally delim-

ited by a prominent ridge. The prezygapophyses are

long, and become more anterodorsally directed passing

distally along the series. They are not connected by a

distinct TPRL at their bases, unlike those of middle cau-

dal vertebrae (see below). The articular surfaces of the

prezygapophyses face dorsomedially in UBB NVM 1-3,

and entirely medially in UBB NVM 1-1. In UBB NVM

1-3 and 1-1, the postzygapophyses are short, with lat-

erally facing surfaces with a circular outline. The neural

spine is low and nearly vertical, being slightly

posterodorsally inclined. Both SPRLs and SPOLs are

present on the neural spine as single (i.e. unbifurcated)

laminae (Csiki, Codrea, et al., 2010, fig. 3D, H). SPRLs

arise on the dorsal aspect of the prezygapophyses lat-

erally, and the next end posteriorly and medially to

meet the PRSL approximately at its midlength. The

SPOLs extend lateral to the POSL. The PRSL and

POSL are wide and rugose laminae that extend over

almost the complete length of the neural spine. The dor-

sal tips of the PRSL and POSL are slightly anteriorly

and posteriorly extended as ‘spurs’, probably for the

attachment of the deep epaxial musculature. In the

plate-like neural spines of UBB NVM 1-3 and 1, a deep

SPRF is present between the SPRLs. A deep SPOF is

also present between the two vertical SPOLs.

Four middle caudal vertebrae are preserved in

Paludititan (UBB NVM 1-46, 1-47, 1-48 and 1-49)

(Csiki, Codrea, et al., 2010, fig. 4A–F). The centra have

a concave to flattened anterior surface. UBB NVM 1-48

is procoelous, whereas UBB NVM 1-49 has a flat pos-

terior articular surface. UBB NVM 1-46 is a slightly

procoelous to amphyplatyan caudal vertebra, and UBB

NVM 1-47 is platycoelous. UBB NVM 1-46 and 1-47

have articular surfaces with circular outlines, whereas

the centrum is transversely compressed in UBB NVM

1-48 and 1-49. The posterior articular surfaces of UBB

NVM 1-46, 1-47 and 1-49 have a heart-shaped outline,

but with a rounded ventral edge (Csiki, Codrea, et al.,

2010, fig. 4C). The ventral surface of the centrum is flat

and slightly transversely constricted in UBB NVM 1-47,

1-48 and 1-49, whereas no transverse constriction is pre-

sent in the similarly flat ventral surface of UBB NVM

1-46. No ridges are present on the ventral surface of

any specimen. A prominent longitudinal ridge is present

close to the midheight of the lateral surface of the cen-

trum in UBB NVM 1-49. All specimens possess a lon-

gitudinal ridge demarcating the junction between the

centrum and the anteriorly located neural arch. In UBB

NVM 1-47 and 1-49, these ridges are approximately

half the length of the base of the neural arch, but they

are more posteriorly extensive in UBB NVM 1-46. The

prezygapophyses are short, oriented almost horizontally,

and are joined at their bases by a well-developed TPRL.

The postzygapophyseal articular facets, preserved in

UBB NVM 1-46 and 49, are ellipsoidal, with their

articular surfaces oriented ventrolaterally. The neural

spine is not complete in any of the four specimens, but

was probably low, long and posteriorly directed, with a

slight dorsal deflection. Two sharp SPRLs delimit a

deep SPRF in UBB NVM 1-49 (Csiki, Codrea, et al.,

2010, fig. 4F). Between the short and almost vertical

SPOLs, a deep SPOF is also present in this specimen.

70 V. D



ıez D

ıaz et al.

Chevrons. Of the 12 chevrons described by Csiki,

Codrea, et al. (2010), 10 could be located by us in the

UBB collections (UBB NVM 1-4, -5, -6, -7, -8, -9, -16,

-17, -18 and -24, Fig. 47). UBB NVM 1-6 is the largest

chevron of the available sample (Fig. 47I–L), whereas

UBB NVM 1-9 seems to have been the smallest chev-

ron, although it is not complete (Fig. 47X). Only the

proximal part of a transversely compressed blade is pre-

served of chevron UBB NVM 1-18. Although the ana-

tomical order of these chevrons is not known on the

basis of articulation with caudal vertebrae or other field

associations, a combination of relative size and morpho-

logical gradients can be used to infer an approximate

sequence. From anterior-most to posterior-most, this

inferred sequence is: UBB NVM 1-16, 1-17, 1-24, 1-6,

1-4, 1-5, 1-9 (N.B. the relative positions of the remain-

ing three available chevrons are more uncertain, but

they are likely to have been located between 1-5 and 1-

9). These elements sample those ranging from near the

anterior end of the tail to approximately its middle

(which is consistent with the seven anterior and four

middle caudal vertebrae preserved in this partial skel-

eton), but are unlikely to represent a continuous series.

If our estimated sequence is correct, it seems that the

chevrons increased in length from UBB NVM 1-16

(Fig. 47Y, Z) to UBB NVM 1-6 (Fig. 47I–L), and then

shortened again towards the middle of the tail.

All preserved chevrons are ‘Y’-shaped in anterior/pos-

terior view (i.e. the haemal canal is not bridged by bone

proximally), although the proximal rami diverge only

slightly from one another. The ratio between the proxi-

modistal length of the haemal canal and that of the

entire chevron varies between 0.36 and 0.57, with

higher ratios in the more anteriorly placed elements.

The anterior-most chevrons (UBB NVM 1-16 and 1-17,

Fig. 47Y–AB) have single articular facets, whereas the

other elements have proximal facets that are divided by

a transverse groove that is anteroposteriorly widest at its

medial end. The resultant anterodorsal articular surface

is slightly larger than its posterodorsal counterpart,

except in UBB NVM 1-7 (Fig. 47N) and 8 (Fig. 47J),

where these facets are equidimensional. As noted above,

such a groove is absent from the chevrons of

Magyarosaurus, but Paludititan shares this feature with

several South American titanosaurs (Poropat et al.,

2016; Powell, 2003; Santucci & Arruda-Campos, 2011),

as well as Lohuecotitan (D



ıez D

ıaz et al., 2016). The

proximal rami of Paludititan chevrons lack the lateral

ridges present in Magyarosaurus and several other tita-

nosaurs (see above). In the anterior-most chevrons

(UBB NVM 1-16 and 1-17), the distal blade has a sub-

circular cross-section. The next two chevrons in our

inferred sequence (UBB NVM 1-24 and 1-6) display the

onset of more transverse compression of the distal

blade, but retain strongly convex (anteroposteriorly) lat-

eral surfaces. This trend continues in the remaining

chevrons, such that the distal blade is transversely com-

pressed and anteroposteriorly expanded, with a rounded

distal end profile. The more anterior chevrons (UBB

NVM 1-16, 1-17, 1-24, 1-6 and 1-4) are nearly straight

in lateral view, but the distal blade starts to curve more

posteriorly relative to the proximal rami from UBB

NVM 1-5 onwards. From this list of anterior chevrons

only UBB NVM 1-24 presents a rounded, slightly

anteriorly located, prominence on the lateral surface at

the junction of the ramus and the blade, probably for

the attachment of the ventral part of the M. caudofemor-

alis longus. Anterior and posterior midline crests are

present along the length of the distal blade of this chev-

ron. However, in UBB NVM 1-5 (Fig. 47H) no bulges

or rugosities are visible, which could indicate that the

distal tip of the M. caudofemoralis longus was located

anteriorly to this section of the tail.

The best-preserved and most complete chevrons in

the sample present the ‘open Y’-shaped morphology

described by Otero et al. (2012, fig. 2).

Clavicle. Two unidentified bones (UBB NVM 1-51, 52,

Fig. 48) were included within the holotype, but not

described by Csiki, Codrea, et al. (2010). Following

Tschopp and Mateus (2013), we tentatively interpret

these specimens as ‘morphotype B’-type clavicles. Both

are slender, elongate, curved elements. These taper at

their lateral (?) tip and expand transversely as they

approach the medial (?) articulation. UBB NVM 1-51

(Fig. 48A, C, E) is more anteroposteriorly compressed

than UBB NVM 1-52 (Fig. 48B, D, F). The medial (?)

half of UBB NVM 1-52 has a convex external surface

(Fig. 48D), whereas the internal one is flat (sensu

Tschopp & Mateus, 2013)(Fig. 48B). The specimen is

gently curved, and the convex edge has a crest that ter-

minates before reaching the lateral (?) tip. The lateral

(?) half has a more subcircular cross-section. The gen-

eral morphology of these specimens contrasts with the

typical ‘L’-shaped clavicles found in some diplodocids

(Schwarz et al., 2007; Tschopp & Mateus, 2013), but

they are not as straight as that in the non-neosauropod

eusauropod Spinophorosaurus (Remes et al., 2009). This

is the first report of a clavicle in a titanosauriform

sauropod.

Ilium. The right ilium UBB NVM 1-13 (Fig. 49A, B)

preserves the acetabular region, as well as portions of

the preacetabulum and postacetabulum (Csiki, Codrea,

et al., 2010, fig. 5). Because the specimen is very

incompletely preserved, it is not possible to determine

the orientation of the preacetabular process, nor is it

Revision of Romanian sauropod dinosaurs 71

possible to be entirely certain of the orientation of the

pubic peduncle relative to the long axis of the ilium. No

triangular fossa is present on the lateral surface of the

upper portion of the pubic peduncle. The distal end of

this peduncle has a comma-shaped outline that tapers in

anteroposterior width medially, with a transversely long,

straight to mildly concave posterior margin. In this

regard, it clearly differs from the distal pubic peduncle

of Petrustitan (see above, Fig. 39). The pubic peduncle

appears to be unusually dorsoventrally elongate in

Paludititan, but this might be a false impression given

by the incomplete nature of the main body of the ilium.

The ischiadic peduncle is unusually well developed in

terms of dorsoventral extent and distal anteroposterior

width compared to other eusauropods (see also Csiki,

Codrea, et al., 2010, fig. 5A, C), which we regard as an

autapomorphic feature of Paludititan. The distal end of

the ischiadic peduncle has a comma-shaped outline,

decreasing in anteroposterior width medially. This trans-

versely convex articular surface is strongly deflected

to face laterally, as well as ventrally and very slightly

posteriorly. No tubercles are present on the lateral

surface of the ischiadic peduncle, such as those

described in Diamantinasaurus,Opisthocoelicaudia and

Rapetosaurus (Borsuk-Białynicka, 1977; Curry Rogers,

2009; Poropat, Upchurch, et al., 2015). There is some

evidence for pneumatic pockets on the broken dorsal

margin of the element, suggesting that Paludititan pos-

sessed the derived pneumatized ilium seen in many

other somphospondylans (Mannion et al., 2013; Wilson

& Sereno, 1998).

Pubis. The right pubis UBB NVM 1-13 (Fig. 49A, B)

is almost complete in terms of length (Csiki, Codrea,

et al., 2010, fig. 5), but its distal end is poorly pre-

served, and the posterior margin is incomplete distal to

the ischiadic articulation. Despite the relevant area being

preserved, albeit fractured, the obturator foramen cannot

be observed. The anterior margin of the proximal end is

too poorly preserved to determine whether or not an

ambiens process was present. The pubic blade is

strongly twisted relative to the proximal end, such that

what would usually be described as the lateral surface

faces primarily posteriorly. This posterolateral surface is

gently convex anteroposteriorly along its proximal

half, but flattens distally. It lacks any distinct

longitudinal ridge and groove, contrasting with the

lateral surface of the pubis of several titanosaurs,

including Epachthosaurus,Isisaurus, saltasaurines and

Uberabatitan (Otero, 2010; Poropat et al., 2016; Powell,

2003; Salgado & Carvalho, 2008). The distal end is too

poorly preserved to unequivocally state that it did not

expand mediolaterally, although such an expansion

seems unlikely.

Ischium. The right ischium UBB NVM 1-13 (Fig. 49A,

B) is essentially complete (Csiki, Codrea, et al., 2010,

fig. 5), with only the very distal end poorly preserved,

and some of the anteroventral margin of the distal blade

missing. The iliac peduncle has a raised area on the post-

erolateral margin of its proximal end, described as a post-

erolateral buttress by Csiki, Codrea, et al. (2010), who

noted that this feature is otherwise only known in a spe-

cimenattributedtoAmpelosaurus. We herein follow

these authors in regarding this as a local autapomorphy

of Paludititan. The medial margin of the acetabulum is

broken away, such that the presence of an anteromedial

flange on the iliac peduncle cannot be ascertained, and it

is also uncertain whether the acetabulum maintained a

constant mediolateral width. It appears that there is no

prominent upturning of the anterodorsal corner of the

acetabulum, such as that which characterizes many titano-

saurs (D’Emic, 2012), although this region is not fully

preserved. On the lateral surface of the ischium, close to

its posterior margin and situated at approximately mid-

height of the pubic plate, a very prominent ridge marks

the attachment of M. flexor tibialis internus III. This

ridge projects laterally and is not associated with a

groove. The long axis of the distal blade, if extrapolated

anteriorly, passes through the pubic articular surface, and

forms a right angle with a line drawn between the anter-

ior margin of the proximal end of the iliac peduncle and

the anterodorsal corner of the pubic peduncle. The ventral

margin of the blade is too incomplete to determine the

nature of the symphysis between the ischia. The distal

end of the ischium is little expanded mediolaterally or

dorsoventrally relative to the more proximal part of the

blade. Based on the orientation of the ischium with the

iliac articular surface held in a horizontal plane, it seems

unlikely that the long axes of the distal end surfaces of

the ischia would have been coplanar. This contrasts with

most somphospondylans (Upchurch, 1998;Wilson&

Sereno, 1998), in which the derived coplanar condition is

present, although Muyelensaurus,Neuquensaurus,

Patagotitan and Rinconsaurus also lack coplanar distal

ischia (Mannion, Upchurch, Jin, et al., 2019).

Femur. The distal end of a femur (UBB NVM 1-53)

was recovered, although it is poorly preserved (Fig. 49C).

It is probably a right element, although this is difficult to

confirm because it is missing a large amount of the distal

articular region, only preserving the posteromedial portion

of what we interpret as the tibial condyle. If correct, then

the tibial condyle is narrower mediolaterally than the

fibular distal condyle, almost certainly displaying the

derived state that characterizes most somphospondylans,

wherein the tibial to fibular condyle width ratio is <0.8

(Poropat et al., 2016). The distal articular surface is

strongly convex anteroposteriorly, curving up onto the

72 V. D



ıez D

ıaz et al.

posterior surface. Although Csiki, Codrea, et al. (2010)

noted that the shaft cross-section is not as anteroposter-

iorly compressed (ECC ¼1.69) as occurs in some titano-

saurs, this fragment does not preserve the midshaft where

the measurements for calculating eccentricity are nor-

mally taken.

Pedal unguals. Two pedal unguals are known; UBB

NVM 1-10 (Fig. 50A–E), a right pedal ungual from

either digit I or II (Csiki, Codrea, et al., 2010); and

UBB NVM 1-11 (Fig. 50F–I), a left pedal ungual from

digit III. The general morphology of both unguals, cres-

cent-shaped and transversely compressed, is typical for

eusauropods (Wilson & Sereno, 1998), and is retained

by most titanosaurs, e.g. Dreadnoughtus,Rapetosaurus

and Mendozasaurus (Curry Rogers, 2009; Gonz

alez

Riga et al., 2018; Ullmann & Lacovara, 2016). It is not

possible to discern the nature of the proximal bevelling

of UBB NVM 1-10, because its proximal articular sur-

face is poorly preserved along its medial half. However,

the proximal articular surface of UBB NVM 1-11 faces

proximolaterally relative to the long axis of the ungual,

such that it would have projected anterolaterally in life,

as is typical for most eusauropods (Poropat et al., 2016;

Wilson & Upchurch, 2009). The proximal ends have

dorsoventrally elongated elliptical outlines (Fig. 50D).

The ventral margin of the proximal end forms a projec-

tion (flexor tubercle) that is somewhat more prominent

in UBB NVM 1-11 than in UBB NVM 1-10. The dorsal

margin of the proximal end is incomplete in the latter

specimen, but is preserved in the former, where it proj-

ects less prominently than the ventral one. In medial/lat-

eral views, the dorsal margin of each ungual is

convex. The ventral surfaces are rugose on the distal

third of the claw, with a well-developed tubercle, as in

other somphospondylans (Canudo et al., 2008; Mannion

et al., 2013), including the titanosaurs Dreadnoughtus

(Ullmann & Lacovara, 2016), Epachthosaurus (Mart

ınez

et al., 2004), Malawisaurus (Gomani, 2005), and

Mendozasaurus (Gonz

alez Riga et al., 2018). Distal to

this ventral tubercle, both the dorsal and the ventral

margins narrow dorsoventrally towards the distal tip in

UBB NVM 1-10. The ventral margin is concave

throughout its length in UBB NVM 1-10 in lateral/med-

ial views, whereas it is only concave proximal to the

ventral tubercle in UBB NVM 1-11. Distal to this tuber-

cle, in the latter ungual, the ventral margin is deflected

to face anteroventrally.

Additional individuals and assemblages

In this section we describe individuals and assemblages

whose elements do not present sufficient similarities

with those referred to the nominal titanosaur taxa

described above to justify referral, and/or whose remains

are not diagnostic enough to erect new taxa.

Nonetheless, they help to understand the titanosaurian

diversity of the Hat¸eg Basin and are worth reporting

here in case more complete and associated material is

found in the future.

Individual H (Green Four titanosaur)

All of the specimens referred to Individual H (see ‘Key

localities and skeletal associations’section) are acces-

sioned as NHMUK R.4891, and comprise a partial dor-

sal vertebra, two partial dorsal ribs, a left scapula,

coracoid and sternal plate, and three right metacarpals.

Dorsal vertebra. This specimen preserves the posterior

end of the centrum and part of the arch (Fig. 51A, B).

The centrum is camellate throughout. Despite the cen-

trum being heart-shaped in posterior view, there is no

ventral midline keel. The posterior end of the lateral

pneumatic opening is preserved. This opening is deep,

although the midline septum appears to have been medi-

olaterally thick, but it is possible that the septum

thinned anteriorly. No further detailed comparisons with

remains attributed to the named titanosaurs of the Hat

,eg

Basin can be made.

Dorsal ribs. Both dorsal ribs are broken into two por-

tions that fit together (Fig. 51C–F). First, we describe

the best-preserved element (Fig. 51C, D), which prob-

ably came from the left side and is likely to have been

either the most anterior rib or a very posterior one,

based on its non-plank-like shaft. The capitulum and

tuberculum are nearly equally prominent, and diverge

from each other at approximately 80. Thus, with the

tuberculum directed dorsally, the capitulum projects

medially and slightly upwards. Both of these articula-

tions are compressed anteroposteriorly, but much of this

is due to crushing. The tuberculum is slightly shorter

and somewhat wider than the capitulum. Although dam-

aged, there appears to have been a prominent vertical

plate-like ridge on the anterior surface of the proximal

end, extending distally from the base of the tuberculum,

and curving slightly medially towards its distal end.

This ridge is deflected anteromedially, creating a groove

between itself and the anterior face of the base of the

capitulum. This ridge fades out rapidly distally and does

not extend onto the shaft. There is damage to the anter-

ior and posterior faces of the proximal end, but there is

some indication of a pocket-like depression on the pos-

terior surface below the base of the tuberculum. In

anterior/posterior views, the lateral margin of the prox-

imal end is concave over the transition from shaft to

tuberculum. Below the proximal end, the shaft starts

Revision of Romanian sauropod dinosaurs 73

with an anteroposteriorly compressed ‘D’-shaped hori-

zontal cross-section, with convex anterior and flat to

medially concave posterior surfaces. This rapidly nar-

rows in transverse width as it extends distally, though it

maintains a cross-section that is wider transversely than

anteroposteriorly throughout its length. The broken dis-

tal portion has an oval cross-section, with an anteropos-

teriorly wider and more rounded medial margin and a

narrower sharper lateral one.

The second, less well-preserved rib preserves the

damaged bases of the tuberculum and capitulum (Fig.

51E, F). Again, this is a left rib, probably from the

anterior part of the series. The proximal shaft is plank-

like, being at least three times wider transversely than

anteroposteriorly. There is again evidence of a pocket-

like structure on the posterior surface of the proximal

end, below the base of the tuberculum. There is a low

rounded ridge, trending in the same direction, as the

plate-like structure seen on the anterior surface of the

proximal end in the previous rib. The distal blade is

strongly compressed anteroposteriorly and again has a

more rounded medial and sharper lateral margin.

Scapula. This specimen is a left scapula missing the

dorsal portion of the acromion and the distal tip of the

blade (Fig. 52A–D). There are large amounts of plaster

reconstruction, especially along the blade: therefore, it is

difficult to be confident about the anatomy of the dorsal

and ventral margins. The acromial plate seems to have

been thin transversely over most of its extent, apart

from the stout region posteriorly formed from the acro-

mial ridge. It is not possible to determine the orientation

of the coracoid articular surface due to incompleteness.

The glenoid articular surface has a dorsoventrally elon-

gated ‘D’-shaped outline (with a flat lateral margin) and

is deflected to face medially as well as anteroventrally.

The medial edge of the glenoid forms a slightly projec-

ting lip. Anterior to the acromial ridge, the lateral

surface of the acromion is gently concave and feature-

less. The medial surface of the acromial plate is moder-

ately concave over most of its preserved extent. The

acromial ridge is anteroposteriorly broad, but not espe-

cially prominent, and is incomplete dorsally. There is no

evidence for a fossa posterior to the acromial ridge. The

angle between the long axes of the acromial ridge and

the blade is acute, approximately 70, but caution is

required because most of the ridge is absent. There is a

small tubercle on the ventral margin of the posterior end

of the acromion; a subtle additional tubercle is also situ-

ated slightly anteriorly, on the medial margin of the ven-

tral surface. No tubercle on the ventral margin of the

anterior end of the blade is evident, but this region is

heavily reconstructed with plaster. There are no

tubercles on the medial surface of the acromion, or in

the region where the acromion meets the blade.

The scapular blade is straight, projecting posteriorly

and slightly medially relative to the acromial plate. It is

dorsoventrally narrow, with no indication of expansion

towards its distal end, but this is possibly due to damage

in this area. The ventral margin of the blade is concave

in lateral view, and the dorsal margin is slightly sinuous,

but we cannot be certain that either morphology is

accurate because of the large amount of plaster recon-

struction. The lateral surface of the proximal third of the

blade is gently convex dorsoventrally. Consequently, the

cross-section through the base of the blade is weakly

‘D’-shaped. The convexity of the lateral surface of the

blade does not create a clear longitudinal ridge, and dis-

tally this surface becomes predominantly flat. The med-

ial surface of the blade is dorsoventrally concave

proximally, but is flat for most of its length. Just distal

to the acromial plate, the dorsal margin of the blade

gives rise to a prominent but short ridge on the medial

surface, forming a short dorsal platform. A similar ridge

characterizes the scapula of several titanosaurs,

Figure 51. Lithostrotia incertae sedis, Individual H, dorsal vertebra and ribs NHMUK R.4891. Dorsal vertebra in A, posterior; and

B, left views. ?Left dorsal rib in C, posterior; and D, anterior views. ?Left dorsal rib in E, anterior; and F, posterior views. Scale bar

equals 50 mm in A, B and 100 mm in C–F.

74 V. D



ıez D

ıaz et al.

including Abditosaurus,Lirainosaurus and saltasaurines

(Sanz et al., 1999; Vila et al., 2022). Unlike the pectoral

girdles of saltasaurines (Cerda et al., 2012), there is no

evidence of pneumaticity.

Coracoid. This specimen is a left coracoid that pre-

serves the glenoid and parts of the main body (Fig. 52E,

F). The only anatomical information that can be gleaned

is that the glenoid articular surface does not extend onto

the lateral surface of the coracoid. In this regard the cor-

acoid is similar to that of several other titanosaurs,

including Lirainosaurus and saltasaurines (Poropat

et al., 2016).

Sternal plate. This specimen was figured by Huene

(1932, pl. 47, fig. 11), who identified it as a left sternal

plate, an identification that we follow here. It is pre-

served in three pieces (along with several small frag-

ments) and lacks its posterior end (Fig. 52G, H). The

lateral margin is concave, whereas the medial margin is

convex. Although incomplete posteriorly, there is a

strong posterolateral expansion: thus, the sternal plate

clearly had the kidney-shape typical of later-diverging

titanosaurs (Upchurch, 1998). The sternal plate thickens

dorsoventrally anteriorly, with a prominent ventral ridge

situated close to the lateral margin. Generally, the dorsal

surface is gently convex anteroposteriorly. There is

evidence for a ridge-like tubercle on the dorsal surface,

situated a short distance from the medial margin and at

approximately one-third of the element length from the

anterior end, but the sternal plate is heavily damaged in

this region.

Metacarpals I–III. These three right metacarpals were

figured by Huene (1932, pl. 46, fig. 10) (Fig. 53). They

are largely complete and articulated, although metacar-

pal I is now separated from the other elements. All three

have been broken into several pieces and subsequently

re-glued and their relative positions differ from that

described by Huene (1932). Although it is possible that

they have been reassembled incorrectly, we follow their

current condition in our interpretation of their identifica-

tion. The metacarpals are straight and expanded at their

ends relative to their shafts. No channel-like ventral

concavities are visible. The distal articular surfaces do

not extend onto the dorsal sides of the metacarpals.

Metacarpal II is the longest of the three, and metacarpal

I is also slightly longer than metacarpal III. This feature

–with metacarpal II being the longest, then metacarpal

I and metacarpal III the shortest of the three –is also

observable in the titanosaur Epachtosaurus, while

Alamosaurus and Opisthocoelicaudia present both meta-

carpal I and II of almost equal proximodistal lengths,

Figure 52. Lithostrotia incertae sedis, Individual H, scapular girdle elements NHMUK R.4891. Left scapula in A, lateral; B, dorsal;

C, medial; and D, ventral views. Left coracoid in E, lateral; and F, medial views. Left sternal plate in G, ventral; and H, dorsal

views. Scale bars equal 100 mm.

Revision of Romanian sauropod dinosaurs 75

and longer than metacarpal III (Poropat, Mannion, et al.,

2015, table 13).

Metacarpal I. The proximal end of this metacarpal has

a‘D’- or comma-shaped outline, with a prominent

dorsolateral projection (Fig. 53A). This profile is created

by: a dorsoventrally elongate, mildly concave, lateral

margin; a straight ventral margin; a convex medial mar-

gin; and a dorsal margin that slopes to face dorsome-

dially. This proximal morphology contrasts with that of

the metacarpal I (MCDRD 152) referred to Petrustitan

(Fig. 38A). In the current specimen, the proximal articu-

lar surface is gently convex in all directions. The ventro-

lateral corner of the proximal end continues distally as a

low rounded ventral ridge, fading out at approximately

two-thirds of the element length from the proximal end,

at which point it is situated equidistant from the medial

and lateral margins. In dorsal view, the medial margin

of the metacarpal is fairly straight whereas the lateral

margin is concave. The metacarpal appears to bow

slightly medially in dorsal view, although breaks in the

shaft mean that this morphology might not be reliable.

There are no marked ridges or bumps on the shaft other

than the ventrolateral ridge noted above. The shaft rap-

idly contracts to form a rounded subtriangular cross-sec-

tion in the proximal part, contrasting with the circular

one in Petrustitan.

A short distance from the distal end, the medial mar-

gin forms a small concavity (clearly visible in both

dorsal and ventral views). The bone texture differs on

either side of this concavity, potentially marking a

healed break. Distally, the shaft becomes subtriangular

in cross-section. In distal view, the metacarpal has a

subtriangular profile, with a long, gently convex dorsal

margin, and sloping, gently concave medial and lateral

margins (Fig. 53E). The ventrolateral corner of this tri-

angle forms a small but distinct process that would have

projected into the central cavity of the manus formed by

all five metacarpals when articulated in life (i.e. the cen-

tral hollow within the tubular manus). This projection

makes the ventral, and especially the lateral margins,

moderately concave in outline.

Metacarpal II. In proximal view, this metacarpal has a

mildly asymmetrical subtriangular shape, with the med-

ial margin slightly longer than the lateral one (Fig.

53A). The dorsal and lateral margins meet at a near

right angle, whereas the dorsomedial corner is rounded,

and the apex of the triangular proximal profile projects

primarily ventrally. The proximal articular surface is

gently convex in all directions. The triangular cross-sec-

tion is maintained over the proximal four-fifths of the

metacarpal, with a prominent ridge along the ventral

margin (Fig. 53B). This ridge terminates close to the

distal end, approximately equidistant from the medial

and lateral margins, at the point where the metacarpal

broadens mediolaterally. In dorsal view, both the medial

and lateral margins of the metacarpal are concave. The

Figure 53. Lithostrotia incertae sedis, Individual H, right metacarpals I to III in articulation, NHMUK R.4891. A, proximal (ventral

towards bottom); B, ventral; C, medial; D, dorsomedial; E, distal (ventral towards top); and F, dorsolateral views. Scale bar equals

50 mm.

76 V. D



ıez D

ıaz et al.

dorsal surface is concave proximodistally, giving the

metacarpal a bowed appearance in medial view. Just

proximal to midlength, there is a prominent, anteropos-

teriorly elongate tubercle on the medial surface –this

extends to the dorsal margin (and is visible in dorsal

view, Fig. 53D), but it is not possible to determine its

ventral extent because the bone surface is missing. The

medial margin of the distal third of the metacarpal

forms a prominent ridge, although this is broken distally

(Fig. 53C). In distal view, the metacarpal has a medio-

laterally elongate ‘D’-shaped outline created by a verti-

cal lateral margin, and sloping dorsal and ventral edges

that form a rounded medial margin (Fig. 53E). The dis-

tal articular surface is strongly convex in all directions.

As in metacarpal I, the ventrolateral corner of the distal

end forms a distinct projection that extends towards the

centre of the tubular manus.

Metacarpal III. This specimen is attached to metacar-

pal II and so aspects of its medial, as well as the latter’s

lateral surface are obscured, especially where they

articulate proximally and distally. There are breaks and

glued repairs throughout the shaft in at least three pla-

ces. The proximal end has a nearly symmetrical, dorso-

ventrally tall, triangular outline, with the apex of this

triangle projecting ventromedially (Fig. 53A). A triangu-

lar cross-section is maintained for approximately the

proximal three-quarters of the metacarpal length,

although the ventral margin is broad, rather than form-

ing a distinct ridge (N.B. this surface is not preserved

along the proximal third). By mid-length the shaft is

more rounded in cross-section, being subrectangular and

slightly transversely compressed. The medial margin of

the metacarpal is fairly straight in dorsal view, whereas

the lateral margin is concave. As in metacarpal II, meta-

carpal III is bowed in medial/lateral view, with the dor-

sal surface concave anteroposteriorly. There are no clear

additional ridges or bumps on the shaft. The distal end

is expanded, especially transversely, compared to the

mid-shaft. In distal view, the metacarpal has a trapez-

oidal outline, created by: a long, flat dorsal margin; a

medial margin that is nearly vertical and dorsoventrally

taller than the lateral margin; a lateral margin that slopes

such that it faces ventrolaterally; and a ventral margin

that is gently concave (Fig. 53E). The distal articular

surface is gently convex dorsoventrally, and mildly con-

cave mediolaterally.

Individual I (Tus

,tea titanosaur)

The most important element preserved for this individ-

ual (see the ‘Key localities and skeletal associations’

section) is the sacrum (currently only observable in ven-

tral view) in articulation with the partial right ilium

(LPB [FGGUB] R.2345), alongside two posterior caudal

vertebrae (LPB [FGGUB] R.2027 and 2028), and four

chevrons (LPB [FGGUB] R.1881, 2025, 2026 and

2184).

Sacrum. The sacrum (LPB [FGGUB] R.2345) com-

prises five vertebrae (Fig. 54). The original sacral verte-

bra 1 is probably missing, given that somphospondylan

sacra usually possess six vertebrae (Upchurch, 1998;

Wilson & Sereno, 1998), although it is possible that

some titanosaur taxa could have presented the plesio-

morphic condition of a sacrum with five vertebrae

(Averianov et al., 2023). The total preserved anteropos-

terior length of the sacrum is c. 290 mm. It is dorsoven-

trally compressed, with this compression also visible in

the horizontal preacetabular process of the ilium. The

centra are fused, although the sutures between them are

visible. The most anterior centrum (S2?) presents a con-

dyle on its anterior articular surface, and the posterior-

most centrum (S6?) has a concave posterior articular

surface. On the ventrally exposed parts of the first pre-

served vertebra (S2?), long oval longitudinal pneumatic

structures can be observed on the lateral and ventral

sides. The ventral surfaces of the centra vary through

the series: only the first one (S2?) is flat, the second

and third (S3? and S4?) have an irregular surface, the

fourth (S5?) presents a narrow and shallow, longitudin-

ally elongated concave surface on the midline bounded

by two poorly developed ridges, and the fifth (S6?) has

a subtriangular outline with a concave area on its poster-

ior part. The lateral edges of this ‘triangle’form smooth

ventrolateral ridges on the posterior-most centrum. The

lateral surfaces of the centra bear anterodorsally placed,

deep pneumatic openings, most clearly observed in the

first preserved sacral vertebra (S2?). In S2?, the prezy-

gapophyses are wide and short, and anterodorsally

directed, with the articular facets facing dorsomedially.

Below the prezygapophyses, a CPRF is present.

No transverse foramina are present between the centra

and the ribs. The sacral ribs are anteroposteriorly com-

pressed and dorsoventrally expanded, especially in their

distal part, where they form the sacricostal yoke and

connect to the ilium. There were probably intercostal

junctions between the ribs forming the sacricostal yoke,

as can be observed between the first (S2?) and second

(S3?) right sacral ribs, whose distal ends are connected.

The presence of intercostal foramina cannot be assessed

because of the state of preservation.

Caudal vertebrae. LPB (FGGUB) 2027 is a posterior

caudal centrum (Fig. 55A–E). It is procoelous, with a

weakly developed posterior condyle. Although slightly

affected by taphonomic deformation, this centrum is

probably genuinely dorsoventrally compressed: the

Revision of Romanian sauropod dinosaurs 77

height to width ratio of the anterior articular surface is

0.61, whereas most of the other titanosaurian posterior

caudal vertebrae from the Hat¸eg Basin show values

closer to 1 (Table S1). This is the only posterior caudal

vertebra to display such an advanced degree of vertical

compression out of all those from the Hat¸eg Basin

studied in this work, as well as in the large sample sur-

veyed by Mocho et al. (2023). Some posterior caudal

vertebrae with strongly dorsoventrally compressed cen-

tra have also been described previously for

Figure 54. Individual I sacrum LPB (FGGUB) R.2345 in ventral view. The dotted lines highlight the ventral ridges. Scale bar equals

100 mm.

Figure 55. Lithostrotia incertae sedis, Individual I, posterior caudal vertebrae and chevrons. Posterior caudal vertebra LPB (FGGUB)

R.2027 in A, anterior; B, left lateral; C, posterior; D, dorsal; and E, ventral views. Posterior caudal vertebra LPB (FGGUB) R.2028

in F, left lateral; G, right lateral; H, posterior; I, ventral; and J, anterior views. Chevron LPB (FGGUB) R.1881 in K, anterior; L,

right lateral; and M, posterior views. Chevron LPB (FGGUB) R.2026 in N, anterior; O, left lateral; and P, posterior views. Chevron

LPB (FGGUB) R.2184 in Q, anterior; R, right lateral; and S, posterior views. Chevron LPB (FGGUB) R.2025 in T, anterior; U,

right lateral; and V, posterior views. Abbreviations: PO, postzygapophysis; PRE, prezygapophysis. Scale bars equal 40 mm.

78 V. D



ıez D

ıaz et al.

Atsinganosaurus (Garcia et al., 2010, fig. 4D–G) and

Lirainosaurus (D



ıez D

ıaz et al., 2013b, fig. 6K–R).

Indeed, a dorsoventral compression of the centrum of

the first posterior caudal vertebra is one of the changes

that can occur in the caudal series, as observed in the

tails of Bonitasaura salgadoi (Gallina & Apestegu

ıa,

2015) and Rinconsaurus caudamirus (P

erez Moreno

et al., 2022). In LPB (FGGUB) 2027, the ventral surface

is slightly convex transversely (Fig. 55E). The base of

the neural arch is placed anteriorly (Fig. 55D).

LPB (FGGUB) 2028 (Fig. 55F–J) is a caudal vertebra

that is somewhat smaller and located more posteriorly

in the vertebral series than LPB (FGGUB) 2027, and

the main difference between them is that the former has

a more subcircular (albeit still somewhat dorsoventrally

compressed) cross-section through the centrum. In LPB

(FGGUB) 2028, the ventral profile of the centrum is

subrectangular (Fig. 55I), contrasting with the more

spool-like shape (i.e. expanded articular ends and con-

stricted central region) that characterizes the posterior

caudal vertebra SZTFH Ob.3105 of Magyarosaurus

(Fig. 25C) and those of Uriash (Fig. 41). The neural

arch is anteriorly placed on the centrum. There are long,

slender, anteriorly directed prezygapophyses (differing

from the more robust ones of Petrustitan,Fig. 35), and

postzygapophyseal facets are still present. The TPRL

extends beyond the anterior margin of the centrum

(clearly observable in ventral view), a morphology that

is not present in Petrustitan.

Chevrons. None of the chevrons are complete (Fig.

55K–V). They have the typical ‘Y’-shaped morphology

(i.e. not bridged proximally), with the exception of LPB

(FGGUB) R.1881, in which there is an incomplete dor-

sal bridge (Fig. 55K–M); however, the clearly asymmet-

rical development of this bony bridge, originating from

the proximal articular face of one ramus, but unevenly

developed and not reaching the articular surface of the

opposite ramus, suggests that it might represent a patho-

logical bony outgrowth instead of a genuine anatomical

feature bridging the haemal canal. It is difficult to assess

the morphology of the proximal articular facets, because

of poor preservation, but they appear to be divided into

anterior and posterior articular surfaces, better observed

in in LPB (FGGUB) R.1881 and R.2025. This is a simi-

lar condition to that seen in Paludititan, but differing

from the chevrons referred to Magyarosaurus (Fig. 26).

In the more anterior chevrons of Individual I, the anter-

ior articular facet faces dorsally and is larger than the

posterior facet, which faces posterodorsally. Those chev-

rons that preserve part of the distal blade demonstrate

that it was transversely compressed, whereas the prox-

imal rami have a more circular cross-section. LPB

(FGGUB) R.2026 (Fig. 55N–P) bears a prominent bulge

on the lateral surface of the proximal ramus, as is also

the case in Magyarosaurus.

Ilium. Only the right ilium is preserved in Individual I

(Fig. 54). It is fairly complete, although further prepar-

ation is needed to fully understand its anatomy. The

ilium is not well preserved where it articulates with the

sacral ribs. The dorsal margin of the preacetabular pro-

cess is gently curved and rounded in lateral view, and

its horizontal orientation and lateral deflection could

have been exaggerated by dorsoventral taphonomic com-

pression. The pubic peduncle is incomplete, obfuscating

a detailed comparison with those of Petrustitan and

Paludititan. Nevertheless, enough of the pubic process

is preserved in Individual I to determine that its base is

robust. The highest part of the dorsal margin of the pre-

acetabular process seems to have been located anterior

to the base of the pubic peduncle, as is typical for tita-

nosauriforms (Upchurch, 1998).

Assemblage J (the Groap

a assemblage)

This assemblage includes axial and appendicular

remains, although not all of them are well preserved.

Part of the material collected at this locality (recovered

mainly during the 1980s) is in the collections of the

MCDRD in Deva, Romania, and was inaccessible to us

during this study. However, more recently collected

specimens are deposited in the collections of the LPB

(FGGUB), and comprise one cervical vertebra (LPB

[FGGUB] R.1073), one dorsal vertebra (LPB

[FGGUB]R.1854), and over 50 caudal vertebrae that

are mostly isolated (e.g. LPB [FGGUB]R.1148,

R.1184, R.1187, R.1191, R.1193, R.1202–R.1217,

R.1232–R.1238, R.1240–R.1245, R.1512, R.1515–

R.1522, R.1540, R.1569, R.1813], but can also

occur occasionally articulated, with four (LPB

[FGGUB] R.1254], or five (LPB [FGGUB] R.1568) ele-

ments, a dorsal rib fragment (LPB [FGGUB] R.1795),

one chevron (LPB [FGGUB] R.1597), three humeri

(LPB [FGGUB]R.1246, R.1257, R.1528), two ulnae

(LPB [FGGUB] R. 1514, R.1598), one radius (LPB

[FGGUB] R.1248), four femora (LPB [FGGUB]R.1220,

R.1511. R.1513, R.1578), four tibiae (LPB [FGGUB]

R.1182, R.1219, R.1252, R.1301), two fibulae (LPB

[FGGUB] R.1221, R.1263), two metapodial elements

(LPB [FGGUB] R.1185, R.1564) and one phalanx (LPB

[FGGUB] R.1222). From this large sample, only the

best-preserved and most informative specimens are

described here.

Cervical vertebra. LPB (FGGUB) R.1073 (Fig. 56)is

an opisthocoelous posterior cervical vertebra, better pre-

served on its right side. Erosion reveals that the internal

tissue structure is camellate. The centrum is slightly

Revision of Romanian sauropod dinosaurs 79

transversely compressed. It has an aEI of 2.08, which is

slightly higher than that of the known posterior cervical

vertebra of Magyarosaurus. There is a weakly devel-

oped midline ridge along the anterior three-quarters of

the ventral surface of the centrum. On either side of this

ridge, the ventral surface is shallowly concave trans-

versely, flattening posteriorly. The parapophysis is ante-

roventrally located on the lateral surface of the centrum

and merges with a posteriorly directed PCPL-like lamina

that extends along the ventrolateral edge of the centrum.

The lateral surface of the centrum bears a deep oval lat-

eral pneumatic opening.

The neural arch occupies most of the centrum

length. As a result of poor preservation, the transition

between the diapophysis and the prezygapophysis is

not clearly discernible. The prezygapophysis is short

and slightly anterodorsally directed, such that it does

not project beyond the anterior edge of the condyle.

The prezygapophyseal articular surface faces dorsome-

dially. The diapophysis is only slightly laterally devel-

oped, although this could be due to poor preservation.

In overall morphology and orientation, the diapophyses

are different in LPB [FGGUB] R.1073 relative to the

cervical vertebra of Magyarosaurus (LPB [FGGUB]

R.2505, Fig. 20); they are shorter, dorsoventrally

thicker, laterally oriented, with a rounded profile in the

former, whereas they are ventrally directed in the lat-

ter. In LPB (FGGUB) R.1073, a prominent PCDL

extends posteroventrally from the diapophysis, and a

short ACDL is also present (Fig. 56F). Posterior to the

PODL and below the postzygapophysis, a deep

POCDF is present. The postzygapophysis is relatively

shorter than that of the cervical vertebra of

Magyarosaurus, and its articular surface (although

eroded) is dorsolaterally oriented. Matrix means that it

is not possible to observe whether a TPRL and TPOL

are present (Fig. 56G).

The summit of the neural spine is not complete,

although it appears to have been low, with a triangular

outline in lateral view. It was probably lower than the

incomplete neural spine of the cervical vertebra of

Magyarosaurus. The dorsoventral extent of the PRSL

cannot be assessed because of the presence of matrix

(although the most distal tip is visible), as is also the

Figure 56. Lithostrotia incertae sedis, Assemblage J, posterior cervical vertebra LPB (FGGUB) R.1073 in A, left lateral; B, anterior;

C, right lateral; and D, posterior views. Interpretive drawings in E, anterior; F, right lateral; and G, posterior views. Abbreviations:

ACDL, anterior centrodiapophyseal lamina; CPOL, centropostygapophyseal lamina; CPRL, centroprezygapophyseal lamina; D,

diapophysis; PCDL, posterior centrodiapophyseal lamina; PO, postzygapophysis; POCDF, posterior centrodiapophyseal fossa;

PODL, postzygodiapophyseal lamina; PRE, prezygapophysis; PRSL, prespinal lamina; SDF, spinodiapophyseal fossa; SPOL,

spinopostzygapophyseal lamina; SPRL, spinoprezygapophyseal lamina. Scale bar equals 100mm.

80 V. D



ıez D

ıaz et al.

case for the POSL (Fig. 56E, G). Both the SPRL and

SPOL are single (i.e. unbifurcated), robust, and verti-

cally oriented in lateral view. The latter lamina shows a

slight ventrolateral orientation in posterior view.

Caudal vertebrae. Anterior caudal

vertebrae. LPB (FGGUB) R.1813 (Fig. 57A–C) is the

most completely preserved anterior caudal vertebra in

this assemblage, but the neural arch and part of the

eroded condyle are obscured by matrix. The centrum is

procoelous and slightly dorsoventrally compressed. Its

anterior surface has a subcircular outline, with the ven-

tral margin shorter than the dorsal one because of trans-

verse compression ventrally. The shape of the posterior

surface is difficult to assess due to damage and concre-

tionary matrix that is adhered to this region, but it

appears to be more quadrangular. The posterior condyle

is dorsally displaced and, although incomplete at its tip,

it is conical and relatively prominent. The anterior half

of the centrum appears to have been more transversely

expanded than the rest of the centrum. The lateral sides

of the centrum are very mildly convex dorsoventrally

and concave anteroposteriorly. The ventral surface is

both longitudinally dorsally arched and transversely con-

cave, with a wide and relatively shallow median groove

bordered by similarly wide and rounded margins (Fig.

57C). This median groove attenuates towards the anter-

ior and posterior ends. The articular facets for the chev-

rons are not preserved because of damage to both

anterior and posterior margins of the ventral surface.

The transverse processes have large rounded bases that

are situated above the junction between the anterior

neural arch and the centrum. There is a well-developed

PRDL that extends laterally, suggesting that this speci-

men probably comes from a relatively anterior position

in the tail. Anterior to the PRDL, there is a PRCDF.

The preserved bases of the prezygapophyses indicate

that they were directed anterodorsally, as in the lecto-

type of Magyarosaurus (SZTFH Ob.3091). Only the

base of the neural spine is preserved, and it is covered

Figure 57. Lithostrotia incertae sedis, Assemblage J, anterior caudal vertebrae. LPB (FGGUB) R.1813 in A, right lateral; B,

anterior; and C, ventral views. LPB (FGGUB) unnumbered in D, right lateral, E, left lateral; F, anterior; G, posterior; H, dorsal; and

I, ventral views. Neural arch LPB (FGGUB) R.1184 in J, left lateral; K, anterior; and L, dorsal views. Abbreviations: PRCDF,

prezygocentrodiapophyseal fossa; PRDL, prezygodiapophyseal lamina. Scale bars equal 50mm.

Revision of Romanian sauropod dinosaurs 81

in hard matrix, hindering observation of its morphology

as well as that of the postzygapophyseal region.

A second, somewhat smaller, procoelous anterior cau-

dal centrum (unnumbered) (Fig. 57D–I) is better pre-

served, but less complete, with breaks marking the

bases of the neural pedicels. It is similar in most

respects to the centrum of R.1813, including the dorsally

wider subcircular anterior surface, and the presence of a

dorsally displaced, conical (albeit less protruding) con-

dyle on the posterior surface. On the ventral surface, the

median groove is relatively deeper, and the bounding

ventrolateral ridges are narrower and more pronounced

(Fig. 57I).

LPB (FGGUB) R.1184 preserves the base of a caudal

neural arch (Fig. 57J–L). The prezygapophyses are short

and anterodorsally directed. A rounded and shallow

fossa is present posteriorly at the base of the neural

spine. A SPRL originates from the angular dorsal mar-

gin of each prezygapophysis, together delimiting a shal-

low longitudinal depression along the anterior base of

the neural spine. As a result, the basal horizontal cross-

section through the spine is approximately ‘Y’-shaped,

with a very short anterior ‘fork’and a distally expanded

base. These features suggest that this specimen comes

from an anterior caudal vertebra, although probably

from a more distal position than R.1813.

Middle caudal vertebrae. Middle caudal vertebrae

are represented by a number of well-preserved speci-

mens in the Groap

a sample, including LPB (FGGUB)

R.1202, R.1203, R.1204, R.1210 and R.1512 (Fig. 58).

Although none of these specimens have their articular

surfaces completely exposed from under the hard

matrix, it is nevertheless clear that their centra were

largely platycelous, with a relatively shallowly exca-

vated anterior surface and a posterior surface that

appears to vary from very mildly convex, and appar-

ently centrally excavated by a depression (R.1202,

R.1204), to flat and centrally hollowed (R.1512). The

centra are anteroposteriorly longer than dorsoventrally

high and markedly transversely compressed, especially

at midlength where the height of the centrum can be

more than 1.5 times its width. Both the anterior and

posterior articular surfaces have tall, evenly rounded

subquadrangular contours. The ventral surface is narrow

and flat centrally, strongly concave in lateral view, and

transversely concave between the ventrolateral ridges

that connect the weak (when preserved and visible)

articulations for the chevrons. The lateral sides of the

centra are dorsoventrally flat, but concave anteroposter-

iorly. A sharp anteroposteriorly oriented ridge, located

at the base of the neural arch pedicels, represents the

remnant of the transverse process. This ridge extends

along almost the entire length of the neural arch

pedicels, and is more pronounced posteriorly. Only the

basal parts of the prezygapophyses can be seen in

R.1202, R.1204 and R.1210; these appear to be rod-like,

and are anteriorly and only very slightly dorsally ori-

ented. Their bases appear to be bridged medially above

the neural canal. Based on partial preservation in differ-

ent specimens, the neural spine was a transversely com-

pressed, dorsoventrally low and posteriorly elongated

blade, overhanging the posterior half of the centrum.

More distally placed in the middle caudal series,

specimens such as LPB (FGGUB) R.1190, R.1191 and

R.1569 (Fig. 59) have very mildly procoelous to platy-

coelous/amphicoelous, spool-like centra. These centra

are not transversely compressed to any significant

amount, with their width and height being approxi-

mately equal at mid-length. These specimens are rela-

tively more elongated than those from more proximal

positions in the middle tail. The articular surfaces are

rounded subquadrangular, often with a central concavity.

The ventral surface is slightly narrowed at mid-length,

but otherwise generally wider than in the more proximal

middle caudal vertebrae. This ventral surface is only

mildly concave both anteroposteriorly and transversely,

with a very weak midline furrow bounded by two ridges

connecting the poorly expressed articular facets for the

chevrons. The ventral furrow is better developed near

the articular ends, but almost flattens out and disappears

at mid-length. The lateral surfaces are weakly concave

longitudinally. Rugose bumps appear below the junction

between the anteriorly placed neural arch and the cen-

trum, marking the position of the even more weakly

developed, vestigial transverse processes. LPB

(FGGUB) R.1191 preserves parts of a low and long

neural spine with a sharp dorsal margin.

Posterior and distal caudal vertebrae. Several iso-

lated caudal vertebrae (LPB [FGGUB] R.1212–R.1215,

1516) (Fig. 60) will be described for Assemblage J.

LPB (FGGUB) R.1516 is a posterior caudal vertebra.

The spool-like centrum is incomplete (damaged at both

ends), so aspects of its morphology and proportions can-

not be ascertained. It has a transversely flat ventral sur-

face that lacks longitudinal ridges, although it shows

very short and shallow depressions near the articular

surfaces. The long neural arch is located anteriorly on

the centrum. A low, weakly rugose longitudinal ridge

marks the arch–centrum junction. The neural spine is

low and long, with a narrow and somewhat sharp hori-

zontal dorsal edge. The other specimens have better-pre-

served centra with a morphology reminiscent of that in

R.1516, but preserve only parts of the neural arch or

only the neural arch pedicels. Their centra are spool-

shaped and relatively more elongated than those from

the middle section of the tail. They document a variety

82 V. D



ıez D

ıaz et al.

of articular surface morphologies, including platycoe-

lous-amphiplatyan (R.1212, R.1215), mildly biconvex

(R.1214), and even weakly opisthocoelous as seen in

specimen R.1213, where in the posterior surface is

mildly excavated centrally and the anterior one has a

centrally placed low and conical condyle. The propor-

tions of these centra also vary, from ones with a quad-

rangular cross-section at mid-length, and articular faces

that are highly rectangular (R.1213), to ones that are

moderately dorsoventrally compressed at both the mid-

section and at their articular surfaces that can be as

much as 1.5 times wider than tall (e.g. R.1215). The

ventral surfaces are slightly arched longitudinally, but

mainly flat transversely, with weak central depressions

bordered by short triangular ridges near the articular

ends. There are no distinct chevron facets. The lateral

surfaces of the centra are convex dorsoventrally, some-

times with a weak longitudinal ridge at the base of the

neural arch pedicels. The neural arches are anteriorly

situated and relatively long, amounting to about half of

the centrum length. One further specimen (R.1216, Fig.

60M–R), although significantly smaller in size, has the

same general centrum morphology, being mildly opis-

thocoelous and dorsoventrally compressed. Its anteriorly

positioned neural arch is better preserved, with relatively

long and slightly anterodorsally projecting

Figure 58. Lithostrotia incertae sedis, Assemblage J, middle caudal vertebrae. LPB (FGGUB) R.1203 in A, right lateral; B, anterior;

C, left lateral; D, posterior; E, dorsal; and F, ventral views. LPB (FGGUB) R.1204 in G, right lateral; H, anterior; I, left lateral; J,

dorsal; and K, ventral views. LPB (FGGUB) R.1512 in L, right lateral; M, anterior; N, left lateral; O, posterior; P, dorsal; and Q,

ventral views. Abbreviation: PO, postzygapophysis. Scale bar equals 50 mm.

Figure 59. Lithostrotia incertae sedis, Assemblage J, middle caudal vertebrae. LPB (FGGUB) R.1191 in A, right lateral; B, anterior;

C, left lateral; D, posterior; E, dorsal; and F, ventral views. LPB (FGGUB) R.1569 in G, right lateral; H, anterior; I, left lateral; J,

posterior; K, dorsal; and L, ventral views. LPB (FGGUB) R.1233 in M, right lateral; N, anterior; O, left lateral; P, posterior; Q,

dorsal; and R, ventral views. Scale bar equals 50mm.

Revision of Romanian sauropod dinosaurs 83

prezygapophyses that are bridged medially along their

preserved length above the neural canal.

Another set of isolated caudal vertebrae (LPB

[FGGUB] R.1236-R.1238), alongside a string of five

vertebrae preserved in articulation (LPB [FGGUB]

R.1568), are identified as distal caudal vertebrae. These

have elongated, dorsoventrally compressed cylindrical

centra that flare out in all directions towards their articu-

lar ends. The centra are elongated, although relatively

less so than those from the preceding, posterior series

(e.g. LPB [FGGUB] R.1214 and R.1215; see Fig. S1).

Their morphology varies from amphiplatyan (R.1237) to

mildly procoelous (R.1238) to biconvex (R.1568). The

ventral and dorsal surfaces are flat transversely,

although the former is somewhat dorsally arched longi-

tudinally. The lateral sides are dorsoventrally convex.

Some of the specimens (e.g. R.1237; Fig. 61A–E) are

slightly asymmetrical, laterally deflected along their

length, so that one of the sides is convex and the oppos-

ite one is concave in ventral/dorsal view. Neural arches

are no longer present in these vertebrae, although one

(but not both) of the pedicels may sometimes be present

in the form of an anteriorly and marginally placed short

and rounded dorsal ridge (e.g. R.1238). In the first two

vertebrae of the articulated series R.1568 (Fig. 61F–H),

this vestigial pedicel is present only on the right side,

appears only on the left side in the third, and may be

missing entirely in the last two caudal vertebrae

(although this is difficult to ascertain because the centra

are still partly embedded in the matrix).

A few minute rod-like elements represent distal-most

caudal vertebrae. These are approximately similar to the

distal ones described previously in being dorsoventrally

compressed cylinders, but are significantly smaller (11–

50 mm in length, and less than 5 mm in width). No other

features, such as traces of the pedicels, can be seen on

these very anatomically ‘simple’elements.

Ulna. The right ulna LPB (FGGUB) R.1598 is complete

and fairly well preserved (Fig. 62). The estimated max-

imum proximal mediolateral width to ulna length ratio

is 0.39, indicating a relatively robust element, as in

Petrustitan. The long axes of the anteromedial and

anterolateral processes form an angle of approximately

80(Fig. 62D). The anteromedial process of the

Figure 60. Lithostrotia incertae sedis, Assemblage J, posterior caudal vertebrae. LPB (FGGUB) R.1214 in A, right lateral; B,

anterior; C, left lateral; D, posterior; E, dorsal; and F, ventral views. LPB (FGGUB) R.1215 in G, right lateral; H, anterior; I, left

lateral; J, posterior; K, dorsal; and L, ventral views. LPB (FGGUB) R.1216 in M, right lateral; N, anterior; O, left lateral; P,

posterior; Q, dorsal; and R, ventral views. Abbreviations: PRE, prezygapophysis. Scale bar equals 50 mm.

Figure 61. Lithostrotia incertae sedis, Assemblage J, distal caudal vertebrae. LPB (FGGUB) R.1237 in A, anterior; B, right lateral;

C, posterior; D, dorsal; and E, ventral views. Five distal caudal vertebrae in articulation LPB (FGGUB) R.1568 in F, dorsal view;

and detail of the first two vertebrae in G, dorsolateral; and H, dorsal views. The arrow indicates the vestigial pedicel. Scale bar

equals 50 mm.

84 V. D



ıez D

ıaz et al.

proximal end is much longer than that of the anterolat-

eral process, similar to Petrustitan (Fig. 36D). The pos-

terior process of the proximal end is only moderately

developed, and its long axis is offset from that of the

anterolateral process, such that it projects posterolater-

ally. Only a weakly developed olecranon process is pre-

sent. The proximal surface of the anteromedial process

is flat, lacking any concave profile. The posterior sur-

face of the proximal end is concave mediolaterally. The

proximal processes extend distally as weakly developed

ridges. A proximodistally oriented ridge is present along

the distal third of the anteromedial surface, for articula-

tion with the radius (Fig. 62A). This ridge extends

obliquely relative to the long-axis of the shaft, such that

it trends laterodistally, similar to the condition in

Magyarosaurus. Posterior to this ridge, the distal-most

portion of the anteromedial surface is gently concave.

The ulna expands posteriorly at its distal end. In distal

view, the ulna has a subtriangular or ‘D’-shaped outline,

with a flat to mildly concave medial margin (Fig. 62C).

The ulna shares similar anatomical features of the distal

end with that of Petrustitan, with a convex articular sur-

face with rounded edges, and little posterior expansion.

Femur. None of the femora are complete. Specimen

LPB (FGGUB) R.1578 is the proximal end of a right

femur. Apart from the dorsomedial orientation of the

femoral head, no other informative anatomical features

can be discerned. LPB (FGGUB) R.1511 (Fig. 63A) and

R.1513 are left femoral shafts, lacking the proximal and

distal ends. The best-preserved specimen (LPB

[FGGUB] R.1220, Fig. 63B-E) is also the smallest one

from the Groap

a sample, lacking the distal condylar

region and the extreme proximal end. Both LPB

(FGGUB) R.1220 and R.1511 were histologically

Figure 62. Lithostrotia incertae sedis, Assemblage J, right ulna LPB (FGGUB) R.1598 in A, anterior; B, lateral; C, distal (lateral

towards top); D, proximal (anterior towards top); E, posterior; and F, medial views. Scale bar equals 50 mm.

Revision of Romanian sauropod dinosaurs 85

sampled by Stein et al. (2010), who assigned them to

HOS 13. The anterior face is more convex transversely

than the flatter posterior one, especially in LPB

(FGGUB) R.1513, because of the wide and well devel-

oped linea intermuscularis cranialis. It is more difficult

to assess the presence of this ridge in LPB (FGGUB)

R.1511, although this structure is potentially represented

by a rugose surface in the middle of the anterior face of

the shaft. In LPB (FGGUB) R.1220, a moderate degree

of lateral bulging of the lateral edge is present near the

proximal end. On the posterior surface, a wide trochan-

teric shelf extends ventrally from the preserved part of

the proximal end, parallel to the lateral edge of the shaft

(Fig. 63D). This shelf ends approximately level with the

distal tip of the fourth trochanter, as in the femur of

Magyarosaurus. The shelf and fourth trochanter are sep-

arated by a shallow concave surface. The fourth tro-

chanter is a poorly developed ridge close to the medial

edge of the shaft (Fig. 63D), differing from the more

developed ones of Magyarosaurus (Fig. 32) and, espe-

cially, Uriash (Fig. 43). In the Assemblage J femora,

there is a rugose oval area located on the medial sur-

face, close to the fourth trochanter. Specimens LPB

(FGGUB) R.1220 and R.1511 have ECC values of 1.47

and 1.45, respectively –higher than in the femora of

Magyarosaurus –whereas that of LPB (FGGUB)

R.1513 is 1.24, similar to Magyarosaurus. All of these

femora show much lower ECC values when compared

to the femur of Uriash.

Phylogenetic analysis

Phylogenetic dataset

We utilized the phylogenetic data matrix presented by

Poropat et al. (2023), which is the most recent iteration

of the matrix originally presented by Mannion et al.

(2013), and which has undergone numerous updates and

expansions (e.g. Mannion, Upchurch, Schwarz, et al.,

2019; Poropat et al., 2016). Character scores for the

existing OTU of Ligabuesaurus leanzai were revised

based on additional remains described by Bellardini

et al. (2022), as were those of Patagotitan mayorum

based on new information presented by Otero et al.

(2020) and personal observations (PDM 2018: MPEF-

PV specimens). We also made a small number of char-

acter score changes to several other existing OTUs.

These changes are all documented in the Supplemental

Material.

We added Magyarosaurus dacus,Paludititan nalat-

zensis and Petrustitan hungaricus as new OTUs based

on the information presented herein and personal obser-

vations. We did not include Uriash kadici given its

incompleteness and the limited number of characters for

which it could be scored (1%). As noted by Gorscak

et al. (2023), one potential driver of inconsistency

between phylogenetic analyses using different datasets

might be variation in spatial sampling of taxa. Those

authors commented upon the underrepresentation of

European and African titanosaurs in most analyses

Figure 63. Lithostrotia incertae sedis, Assemblage J, femora. A, left femur LPB (FGGUB) R.1511 in anterior view. Right femur

LPB (FGGUB) R.1220 in B, anterior; C, medial; D, posterior; and E, lateral views. Scale bar equals 100mm.

86 V. D



ıez D

ıaz et al.

(including previous iterations of the data matrix used in

the present study), as well as poor sampling of

Australian taxa in datasets other than the one used

herein. Consequently, we also included a further 23

Cretaceous titanosaurian OTUs. These comprise taxa

from Africa, Europe and South America, scored based

on a combination of personal observations and the pub-

lished literature, as documented in Table 1. Although

we do not disagree a priori with the conclusions of

Silva Junior, Martinelli, Marinho, et al. (2022) that the

latest Cretaceous Brazilian species Trigonosaurus pricei

might be a junior synonym of the sympatric species

Baurutitan britoi, the presented evidence is limited in

terms of identified anatomically overlapping autapomor-

phies. As such, we retain these two taxa as separate

OTUs. Following Silva Junior, Martinelli, Marinho, et

al. (2022), we exclude the caudal vertebrae previously

referred to the paratype of Trigonosaurus pricei (MCT

1719-R), but that are now assigned to Caieiria allocau-

data, and analyse this taxon as an additional OTU. In

total, the updated data matrix consists of 152 OTUs.

Our resultant data matrix therefore goes some way to

alleviating the problem highlighted by Gorscak et al.

(2023), such that our data matrix includes the majority

of named titanosaur taxa from Africa, Asia, Australia,

Europe, Indo-Madagascar and North America, as well

as a large sampling of South American taxa (30 taxa).

In total, we sample 50 unequivocal titanosaurs, with a

further 40 OTUs representing additional somphospondy-

lans; furthermore, a large proportion of this subset of

taxa (>60%) is scored first-hand based on personal

observations.

Our character sampling universe is also extensive,

building upon previous iterations of this data matrix.

Taxa are scored for 570 phylogenetically informative

characters (see the Supplemental Material for the com-

plete character list), with a large proportion of these not

included in other data matrices. Of these, we modified

character 122, which pertains to the morphology of the

lateral excavation of the cervical centra, separating

depth (a revised character 122, with new scores docu-

mented in the Supplemental Material) and division of

this feature (newly added character 557). Also, charac-

ters 558–570 are newly added to the dataset, based on

the literature and personal observations:

C122. Postaxial cervical centra, lateral surfaces: lack an

excavation or have a shallow fossa (0); possess a deep

foramen with clearly defined margins (1) (McIntosh,

1990; Russell & Zheng, 1993; Upchurch, 1995,1998;

revised here).

C557. Postaxial cervical centra, lateral surfaces: undiv-

ided (0); fossa or foramen divided into separate portions

by one or more laminae (1) (McIntosh, 1990; Russell &

Zheng, 1993; Upchurch, 1995,1998; revised here).

C558. Middle–posterior cervical centra, paired pneu-

matic fossae on the anterior third of the ventral surface,

level with the parapophyses: absent (0); present (1)

(new character: based on (Coria et al., 2013; Gorscak

et al., 2014).

C559. Posterior cervical centra, posterior margin of par-

apophysis–capitulum suture forms a tubercle: absent (0);

present (1) (new character: based on Gorscak et al.,

2014; note that this is most clearly seen in ventral

view).

C560. Postaxial cervical neural arches, prezygapophy-

seal articular surfaces situated dorsal to the anterior por-

tion of the diapophyses: absent (0); present (1) (Salgado

et al., 1997; modified here).

C561. Middle–posterior cervical neural spines, ‘bulbous’

lateral expansions at apex: absent (0); present (1)

(Campos et al., 2005; Gorscak & O’Connor, 2016;

modified here; note that these expansions are not formed

by laminae).

C562. Middle–posterior caudal vertebrae, internal tissue

structure: solid (0); camellate (1) (new character: based

on Cerda et al., 2012).

C563. Scapulocoracoid, internal tissue structure: solid

(0); camellate (1) (new character: based on Cerda et al.,

2012).

C564. Scapular blade, ventrally located ridge or tuberos-

ity on medial surface of proximal half: absent (0); pre-

sent (1) (Sanz et al., 1999).

C565. Humerus, medial margin of anterior surface

forms a ridge (‘anteromedial arm’) along the proximal

half: absent (0); present (1) (new character: based on

Gorscak et al., 2017; Mannion & Otero, 2012).

C566. Femur, fibular condyle, posterior surface: single

condyle, with the lateral portion of the posterior surface

merging smoothly into the lateral surface of the distal

end (0); divided, forming two distinct condylar proc-

esses (i.e. a well-developed epicondyle) (1) (Beeston

et al., 2024; Carballido et al., 2017; Sekiya, 2011; modi-

fied here).

C567. Tibia, anteromedial margin of distal quarter forms

a sharp, prominent ridge, delimiting concave anterior

and medial surfaces: absent (0); present (1) (new charac-

ter: based on (D

ıez D

ıaz et al., 2013a,2018).

C568. Tibia and fibula, distal articulation, tibia with a

laterally expanded distal end (extending further laterally

than the remainder of the tibia, creating a concave

Revision of Romanian sauropod dinosaurs 87

margin along the distal half of the tibia in anterior

view) that articulates with a concave medial surface on

the distal end of the fibula: absent (0); present (1) (new

character).

C569. Fibula, posterodistally oriented accessory ridge

on lateral surface of fibula, extending from mid-length

to approximately one-quarter of fibula length from distal

end: absent (0); present (1) (new character).

C570. Osteoderms, keeled morphology (subcircular to

oval in external or internal view and bear one or more

longitudinal keels on their internal surfaces): absent (0);

present (1) (Curry Rogers, 2005;D’Emic et al., 2009).

Phylogenetic analysis

Phylogenetic analyses under a parsimony framework

were run in TNT v1.6 (Goloboff et al., 2008; Goloboff

& Morales, 2023). Following previous iterations of this

data matrix (e.g. Mannion et al., 2013; Mannion,

Upchurch, Schwarz, et al., 2019; Poropat et al., 2023),

17 of the 574 characters were treated as ordered (char-

acters 11, 14, 15, 27, 40, 51, 104, 147, 148, 195, 205,

259, 297, 426, 435, 472 and 510; note that revision to

character 122 means that this is no longer an ordered

character) and eight OTUs identified as highly unstable

were excluded a priori (Astrophocaudia,Australodocus,

Brontomerus,Fukuititan,Fusuisaurus,Liubangosaurus,

Malarguesaurus and Mongolosaurus). Under a ‘New

Table 1. List of taxa added to the data matrix (in addition to the three Romanian taxa), along with their spatiotemporal distribution

and the basis for their scoring.

Taxon Spatiotemporal distribution Reference/source

Africa

Mansourasaurus shahinae Middle Campanian, Egypt Sallam et al. (2018)

Mnyamawamtuka moyowamkia Aptian to Cenomanian, Tanzania Gorscak and O’Connor (2019)

Paralititan stromeri Cenomanian, Egypt Smith et al. (2001)

Rukwatitan bisepultus Cenomanian to Campanian, Tanzania Gorscak et al. (2014)

Shingopana songwensis Cenomanian to Campanian, Tanzania Gorscak et al. (2017)

Europe

Abditosaurus kuehnei Early Maastrichtian, Spain Vila et al. (2022); B. Vila and A. Sell

es, pers.

comm.

Atsinganosaurus velauciensis Late Campanian, France Garcia et al. (2010); D

ıez D

ıaz et al. (2018);

VDD pers. obs.

Lirainosaurus astibiae Late Campanian, Spain Sanz et al. (1999); D

ıez D

ıaz et al. (2011); D

ıez

D

ıaz et al., 2013a,2013b); VDD, PDM & PU

pers. obs. (2009)

Lohuecotitan pandafilandi Late Campanian to early

Maastrichtian, Spain

D

ıez D

ıaz et al. (2016); VDD pers. obs.

South America

Argyrosaurus superbus Coniacianh to Campanian, Argentina Mannion and Otero (2012)

Arrudatitan maximus Late Campanian to early

Maastrichtian, Brazil

Santucci and Arruda-Campos (2011); Silva

Junior, Martinelli, Marinho, et al. (2022)

Baurutitan britoi Maastrichtian, Brazil Kellner et al. (2005); PDM pers. obs. (2019)

Bonatitan reigi Late Campanian to early

Maastrichtian, Argentina

Martinelli and Forasiepi (2004); Salgado et al.

(2015); PDM pers. obs. (2013, 2018)

Caieiria allocaudata Maastrichtian, Brazil Campos et al. (2005); Silva Junior, Martinelli,

Iori, et al. (2022); PDM pers. obs. (2019)

Dreadnoughtus schrani Campanian to Maastrichtian,

Argentina

Lacovara et al. (2014); Ullmann and Lacovara

(2016); Voegele et al. (2017,2020)

Elaltitan lilloi Coniacian, Argentina Mannion and Otero (2012); PDM pers. obs.

(2013)

Gondwanatitan faustoi Late Campanian to early

Maastrichtian, Brazil

Kellner and Azevedo (1999); PDM pers. obs.

(2009)

Narambuenatitan palomoi Early to middle Campanian, Argentina Filippi et al. (2011); Paulina Carabajal et al.

(2020); PDM pers. obs. (2014)

Neuquensaurus australis Early to middle Campanian, Argentina Salgado et al. (2005); Otero (2010); D’Emic

and Wilson (2011); PDM pers. obs. (2013,

2018)

Overosaurus paradasorum Santonian, Argentina Coria et al. (2013); PDM pers. obs. (2014)

Pellegrinisaurus powelli Early to middle Campanian, Argentina Cerda et al. (2021); VDD pers. obs. (2011)

Trigonosaurus pricei Maastrichtian, Brazil Campos et al. (2005); PDM pers. obs. (2019)

Uberabatitan ribeiroi Maastrichtian, Brazil Salgado and Carvalho (2008); Silva Junior et al.

(2019)

88 V. D



ıez D

ıaz et al.

Technology Search’, we used the ‘Stabilize Consensus’

option with sectorial searches, drift and tree fusing.

After five rounds of consensus stabilizing, the resultant

trees were used as the starting topologies for a

‘Traditional Search’, utilizing tree bisection–reconnec-

tion. Two versions of the analysis were run: one in

which equal character weighting (EQW) was applied,

and a second in which extended implied weighting

(EIW) was implemented (Goloboff, 2014; Goloboff

et al., 2018), using a k-value of 9 (see Mannion,

Upchurch, Jin, et al., 2019; Tschopp & Upchurch,

2018). As with recent iterations of this data matrix

(Poropat et al., 2021,2023), two further OTUs (the

‘Cloverly titanosauriform’and Ruyangosaurus) were

excluded a priori from the analysis applying equal char-

acter weighting, although they were retained in the

extended implied weighting analysis.

Phylogenetic results

Our EQW analysis results in more MPTs than TNT can

hold, with a large polytomy for much of Titanosauria.

As such, we used the Iterative PCR approach of Pol and

Escapa (2009) to identify additional unstable taxa, with

this implemented via the IterPCR option in TNT

(Goloboff & Szumik, 2015). This was supplemented by

several preliminary analyses, as well as use of the

Pruned Trees option in TNT. These combined

approaches highlighted Puertasaurus and five of the

newly incorporated taxa (Arrudatitan,Caieiria,

Paralititan,Shingopana and Trigonosaurus) as unstable.

Exclusion of these six OTUs, along with the 10 OTUs

from the initial analysis, a priori, yields 304,128 MPTs

of length 2996 steps (CI ¼0.200, RI ¼0.613), and

results in a fairly well-resolved strict consensus tree

(Fig. 64). Relationships outside of Titanosauria are

largely congruent with those produced through previous

iterations of this data matrix, but there are numerous dif-

ferences in terms of titanosaurian interrelationships,

including: (1) Diamantinasauria is recovered as the sister

taxon of Titanosauria; (2) Malawisaurus is the sister

taxon of a clade composed of Andesaurus and several

East Asian sauropods (Baotianmansaurus,Daxiatitan,

Dongyangosaurus,Huabeisaurus,Xianshanosaurus),

with Rapetosaurus þAeolosaurini (¼Aeolosaurus þ

Gondwanatitan) as the successive sister taxon to this

grouping (the preliminary analysis also suggests that

Arrudatitan and Caieiria are additional members of the

Aeolosaurini þRapetosaurus clade); Mnyamawamtuka

is the earliest diverging member of this clade, which is

the sister taxon to a clade comprising all other titano-

saurs (¼Eutitanosauria sensu Carballido et al., 2022);

(3) the usual members of Lognkosauria are divided into

two clades, with Abditosaurus and Atsinganosaurus

clustering with Argentinosaurus and Patagotitan,

whereas Lohuecotitan and Paludititan form a more

phylogenetically nested clade with Futalognkosaurus,

Mendozasaurus and Notocolossus, and neither clade

clusters with Rinconsauria (which includes Baurutitan,

Normanniasaurus, and Rukwatitan, as well as the two

clade specifiers) to form Colossosauria; (4)

Nemegtosaurus is recovered as a relatively early diverg-

ing titanosaur and does not cluster with Tapuiasaurus,

with the latter taxon instead forming a clade with

Dreadnoughtus,Epachthosaurus and Uberabatitan; (5)

Petrustitan is most closely related to Narambuenatitan

and Pitekunsaurus, with these taxa forming a clade with

Antarctosaurus,Argyrosaurus,Jainosaurus and Vahiny;

(6) Bonatitan and Elaltitan are sister taxa and are the

earliest diverging members of the clade that includes

saltasaurids, which comprises Neuquensaurus þ

Saltasaurus (¼Saltasaurinae) as the sister taxon to a

polytomy formed of Isisaurus,Opisthocoelicaudia and

Overosaurus (¼Opisthocoelicaudiinae); Alamosaurus,

Pellegrinisaurus,Magyarosaurus, and a clade consisting

of Lirainosaurus þMansourasaurus, form successively

distant sister taxa to Saltasauridae (our preliminary anal-

yses suggest that Shingopana and Trigonosaurus are

also part of this broader clade). Bremer supports are low

across the tree (values of 1–4), with values of 1 or 2 for

most titanosaurian clades, although values of 3 charac-

terize Titanosauria, the early diverging titanosaur clade

that includes Andesaurus and Mnyamawamtuka, as well

as Saltasaurinae.

Under EIW, our analysis produces 90,720 MPTs of

length 154.7 steps (CI ¼0.194, RI ¼0.599), with a

generally well-resolved strict consensus tree (Fig. 65).

As with the EQW analysis, relationships outside of

Titanosauria are largely unchanged from those produced

using the most recent previous iteration of this data

matrix (Poropat et al., 2023). This includes the recovery

of Diamantinasauria as a clade closely related to, but

outside of, Titanosauria. Relationships at the base of

Titanosauria are more consistent with recent iterations

of this data matrix than the results from the EQW ana-

lysis, with Andesaurus clustering with only two other

taxa (Huabeisaurus and Ruyangosaurus), and most tita-

nosaurs recovered within Lithostrotia as a result of

Malawisaurus occupying its ‘typical’position, such that

Lithostrotia excludes Andesaurus. Daxiatitan and

Xianshanosaurus are sister taxa, forming a clade with

Malawisaurus. As with the EQW analysis, Aeolosaurini

occupies a notably different position to its usual place-

ment, recovered as a diverse clade that is outside

Eutitanosauria (sensu Carballido et al., 2022).

Aeolosaurini is divided into two subclades: one is com-

posed of Aeolosaurus and Arrudatitan as sister taxa,

Revision of Romanian sauropod dinosaurs 89

with Rapetosaurus and Nemegtosaurus þTapuiasaurus

recovered as successively more distant taxa; the second

clade comprises Caieiria and Overosaurus as sister taxa,

with Gondwanatitan,Trigonosaurus,Shingopana and

Uberabatitan placed as successively more distant taxa.

Argyrosaurus is placed as the sister taxon to

Eutitanosauria. As with the EQW analysis, the taxa usu-

ally recovered within Lognkosauria form two clades,

and neither of these clades clusters with Rinconsauria.

Argentinosaurus,Patagosaurus and Puertasaurus form

a clade with Petrustitan, whereas Futalognkosaurus,

Mendozasaurus and Notocolossus form a more phylo-

genetically nested clade with Antarctosaurus,

Jainosaurus and Vahiny, as well as Lohuecotitan and

Paludititan. Other early diverging eutitanosaurian clades

consist of Narambuenatitan as the sister taxon of

Mnyamawamtuka þPitekunsaurus, as well as another

distinct lineage in which Dreadnoughtus and

Epachthosaurus form a monophyletic group. As

well as the clade specifiers, Rinconsauria includes

Figure 64. Time-calibrated agreement subtree based on the equal weights analysis, showing the phylogenetic relationships and

palaeobiogeographical distribution of Titanosauria. Romanian genus names emboldened.

90 V. D



ıez D

ıaz et al.

Figure 65. Time-calibrated agreement subtree based on the extended implied weights analysis, showing the phylogenetic

relationships and palaeobiogeographical distribution of Titanosauria. Romanian genus names emboldened.

Revision of Romanian sauropod dinosaurs 91

Rukwatitan, forming the sister taxon to a clade

comprising Normanniasaurus þ(Dongyangosaurus þ

Jiangshanosaurus). Rinconsauria is the sister taxon to a

clade of the remaining titanosaurs in the data matrix.

Bonatitan þElaltitan forms a clade with

Lirainosaurinae, which consists of Abditosaurus as the

sister taxon to Atsinganosaurus þLirainosaurus. This is

in turn the sister taxon to a clade comprising Paralititan

þ(Isisaurus þSaltasauridae). Within Saltasauridae, we

recover Mansourasaurus as the sister taxon to

Neuquensaurus þSaltasaurus, forming Saltasaurinae. A

relatively diverse Opisthocoelicaudiinae is composed of

Opisthocoelicaudia,Alamosaurus,Pellegrinisaurus and

Baurutitan þMagyarosaurus as successively more

phylogenetically nested taxa.

Estimation of body size and mass

Table 2 summarizes the body mass and size estimations

for Magyarosaurus dacus and Uriash kadici obtained by

the equations of Benson et al. (2018), Packard et al.

(2009) and Seebacher (2001), and the humeri and fem-

ora used for those calculations. For Magyarosaurus

dacus we have obtained masses between 660 and

972 kg. The mass calculated by Benson et al. (2018) for

Magyarosaurus (750 kg) falls within the ranges obtained

in this work, being closer to the values calculated for

the specimens referred to Assemblage G. Benson et al.’s

(2018) estimation was based on specimens that we are

not currently able to assign to Magyarosaurus dacus

with confidence, but their value is consistent with ours.

Body length ranges from 2.16 to 2.82 m for this taxon.

Thus, Magyarosaurus dacus is still one of the smallest

titanosaur taxa described so far, even within the Late

Cretaceous of Europe (see D

ıez D

ıaz et al., 2021).

Concerning Uriash kadici, the body mass estimates

range from 5 to 8 tonnes, with lengths between 8.83

and 11.87 m. Benson et al. (2018) calculated a body

mass of 10.5 tonnes for the titanosaur Epachthosaurus,

whose femur has a length of 988 mm. The Uriash holo-

type Individual C could have had a body size similar to

this taxon.

Discussion

Titanosaurian diversity in the Hat

,eg Basin

This work provides an extensive and updated revision

of numerous sites from the continental uppermost

Cretaceous of the Hat

,eg Basin and surrounding regions,

and the titanosaurian individuals and assemblages found

therein, but it also highlights the complexity in object-

ively assessing the actual diversity of this group of sau-

ropods in this area. With more than 20 individuals/

monospecific assemblages identified and described here,

four taxa can be confidently recognized at present:

Magyarosaurus dacus,Paludititan nalatzensis,

Petrustitan hungaricus n. gen., and Uriash kadici n.

gen. n. sp. These four taxa exhibit significant size

Table 2. Body size and mass of the titanosaurian taxa from the Hat

,eg Basin described in this work. Maximum and minimum sizes

were calculated using results from the formulae of both Packard et al. (2009) and Benson et al. (2018). It is important to remember

that calculations derived from assemblages only provide approximate results, as their elements do not belong to a single individual.

Specimens

PerH þF

(mm)

3.352 

PerH þF

2.125

(gr)

(Packard

et al.,

2009)

Weight

(kg)

(Packard

et al.,

2009)

Weight

(kg)

Benson

et al.

(2018)

M/214.44 ¼L

1.46

(Seebacher,

2001)

Magyarosaurus

dacus

Assemblage A Left humerus (MBFSZ

Ob.3089) þleft femur

(MBFSZ Ob.3088)

330? 753,609.1 753.61 659.77 2.37 m 2.16 m

Individual E Left humerus (LPB

[FGGUB] R.1047) þ

left femur (LPB

[FGGUB] R.1046)

372 972,092.86 972.09 917.11 2.82 m 2.71 m

Assemblage G Right humerus (LPB

[FGGUB] R.2506) þ

right femur (LPB

[FGGUB] R.2507)

341.5 810,511.65 810.51 724.92 2.49 m 2.3 m

Uriash kadici Individual C Right humerus (MBFSZ

Ob.3104) þright

femur (MBFSZ

Ob.3103)

816 5,159,951.23 5159.95 7947.6 8.83 m 11.87 m

Abbreviations: L, body length; M, body mass; PerH 1F, sum of the perimeters of the humerus and femur in mm.

92 V. D



ıez D

ıaz et al.

Table 3. List of the specimens described and figured by Huene (1932) and their referral to individuals and assemblages in this

work.

Huene (1932) This work

Taxon Specimen no. Element Individual/assemblage Taxon Notes

M. dacus NHMUK R.3861a Dorsal vertebra Not assigned Titanosauria indet

NHMUK R.3851 Anterior caudal

vertebra

Not assigned Titanosauria indet

SZTFH Ob.3091 Middle caudal

vertebra

Assemblage A Magyarosaurus

dacus

(lectotype)

Figured and

assigned to

Titanosaurus

dacus by

Nopcsa, 1915.

SZTFH v.10339 Posterior caudal

vertebra

Not assigned Titanosauria indet

NHMUK R.3849 Humerus Individual L Titanosauria indet

SZTFH Ob.3099 Ulna Assemblage M Titanosauria indet

SZTFH Ob.3086a Fibula Assemblage A Magyarosaurus

dacus

M. transsylvanicus NHMUK R.3896 2 articulated dorsal

vertebrae

Not assigned Titanosauria indet

SZTFH v.10341 Anterior caudal

vertebra

Not assigned Titanosauria indet

Cd. Pl.46 fig. 5 Anterior caudal

vertebra

Not assigned Titanosauria indet

SZTFH v.13520b Posterior caudal

vertebra

Not assigned Titanosauria indet

SZTFH Ob.4215b Posterior caudal

vertebra

Not assigned Titanosauria indet Figured and

assigned to

Titanosaurus

dacus by

Nopcsa, 1915.

SZTFH v.13469 Anterior caudal

vertebra

Not assigned Titanosauria indet

SZTFH v.10342 Anterior caudal

vertebra

Not assigned Titanosauria indet

SZTFH v.13520a Posterior caudal

vertebra

Not assigned Titanosauria indet

SZTFH v.13492 Humerus Assemblage M Titanosauria indet

SZTFH Ob.3100 Ulna Assemblage A Magyarosaurus

dacus

SZTFH R.4891 Metacarpals I–

II–III

Individual H Titanosauria indet

SZTFH Ob.3086b Fibula Assemblage A Magyarosaurus

dacus

SZTFH Ob.3102 Fibula Assemblage A Magyarosaurus

dacus

M. hungaricus NHMUK R.3833 Fibula Individual B Petrustitan

hungaricus

Tentatively assigned

M. dacus ? SZTHF unknown Cervical neural

arch

?? ?? No specimen

number given,

we were unable

to identify this

specimen in the

Budapest

collection

M. hungaricus ? NHMUK R.4891 2 dorsal ribs Individual H Titanosauria indet

M. hungaricus ? SZTFH Ob.3130b Dorsal rib Not assigned Titanosauria indet

M. dacus or

M. transylvanicus ?

NHMUK R.3898 Sacrum Titanosaur?

M. hungaricus ? SZTFH Ob.3090b Middle caudal

vertebra

Individual C Uriash kadici

(Continued)

Revision of Romanian sauropod dinosaurs 93

disparities, in addition to their anatomical differences,

ranging from a dwarf taxon weighing less than 1 tonne

and achieving an adult body length of no more than 3

metres, to a large-bodied taxon weighing up to approxi-

mately 10 tonnes and a length of nearly 12 metres.

We have demonstrated the validity and diagnostic

nature of Magyarosaurus dacus –one of the first titano-

saurs to be described –with 11 autapomorphies distrib-

uted across the axial and appendicular skeleton. Nopcsa

(1915) did not describe diagnostic features for this spe-

cies, although he gave a brief description of some axial

remains, as well as the scapula and the femur. Huene

(1932), on the other hand, did provide descriptions,

measurements and figures of several individual skeletal

elements from Transylvania that allowed him to differ-

entiate between two approximately similarly sized spe-

cies of his newly erected genus Magyarosaurus,M.

dacus and M. transsylvanicus, with a third, larger taxon

referred to the same genus only tentatively as M. (?)

hungaricus (see the ‘History of titanosaur discoveries

and research in Transylvania’section). In summary,

Huene (1932) regarded M. dacus as a more robustly

built taxon compared to M. transsylvanicus, based on

the caudal vertebrae, humerus, ulna, metacarpals, and

fibula, with the bones of the former species also

assessed to have more prominently marked muscle

scars. By contrast, he differentiated M. (?) hungaricus

from M. dacus and M. transsylvanicus mainly on the

basis of its larger size and the distinct anatomy of its

fibula. Subsequently, the validity of these three

Transylvanian titanosaur taxa has been either accepted

(e.g. Curry Rogers, 2005; Steel, 1970) or tentatively

refuted (e.g. Le Loeuff, 1993; McIntosh, 1990;

Upchurch et al., 2004), although usually without sup-

portive argumentation, with Upchurch et al. (2004) also

noting that Magyarosaurus itself required a major

revision.

In the current work, in line with Huene’s(1932,p.

274) tentative suggestion concerning the potential gen-

eric distinctiveness of M. dacus and M. (?) hungaricus,

we have identified enough differences between these

species to establish the latter as a new genus,

Petrustitan. However, the situation is markedly different

in the case of M. transsylvanicus. As noted above,

McIntosh (1990) and Le Loeuff (1993) considered this

species as invalid, although without clear arguments pre-

sented in support of such a decision. Our re-study of the

Transylvanian titanosaur material presents evidence that

helps to resolve this issue more decisively. According to

Huene (1932), although the two taxa were remarkably

similar to each other both in their dimensions and over-

all morphology, two lines of evidence supported their

Table 3. (Continued).

Huene (1932) This work

Taxon Specimen no. Element Individual/assemblage Taxon Notes

M. hungaricus ? SZTFH Ob.3090d Posterior caudal

vertebra

Individual C Uriash kadici

M. hungaricus ? NHMUK R.3858 Distal caudal

vertebra

Not assigned Titanosauria indet

M. dacus ? SZTFH v.10345 Distal caudal

vertebra

Not assigned Titanosauria indet

M. hungaricus ? NHMUK R.4891 Sternum Individual H Titanosauria indet

M. hungaricus ? NHMUK unnumbered Coracoid Not assigned Titanosauria indet

M. hungaricus ? NHMUK unnumbered Ulna Not assigned Titanosauria indet

M. hungaricus ? NHMUK unnumbered Metacarpal III Not assigned Titanosauria indet

M. hungaricus ? NHMUK R.3852 Right pubis Individual P Titanosauria indet

M. hungaricus ? NHMUK unnumbered Distal femur Not assigned Titanosauria indet No specimen

number given,

hard to identify.

M. dacus ? NHMUK R.3856 Femur Not assigned Titanosauria indet

M. hungaricus ? NHMUK R.3853 Tibia Individual B Petrustitan

hungaricus

(para-lectotype)

M. hungaricus ? NHMUK R.3850 Tibia Not assigned Titanosauria indet

M. hungaricus ? NHMUK unnumbered Tibia Not assigned Titanosauria indet Not figured, not

described, no

specimen

number given -

cannot be

identified.

Note: In bold, those specimens assigned to a specific Individual, Assemblage or taxon.

94 V. D



ıez D

ıaz et al.

separation: the more robust nature of the bones of

M. dacus, displaying better-developed muscle scars,

compared to M. transsylvanicus; and the morphology of

the caudal vertebral centra. Regarding the morphological

differences between the caudal vertebrae, Huene (1932)

alluded to the presence of two different morphotypes

that he considered could not be present in the tail of a

single species: (1) a morphotype with centra possessing

subparallel and slightly concave lateral surfaces, such

that the dorsal and ventral surfaces of the centrum are

of similar width (considered to characterize M. dacus);

and (2) one with slightly convex lateral surfaces that

converge ventrally, such that the ventral surface is nar-

rower than the dorsal one (corresponding to M. transsyl-

vanicus). However, our results indicate that both the

diagnostic nature and conspecificity of the purported

M. transsylvanicus specimens are far from being firmly

established. Indeed, several of the specimens that Huene

(1932) referred to this taxon are identified here as

belonging to different key individuals and assemblages

(some of which we can even refer to other nominal

taxa). Moreover, some of these specimens actually

belong to individuals/assemblages whose other skeletal

elements were referred by Huene (1932)toM. dacus

(e.g. assemblages A and M) or, tentatively, even to

M. (?) hungaricus (Individual H). Such a mixing of

specimens makes M. transsylvanicus chimaeric, and ren-

ders the purported differences in relative robustness of

the appendicular elements (considered to be diagnostic

by Huene, but here recognized to occur within the same

individual/assemblage) taxonomically meaningless.

Other specimens referred to M. transsylvanicus, most of

them isolated caudal vertebrae, have been shown by us

to be currently taxonomically indeterminate, often being

identified no more precisely than ‘Titanosauria indet.’

(see Table 3). It is also worth emphasizing that our sur-

vey of associated titanosaur material demonstrates that

caudal vertebrae belonging to the same taxon (even to

the same individual) can display both dacus-type paral-

lel-sided and transsylvanicus-type wedge-shaped mor-

phologies, depending on their position within the tail

(e.g. Individual E and Assemblage J) –an observation

that further reduces the support for separating M. dacus

and M. transsylvanicus. To conclude, all currently

available information argues against maintaining

M. transsylvanicus as a valid species, so here we follow

McIntosh (1990), Le Loeuff (1993), and Upchurch et al.

(2004) in considering this species as invalid and, at least

in part, a junior synonym of M. dacus.

Regarding the erection of the new genus and species

Uriash kadici, we consider this justifiable and useful

despite the fact that the material from Individual C is

scarce and poorly preserved. The type material belongs

to a single individual and displays eight autapomorphies,

which are widely distributed among the recovered skel-

etal remains and are absent in the corresponding over-

lapping elements of Magyarosaurus,Paludititan, and

Petrustitan. Moreover, Uriash can be definitively distin-

guished from other Late Cretaceous Transylvanian taxa

because of its larger size, which cannot be accounted

for by interpreting Individual C as a more mature mem-

ber of Magyarosaurus,Paludititan or Petrustitan,as

was suggested previously by Le Loeuff (2005a; see

below). Indeed, Uriash is one of the largest titanosaurs

recovered to date from the Upper Cretaceous of Europe,

surpassed in size only by Abditosaurus. All these rea-

sons support the recognition of Individual C as a new

taxon.

As previously indicated, there are three further indi-

viduals/assemblages (Individuals H and I, and

Assemblage J) that provide important information on

the anatomy of Transylvanian titanosaurs, even though

there is currently insufficient evidence to either refer

them to one of the four named Hat

,eg taxa or

make them the basis of new taxa. Of these three sets of

specimens, Assemblage J is the only one with several

elements that can be compared with any of the named

taxa, and it presents anatomical similarities with both

Petrustitan (in the ulna) and Magyarosaurus (in the

femur; note, however, that the femur of Petrustitan is

currently unknown).

Similarly, because of the lack of sufficient compara-

tive data and diagnostic features for the remainder of

the individuals and assemblages identified in this survey

(see descriptions and comments in the Supplemental

Material), we refrain from referring them to any of the

four named Transylvanian titanosaur taxa recognized

here, or using them to erect new taxa, at least for the

time being. Thus, we maintain a working hypothesis of

a minimum diversity of four titanosaurian taxa in the

Maastrichtian of the Hat

,eg Basin, pending further dis-

coveries. These conclusions are broadly congruent with

those presented by Mocho et al. (2023), in which they

suggested that there is evidence for four different caudal

vertebral morphotypes for the titanosaurs of the Hat

,eg

and Transylvanian basins. However, it is important to

note that some of the differences between morphotypes

recognized by those authors might represent serial vari-

ation along the tail, rather than being indicators of taxo-

nomic separation (see the ‘Systematic palaeontology’

section, and the Supplemental Material).

One final comment regarding the diversity of Hat

,eg

titanosaurs concerns their spatial and temporal distribu-

tion. According to our survey of the occurrences of indi-

viduals/assemblages that can be positively referred to

one or another of the currently recognized four nominal

Revision of Romanian sauropod dinosaurs 95

taxa, these appear to show spatiotemporally distinct dis-

tributions (Figs 1,8). Specimens of Magyarosaurus

dacus occur between Ciula Mic

a and V

alioara in the

north-western Hat

,eg Basin, in beds assigned to the

lowermost part of the middle member of the Densus

Ciula Formation, and also at Pui in the central-eastern

Hat

,eg Basin, in beds belonging to the lower part of the

‘Pui beds’cropping out in this area (Csiki-Sava et al.,

2016; Therrien, 2005). Uriash is currently only known

from east of V

alioara in the north-eastern Hat

,eg Basin,

also from the lower part of the middle member of the

Densus

,-Ciula Formation, albeit from beds that are mar-

ginally younger than those yielding the remains of

Magyarosaurus (e.g. Botfalvai et al., 2021). All of these

occurrences are probably early Maastrichtian in age

(some perhaps as old as earliest Maastrichtian), and thus

fall into chronostratigraphical Tier 2 (or tiers 1/2 to 2)

of the Transylvanian faunal assemblages defined by

Csiki-Sava et al. (2016).

By contrast, the holotype and only known specimen

of Paludititan nalatzensis was discovered in the lower

part of the ‘R^

aul Mare beds’(see Csiki-Sava et al.,

2016), corresponding to chronostratigraphical Tier 3 (i.e.

late early to early late Maastrichtian in age). Botfalvai

et al. (2021) reported currently taxonomically indeter-

minate titanosaur specimens from site K2 near V

alioara

(here designated Individual V) that they assessed as

potentially referable to Paludititan or a closely related

taxon, despite coming from beds roughly contemporan-

eous with those yielding specimens of Magyarosaurus

dacus. Their suggestion was based on the presence of

non-procoelous caudal vertebrae within the middle part

of the tail of the K2 titanosaur, previously considered a

diagnostic feature of Paludititan (Csiki, Codrea, et al.,

2010). However, our survey of Hat

,eg titanosaur material

has revealed the presence of this feature in several other,

currently taxonomically indeterminate titanosaur speci-

mens (e.g. Assemblage J) and it can no longer be used

to unequivocally support referral to Paludititan.As

such, this removes the evidence supporting the possible

co-occurrence of Paludititan and Magyarosaurus within

the same unit/time interval.

Finally, specimens of Petrustitan hungaricus are

documented solely from the Sibis

,el Valley outcrops of

the S^

ınpetru Formation, although whether these come

from Tier 2 (earliest Maastrichtian), Tier 3 (late early to

early late Maastrichtian), or span the Tier 2–3 intervals,

is currently uncertain because of the scarcity of con-

straints on the relative stratigraphical positions of these

deposits compared to those from around V

alioara, Pui

and N

alat

,-Vad (Csiki-Sava et al., 2016). Nevertheless, it

appears likely that these Sibis

,el Valley occurrences are

somewhat older than that of Paludititan, although they

may be at least in part broadly synchronous with

Magyarosaurus and/or Uriash (Csiki-Sava et al., 2016;

Fig. 8).

To conclude, the currently recognized higher taxic

diversity of Hat

,eg titanosaurs (at least four different

taxa) does not translate automatically into a high diver-

sity of sympatric titanosaurs; to the contrary, it appears

that these taxa had largely non-overlapping areal and/or

temporal ranges, possibly representing members of dis-

tinct chrono- and/or eco-faunas. However, it should be

borne in mind that this clear-cut picture of differential

titanosaur distribution in the Hat

,eg Basin may result at

least in part from our current inability to refer further

individuals to one or another of these nominal taxa from

among the large amount of known titanosaur remains

that for the moment are left indeterminate.

Indeed, each of these nominal taxa are currently

known from a very small number (1 to at most 3) of

occurrences, which suggests that we might greatly

underestimate the range of their temporal (and probably

also spatial) distribution (e.g. Marshall, 1990). Thus,

their perceived chronostratigraphical and geographical

separateness may in fact simply be a by-product of their

currently limited fossil record instead of reflecting a

real-life distributional pattern. These adverse effects of a

limited fossil record may be alleviated in the future

through the discovery of new, more complete, associated

and diagnostic specimens, which should improve our

ability to refer further remains to these taxa.

The phylogenetic affinities of latest Cretaceous

European titanosaurs

In both sets of phylogenetic analyses, Magyarosaurus

dacus,Paludititan nalatzensis and Petrustitan hungari-

cus are not closely related to one another (note that

Uriash kadici was not included in these analyses

because of its incompleteness, but there is evidence to

suggest it is also not closely related to the other three

taxa –see below). This lends further support to our

retention of Paludititan nalatzensis as a taxon distinct

from Magyarosaurus dacus, and our removal of

‘Magyarosaurus’(?) hungaricus from this genus and

the erection of Petrustitan hungaricus.

As a result of its uncertain taxonomic status and chi-

maeric nature, Magyarosaurus has largely been

excluded from previous phylogenetic analyses. The sole

exception is the study of Curry Rogers (2005), along

with subsequent analyses that reran this matrix (e.g.

Csiki, Codrea, et al., 2010; Mannion & Upchurch,

2011). The position of Magyarosaurus was poorly con-

strained in these studies: such results potentially reflect

the impact of phylogenetic character scores obtained

from remains that can no longer be attributed with any

96 V. D



ıez D

ıaz et al.

certainty to this taxon. Here, Magyarosaurus is placed

either as a member of, or a close relative of,

Saltasauridae.

In their original description of Paludititan, Csiki,

Codrea, et al. (2010) recovered it in a range of positions

dependent on the underlying phylogenetic dataset,

including placing it as an early diverging titanosaur and

as a saltasaurid. A sister taxon relationship between

Paludititan and the Spanish genus Lohuecotitan was

identified by D



ıez D

ıaz et al. (2018), with this clade

positioned close to the ‘base’of Lithostrotia. A similar

position for Paludititan was supported by Navarro et al.

(2022), although those authors identified Lohuecotitan

as being phylogenetically deeply nested within the

Lirainosaurinae instead. Sallam et al. (2018) placed both

Lohuecotitan and Paludititan in a clade comprising

Lirainosaurinae and closely related titanosaurs, with

broadly similar results in other iterations of this matrix

(Gorscak et al., 2023; Gorscak & O’Connor, 2016,

2019; Vila et al., 2022). In our analyses, Paludititan is

consistently recovered either as a member of

Lognkosauria, or just outside of this clade, albeit still

closely related to Lohuecotitan. As such, our results dif-

fer, at least with respect to the wider phylogenetic posi-

tion of this taxon, from those of all previous studies,

while seemingly upholding a close Paludititan–

Lohuecotitan relationship.

The phylogenetic position of Petrustitan has not been

previously evaluated or discussed. This taxon has a less

consistent placement than Magyarosaurus and

Paludititan, although it is most closely related to South

American early diverging eutitanosaurian taxa in both

sets of analyses. Whereas Petrustitan groups with

Narambuenatitan and Pitekunsaurus in our EQW ana-

lysis, it clusters with Argentinosaurus,Patagotitan and

Puertasaurus in the EIW analysis.

Although it is not possible to quantitatively test the

phylogenetic affinities of Uriash at this stage, at least

one feature of this taxon is otherwise known only in

Gondwanan lithostrotian titanosaurs. This pertains to the

‘anteromedial arm’that characterizes the proximal half

of the humerus, which is a feature that had previously

been identified in the Argentinean genus Argyrosaurus

(Mannion & Otero, 2012), as well as the African genera

Paralititan and Rukwatitan (Gorscak et al., 2017), and

which we newly document in Gondwanatitan and

Muyelensaurus, from Brazil and Argentina, respectively

(PDM pers. obs.). As these Gondwanan taxa are recov-

ered in different parts of our trees, it is not possible to

narrow down the phylogenetic affinities of Uriash, but

it is noteworthy that none of the other European taxa

cluster with any of these five Gondwanan genera,

providing tentative support to the idea that Uriash

belongs to a distinct lineage.

Most previous studies have tended to place the late

Campanian Spanish genus Lirainosaurus either within,

or close to, the saltasaurid radiation (e.g. Calvo, Porfiri,

et al., 2007; Calvo, Gonz

alez Riga, et al., 2007; Curry

Rogers, 2005; Gorscak & O’Connor, 2016; Navarro

et al., 2022; Sanz et al., 1999; Upchurch et al., 2004),

although it has been recovered as an early diverging

lithostrotian or eutitanosaurian in some analyses (D

ıez

D

ıaz et al., 2018; Gallina & Apestegu

ıa, 2011; Salgado

et al., 2015). The contemporaneous French genus

Atsinganosaurus has been recovered as a member of

Lirainosaurinae in some studies (D

ıez D



ıaz et al., 2018,

2021; Navarro et al., 2022), whereas it has clustered

with Lognkosauria in others (Gorscak et al., 2023;

Gorscak & O’Connor, 2019; Sallam et al., 2018; Vila

et al., 2022). The early Maastrichtian Spanish taxon

Abditosaurus has previously been recovered as a salta-

saurid (Gorscak et al., 2023; Vila et al., 2022), without

close affinities to any other European titanosaur. In our

EQW analysis, Lirainosaurus is closely related to

Saltasauridae, forming a clade with the African taxon

Mansourasaurus, whereas Abditosaurus and

Atsinganosaurus form an early diverging lithostrotian

clade with the South American titanosaurs

Argentinosaurus and Patagotitan. By contrast, in our

EIW analysis, Atsinganosaurus is the sister taxon of

Lirainosaurus, forming a lirainosaurine clade with

Abditosaurus, with these taxa just outside the saltasaurid

radiation. The overall position of Lirainosaurus in our

analyses is broadly congruent with previous studies that

have supported close affinities with saltasaurids,

whereas the inconsistent placement of Atsinganosaurus

mirrors the contrasting positions recovered by other

authors. Both of our analyses support novel positions

for Abditosaurus. The incorporation of the early

Maastrichtian French titanosaur Ampelosaurus, which is

known from multiple individuals, preserving much of

the skeleton (Le Loeuff, 1995,2005b), might help to

better constrain the relationships of these western

European taxa. Ampelosaurus has been recovered as a

lirainosaurine in certain previous studies (D

ıez D



ıaz

et al., 2018,2021; Navarro et al., 2022; Vila et al.,

2022), but was excluded here from the phylogenetic

analyses given that it probably represents a chimaera

and is in need of revision (VDD pers. obs.).

Broader implications for the evolutionary

relationships of titanosaurs

There are a large number of additional novel results per-

taining to the evolutionary relationships of Titanosauria.

Here, we focus on several results that are broadly

Revision of Romanian sauropod dinosaurs 97

consistent across equal and extended weights analyses,

and that have implications for titanosaurian systematics.

Using a previous iteration of this data matrix, Poropat

et al. (2023) recovered Diamantinasauria as a titanosau-

rian clade in their EQW analysis, but it fell outside of

Titanosauria when EIW was applied. In the present

study, both analyses support a non-titanosaurian place-

ment for this clade (see also Beeston et al., 2024). With

Diamantinasaurus now one of the most completely

known somphospondylans (Beeston et al., 2024;

Hocknull et al., 2009; Poropat, Upchurch, et al., 2015;

Poropat et al., 2016,2021,2022,2023; Rigby et al.,

2022), this taxon potentially plays a pivotal role in

determining the polarity and acquisition of character

states at the base of Titanosauria.

In all of our analyses, the commonly recovered mem-

bers of Colossosauria (see Carballido et al., 2022) are

polyphyletic, forming three non-closely related clades.

Lognkosauria includes its two clade specifiers,

Futalognkosaurus and Mendozasaurus, as well as

Notocolossus and the European taxa Lohuecotitan and

Paludititan, but excludes Argentinosaurus,Patagotitan

and Puertasaurus. The latter three taxa form a clade of

early diverging lithostrotians, whereas Lognkosauria is

more deeply phylogenetically nested within Lithostrotia

(Figs 64,65). This disruption of Lognkosauria contrasts

with most recent analyses that have sampled a broad

array of these taxa, including previous iterations of this

data matrix (e.g. Agnolin et al., 2023; Carballido et al.,

2017; Gonz

alez Riga et al., 2018; Hechenleitner et al.,

2020; Mannion, Upchurch, Jin, et al., 2019), although

not all analyses have supported the recovery of a diverse

Lognkosauria (e.g. Gorscak & O’Connor, 2016,2019;

Navarro et al., 2022). The inclusion of non-South

American taxa within Lognkosauria is not novel to our

analysis. Previous iterations of the data matrix used

herein provided some support for a position at the ‘base’

of Lognkosauria for the Early Cretaceous French sauro-

pod Normanniasaurus (Averianov et al., 2021;

Mannion, Upchurch, Jin, et al., 2019), although this is

not supported in the present study. Material assigned to

the latest Cretaceous North American genus

Alamosaurus was recovered within this clade by some

authors (Navarro et al., 2022; Tykoski & Fiorillo,

2017), a configuration which requires further testing.

Lerzo et al. (2021) also found support for a lognkosau-

rian position for the early Late Cretaceous Dzharatitanis

from Uzbekistan, although we contend that this taxon is

much more likely to be a close relative of approximately

contemporaneous somphospondylans from East Asia,

such as Dongyangosaurus (e.g. see Mannion, Upchurch,

Jin, et al., 2019; Sues et al., 2015), to which those

authors made no comparisons. Nevertheless, our

consistent recovery of the European taxa Lohuecotitan

and Paludititan, as well as that of some Indo-

Madagascan taxa (Jainosaurus and Vahiny) in our EIW

analysis, within this clade provides support for a more

widespread distribution of Lognkosauria.

A close relationship between Lognkosauria and

Rinconsauria has also been found in a large number of

studies (e.g. Carballido et al., 2017; Gallina &

Apestegu



ıa, 2011; Gallina & Otero, 2015;Gonz



alez Riga

et al., 2018; Hechenleitner et al., 2020; Navarro et al.,

2022;P



erez Moreno et al., 2023; Salgado et al., 2015),

leading Gonz

alez Riga et al. (2019)toerect

Colossosauria to unite the two clades. By contrast, the

analyses of Gorscak and O’Connor (2016) and subse-

quent iterations of that data matrix (e.g. Gorscak &

O’Connor, 2019; Sallam et al., 2018; Vila et al., 2022)

recovered Rinconsauria and Lognkosauria as distantly

related lineages, with the latter clade more closely related

to Saltasauridae. Our analyses also support their phylo-

genetic separation, but here it is Rinconsauria that is con-

sistently placed as more closely related to Saltasauridae.

Both sets of our analyses recover the African titanosaur

Rukwatitan within Rinconsauria for the first time, with

unequivocal members otherwise represented by the two

South American clade specifiers, Rinconsaurus and

Muyelensaurus. The Early Cretaceous European taxon,

Normanniasaurus, is placed either as an additional rin-

consaurian or a close relative, providing further support

for a broader spatiotemporal distribution of this clade.

The relationships of the South American titanosaur

Aeolosaurus have been inconsistent across previous

studies (Carballido et al., 2022), with most analyses

placing it close to or within Rinconsauria (e.g. Calvo,

Porfiri, et al., 2007; Calvo, Gonz

alez Riga, et al., 2007;

Gorscak & O’Connor, 2019; Hechenleitner et al., 2020;

Navarro et al., 2022; Poropat et al., 2023; Salgado

et al., 2015; Santucci & Arruda-Campos, 2011), and

some supporting closer affinities with saltasaurids

(Gonz

alez Riga et al., 2018; Poropat et al., 2016).

Nevertheless, regardless of its relationship with other

taxa, in all cases, Aeolosaurus has been recovered as a

relatively later-diverging titanosaur. Our analyses there-

fore stand out notably in placing Aeolosaurus and

closely related taxa (‘aeolosaurines’) as a clade of early

diverging titanosaurs. Aeolosaurini is currently defined

as the most recent common ancestor of Aeolosaurus and

Gondwanatitan and all of its descendants (Carballido

et al., 2017; Franco-Rosas et al., 2004). Our EQW ana-

lysis restricts Aeolosaurini to the clade specifiers and

probably Arrudatitan and Caieiria, all of which are

from South America, with the Malagasy titanosaur

Rapetosaurus recovered as the sister taxon to the clade.

A more diverse Aeolosaurini is recovered in our EIW

98 V. D



ıez D

ıaz et al.

analysis, including additional South American taxa (e.g.

Tapuiasaurus), Rapetosaurus, the African titanosaur

Shingopana, and the Mongolian taxon Nemegtosaurus.

The position of Shingopana as a member of

Aeolosaurini is consistent with some of the few analyses

that have previously considered this taxon (Gorscak

et al., 2017; Gorscak & O’Connor, 2019). Although the

recovery of Nemegtosaurus as a member of

Aeolosaurini may appear extremely unexpected, it has

often been positioned as the sister taxon to

Tapuiasaurus, a South American taxon that had not

been recovered in this part of the tree prior to our analy-

ses. However, this particular grouping may potentially

be driven by a ‘monophyly of the preserved’effect,

rather than a genuinely close relationship (see e.g.

D’Emic, 2012; Poropat et al., 2023; Wilson et al.,

2016). Furthermore, neither Nemegtosaurus nor

Tapuiasaurus are recovered as closely related to

Aeolosaurini in our EQW analysis.

The analyses of Carballido et al. (2022) suggested

that Colossosauria was one of the more stable groups

inside Titanosauria, and those authors redefined it as the

most inclusive clade containing Patagotitan, but not

Saltasaurus. They also used Patagotitan as a clade spe-

cifier for two already existing, but largely unused clade

names, Eutitanosauria and Saltasauroidea. Carballido

et al. (2022) redefined Eutitanosauria as the most recent

common ancestor of Patagotitan and Saltasaurus and

all of its descendants, and Saltasauroidea as the most

inclusive clade containing Saltasaurus, but not

Patagotitan. The results from our analyses do not have

a substantial impact upon the content of Eutitanosauria,

the most notable outcome being the placement of

Aeolosaurus outside of this clade. However,

Colossosauria is now a depauperate clade, consisting of

just a few taxa, whereas nearly all eutitanosaurians

belong to Saltasauroidea. In parallel, Navarro et al.

(2022) proposed an alternative definition for

Saltasauroidea, as the most recent common ancestor of

Lirainosaurus,Opisthocoelicaudia and Saltasaurus, and

all of its descendants. Under this definition, the saltasau-

roid clade in our analyses would be much closer to the

taxonomic composition envisaged by Carballido et al.

(2022). However, rather than choosing between defini-

tions, or proposing further changes, we suggest that fur-

ther work is first required to test the robustness and

consistency of these different alternative phylogenetic

topologies.

Biogeographical implications for the assembly of

the European titanosaur fauna

A survey of the literature suggests that there are three

key hypotheses that could account for the presence of

titanosaurs in Europe during the latest Cretaceous:

Hypothesis 1 [H1] –sauropods died out in Europe dur-

ing the early Late Cretaceous and then re-invaded from

Africa in the Campanian–Maastrichtian; Hypothesis 2

[H2] –Early Cretaceous European titanosaurs gave rise

to lineages that persisted into the latest Cretaceous; and

Hypothesis 3 [H3] –titanosaurs were involved in faunal

exchange between Europe and Asia during the Late

Cretaceous. H2 can be regarded as having two variants:

H2.1 –Early Cretaceous European titanosaurs were

members of Gondwanan lineages that invaded the for-

mer area during the Barremian–Albian; and H2.2 –

Titanosauria, or at least some of its major clades, origi-

nated in Eurasia and gave rise to the Early Cretaceous

European fauna ‘in situ’. It should be noted that these

hypotheses are not necessarily mutually exclusive: with

multiple distinct titanosaur lineages present in latest

Cretaceous European faunas, it is conceivable that dif-

ferent groups had distinct biogeographical histories (e.g.

D

ıez D



ıaz et al., 2018). Moreover, some aspects of these

hypotheses can be combined: for example, titanosaurs

could have crossed from Africa to Europe (in the late

Early Cretaceous and/or the latest Cretaceous) and then

dispersed from the latter area to Asia (e.g. Upchurch, in

press).

Our revised Romanian titanosaur taxonomy and

updated phylogeny has produced evolutionary trees con-

taining seven Late Cretaceous European sauropod gen-

era (as well as Normanniasaurus from the Albian)

distributed across five (EIW) or seven (EQW) lineages,

providing a fresh opportunity to evaluate these biogeo-

graphical hypotheses. We have not carried out a quanti-

tative phylogenetic biogeographical analysis (such as

BioGeoBEARS; Matzke, 2013,2014): this is partly

because this would substantially increase the length of

an already long study, and partly because we believe

that such analysis should wait until further planned

work on titanosaurian phylogeny is completed. Instead,

below we first offer some caveats regarding fossil

record sampling, and then explore the key biogeograph-

ical hypotheses in more detail and determine to what

extent they are supported or contradicted by our new

insights into European sauropod lineages.

Fossil record sampling. Before examining individual

titanosaur relationships and assessing their biogeographical

implications, an important caveat concerning fossil

record sampling must be discussed. In our assessments

below, we make the assumption that a ghost lineage

occupied the same geographical area as its terminal

taxon. This is reasonable in terms of Occam’sRazor–

without additional evidence, we cannot justify more

complex assumptions of area occupancy. Thus, if we

have a European taxon X of Maastrichtian age that is

Revision of Romanian sauropod dinosaurs 99

the sister taxon of a South American taxon Y of Albian

age, the implied ghost range would be inferred as occu-

pying Europe from the Albian to the Maastrichtian.

The problem is that such assumptions can easily be

invalidated by uneven fossil record sampling. If, for

example, taxon Z from Campanian age deposits in

Africa is more closely related to taxon X than is taxon

Y, then the lineage extending from X þZbacktothe

Albian might have included taxa living in Europe, or

Africa, or both. Failure to sample taxon Z could thus

have a significant impact on the inferred biogeograph-

ical history. This is more than just a hypothetical

example: the Late Cretaceous dinosaur fossil record of

Africa is particularly poorly sampled (Gorscak et al.,

2023; Krause et al., 2019; Mannion & Barrett, 2013;

Sereno & Brusatte, 2008; Tortosa et al., 2014;Vila

et al., 2022; Upchurch, in press), as illustrated by the

relatively low numbers of such taxa in our data set.

The deleterious effects of sampling failures can be par-

tially alleviated by applying quantitative phylogenetic

biogeographical methods such as BioGeoBEARS, but

even these approaches are not immune to sampling

issues (e.g. Mannion, Upchurch, Jin, et al., 2019;

Poropat et al., 2016). This issue, combined with the

observation that our EQW and EIW trees have different

topologies and therefore imply different biogeograph-

ical histories, demonstrates the need to regard the fol-

lowing discussion as presenting working hypotheses

that should be tested through new specimens and fur-

ther work on titanosaur phylogenetic relationships.

H1: Late Out of Africa, and the ‘sauropod

hiatus’.During the 1980s and 1990s, two linked ideas

were proposed to explain the presence of titanosaurs in

latest Cretaceous European faunas. First, several studies

noted the apparent absence of sauropods in North

America and Europe throughout most of the Late

Cretaceous, the so-called ‘sauropod hiatus’(Le Loeuff,

1993; Le Loeuff & Buffetaut, 1995; Lucas & Hunt,

1989; see reviews in D’Emic et al., 2010; Mannion &

Upchurch, 2011; Upchurch, in press). Second, if the hia-

tus hypothesis is correct, then the latest Cretaceous

European titanosaurs must have invaded from another

region during the Campanian and/or Maastrichtian.

Titanosaurs have long been regarded as an essentially

Gondwanan radiation (e.g. Wilson & Upchurch, 2003;

see also Mocho et al., 2019), and, even today with

much better sampling from Laurasia (e.g. Mannion,

Upchurch, Jin, et al., 2019), the clade remains domi-

nated by taxa from South America, augmented by taxa

from Indo-Madagascar, with recently described contribu-

tions from Africa and Australia (e.g. Carballido et al.,

2017; Curry Rogers, 2009; Curry Rogers & Wilson,

2014; Gallina et al., 2022; Gorscak et al., 2023;

Gorscak & O’Connor, 2016; Hocknull et al., 2009;

Poropat et al., 2016; Sallam et al., 2018; Santucci &

Filippi, 2022; Wilson et al., 2009). Consequently, the

latest Cretaceous European titanosaurs have been inter-

preted as invaders from Gondwana, and the palaeogeo-

graphically most probable source area for such

dispersals has been identified as Africa (Astibia et al.,

1990; Buffetaut, 1989; Buffetaut et al., 1988; Csiki,

1997; Csiki-Sava et al., 2015; Ezcurra & Agnolin, 2012;

Gheerbrant & Rage, 2006; Le Loeuff, 1991,1993,

1997; Le Loeuff & Buffetaut, 1995; Novas et al., 2013;

Pereda-Suberbiola, 2009; Rage, 1996,2002; Vila et al.,

2022; Weishampel et al., 2010).

More recently, the validity of the sauropod hiatus has

been questioned (Mannion & Upchurch, 2011), espe-

cially regarding Europe (e.g. Pereda-Suberbiola, 2009;

Wilson & Upchurch, 2003). Although sauropods remain

unknown in North America from the late Cenomanian

to end-Campanian (inclusive) (D’Emic et al., 2010),

there is now strong evidence that they were present in

Europe during this time interval. In particular, a number

of Turonian–Santonian tracksites, as well as a tooth,

indicate the presence of sauropods in Europe at this

time (Mezga et al., 2006; Nicosia et al., 2007;}

Osi

et al., 2017b; Solt et al., 2020). This opens up the possi-

bility that at least parts of Europe were occupied by sau-

ropods continuously from the Early Cretaceous through

to the Maastrichtian, and that at least some of the latest

Cretaceous titanosaurs could be descendants of earlier

European forms rather than recent immigrants from

Africa or elsewhere (see below).

Palaeogeography provides another line of evidence

that potentially mitigates against the latest Cretaceous

‘out of Africa’explanation for European titanosaurs.

Recent palaeogeographical maps depict an extensive

marine barrier between Africa and various European

islands during the Campanian–Maastrichtian (Longrich

et al., 2021,2024; Scotese, 2016). This does not neces-

sarily falsify H1 if sauropods were capable of trans-

oceanic dispersal. Some workers have advocated that

such dispersals were feasible for sauropods and other

large-bodied dinosaurs (e.g. Dunhill et al., 2016;

Longrich et al., 2021,2024), but others have considered

this unlikely (Mannion, Upchurch, Schwarz, et al.,

2019; Poropat et al., 2016; Xu et al., 2018; Upchurch,

in press). Ezcurra and Agnolin (2012) and Vila et al.

(2022) argued for a landbridge linking Africa and

Europe via a re-emergent Apulian carbonate platform

during the latest Cretaceous. However, a continuous

Europe–Africa land route at this time seems unlikely

given current palaeogeographical data (Gheerbrant &

Rage, 2006; Longrich et al., 2021,2024; Upchurch, in

press). Dispersals of sauropods via ‘island hopping’

100 V. D



ıez D

ıaz et al.

seem more feasible, especially at times of lower sea

level, such as in the mid-Campanian and at the

Campanian/Maastrichtian boundary (Csiki-Sava et al.,

2015; Gheerbrant & Rage, 2006; Le Loeuff, 1991;

Longrich et al., 2021,2024;}

Osi, Apestegu

ıa, et al.,

2010; Vila et al., 2022), although this would still require

sauropods to have crossed some open bodies of water.

Recent support for the occurrence of latest Cretaceous

sauropod dispersals between Europe and Africa has

come from studies by Sallam et al. (2018), Vila et al.

(2022), and Gorscak et al. (2023). Sallam et al. (2018)

and Gorscak et al. (2023) described the titanosaurs

Mansourasaurus and Igai, respectively, from mid-

Campanian deposits in Egypt, and assessed their phylo-

genetic relationships and biogeographical histories via

Bayesian tip-dating methods and BioGeoBEARS.

Sallam et al.’s(2018) results suggested that

Mansourasaurus was nested in a latest Cretaceous clade

of European and Asian titanosaurs, and its ancestral lin-

eage had probably dispersed from Europe at approxi-

mately 75 Ma. This implies the existence of a viable

crossing between Europe and Africa, and it is interesting

that the inferred timing of this dispersal lies close to the

mid-Campanian KCa6 sea-level lowstand event (see

Haq, 2014) mentioned above. Gorscak et al. (2023)

found a similar topology, although they estimated that

the divergence of Mansourasaurus occurred at 80 Ma,

disrupting the previously proposed correlation with the

KCa6 sea-level lowstand; nevertheless, such an earlier

dispersal coincides with the more substantial earliest

late Campanian (80 Ma) KCa3 sea-level drop of (Haq,

2014). Our results are partially compatible with these

previous works. In particular, the EQW topology

includes Lirainosaurus and Mansourasaurus as sister

taxa (see also Navarro et al., 2022), supporting dispersal

between Europe and Africa prior to the late Campanian

(though the direction of this dispersal is currently uncer-

tain). By contrast, Mansourasaurus is the sister taxon of

latest Cretaceous South American saltasaurines in our

EIW trees. If the latter is correct, this would remove the

necessity to infer a latest Cretaceous Europe to Africa

dispersal event, and instead the presence of

Mansourasaurus could be interpreted as part of a South

America/Africa vicariance pattern resulting from their

final separation at 100 Ma (see Gorscak & O’Connor,

2016; Upchurch, in press [and references therein]). Igai

was placed as the sister taxon of a European lineage

(LirainosaurusþLohuecotitan) in the analyses of

Gorscak et al. (2023), with their divergence time esti-

mated as close to the Santonian/Campanian boundary.

As Igai is not included in our phylogenetic data set, we

provisionally accept Gorscak et al.’s(2023) view that it

supports dispersal between Europe and Africa during

the latest Cretaceous: we note, however, that the 95%

highest posterior density for this node age extends as far

back as the earliest Late Cretaceous and thus does not

definitively contradict H2.1.

Perhaps the strongest recent case for latest Cretaceous

dispersal of titanosaurs from Africa to Europe was pre-

sented by Vila et al. (2022). That study described the

early Maastrichtian Ibero-Armorican titanosaur

Abditosaurus and found support for a sister taxon rela-

tionship with the Cenomanian African form Paralititan.

BioGeoBEARS ancestral area estimation suggested that

the most recent common ancestor (MRCA) of

Paralititan and Abditosaurus lived either in Africa or

was widespread across the latter continent and Europe,

and that the ancestor of the lineage leading solely to

Abditosaurus occurred in Africa, or Europe, or both.

Vila et al. (2022) interpreted these results as supporting

dispersal from Africa to Europe during an earliest

Maastrichtian (70.6 Ma) sea-level lowstand. Additional

support for this scenario, based on sauropod body mass

variation and their egg fossil record, was also presented

by Vila et al. (2022), and linked to faunal turnover in

the middle part of the Maastrichtian. Essentially, Vila

et al. (2022) suggested that Europe was characterized by

a Campanian–early Maastrichtian sauropod fauna com-

prising small-bodied island dwarf forms (such as

Ampelosaurus,Atsinganosaurus,Lirainosaurus and

Lohuecotitan) that probably produced the endemic

oospecies Megaloolithus aureliensis and Megaloolithus

siruguei. During the early Maastrichtian, larger-bodied

titanosaurs from Africa, such as the Abditosaurus lin-

eage, invaded western Europe and replaced the dwarfed

forms as part of the ‘Maastrichtian Dinosaur Turnover’

(MDT) that also involved a shift away from sauropod-

dominated faunas to ones with a higher abundance and

diversity of ornithopods (especially lambeosaurine

hadrosaurs) (Csiki-Sava et al., 2015; Mocho et al.,

2019). Moreover, Abditosaurus was found in association

with the oospecies Fusioolithus baghensis. This oospe-

cies, and also Megaloolithus mamillare, are known from

Campanian and Maastrichtian age deposits in South

America and India (and potentially also Africa in the

case of the former egg type), but do not occur in

Europe until the early Maastrichtian. Vila et al. (2022)

therefore proposed that the transition from the

Megaloolithus aureliensisþMegaloolithus siruguei

assemblage to the Fusioolithus baghensisþMegaloolithus

mamillare one in Europe during the early Maastrichtian

marks the invasion of non-dwarfed titanosaurs from

Gondwana. Given that Abditosaurus is an Ibero-

Armorican taxon, the focus of Vila et al.’s(2022)

study was on that area; however, they speculated that

their invasion and replacement scenario might also

Revision of Romanian sauropod dinosaurs 101

apply to the Hat

,eg region where it could explain the

presence of dwarf forms such as Magyarosaurus and

large ones such as the taxon represented by Uriash

(but see below). In addition, although not specifically

highlighted by Vila et al. (2022), the occurrence of

the Transylvanian Paludititan, recovered in their

study as the sister-taxon of the north African

Mansourasaurus, could represent a potential second

example of such an early Maastrichtian out-of-Africa

titanosaur dispersal.

Vila et al.’s(2022) study provides an elegant and uni-

fying explanation of several disparate aspects of latest

Cretaceous European dinosaur biogeography. However,

closer scrutiny reveals a number of difficulties, espe-

cially in light of our revision of Hat

,eg titanosaur tax-

onomy and broader phylogenetic results. First, Vila

et al.’s(2022) own BioGeoBEARS results are more

compatible with H2.1 than H1: most estimates suggest

that Abditosaurus was descended from an ancestor that

lived in Europe, or Europe þAfrica, during the

Cenomanian. However, in any case, an augmented and

updated version of the Vila et al. (2022) data set was

analysed by Gorscak et al. (2023), resulting in the dis-

ruption of the Paralititan-Abditosaurus clade, with the

latter genus instead being placed as the sister taxon of

the South American early–middle Campanian titanosaur

Pellegrinisaurus. Second, our phylogenetic results also

do not support a sister taxon relationship between

Abditosaurus and an African form. Our EQW trees

place Abditosaurus as a member of an endemic latest

Cretaceous European lirainosaurine clade (with

Atsinganosaurus and Lirainosaurus), which is the sister

taxon of a South American lineage comprising the

Coniacian taxon Elaltitan and the late Campanian–early

Maastrichtian genus Bonatitan. These relationships are

not decisive with regard to choosing between H1 and

H2.1: the latter hypothesis would be favoured if we

assume that terrestrial dispersal between South America

and Africa ended around 100 Ma, when the Central

Atlantic Ocean severed their last remaining land connec-

tion (see Upchurch, in press and references therein), but

H1 can still be supported if we accept the feasibility of

transoceanic dispersal for sauropods or the presence of

land connections persisting much later into the

Cretaceous (e.g. Ezcurra & Agnolin, 2012; Sereno et al.,

2004). By contrast, our EIW trees place Abditosaurus as

the sister taxon of the late Cenomanian Argentinosaurus:

given the minimum age of their MRCA, such a relation-

ship is more consistent with H2.1. Third, significant spa-

tiotemporal gaps in the fossil record of sauropod eggs

mean that it is difficult to test the proposition that the pro-

ducers of Fusioolithus baghensis and Megaloolithus

mamillare originated in Gondwana and dispersed into

Europe during the early Maastrichtian. For example, des-

pite the presence of taxa such as Alamosaurus in North

America and Opisthocoelicaudia in East Asia during the

Maastrichtian, no fossil eggs referable to titanosaurs are

known from these regions at present. Finally, our revised

taxonomy and chronostratigraphical framework for Hat

,eg

titanosaurs does not lend support to the proposal that

large-bodied forms replaced dwarf forms, and, when con-

sidered together with data concerning other dinosaurian

taxa, also casts some doubt on the occurrence of a MDT

in Transylvania (see the ‘Body size evolution and island

dwarfism in Transylvanian titanosaurs’section below).

Aside from the LirainosaurusþMansourasaurus sister

taxon relationship in the EQW trees (see above), the stron-

gest (although still relatively weak) support for H1 in our

results involves Magyarosaurus and Petrustitan.Inthe

EQW trees, Magyarosaurus is the sister taxon of the wide-

spread clade containing Pellegrinisaurus,Alamosaurus,

saltasaurines, Isisaurus,andOpisthocoelicaudia,withallof

these taxa being Campanian or Maastrichtian in age.

However, given that this clade contains both Laurasian and

Gondwanan forms, it is difficult to determine the ancestral

area for the lineage leading to Magyarosaurus without

applying a more quantitative method. In the EIW trees,

Magyarosaurus is nested in a South American clade as the

sister taxon of the Maastrichtian Brazilian taxon

Baurutitan; and in the EQW trees, Petrustitan is the sister

taxon of a clade containing the early–middle Campanian

South American taxa Narambuenatitan and Pitekunsaurus.

Although no African taxa are associated with these portions

of our trees, some support for H1 can be claimed on the

basis that Magyarosaurus and Petrustitan typically have

their closest relatives in latest Cretaceous South American

faunas, implying potential Gondwana to Europe dispersals

during this interval. Such an interpretation requires either:

direct dispersal to Europe from South America (or via

Africa) during the latest Cretaceous, which seems unlikely

given the width and depth of the North and Central Atlantic

in the latest Cretaceous (Upchurch, in press and references

therein); or a substantial amount of sampling failure in the

European, South American, and especially African Late

Cretaceous terrestrial fossil records, which is a more plaus-

ible explanation (see above).

H2: Descendants of an Early Cretaceous

European fauna. The fossil record clearly demonstrates

the presence of titanosaurs in Early Cretaceous faunas

in Europe and western Asia, including Volgatitan from

the upper Hauterivian of western Russia (Averianov &

Efimov, 2018), fragmentary remains from the Barremian

Wessex Formation of the Isle of Wight, UK (D’Emic,

2012; Le Loeuff, 1993; Mannion et al., 2013; Upchurch

et al., 2011; Wilson & Upchurch, 2003), a

Malawisaurus-like taxon from the upper Aptian–lower

102 V. D



ıez D

ıaz et al.

Albian of Italy (Dal Sasso et al., 2016), and

Normanniasaurus from the lower Albian of France (Le

Loeuff et al., 2013). Given the evidence cited above for

the presence of sauropods in Europe during the

Turonian–Santonian, it is conceivable that latest

Cretaceous titanosaurs represent direct descendants of

these Early Cretaceous European faunas. The latter fau-

nas appear to include immigrants from Gondwana, as

indicated by several recent discoveries and phylogenetic

studies (Dal Sasso et al., 2016; Gorscak & O’Connor,

2016; Holwerda et al., 2018; Mocho et al., 2019;

Weishampel et al., 2010; Upchurch, in press), giving

rise to the H2.1 variant of H2. For example, the

Bayesian phylogenies and BioGeoBEARS analyses pre-

sented by Gorscak and O’Connor (2016) and Sallam

et al. (2018) supported the presence of a latest

Cretaceous Eurasian clade whose ancestral lineage dis-

persed from Gondwana at approximately 93 Ma.

Moreover, Dal Sasso et al. (2016) and Mocho et al.

(2019) reported titanosaur remains from the Albian of

Italy and Cenomanian of Spain, respectively, and

showed that these taxa had close affinities with

Gondwanan lineages. These putative dispersal events

coincide with the probable presence of a land connec-

tion (via the Apulian Route) during the Barremian–early

Albian.

The most frequently observed pattern in our results is

one in which one or more latest Cretaceous European

taxa are the sister taxa of a Gondwanan (typically South

American) form that occurred during the late Early or

early Late Cretaceous. This pattern is indeed most con-

sistent with H2.1. Such lineages include: (1) the

Abditosaurus-Argentinosaurus clade in the EIW tree

(see above); (2) the late Campanian Atsinganosaurus

(EQW) or the early–early late Maastrichtian Petrustitan

(EIW) clustering with the late Albian Patagotitan and

late Cenomanian Argentinosaurus; (3) the late

Campanian–early Maastrichtian Lohuecotitan and the

‘mid’-Maastrichtian Paludititan clustering with late

Turonian–Santonian members of Lognkosauria (EQW

and EIW); and (4) the lirainosaurine clade which

diverged from a South American lineage (including

Elaltitan) prior to the Coniacian (EQW). The early

Albian Normanniasaurus is the sister taxon of the puta-

tively Cenomanian African taxon Rukwatitan (EQW),

and, although not directly informative regarding the lat-

est Cretaceous sauropods of Europe, does also support

faunal exchange between these two areas in the late

Early Cretaceous. Recently, the possibility that

Rukwatitan is Campanian in age has been raised

(Widlansky et al., 2018), but the Albian age of

Normanniasaurus would still favour a late Early

Cretaceous Europe–Africa dispersal rather than a latest

Cretaceous one.

Although not mutually exclusive of Barremian–

Albian dispersals from Africa to Europe, it is also con-

ceivable that at least some components of the Early

Cretaceous European titanosaur fauna had occupied that

region since the earliest Cretaceous or even the Late

Jurassic (H2.2) (see e.g. Averianov & Efimov, 2018).

Indeed, the observation that the stratigraphically oldest

members of Titanosauria are typically Eurasian (with

the exception of the late Berriasian–Valanginian

Ninjatitan from Argentina; Gallina et al., 2021), could

be interpreted as indicating that this clade originated

there. If correct, then at least some of the latest

Cretaceous European titanosaur lineages might be direct

descendants of this initial radiation, either persisting in

situ in Europe or invading from Asia (see H3 below).

For example, the late Hauterivian Volgatitan from

Russia (Averianov & Efimov, 2018) potentially predates

the sea-level lowstands that apparently generated the

Apulian landbridge during the Barremian–early Albian

(Upchurch, in press): as such, the occurrence of this

Russian taxon is more consistent with H2.2 than H2.1.

Similarly, stratigraphically older taxa such as the

Valanginian Tengrisaurus (from Asiatic Russia;

Averianov et al., 2021) and Volgatitan have been placed

in early diverging positions in the titanosaurian clades

Lognkosauria, Colossosauria and Rinconsauria, suggest-

ing that these groups originated in Asia according to

Averianov and Efimov (2018) and Averianov et al.

(2021). Moreover, as noted above, there are several

instances of apparent dispersal involving Europe and

Africa during the middle or latest Cretaceous that can

be interpreted as Laurasian lineages invading

Gondwana.

Although a Eurasian origin of titanosaurs around

approximately the Jurassic/Cretaceous boundary cannot

be ruled out, there are some lines of evidence that sug-

gest this is a less likely scenario than a Gondwanan ori-

gin. In particular, there are far more Gondwanan

titanosaurs than Laurasian ones, and this results in most

of the Eurasian taxa being deeply nested within

Gondwanan clades. Such a recurrent topology might be

interpreted as the result of a sampling bias, but this

explanation seems unlikely given that current data sug-

gest that any such bias is in favour of Laurasia. For

example, inspection of The Paleobiology Database

(https://paleobiodb.org/, accessed on 15 August 2023),

indicates that 84% of Cretaceous dinosaur-bearing col-

lections, and 82% of Cretaceous tetrapod-bearing collec-

tions, have been recovered from Laurasian landmasses.

In short, although H2.2 cannot be rejected at this time

(especially given the difficulties of estimating

Revision of Romanian sauropod dinosaurs 103

phylogenetic relationships accurately for the often very

fragmentary earliest titanosaur specimens), a Eurasian

origin for titanosaurs followed by numerous earliest

Cretaceous dispersals to Gondwana seems less likely

than other scenarios based on current evidence.

H3: Europe-Asia faunal interchange in the Late

Cretaceous. Putative dispersals between Europe and

Asia during the Late Cretaceous have played an impor-

tant role in explaining several aspects of dinosaurian

biogeographical history, although this has largely

applied to non-sauropod groups such as dromaeosaurid

theropods, lambeosaurine hadrosaurs, and ceratopsians

(Csiki, Vremir, et al., 2010; Gierli

nski, 2015; Prieto-

M

arquez et al., 2013; Prieto-Marquez & Wagner, 2009;

}

Osi, Butler, et al., 2010; Sell

es et al., 2021; see review

in Upchurch in press). Recently, however, several

phylogenetic studies have also found evidence for a lat-

est Cretaceous opisthocoelicaudine clade comprising

European and Asian taxa (Gorscak & O’Connor, 2016;

Gorscak et al., 2023; Sallam et al., 2018; Vila et al.,

2022). Divergence time estimation and BioGeoBEARS

analysis by Sallam et al. (2018) generated a biogeo-

graphical scenario that fits well with our current under-

standing of palaeogeographical history. Essentially,

having crossed from Gondwana at around 93 Ma (see

above), a lineage of titanosaurs apparently gave rise to a

clade that became widespread across Eurasia in the mid-

Santonian (85 Ma). This geodispersal event coincided

approximately with a late Coniacian–early Santonian

drop in sea level that created a land connection between

Europe and Asia across the Turgai Sea (Baraboshkin

et al., 2003; Longrich et al., 2021,2024; Poropat et al.,

2016; Upchurch, in press). Subsequently, it appears that

this clade underwent two Europe/Asia vicariance events

in the early part of the Campanian (78–80 Ma),

reflecting increasing sea levels during the late Santonian

and early Campanian that severed the land connection

across the Turgai Sea and made the latter a more sig-

nificant marine barrier to terrestrial dispersal

(Baraboshkin et al., 2003; Upchurch, in press).

However, evidence for a Eurasian opisthocoelicaudine

clade is not present in either the EQW or EIW topolo-

gies. Indeed, the only sister-taxon relationships between

European and Asian titanosaurs are seen in the EIW

trees, involving taxa of Early/Late Cretaceous boundary

age (i.e. the Albian Normanniasaurus from France, and

the Cenomanian Dongyangosaurus and

Jiangshanosaurus from China) (see also Averianov

et al., 2021, who recognized a close relationship

between Normanniasaurus and Tengrisaurus). Although

this represents a Eurasian clade of titanosaurs, it is not

currently known to have given rise to any latest

Cretaceous forms. Moreover, the geodispersal event(s)

that would have produced this Eurasian clade predate

the Coniacian–Santonian sea-level lowstand highlighted

by Upchurch (in press). Therefore, this mid-Cretaceous

biogeographical pattern could be part of a more general

phenomenon among dinosaurs whereby an Aptian–

Albian sea-level lowstand established land connections

between Europe and Asia, Europe and Africa, and

potentially also Europe and North America, allowing

multiple dispersal events (e.g. Upchurch, in press;

Upchurch et al., 2002).

Summary. Part of the difficulty in understanding the

origins of the Campanian–Maastrichtian European tita-

nosaur faunas stems from placing dispersal from

Gondwana in opposition to the idea of Early Cretaceous

European/Eurasian relicts persisting into the latest

Cretaceous. However, these two concepts can be recon-

ciled by postulating Barremian–Albian dispersals from

Africa into Europe (and vice versa), followed by sur-

vival of those lineages through the ‘sauropod hiatus’

into the latest Cretaceous (H2.1). Notwithstanding issues

of uneven sampling and a particularly poor Late

Cretaceous African terrestrial fossil record, H2.1 is most

consistent with our results. This scenario side-steps

ongoing controversies over the extent of marine barriers

between Europe and Africa, and between the latter and

South America, during the latest Cretaceous, and also

avoids the necessity of inferring that titanosaurs under-

went multiple long-distance transoceanic dispersals.

Moreover, H2.1 also matches the growing evidence for

continuous occupation of the European region by titano-

saurs from at least the Barremian through to the

Maastrichtian. Yet, the combination of titanosaur egg

chronostratigraphical data, occurrence of several epi-

sodes of sea-level lowstands in the Campanian and ear-

liest Maastrichtian, and the phylogenetic relationships of

Mansourasaurus and Igai, mean that it would be prema-

ture to fully reject H1. The evidence supporting these

different ideas tends to wax or wane depending on

which phylogeny is preferred. For example, our phylo-

genetic results are clearly more consistent with H2.1,

and offer only sparse and weak support for H1 and H3.

By contrast, phylogenies based on the Gorscak and

O’Connor (2016) family of data matrices tend to support

a Eurasian clade (H3), although usually in a way that is

also compatible with H2.1 (e.g. lineages crossing from

Gondwana into Europe during the mid-Cretaceous and

then dispersing from Europe into Asia during the Late

Cretaceous). Thus, establishing a consensus on the rela-

tionships and divergence times of European and Asian

titanosaurs, and improved sampling of the critical

Cenomanian–Santonian time interval in both Europe and

Africa, will be central to resolving these biogeographical

debates.

104 V. D



ıez D

ıaz et al.

Titanosaur colonization of the latest Cretaceous

Hat

,eg Island

The absence of any close phylogenetic relationship

between Magyarosaurus,Paludititan,Petrustitan (Figs.

64,65), and probably Uriash, combined with the

absence of any pre-Maastrichtian titanosaur remains on

the Transylvanian landmass (e.g. B

alc et al., 2024;

Vremir et al., 2014), suggests that representatives of

each of these lineages migrated into this insular area

after the Campanian/Maastrichtian boundary, via four

separate dispersal events. Although the details of these

dispersals remain contentious (see the previous section),

it is nonetheless clear that the assembly of the Hat

,eg

Island titanosaur fauna was a protracted process involv-

ing several different episodes of biotic connection with

other areas. Most intriguingly, the near-absence of any

particularly close phylogenetic relationships between

most latest Cretaceous Transylvanian and Ibero-

Armorican titanosaurs also suggests that there may had

been little faunal exchange between these two parts of

the Late Cretaceous European Archipelago, and that

most of the Transylvanian and Ibero-Armorican titano-

saurs were derived from different source areas.

This multi-step assembly scenario, fuelled by recur-

rent faunal exchange events, potentially characterizes

the Hat

,eg Island faunas overall, as was recently hinted

at in the description of a new Transylvanian rhabdodon-

tid by Augustin et al. (2022). Thus, despite ample evi-

dence for a high level of endemism in the Hat

,eg Island

assemblages (e.g. Csiki & Grigorescu, 2007; Csiki-Sava

et al., 2015; Weishampel et al., 2010), which is rein-

forced by our results, it appears that this area may not

have been as isolated as previously thought (e.g.

Weishampel et al., 2010) and that occasional faunal con-

nections were possible throughout the latest Cretaceous.

Although this combination of endemism and recurrent

episodes of faunal connection might appear counter-

intuitive at first glance, similar patterns have been docu-

mented in the Cenozoic insular fossil record (see exam-

ples in e.g. van der Geer et al., 2021). Such instances

usually result from a synergic interaction between

repeated episodes of: (1) areal connections being estab-

lished among previously isolated areas due to short-term

eustatic changes and/or tectonic events (which promote

range expansion of previously isolated faunal elements);

and (2) subsequent rapid evolution of the newly intro-

duced faunal elements (a characteristic of insular evolu-

tion; e.g. Cucchi et al., 2014; Keogh et al., 2005;

Millien, 2006; van der Geer et al., 2021 and references

therein) once connections are severed again and some

degree of more stringent biotic isolation is re-estab-

lished. The relatively large number of eustatic sea-level

fall episodes that took place during the first half of the

Maastrichtian (four in fewer than 4 million years; Haq,

2014), all of which are documented in Europe, com-

bined with the tectonically active nature of the entire

eastern European area, including the surroundings of the

Transylvanian landmass (e.g. Csiki-Sava et al., 2016;

Schmid et al., 2020; van Hinsbergen et al., 2020), and

references therein), would have created the ideal setting

for such a palaeobiogeographical evolutionary scenario.

Body size evolution and island dwarfism in

Transylvanian titanosaurs

Dwarfism. One of the most striking features of the

Transylvanian titanosaurs, excluding Uriash, is their

small size compared to most other sauropods. Nopcsa

(1914; see also Nopcsa, 1923a) noted a similar phenom-

enon among other sympatric herbivorous dinosaurs

(ornithopod and ankylosaur ornithischians) and linked it

to an isolated island habitat. He hypothesized that the

small titanosaurs were insular dwarfs, Mesozoic correla-

tives of the Plio-Pleistocene dwarf elephants of the

Mediterranean islands. The dwarfed status of the Hat

,eg

titanosaurs was supported by morphometric regression

analyses of their humeri by Jianu and Weishampel

(1999), who regarded it as a by-product of paedomorph-

osis. This interpretation was, however, subsequently

questioned by Le Loeuff (2005a, fig. 1a), who noted the

presence of large skeletal elements (including the

humerus of Uriash) among the titanosaurian fossils

from the Hat

,eg Basin, demonstrating that not all

Transylvanian titanosaur specimens were of small size.

Le Loeuff (2005a) proposed that the relatively small

specimens were merely juveniles of a larger-bodied

taxon, and that the observed abundance of small ele-

ments could be explained by taphonomical and/or

palaeoecological biases. In order to resolve this issue,

Stein et al. (2010; see also Benton et al., 2010) con-

ducted an osteohistological survey of a large sample of

Transylvanian titanosaur limb bones, concluding that the

most common remains did indeed belong to dwarfed,

adult individuals. Nevertheless, the presence of a larger-

bodied taxon in this sample was also suggested based

on distinctive osteohistological traits shown by one of

the sampled elements, reigniting a long-lasting thread of

taxonomic discussions concerning these titanosaurs.

A total of 22 appendicular specimens were sampled

by Stein et al. (2010, see table 1), and the new referrals

presented in the current work provide an opportunity to

review and update their data and conclusions. Of the

originally sampled elements: 16 are currently regarded

as indeterminate/unidentified titanosaurs; five can be

referred to Magyarosaurus dacus –the humerus SZTFH

Ob.3089 and the fibula SZTFH Ob.3086 of Assemblage

A (representing the type material of this taxon), and the

Revision of Romanian sauropod dinosaurs 105

humerus LPB (FGGUB) R.1047 and the femora LPB

(FGGUB) R.1046 and R.1992 of Individual E; and one

belongs to Uriash kadici –the humerus SZTFH

Ob.3104 (Individual C). Stein et al. (2010) highlighted

significant osteohistological differences between SZTFH

Ob.3104 and those referred here to Magyarosaurus

dacus, and indicated that this humerus cannot be placed

on the same growth trajectory with those referred to the

latter taxon. Indeed, although being markedly larger, the

humerus of Uriash kadici (SZTFH Ob.3104) displays an

earlier histological ontogenetic stage (HOS11) than the

elements referred to Magyarosaurus dacus (HOS13 for

the humerus LPB [FGGUB] R.1047; HOS14 for the

femora LPB [FGGUB] R.1046 and R.1992 of Individual

E, and the elements sampled from Assemblage A), mak-

ing Uriash an ontogenetically younger titanosaur indi-

vidual than the sampled material of M. dacus.

Magyarosaurus has the best skeletal representation of

any Transylvanian titanosaur, allowing robust body size

and mass estimates for several individuals using the for-

mulae proposed by Benson et al. (2018), Packard et al.,

(2009), and Seebacher (2001) (see Table 2). With an

estimated total length of less than 3 metres,

Magyarosaurus is the smallest adult sauropod currently

known, smaller than the Brazilian dwarf titanosaur spe-

cies Ibirania parva (5.7 metres, Navarro et al., 2022),

and the latest Cretaceous Ibero-Armorican taxa

Lirainosaurus and Atsinganosaurus, which reached body

lengths of 4–9 metres (D



ıez D

ıaz et al., 2021,table 3).

Moreover, the adult body mass estimate for M. dacus of

under 1 tonne has not been observed for any other tita-

nosaur (Benson et al., 2018). All of this evidence con-

firms that Magyarosaurus was the product of a dramatic

dwarfing process. However, it should be noted that the

body length value of 2.5–2.7 metres for Magyarosaurus

would indicate a remarkably gracile, long-limbed body

shape for this animal, with an estimated hip height of

approximately 1 metre based on the summed lengths of

its hind limb elements (Table 2). These could be

regarded as unrealistic body proportions for a sauropod,

perhaps indicating that the formulae used here for calcu-

lating titanosaur body lengths have problematically wide

error margins when applied to particularly small taxa.

However, comparable changes in overall body shape

(relative elongation of limbs, and overall more gracile

posture) have also been reported for other Late

Cretaceous European island-dwelling dinosaurs ( }

Osi &

Mak

adi, 2009), suggesting that such body shape modifi-

cations might have arisen convergently as a result of

adaptation to the demands of island life.

Elements from two other key titanosaur individuals/

assemblages discussed in this study were also sampled

by Stein et al. (2010), including several limb bones

from Assemblage J in the LPB (FGGUB) collection

(humerus [R.1246], ulna [R.1598], femora [R.1220,

R.1511], and tibia [R.1252]), and both elements referred

to Assemblage M (humerus [SZTFH v.13492], and ulna

[SZTFH Ob.3099]). In the case of Assemblage J, the

osteohistological maturity of the elements varies

between HOS12 (tibia) to HOS14 (humerus, ulna), des-

pite all the sampled elements being even smaller than

those of Magyarosaurus. The same is also true for

Assemblage M, where both elements display very

advanced osteohistological stages (HOS13–14), for a

body size comparable to that of Magyarosaurus. Stein

et al. (2010) implicitly considered all these elements

referable to Magyarosaurus dacus, because they could

be easily accommodated within the same growth trajec-

tory. However, here we are unable to unequivocally

refer these assemblages to M. dacus based on shared

autapomorphies, and thus they are currently left as inde-

terminate titanosaurs. Nevertheless, these occurrences

show that the advanced degree of dwarfing documented

in Magyarosaurus was potentially more widespread

among Transylvanian titanosaurs.

The specimens referred to Petrustitan have dimen-

sions close to those of Magyarosaurus (Table S1),

although Petrustitan may had been somewhat larger, in

line with the original proposal of Huene (1932). The

largest Petrustitan specimen (Individual B, the type

material) is estimated to have been approximately 16–

20% larger in linear dimensions than the largest speci-

men referred to Magyarosaurus (Individual E), assum-

ing a more or less isometric scaling between the two

taxa. This would translate into a body length estimate of

c. 3.2 metres for Petrustitan, a figure that still places it

at the lower end of the titanosaur body size spectrum,

even when compared to the other generally small-bodied

European titanosaurs. However, because no histological

analyses have been carried out for any Petrustitan speci-

mens, it is currently unknown whether they belong to

immature or adult individuals, and thus their attribution

to a dwarf species must be treated with caution at

present.

The elements of Paludititan that overlap anatomically

with Magyarosaurus (mainly dorsal and caudal verte-

brae) similarly indicate a somewhat larger body size for

the former taxon (3.5–4 metres in length), albeit still

markedly smaller than the largest known specimens of

Lirainosaurus and Atsinganosaurus (4–9 metres in

length). Furthermore, the degree of fusion between

neural centra and arches in both the dorsal and caudal

vertebrae of the only known individual of Paludititan

suggests that it most probably represents an ontogenetic-

ally advanced growth stage. However, in the absence of

osteohistological data, its adult status cannot be

106 V. D



ıez D

ıaz et al.

confirmed unequivocally at present. To summarize, both

Petrustitan and Paludititan potentially represent further

examples of small-sized titanosaurs present on the

Transylvanian landmass, although whether they also dis-

played the extreme degree of dwarfism recognized in

Magyarosaurus and assemblages J and M cannot cur-

rently be ascertained. Nevertheless, the available evi-

dence (e.g. absolute size of the largest skeletal elements

belonging to their different referred individuals) sug-

gests only marginally larger adult body sizes for the for-

mer two taxa when compared to Magyarosaurus.

Nopcsa (1914) proposed that the insular environment

of Hat

,eg Island was the main cause of the dwarfing dis-

played by the Transylvanian dinosaurs. Subsequently,

the nature of this phenomenon has become much better

understood with the discovery and interpretation of a

large number of Recent and fossil examples of distinct-

ive body size changes occurring on islands. Island

dwarfism is part of what is often termed the ‘island

rule’, according to which large-sized animals tend to

evolve towards smaller body sizes, and small-sized ani-

mals tend to evolve towards larger body sizes, both sup-

posedly converging on a hypothetical ‘optimal body

size’characteristic for a given clade and ecological type

(e.g. Ben



ıtez-L

opez et al., 2021; Foster, 1964;

Lomolino, 2005). Several different mechanisms have

been proposed to account for these patterns (e.g. see the

synthesis by van der Geer et al., 2021). These mecha-

nisms are not necessarily mutually exclusive and

include: adaptation to resource (area, food) limitations

on islands; ecological release (from predators and/or

competitors); niche expansion and/or shifts resulting

from adaptations to the particulars of the insular envi-

ronments; and optimization/change of life-history traits

(see also Benton et al., 2010). In addition, energetic

constraints related to movement across the potentially

more rugged habitats of a tectonically active island –

such as that reconstructed for the Transylvanian land-

mass –may have also promoted body size reduction

(e.g. Ruxton & Wilkinson, 2014). Although the identifi-

cation of the precise mechanisms driving dwarfism in

Transylvanian titanosaurs remains problematic, the iso-

lated insular environment (e.g. Benton et al., 2010;

Csiki-Sava et al., 2015; Nopcsa, 1915,1923a;

Weishampel et al., 1991,2010; but see dissenting opin-

ions by Jianu & Boekschoten, 1999; Krause et al.,

2020) and its unbalanced faunal content (e.g. Csiki-Sava

et al., 2015,2016) almost certainly resulted in both

resource limitations and ecological release (see below).

Larger-bodied titanosaurs. Uriash kadici is estimated

to have had a body mass of between 5 and 8 tonnes,

and a body length close to 12 metres (Table 2). This

means that Uriash is the largest titanosaurian taxon

known from the Hat

,eg Basin, and surpassed the max-

imum values reached by most other Late Cretaceous

European titanosaurs (e.g. D

ıez D



ıaz et al., 2021), with

the exception of Abditosaurus (with estimates of 14

tonnes and 17.5 meters in length) (Vila et al., 2022).

The presence of large-sized titanosaurs such as

Uriash is noteworthy and requires explanation because

it appears to contradict (or at least undermine) the sup-

posed action of the ‘island rule’upon these faunas.

Several hypotheses have been put forward to account

for the presence of these ’outlier’large-bodied titano-

saurs within dinosaur assemblages that were otherwise

characterized by the dominance of small-sized (and

potentially dwarfed) herbivorous taxa (e.g. Benton et al.,

2010;}

Osi et al., 2014; Stein et al., 2010). In particular,

Stein et al. (2010, p. 9261) proposed several scenarios

that might account for the presence of larger-bodied tita-

nosaurs in this fauna: S1, sea-level drops and/or tectonic

events led to the expansion of the emergent land area,

relaxing the potential constraints of insular resource

limitation, and allowing newly introduced and large-

bodied taxa to retain their body sizes (see also Vila

et al., 2022); S2, ‘accidental’occurrences of stray indi-

viduals of larger-bodied taxa that originated from larger

nearby landmasses, either as the result of chance over-

seas dispersal or as washed-up carcasses; and S3, large-

bodied Hat

,eg taxa represent members of the earliest

stages of an immigration wave. Scenario 3 has two var-

iants: S3.1, the large-bodied immigrants eventually gave

rise to dwarfed forms because of the operation of

island-rule processes; and S3.2, the large-bodied immi-

grants eventually went extinct because the limitations on

resources meant that Hat

,eg Island could not support

them (see also Benton et al., 2010). These scenarios are

not mutually exclusive, and all of these processes may

have operated to some extent to bring about the pres-

ence of larger-body forms on Hat

,eg Island.

Nevertheless, our revised understanding of the tax-

onomy, geographical distributions, and stratigraphical

ranges, of Transylvanian titanosaurs, provides an oppor-

tunity to re-evaluate these scenarios.

In our view, beyond the statistically highly improb-

able discovery of the remains of some accidentally

stranded animals (whether arriving alive or as floating

carcasses), S2 can be reasonably dismissed, given the

continuous (albeit rare) occurrences of large-sized tita-

nosaur individuals throughout the Maastrichtian on the

Transylvanian landmass (see the ‘Did the “Maastrichtian

Dinosaur Turnover”occur in the Hat

,eg Basin?’section).

Rejection of S2 is also consistent with the lack of

palaeogeographical data supporting the existence of sig-

nificantly larger landmasses populated by large-sized

titanosaurs in close proximity to the latest Cretaceous

Revision of Romanian sauropod dinosaurs 107

Hat

,eg Island. Indeed, the only currently documented

example of a geographically close titanosaur occurrence

that comes from a sizeable emergent area in the western

Balkans is both significantly older (dated close to the

Santonian/Campanian boundary) and very probably per-

tains to a small-sized, potentially dwarfed taxon

(Nikolov et al., 2020).

Scenario S3 similarly requires the fortuitous (and

palaeontologically highly improbable) sampling of the

earliest settling stages of a large-bodied immigrant tita-

nosaur population (i.e. Uriash). In addition, as with S2,

the temporally recurrent occurrence of large-sized titano-

saur fossils in different areas of the Transylvanian land-

mass mitigates against a scenario in which we have

obtained a fortuitous glimpse of the potentially strati-

graphically very narrow window between the earliest

arrival of large body forms and their subsequent dwarf-

ism/extinction. Under S3.1, the original large-bodied

taxon is expected to have undergone dwarfing over time

according to the ‘island rule’, leading to subsequent

establishment of either a population displaying smaller-

body size on average, or a smaller-bodied descendant

taxon; for example, van der Geer et al. (2021) presented

several instances of such outcomes in Cenozoic island

mammals. However, neither of these possible outcomes

are supported in the case of the Hat

,eg Basin titanosaurs

because Uriash does not seem to be particularly closely

related to (i.e. con-specific with or ancestral to) any of

the stratigraphically younger and smaller-sized titano-

saurs known from Hat

,eg Island. Scenario S3.2 would

require extinction of these immigrant larger-bodied taxa

rather shortly after their introduction on Hat

,eg Island, in

order to prevent preservation of any further individuals.

Moreover, given the recurrent presence of large titano-

saur specimens in the Hat

,eg fauna, S3.2 would require

such local extirpations to occur repeatedly during the

Maastrichtian, such that none of these large-sized titano-

saurs were able to permanently establish themselves as

components of the Hat

,eg Island palaeofaunas. Again,

such a pattern has been documented in Cenozoic placen-

tal mammals by van der Geer et al. (2021). However,

S3.2 is difficult to test in the case of Transylvanian tita-

nosaurs because we lack the data needed to distinguish

between genuine absence resulting from extinction as

opposed to pseudo-absence caused by factors such as

sampling failures, lack of stratigraphical resolution, and/

or taxonomic uncertainty.

Of the scenarios proposed to date, S1 appears to be

the palaeobiogeographically, palaeogeographically and

palaeoecologically soundest explanation for the presence

of large-bodied titanosaurs on the Transylvanian land-

mass, whereby expansion of the emergent land area

would have provided suitable accommodation for large-

bodied late-arriving immigrants. This is consistent with

palaeogeographical hypotheses for the evolution of the

Hat

,eg Island that propose a progressive enlargement of

its land area from the late Campanian through to the

Maastrichtian (e.g. Csiki-Sava et al., 2016; Vremir

et al., 2014). Enlargement of Hat

,eg Island is potentially

explained by latest Cretaceous mountain-building events

in the circum-Transylvanian Basin region that suggest

generalized uplift (e.g. Kr

ezsek & Bally, 2006;

Willingshofer et al., 2001). Nevertheless, the quasi-con-

tinuous coexistence of small and large Transylvanian

titanosaurs from the beginning of the Maastrichtian

onwards (see the ‘Did the “Maastrichtian Dinosaur

Turnover”occur in the Hat

,eg Basin?’section) suggests

that increase in land area and corresponding resources

may not have been the main driver behind the success-

ful settlement of large-sized titanosaurs.

In short, scenarios S2 and S3.1 are plausible, but rely

on chance observations of rare events and therefore

appear statistically improbable. Scenarios S1 and 3.2

appear to be more probable and are supported by analo-

gous dwarfing of placental mammals during the

Cenozoic, but S3.2 is difficult to test. None of these pre-

viously proposed scenarios fully account for the appar-

ent coexistence of both small- and large-bodied

titanosaurs in the same Hat

,eg faunas, and we therefore

propose two new scenarios. The first of these, S4,

relates to ecological exclusion. Case studies of insular

body size evolution in large placental mammals (e.g.

proboscideans) suggest that drastic body size reduction

is prohibited whenever smaller-sized herbivores are pre-

sent alongside the invading larger-bodied proboscideans

(see examples discussed in van der Geer et al., 2021).

This is because dwarfing of originally large megaherbi-

vores results in increased competition from incumbent

members of the mesoherbivore guild. These cases sug-

gest that important potential drivers of island dwarfing

(e.g. resource limitations, release from predator pres-

sure) may be trumped on occasion by the adverse

effects of ensuing ecological competition. Thus, larger-

sized titanosaurs that invaded the Hat

,eg Island region

would have been sympatric with several smaller-bodied

herbivorous dinosaurs (e.g. ornithopods, dwarfed titano-

saurs). According to S4, any attempt by these large-bod-

ied sauropods to reduce their body size might have been

evolutionarily counter-productive because smaller bod-

ied herbivorous niches were already occupied, regardless

of the concurrent action of other mechanisms that would

otherwise have promoted island dwarfing. Ecological

exclusion, itself driven in part by effects of the ‘island

rule’upon other contemporaneous herbivores, was

potentially a key factor allowing the quasi-continuous

108 V. D



ıez D

ıaz et al.

existence of large-bodied titanosaurs alongside smaller

ones on Hat

,eg Island.

Our final scenario, S5, is that the dwarf titanosaurs

are not the product of island dwarfing at all, or at least

this dwarfing did not occur on Hat

,eg Island. Under this

scenario, dwarfing occurred stratigraphically earlier

among several lineages, and the small-bodied titanosaurs

on Hat

,eg Island are the descendants of existing dwarfed

ancestors. This would explain why Uriash also lived in

the same environment, without reducing its body size.

This scenario would be in keeping with the fact that

most latest Cretaceous Ibero-Armorican titanosaurs are

also small-bodied, and that at least one contemporan-

eous titanosaurian lineage also shows evidence of a

reduction in body size, namely South American salta-

saurines (e.g. Navarro et al., 2022; Powell, 2003), dem-

onstrating that dwarfism is relatively widespread

amongst Titanosauria (see also D’Emic, 2023). A simi-

lar explanation has been proposed recently to explain

the presence of at least some of the small-bodied latest

Cretaceous hadrosauroids from of Europe, i.e. that these

are members of an existing lineage of smaller taxa

(Chiarenza et al., 2021). Given that our sampling of pre-

late Campanian titanosaurs from Europe (and elsewhere)

is scarce (Mannion & Upchurch, 2011), we cannot rule

out that unsampled taxa ancestral to the Transylvanian

titanosaurs had already undergone body size reductions

(e.g. see the material described by Nikolov et al., 2020,

of a small-bodied titanosaur from the Santonian/

Campanian boundary beds of the western Balkans).

In summary, although we accept that both enlarge-

ment of the available land area (S1) and/or episodes of

local extinction (S3.2) potentially contributed to control-

ling the body size evolution of Transylvanian titano-

saurs, we contend that S4 (ecological exclusion) and/or

S5 (pre-Hat

,eg Island dwarfing) are the more likely scen-

arios for explaining the observed body size disparity.

Did the ‘Maastrichtian Dinosaur Turnover’

occur in the Hat

,eg Basin?

The Maastrichtian Dinosaur Turnover (MDT) was first

proposed as a rather abrupt transition from rhabdodon-

tid- and titanosaur-dominated to hadrosauroid-dominated

faunas that took place in the Ibero-Armorican domain

around the early/late Maastrichtian boundary (e.g. Le

Loeuff et al., 1994). Subsequently, the timing, extent,

and impact of the turnover have been further refined,

painting a more nuanced picture of this transition (e.g.

Fondevilla et al., 2016; Vila et al., 2012,2016). For

example, the faunal replacement is now estimated to

have been protracted rather than abrupt, extending over

a time interval of 2.5–2.8 million years, from the middle

of the early Maastrichtian to the early part of the late

Maastrichtian (i.e. the early part of magnetochron C31r

to magnetochron C30r; Fondevilla et al., 2019).

Moreover, a survey of the Ibero-Armorican latest

Cretaceous fossil record by Vila et al. (2012) suggested

that both small- and larger-sized titanosaurs were present

up until the very end of the Maastrichtian, alongside sev-

eral hadrosauroid lineages, whereas rhabdodontids and

nodosaurids had apparently become extinct by the end of

the early Maastrichtian (Fondevilla et al., 2019). Vila

et al. (2022) proposed a variant of the MDT in which

small-bodied (potentially dwarfed) titanosaurs formed

endemic Campanian–early Maastrichtian faunas in Ibero-

Armorica and possibly elsewhere in Europe, followed by

the invasion of larger-bodied forms such as Abditosaurus

from outside Europe during the early Maastrichtian.

These authors further supported this scenario with evi-

dence from the sauropod egg fossil record, and tentatively

extended this MDT scenario to include Transylvania as

well as Ibero-Armorica (see above). For example, these

authors noted that the larger Transylvanian forms

appeared to be stratigraphically younger than smaller-bod-

ied forms such as Magyarosaurus, based on the conclu-

sions of Botfalvai et al. (2021).

Our updated review of the Transylvanian (and espe-

cially, the Hat

,eg Basin) titanosaur record provides an

opportunity to address the scenario put forward by Vila

et al. (2022) in more depth. In particular, our survey

does not support Botfalvai et al.’s(2021) suggestion of

a somewhat later date for the first occurrence of large-

bodied titanosaurs compared to smaller-sized taxa in the

Transylvanian area, and specifically in the Hat

,eg Basin.

Botfalvai et al.’s(2021) assertion was based on the pres-

ence of large-sized forms (here identified as Uriash)

recorded at Kadi

c’s locality VI at a marginally higher

stratigraphical position compared to those of smaller-

sized titanosaurs discovered at Kadic’s localities I

(Assemblage A, the lectotype of Magyarosaurus) and II

(Individual K), and at site K2 (Individual V) (see more

details in Botfalvai et al., 2021). However, current evi-

dence suggests that unlike Ibero-Armorica, where titano-

saurs were already both common and taxonomically

diverse by the late Campanian (e.g. D

ıez D



ıaz, 2022),

members of this clade appeared on the Transylvanian

landmass only later, probably after the Campanian/

Maastrichtian boundary (e.g. B

alc et al., 2024; Vremir

et al., 2014). Furthermore, small- and large-bodied

Transylvanian titanosaurs were penecontemporaneous

members of the same early Maastrichtian faunas, with

Magyarosaurus and Uriash being approximately sym-

patric. Admittedly, Vila et al.’s(2022) hypothesis does

allow for limited spatiotemporal overlap between the

stratigraphically older dwarf and younger invading

large-bodied titanosaurs in the Ibero-Armorican area, so

Revision of Romanian sauropod dinosaurs 109

this Magyarosaurus-Uriash sympatry by itself does not

falsify the replacement scenario for the Hat

,eg region.

However, these taxa appear to be replaced by

Paludititan during the middle part of the Maastrichtian,

so it must be acknowledged that the stratigraphically

youngest of the currently known named Transylvanian

taxa was relatively small.

More importantly, there seems to be no indication for

a clear-cut body size-related taxon replacement in the

Transylvanian assemblages when the entire titanosaur

fossil record is considered (i.e. irrespective of whether

or not specimens have been referred to a named taxon).

Indeed, our review of the Hat

,eg Basin material has

revealed the presence of (admittedly, rare) titanosaur

remains belonging to large-sized individuals in localities

that are penecontemporaneous with those yielding the

stratigraphically earliest small-sized titanosaurs. Among

these remains is a relatively large anterior caudal cen-

trum from Densus

,(LPB [FGGUB] R.1787; Fig. 66).

Unfortunately, there are no precise locality data for

this specimen, collected in the first half of the twentieth

century (based on the oldest available label associated

with it). Nevertheless, continental beds cropping out

near Densus

,village have been referred either to the

lower or to the basal-most part of the middle member of

the Densus

,-Ciula Formation (Csiki-Sava et al., 2016;

Grigorescu, 1992; Laufer, 1925; Nopcsa, 1905), suggest-

ing that its age is earliest Maastrichtian, possibly

approximately coeval with Magyarosaurus dacus.A

second example of a relatively early occurrence of a

large-sized titanosaur is the scapular blade NHMUK

R.11144 (Fig. 66) that apparently comes from the same

locality as the (probably) small-sized Individual K (site

II of Kadi

c, 1916; see the ‘Key localities and skeletal

associations’section), and from approximately the same

place and stratigraphical level as the small-sized

Individual V from site K2 reported by Botfalvai et al.

(2021). Based on these occurrences, it thus appears that

both small taxa (such as the dwarfed Magyarosaurus)

and large-bodied titanosaurs made their appearance con-

temporaneously in the Hat

,eg Basin near the beginning

of the early Maastrichtian. Subsequently, as noted by

Botfalvai et al. (2021), small- and large-bodied forms

continued to coexist until the latest Maastrichtian in the

south-western Transylvanian Basin (e.g. Csiki &

Vremir, 2011; Csiki-Sava et al., 2012). In conclusion,

contrary to previous suggestions (Vila et al., 2022),

there is no indication of a body-size related titanosaur

turnover, involving replacement of dwarfed taxa with

larger-sized ones, in the uppermost Cretaceous of the

Transylvanian area.

Moreover, a brief survey of the entire dinosaur fossil

record from the Transylvanian landmass fails to identify

any significant higher-level faunal turnover similar to

that reported in Ibero-Armorica during the

Maastrichtian. The Transylvanian faunas were generally

dominated by co-occurring titanosaurs and rhabdodontid

ornithopods throughout the Maastrichtian, whereas

Figure 66. Lithostrotia incertae sedis, large titanosaurian remains from the Maastrichtian of the Hat

,eg Basin. Anterior caudal

centrum from Densus

,(LPB [FGGUB] R.1787) in A, right lateral; B, anterior; C, left lateral; D, posterior; E, dorsal; and F, ventral

views. Left scapular blade NHMUK R.11144 from Assemblage K in G, lateral; H, medial; and I, dorsal views; J, detail of the label.

Scale bar equals 50 mm in A–F and 100 mm in G–I.

110 V. D



ıez D

ıaz et al.

hadrosauroids, although present, constituted less impor-

tant components (e.g. Csiki, Grigorescu, et al., 2010;

Csiki-Sava et al., 2015,2016; Vremir et al., 2015), espe-

cially when compared to their relative abundance and

diversity in post-MDT Ibero-Armorican faunas (e.g.

Blanco et al., 2015; Cruzado-Caballero et al., 2014;

Fondevilla et al., 2019; Marmi et al., 2016;P



erez-Pueyo

et al., 2021). Moreover, rhabdodontids, hadrosauroids

and titanosaurs were already members of the same local

assemblages by the early part of the early Maastrichtian

(e.g. Botfalvai et al., 2021; Csiki-Sava et al., 2016;

Nopcsa, 1915), and this clade-level faunal composition

seems to have continued largely uninterrupted into the

late Maastrichtian (e.g. Csiki-Sava et al., 2016; Smith

et al., 2002; Vremir et al., 2015), including the survival

of the otherwise rare nodosaurids (e.g. Codrea et al.,

2002; Vasile et al., 2011), again in stark contrast with

patterns reported from the Ibero-Armorican landmass

(e.g. Fondevilla et al., 2019;P



erez-Pueyo et al., 2021;

Vila et al., 2016). In fact, in some of the stratigraphic-

ally highest-occurring faunal assemblages from the

south-western Transylvanian (e.g. Vremir et al., 2015)

and Hat

,eg (Brusatte et al., 2017; Csiki-Sava et al.,

2016; Smith et al., 2002) basins, hadrosauroids are

either absent or very rare, whereas there are numerous

titanosaur remains, often associated with rhabdodontids.

Thus, it is currently unclear whether any dwarf to

large-bodied titanosaur transition, and the MDT more

generally, occurred either in the Hat

,eg region or on the

Transylvanian landmass overall –in fact, all current evi-

dence points to the contrary. Nevertheless, it is entirely

conceivable that lower-level compositional changes did

take place during the Maastrichtian, such as the apparent

replacement of the early Maastrichtian Magyarosaurus

by the ‘middle’Maastrichtian Paludititan (Fig. 8).

Conclusion

Since the first Late Cretaceous titanosaurian sauropod

dinosaur remains were found in Transylvania, and espe-

cially in the Hat

,eg Basin, towards the end of the nineteenth

century, an extensive number of new sites, isolated speci-

mens and partial skeletons have been found and described,

greatly advancing our understanding of these faunas.

However, a detailed study of these sites and skeletal

remains as a whole has never been carried out, with a num-

ber of taxonomic problems preventing any understanding

of the diversity and phylogenetic affinities of the Hat

,eg

Basin titanosaurs. Following our systematic revision, we

now recognize a minimum of four titanosaurian taxa:

Magyarosaurus dacus,Paludititan nalatzensis,

Petrustitan hungaricus n. gen. and Uriash kadici n. gen. n.

sp. This diversity was likely even higher, as evidenced by

the substantial amount of associated material that cannot

currently be adequately diagnosed as a new taxon, and yet

display differences from these four named taxa. Our phylo-

genetic analyses show that these Transylvanian titanosaurs

present particularly close relationships with Gondwanan

taxa: Magyarosaurus is recovered either as a member or a

close relative of Saltasauridae; Paludititan has affinities

with Lognkosauria, along with the approximately contem-

poraneous Spanish titanosaur Lohuecotitan;Petrustitan is

most closely related to South American early diverging

eutitanosaurian taxa; and Uriash shares one feature

uniquely with Gondwanan titanosaurs. These analyses also

strengthen the palaeobiogeographical hypothesis that the

latest Cretaceous European titanosaurs were members of

Gondwanan lineages that invaded the former area during

the Barremian–Albian. Since its initial discovery,

Magyarosaurus dacus has been identified as a dwarfed

sauropod, with island dwarfism proposed as an explanation

for the diminutive size of this species and other dinosaurs

on Hat

,eg Island. Whereas Paludititan and Petrustitan are

also small-bodied sauropods, Uriash was an order of mag-

nitude heavier and represents one of the largest titanosau-

rian taxa found in the Late Cretaceous of Europe. We

interpret the presence of this body-size disparity as either

evidence that large-bodied taxa were ecologically

excluded from body-size reduction by competition from

small-bodied titanosaurs, or that dwarfing occurred strati-

graphically earlier among several lineages and the small-

bodied titanosaurs on Hat

,eg Island are the descendants of

existing dwarfed ancestors. By contrast with some previ-

ous studies, we find no indication of a body-size related

titanosaur turnover, involving replacement of dwarfed taxa

with larger-sized ones, in the uppermost Cretaceous of the

Transylvanian area.

Acknowledgements

This project was funded by the Jurassic Foundation

(VDD), a Royal Society International Exchange award

(IESnR1n180088; PDM and ZCs-S) and a Royal

Society University Research Fellowship (UF160216,

URFnRn221010; PDM); it was also supported by the

Romanian Ministry of Research, Innovation and

Digitalization CNCS-UEFISCDI project PN-III-P4-ID-

PCE-2020-2570, within PNCDI III and the Hungarian

National Research, Development and Innovation Office

project NKFIH OTKA FK 146097 (ZCsS). We would

like to thank S. Maidment, P. Barrett and S. Chapman

(Natural History Museum London, UK), L. Kordos and

L. Mak

adi (Budapest Mining and Geological Survey of

Hungary), V. Codrea (Babes¸-Bolyai University of Cluj-

Napoca, Romania), C.-M. Jianu and S. Burnaz (Muzeul

Revision of Romanian sauropod dinosaurs 111

Civilizatiei Dacice si Romane, Deva, Romania), R.

Costa da Silva (Museu de Ci^

encias da Terra, Servic¸o

Geol

ogico do Brasil), L. Filippi (Museo Municipal

‘Argentino Urquiza’), and M. Reguero (Museo de la

Plata, Argentina) for access to the collections under

their care. We are also grateful to B. Vila and A. Sell

for providing information on Abditosaurus. We also

want to thank Scott Hartman for providing the

titanosaur skeletal outline used in figures. We are

grateful for the helpful comments provided by the editor

Zerina Johanson and the reviewers Alejandro Otero and

Julian Silva Junior, which have improved a first version

of this manuscript. We acknowledge the Willi Hennig

Society, which has sponsored the development and free

distribution of TNT.

Disclosure statement

No potential conflict of interest was reported by the

author(s).

Supplemental material

Supplemental material for this article can be accessed

online here: https://doi.org/10.1080/14772019.2024.

2441516.

ORCID

Ver

onica D

ıez D

ıaz http://orcid.org/0000-0002-9840-

9829

Philip D. Mannion http://orcid.org/0000-0002-9361-

6941

Zolt

an Csiki-Sava http://orcid.org/0000-0001-7144-

0327

Paul Upchurch http://orcid.org/0000-0002-8823-4164

References

Agnolin, F. L.,Gonz

alez Riga, B. J.,Aranciaga Rolando,

A. M.,Rozadilla, S.,Motta, M. J.,Chimento, N. R.,&

Novas, F. E. (2023). A new giant titanosaur (Dinosauria,

Sauropoda) from the Upper Cretaceous of Northwestern

Patagonia, Argentina. Cretaceous Research, 105487.

https://doi.org/10.1016/j.cretres.2023.105487

Antonescu, E.,Lupu, D.,&Lupu, M. (1983). Correlation

palinologique du Cr

etac

e terminal du sud-est des Monts

Metaliferi et des Depressions de Hat¸eg et de Rusca

Montan

a. Anuarul Institutului de Geologie S¸i Geofizic

59(3), 71–77.

Apesteguía, S. (2005). Evolution of the titanosaur metacarpus.

In V. Tidwell, & K. Carpenter (Eds.), Thunder lizards:

The Sauropodomorph dinosaurs (pp. 321–345). Indiana

University Press.

Astibia, H.,Buffetaut, E.,Buscalioni, A. D.,Cappetta, H.,

Corral, C.,Estes, R.,Garcia-Garmilla, F.,Jaeger, J. J.,

Jimenez-Fuentes, E.,Le Loeuff, J.,Mazin, J. M.,Orue-

Etxebarria, X.,Pereda-Suberbiola, J.,Powell, J. E.,

Rage, J. C.,Rodriguez-Lazaro, J.,Sanz, J. L.,&Tong,

H. (1990). The fossil vertebrates from La~

no (Basque

Country, Spain); new evidence on the composition and

affinities of the Late Cretaceous continental faunas of

Europe. Terra Nova,2(5), 460–466.

Augustin, F. J.,Bastiaans, D.,Dumbrav

a, M. D.,&Csiki-

Sava, Z. (2022). A new ornithopod dinosaur,

Transylvanosaurus platycephalus gen. et sp. nov.

(Ornithischia). from the Upper Cretaceous of the Hațeg

Basin, Romania. Journal of Vertebrate Paleontology,

42(2), e2133610. https://doi.org/10.1080/02724634.2022.

2133610

Augustin, F. J.,Matzke, A. T.,Csiki-Sava, Z.,&

Pfretzschner, H.-U. (2019). Bioerosion on vertebrate

remains from the Upper Cretaceous of the Hat¸eg Basin,

Romania and its taphonomic implications.

Palaeogeography, Palaeoclimatology, Palaeoecology,534,

109318.

Averianov, A.,&Efimov, V. (2018). The oldest titanosaurian

sauropod of the Northern Hemisphere. Biological

Communications,63(3), 145–162. https://doi.org/10.21638/

spbu03.2018.301

Averianov, A.,Podlesnov, A.,Slobodin, D.,Skutschas, P.,

Feofanova, O.,&Vladimirova, O. (2023). First sauropod

dinosaur remains from the Early Cretaceous Shestakovo 3

locality, Western Siberia, Russia. Biological

Communications,68(4), Article 4. https://doi.org/10.

21638/spbu03.2023.404

Averianov, A. O.,Sizov, A. V.,&Skutschas, P. P. (2021).

Gondwanan affinities of Tengrisaurus, Early Cretaceous

titanosaur from Transbaikalia, Russia (Dinosauria,

Sauropoda). Cretaceous Research,122, 104731.

B

alc, R.,Bindiu-Haitonic, R.,K€

ovecsi, S.-A.,Vremir, M.,

Ducea, M.,Csiki-Sava, Z.,T¸ab

ar

a, D.,&Vasile, S

(2024). Integrated biostratigraphy of Upper Cretaceous

deposits from an exceptional continental vertebrate-bearing

marine section (Transylvanian Basin, Romania) provides

new constraints on the advent of ‘dwarf dinosaur’faunas

in Eastern Europe. Marine Micropaleontology,187,

102328.

Baraboshkin, E. Y.,Alekseev, A. S.,&Kopaevich, L. F.

(2003). Cretaceous palaeogeography of the north-eastern

Peri-Tethys. Palaeogeography, Palaeoclimatology,

Palaeoecology,196(1–2), 177–208.

Beeston, S.,Poropat, S.,Mannion, P.,Pentland, A.,

Enchelmaier, M.,Sloan, T.,&Elliott, D. (2024).

Reappraisal of sauropod dinosaur diversity in the Upper

Cretaceous Winton Formation of Queensland, Australia,

through 3D digitisation and description of new specimens.

PeerJ,12, e17180, https://doi.org/10.7717/peerj.17180

Bellardini, F.,Coria, R. A.,Pino, D. A.,Windholz, G. J.,

Baiano, M. A.,&Martinelli, A. G. (2022). Osteology

and phylogenetic relationships of Ligabuesaurus leanzai

(Dinosauria: Sauropoda) from the Early Cretaceous of the

Neuqu

en Basin, Patagonia, Argentina. Zoological Journal

112 V. D



ıez D

ıaz et al.

of the Linnean Society.https://doi.org/10.1093/

ZOOLINNEAN/ZLAC003

Ben

ıtez-L

opez, A.,Santini, L.,Gallego-Zamorano, J.,Mil

B.,Walkden, P.,Huijbregts, M. A. J.,&Tobias, J. A.

(2021). The island rule explains consistent patterns of

body size evolution in terrestrial vertebrates. Nature

Ecology & Evolution,5(6), 768–786. https://doi.org/10.

1038/s41559-021-01426-y

Benson, R. B. J.,Hunt, G.,Carrano, M. T.,&Campione,

N. (2018). Cope’s rule and the adaptive landscape of

dinosaur body size evolution. Palaeontology,61(1), 13–

48. https://doi.org/10.1111/pala.12329

Benton, M. J.,Csiki, Z.,Grigorescu, D.,Redelstorff, R.,

Sander, P. M.,Stein, K.,&Weishampel, D. B. (2010).

Dinosaurs and the island rule: The dwarfed dinosaurs from

Hat¸eg Island. Palaeogeography, Palaeoclimatology,

Palaeoecology,293(3), 438–454. https://doi.org/10.1016/j.

palaeo.2010.01.026

Benton, M. J.,Forth, J.,&Langer, M. C. (2014). Models

for the rise of the dinosaurs. Current Biology, 24(2).

https://doi.org/10.1016/j.cub.2013.11.063

Blanco, A.,Prieto-M

arquez, A.,&De Esteban-Trivigno, S.

(2015). Diversity of hadrosauroid dinosaurs from the Late

Cretaceous Ibero-Armorican Island (European

Archipelago) assessed from dentary morphology.

Cretaceous Research,56, 447–457.

Bojar, A.-V.,Halas, S.,Bojar, H.-P.,Grigorescu, D.,&

Vasile, S. (2011). Upper Cretaceous volcanoclastic

deposits from the Hat¸eg Basin, South Carpathians

(Romania): K-Ar ages and intrabasinal correlation.

Geochronometria,38(2), 182–188. https://doi.org/10.2478/

s13386-011-0023-8

Bonaparte, J. F.,&Coria, R. A. (1993). A new and huge

titanosaur sauropod from the Rio Limay Formation

(Albian-Cenomanian) of Neuqu

en Province, Argentina.

Ameghiniana,30, 271–282.

Bonaparte, J. F.,Gonz

alez Riga, B. J.,&Apesteguia, S.

(2006). Ligabuesaurus leanzai gen. et sp. nov.

(Dinosauria, Sauropoda), a new titanosaur from the Lohan

Cura Formation (Aptian, Lower Cretaceous) of Neuqu

en,

Patagonia, Argentina. Cretaceous Research,27(3), 364–

376. https://doi.org/10.1016/j.cretres.2005.07.004

Borsuk-Białynicka, M. (1977). A new camarasaurid sauropod

Opisthocoelicaudia skarzynskii gen. n., sp. n. from the

Upper Cretaceous of Mongolia. Palaeontologia Polonica,

37(5), 5–64.

Botfalvai, G.,Csiki-Sava, Z.,Grigorescu, D.,&Vasile, S¸.

(2017). Taphonomical and palaeoecological investigation

of the Late Cretaceous (Maastrichtian) Tus¸tea vertebrate

assemblage (Romania; Hat¸eg Basin)-insights into a unique

dinosaur nesting locality. Palaeogeography,

Palaeoclimatology, Palaeoecology,468, 228–262.

Botfalvai, G.,Csiki-Sava, Z.,Kocsis, L.,Albert, G.,

Magyar, J.,Bodor, E. R.,T¸ab

ar

a, D.,Ulyanov, A.,&

Mak

adi, L. (2021). ‘X’marks the spot! Sedimentological,

geochemical and palaeontological investigations of Upper

Cretaceous (Maastrichtian) vertebrate fossil localities from

the V

alioara valley (Densus¸-Ciula Formation, Hat

,eg

Basin, Romania). Cretaceous Research,123, 104781.

https://doi.org/10.1016/j.cretres.2021.104781

Brusatte, S. L.,Dumbrav

a, M.,Vremir, M.,Csiki-Sava, Z.,

Totoianu, R.,&Norell, M. A. (2017). A catalog of

Zalmoxes (Dinosauria: Ornithopoda) specimens from the

Upper Cretaceous N

alat¸-Vad Locality, Hat¸eg Basin,

Romania. American Museum Novitates,2017(3884), 1–36.

Buffetaut, E. (1989). Archosaurian reptiles with Gondwanan

affinities in the Upper Cretaceous of Europe. Terra Nova,

1(1), 69–74. https://doi.org/10.1111/j.1365-3121.1989.

tb00328.x

Buffetaut, E.,Mechin, P.,&Mechin-Salessy, A. (1988). Un

dinosaure th

eropode d’affinit

es gondwaniennes dans le

Cr

etac

e sup

erieur de Provence. Comptes Rendus de

l’Acad

emie Des Sciences. S

erie 2, M

ecanique, Physique,

Chimie, Sciences de l’univers, Sciences de La Terre,

306(2), 153–158.

Calvo, J. O. (2014). New fossil remains of Futalognkosaurus

dukei (Sauropoda, titanosauria) from the Late Cretaceous

of Neuqu

en, Argentina. 4th International Palaeontological

Congress, “The History of Life: A View from the Southern

Hemisphere”, Abstract Volume, 325.

Calvo, J. O.,&Gonz

alez Riga, B. J. (2003). Rinconsaurus

caudamirus gen. et sp. nov., a new titanosaurid

(Dinosauria, Sauropoda) from the Late Cretaceous of

Patagonia, Argentina. Revista Geol

ogica de Chile,30(2),

333–353. https://doi.org/10.4067/S0716-

02082003000200011

Calvo, J. O.,Gonz

alez Riga, B. J.,&Porfiri, J. D. (2007).

A new titanosaur sauropod from the Late Cretaceous of

Neuqu

en, Patagonia, Argentina. Arquivos Do Museu

Nacional,65(4), 485–504.

Calvo, J. O.,Porfiri, J. D.,Riga, B. J. G.,&Kellner,

A. W. A. (2007). Anatomy of Futalognkosaurus dukei

Calvo, Porfiri, Gonz

alez Riga & Kellner, 2007

(Dinosauria, Titanosauridae) from the Neuqu

en Group

(Late Cretaceous), Patagonia, Argentina. Archivos Do

Museu Nacional Do Rio de Janeiro,65(4), 511–526.

Calvo, J. O.,&Salgado, L. (1995). Rebbachisaurus tessonei

sp. nov. A new Sauropoda from the Albian-Cenomanian

of Argentina; new evidence on the origin of the

Diplodocidae. Gaia,11,13–33.

Campione, N. E.,&Evans, D. C. (2012). A universal scaling

relationship between body mass and proximal limb bone

dimensions in quadrupedal terrestrial tetrapods. BMC

Biology,10, 60. https://doi.org/10.1186/1741-7007-10-60

Campione, N. E.,Evans, D. C.,Brown, C. M.,&Carrano,

M. T. (2014). Body mass estimation in non-avian bipeds

using a theoretical conversion to quadruped stylopodial

proportions. Methods in Ecology and Evolution,5(9), 913–

923.

Campos, D.,de A.,Kellner, A. W.,Bertini, R. J.,&

Santucci, R. M. (2005). On a titanosaurid (Dinosauria,

Sauropoda) vertebral column from the Bauru group, Late

Cretaceous of Brazil. Arquivos Do Museu Nacional,63(3),

565–593.

Canudo, J. I.,Royo-Torres, R.,&Cuenca-Besc

os, G.

(2008). A new sauropod: Tastavinsaurus sanzi gen. et sp.

nov. from the Early Cretaceous (Aptian) of Spain. Journal

of Vertebrate Paleontology,28(3), 712–731. https://doi.

org/10.1671/0272-4634(2008)28[712:ANSTSG]2.0.CO;2

Carballido, J. L.,Marpmann, J. S.,Schwarz-Wings, D.,&

Pabst, B. (2012). New information on a juvenile sauropod

specimen from the Morrison Formation and the

reassessment of its systematic position. Palaeontology,

55(3), 567–582. https://doi.org/10.1111/j.1475-4983.2012.

01139.x

Carballido, J. L.,Otero, A.,Mannion, P. D.,Salgado, L.,&

Moreno, A. P. (2022). Titanosauria: a critical reappraisal

Revision of Romanian sauropod dinosaurs 113

of its systematics and the relevance of the South American

record. In A. Otero, J. L. Carballido, & D. Pol (Eds.),

South American Sauropodomorph Dinosaurs: Record,

Diversity and Evolution (pp. 269–298). Springer

International Publishing. https://doi.org/10.1007/978-3-

030-95959-3_8

Carballido, J. L.,Pol, D.,Otero, A.,Cerda, I. A.,Salgado,

L.,Garrido, A. C.,Ramezani, J.,C

uneo, N. R.,&

Krause, J. M. (2017). A new giant titanosaur sheds light

on body mass evolution among sauropod dinosaurs.

Proceedings of the Royal Society B: Biological Sciences,

284(1860), 20171219. https://doi.org/10.1098/rspb.2017.

1219

Cerda, I. A.,Paulina Carabajal, A.,Salgado, L.,Coria,

R. A.,Reguero, M. A.,Tambussi, C. P.,&Moly, J. J.

(2011). The first record of a sauropod dinosaur from

Antarctica. Die Naturwissenschaften,99(1), 83–87. https://

doi.org/10.1007/s00114-011-0869-x

Cerda, I. A.,Salgado, L.,&Powell, J. E. (2012). Extreme

postcranial pneumaticity in sauropod dinosaurs from South

America. Palaontologische Zeitschrift,86(4), 441–449.

https://doi.org/10.1007/s12542-012-0140-6

Cerda, I.,Zurriaguz, V. L.,Carballido, J. L.,Gonz

alez, R.,

&Salgado, L. (2021). Osteology, paleohistology and

phylogenetic relationships of Pellegrinisaurus powelli

(Dinosauria: Sauropoda) from the Upper Cretaceous of

Argentinean Patagonia. Cretaceous Research,128,

104957. https://doi.org/10.1016/j.cretres.2021.104957

Chiappe, L. M.,Coria, R. A.,Dingus, L.,Jackson, F.,

Chinsamy, A.,&Fox, M. (1998). Sauropod dinosaur

embryos from the Late Cretaceous of Patagonia. Nature,

396(6708), 258–261. https://doi.org/10.1038/24370

Chiappe, L. M.,Salgado, L.,&Coria, R. A. (2001).

Embryonic skulls of titanosaur sauropod dinosaurs.

Science,293(September), 2444–2446. https://doi.org/10.

1126/science.1063723

Chiarenza, A. A.,Fabbri, M.,Consorti, L.,Muscioni, M.,

Evans, D. C.,Cantalapiedra, J. L.,&Fanti, F. (2021).

An Italian dinosaur Lagerst€

atte reveals the tempo and

mode of hadrosauriform body size evolution. Scientific

Reports,11(1), 23295. https://doi.org/10.1038/S41598-021-

02490-X

Chure, D.,Britt, B. B.,Whitlock, J. A.,&Wilson, J. A.

(2010). First complete sauropod dinosaur skull from the

Cretaceous of the Americas and the evolution of sauropod

dentition. Die Naturwissenschaften,97(4), 379–391.

https://doi.org/10.1007/s00114-010-0650-6

Codrea, V. A.,&Godefroit, P. (2008). New Late Cretaceous

dinosaur findings from northwestern Transylvania

(Romania). Comptes Rendus Palevol,7(5), 289–295.

Codrea, V.,Godefroit, P.,&Smith, T. (2012). First

discovery of Maastrichtian (latest Cretaceous) terrestrial

vertebrates in Rusca Montan

a Basin (Romania). In P.

Godefroit (Ed.), Bernissart Dinosaurs and Early

Cretaceous Terrestrial Ecosystems (pp. 570–581). Indiana

University Press.

Codrea, V.,&M

arginean, R. (2007). A catalogue of fossil

vertebrates from Aiud Natural Sciences Museum. Oltenia,

Studii S¸i Comunic

ari,S¸tiint¸e Naturale,23, 177–186.

Codrea, V.,Murzea-Jipa, C.,&Venczel, M. (2008). A

sauropod vertebra at R^

apa Ros¸ie (Alba district). Acta

Palaeontologica Romaniae,6,43–48.

Codrea, V.,Smith, T.,Dica, P.,Folie, A.,Garcia, G.,

Godefroit, P.,&Van Itterbeeck, J. (2002). Dinosaur egg

nests, mammals and other vertebrates from a new

Maastrichtian site of the Hat¸eg Basin (Romania). Comptes

Rendus Palevol,1(3), 173–180.

Codrea, V. A.,&Solomon, A. (2012). Peculiar fossilization

and taphonomy in Maastrichtian terrestrial deposits of Pui

(Hat¸eg Basin, Romania). Studii S¸i Cercet

ari (Geologie-

Geografie),17,51–69.

Codrea, V.,Vremir, M.,Jipa, C.,Godefroit, P.,Csiki, Z.,

Smith, T.,&F

arcas¸, C. (2010). More than just Nopcsa’s

Transylvanian dinosaurs: A look outside the Hat¸eg Basin.

Palaeogeography, Palaeoclimatology, Palaeoecology,

293(3–4), 391–405.

Coria, R. A.,Filippi, L. S.,Chiappe, L. M.,Garc

ıa, R.,&

Arcucci, A. B. (2013). Overosaurus paradasorum gen. et

sp. nov., a new sauropod dinosaur (Titanosauria:

Lithostrotia) from the Late Cretaceous of Neuqu

en,

Patagonia, Argentina. Zootaxa,3683(4), 357–376. https://

doi.org/10.11646/zootaxa.3683.4.2

Cruzado-Caballero, P.,Ruiz-Ome~

naca, J. I.,Gaete, R.,

Riera, V.,Oms, O.,&Canudo, J. I. (2014). A new

hadrosaurid dentary from the latest Maastrichtian of the

Pyrenees (north Spain) and the high diversity of the duck-

billed dinosaurs of the Ibero-Armorican Realm at the very

end of the Cretaceous. Historical Biology,26(5), 619–630.

https://doi.org/10.1080/08912963.2013.822867

Csiki, Z. (1997). Leg

aturi paleobiogeografice ale faunei de

vertebrate continentale Maastrichtian superioare din

Bazinul Hat

¸eg. Nimphaea,23,45–68.

Csiki, Z. (1999). New evidence of armoured titanosaurids in

the Late Cretaceous. Magyarosaurus dacus from the

Hat

,eg Basin (Romania). Oryctos,2,93–99.

Csiki, Z. (2005). Systematics, taphonomy and paleoecology of

the Maastrichtian microvertebrates and saurischian

dinosaurs from the Hat¸eg Basin. [Unpublished PhD

thesis]. University of Bucharest.

Csiki, Z. (2006). Insect borings in dinosaur bones from the

Maastrichtian of the Hat¸eg Basin, Romania –

Paleoecological and paleoclimatic implications. In Z. Csiki

(Ed.), Mesozoic and Cenozoic Vertebrates and

Paleoenvironments. Tributes to the Career of Dan

Grigorescu (pp. 95–104). Ars Docendi.

Csiki, Z.,Barnes, R. N.,&Upchurch, P. (2011). The

revision of the genus Magyarosaurus, and titanosaur

diversity in the Maastrichtian of the Hat

,eg Basin,

Romania –preliminary results. In R. Forrest (Ed.),

Abstracts of Presentations, 59th Annual Symposium of

Vertebrate Palaeontology and Comparative Anatomy (p.

10), Lyme Regis, UK.

Csiki, Z.,Codrea, V.,Jipa-Murzea, C.,&Godefroit, P.

(2010). A partial titanosaur (Sauropoda, Dinosauria)

skeleton from the Maastrichtian of N

alat¸-Vad, Hat¸eg

Basin, Romania. Neues Jahrbuch Fur Geologie Und

Palaontologie –Abhandlungen,258(3), 297–324. https://

doi.org/10.1127/0077-7749/2010/0098

Csiki, Z.,&Grigorescu, D. (2004). Maastrichtian sauropods

of the Hat

,eg Basin, Romania. Program of the 2nd

Meeting of EAVP (p. 10), Brno. https://doc.rero.ch/record/

16292/files/PAL_E568.pdf#page=10

Csiki, Z.,&Grigorescu, D. (2006). The revision of

“Magyarosaurus”hungaricus Huene (Dinosauria:

Sauropoda) from the Hat¸eg Basin. Extended Abstracts, 8th

Mining, Metallurgy and Geology Conference, EMT,65,

70.

114 V. D



ıez D

ıaz et al.

Csiki, Z.,&Grigorescu, D. (2007). The “Dinosaur Island”–

new interpretation of the Hat¸eg Basin vertebrate fauna

after 110 years. Sargetia, 20,5–26.

Csiki, Z.,Grigorescu, D.,Codrea, V.,&Therrien, F.

(2010). Taphonomic modes in the Maastrichtian

continental deposits of the Hat¸eg Basin, Romania –

Palaeoecological and palaeobiological inferences.

Palaeogeography, Palaeoclimatology, Palaeoecology,

293(3–4), 375–390. https://doi.org/10.1016/j.palaeo.2009.

10.013

Csiki, Z.,Grigorescu, D.,&Weishampel, D. B. (2007). A

new titanosaur sauropod (Dinosauria: Saurischia) from the

Upper Cretaceous of the Hat¸eg Basin (Romania). In J. Le

Loeuff (Ed.), Abstracts Volume, 5th Annual Meeting of the

European Association of Vertebrate Palaeontologists

(EAVP) (p. 17), Carcassone, France.

Csiki, Z.,Ionescu, A.,&Grigorescu, D. (2008). The

Budurone microvertebrate site from the Maastrichtian of

the Hat¸eg Basin-flora, fauna, taphonomy and

paleoenvironment. Acta Palaeontologica Romaniae,6,49–

66.

Csiki, Z.,&Vremir, M. (2011). A large-sized (?)Late

Maastrichtian titanosaur from R^

apa Ros¸ie, Sebes¸. In Z.

Csiki (Ed.), Abstract Book, 8th Romanian Symposium of

Paleontology, Bucharest (Romania) (pp. 28–29). Ars

Docendi.

Csiki, Z.,Vremir, M.,Brusatte, S. L.,&Norell, M. A.

(2010). An aberrant island-dwelling theropod dinosaur

from the Late Cretaceous of Romania. Proceedings of the

National Academy of Sciences,107(35), 15357–15361.

https://doi.org/10.1073/pnas.1006970107

Csiki-Sava, Z.,Buffetaut, E.,}

Osi, A.,Pereda-Suberbiola,

X.,&Brusatte, S. L. (2015). Island life in the Cretaceous

–Faunal composition, biogeography, evolution, and

extinction of land-living vertebrates on the Late

Cretaceous European archipelago. ZooKeys 469,1–161.

https://doi.org/10.3897/zookeys.469.8439

Csiki-Sava, Z.,Butiseac

a, G.,Grigorescu, D.,Dumbrav

M.,&Vasile, S¸. (2012). “A new piece to the puzzle”–

titanosaur skeletal remains from the Tus¸tea dinosaur

nesting site and its surroundings, Hat¸eg Basin, Romania.

In R. Royo-Torres, F. Gasc

o & L. Alcal

a (Coord.), 10th

Annual Meeting of the European Association of Vertebrate

Palaeontologists. <Fundamental! 20,49–51.

Csiki-Sava, Z.,Vasile, S

,.,Grigorescu, D.,&Vremir, M.

(2018). Mind the gap! Significance of a new latest

Cretaceous fossiliferous site in the northern Hat

,eg Basin,

Romania. In M. Marzola, O. Mateus, & M. Moreno-

Azanza (Eds.), Abstract Book of the XVI Annual Meeting

of the European Association of Vertebrate Palaeontology

(p. 53). Faculdade de Ci^

encias e Tecnologia da

Universidade Nova de Lisboa.

Csiki-Sava, Z.,Vremir, M.,Vasile, S¸.,Brusatte, S. L.,

Dyke, G.,Naish, D.,Norell, M. A.,&Totoianu, R.

(2016). The East Side Story –The Transylvanian latest

Cretaceous continental vertebrate record and its

implications for understanding Cretaceous-Paleogene

boundary events. Cretaceous Research,57, 662–698.

https://doi.org/10.1016/j.cretres.2015.09.003

Cucchi, T.,Barnett, R.,Mart

ınkov

a, N.,Renaud, S.,

Renvois

e, E.,Evin, A.,Sheridan, A.,Mainland, I.,

Wickham-Jones, C.,&Tougard, C. (2014). The

changing pace of insular life: 5000 years of

microevolution in the Orkney vole (Microtus arvalis

orcadensis). Evolution,68(10), 2804–2820.

Curry Rogers, K. (2009). The postcranial osteology of

Rapetosaurus krausei (Sauropoda: Titanosauria) from the

Late Cretaceous of Madagascar. Journal of Vertebrate

Paleontology,29(4), 1046–1086.

Curry Rogers, K. A. (2005). Titanosauria: a phylogenetic

overview. In K. A. Curry Rogers & J. A. Wilson (Eds.),

The sauropods: Evolution and paleobiology (pp. 50–103),

University of California Press.

Curry Rogers, K. A.,&Wilson, J. A. (2014). Vahiny

depereti, gen. et sp. nov., a new titanosaur (Dinosauria,

Sauropoda) from the Upper Cretaceous Maevarano

Formation, Madagascar. Journal of Vertebrate

Paleontology,34(3), 606–617. https://doi.org/10.1080/

02724634.2013.822874

Dal Sasso, C.,Pierangelini, G.,Famiani, F.,Cau, A.,&

Nicosia, U. (2016). First sauropod bones from Italy offer

new insights on the radiation of Titanosauria between

Africa and Europe. Cretaceous Research,64,88–109.

https://doi.org/10.1016/j.cretres.2016.03.008

D’Emic, M. D. (