The fallow deer Dama celiae sp. nov. with two-pointed antlers from the Middle Pleistocene of Madrid, a contemporary of humans with Acheulean technology

We describe fossils of a new species of fallow deer, Dama celiae. It is the end member of the lineage Dama farnetensis–D. vallonnetensis–D. roberti–D. celiae, which reduced the number of points of the antler from four to two, while the parallel lineage leading to the living fallow deer evolved more complex and palmate antlers. The fossils are from localities Pedro Jaro I and Orcasitas in the + 25–30-m terrace of the Manzanares river, which is correlated to MIS9 (337–300 ka) and which also yielded fossils of Megaloceros matritensis, a recently named species, end member of a lineage that survived longer than previously believed. A younger terrace of the Manzanares yielded remains of Haploidoceros, a rare deer known from two older localities in southern France and one younger locality in Spain. So many rare deer species in this valley indicates either endemism and a very special environment or that the record of fossil deer is much less known than generally assumed. Until recently, the European Middle Pleistocene record of deer had only one middle-sized species at a time. Now, it appears that there were up to three contemporaneous species of the size of a fallow deer. Acheulean lithic assemblages have been documented from the same sites as Dama celiae. This species was contemporaneous to Neanderthals with Acheulean culture. Cut marks suggest that it was consumed by them and probably was hunted.


Introduction
Traditionally, it was considered that the diversity of Cervidae in the later part of the Middle Pleistocene of Europe was low with only one species at a time that had the size of a fallow deer (e.g., Kurtén, 1968;Faure & Guérin 1992;Delpech & Guérin 1996;Van der Made 1999). More recently, the existence of a new genus with a species of this size in the Middle and Late Pleistocene of Europe became accepted (Croitor et al. 2008;Sanz et al. 2014; Van der Made & Mazo 2014) and descendant of the giant deer Megaloceros savini was shown to have survived into the later part of the Middle Pleistocene of the Manzanares valley with the species Megaloceros matritensis (Van der Made 2019). Here, we describe another new species of deer from the same terraces of the Manzanares valley, Madrid, which adds to the cervid diversity and contributes to our understanding of the evolution of the fallow deer.
There is consensus that the living species, Dama dama and Dama mesopotamica, are closely related to the fossil species (or subspecies) Dama clactoniana. All of them have palmate antlers. But when tracing the ancestry further back, consensus disappears. Different authors trace the palmate fallow deer back to some of about ten Early Pleistocene European named species or subspecies of similar size, with branching antlers (e.g., Fig. 1). These or other authors placed these species in the genera Dama, Cervus, Axis, Rusa, Pseudodama, or Euraxis. Part of the problem is that the early Middle Pleistocene and late Early Pleistocene record of fallow deer is less dense. Many species of fallow deer have been named, quite some of them being synonyms, but the species from this period were among the last ones to be named (Table 1). These are Dama vallonnetensis (Lumley et al. 1988) and Dama roberti (Breda & Lister 2013). Several models of the evolution of the fallow deer have been published, which differ in the position of these species, either directly ancestral to the palmate species or on a side branch.
It is the aim of this paper to (1) describe and name the fallow deer from the sand pits Pedro Jaro I and Orcasitas in the Manzanares valley, Madrid; (2) discuss the phylogenetic position of this species and of D. vallonnetensis and D. roberti relative to the fallow deer with palmate antlers; and (3) present the Acheulean archaeological sites of Pedro Jaro I and Orcasitas.
Praemegaceros, and Eucladoceros. No tree in a temporal context was presented.
Di Stefano and Petronio (2002) considered the palmate species of Dama to be descendants of a lineage with species with three-pointed antlers (here: points or tines A, B, C), which they placed in Rusa (Fig. 1). They placed a parallel lineage in Axis, which acquired a fourth point (here: A, B, C, D), and then gave rise to A. farnetensis, which survived till about 500 ka. Croitor (2006Croitor ( , 2018 believed that the Dama e. eurygonos-D. e. farnetensis lineage gave rise to three lineages, leading to D. vallonnetensis and the two living species (Fig. 1), which implies the loss of the middle tine (point B) in D. vallonnetensis, while the palmate species retained it. In his opinion, D. roberti is based on juvenile antlers of D. clactoniana. He believed most of the Early Pleistocene European fallow deer-sized species not to be closely related to palmate Dama and placed them in Cervus and Metacervoceros. Breda and Lister (2013) presented a model in which the palmate species of Dama evolved passing through a threepointed stage and a four-pointed stage, lost the middle tine (or tine B) (in "P." farnetensis, D. vallonnetensis, D. roberti), and regained it in the palmate species. They also suggested the alternative that D. roberti evolved directly from a species with three points (A, B, C) or that the palmate species evolved directly from a species retaining the middle tine.
Van der Made and Mazo (2014), Van der Made (2015), and Van der Made et al. (2017) considered Dama vallonnetensis and Dama roberti to be on a lineage that reduced the tines and that the palmate species evolved from Dama nestii or Dama nestii farnetensis. Van der Made (2015) and Van der Made et al. (2016 recognized the palmate species Dama peloponesiaca and D. aff. peloponesicaca, not recognized by the other authors ( Fig. 1).

Nomenclatorial issue
One of the species discussed here is Dama roberti. Its nomenclatorial status has given rise to some discussion. Robert (1829) described two skeletons under the name "Cerv. Dama Polignacus": one of a large individual and one of a small individual and ended his description with the question whether these size differences indicated two different species or two races of the same species and stated that new finds will allow to decide. He did not formally indicate a holotype. Pomel (1853) introduced the new name Cervus roberti for fossils previously described by Robert under the name Cervus Dama polignacus. Pomel did not indicate whether he referred to the large or small skeleton, but his description included two features that are present in the antler of the small deer and that are different in the antler of the large deer and one feature that Robert (1829) indicated to be present in the large deer, but not in the small one. It is not clear whether Pomel (1853) believed there were two species and named the second one, or that he intended to introduce a replacement name and for what reason. Pomel's (1853) name was used only a few times. Scrope (1858, p. 224) reproduced Pomel's (1853) faunal list, including Cervus roberti, but did not comment on this species. Lydekker (1885) merely mentioned it as a synonym of C. polignacus. De Stefano (1914) stated that the fossils of Cervus roberti, C. polignacus, and other species were too incomplete and fragmentary to know whether some of them are synonyms. Azzaroli (1953) unambiguously applied Robert's (1829) name Cervus polignacus to the large skeleton and Pomel's (1853) name Cervus roberti to the small skeleton. This seems to satisfy article 74.5 of the ICZN for the selection of a lectotype for each of the two species. Breda and Lister (2013) named a species Dama roberti in honor of two persons named Robert, Robert Mutch, who collected the holotype from Pakefield, and Félix Robert, who collected the paratype from Soleilhac. Van der Made et al. (2017) noted (1) that there was already a species called Cervus Dama roberti, that according to Azzaroli (1953), was based on the specimen, that Breda and Lister (2013) later designated paratype of their D. roberti, and (2) the correct name for a species named in honor of both persons called Robert would be Dama robertorum. Lister and Breda (2016) indicated that the ICZN (1999, article 33.2) does not allow correcting an incorrectly latinized name. They also opted for the interpretation that C. roberti was meant to be a replacement name of C. D. polignacus and that thus both have the same type specimen (ICZN 1999, article 72.2). Whatever the interpretation of the nomenclatorial rules, the name Dama roberti applies to the deer from Soleilhac and Pakefield.
Here, we describe and name the new species Dama celiae, which we believe to be the descendant of D. roberti and which was found in two Acheulean archaeological sites from Madrid.

Geomorphological and chronological framework
The Pedro Jaro I and Orcasitas sand quarries were located in the SE of Madrid, in the valley of the Manzanares, in the South Sub-plateau of the Iberian Peninsula. The Manzanares, tributary of the Jarama, which flows again into the Tagus river, is a lowsinuosity river embedded in an asymmetrical valley in which two terrace systems can be distinguished (Pérez-González and Uribelarrea 2002). Isostatic rising, block adjustment, and litological structured controls (Alia 1960;Pérez-González 1971Silva et al. 1988a and b) or climatic and eustatic fluctuations (Silva et al. 2017) interfered with their formation. In the upper course of the Manzanares River valley, up to 13 stepped terraces can be distinguished (between + 4-5 and + 95 m), with thicknesses of up to 6-7 m (Pérez-González 1994). In the lower course of the Manzanares river, downstream of the district of Villaverde and until the confluence with the Jarama, the terraces (+ 25-30 m, + 18 + 20 m, and + 12-15 m) are affected by synsedimentary subsidence because of the substratum of evaporitic rocks, which gives rise to a thickening of the alluvial deposits (Uribelarrea 2008), known as the Complex Terrace of Butarque (Goy et al. 1989). Lithic industry and faunal assemblages have been found in the sand and gravel bars, as well as in flood plain facies of the stepped terraces and mainly in the Complex Terrace of Butarque, resulting in one of the largest concentrations of Paleolithic sites known in Europe Villa 1990, 2006;Pérez-González 2000-2001;Rubio-Jara et al. 2016;Rubio-Jara and Panera 2019;Yravedra et al. 2019a).
The Pedro Jaro I and Orcasitas sand quarries (Quero 1994;Gamazo 2002) were located at the stepped terrace + 25-30 m of the Manzanares River valley (Fig. 2) (Rubio-Jara and Panera 2019), which has an equivalent morphological position to Unit Arganda I at the Arganda Complex Terrace (≈T + 30-32 m) of the Jarama River, dated between MIS11 and MIS9 according to AAR analysis and biostratigraphy (López Martínez 1980;Sesé et al. 2011a and b;Panera et al. 2011), and probably within MIS9 according to ESR dating (Moreno et al. 2019). Likewise, the T + 25-30 m of the river Tagus valley has been dated to the end of the MIS9 or early MIS8 by IRSL (López-Recio et al. 2015).

Arenero de Pedro Jaro I
The Arenero (= sandpit) Pedro Jaro I was located in the neighborhood of Los Rosales, on the left of the road from Madrid to Andalucía at km 7.3 (Gamazo 2002: 374). It is situated in deposits of the + 25-30-m terrace of the river Manzanares (Fig. 2). The same person owned also another sandpit, situated between km 4 and 5 of the road to San Martín de la Vega (Gamazo 1982, Fig. 1, number 14;Gamazo 2002: 375). This one is now called Arenero de Pedro Jaro II and should not be confused with number I. Sesé and Soto (2002b) gave a faunal list under the heading Pedro Jaro, but this is of the other sandpit (Pedro Jaro II). The faunal list of Pedro Jaro I includes caballoid horse Equus cf. ferus, Equus cf. hydruntinus, Rhinocerotidae cf.
The finds preserved in the Museo Arqueológico Nacional (hereinafter MAN) and Museo de San Isidro (hereinafter MSI) have been virtually unpublished until this study. The lithic industry (448 pieces) is ascribed to the Acheulean technocomplex, because of the presence of handaxes, choppers, and flakes with lengths of over 10 cm and because levallois is rare (Fig. 4). The presence of pieces of all the technological groups that constitute the operative chain proves that the collection of lithic elements on site was not selective but systematic ( Table 2).
The best-represented technological group is that of flakes, 61.5% of the total series, of which 16.5% are retouched, although the relevant presence of large cutting tools (LCTs)-20%-stands out.
Among the cores (average dimensions 89 × 74 × 41 mm), the dominant reduction processes are the elementary ones with 20 pieces, some of them opportunists. There is also the presence of some pieces with polyhedral bifacial and multifacial schemes. However, a significant presence of cores with organized reduction processes (13) is  Instituto  Geológico y Minero de España,  IGME-CSIC) documented, fundamentally from the discoid group, and to a lesser extent peripheral monopolar debitage or Quina concept (cf. Turq 2000: 316), characterized by the secant or perpendicular exploitation of a main surface which may be cortical or débitage. Although its presence might seem rare for the Acheulean, at the Thiongo Korongo site, with a chronology of 1.3. Ma, there is a high percentage (20%) of these cores from the Lower Floor (Santonja et al. 2014), and they were also recorded at TKSF (Rubio-Jara et al. 2017). Approximately 5% of the total of the complete flakes are from centripetal or discoid cores.
The f lake tools have average dimensions of 59 × 50 × 17 mm and all of them are made on flint, except two pieces of quartzite. The best-represented types are scrapers and retouched flakes (the two add up to 60% of the flake tools). Among the scrapers, the transverse ones The temporal distribution of relevant species of large mammals discussed in the text. On the left, the temporal scale in millions of years (Ma), the difference in Northern Hemisphere air temperature with the present (dNHT in Kelvin), after Van de Wal et al. (2011), and the Marine Isope Stages (MIS). On the right, temporal distribu-tion of the species (thick lines) and their presence in selected localities (solid squares) or possible presence (open squares). The selected localities are preferentially those with a first or last occurrence (left; save if this is latest Pleistocene or Holocene) and in the Manzanares sandpits (on the right)
The predominance of convergent scrapers and the presence of two Mousterian Points show the presence of symmetrical points in the toolkit.
Among t he handaxes (average dimensions 139 × 79 × 42 mm), the most commonly used support is flint flakes, although in five pieces (of which three have flake as support) the raw material is quartzite. Thick handaxes (virtually 90%), dominate over the flat ones. A good deal of the first ones (12) are partial handaxes with sparse bilateral and bifacial shaping, which tends to concentrate on the apical area to create pointy tips, whereas others (actually, 10) have been shaped by means of broad invasive removals, with more intense bilateral and bifacial shaping, giving rise to amygdaloid silhouettes, oval in some cases, with edges sometimes retouched and pointy tips, and in four pieces, straight edges. The flat handaxes, all made on flint and mainly crudely made, show subtriangular silhouettes with pointy tips. The intention of creating pointed pieces can also be observed in the shapping of the trihedrals (average dimensions 144 × 71 × 48 mm), which are well represented in the series. Generally, shapping is less invasive, and they are made on large flint flakes, except for one piece on quartzite. The cleavers on flake (average dimensions 144 × 93 × 50 mm), although fewer (6), are also significant. They are grouped into types II and V (Tixier 1956;Inizian et al., 1995: pp. 55-57), and show symmetrical silhouettes, transverse straight of slightly convex edges, and bilateral shapping aimed to thin down thicker areas. Finally, the large number of heavy-duty tools (24 pieces) made from large flint flakes (average dimensions 111 × 95 × 38 mm) with shapping aimed to obtaining pointy tips, sometimes with handaxe-like morphologies, together with edges that shape macrotools assimilable to scrapers, denticulates, or broad backed knives, is outstanding.

Arenero de Orcasitas
The sand quarry of Orcasitas was located at km 7 along the road from Madrid to Andalucía (Quero 2002) and very close to the Pedro Jaro I sandpit (Fig. 2), but today, the site corresponds to number 10 in the street Menas Albas of the neighborhood Orcasitas of Madrid. It is in deposits of the + 25-30-m terrace of the river Manzanares valley (Pérez-González and Calvo 1989).
In 1959, an elephant skull was found and excavated. The excavation also yielded Acheulean and, from a different level, Bronze-age lithichs and pottery (Quero 1994). Mazo (1994) described the skull. Sesé and Soto (2002b) gave the following faunal list: Elephas antiquus, Rhinocerotidae indet., Equus caballus, Cervus elaphus, Dama sp., and Bos primigenius. Van der Made (2019) described part of the deer fossils as a new species: Megaloceros matritensis. None of the deer fossils are now assigned to Cervus elaphus. The fossil collection includes part of the skull of Bos primigenius. However, metacarpal MSI O/62 7925 has a morphology that goes better with Bison. Several partial skulls of Bos primigenius are known from the Manzanares terraces, though a photograph from 1935 by Royo in the collections of the Museo Nacional de Ciencias Naturales (MNCN) attests that a Bison skull from the "arenero del Barbas" existed, but the specimen was accidentally lost (Sesé & Soto 2002b). The presence of both Bos and Bison is thus possible. The updated faunal list includes Elephas (Palaeoloxodon) antiquus, Equus cf. ferus, Rhinocerotidae, Bos primigenius, Bison sp., cf. Capreolus, Dama celiae, and Megaloceros matritensis (Fig. 3). Some subrecent teeth in the MAC belong to Bos taurus (domestic cattle) and very probably do not derive from the + 25-30-m terrace. This terrace is correlated to MIS9. This is not contradicted by the fauna, but most of the species existed well before and after MIS9 (Fig. 3). An exception is Megaloceros matritensis, which is the last member of a lineage and occurs in about 10 localities or levels in the + 25-30-m terrace of the Manzanares. No younger occurences are known, but then the species was recognized only recently. The species is a descendant of Megaloceros savini, still present in Cúllar de Baza 1 (Van der Made 2019), dated to 476 ± 24 ka (Ortiz et al. 2000). This supports an age for Orcasitas that is much younger than that of localities with Dama roberti.
The 96 pieces of lithic tool from Orcasitas studied in this paper are unpublished and are deposited at MAN. These lithic tools can be ascribed to the Acheulean technocomplex, and consist mainly of flakes, which add up to 81.26% of the whole, including retouched flakes ( Table 3). The limited presence of cores (just two) may be due to a selective collection of the finds. One is a bipolar unifacial core on flake, and the other a centripetal bifacial retouched core. The group of flake tools consists of seven flint pieces: five retouched flakes-in one of them the support is a large flake, a notch, and a retouched notch. Among the LCTs (average dimensions 140 × 86 × 42 mm), there are four handaxes: one of flint and a fourth one, very rounded, of quartzite. The supports used consist of large flakes in two cases, one of quartzite and the other one a flint cobble-slab. The handaxes are thick, with little bilateral and bifacial shapping, made by means of wide removals aimed at shapping a pointed end, except in one piece that has a fractured tip. The bifacial shapping extends to the bases, which are sharp, except in one case that is cortical. The silhouettes are amygdaloid, except in one piece that is oval. One of the handaxes was intensely retouched. The four cleavers, all made on large flint flakes, show suppressed butts and removals, in some cases with bifacial shapping, in the thicker areas that result in regular and symmetrical silhouettes. Two of those are type II cleavers, one with use retouch in the edge, and the other two are of types V and VI, the last one on kombewa flake (classification of Tixier 1956;Inizan et al. 1999). There are four heavy-duty tools, three of which are large flakes with pointed tips. The same happens with the two trihedrals, one made on an angular flint cobble and the other one on a highly cortical quartzite flake. The scarce shapping is focused on the apical third of the object to create a point. This is an Acheulean association because of the presence of handaxes, choppers, and flakes with lengths of over 10 cm and because levallois is absent (Fig. 4).

Transfesa
A few specimens from Transfesa are tentatively assigned to Dama celiae. TRANSFESA (later called TAFESA) near Villaverde (Fig. 2) was an enterprise dedicated to transport and logistics. In 1958, the skull of an elephant was found on the terrain of this company (Meléndez 1958). In addition to the skull, other remains of mammals were excavated at a higher stratigraphic level (De Andrés & Aguirre 1975) and pollen was sampled from sediments associated to the skull and from a core (Menéndez Amor & Florschütz 1963). An excavation at a level 6 m above the elephant skull and at about 75 m distance was carried out in 1996. The materials collected during the first excavations are kept in the Museo Nacional de Ciencias Naturales (hereinafter MNCN), and those of the later excavation in the Museo Arqueológico Regional (hereinafter MAR) and were described in a monograph (Baena & Baquedano Beltrán, 2010). The pollen comes from a level that has been called Mz0, the elephant skull comes from Mz1, the excavation in 1996 was in Mz2, and the other early excavations in Mz3 (Silva et al. 2010). Theese units form part of the + 25-30-m terrace of the Manzanares (Rubio et al. 2016), which is correlated to the + 30-32-m terrace of the Jarama, which is dated and correlated to MIS10-9 (Panera et al. 2011;Moreno et al. (2019);. Aguirre (1989) classified the fossils from Mz1 as Elephas antiquus, Sus cf. scrofa, and Cervidae indet., and from Mz3 as Dolichodoryceros savini and Equus sp. A skull fragment of a deer from Mz3 was described in detail as Praedama (De Andrés & Aguirre 1975). Sesé and Soto (2002b) gave the faunal list (for Mz1 and Mz3): Elephas (Palaeoloxodon) antiquus plathyrhinchus, Equus caballus, Sus cf. scrofa, "Cervus" sp., Praemegaceros sp. (= Praedama sp.), and Bovidae indet. The fossils collected from Mz2 were assigned to cf. Elephas antiquus, Megaloceros savini, Bos/ Bison, Microtus sp., Oryctolagus cuniculus, and cf. Mauremys leprosa (Mazo 2010;Sesé, 2010). We do not know on which fossils the mention of Sus was based, while part of the deer fossils became type specimens of a new species and the remaining deer fossils were assigned to a fallow deer (Van der Made 2019). The updated faunal list is Elephas (Palaeoloxodon) antiquus, Equus cf. ferus, Megaloceros matritensis, Dama celiae?, Bos/Bison, Microtus sp., Oryctolagus cuniculus, and cf. Mauremys leprosa (Fig. 3). The pollen was interpreted as indicative of a steppic environment (Menéndez Amor & Florschütz 1963). These species do not contradict correlation of the terrace to MIS10-9, but most of them are not informative with this precision. The deer fossils were initially assigned to Megaloceros savini, which led to an overestimation of the age of site, but when it was recognized that these fossils represented M. matritensis, a descendant of M. savini, nothing contradicted a correlation to MIS9-10 (see Van der Made 2019).
The lithics from unit Mz1were interpreted as middle Acheulean (Aguirre 1989), and those from unit Mz2 were described and figured in detail and classified as Acheulean  (Baena et al. 2010b, a). The substantial presence of cobbles (36%) in the assemblage of unit Mz2, which consists of 651 pieces, is remarkable. These are mainly made on quartz but also on quartzite. Only six of them show percussion marks, which suggest their use as hammerstones. The presence of these raw materials contrasts with the hegemonic use of flint (92%) in the knapping items. The flake production is focused mainly on bifacial and multifacial operative schemes, and to a lesser extent on orthogonal and discoid schemes, with no presence of Levallois, suggesting an expeditious assemblage. Macrotools (20%) and other tools on flake (30%), among which denticulates dominate followed by scrapers, notches, and becs, are well represented. There are only two cleavers on flake, in contrast with the 22 handaxes and trihedral-picks present, which are thick, with sinuous edges and barely symmetrical profiles.

Arenero de Manuel Soto
A fragment of a metatarsal from the sandpit of Manuel Soto could belong to Dama celiae. As sandpits changed owners and the same person may have owned more than one sandpit, there may be some confusion with the names. Gamazo (1982, Fig. 1) indicated the position of sandpits under the names Arenero de Manuel Soto (number 3), another one also called Arenero de Soto, Arenero de Ramón Soto, or Arenero de Nicomedes (number 19), and a third one under the name Arenero de Adrián Rosa, Aporta o Nicomendes (no. 24). Gamazo (2002) indicated that the sandpit of Manuel or Ramón Soto or Nicomedes (near km 1.6 of the road to San Martín de la Vega) corresponds to number 19 and has the code AM, that Arenero de Soto (near km 1.5 of the same road and should be no. 3) has codes ASO and CASO, while she gave no code for number 24. Fossils in the MSI with code AN are indicated to come from Arenero de Nicomedes and should correspond to no. 24.
We identified other fossils with the code AM (corresponding to no. 19) as Panthera leo (or P. spelaea), Equus cf. ferus, Rhinocerotidae indet., Bison sp., Dama celiae?, and Megaloceros matritense (Fig. 3), and from Arenero de Nicomedes with code AN as Megaloceros matritensis (= "Cervus sp.") and Bovini indet. However, Sesé and Soto (2002b)  The fossil which concerns us here is labeled to come from Arenero de Manuel Soto and has a collection number including the code AM and should be from sandpit number 19, but we cannot exclude number 3. Sandpit no. 19 is in the stepped terraces, while no. 3 is in the Complex Terrace of Butarque and its age is more difficult to assess on the basis of geomorphology. In any case, the association with code AM includes both the fallow deer metatarsal and Megaloceros matritensis, which otherwise is only known from the + 25-30-m terrace, correlated to MIS9-10.

Material and methods
The fossils from Arenero de Pedro Jaro I are stored in the MAN and MSI, and those from Arenero de Orcasitas are stored in the MSI, MAN and a few specimens in the MAR. The lithics from Pedro Jaro I and Orcasitas, analyzed in this paper, come from archeopaleontological interventions carried out during the monitoring and control of the exploitation of the Manzanares sand quarries during the 1950s and 1960s, and have never been published before. The specimens from Transfesa are stored in the MNCN and MAR, and those from Manuel Soto in the MSI. The fossil specimens are compared with others, both recent and fossil indicated in Table 4. The acronyms of the collections where those specimens were studied are indicated in Table 5. The methods of study of the Cervidae and the way of measuring (Table 6) are similar to those used by Van der Made and Tong (2008). The measurements in the figures and in Tables 7 and 8 are in millimeters.  backwards and curving laterally and again backwards, without important upward curvature and without signs of further bifurcation, important flattening or palmation. The bifurcation between brow tine and main beam is situated low above the burr.

Taxonomy and description
Differential diagnosis Dama celiae differs from Dama roberti and other species of Dama in having a long main beam without bifurcation or the flattening that preceedes such a bifurcation (or if it would have had such a bifurcation, it would be situated much further from the burr).
Dama celiae differs from Haploidoceros in that the bifurcation between brow tine and main beam is situated much lower above the burr and that the main beam does not curve upward and then foreward. Dama celiae differs from R. eldii in that the main beam of the antler does not curve upward in its distal part and in not having minor tines.

Material
MAN 73/58/PJ-39, 21, 26-three fragments of a left antler that fit together: basal part (39), main beam (21, 26) from Pedro Jaro I. Figure 5 (1). MAN 73/58/PJ-24-part of the main beam of the right antler (same individual as above) from Pedro Jaro I. MAN 73/58/PJ-25, 27, 33, 34, 35, 75-minor fragments belonging to the previous two antlers from Pedro Jaro I.  Antero-posterior diameter of the pedicles, just below the burr DTp Transverse diameter of the pedicles, just below the burr DAPr Antero-posterior diameter of the burr DTr Transverse diameter of the burr DAPb Antero-posterior diameter of the antler, just above the burr DTb Transverse diameter of the antler, just above the burr Hint1 Height of the first bifurcation, measured at the medial side and from below the burr Hint2 Height of the first bifurcation, measured at the medial side and from the upper surface of the burr Hext1 Height of the first bifurcation, measured at the lateral side and from below the burr Hext2 Height of the first bifurcation, measured at the lateral side and from the upper surface of the burr Hmid Height of the middle tine measured from the middle of the lower side of the burr till the bifurcation Hback Height    MAN 3/58/ORC-1-right frontal and antler with brow tine and initial part of the main beam from Orcasitas. Figure 5 (5).
MAN 73/58/ CA K7/7-8-left frontal, antler broken of just above the burr, right frontal of the same individual with basal part of the antler, various fragments, none of them being palmate, from Orcasitas. Figure 5 ( Figure 5 (7).
MNCN 62,332-a proximal and a distal fragment of a left radius from Transfesa unit Mz1.
MSI AM61/3877-proximal fragment of a right metatarsal from Arenero de Manuel Soto.

Description and comparison
The specimens from Orcasitas and Pedro Jaro I share the following morphologies: massive and long upward-curving brow tine, initial parts of the brow tine and main beam in a straight line or with a very obtuse angle, bifurcation between the brow tine and main beam very low above the burr, and regular pearling of the burr and groves on the surface as in Dama (the latter feature as described by Lister 1996), long main beam without bifurcation, directed backwards and then curving outwards without any tine originating at this place. The similarities between these specimens and the difference with other species indicate that they belong to the same species.
In some cases, half the width of the skull below the pedicle could be measured, allowing to estimate the approximate width. It is about 10 cm (Table 7). This is a middle-sized species.
The pedicle of the antler is situated close to the posterior half of the orbital (Fig. 53a). This is a common position in the Cervinae, while in the Capreolinae, the pedicle is situated further back above the brain case. The suture between frontal and parietal is situated just behind the pedicle and is directed transversely. The pedicles are oriented parallel (Fig. 53c), while in Haploidoceros, they tend to be divergent.
The burr has a very regular "pearling." This is more like in Dama than in Cervus, where the pearling may be more irregular. The surface of the antler has shallow groves, but no pearling, and resembles more that in Dama than in Cervus (Lister 1996).
The left antler from Pedro Jaro I, the holotype (Fig. 5  (1)), is a nearly complete antler. The main beam is long without any sign of bifurcation and is directed backwards and curves outward and then again backward. The preserved part is 58 cm from the middle of the burr, measured in a straight line and 62 cm along the outer curvature. A long part of the beam of the right antler, which is identical in its curvature, is also preserved. Its length is 62 cm, measured along the curvature, and 59 cm in a straight line. Its lower part is already some distance of the burr, suggesting that at about 64 cm from the burr there was still nothing to be seen of the flattening that preceedes a bifurcation. At this place, the diameters of the section become smaller, suggesting that the tip would be close. Fragments of the main beam that fit on MAN C37/V6 from Orcasitas show an identical outward curvature. There is no important upward curvature in the preserved part. At some distance from the burr, the section of the main beam is oval with a clearly wider transverse than antero-posterior or dorso-ventral section. There is no indication of any bifurcation save for that of the brow tine. Haploidoceros mediterraneus (Fig. 6 (2) is one of the few deer with a long not bifurcated main beam, but it is directed more upward and curves forward. The angle between brow tine and main beam is much sharper and the bifurcation placed much higher above the burr (Fig. 6). Rucervus eldi (or Panolia eldi; Fig. 6 (3) has a long main beam, but with minor tines coming off and with a distal upward curvature. It is similar to the specimens from Madrid in that the bifurcation of the brow tine and main beam is situated very low above the burr and that they make a very obtuse angle (Fig. 6).
The middle-sized to large deer like Metacervoceros pardinensis, Praeelaphus perrieri, and Axis? perolensis (Fig. 6 (1) have branching antlers with three or four points (Heintz 1970). The latter differs from the fallow deer from Madrid in having a much higher bifurcation (Fig. 6). Azzaroli (1992) named the species Pseudodama lyra, and the holotype is a frontal with both complete antlers with tree points from Ponte a Elsa. The antlers are of a general primitive type, which occurs also in other species of this age that are no fallow deer, and we are not aware of any associated dental or postcranial specimens that allow the confirmation of the generic classification. It differs from the deer from Madrid in having three points and a high first bifurcation with a sharp angle between brow tine and main beam. In these features, it is similar to D. rhenana.
The Dama-like deer are a very large group and include species with three points. This is the case with Dama rhenana, Dama philisi ( Fig. 7 (1) and (2), Dama philisi valliensis (Heintz 1970), which differ from the deer from Pedro Jaro I and Orcasitas in having a high first bifurcation (Fig. 8) with a sharp angle.
The type of Dama farnetensis has three points on one side and four on the other, differing thus from the deer from Pedro Jaro I and Orcasitas. The species has a first bifurcation with an obtuse angle ( Fig. 7 (5)) and in some specimens a bifurcation that is proportionally as low as in the deer from Madrid (Fig. 8).
We do not know the number of points of the antler of Dama vallonnetensis in its type locality Vallonnet, but this species has antlers with a bifurcation as low as in the deer from Madrid (Fig. 8) and with a similar very obtuse angle ( Fig. 9 (1)). The deer from TE9 was assigned to Dama vallonnetensis (Rosas et al. 2001;Van der Made 2013). This antler has a long main beam, which becomes flattened and still more distally bifurcated. It has three points, but it is similar to those from Madrid in the shape and position of the first bifurcation.
The main beam of D. roberti from Pakefield becomes flattened at some distance of the burr (Fig. 8 (9)), as the antler from TE9, but there is no second bifurcation. In a specimen from Soleilhac (Fig. 8 (10)), there is a similar flattening and then a narrowing and the main beam ends in a single point. This is a two-pointed antler, but the specimens from Pakefield and Soleilhac differ from those from Pedro Jaro I in having a marked flattening. Antlers of this species from Soleilhac, West Runton, and Bacton assigned to this species ( Fig. 9 (2, 3)) have a bifurcation with a low position above the burr and a variable angle. The main beam of the antler from Soleilhac is straight, and the ones from Pakefield are only slightly curved, but the antler from Pedro Jaro I curves strongly outward and then backward again.
The remaining species of Dama differ from the deer from Madrid in having palmate antlers (Figs. 7 (8,9,11) and 9 (6, 7); Melentis 1966a, plate III Fig. 4). While D. peloponesiaca and D. aff. peloponesiaca ( Fig. 9 (6, 7)) tend to have the bifurcation in much higher poisitions, relative to the size (DAPb) of the antler than the deer from Madrid (Fig. 9). Dama clactoniana and Dama dama have the first bifurcation in a position similar to that of the deer from the Manzanares valley (Fig. 8). The curvature of the main beam in the deer from Madrid is similar to that in the living fallow deer, in curving outward and then again more parallel to the median plane. Dama clactoniana and D. dama have the origin of the middle tine where the main beam curves backward and the origin of the back tine close to the place where the antler curves upward and well after the place where the main beam starts to widen into what becomes the palmation. In the specimen from Pedro Jaro I, there is no middle tine at the position where the antler curves backward and the preserved pard does not include a back tine, palmation, or upward curvature. The preserved part of the main beam is long, and if  with species and samples of Dama-like deer with branching antlers (above) and with palmate antlers (below). Provenance of data as indicated in Table 4. Approximate temporal distribution of these species there would be a second bifurcation, the resulting Hm or Hback would be > 58 cm in a straight line and > 62 cm along the curvature and much higher than in species with a similar low first bifurcation (Figs. 10 and 11). The brow tine is large in all specimens from Pedro Jaro I and Orcasitas; it is long and has a large diameter. It is initially horizontal and then curves up very much and even backwards. The point may be directed medially ( Fig. 5 (4a,  4c)). Also, Dama roberti has such large brow tines in Pakefield ( Fig. 7 (7)), Soleilhac (Figs. 7 (10) and 9 (4, 5)), and Bacton ( Fig. 9 (2)) and this is also the case in Dama vallonnetensis ( Fig. 9 (1)). Dama farnetensis has a long brow tine, but it is directed more outwards and less upwards (Fig. 7  (5)). All these species tend to have low first bifurcations. The palmate species also have low first bifurcations, but their brow tines tend to be smaller, shorter, and with smaller diameters at the base. This is so in Dama peloponesiaca ( Fig. 9 (6) is just one out of many specimens) and Dama aff. peloponesiaca (Fig. 9 (7)). Dama clactoniana (Fig. 7  (9)) and D. d. geiselana (Fig. 7 (11)) may have large brow tines, but the upward curved part is not as long as in Pedro Jaro and Orcasitas. Recent Dama dama tend to have much smaller and less cuved brow tines and Dama mesopotamica has the brow tine very reduced or completely lost ( Fig. 7 (8)) and the first large tine that can be seen is homologous to the second tine in Dama dama.
Two scapulas from Transfesa were too small for Megaloceros matritensis (Van der Made 2019, Fig. 18). The best preserved specimen ( Fig. 12 (2)) has a tuberosity that is pointed in lateral view, as in Dama, with a nearly straight anterior edge, and slightly hooked in distal view. The articular surface is slightly flattened at the lateral side. Lister (1996) used these features to distinguish between Cervus and Dama. These specimens have the right size for Dama. In side view, the scapulas have a constriction just above the articulation. In Dama and the specimens from Transfesa, the diameter there (DAPn) is relatively less compared to the other measurements, than in Cervus elaphus, Haploidoceros, and Megaloceros matritensis from Orcasitas (Fig. 12). The transverse diameter at this place is not clearly different between the species. This morphological feature confirms that the scapulas belong to Dama. A radius from Transfesa ( Fig. 12 (1)) is smaller than the radii of Cervus elaphus, and its size fits Dama (Fig. 12). The edge between the facets for the lunar and ulnar is angled as in Dama (a feature  Lister 1996). A proximal metatarsal from Arenero de Manuel Soto is of cervid morphology and the approximate size of a fallow deer (Fig. 12). The proximal articular surface is damaged. If these remains belong to D. celiae, they suggest a large body weight, comparable to that of D. clactoniana and (on average) more than D. d. geiselana and considerably more than the living Dama dama and D. mesopotamica, which weigh 35-80 kg and 70-140 kg, respectively (Wilson & Mittermeier, 2011).

Taxonomic discussion
The origin of the pedicle is not so far posterior as in the Capreolinae. The main beam is much longer than in the Muntiacinae and primitive Miocene species of deer, which are smaller (as seen by the diameter of the pedicle or estimated skull width). This leaves the Cervinae. The low first bifurcation (compared to other measurements of the antler) is well documented in the specimens from Madrid. It is lower than in most of the European Cervinae (Haploidoceros, Eucladoceros, Croizetoceros, Arvernoceros, Praeelaphus, Metacervocerus, Cervus) and living species from Asia, Axis, Rusa, Elaphurus, and Rucervus (with the exception of R. eldi). Such a low first bifurcation occurs only in the Cervinae in some species of Dama (Fig. 7), Megaloceros, Megaceroides, and Rucervus eldi (or Panolia eldi). The species of Megaloceros and Megaceroides are much larger (as can be seen in the diameters of the pedicle or the estimated width of the skull). Rucervus eldi (Fig. 6 (3)) is different in that the main beam of the antler curves much upward in its distal part and in that it has very small points situated on the posterior side of the main beam. These comparisons show the deer from Madrid to be different from all Pleistocene deer of Europe, save for some species of Dama. The specimens from Pedro Jaro I and Orcasitas have the perlation of the burr and the regular shallow groves on the brow tine and main beam as in Dama (Lister 1996) and also the parallel pedicles. They share the low first bifurcation with Dama vallonnetensis, D. roberti, D. clactoniana, and D. dama, as well as D. farnetensis in part.
The deer from Madrid does not have a second bifurcation closer than 58-64 cm to the burr, while at this place, the diameters of the main beam decreased and started the formation of a point. D. clactionana and D. dama have the second bifurcation at 10-30 cm from the burr (Fig. 10). Dama vallonnetensis from type locality Vallonnet is not known in this respect, but from Atapuerca TE9 (Fig. 8 (6)), a second bifurction is known. In TE9, it occurs about 60 cm from the  Table 4. The lines indicate average proportions burr, but some 20 cm lower at about 40 cm from the burr, the main beam starts to widen (in antero-posterior direction). There is no sign of this in the specimens from Pedro Jaro I. There is also no sign of a palmation, which occurs in Dama clactoniana,D. dama,D. mesopotamica,and D. peloponesiaca (Figs. 8 (8,9,11) and 9 (6, 7); Melentis 1966a, plate 3, Fig. 4). There is no sign of a flattening of the distal part of the main beam as seen in D. roberti (Fig. 7 (7, 10)).
We can reasonably assume that the deer from the Manzanares valley is different from the other known species of Cervidae. Therefore, we give this species the new name Dama celiae. A reconstruction of how a male of the species may have looked like is given in Fig. 13.

Evolution of the fallow deer
In a broad sense, the evolution of the fallow deer has been seen as the increase of complexity of the antler, with an increase from three points (as in Dama rhenana and D. philisi) to four points (as in D. nestii) and then part of the antler developing into a palmation. However, to know more precisely how this evolution occurred, it is necessary to know which tine is which and what happened to each tine.
The different species of fallow deer with branching antlers have mostly three or four points (or tines coming off the  Table 4. For Pedro Jaro I, the minimum height is given in the hypothetical case that it would have had a "back" tine  Fig. 9), while some later species have again three tines. Breda and Lister (2013) discussed several scenarios of the evolution of the fallow deer and mentioned the possible reduction of the middle tine, and that the type specimen of D. farnetensis has four tines on the right side and three on the left side. The second tine on the right side (B) is small, and on the same place on the left antler, there is no tine at all. So, the left antler has tines on positions A, C, and D. Biometry and morphology provide some additional arguments for the identification of the middle tine.
The position of a middle tine is not random, but is proportional to the rest of the antler. As we can see in Figs. 6 and 7, the position of the brow tine varies much between species, but within each species, the variation is much more limited. We can observe that there is a tendency for the successive species to have the brow tine in progressively lower positions. This is a common evolutionary tendency in the Cervinae and has been documented for Dama (Van der Made 1999; Van der Made et al. 2016Made et al. , 2017, Megaloceros, and Sinomegaceros (Van der Made & Tong 2008;Van der Made 2015 and is also known from Megaceroides (Praemegaceros), while in Eucladoceros there is no clear tendency, but the different species differ in this feature (Van der Made & Dimitrijevic 2015). Though there are fewer data on the position of the middle tine (B) than on the brow tine (A), in successive samples or species, it also tends to originate in a progressively lower position. This is documented in Megaloceros (Van der Made & Tong 2008), but it occurs also in Dama. The positions in which the different tines originate from the main beam are not independent but are related: if the brow tine originates lower, the middle tine also originates lower (compare Figs. 8 and 10). The antler of Dama vallonnetensis from Atapuerca TE9 has suffered post-depositional deformation, but the approximate position of the second bifurcation could be estimated and is indicated in Fig. 10. Given the low position of the brow tine in TE7 and TE9 (Fig. 8), we would expect the position of the middle tine to be as low as in Dama clactoniana and Dama dama, or at least as low as in Dama farnetensis. Nevertheless, the bifurcation is much higher than the bifurcation that marks the origin of the middle tine in any of the other species and it seems much more likely that the middle tine is lost and the second bifurcation in TE9 is homologous to the third bifurcation in Dama nestii.
Morphology supports the idea that the middle tine (B) is reduced: the origin of this tine is situated near to the place where the main beam curves from an outward direction to a posterior or upward direction. In any case, the middle tine is not situated more distally than this curvature. Such a curvature is not observed in the specimen from Atapuerca TE9, which could be due to taphonomical processes involving deformation and this may also be the case in Soleilhac. In Pakefield, such a curvature is observed, though it is not very pronounced (Breda & Lister 2013, SI Fig. 3), while in Pedro Jaro I, it is very pronounced. In none of the two cases is there any sign of a middle tine (B) at this place.
A middle tine may occasionally be absent. It is a possibility that by chance, the middle tine is absent in the antlers from Atapuerca TE9, Pakefield, Soleilhac, and Pedro Jaro I, but that it was present in the majority of the individuals of these populations. This is possible, but is it also probable? In the living Dama dama, the middle tine (B) may be lacking. Of 155 measured recent antlers from Doñana, two lack the middle tine (EBD 11-3-89, A5 andEBD 2649 M, S27, 1981), and two shed specimens of the same individual lack the middle tine and brow tine (EBD A44), but have the backward-directed tine at the beginning of the palmation. These four are about 2.5% of the sample, and the two which lack only the middle tine are about 1.3%. If the proportion of antlers with a lacking middle tine was more or less similar in the fossil species, it seems unlikely that all the antlers from Atapuerca TE9, Pakefield, Soleilhac, and Pedro Jaro I, which are sufficiently preserved to expect a middle tine, lack it just for chance.
Morphology, biometrics, and statistics favor the first explanation offered by Breda and Lister (2013): the evolutionary loss of the middle tine (B) in Dama vallonnetensis and D. roberti, but also in D. celiae. The other proposal they made is that the flattening near the end of the antler from Soleilhac and in the antlers from Pakefield evolved into the palmation of Dama clactoniana and the living species. This, however, seems to be a reversal of the evolution that led to the D. roberti antler.
The antler of Dama vallonnetensis from Atapuerca TE9 is complete, though it is deformed (Fig. 8 (6)). The two distal tines (C and D) diverge at a very sharp angle and the area between them is a bony plate till the two tines are at a distance. The divergence of the tines is noted approximately 20 cm below the actual bifurcation and roughly at 40 cm from the burr. Just below the separation, the bony plate is over 10 cm wide. The two tines are also flattened at their base. The posterior tine measures 36.6 × ≈ 24 mm at a short distance above the bifurcation. The sections of these tines become more circular near the tips. These tines have lengths from the bifurcation of 46 and 48 cm. Dama roberti from Soleilhac has a main beam with a length of 57 cm. It becomes wide (up to 67 mm) and flat near the tip and then round again without bifurcating (Breda & Lister 2013). This widening may have started at a similar height as in TE9. It seems likely that the morphology of the antlers of Dama roberti evolved from the morphology of the antler from TE9 by shortening the anterior tine (C) from 48 to about 10 cm and by completely reducing the posterior tine (D). An evolution from the antler of Dama roberti to the one of Dama celiae would involve the continuation of the trend of reduction of tines and the loss of the flattened part.
Many of the antlers of Dama clactoniana do not have a wide palmation but do have several branches. It is easy to imagine the perfectly palmate antlers of Dama dama to have originated from filling in the spaces between these branches by bony plate or by a palmate structure growing from the back of tine C. In the more developed Dama clactoniana antlers (as in the skull from Swanscombe), many long flat tines may originate from the palmation, while in Dama dama, the predominant situation is that the tines originating from the palmation are very short. An evolution from D. roberti to D. clactoniana and D. dama would require that after these tines were reduced (in D. roberti), they started again to grow from the flattened area (in D. clactoniana) and then, when the palmation increased in size, they decreased again (in D. dama). These are repeated reversals.
The antler from Pedro Jaro I does not have a palmation, at least not in the part that is preserved. Could there have been a more distal palmation? The anterior edge of the palmation in Dama dama and Dama clactoniana is smooth, but at the back there are little points or not so little flattened tines. The most proximal of these is separate from the palmation and is tine D. In German, it is called Dorn or Sporn (Ueckermann & Hansen 2002) or Wolfssprosse (Pfeiffer 1999, Fig. 14) and in English "back tine." The base of this tine can be easily and objectively identified, and its distance to the burr (Hback) can be measured as an indication of the beginning of the palmation. Figure 11 shows the position of the "Dorn" relative to other measurements of the antler. It is interesting to note that Dama clactoniana has different proportions from Dama dama and that the palmation starts in a clearly higher position. This is a little surprising, since the brow and middle tines are in a similar position in both species. In any case, even in Dama clactoniana, the palmation starts closer to the burr than what could be possible in Pedro Jaro I.
The most likely scenario is of an evolution of a threepointed (A, B, C) Dama (D. rhenana, D. philisi, D. valliensis) to the four-pointed (A, B, C, D) Dama nestii and then the starting reduction of the middle tine (B) in Dama farnetensis, complete lack of middle tine in D. vallonnetensis (retaining A, C, D), reduction of the most distal tine (D) in D. roberti but retention of a flattened area, and finally a simple main beam with out flattening or bifurcation in Dama celiae (retaining A and C). The tines moved to a lower position in this lineage, and this is particularly so for the brow tine (A). It seems palmation evolved (from the back of C) in parallel to this reduction and was present already in Dama peloponesiaca before D. clactoniana appeared. In these species of Dama, the first bifurcation, and possibly other structures as well, moved toward the burr and the brow tine (A) reduced (in D. dama) or nearly completely disappeared (in D. mesopotamica).

Diversity of fallow deer-sized species in the Middle Pleistocene
Until recently, the predominant idea was that there were few Dama-sized deer in the Middle and Late Pleistocene of Europe. The living fallow deer were considered to have evolved from Dama clactoniana (or D. dama clactoniana), which was preceded by Dama nestii, whether related or not (e.g., Kurtén, 1968;Faure & Guérin 1992;Delpech & Guérin 1996;Van der Made 1999). In general, Dama clactoniana was traced back till the beginning of the Middle Pleistocene or even earlier. With one species at a time, this is a very low diversity of deer of this size, but the idea of a simple lineage or succession of few species started to change (Fig. 1) and various models were published in which species of Dama or of this size were contemporaneous and even sympatric (Figs. 1, 14).

Dama clactoniana
If Dama clactoniana is not a direct descendant of D. roberti or D. vallonnetensis, and if there is no directly ancestral form known from W Europe, then it seems that it evolved outside this area and dispersed into it later. The date of the appearance of Dama clactoniana varies widely according to the authors and is given as Holsteinian (about MIS9-11) (Sickenberg 1965), possibly before the Holsteinian (Kurtén, 1968), zone 21 (750-600 ka, Cromerian; Guérin 1982;Faure & Guérin 1992), and 1.3 Ma (Di Stefano & Petronio 2002). Lister (1986) considered the first appearance of Dama dama (including D. d. clactoniana) to be in the early Cromerian and perhaps Beestonian. Croitor (2006Croitor ( , 2018 considered the antlers from Soleilhac and Pakefield (on which the species D. roberti is based) as juveniles of Dama clactoniana, which implies a first appearance not later than MIS17 (712-676 ka). In the scenario that D. clactoniana evolved from D. roberti (Breda & Lister 2013), the transition must be after Soleilhac (MIS15, 563-621 ka) and older records are discarded. Dama clactoniana is present in Clacton, Swanscombe, and Bilzingsleben (Fig. 14). Flattened tines from Atapuerca TD10 ( Fig. 9 (8)) show the species was present in Spain between 400 and 300 ka, but many other sites with remains of a fallow deer, but without a clear palmation or middle tine, are dubious now. Melentis (1966a) described the deer fossils from Megalopolis, which had been excavated in 1902, and assigned part of them to Cervus (Dama) somonensis. Instead of this name, today, the name Dama clactoniana is used (Sickenberg 1965). The material includes a part of a palmation (Melentis 1966a, Pl. 3, Fig. 4). Sickenberg (1975) named Cervus s.l. peloponesiacus on the basis of material collected from the same basin. This species was ignored for a long time. There is much more material of this species in the University of Athens than described by Melentis and includes also flattened tines, which appear to have originated at the back of a palmation ( Fig. 9 (6b-c)) and many antler bases, with a first bifurcation that is higher than in Dama dama or D. clactoniana (Fig. 8). The latter difference was given importance, and the species D. peloponesiaca was recognized as valid (Van der Made 2011Made , 2013Made , 2015Van der Made et al. 2014Van der Made & Mazo 2014).

Dama peloponesiaca
The age of the older collections is not well known. Melentis (1966a, b, c) indicated the deer and Equus abeli come from a locality near the village of Mussakla, while the rhinos come from a site near the village Issoma Karyon. Of other remains (Sus scrofa, Hippopotamus, Bovini), the exact provenance was not indicated. The Equus is of the caballoid type, and this type appeared around 600 ka in Europe. The Cervus elaphus is smaller than in Voigtstedt and Süssenborn, pointing to a similar maximum age (for size changes in Cervus elaphus, see Van der Made et al. 2014. The fossils described by Sickenberg (1975) come from different sites in the Marathousa Beds, and the faunal list he gave includes Mammuthus meridionalis, Praemegaceros verticornis, and Bubalus marathousae. The proboscidean would suggest an age of 700 ka or older. Bubalus murrensis is known to have dispersed to Europe during MIS11, 9, and probably 5 (Von Koenigswald et al. 2019). Alternatively, Bubalus marathousae is the same as Bos galerianus or Hemibos galerianus Martínez-Navarro & Palombo 2004), implying that Bubalus dispersed even earlier to Europe and that the age of at least part of the Sickenberg collection may be older than MIS11. Material recently excavated was described as "Cervus" peloponesiacus, and its age was assumed to be early Middle Pleistocene (Athanassiou 2018). The Marathousa member was deposited during various 100ka eccentricity cycles covering a period from at least 900 to 300 (Van Vugt et al. 2001;Okuda et al. 2002).
A similar but larger species is present in Azokh level V, dated to around 300 ka or less and correlated to MIS9 (Van der Made et al. 2016;Fernández-Jalvo et al. 2016), perhaps MIS8 is also possible for this level (Fig. 14). Fallow deer of this type may have lived for a long time in SE Europe or the Middle East. Bonifay (1967) named Euctenoceros mediterraneus on the basis of material from Lunel Viel. For a long time, this species was ignored, until much more material was published and it was placed in a different genus (Croitor et al. 2008). Now it is recognized from several other localities including Preresa in Manzanares Valley (Fig. 14;Croitor et al. 2008;Sanz et al. 2014;Van der Made & Mazo 2014). This is a species which is not closely related to Dama, but it is of similar size. Its first appearance is either at Lunel Viel or Igue des Rameaux, which have estimated ages of 350-300 ka, and the last appearance is at Cova del Rinoceront, which is slightly older than 83.8 ± 5.9 ka (Sanz et al. 2014).

Dama mesopotamica
This species has been proposed to be a descendant of Dama clactoniana (Di Stefano 1995, 1996, and this seems plausible. The oldest dated and certain Dama mesopotamica is from Tabun E, dated to 213 ± 46 ka and correlated to MIS7, while still older remains are of uncertain specific identification (Di Stefano 1995). Di Stefano (1996) named the subspecies Dama clactoniana mugharensis, based on fossils from Tabun E, and classified the living mesopotamian fallow deer as D. clactionana mesopotamica. Here, we prefer to classify D. clactionana and D. mesopotamica as different species.

Dama vallonnetensis-roberti-celiae
The long lineage leading to Dama celiae was found in Spain in localities of different ages (Atapuerca TE7, TE9, Chaparral, Cueva Victoria, Pedro Jaro I, Orcasitas), and its presence may have been continuous. The impression is that this was a widespread lineage living in what is now the UK, France, Italy, and Spain, but that the geographic distribution contracted, with the last records being Spanish. However, this changes if D. celiae is recognized elsewhere. Dama clactoniana was described from Aridos, just outside the Manzanares valley (Soto, 1980). The antler has the beginning of a large brow tine, but it is not known whether there was a middle tine or palmation. Fallow deer from Torralba and Ambrona have been classified as Dama sp., Dama cf. clactoniana, and Dama cf. dama (see Sesé & Soto 2005). They have large upward-curving brow tines, but the distal part of the antler does not seem to be known. It is of interest to note that Sardella et al. (1998), Di Stefano and Petronio (2003), and Mancini et al. (2006Mancini et al. ( , 2008 believed that "Pseudodama" survived till about 450 ka, because of antlers with large upward-curving brow tines. Again, the distal part of the antlers is not known. Dama clactoniana is cited from Arago. There are several specimens (LPTUP F13/FEJ12M/341, I18/IJAJ/36693670) with brow tines that recall those of Dama roberti and D. celiae, but we do not know whether the middle tine or palmation is documented there. There is a large antler of Dama from Mosbach (IQW Mosbach 159); perhaps, it is from Mosbach 2, but it is only the basal part.

Number of contemporaneous/sympatric species.
Dama clactoniana was present in Atapuerca TD10 and probably also Atapuerca TG10-11 at a distance of about 220 km, as the crow flies, from Orcasitas. These lineages overlapped during 150-100 ky. Though the two lineages may have been ecologically exclusive, they must have lived geographically very close. Both lineages overlap in time with Haploidoceros, which is known of this time from the south of France, while the Spanish records of this genus (in Preresa in the Manzanares valley and Cova del Rinoceront in Catalonia) are coeval to those of Dama dama (in Pinilla del Valle in the north of the province of Madrid). This means a maximum of three coeval lineages in the south of France and North and Central Spain during MIS9 and a prolonged presence of at least two lineages from at least 400 ka till 85 ka. These lineages coincide in time with the D. peloponesiaca lineage in the Middle East and also Greece and afterwards with the D. mesopotamica lineage. This means that there may have been up to four coeval Dama-sized species in western Eurasia. This is a much greater diversity than previously believed.
Recently, the species Megaloceros matritensis was described from terraces of the Manzanares River (Van der Made 2019). It is the end of a lineage, which was believed to have gone extinct when Megaloceros giganteus appeared in Europe, around 400 ka. However, this lineage survived longer and both species were contemporary. The fossils of this species come from the same terrace as Dama celiae. In both cases, a lineage survived longer, resulting in a higher known biodiversity in Spain.

The landscape of the Manzanares valley
The landscape, in which Dama celiae lived, started to form many millions of years earlier. During the Miocene, the Madrid basin was endoreic; it had no connection to the sea, and any runoff was toward the center of the basin, where the water evaporated, forming thick gypsum deposits (Calvo Sorando et al., 2004). Not later than the Early Pleistocene, the Tagus cut back as far as the Madrid basin, connecting it with the Atlantic Ocean (Silva et al. 2017). The lower part of the Manzanares River cuts into the sediments that were deposited in the Miocene: gray clays and gypsum and massive gypsum, forming steep walls. The terraces are mostly on the right side and in the distal part are mostly made of sands, marls, and clays. Pollen preserved in terraces of this age (Gil-García et al. 2018) permits the reconstruction of the vegetation. Close to the river grew willow (Salix), alder (Alnus), and elm (Ulmus), as well as birch (Betula), hazel (Coryulus), chestnut (Castanea), and evergreen oak (Quercus). In oxbow lakes or wet areas on the floodplain, there were Epilobium, Typha, Myryophillum, Ranunculaceae, Juncaceae, and Nymphaeceae. Pinus and Cuprecsaceae grew in the higher and drier areas like the previous terraces and Miocene sediments. Steppe herbs, such as Artemisia, grew on the higher flats. Gypsophile plants grew on the areas where the Miocene gypsum was exposed. A large part of the Madrid basin had similar environments along the Jarama, Henares, Tajuña, and Tajo (Tagus) rivers and the high flats between them.
The species of large mammals that were contemporaries of Dama celiae (Fig. 3) include the ubiquitous species Elephas (= Palaeoloxodon) antiquus, Equus ferus, and Bos primigenius. They are common in sites in the Manzanares and Jarama valleys, such as Áridos 1 and 2 (Villa 1990;Yravedra et al. 2010Yravedra et al. , 2019b and Tafesa/Transfesa (Sesé & Soto 2002 a, b;Mazo 2010;Yravedra 2010;Yravedra et al. 2019b), but also in sites of similar age from the Central Iberian Peninsula, such as Torralba (Shipman & Rose 1983;Villa 1990) and Ambrona (Shipman & Rose 1983;Villa et al. 2005). The species that lived in the Manzanares valley at the time of Dama celiae (Fig. 3) used different parts of the habitat of the Manzanares valley. The wild boar Sus scrofa, auerochs Bos primigenius (very well represented by fossils), and caballoid horses (Equus torralbae or E. ferus torralbae) may have used predominantly the floodplains, while bisons (probably Bison schoetensacki) and Equus hydruntinus may have used predominantly the higher and drier areas between the river valleys. The steppe rhinoceros Stephanorhinus hemitoechus has recently been interpreted to be less of a grazer than its vulgar name suggests (Van Asperen & Kahlke 2015) and may have used the floodplains and also the open areas between them. The straight-tusked elephant Elephas antiquus is known from many fossils from the Manzanares terraces and was a prominent member of the fauna of that time. A lion (Panthera leo/spelaea) was part of this fauna and possibly also a small wolf (Canis mosbachensis?). Several of these species appear in Fig. 15, a reconstruction of the environment, close to the river, where all species may have occurred, if only for drinking. Which of the different environments of the valley and higher plain was used preferentially by Dama celiae?
Not only Dama celiae, but also Megaloceros matritensis was only recently recognized as a distinct species. Both come from the same area and date to the period between 365 and 295 ka. Both are end members of a lineage; one lineage was believed to be extinct several hundred thousands of years earlier, and the other was believed to have evolved into a different species. There are two possible explanations for the fact that these species are not known from other areas in Europe: they became known only recently and therefore have not yet been detected in other areas, or their distribution had become very restricted. The time around 300 ka is particularly well known from many fossil localities in Europe, and in view of this, it is strange that these species were not detected earlier. Dama clactoniana, Cervus elaphus, and Megaloceros giganteus are common and well known from localities of this age, but the first is not known from the Manzanares valley and the latter two species are only known from the younger terraces. A similar case is Haploidoceros mediterraneus, which is also known from a restricted area from the south of France to Spain. Though already published in 1967 (Bonifay 1967), it was not detected from other parts of Europe during more than half a century. Perhaps, the environments of SW Europe or perhaps of Mediterranean Europe were so different from the rest of Europe to have three endemic species of deer.
Megaloceros matritensis is known from ten localities or levels, and Dama celiae is known from two or perhaps four localities in the Manzanares valley and must have been common elements of the fauna of that time and place. This might suggest that these two species were adapted to the particular environment of this area and were rare in environments better suited to D. clactoniana, C. elaphus, and Megaloceros giganteus. Megaloceros matritensis has many peculiar masticatory adaptations, which raised the question, whether they might be adaptations to particular environments of the Madrid basin (Van der Made 2019), such as the areas with steppic or gypsophyle plants. We do not know the masticatory apparatus of Dama celiae and cannot know its masticatory adaptations, but their co-occurrence suggests that it might have used the peculiar habitats in the Madrid basin as M. matriteneis. Dental wear or stable isotope analysis might contribute information that helps to resolve some of these questions.

Acheulean activities
Pedro Jaro I and Orcasitas add up to the not-very-rich list of Middle Pleistocene open-air sites that have provided lithic industries associated with faunal remains ). These two unpublished assemblages, preserved in the MAN, can be described as Large Flake Acheulean lithic industries on flint (Tables 2 and 3, Fig. 4), and were developed in a moment of expansion of the Acheulean technocomplex throughout the peninsular interior. In the Manzanares valley, this techno-complex is characterized by the hegemonic use of flint, which in Pedro Jaro I and Orcasitas is over 96%, with residual presence of quartzite and anecdotical of quartz, raw materials mainly used for shaping large cutting tools (LCTs). The preference for flint in the industries of the Middle Pleistocene of the Manzanares valley, with a generalized use of large flakes for the elaboration of macro-tools, is a singular characteristic within the panorama of the peninsular Acheulean where the macro-tools are mainly made of quartzite. The selection of flint represented in the lithic assemblages does not seem to be justified in the fluvial loads of the river Manzanares as those consist mainly of quartz. Also, the sizes of these large flakes do not fit the percentiles of the flint located in the gravel and cobble bars that constitute the fluvial loads of the river Manzanares (Goy et al. 1989). They would more likely come from Miocene outcrops found in the interfluvium of the rivers Manzanares and Jarama, about 10 km away from the concentration of sites in the Manzanares valley (Bárez et al. 2016). In Pedro Jaro I and Orcasitas, where the LCTs are well represented, the selection of flint supports could be determined by functional criteria with a macro-tool that presents some variability: pointed handaxes with straight edges, cleavers on flake, heavy-duty tools, and trihedrals. Nevertheless, both in the pieces with little and in those with intense shaping, the search for pointed ends predominates. This characteristic can be observed among the handaxes, in the tools shaped on large flakes, and in the frequent trihedral picks of Pedro Jaro I. The shaping of points is sometimes carried out by selecting supports with already preconfigured morphologies, as can be seen in the handaxe preforms and in several trihedrals from Pedro Jaro I. LCTs with re-shaped tips are also present. In general, those characteristics are shared with other lithic assemblages of the same terrace, such as Tafesa (Baena & Baquedano, 2010) and Transfesa. In the management of raw materials in Pedro Jaro I and Orcasitas, the use of recycled supports is also detected, again an element that can be extended to many deposits in the Manzanares valley, as well as the elemental reduction processes that coexist with other organized of discoid and centripetal types.

Interaction with humans?
The finds from Pedro Jaro I and Orcasitas analyzed here have not been retrieved from a systematic excavation, and therefore, the available information is limited. Both lithic artifacts and fossils have been recovered, but the way they are associated is not known. Only 11% of the lithic artifacts from Pedro Jaro I show edges and surfaces with intense rounding. The faunal remains of the two sites do not present alterations related to fluvial transport such as abrasion, polishing, or rounding on bone surface, so fluvial transport was null or played a minor role in the formation of the bone assemblage of both sites. No carnivore tooth marks have been documented on any of the faunal remains at the two sites. However, cut marks have been documented that evidence anthropic activity on a rhinoceros femur (probably Stephanorhinus hemitoechus) from Pedro Jaro I (Yravedra et al. 2019a), and in addition, ten cut marks have been found on a mandible of Equus cf. ferus from this site. Seven cut marks have been found on the external face of a fragment of a medium-sized deer rib from Orcasitas, attributable to Dama celiae or Megaloceros matritensis (Fig. 16). In any case, medium-sized ungulates, like most cervids, fall within the trophic spectrum of humans and D. celiae was a meat resource that was exploited by the Middle Pleistocene humans with Acheulean technology in the same way as other medium-sized ungulates. At this time, spears existed already as several were found at Schöningen (Thieme 1997). These specimens were made of the stem of small trees of Picea and Pinus (Thieme 1998), and Pinus grew not far from Pedro Jaro I and Orcasitas (Gil-García et al. 2018). In their features, these spears resemble those for females in Olympic contests (220 m long, 600 g weight) and a copy has been tested. It could be used to throw straight, hitting with great force, or long distance (> 60 m) with a curved trajectory. With an average impact speed of 85 km/h, the spear penetrated 23 cm in a gelatine block with a consistency resembling the body of an animal (Rieder 2000). Such spears probably formed part of the Acheulean culture, while propulsors, arrows, and bows form part of the Upper Paleolithic (Cattelain 1997) and were not present at this time. An alternative possibility is the use of lithic spear points for hunting, which are documented from other Middle Paleolithic sites (Shea 1988;Shea et al. 2001;Villa et al. 2009), and we have observed some Mousterian Points and convergence scrapers in Pedro Jaro I.
We cannot be sure whether the cut-marked rib from Orcasitas derives from a hunted or a scavenged individual. However, the presence of cut marks (which are associated to filleting) and the absence of carnivore activity in the site suggest hunting by early Neanderthals like in other open-air Middle Pleistocene localities such as Cuesta de la Bajada (Domínguez-Rodrigo et al. 2015), Estanque Tormentas (Yravedra et al 2019a, b), and Tafesa (Yravedra 2010). The technique used in the hunting of these animals could be using spears like those from Schöningen (Thieme 1997) or lithic spear points similar to other regions with mousterian points or convergent scrapers (Shea 1988;Villa et al. 2009); however, only detailed micro wear studies could allow moving forward on this line of interpretation.

Conclusions
The study of fallow deer fossils from the valley of the Manzanares led to the following conclusions: -The fossils from the sandpits of Pedro Jaro I and Orcasitas belong to a new species, which we call Dama celiae. -This species is characterized by two-pointed antlers with a bifurcation between the brow tine and main beam with a blunt angle and a low position above the burr. The brow tine is large and curves upward. The main beam is very long, directed backwards, and curves outward and then backward again; it has no flattened area as in D.
roberti. -This morphology is the result of a reduction of the points of the antler in the lineage Dama farnetensis-D. vallonnetensis-D. roberti-D. celiae. -Dama celiae lived more or less around 300 ka and was contemporary to Haplodoiceros, which also had twopointed antlers, and the palmate Dama clactoniana. The diversity of deer of this size was greater than previously believed. -The occurrence of Dama celiae, Megaloceros matritensis, and Haplodoiceros in the Manzanares valley, while they are not known or rare in the rest of Europe, raises the question whether they were endemic and adapted to the particular environment and climate of this area, or whether their fossils have not yet been detected in other parts of Europe. -Dama celiae was also contemporary to the Neanderthals and acheulean culture and probably was hunted by them.

Declarations
Ethical approval Not applicable.

Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes 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/.