Dietary Reconstruction of Pliocene–Pleistocene Mammoths and Elephants (Proboscidea) from Northern Greece Based on Dental Mesowear Analysis

Abstract

Dental wear analyses of extinct animals offer key insights into their dietary preferences and in turn contribute substantially to palaeoenvironmental reconstructions, leading to more accurate interpretations about past ecosystems. This study employs dental mesowear analysis on Pliocene and Pleistocene elephants and mammoths from several localities in Northern Greece (Ptolemais Basin, Mygdonia Basin, Drama Basin, and the Neapolis-Grevena Basin), aiming to classify them into three main dietary categories (browsers, mixed-feeders, grazers) and investigate potential niche partitioning. The method relies on documenting the wear pattern of molar surfaces through angle measurements on the enamel ridges, which reflect the average annual diet of the examined taxon and in turn the annual ecological conditions of the studied area. Prior to the palaeodietary study and in order to ensure the taxonomic attribution of the examined specimens, a taxonomic review was conducted which confirmed the presence of the mammoths Mammuthus rumanus, Mammuthus meridionalis (southern mammoth), and Mammuthus trogontherii (steppe mammoth), and the European straight-tusked elephant Palaeoloxodon antiquus. Dental mesowear results indicate a grazing diet for M. (cf.) rumanus, a mainly browsing diet for M. meridionalis but mixed-feeding to grazing for the subspecies Mammuthus meridionalis vestinus, a grazing one for M. trogontherii, and a wide diet spectrum for P. antiquus, including browsing, mixed-feeding and grazing, depending on the locality. This study expands our knowledge on the palaeoecology of Greek proboscideans and further highlights the importance of mesowear analysis on proboscidean teeth for palaeodietary and palaeoenviromental inferences.

1. Introduction

The fossil record of proboscideans showcases a remarkable evolution, especially during the Pleistocene, which is generally interpreted as a response to environmental shifts and biotic interaction dynamics [1]. These changes were particularly prominent during key climatic transitions, such as the notable climatic shifts over the last 2.6 million years [1] that tremendously affected the evolution of Elephantidae, including its members in the European continent.

The family Elephantidae (genera Stegotetrabelodon, Primelephas, Loxodonta, Mammuthus, Elephas, and Palaeoloxodon) is part of the third main radiation of Proboscidea that was initiated at the end of the Miocene and continued during the Pliocene–Pleistocene [2,3]. The emergence of Elephantidae is generally linked to the major climatic changes towards more open and arid environmental conditions, including the expansion of C₄ plants, namely grasses and sedges [4]. These environmental changes resulted in the reduction of mammutids and gomphotheriids, which in turn “released” ecological niches that were free for elephantids to occupy, if certain anatomical adaptations to a grazing lifestyle were acquired [5]. These modifications included the reconfiguration of craniodental anatomy, transitioning from the grinding-shearing mastication observed in gomphotheriids to a specialized fore-aft power-shearing movement of the jaws during feeding, the shortening of the rostral parts of the cranium and the mandible, the elevation of parietals and occipitals and the cranium in general, the loss of lower tusks, the shift from loph(id)s to lamellae (plates) in the cheek teeth, increased number and packing of the enamel plates in the molars (higher lamellar frequency), the higher crown of the molars (hypsodonty), the thinning and more complex folding of the enamel, and the delayed sequential, oblique eruption of cheek teeth (e.g., [5,6,7,8]).

The elephantid molars are comprised of the root and the crown, the latter formed by dentine plates, surrounded by enamel, and joined together with cement. The plate number depends on the molar position in the dental series, the wear stage, and the taxon. Differentiations in the number of plates, lamellar frequency, enamel thickness and hypsodonty imply dietary differences, as well as differences in the preferred residences [9]. The change in the shape of the molars’ occlusal surface is caused both by the contact between the teeth of the upper and lower jaw during mastication, and simultaneously by the plant parts (fibers and phytoliths) included in the animal’s diet, as well as by dust particles and gravel [10]. The natural properties of the plants consumed by proboscideans tend to wear the molars, and eventually reduce their height, thus reducing their functionality in the long run [10]. Both the pattern and the amount of molar wear differ among taxa and even between individuals of the same taxon. These patterns are affected mainly by two different processes, attrition (tooth-to-tooth contact) and abrasion (tooth-to-food contact), which, although independent, occur simultaneously from the grinding that occurs due to the jaw movement during the mastication process [11].

In general, the dental wear grade of herbivore molars is heavily influenced by their dietary preferences and, therefore, provides crucial information on both the palaeodiets of extinct species and the palaeoenvironment of certain areas at specific time periods [12]. Dental wear patterns reveal the type of food items consumed by an individual/taxon, thus, their study methods constitute an important tool in determining palaeodiets, particularly because, as taxon-independent methods, they are assumed to be unaffected by the tooth morphology variations of the different species [13].

Two main methods are commonly employed, the dental micro- and mesowear analysis. The latter method (originally developed by [14]; see ref. [15] and references therein) involves the macroscopic examination of the worn cusps’ apices (or the worn lamellae in the case of elephantid molars [16]) in order to illuminate the average dietary signal of herbivorous mammals. The method can be applied relatively easily and quickly (even for a large sample of molars) because it requires rather simple scoring/measurements in each molar, in contrast with the more time-consuming microscopic examination performed in dental microwear analysis. Dental micro- and mesowear capture diet information at different time scales. Mesowear corresponds to a longer temporal scale of months–years, providing thus an estimation of the overall diet and of the generalized annual ecological conditions; on the other hand, microwear captures only a diet snapshot of the last days or weeks of the animals’ life (e.g., [14,17,18]). As such, the combination of dental micro- and mesowear (along with isotopic analysis) can provide precise results for the dietary habits of a mammal (and consequently of the studied areas’ palaeoenvironment), therefore allowing the differentiation of short-term (seasonal) and long-term (average annual) diets [18].

The present study employs the dental mesowear angle analysis on proboscidean molars from several Pliocene–Pleistocene localities in Northern Greece in order to illuminate their dietary habits and ecological adaptations. The ultimate goal of this research is to provide insights into the palaeoecology and evolutionary dynamics of proboscideans, correlating dental wear with environmental conditions and possible dietary shifts. Before applying the mesowear analysis and in order to secure the taxonomic attributions of the examined specimens, particularly those belonging to the older collections of the Museum of Geology-Palaeontology-Palaeoanthropology of Aristotle University of Thessaloniki (LGPUT), an updated taxonomic review is provided including brief descriptions and basic metrical comparisons of the specimens.

2. Materials and Methods

2.1. Material and Fossiliferous Localities

The material examined for mesowear analysis consists of molars (m1–m3 refer to the lower molars, and M1–M3 to the upper ones) belonging to the mammoths Mammuthus rumanus, Mammuthus meridionalis and Mammuthus trogontherii, and the elephant Palaeoloxodon antiquus. It originates from several Pliocene–Pleistocene localities of Northern Greece (Figure 1) which belong to the Neogene–Quaternary systems of the Ptolemais Basin (Sotiras, Amyntaio, Pentavryssos), Mygdonia Basin (Apollonia-1, Kalamoto-1), Drama Basin (Philippi, Symvoli), and Neapolis-Grevena Basin (Tsotylion, Kapetanios, Polylakkos, and the surroundings of Siatista). Information on the geology and stratigraphy of the basins and the localities can be found in Appendix A. The majority of the specimens are housed at LGPUT, but several are stored at HPCS, PHP, and AMPG. The Kalamoto-1 material is part of the LGPUT collections, but it is stored at the Palaeontological Exhibition of Kalamoto.

2.2. Methods

Dental terminology used for the description of the specimens follows [19]. Mandibular measurements are according to [20,21]. Dental measurements were taken after [22] with a digital caliper. Measurements include length (L), width (W), enamel thickness (ET), lamellar frequency (LF) and plate number (PN). On the molars where cement is not preserved, a 5 mm estimate was added to the W value in order to be more accurately compared. Because several of the specimens studied are segments of the original specimen, L was not used for the species attribution. Comparative measurements were taken from [6,23,24,25]. In the main text, we provide the provenance, the available chronological information, the taxonomic attribution and basic remarks for each specimen; detailed descriptions are provided in Appendix B.

For the dental mesowear angle analysis, we followed the methodology of [16]. Specifically, mesowear angles were measured on the occlusal surface of each molar’s lamellae, by placing the contour gauge vertically on the occlusal surface of each selected lamella and pressing it against the surface of the tooth that was in wear, along the lingual half of the molar (not along the widest part of the lamellae in the center of the molar). As a result, the lamellae’s valleys were distinctively portrayed on top of the contour gauge as angles (Figure 2). To obtain the most representative measurements possible, the angles from the three central complete enamel loops of each specimen were calculated (with the exception of two molars of M. meridionalis from Tsotylion where worn enamel rings were measured to increase the available sample). The mesial-most or distal-most enamel bands were avoided, because the former are often too worn and the latter not worn enough for mesowear analysis [16].

The angles were then photographed separately from the tooth using a camera stand to prevent parallax errors. The teeth and the contour gauge were photographed using the same camera settings for each photograph. The contour gauge images were digitally measured and analyzed using ImageJ software 1.54g ([26]; https://imagej.net/ij/). Then the mean angle was calculated by averaging the three measurements of each specimen. For mandibles and maxillae bearing both left and right molars (e.g., m3 sin and m3 dex), the average angle value of the two was used. Similarly, for specimens preserving four molars (e.g., m3 sin, m3 dex, M3 sin and M3 dex) we used the average of the four mean mesowear angle measurements to represent the individual. In the case of fragmentary or badly preserved molars, a minimum of two central lamellae was used for calculating the mesowear angle in order to increase the available sample. Each individual was classified as browser (mesowear angle <106°; <10% grass in diet; C₃-dominated diet), mixed-feeder (106–117°) or grazer (>117°; >70% grass in diet; C₄-dominated diet) [16,27]. The results are given in Table 1 and Table 2. Finally, the results were compared to corresponding angle ranges of [11,16].

The results of the mesowear analysis for localities featuring only one or two specimens were statistically compared using z-scores. Z-scores were computed as z = (x − m)/SD, where x is the mean mesowear angle of the examined specimen, and m and SD, the mean value and standard deviation of the comparative sample (with ≥3 specimens), respectively. Box-and-whisker plots and statistical computations were performed with PAST v. 4.04 ([28]; https://www.nhm.uio.no/english/research/resources/past/).

In order to control the degree of dental wear and progression of the molars measured for the mesowear analysis, we estimated the age at death of the individual in African Equivalent Years (AEY) following the dental-wear-based age criteria for the extant African savannah elephant Loxodonta africana [29,30]. We also used the five, equal length, age intervals of [30,31], each representing ~20% of maximum expected lifespan: 1. 0–12 AEY, 2. 13–24 AEY, 3. 25–36 AEY, 4. 37–48 AEY, and 5. 49–60 AEY (Table 2). These intervals roughly correspond to immature, adolescent, early prime, late prime and old individuals, respectively. In fragmentary specimens a broad range of age groups, estimated AEY and age intervals are given.

Most of the studied molars are isolated, i.e., they are not part of complete mandibles preserving both right and left hemimandibles or maxillae preserving both the right and left side. This is rather common in older/historical collections as proboscidean teeth were randomly collected from the surface or from exposed sections. Additionally, in fossiliferous sites proboscideans are usually represented by isolated teeth and generally the representation of proboscideans in the faunal composition is low. Consequently, most of the dental mesowear studies deal with isolated specimens (e.g., [11,16,32,33]) and only few (e.g., [34]) have examined molars of both sides of the jaw.

One limitation that may influence the accuracy of the dental mesowear measurements, particularly those conducted on elephantid molars, is the fact that the molars (both intraspecifically and intrapopulationally) are commonly morphologically variable (thus affecting the configuration of enamel on the wear surface) due to the forces that act on them during their long formation history [35]. In some cases, these forces may result in anomalies (malformations) in dental structure and molar position within the jaw [36] (see also ref. [34]). Additionally, dietary stress including nutritional deficiencies may also lead to dental pathologies (e.g., ref. [37] and references therein), which also in this case may affect the mesowear angle. With the exception of the upper molar KAL-128 from Kalamoto-1 which shows unusual subdivision and arrangement of the enamel wear figures (yet with otherwise normal wear), all molars examined here do not show any evidence of abnormal wear, dental deformations or atypical molar position on the jaw, at least to the severe extent described in the literature, e.g., [34,36,37,38,39,40], that would affect the accuracy of the measurements. As such all described molars are included in the analysis.

3. Taxonomic Remarks

Order Proboscidea Illiger, 1811

Family Elephantidae Gray, 1821

Subfamily Elephantinae Gray, 1821

Genus Mammuthus Brookes, 1828

Mammuthus rumanus (Stefanescu, 1924)

$\phantom{}$

Locality: Tsotylion, Neapolis-Grevena Basin, Western Macedonia.

Age: Precise locality and stratigraphic horizon unknown, possibly Upper Pliocene.

Material examined: Partial right maxilla with M2 and M3, LGPUT-MGPP-VP-04 (previously labeled as MP-04; Figure 3a).

Remarks: The maxilla fragment MGPP-VP-04 was described in detail in [41]. Taking into consideration the low LF, the thick enamel, and the low hypsodonty index, the authors attributed the specimen to Mammuthus sp. (primitive Eurasian morph). Here it is included within M. rumanus for the sake of simplification. For the mesowear analysis, the two distal lamellae of the M2 and the mesial one of the M3 were measured.

$\phantom{}$

Mammuthus cf. rumanus

$\phantom{}$

Locality: Polylakkos, Neapolis-Grevena Basin, Western Macedonia.

Age: Stratigraphic horizon unknown, possibly Upper Pliocene.

Material examined: Left m1/m2 fragment, LGPUT-MGPP-VP-07; right m2/m3 fragment, LGPUT-MGPP-40 (Figure 3b,c).

Remarks: MGPP-VP-07 was originally attributed to Archidiskodon cf. meridionalis [42], while later it was assigned to Mammuthus cf. rumanus [43]. Taking into consideration primitive traits such as the increased ET and the loosely spaced lamellae (low LF), outside the range of M. meridionalis (Table S2), the specimen is indeed closer to M. rumanus. Given its fragmented nature, the molar is tentatively attributed to either m1 or m2, due to the lack of tapering in the distal end.

Similarly, MGPP-VP-40 has also primitive traits such as high ET and low LF, which fall outside the range of M. meridionalis m2 (Table S1). Given the lack of data regarding M. rumanus m2, and the relatively small width of the specimen, MGPP-VP-40 is identified here as either a m2 or a m3. For the purpose of mesowear analysis both MGPP-VP-07 and 40 are treated as M. rumanus.

$\phantom{}$

Mammuthus meridionalis (Nesti, 1825)

$\phantom{}$

Locality: Tsotylion, Neapolis-Grevena Basin, Western Macedonia.

Age: Precise locality and stratigraphic horizon unknown (Lower Pleistocene based on the presence of M. meridionalis).

Material examined: Left M3 fragment, LGPUT-MGPP-VP-16; left m2 fragment, LGPUT-MGPP-VP-17; right m2 fragment, LGPUT-MGPP-VP-43 (previously labeled as MP-Y) (Figure 3d–g).

Remarks: MGPP-VP-16 was attributed to M. cf. trogontherii [43], based on the W/LF value, which indeed exceeds slightly the M. meridionalis upper limit. However, taking into consideration the rather thick enamel (3.4 mm) and the high occlusal angle of 130° (the angle that is formed between the tangent lines on the worn and unworn occlusal surfaces of the molar in lateral view; Figure 3e) that exceeds the upper limit known in M. trogontherii from its type locality Süßenborn in Germany [24] (Figure 3, and references therein), it probably belongs instead to M. meridionalis, a species already present at Tsotylion.

The morphological and metric traits of MGPP-VP-43 (Table S2) allow for an attribution to M. meridionalis, in line with its original attribution [42].

MGPP-VP-17 was attributed to M. meridionalis [43]. Indeed W, LF and ET are within the known range of M. meridionalis (Table S2) and support the allocation to this species.

$\phantom{}$

Locality: Philippi, Drama Basin, Eastern Macedonia.

Age: Precise locality and stratigraphic horizon unknown (Lower Pleistocene based on the presence of M. meridionalis).

Material examined: Right m3 fragment, LGPUT-MGPP-VP-18; left m3 fragment, LGPUT-MGPP-VP-30 (Figure 3h,i).

Remarks: Both MGPP-VP-18 and MGPP-VP-30 were identified as either m2 or m3 due to their fragmented nature [43]. Judging from the metrical and morphological traits, MGPP-VP-18 is more probable an m3 and is tentatively treated as such. The marked tapering in the distal end of MGPP-VP-30 indicates also in this case an m3. Regarding their taxonomy, ET, LF and W of both specimens fall within the ranges of M. meridionalis (Table S3).

$\phantom{}$

Locality: Symvoli, Drama Basin, Eastern Macedonia.

Age: Precise locality and stratigraphic horizon unknown (Lower Pleistocene based on the presence of M. meridionalis).

Material examined: Left M3, AMPG-1964/449 (Figure 3j).

Remarks: Based on its low LF and thick enamel, AMPG-1964/449 was originally attributed to the new subspecies Archidiskodon meridionalis proarchaicus [44]. However, this molar exhibits morphological and metrical similarities to already published samples of M. meridionalis, therefore lacking distinctive features that would allow for a distinction at subspecific level, rendering A. m. proarchaicus synonymous to M. meridionalis [45].

$\phantom{}$

Locality: Kapetanios, Neapolis-Grevena Basin, Western Macedonia.

Age: Lower Pleistocene.

Material examined: Right m3, LGPUT-KAP-1 (Figure 3k).

Remarks: Based on metrical similarities with the typical M. meridionalis from Upper Valdarno (Italy), KAP-1 was attributed to this species and evolutionary stage [46]. Despite that it was originally identified as a left m3 [46], it is identified here as a right one based on the curvature.

$\phantom{}$

Locality: Apollonia-1, Mygdonia Basin, Central Macedonia.

Age: Lower Pleistocene (late Villafranchian/Epivillafranchian).

Material examined: Left M3 fragment, LGPUT-APL-686B; right hemimandible with m3, LGPUT-APL-716; left m3, LGPUT-APL-687 (Figure 3l–n).

Remarks: The specimens were described in [24]. Mandibular and dental traits permit their attribution to M. meridionalis. Based on the presence of more advanced features compared to the typical Mammuthus meridionalis meridionalis from Upper Valdarno, the material was assigned to Mammuthus meridionalis vestinus.

$\phantom{}$

Locality: Kalamoto-1, Mygdonia Basin, Central Macedonia.

Age: Lower Pleistocene (late Villafranchian/Epivillafranchian).

Material examined: Right M1, LGPUT-KAL-128; lower molar fragment, LGPUT-KAL-85; molar fragment, LGPUT-KAL-117 (Figure 4).

Remarks: Mammoth remains from Kalamoto-1, including KAL-85 and an almost complete tusk, were originally studied in [47] and attributed to M. meridionalis. The M1 KAL-128 and the molar fragment KAL-117 are for the first time studied here and show morphological and metrical traits that further support the attribution to the southern mammoth. The unusual subdivision and arrangement of the enamel wear figures in KAL-128 is observed also in other M. meridionalis specimens, e.g., [48] (pl. 3, Figure 4) and [49] (pl. 9, Figure 3).

$\phantom{}$

Mammuthus trogontherii (Pohlig, 1885)

$\phantom{}$

Locality: Sotiras, Ptolemais Basin, Western Macedonia.

Age: Middle Pleistocene based on the presence of M. trogontherii; LGPUT-MGPP-VP-42 originates from the Proastio Formation of the Ptolemais Basin; for the other molars the precise locality and stratigraphic horizon are unknown.

Material examined: Right M2 fragment, LGPUT-MGPP-VP-45 (previously labeled as MP-SOT-S); left m3, LGPUT-MGPP-VP-46 (previously labeled as MP-SOT-6); left hemimandible with m2, LGPUT-MGPP-VP-42 (previously labeled as MP-diadrsin); mandible with right and left m2, LGPUT-MGPP-VP-05; right m3, LGPUT-MGPP-VP-44 (previously labeled as MP-SOT-Z) (Figure 3o–q and Figure 5).

Remarks: MGPP-VP-42 was first reported as Mammuthus cf. meridionalis [50] and was later attributed to M. trogontherii [43]. Indeed, given that W, ET and LF fall within the range of M. trogontherii m2 from Süßenborn (Table S2), it is attributed to this species.

All available measurements (W, ET, LF) for MGPP-VP-45 (Table S1) are in agreement with the attribution to M. trogontherii [43]. Specifically, the LF (6.46) differentiates it from both M. meridionalis (the upper limit of LF is 5.7 for M2) and M. primigenius (lower limit of LF is 7.7 for M2).

In MGPP-VP-46, W, ET and LF fall within the ranges of both M. meridionalis and M. trogontherii, with the PN being the decisive parameter. More specifically, the 18 plates present on the occlusal surface exceed the upper limit of the m3 for M. meridionalis, permitting an attribution to M. trogontherii.

MGPP-VP-05 was previously attributed to P. antiquus [43]. However, the high W and LF values of the specimen (Table S1), fall outside of the corresponding P. antiquus ranges, while aligning closer to M. trogontherii. Moreover, the presence of the medial mental foramina (MMF) in the lingual side of both hemimandibles (for this character see [51]), provides further evidence for the new attribution.

Similarly, MGPP-VP-44 was previously attributed to P. antiquus [43]. The molar is characterized by increased width exceeding the upper limit of the m3 of P. antiquus but within the range of M. trogontherii m3, a species already found at Sotiras. Neither MGPP-VP-44 nor MGPP-VP-05 bear pointed midline sinuses or the “dot-dash-dot” wear pattern of the distal end enamel rings, both distinctive characteristics of P. antiquus [23,45,52].

$\phantom{}$

Locality: Philippi, Drama Basin, Eastern Macedonia.

Age: Middle Pleistocene (based on the presence of M. trogontherii).

Material examined: Left m3, LGPUT-MGPP-VP-21; right m3, LGPUT-MGPP-VP-22 (Figure 3r,s).

Remarks: MGPP-VP-21 was first described by [42] and MGPP-VP-22 by [53]; both were attributed to M. trogontherii. Indeed, W, LF and ET fall within the ranges for M. trogontherii m3 (Table S3) and support the attribution to this species.

$\phantom{}$

Genus Palaeoloxodon (Matsumoto, 1924)

Palaeoloxodon antiquus (Falconer and Cautley, 1847)

$\phantom{}$

Locality: Tsotylion, Neapolis-Grevena Basin, Western Macedonia.

Age: Precise locality and stratigraphic horizon unknown (Middle–Upper Pleistocene based on the presence of P. antiquus).

Material examined: Mandible with left and right m2 and associated left and right M2; LGPUT-MGPP-VP-02 (Figure 6).

Remarks: Morphological traits of the MGPP-VP-02 molars such as the presence of pointed midline sinuses, the narrow shape of the occlusal surface and the dot-dash-dot pattern in the slightly worn lamellae combined with metric traits (ET, LF and W values) within the known range of P. antiquus for the m2 and M2 (Table S2) allow for an attribution to this species. This is also supported by the absence of a medial mental foramen in the mandible (see [51]).

$\phantom{}$

Locality: Pentavryssos, Ptolemais Basin, Western Macedonia.

Age: Precise locality and stratigraphic horizon unknown (Middle–Upper Pleistocene based on the presence of P. antiquus).

Material examined: Left maxillary fragment with M3, PHP/- (Figure 7d).

Remarks: The metrical values of the Pentavryssos molar are within the range of both P. antiquus and M. trogontherii. The attribution to P. antiquus is justified by the presence of midline sinuses and the “dot-dash-dot” pattern at the distal end of the molar which is at an early stage of wear.

$\phantom{}$

Locality: Philippi, Drama Basin, Eastern Macedonia.

Age: Precise locality and stratigraphic horizon unknown (Middle–Upper Pleistocene based on the presence of P. antiquus).

Material examined: Right m2, LGPUT-MGPP-VP-27 (Figure 7b).

Remarks: MGPP-VP-27 was previously attributed to Mammuthus trogontherii [43]. Judging from morphological (presence of pointed midline sinuses, intensively folded enamel of the lamellae) and metrical (narrow crown; the width stands below the lower limit for the m2 of M. trogontherii from Süßenborn) traits, the molar probably belongs instead to P. antiquus.

$\phantom{}$

Locality: Wider region of Siatista, Neapolis-Grevena Basin, Western Macedonia.

Age: Precise locality/localities and stratigraphic horizon(s) unknown (Middle–Upper Pleistocene based on the presence of P. antiquus).

Material examined: Left M2, HPCS-SIA-2 (Figure 7c), left and right M2, HPCS-SIA-12, m3, HPCS-SIA-21, right m3, HPCS-SIA-1, right M3, HPCS-SIA-5, right M3, HPCS-SIA-8, right M3, HPCS-SIA-10.

Remarks: The specimens of the Siatista collection were presented by [54]. The examined molars of P. antiquus share the morphological characteristics of this species that include the pointed midline sinuses, the increased hypsodonty and the narrow lamellae.

$\phantom{}$

Locality: Amyntaio, Ptolemais Basin, Western Macedonia.

Age: early Middle Pleistocene.

Material examined: Maxilla with right and left M3, PHP-AME-011, and associated mandible with right and left m3, PHP-AME-012 (Figure 7a).

Remarks: The maxilla and the mandible belong to the partial elephant skeleton of P. antiquus described in detail in [55].

4. Dental Mesowear Analysis—Results

Dental mesowear angle analysis was performed on 40 molars (17 upper, and 23 lower molars), representing 31 individuals. The dataset includes 4 molars (3 individuals) of M. rumanus, 10 molars (10 individuals) of M. meridionalis, 8 molars (7 individuals) of M. trogontherii, and 18 molars (11 individuals) of P. antiquus.

The dental mesowear results, categorized by locality and taxon are given in Table 1 and illustrated in Figure 8; Table 2 provides the mean mesowear angle and dietary category for each specimen accompanied by the age group and the corresponding age in AEY.

Figure 8. Box-and-whisker plots of mean mesowear angles (°) of Mammuthus spp. and Palaeoloxodon antiquus from several localities of Greece. Red horizontal lines represent the boundaries between the dietary categories (106° for browsers/mixed-feeders, and 117° for mixed-feeders/grazers; [16,27]), black horizontal lines represent the median values, boxes represent the 25 and 75 percentiles (interquartile range), jitters represent each individual specimen’s mean mesowear angle and whiskers represent the maximum–minimum values. Abbreviations: TSO, Tsotylion; POL, Polylakkos; PHI, Philippi; KAP, Kapetanios; SVL, Symvoli; APL, Apollonia-1; KAL, Kalamoto-1; SOT, Sotiras; AMN, Amyntaio; PEN, Pentavryssos; SIA, Siatista. Numbers in parentheses represent the number of measured specimens. Colors in boxes: light blue (Mammuthus rumanus), yellow (Mammuthus meridionalis), red (Mammuthus trogontherii) and green (Palaeoloxodon antiquus). Silhouette images from PhyloPic (phylopic.org) or redrawn and modified from reconstructions (not to scale).

Figure 8. Box-and-whisker plots of mean mesowear angles (°) of Mammuthus spp. and Palaeoloxodon antiquus from several localities of Greece. Red horizontal lines represent the boundaries between the dietary categories (106° for browsers/mixed-feeders, and 117° for mixed-feeders/grazers; [16,27]), black horizontal lines represent the median values, boxes represent the 25 and 75 percentiles (interquartile range), jitters represent each individual specimen’s mean mesowear angle and whiskers represent the maximum–minimum values. Abbreviations: TSO, Tsotylion; POL, Polylakkos; PHI, Philippi; KAP, Kapetanios; SVL, Symvoli; APL, Apollonia-1; KAL, Kalamoto-1; SOT, Sotiras; AMN, Amyntaio; PEN, Pentavryssos; SIA, Siatista. Numbers in parentheses represent the number of measured specimens. Colors in boxes: light blue (Mammuthus rumanus), yellow (Mammuthus meridionalis), red (Mammuthus trogontherii) and green (Palaeoloxodon antiquus). Silhouette images from PhyloPic (phylopic.org) or redrawn and modified from reconstructions (not to scale).

Table 1. Sample means, ranges of individual mean mesowear angles and dietary categories (Gr: grazer; Mf: mixed-feeder; Br: browser) of the proboscidean specimens from the examined Pliocene–Pleistocene localities of Greece. In bold is the dietary classification inferred from the mean mesowear angle.

Species	Locality	n	Mean Mesowear Angle (°)	Mean Mesowear Angle Range (°)	Dietary Category
Mammuthus rumanus	Tsotylion	1	128.5	–	Gr
Mammuthus cf. rumanus	Polylakkos	2	124.3	123.9–124.6	Gr
Mammuthus meridionalis	Tsotylion	3	101.5	95.3–105.9	Br, Mf
Mammuthus meridionalis	Philippi	2	105.7	105.3–106.0	Br
Mammuthus meridionalis	Kapetanios	1	96.9	–	Br
Mammuthus meridionalis	Symvoli	1	104.5	–	Br
Mammuthus meridionalis	Apollonia-1	3	119.2	113.3–123.7	Mf, Gr
Mammuthus meridionalis	Kalamoto-1	3	119.8	115.7–122.7	Mf, Gr
Mammuthus trogontherii	Sotiras	5	115.5	105.4–122.9	Br, Mf, Gr
Mammuthus trogontherii	Philippi	2	125.5	123.7–127.3	Gr
Palaeoloxodon antiquus	Philippi	1	112.1	–	Mf
Palaeoloxodon antiquus	Amyntaio	1	105.1	–	Br
Palaeoloxodon antiquus	Siatista	7	101.1	91.0–117.9	Br, Mf, Gr
Palaeoloxodon antiquus	Tsotylion	1	121.9	–	Gr
Palaeoloxodon antiquus	Pentavryssos	1	100.6	–	Br

Table 1. Sample means, ranges of individual mean mesowear angles and dietary categories (Gr: grazer; Mf: mixed-feeder; Br: browser) of the proboscidean specimens from the examined Pliocene–Pleistocene localities of Greece. In bold is the dietary classification inferred from the mean mesowear angle.

Species	Locality	n	Mean Mesowear Angle (°)	Mean Mesowear Angle Range (°)	Dietary Category
Mammuthus rumanus	Tsotylion	1	128.5	–	Gr
Mammuthus cf. rumanus	Polylakkos	2	124.3	123.9–124.6	Gr
Mammuthus meridionalis	Tsotylion	3	101.5	95.3–105.9	Br, Mf
Mammuthus meridionalis	Philippi	2	105.7	105.3–106.0	Br
Mammuthus meridionalis	Kapetanios	1	96.9	–	Br
Mammuthus meridionalis	Symvoli	1	104.5	–	Br
Mammuthus meridionalis	Apollonia-1	3	119.2	113.3–123.7	Mf, Gr
Mammuthus meridionalis	Kalamoto-1	3	119.8	115.7–122.7	Mf, Gr
Mammuthus trogontherii	Sotiras	5	115.5	105.4–122.9	Br, Mf, Gr
Mammuthus trogontherii	Philippi	2	125.5	123.7–127.3	Gr
Palaeoloxodon antiquus	Philippi	1	112.1	–	Mf
Palaeoloxodon antiquus	Amyntaio	1	105.1	–	Br
Palaeoloxodon antiquus	Siatista	7	101.1	91.0–117.9	Br, Mf, Gr
Palaeoloxodon antiquus	Tsotylion	1	121.9	–	Gr
Palaeoloxodon antiquus	Pentavryssos	1	100.6	–	Br

Both molars of M. cf. rumanus from Polylakkos, MGPP-VP-40 and MGPP-VP-07, reveal grazing feeding patterns with mean mesowear angles (MMA) 123.9° and 124.6°, respectively. Similar feeding traits are noticed in the M. rumanus specimen from Tsotylion (MMA = 128.5°).

The molars of M. meridionalis from Tsotylion, Philippi, Kapetanios and Symvoli reveal a browsing diet, ranging from pure browsing to the browsing/grazing boundary (MMA: 96.9–105.6°). On the other hand, M. meridionalis from Apollonia-1 and Kalamoto-1 shows the highest angles among M. meridionalis from Greece, ranging between 113.3° and 123.7° and revealing a mixed-feeding diet with a significant grazing component, i.e., a diet including mainly grasses and a percentage of softer plant material.

Mammuthus trogontherii provides as a whole higher MMA than M. meridionalis and covers the whole mixed-feeding and most of the grazing spectrum (the average MMA from all localities is 119.2°). In particular, both specimens from Philippi, and MGPP-VP-44, 45 and 46 from Sotiras show a clear preference towards grazing. On the contrary, MGPP-VP-42 and MGPP-VP-05 from Sotiras with MMA of 105.7° and 105.4°, respectively, show marginally browsing to browsing/mixed-feeding traits. As such the Sotiras collection can be separated into two different clusters, one within the grazing range (MGPP-VP-44, 45 and 46) and the other within the browsing/mixed-feeding one (MGPP-VP-42 and 05).

The MMA for Palaeoloxodon antiquus vary from 91° (SIA-12 from Siatista) to 121.9° (MGPP-VP-02 from Tsotylion), i.e., from pure browsing to grazing, with a MMA of 109.2° (mixed-feeding) across all localities. Therefore, P. antiquus from Northern Greece reveals a wide range of dietary categories, depending on the locality, and additionally shows marked intra- and inter-population variation. Overall, the Siatista molars exhibit the lowest MMA, primarily consisting of individuals classified as browsers, except SIA-2 that falls on the boundary between mixed-feeders and grazers. Likewise, the Pentavryssos molar (100.6°) is also classified as a browsing one. The Amyntaio individual (105.1°) falls approximately at the boundary between mixed-feeders and browsers and therefore can be classified as browse-dominated mixed-feeder. On the other hand, the Philippi molar (112.1°) is plotted within mixed-feeders.

Table 2. Examined elephantid specimens with their mean mesowear angle, dietary category (Gr: grazer; Mf: mixed-feeder; Br: browser), classification into the age groups of [29], and corresponding age in African Equivalent Years (AEY) and age intervals according to [30,31]; 1: 0–12 AEY, 2: 13–24 AEY, 3: 25–36 AEY, 4: 37–48 AEY, 5: 49–60 AEY. MGPP-VP specimens are housed at LGPUT.

Species	Locality	Specimen	Tooth/Teeth Measured	Mean Mesowear Angle (°)	Dietary Category	Group [29]	Age (AEY) [30]	Age Intervals [30]
M. rumanus	Tsotylion	MGPP-VP-04	M2 dex, M3 dex	128.5	Gr	XXI	33–35	3
M. cf. rumanus	Polylakkos	MGPP-VP-07	m1/m2 sin	124.6	Gr	–	–	–
M. cf. rumanus	Polylakkos	MGPP-VP-40	m2/m3 dex	123.9	Gr	–	–	–
M. meridionalis	Apollonia-1	APL-716	m3 dex	113.3	Mf	XXV	47–48	4
M. meridionalis	Apollonia-1	APL-686B	M3 sin	123.7	Gr	XX	31–33	3
M. meridionalis	Apollonia-1	APL-687	m3 sin	120.6	Gr	XXVII	52–53	5
M. meridionalis	Kalamoto-1	KAL-128	M1 dex	115.7	Mf	XI	10–14	1–2
M. meridionalis	Kalamoto-1	KAL-85	m	122.7	Gr	–	–	–
M. meridionalis	Kalamoto-1	KAL-117	m/M	120.9	Gr	–	–	–
M. meridionalis	Kapetanios	KAP-1	m3 sin	96.9	Br	XXIV	42–45	4
M. meridionalis	Philippi	MGPP-VP-18	m3 dex	105.3	Br	XXII–XXVI	36–50	3–5
M. meridionalis	Philippi	MGPP-VP-30	m3 sin	106.0	Br	XXIII–XXIV	39–45	4
M. meridionalis	Symvoli	AMPG-1964/442	M3 sin	104.5	Br	XX–XXI	31–35	3
M. meridionalis	Tsotylion	MGPP-VP-43	m2 dex	105.9	Br	XIV–XV	16–17.7	2
M. meridionalis	Tsotylion	MGPP-VP-17	m2 sin	103.3	Br	XII–XIII	14–15.5	2
M. meridionalis	Tsotylion	MGPP-VP-16	M3 sin	95.3	Br	XXII–XXIII	36–42	3–4
M. trogontherii	Philippi	MGPP-VP-21	m3 sin	123.7	Gr	XXIV	42–45	4
M. trogontherii	Philippi	MGPP-VP-22	m3 dex	127.3	Gr	XXV	47–48	4
M. trogontherii	Sotiras	MGPP-VP-45	M2 dex	120.9	Gr	XVI–XVII	19–23	2
M. trogontherii	Sotiras	MGPP-VP-42	m2 sin	105.7	Br	XVI–XVII	19–23	2
M. trogontherii	Sotiras	MGPP-VP-46	m3 sin	122.8	Gr	XXII	36–38	3–4
M. trogontherii	Sotiras	MGPP-VP-05	m2 dex, m2 sin	105.4	Br	XVII	23	2
M. trogontherii	Sotiras	MGPP-VP-44	m3 dex	122.9	Gr	XXV–XXVI	47–50	4–5
P. antiquus	Pentavryssos	PHP/-	M3 sin	100.6	Br	XXV	47–48	4
P. antiquus	Amyntaio	PHP-AME-011, 012	M3 dex, M3 sin, m3 dex, m3 sin	105.1	Br	XXIV–XXV	42–48	4
P. antiquus	Siatista	HPCS-SIA-2	M2 sin	117.9	Gr	XVII–XXI	23–35	2–3
P. antiquus	Siatista	HPCS-SIA-1	m3 dex	105.9	Br	XIX–XXI	29–35	3
P. antiquus	Siatista	HPCS-SIA-8	M3 dex	101.2	Br	XXII–XXVII	36–53	3–5
P. antiquus	Siatista	HPCS-SIA-21	m3	100.2	Br	XXV–XXVII	47–53	4–5
P. antiquus	Siatista	HPCS-SIA-10	M3 dex	97.2	Br	XXII–XXVII	36–53	3–5
P. antiquus	Siatista	HPCS-SIA-5	M3 dex	94.3	Br	XXII–XXVII	36–53	3–5
P. antiquus	Siatista	HPCS-SIA-12	M2 dex, M2 sin	91.0	Br	XVI	19–21	2
P. antiquus	Tsotylion	MGPP-VP-02	M2 dex, M2 sin m2 dex, m2 sin	121.9	Gr	XVI	19–21	2
P. antiquus	Philippi	MGPP-VP-27	m2 dex	112.1	Mf	XV	17.7	2

Table 2. Examined elephantid specimens with their mean mesowear angle, dietary category (Gr: grazer; Mf: mixed-feeder; Br: browser), classification into the age groups of [29], and corresponding age in African Equivalent Years (AEY) and age intervals according to [30,31]; 1: 0–12 AEY, 2: 13–24 AEY, 3: 25–36 AEY, 4: 37–48 AEY, 5: 49–60 AEY. MGPP-VP specimens are housed at LGPUT.

Species	Locality	Specimen	Tooth/Teeth Measured	Mean Mesowear Angle (°)	Dietary Category	Group [29]	Age (AEY) [30]	Age Intervals [30]
M. rumanus	Tsotylion	MGPP-VP-04	M2 dex, M3 dex	128.5	Gr	XXI	33–35	3
M. cf. rumanus	Polylakkos	MGPP-VP-07	m1/m2 sin	124.6	Gr	–	–	–
M. cf. rumanus	Polylakkos	MGPP-VP-40	m2/m3 dex	123.9	Gr	–	–	–
M. meridionalis	Apollonia-1	APL-716	m3 dex	113.3	Mf	XXV	47–48	4
M. meridionalis	Apollonia-1	APL-686B	M3 sin	123.7	Gr	XX	31–33	3
M. meridionalis	Apollonia-1	APL-687	m3 sin	120.6	Gr	XXVII	52–53	5
M. meridionalis	Kalamoto-1	KAL-128	M1 dex	115.7	Mf	XI	10–14	1–2
M. meridionalis	Kalamoto-1	KAL-85	m	122.7	Gr	–	–	–
M. meridionalis	Kalamoto-1	KAL-117	m/M	120.9	Gr	–	–	–
M. meridionalis	Kapetanios	KAP-1	m3 sin	96.9	Br	XXIV	42–45	4
M. meridionalis	Philippi	MGPP-VP-18	m3 dex	105.3	Br	XXII–XXVI	36–50	3–5
M. meridionalis	Philippi	MGPP-VP-30	m3 sin	106.0	Br	XXIII–XXIV	39–45	4
M. meridionalis	Symvoli	AMPG-1964/442	M3 sin	104.5	Br	XX–XXI	31–35	3
M. meridionalis	Tsotylion	MGPP-VP-43	m2 dex	105.9	Br	XIV–XV	16–17.7	2
M. meridionalis	Tsotylion	MGPP-VP-17	m2 sin	103.3	Br	XII–XIII	14–15.5	2
M. meridionalis	Tsotylion	MGPP-VP-16	M3 sin	95.3	Br	XXII–XXIII	36–42	3–4
M. trogontherii	Philippi	MGPP-VP-21	m3 sin	123.7	Gr	XXIV	42–45	4
M. trogontherii	Philippi	MGPP-VP-22	m3 dex	127.3	Gr	XXV	47–48	4
M. trogontherii	Sotiras	MGPP-VP-45	M2 dex	120.9	Gr	XVI–XVII	19–23	2
M. trogontherii	Sotiras	MGPP-VP-42	m2 sin	105.7	Br	XVI–XVII	19–23	2
M. trogontherii	Sotiras	MGPP-VP-46	m3 sin	122.8	Gr	XXII	36–38	3–4
M. trogontherii	Sotiras	MGPP-VP-05	m2 dex, m2 sin	105.4	Br	XVII	23	2
M. trogontherii	Sotiras	MGPP-VP-44	m3 dex	122.9	Gr	XXV–XXVI	47–50	4–5
P. antiquus	Pentavryssos	PHP/-	M3 sin	100.6	Br	XXV	47–48	4
P. antiquus	Amyntaio	PHP-AME-011, 012	M3 dex, M3 sin, m3 dex, m3 sin	105.1	Br	XXIV–XXV	42–48	4
P. antiquus	Siatista	HPCS-SIA-2	M2 sin	117.9	Gr	XVII–XXI	23–35	2–3
P. antiquus	Siatista	HPCS-SIA-1	m3 dex	105.9	Br	XIX–XXI	29–35	3
P. antiquus	Siatista	HPCS-SIA-8	M3 dex	101.2	Br	XXII–XXVII	36–53	3–5
P. antiquus	Siatista	HPCS-SIA-21	m3	100.2	Br	XXV–XXVII	47–53	4–5
P. antiquus	Siatista	HPCS-SIA-10	M3 dex	97.2	Br	XXII–XXVII	36–53	3–5
P. antiquus	Siatista	HPCS-SIA-5	M3 dex	94.3	Br	XXII–XXVII	36–53	3–5
P. antiquus	Siatista	HPCS-SIA-12	M2 dex, M2 sin	91.0	Br	XVI	19–21	2
P. antiquus	Tsotylion	MGPP-VP-02	M2 dex, M2 sin m2 dex, m2 sin	121.9	Gr	XVI	19–21	2
P. antiquus	Philippi	MGPP-VP-27	m2 dex	112.1	Mf	XV	17.7	2

Considering all the examined samples and localities, the results show a grazing diet for M. (cf.) rumanus (although the sample is limited), browsing to mixed-feeding for M. meridionalis but mixed-feeding to grazing for the late representatives from Apollonia-1 and Kalamoto-1, mixed-feeding to grazing for M. trogontherii, and a wide range for P. antiquus covering all dietary spectrum.

In order to statistically examine whether significant differences exist between the dietary preferences of individuals among the localities, with only one or two specimens available, the z-scores were calculated (

Table 3. Calculated z-scores for the mean mesowear angles (°) of proboscideans from Greek localities. When z > |1.96|, the null hypothesis that the studied specimen fits within the variation of the comparative sample can be rejected at p < 0.05; in such cases, the z-score is typed in bold.

	P. antiquus Siatista	M. meridionalis Tsotylion	M. meridionalis Apollonia-1	M. meridionalis Kalamoto-1	M. trogontherii Sotiras
M. rumanus, Tsotylion (MGPP-VP-04)	3.10	4.89	1.74	2.40	1.42
M. cf. rumanus, Polylakkos (MGPP-VP-07)	2.66	4.18	1.01	1.33	0.99
M. cf. rumanus, Polylakkos (MGPP-VP-40)	2.58	4.05	0.88	1.14	0.91
M. meridionalis, Kapetanios (KAP-1)	−0.47	−0.83	−4.18	−6.29	−2.04
M. meridionalis, Philippi (MGPP-VP-18)	0.47	0.69	−2.60	−3.98	−1.12
M. meridionalis, Philippi (MGPP-VP-30)	0.55	0.81	−2.47	−3.79	−1.04
M. trogontherii, Philippi (MGPP-VP-21)	2.55	4.02	0.84	1.08	0.89
M. trogontherii, Philippi (MGPP-VP-22)	2.96	4.67	1.52	2.07	1.28
P. antiquus, Philippi (MGPP-VP-27)	1.24	1.92	−1.33	−2.11	−0.38
P. antiquus, Pentavryssos (PHP/-)	−0.06	−0.16	−3.48	−5.27	−1.63
P. antiquus, Tsotylion (MGPP-VP-02)	2.35	3.69	0.51	0.59	0.69
P. antiquus, Amyntaio (AME-011, 012)	0.45	0.65	−2.64	−4.03	−1.14

). Among the significant differences that the analysis showed are (a) M. rumanus (Tsotylion) and M. cf. rumanus (Polylakkos) from the Tsotylion M. meridionalis, (b) the Kapetanios and Philippi M. meridionalis from those of Apollonia-1 and Kalamoto-1, (c) the Philippi M. trogontherii from P. antiquus from Siatista, and (d) the Tsotylion P. antiquus from that of Siatista.

Most of the studied molars are isolated specimens and only a few preserve both the right and left molars, and/or the upper and lower jaw of the same individual in order to evaluate the difference in wear between them. For these specimens the results show that there is no significant difference in the mesowear angles. Specifically, in the mandible MGPP-VP-05 from Sotiras and the upper molars from Amyntaio, all have mesowear angle range differing less than 5° between sides, depicting a consistent pattern of mastication and an even wear pattern between the two sides of the lower or the upper jaw. Additionally, in all the above specimens the right molar provides a greater mesowear angle. The deviation between left and right molar mean mesowear angles is slight and does not lead to contradictory results about the dietary preferences. A slightly higher angle range is observed in the specimens MGPP-VP-02 from Tsotylion and SIA-12 from Siatista. The angle disparity in these specimens ranges from 5.3° to 12.4°, revealing a preference of these individuals to chew from the left or the right side. The highest fluctuation is observed in the upper molars MGPP-VP-02 from Tsotylion where the difference is ca. 16°, with the right side being more worn.

Regarding differences between the preferred mastication side, two individuals with all four molars were available, MGPP-VP-02 and the Amyntaio partial skull. In MGPP-VP-02, the molars on the left side have a MMA of 114.7°, whereas the right-side ones 129.1°. On the Amyntaio skull the molars of the left side have a MMA of 100.7°, whereas the right-side ones 109.0°. In both these specimens the right-side molars are more worn and present a higher MMA, indicating a preference towards right-side mastication.

5. Dental Mesowear Analysis—Comparisons and Discussion

The dental mesowear angle analysis showed great variation in the feeding habits of the examined Pliocene–Pleistocene proboscideans from Northern Greece, which overall cover the entire spectrum of dietary categories, ranging from pure browsing to pure grazing individuals. MMA range from 91° in SIA-12 of P. antiquus from Siatista to 128.5° in M. rumanus from Tsotylion.

Due to the scarcity of M. rumanus molars in Europe comparative data regarding its feeding habits are limited to two localities, Red Crag (UK) and Montopoli (Italy). In order to expand the database, the two M. cf. rumanus specimens from Polylakkos are treated in the MMA analysis as M. rumanus. Based on dental microwear analysis, the specimens from Red Crag and Montopoli revealed a mostly browsing behavior [13,56] in accordance with the low hypsodonty that characterizes M. rumanus and the generally wet and warm climate during the Pliocene in Europe. In contrast, the Greek M. rumanus molars from Tsotylion and Polylakkos show a grazing diet (however, it should be noted that MGPP-VP-04 is heavily worn and the measurement should be considered with caution).

It is worth noting that Tsotylion and Polylakkos are situated near Milia (Grevena Basin), a locality that documents the mammutid Mammut borsoni and the gomphothere Anancus arvernesis, and is dated to the Upper Pliocene (early Villafranchian; [57,58]). The presence of M. (cf.) rumanus at the nearby sites of Tsotylion and Polylakkos might indicate a comparable age. The isotopic analysis on M. borsoni molars from Milia revealed a browsing diet [59]. A browsing diet is also revealed for both M. borsoni and A. arvernensis based on dental mesowear and microwear analyses from several European localities [11,13]. As such a hypothesis could be that in the potential coexistence of M. rumanus with M. borsoni and A. arvernensis in the wider area of Tsotylion, Polylakkos and Milia, the latter two species could have occupied the ecological niche of the softer plant material, leaving the grassier areas (i.e., those with more abrasive plant material) as the only available ecospace for M. rumanus. This potential trophic partitioning could possibly explain the high values of MMA of M. rumanus in Tsotylion and Polylakkos.

An alternative or complementary scenario that could explain the browsing adaptations of the Red Crag and Montopoli M. rumanus and the more grazing one of the Greek specimens might be related to chronological differences. Despite the fact that most of the M. rumanus-bearing localities are not securely dated, the slightly more derived morphological and metrical traits of the Red Crag and Montopoli mammoths than the M. rumanus specimens from Tsotylion [41], might point to a slightly younger age. Red Crag and Montopoli are dated to 2.6–2.4 Ma [60,61,62,63,64] and therefore their mammoths could belong to some of the last M. rumanus populations that might represent a shift to a more browsing diet that took place during the north-western dispersal of M. rumanus. On the other hand, the likely earlier Greek specimens might have retained the more mixed-feeding/grazing signal of the African primitive mammoths. Isotopic and dental mesowear analyses on Mammuthus subplanifrons from several African sites dated in the interval ca. 6.0–3.5 Ma reveal a mixed-feeding/grazing diet [27,65,66]. Alternatively, local environmental factors, and accordingly dietary adaptations related to the dietary plasticity of proboscideans might be the reason for the different signal obtained for the Greek and Red Crag/Montopoli specimens. More dietary research on primitive mammoths both from Greece and the rest of Europe is certainly needed to provide safer inferences on M. rumanus feeding habits.

The populations of M. meridionalis in Europe covered the entire dietary spectrum, from localities featuring mainly pure browsers to some others consisting solely of grazers, which is confirmed both by dental microwear and mesowear analyses [11,13]. This ecological flexibility is attributed to the large sheer size of this species, a trait that granted access to both grasses and soft leaves from the top of the trees [11], as well as to the prolonged time period during which M. meridionalis was the sole proboscidean in Europe [9], and has survived under diverse climatic conditions in both regional and European scale. It is worth noting, that the observed dietary variation is only present on a local or regional scale, since the localities documenting M. meridionalis have very low intra-population variation [13].

Similar results arise for M. meridionalis from the Greek localities Tsotylion, Kapetanios, Philippi and Symvoli which comprise individuals in the browsing spectrum. On the other hand, the Apollonia-1 and Kalamoto-1 individuals fall into the mixed-feeding and mostly into the graze-dominated category, i.e., their diet incorporates a significant amount of C₄ plants.

Apollonia-1 fauna (biochronologically dated to ca. 1.2 Ma; ref. [67] and references therein) exhibits several distinctive faunal traits, such as the predominance in the bovid assemblage of caprines and early bisons, and the remarkable increase in the relative body size of herbivores, e.g., the large-sized species of Equus, the giant cervid Praemegaceros, and the large-sized ovibovines and bovines [68,69,70]. The presence of the semiaquatic Hippopotamus (present at the roughly coeval and of the same stratigraphic position localities of Ravin of Voulgarakis and Kalamoto-1) coupled with taxa adapted to drier/colder environments such as Soergelia and Praeovibos, indicate moderate habitats for Apollonia-1 [71]. The Apollonia-1 faunal assemblage is dominated by mixed-feeders and grazers, the browsers are notably reduced, and forest-dwelling taxa are rare, overall indicating a mainly open grassy environment [69,72]. Based on dental mesowear analysis and morphometric aspects of the metapodials, the Apollonia-1 bison (Bison (Eobison) cf. degiulli) shows a specialization into more intensive grazing that could correspond to a drier climate and presence of a developed grassy cover with relatively high grit component [73]. A comparable environment can be hypothesized also for the approximately contemporary, geographically adjacent and of similar depositional context (Platanochori Fm) Kalamoto-1. The presence at this site of Hippopotamus antiquus, the European pond turtle Emys orbicularis and of freshwater mollusks [47,74] indicates the existence of permanent freshwater bodies under temperate climatic conditions [75,76].

The more open environment of Apollonia-1 is in agreement with the presence of Mammuthus meridionalis vestinus [24], a subspecies that is proposed to have inhabited savannah parkland with tall trees, shrubs and grasses [77], and its dimensions, larger than the typical morph (Mammuthus meridionalis meridionalis) fit with the general size increase of herbivores at Apollonia-1. Therefore, the higher MMA obtained for the Apollonia-1 mammoth compared to the other M. meridionalis from Greece, which reflect a higher grazing component in its diet, are in accordance with both the environments that is hypothesized that this subspecies inhabited, as well as with the reconstruction of the Apollonia-1 landscape. The available mammoth material from Kalamoto-1 does not allow for a subspecific attribution, yet Kalamoto-1 and Apollonia-1 are thought to be roughly coeval. In addition, there is agreement in the MMA values and in turn of the dietary preferences of their mammoths. Hence, it is plausible that the same taxon is present at both sites.

In the majority of European localities M. trogontherii presents mixed-feeding patterns with a noteworthy preference for grazing on C₄ grasses [11], while the single sample from West Runton (UK) has a mesowear angle of 143° that ranks it among the pure grazers. Similarly, dental microwear analysis reveals in most localities mixed-feeding with only the specimens from Pakefield and Trimingham (UK) bearing patterns typical for grazers [13]. The latter authors also note that the absence of puncture pits reveals a lack of fruit in the diet.

The majority of M. trogontherii from Northern Greece (Philippi and Sotiras) present grazing behavior, matching both the species’ cranio-dental adaptations (shortening of the skull, increased hypsodonty of the molars) to more open habitats and its general trend to such feeding traits [11]. None of the M. trogontherii from Greece shows a pure browsing diet. Two specimens from Sotiras (MGPP-VP-05 and 42) plot at the boundary between mixed-feeding and browsing. The sample from this locality shows a clear distinction in mesowear angles, leading to the identification of two distinct dietary groups (one browsing and one grazing). Considering the uncertain stratigraphic origin of the specimens, the presence of multiple populations, exhibiting different dietary preferences and temporal ranges within the locality, is possible.

Palaeoloxodon antiquus exhibits on average browse-dominated mixed-feeding behavior, similar to today’s African savannah elephants [11]. However, in relatively open, savannah-like palaeoenvironments, such as the Ilford region in UK during the Marine Isotope Stage (MIS) 7, it demonstrates a mixed-feeding-pattern with a greater emphasis on grass consumption [11]. In the Happisburgh Formation, molars of P. antiquus were discovered alongside slate slabs containing plant fossils of Pinus sylvestris (Scots pine), Picea excelsa (Norway spruce), Betula alba (Silver birch), and Alnus glutinosa (Black alder). This flora suggests a preference for browsing among these animals, different from the Pleistocene populations of Waverley Wood, where mesowear angle analysis indicates a mixed-feeding behavior in more open and moist grasslands [11]. Such fluctuations are also confirmed by dental microwear analysis, even though most of this method’s results reveal a general tendency towards grazing [13]. Palaeoloxodon antiquus shows a significant dietary diversity in the various investigated localities [10]. For example, microwear analyses show browsing patterns in Taubach (Germany; MIS 5e), mixed-feeding patterns in UK localities and more grazing patterns in the Megalopolis Basin (Greece; old collection). In the locality Marathousa 1 of the Megalopolis Basin, where a partial skeleton of P. antiquus was discovered within sediments correlated to the glacial MIS 12 [78,79], the isotopic analysis revealed that the elephant subsisted in a C₃-dominated open woodland environment under relatively stable climatic conditions [80], in accordance with the other faunal and floral evidence [81,82] (and references cited in both). Similarly, dental microwear analysis of the P. antiquus molars from Poggetti Vecchi (Italy; ca. 179 ka, MIS 6 [83]), reveal grazing dietary preferences, while the mesowear angle analysis indicates a browse dominated mixed-feeding diet. The combined results for the Poggetti Vecchi specimens reveal a mixed-feeding behavior, with seasonal shifts toward grass consumption, in order to adapt to local environmental changes [32].

Overall, a wide range of MMA and accordingly of feeding habits is observed in the Pliocene–Pleistocene proboscideans from the several localities of Northern Greece. Approximately half of the specimens reveal browsing diet (17 out of 31 individuals), with the rest classified as mixed-feeders (2 individuals) and grazers (12 individuals). Thus, a significant interspecific variation is observed. On the other hand, a high intraspecific variation in a specific locality is limited to two sites, Sotiras (five specimens, yet from unknown stratigraphic horizon) and Siatista (seven specimens, surface findings from several sites). The MMA in the Siatista specimens span from 91.0° to 117.9° (standard deviation 8.8). Accordingly, the lowest angle from Sotiras of 105.4° (MGPP-VP-05) plots as browser whereas the highest one of 122.9° (MGPP-VP-44) as grazer (standard deviation 9.2); however, the small sample size does not allow for safe conclusions.

Data from the rest of Europe reveal a significant difference in mesowear angles and eating habits between P. antiquus and M. trogontherii, which coexisted in the same area (Ilford) during the same period (MIS 7). The mesowear analysis for P. antiquus indicates a browse-dominated mixed-feeding pattern, whereas that for M. trogontherii suggests a grass-dominated mixed-feeding behavior [11]. Therefore, in such cases of coexistence, the only possibility of surviving is for one of them to alter its feeding habits. This is known as niche partitioning and allows several taxa to inhabit the same region by utilizing different parts of the environment. Therefore, it can be reasonably assumed that when P. antiquus and M. trogontherii occasionally coexisted, they presented different feeding preferences, thus avoiding/reducing trophic conflicts. In Νorthern Greece, the only herein studied locality that features both P. antiquus and M. trogontherii is Philippi [note that both M. trogontherii and P. antiquus are reported from Sotiras (present study; [84,85]), however, the sample of P. antiquus was not examined here]. In Philippi both M. trogontherii individuals reveal pure grazing diet (123.7° and 127.3°), while the only P. antiquus specimen is sorted as a mixed- feeder (MMA 112.1°). Yet, the unknown stratigraphic position of the specimens makes any discussion about niche partitioning speculative.

6. Conclusions

Dental mesowear angle analysis comprises a valuable tool in understanding the dietary preferences and behavior of elephantids and therefore contributes to palaeoenvironmental reconstructions and interpretations. The findings presented in this study contribute to the growing body of knowledge on the palaeoecology of Greek proboscideans, emphasizing the need for the application of several methods for more accurately reconstructing their dietary habits.

Considering all samples and localities of Northern Greece, the results show a rather grazing diet for M. rumanus (although the sample is limited), a mostly browsing diet for M. meridionalis (except of the Apollonia-1 mammoth), grazing for M. trogontherii (apart from the browsing cluster of the Sotiras specimens) and a wide range (but mostly browsing) for P. antiquus covering all the dietary spectrum.

Mammuthus (cf.) rumanus from Polylakkos and Tsotylion, and M. trogontherii from Philippi show the highest mean mesowear angle among Greek sample and reveal a grass-dominated diet. On the other hand, P. antiquus from Siatista show the lowest mean mesowear angle and pure browsing preferences.

The southern mammoth from Apollonia-1 and Kalamoto-1 displays the highest mean mesowear angle among M. meridionalis in Greece, which indicate a diet with a significant grass component. This is in agreement with the more open character of Apollonia-1 and the presence in the faunal assemblage of open-dwellers and mixed-feeders/grazers.

The wide dietary range of P. antiquus reflects the ability of the species to exploit a diverse array of resources and indicates the capacity of the species to cope with the vegetational shifts triggered by the climatic fluctuations (glacial/interglacial periods) of the Middle Pleistocene.

The application of dental mesowear analysis in additional specimens/localities from Greece and other European localities will contribute to the enhancement of the comparisons and the statistical analyses and strengthen the interpretations. Together with the application of diverse palaeoecological proxies, it will yield further insights into proboscidean dietary preferences and behavior.