First Dinosaurs from Saudi Arabia

Dinosaur remains from the Arabian subcontinent are exceedingly rare, and those that have been documented manifest indeterminate affinities. Consequently the discovery of a small, but diagnostic, accumulation of elements from Campanian-Maastrichtian (∼75 Ma) deposits in northwestern Saudi Arabia is significant because it constitutes the first taxonomically identifiable dinosaur material described from the Arabian Peninsula. The fossils include a series of possible lithostrotian titanosaur caudal vertebrae, and some isolated theropod marginal teeth that share unique character states and metric parameters (analyzed using multivariate statistical methods) with derived abelisaurids – this is the first justifiable example of a non-avian carnivorous dinosaur clade from Arabia. The recognition of titanosaurians and abelisaurids from Saudi Arabia extends the palaeogeographical range of these groups along the entire northern Gondwanan margin during the latest Cretaceous. Moreover, given the extreme paucity of coeval occurrences elsewhere, the Saudi Arabian fossils provide a tantalizing glimpse into dinosaurian assemblage diversity within the region.

Virtually nothing has been reported on dinosaurs from Saudi Arabia. Hughes and Johnson ( [13] p. 59, Fig. 11 and in text on p. 60) briefly mentioned a confidential Saudi Aramco report (''Milner, A., N. Morris and P. Jeffery. 1993. Report on Macrofossils from the Kingdom of Saudi Arabia. Natural History Museum, London, Confidential report for Saudi Aramco'') that identified bone fragments of a ''sauropod dinosaur, possibly a titanosaurid'' from the Adaffa Formation, an Upper Cretaceous unit that crops out in the Midyan Peninsula region along the northeastern coast of the Red Sea (Fig. 1). Grainger ([14] p. 153) also noted some additional ''tentatively confirmed'' dinosaur bones, together with other vertebrate remains, collected from the Adaffa Formation in 2004-2008 by a joint team from the Saudi Geological Survey (SGS) and Egyptian Geological Museum. Subsequent appraisal of this material by Kear et al. [15,16] documented a primarily marine fauna incorporating: indeterminate anacoracid? sharks; actinopterygians 2 lepisosteids, pycnodontiforms, pachycormids (cf. Protosphyraena) and teleosts (cf. Enchodus sp.); ceratodont lungfish (Ceratodus sp.); bothremydid turtles; dyrosaurid crocodilians; an elasmosaurid plesiosaur; plioplatecarpine mosasaurs and the widespread mosasaurine Prognathodon; as well as a small aquatic varanoid (cf. Pachyvaranus). A few dinosaur bones and teeth were also recovered from the deposit and are presented in this paper. These fossils are important because they represent the first definitive dinosaurian remains described from the Kingdom of Saudi Arabia.
No permits were required for the described study, which complied with all relevant regulations.

Geological Setting
The Adaffa Formation is a thick sequence of cross-bedded, quartz arenite sandstones with basal conglomerates, and uppermost thin marl, siltstone, and fine-grained sandstone layers. It forms the lower-most unit of the Late Cretaceous to Paleogene Suqah Group, a series of pre-Red Sea rift strata that unconformably overly Paleozoic basement rocks in the Usfan region (Jeddah Basin) near Jeddah, and northwards into the Midyan region (Ifal Basin) of far northwestern Saudi Arabia [13]. Detailed stratigraphic assessments [13,17] and palynological studies indicate an early Campanian2early Maastrichtian age [18].
The Adaffa Formation vertebrate macrofossils occur in two restricted graben structures, the Aynunah and Sharmah troughs. The individual elements are disarticulated and randomly distributed through thin limonitic beds near the top of the main sandstone sequence. The bone surfaces exhibit extensive surface abrasion (e.g. edge rounding and decortication) consistent with damage by wave action and/or currents prior to burial. This accords with the inferred supratidal marine to fluviatile depositional settings, with coarse clastic input from braided river outlets fed by periodic flash floods [18].

Results and Discussion
Seven caudal vertebrae from a sauropod (SGS 0188, SGS 0213, SGS 0342, SGS 0366, SGS 0422, plus two additional unregistered fragments), and two theropod marginal teeth (SGS 0061, SGS 0090), were recovered during an exhaustive excavation of a small limonitic sandstone exposure (,10 m 2 ) within the Aynunah Trough, about 11 km northeast of Al Khuraybah ([ Fig. 1]; see Hughes and Johnson ([13] p. 60, Fig. 12 for a photograph of the site). These elements were found intermixed with other vertebrate remnants and numerous wood fragments, presumably sorted by turbulent water action. There was no obvious association between individual skeletal components; although, compatible anatomical positioning, size, ontogentic stage, and taxonomic affinities of at least the dinosaur material suggests derivation from single animals. All of the specimens were accessioned into the Paleontological Collection of the Saudi Geological Survey, Jeddah, Kingdom of Saudi Arabia.
The sauropod vertebrae appear to form a continuous series from the posterior-distal caudal region. Unfortunately, most of the bones are badly weathered and comprise only broken parts of the centra. However, one specimen (SGS 0366 [ Fig. 2A2C]) is relatively complete and retains a neural arch. Dimensions of SGS 0366 are: centrum length = 105 mm; centrum width across the anterior articular surface = 66 mm; lateral height of the anterior articular surface = 56.5 mm; maximum vertebral height including neural arch = 133 mm. The recovered centra are all cylindrical in outline and clearly procoelous, a classic feature of titanosaurians [19]. Where discernible, the ventral surface is flat and exhibits raised areas on both the anterior and posterior ends for accommodation of the chevron facets. The neural arch is anteriorly positioned (compatible with titanosauriforms [20,21]) and the prezygapophyses, although broken and heavily weathered, would have projected anterodorsally. The spinoprezygapophyseal laminae are not fused and the prespinal lamina is present as a low, near horizontal ridge. The neural spine is elongate and posteriorly inclined reminiscent of titanosaurian taxa such as Isisaurus and Neuquensaurus [22]. The postzygaphophyses are weakly delineated. . The gently convex mesial face is rounded and lacks a carina. Conversely, the distal profile is clearly linear, consistent with the distinctive maxillary-dentary teeth of abelisaurid ceratosaurians [23]. The distal carina is straight and comprises labio-lingually broad denticles (average density [DAVG] = 10.9/5 mm) with basally angled interdental sulci (resembling derived carnotaurines [23,24]). Both the labial and lingual tooth surfaces bear apically converging longitudinal ridges, which are more pronounced distally and occasionally branch towards the base [ Fig. 2F]. Similar vertical enamel ridging or fluting has been reported in dromaeosaurids [25], spinosaurids [26], and ceratosaurians [27] including the abelisauroid Masiakasaurus [28].

Analysis
The Adaffa Formation dinosaur remains are fragmentary but can be unambiguously referred to typical Late Cretaceous Gondwanan lineages based on discrete phylogenetic character states: Titanosauria [29] and Lithostrotia [21], diagnosed by the presence of procoelous caudal centra; and Abelisauridae, characterized by a centrally positioned tooth apex with strongly curved mesial, and straight distal profiles [30]. Baso-apically trending ridges and interdental sulci can also be variably developed in derived forms [23]. However, because some qualitative theropod tooth characters are known to be phylogenetically ambiguous [31], we conducted a series of morphometric analyses to corroborate our hypothesized affinity for the Adaffa Formation specimens, and to test their proportional similarities relative to other non-avian theropods. Measurements of SGS 0090 and SGS 0061 [ Table 1] were added to the most taxon-rich matrix of theropod dental metrics obtainable from the literature [32] with taxonomic modifications introduced by Smith and Lamanna [24]. Unfortunately, a compatible metric data set was not available for sauropod postcranial elements, preventing quantitative evaluation of affinities; primary compilation of such information from original fossils and/or the literature was also beyond the scope of this paper.
We log 10 -transformed (see rationalization in Samman et al. [33] and references therein) the theropod tooth measurements and subjected them to a series of multivariate statistical analyses designed to best accommodate our small sample size and its  Parameters follow Smith et al. [32]. Abbreviations: CBL, crown base mesial-distal length; CBW, crown base labio-lingual width; CH, crown height from apex to distal enamel base; AL, apical length from medial enamel base; CBR, crown base ratio CBW/CBL; CHR, crown height ratio CH/CBL; CA, crown angle from mesial base to crown; DA, number of denticles/5 mm ( = density) at the apical section of the distal carina; DC, distal denticle density at the mid-crown; DB, distal denticle density at the crown base; DAVG, average distal denticle density/5 mm. doi:10.1371/journal.pone.0084041.t001 (2) a parametric Multivariate analysis of variance (MANOVA) and Canonical variates analysis (CVA) utilizing the total specimen data set organized into 10 phylogenetically defined [34,35,36,37,38] family-level clades [groupings listed in Fig. 3

]; (3) Euclidean and
Neighbor-joining cluster analyses of all parameters averaged over these same families; and (4), a non-parametric one-way MAN-OVA (NPMANOVA), coupled with (5) a non-parametric one-way Analysis of similarities (ANOSIM), both manipulating the total data set divided into 14 stratigraphic source unit categories (collated from [39]). All calculations were conducted in PAST [40].

(3) Cluster Analyses
We also employed a series of bootstrapped (1000 replicates) linear Euclidean cluster analyses to visualize the placement of SGS 0090 + SGS 0061 against a distance matrix of values averaged (so as to minimize sensitivity inherent in uneven sample sizes) over non-avian theropod family-level clades [see legend in Fig. 3].

(4) ANOSIM
As a cross-correlation for the NPMANOVA results, we also calculated an ANOSIM (using Bray-Curtis and Euclidean distance measures) to compare distances both within and between our designated stratigraphic groupings. This derived significant distance (p,0.05) between SGS 0090 + SGS 0061 and Liliensternus,

Conclusions
Despite being fragmentary, the Adaffa Formation dinosaur remains are justifiably referable to well-known Late Cretaceous taxa: Titanosauria, the globally dominant group of Campanian-Maastrichtian sauropods [47], robustly diagnosed by the presence of procoelous caudal centra [29] (alternatively this has been considered a synapomorphy for the constituent clade Lithostrotia [21]); and Abelisauridae, whose latest Cretaceous African distribution is substantiated by rare teeth [24] uniquely possessing a centrally positioned apex with strongly curved mesial, and straight distal profiles [30]. Closely compatible dental morphometry derived using an established data set [24,32,45] and methodologies [33,41,43,46] contributes further support for this theropod classification.
The combined evidence of testable phylogentic character states and metric similarities, we believe, provides a rigorous basis for our taxonomic assignments, even when reliant upon a few incomplete specimens. This is important because dinosaur material from the Arabian Peninsula and Levant is otherwise limited to isolated traces, or rare assemblages recognized only from non-diagnostic body fossils [3,4] and track ways [11]. Given this dearth of co-occurring remains, nothing has yet been gleaned of Arabian dinosaur diversity other than the sympatric presence of indeterminate ornithopods, sauropods, and theropods during the Maastrichtian [3,4]. The recovery of demonstrably coeval titanosaurian (possibly lithostrotian) and derived abelisaurid remains in the Adaffa Formation of Saudi Arabia therefore provides the first taxonomic verification of faunal composition within the region. Moreover, it brings to light the only definitively identifiable example of a non-avian theropod dinosaur clade from the Arabian subcontinent, and one that shows closest compatibility with penecontemporaneous faunas in Africa [24] and Madagascar [23].
The Arabian Peninsula was contiguous with the main North African landmass during the Late Cretaceous [see Fig. 1], and would have experienced uniform equatorial climates and vegetational regimes [18]. Phylogenetic coherence of the Adaffa Formation dinosaur remains with quintessential northern Gondwanan faunal elements is therefore not surprising. However, the Afro-Arabian record of titanosaurians and abelisaurids is not only extremely poor, but also mainly restricted to the pre-Cenomanian Cretaceous [48,30]. Thus the Saudi Arabian fossils, together other finds from the latest Cretaceous (Maastrichtian) of Morocco [49,50], Jordan [2], and Egypt [24], provide important evidence of the palaeogeographical ubiquity of these taxa along the northern Gondwanan margin towards the end of the Mesozoic.