An enigmatic giraffid from the latest Miocene of Italy: Taxonomy, affinity, and paleobiogeographic implications

The origin and evolution of endemic species characterizing the Oreopithecus-faunal assemblages of the Tusco-Sardinian archipelago remain a matter of debate. An emblematic case is the enigmatic giraffid Umbrotherium azzarolii, represented by a single specimen from the type locality of Casteani (Tuscany) and by several isolated teeth and fragmentary mandibles from the locality of Fiume Santo (Sardinia). An exhaustive diagnosis of Umbrotherium has not been firmly established, and its systematic and phylogenetic position remain unresolved. Unpublished remains of giraffids, including an almost complete mandible, several isolated teeth, and other cranial remains are described for the first time in the present work. The specimens were collected from the locality of Botro della Canonica (Pisa), located at the northernmost portion of the Tusco-Sardinian archipelago. The new material sheds light on the morphological and morphometric variability of Umbrotherium, thereby enabling a comparison between specimens collected from different Tusco-Sardinian Miocene localities spanning from the V1 to the V2 Oreopithecus-Zone Faunas and allowing the establishment of the new species U. engesserii sp. nov. from Fiume Santo (Sardinia). This study also reveals that Umbrotherium was more closely related to Decennatherium than to other Late Miocene continental giraffids, suggesting a dispersal of its ancestor from the Iberian Peninsula. Accordingly, a new paleogeographic and biochronological framework is proposed herein for the Tusco-Sardinian archipelago, hypothesizing a fragmentation of the area into several domains, with sporadic reconnections, and the establishment of different faunal assemblages.


Introduction
Since the description of the Oreopithecus dental material (Gervais 1872), numerous studies have been aiming to investigate the evolution of fossil endemic mammals from the Late Neogene of the so-called Tusco-Sardinian archipelago, in the Western Mediterranean Basin (Weithofer 1888(Weithofer , 1889Rook et al. 2000;Abbazzi et al. 2008;Rook 2016). Despite these efforts, no conclusive evidence has yet been provided that confidently addresses the debate about the origin of some of the species characterizing the Oreopithecus-faunal assemblages, such as the enigmatic giraffidlike Umbrotherium azzarolii. At present, the only occurrence of U. azzarolii is the type specimen from Casteani (around 8.3-7.7 Ma; V1 in the Oreopithecus-Zone Faunas) in Tuscany and from the slightly younger locality of Fiume Santo (ca. 7.1-6.7 Ma; V2 in the Oreopithecus-Zone Faunas) in Sardinia (Hürzeler and Engesser 1976;Rook et al. 2000Rook et al. , 2011Abbazzi et al. 2008;Rook 2016). Despite the presence 1 3 of dentition and some postcranial elements, the identification of Umbrotherium as a giraffid has not been confirmed, and its systematic and phylogenetic position remained unresolved.
Archival research at the Natural History Museum of Basel (NMB) has enabled the rediscovery of an unpublished fauna from the latest Miocene lignite levels (V1) of Botro della Canonica (BC), 60 km south of Pisa, that contains the cranial remains of a giraffid. The new specimens are described here and compared with the type material from Casteani and the material collected from Fiume Santo to provide new considerations on the taxonomy and affinities of this endemic mammal and the relative paleobiogeographic implications.
Archival research at the Natural History Museum of Basel allowed the collection of information about the provenance of some fossil large mammals from a locality at the northmost boundary of the Tusco-Sardinian Miocene paleobioprovince: Botro della Canonica (Pisa) (Fig. 1). This locality was first mentioned by Del Campana (1918), who recorded the presence of a few isolated teeth of Maremmia; no other records or specimens were subsequently published from this locality. The name corresponds to a small creek located in the municipality of Montecatini Val di Cecina (Pisa), close to the medieval village of Sassa (Fig. 1). Fieldwork conducted at Botro della Canonica enabled the identification of the outcropping lignite level from which the fossils had been collected. The fossil assemblage is still under study, and it is characterized by the presence of Chelonia indet., Oreopithecus bambolii, Etruria sp., Tyrrhenotragus gracillimus, and Maremmia haupti. Overall, the fauna can be referred to the V1 faunal complex and can be biochronologically correlated with the well-known localities of Casteani, Montemassi, and Ribolla (Fig. 1).

Materials and methods
The described material is stored in the Naturhistorisches Museum of Basel (NMB), and at the Department of Earth Sciences, University of Florence (DST). Comparisons were made with type material from Casteani, stored at the Natural History Museum, Geology and Palaeontology section, University of Florence (IGF); the specimens described by Abbazzi et al. (2008) and stored at "Soprintendenza Archeologia Belle Arti e Paesaggio per le Province di Sassari e Nuoro"; and Late Miocene representatives of fossil Giraffidae from both published papers (Geraads 1978;Hamilton   1978; Kostopoulos and Koufos 2006;Solounias 2007;Kostopoulos 2009;Ríos et al. 2017;Parizad et al. 2020;Iliopoulos and Roussiakis 2022;Laskos and Kostopoulos 2022), as well as personal observations. Dental nomenclature is based on a previous study by Bärmann and Rössner (2011) (Fig. 2). The studied specimens were measured with a digital caliper 0-150 mm/0.01, and the measurements are reported in Online Resource 1. Several of the studied specimens were acquired using the structured blue LED light 3D scanners Artec Eva and Artec Space Spider. A selection of downloadable 3D models is available in Online Resource 2.
Institutional abbreviations: Bac, Baccinello collection housed in the Naturhistorisches Museum, Basel; DST-FS, Fiume Santo Collection, provisionally housed at the Department of Earth Sciences, University of Florence; FS, Fiume Santo Collection, Soprintendenza Archeologia Belle Arti e Paesaggio per le Province di Sassari e Nuoro; IGF, Natural History Museum, Geology and Palaeontology section, University of Florence; NMB, Naturhistorisches Museum, Basel.
Emended diagnosis. Middle-sized ruminant with fairly brachydont dentition. Markedly rugose enamel wall. DP4 is fully molarized. Both DP3 and DP4 lack strong labial and lingual cingula. DP3 with strong and anterolabially directed parastyle, anterior and posterior lobes are lingually convergent. Protocone and hypocone are lingually V-shaped on DP4. P2 squared, with a posterolabial crista slightly concave, anterior and posterior styles prominent and developed till the base of the crown. P3 and P4 rectangular in outline, wider than long, with rounded lingual wall. The i1 larger than i2 and i3. Lower molars with the presence of ectostylids.
Description. The teeth have a markedly rugose enamel. DP3 (NMB306a16-3.10.93) is trapezoidal labially longer than lingually (Fig. 3b). There is a faint lingual cingulum below the lingual cones. The paracone fold and parastyle are well-developed. The preprotocrista reaches the parastyle; it is slightly backwards directed and the protocone is lingually rounded. The anterior fossa is bilobate in its anterior margin. The premetaconulecrista does not reach the postparacrista. The metaconule is lingually rounded. DP4 is complete and trapezoidal-shaped; the parastyle is well developed and anteriorly elongated; the paracone fold is more prominent than the metacone fold (Fig. 3b). A mesial cingulum is present. The mesostyle and the postmetacrista bend labially. The preprotocrista and parastyle are connected at this stage of wear, the protocone is V-shaped and asymmetric, whereas the metaconule is wide V-shaped. The postprotocrista is not connected with the premetaconulecrista. The premetaconulecrista and the postmetaconulecrista do not reach the labial cristae. M1 has a wide U-shaped labial profile of the posterior lobe, with the metacone fold faint (Fig. 3b). The paracone fold is slightly developed, and the labial profile of the anterior lobe is almost flat. The postmetacrista and the mesostyle are thin and bend labially. The parastyle is thin and long, almost reaching the metacone fold of DP4. The protocone and metaconule are V-shaped and the lingual cristae do not connect with the labial cristae. The postprotocrista and premetaconulecrista are separated. P2 (NMB306a22) is almost squared, slightly wider than long, with a wide and anteriorly located lingual cone (Fig. 3c). The fossa is long and narrow, and it bends labially on its anterior portion. A small enamel fold is present on the lingual-posterior margin of the fossa. The anterior and posterior styles are prominent and reach the base of the crown (Fig. 3c). The labial cone is large, and the posterolabial crista is slightly concave. P3 (NMB306a22) lacks the labial wall (Fig. 3d). The lingual cone is located on the anterior half of the lingual side of the tooth, and the lingual wall displays a faint groove on its posterior half. The fossa is long and narrow, and it bends anterior-labially on its anterior half. The right mandible NMB306a22 preserves the incisor corpus and the ramus (Fig. 3e). The height of the ramus is regular and slightly increases below the m2-m3. The mental foramen is in the middle of the diastema (Fig. 3e). In dorsal view, the diastema has a sharp ridge, and the corpus is wider in its posterior portion. The incisor corpus preserves the c, i1, i2, and i3 (Fig. 3e). The i1 is spatulate and wider than i2 and i3, which are similar in size and rectangular-shaped. The c is asymmetric, with an anterior-labial side taller than the posterolabial one. The posterior lobe has a small cingulum on its labial-posterior side. On p2 (Fig. 3e), the anterior stylid is short and connected to the mesolingual conid by a straight anterolabial cristid. The anterior valley is relatively wide and concave in occlusal view. The p3 lacks a metaconid; the hypoconid and entoconid are fused and are obliquely oriented (Fig. 3e). The entoconid joins the protoconid. The labial wall of the hypoconid is relatively sharp. The anterior valley is wide and V-shaped in occlusal view. The p4 is molarized. The lingual wall of the metaconid is convex in occlusal view, whereas that of the entoconid is relatively flat (Fig. 3e). The anterior fossa is crescent shaped. The entoconid and hypoconid are connected at this stage of wear and the labial wall of the hypoconid is rounded. A small ectostylid is present on the labial side. The m1 is relatively worn out; the lingual cones are V-shaped and separated by a V-shaped groove; the hypoconid is asymmetric and more posteriorly placed while the protoconid is symmetric. The labial cones are slightly convex in occlusal view and are separated by a wide and faint groove. The anterior and posterior valleys are narrow and crescent shaped. The first and second lobes of the m3 resemble the m1, but the lingual cones are sharper. At this stage of wear, the protoconid is connected to the entoconid. The hypoconulid is rounded and well-developed. An ectostylid is present on m1 and m3.
NMB306a06 is a fragment of right mandible with p2-m2 and fragments of m3. The p2 is subtriangular in shape; the anterior stylid is evident and connected to the mesolingual conid by a straight anterolabial cristid. The mesolingual conid is large and posteriorly bifid at this stage of wear. The anterior valley is wide and regularly concave in occlusal view, while the posterior valley is narrow. The posterolabial and posterolingual conids are connected. The p3 is not molarized and lacks a metaconid, fusion has occurred between hypoconid and entoconid, the entoconid joins the protoconid; the paraconid and parastylid are well developed. The labial wall of the hypoconid is sharp and a deep, and a relatively wide labial groove divides the tooth into a small posterior lobe and a long and wide anterior lobe. The anterior valley is V-shaped in occlusal view. The p4 is molarized, with the metaconid in the form of an anterior-posterior wall, which joins the paraconid and lacks the transverse connection to the protoconid. A small anterior fossa is visible on the worn tooth. The entoconid and hypoconid are connected on the worn p4, and the labial wall of the hypoconid is rounded.
A deep groove in the labial side divides p4 into two unequal parts, anterior and posterior lobes, with the posterior one being strongly reduced. The m1 is much worn; the lingual conids are rounded and separated by a V-shaped groove; the hypoconid is asymmetric and more posteriorly placed, while the protoconid is symmetric. The labial conids are slightly convex in occlusal view. The anterior and posterior fossae have a crescent shape. The m3 on NMB306a06 preserves only the hypoconulid, which is rounded and well-developed, and the second lobe, with a sharp hypoconid and an almost lingually flat entoconid. The second and third lobes are lingually separated by a clear step and labially by a V-shaped groove. NMB306a21 is a fragment of right mandible with p4-m3 and a damaged m2 (Fig. 3f). The morphology of the teeth is similar to that of NMB306a22 and p4; m1 and m3 bear ectostylids.
Remarks. A link between the endemic Sardomeryx from Sardinia and Umbrotherium is discarded at present. Sardomeryx is smaller than Umbrotherium (Online Resource 1) and possesses some derived features, such as a tendency toward high crowned dentition and a shortened premolar toothrow (Mennecart et al. 2019). The non-molarized p3 is a feature that precludes a link to the Paleotraginae, but, according to Abbazzi et al. (2008), a primitive p3 is also documented in Decennatherium and Helladotherium, while according to Hamilton (1978) it is pretty common among many other giraffids. However, the deciduous teeth of Helladotherium and all other derived Sivatheriinae display labial and lingual cingula and well-developed folds and spurs (Colbert 1935;Geraads 1978Geraads , 1986Hamilton 1978;Geraads and Güleç 1999;Kostopoulos and Koufos 2006;Kostopoulos 2009), contrary to Decennatherium, Samotherium, and palaeotragines, which display simpler morphology. The anterior lobe of DP3 is fully molariform in U. azzarolii and in Samotherium and Decennatherium rex, but not fully molariform in Paleotragus (Kostopoulos 2009;Ríos et al. 2017). In S. boissieri, the anterior and posterior lobes tend to converge lingually on DP3 (Kostopoulos 2009), similar to the specimen from Botro della Canonica. In D. pachecoi, as well as in Umbrotherium, DP3 has the premetaconulecrista that does not reach the postparacrista and the parastyle is strong; the anterior lobe is longer than the posterior one and both are lingually convergent (Morales and Soria 1981). DP4 is fully molarized in Samotherium, Decennatherium, Paleotragus, and U. azzarolii, whereas in P. rouenii, it has an angular lingual protocone and a feeble hypoconal spur. Both DP3 and DP4 are missing strong labial and lingual cingula in paleotragines, as well as in U. azzarolii. According to Kostopoulos and Koufos (2006) and Kostopoulos (2009), the large-sized Paleotragus species (P. coelophrys, P. expectans, and P. quadricornis) have a more labially placed parastyle on DP3, a functional style between protocone and hypocone on DP3-DP4, and a rounded lingual protocone and hypocone on DP4. The P2 of P. rouenii is sub-squared, with a strong parastyle and paracone that fuse at the base of the tooth, and the protocone and hypocone are barely distinguished by a shallow groove, similar to the P2 from Botro della Canonica, while in P. quadricornis the parastyle is reduced (Kostopoulos 2009). The P2 of Paleotragus sometimes displays an enamel fold on the posterior part of the central fossa (Kostopoulos 2009). In P. rouenii, the P3 and P4 are similar in shape and lingually rounded, and P. quadricornis has an incipient lingual bilobation on both teeth, as also observed on the P3 from Casteani and Botro della Canonica. This feature also seems to be present in Decennatherium rex (Ríos et al. 2017). In S. boissieri, the parastyle and paracone are equally strong on the upper premolars. In Samotherium and Decennatherium rex, the P2-P3 are wider or slightly wider than long, almost rounded in shape, while in D. pachecoi, the P4 is longer than wide, and the labial cones are weak, similar to Umbrotherium. On the upper molars of P. roueni, the mesostyle is stronger than the parastyle, and in P. quadricornis, the upper molars have a rounded hypocone and protocone, contrary to Umbrotherium. The M1 in D. pachecoi has a strong parastyle, as observed for Botro della Canonica. Angular lingual cones are documented in D. pachecoi (Ríos et al. 2016(Ríos et al. , 2017, as are the non-molarized p3 and the very rugose enamel. Holotype. FS1995#0342, a left partial mandible with p3-m3, figured in Abbazzi et al. (2008: text- fig. 8G-I), housed in the "Soprintendenza Archeologia Belle Arti e Paesaggio per le Province di Sassari e Nuoro". Locality, horizon, and age. Fiume Santo (northern Sardinia), faunal assemblage V2; Late Miocene, late Tortonian, late Turolian, late MN12-earliest MN13.
Diagnosis. Smaller species of the genus. P2 with a weak parastyle and a small and narrow lingual wall. All the lower premolars are shorter than U. azzarolii. Lower toothrow shorter than in U. azzarolii. Ectostylid absent on m1-m3. The lower molars display sharper entoconid and metaconid in lingual view than in U. azzarolii.
Description. The specimens were partially described by Abbazzi et al. (2008). Only a few characters can be added. The isolated canine is bilobate with a large anterior lobe and a small posterior one, similar to the specimen from Botro della Canonica. On P2, the parastyle is very weak; it does not reach the occlusal surface on little worn teeth and does not reach the base of the tooth. The lingual cone is small and posteriorly located on the lingual wall.
Remarks. The specimens from Fiume Santo exhibit the general Umbrotherium features but differ from the type species, being smaller (Online Resource 1; Fig. 5) and slightly morphologically different. In particular, the lower and upper teeth are more hypsodont than in U. azzarolii and have a simpler morphology with the absence of ectostylids on lower molars and a weaker parastyle on P2. The sharper lingual cones may be related to an advanced degree of endemism (increase of hypsodonty), as observed in other taxa (Mennecart et al. 2019).

Discussion
The remains originally referred to Umbrotherium by Hürzeler and Engesser (1976) were not formally described, rendering it a nomen nudum according to the ICZN (1999). In 2008, Abbazzi and coauthors provided an extensive description and formal definition of Umbrotherium based on fragmentary mandibles and isolated teeth collected from Fiume Santo (Abbazzi et al. 2008). The authors suggested an attribution to the Giraffidae family, despite the fact that the main diagnostic characters, such as the bilobate lower canines and ossicones, were not documented from either southern Tuscany or Fiume Santo. Unfortunately, the comparison between Fiume Santo and the type material from Casteani did not reveal any significant differences in morphology between the two samples, mainly because of the absence of lower dentition at Casteani. Furthermore, the differences in size within Umbrotherium were interpreted as sexual dimorphism by Abbazzi et al. (2008).
The abundant material from Botro della Canonica allows an extensive comparison with both Casteani and Fiume Santo, as well as a better comparison with other late Miocene European giraffids.
The presence of non-molarized p3s, the morphology of the upper premolars, and the morphological features detected on upper deciduous teeth suggest an affinity between Umbrotherium and Decennatherium, in particular D. pachecoi, rather than with Palaeotragus or Samotherium. Considering the common characters shared between D. pachecoi and Umbrotherium, we suggest a close relationship between these two taxa; however, further phylogenetic analyses would be helpful to validate this hypothesis. Decennatherium pachecoi is documented in the Iberian Peninsula during the Vallesian (mainly MN9-10), and the dispersal by the ancestor of Umbrotherium in Tusco-Sardinian archipelago should have taken place at least before or during the Vallesian, probably from the West (Fig. 6).
The upper permanent dentition from Botro della Canonica resembles the type material from Casteani both in proportions and morphology. The isolated P2 from Fiume Santo instead displays some differences with Botro della Canonica, showing a different degree of development of the parastyle and the lingual wall. The lower teeth from Fiume Santo also display some differences in size (in particular in the length of the toothrow),  and a few differences in morphology, such as the absence of ectostylids. These characters can be related to a more advanced stage of endemism in the species from Fiume Santo and seem to be in agreement with the evolutionary transformation of endemic forms from V1 to V2 assemblages in the Tusco-Sardinian domain (e.g., Paludolutra campanii from Tyrrhenolutra helbingi, and Maremmia lorenzi from Maremmia haupti). The V2 assemblage from Fiume Santo is indeed characterized by species that show a more pronounced degree of endemic evolution with respect to the V1 assemblage from Tuscany. At Fiume Santo, the suid M3, referred to as Eumaiochoerus cf. E. etruscus, lacks the three grooves in the protocone and hypocone, and which are present and shallow in the paracone and metacone. These characters are interpreted as the result of increasing enamel thickness, a derived feature in suids under insular condition (Made 1999). The Maremmia from Fiume Santo resembles M. lorenzi from the V2 of Tuscany, but it appears to be slightly larger when compared with the material from Baccinello-Cinigiano. For this reason it was referred to as Maremmia cf. M. lorenzi (Abbazzi et al. 2008).
All this evidence suggests that the Fiume Santo assemblage could be intermediate between the classical V1 and V2 assemblages from Tuscany, but also suggests the existence of separated areas within the Tusco-Sardinian Paleobioprovince (Fig. 6). The new results obtained from the study of Paludotona highlighted the existence of a geographic barrier between the Fiume Santo area (Sardinian subdomain) and the Tuscany area (Tuscan subdomain), as was also hypothesized by Casanovas-Vilar et al. (2011). Furthermore, according to Angelone et al. (2017), morpho-dimensional differences between Paludotona etruria and Paludotona minor are indicative of a further fragmentation into two or more insular domains within the Tuscan subdomain (Angelone et al. 2017) as previously suggested by Engesser (1989) regarding the occurrence and intrageneric morphological comparison of Anthracomys. The fragmentation of the Tuscan subdomain may explain the absence of Umbrotherium in the Baccinello-Cinigiano Basin and its occurrences only in the northern part of the Tuscan subdomain (i.e., Botro della Canonica and Casteani) (Fig. 6). The presence of Umbrotherium at Fiume Santo is also suggestive of a temporary connection between the northern Tuscan subdomain and the Sardinian subdomain at the time of the arrival of Eumaiochoerus.

Conclusion
New fossil material collected from the late Miocene of Tuscany and from the V1 assemblage at Botro della Canonica (Pisa) is described here for the first time. Most of this collection include cranial elements and isolated teeth belonging to the enigmatic giraffid Umbrotherium. Among the studied specimens, bilobate canines allowed a reference to the family Giraffidae, while deciduous teeth and new observations on upper and lower teeth allowed a linkage between Umbrotherium and the Iberian giraffid Decennatherium pachecoi. Furthermore, the new material from Botro della Canonica, together with that from Casteani, display a larger size and a different morphology with respect to the material collected at Fiume Santo, suggesting the presence of two endemic species with different levels of endemism: U. azzarolii from the Tuscan subdomain and U. engesseri sp. nov. from the Sardinian subdomain. The new results derived from the study of the Botro della Canonica giraffid would open up a new hypothesis regarding the colonization route of the Tusco-Sardinian archipelago and its paleogeographic structure. New fieldwork activities at Botro della Canonica and the revision of the material from the other Tuscan localities, such as Serrazzano and Montebamboli, are required for better investigation of the paleogeographic assessment of the area and the origin of other endemic mammals, such as the genera Maremmia and Etruria.
were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.