Déjà vu: a reappraisal of the taphonomy of quarry VM4 of the Early Pleistocene site of Venta Micena (Baza Basin, SE Spain)

Patterns of species abundance

In their analysis of the fossil assemblage of VM4, Luzón et al.¹ indicate that herbivorous taxa comprise the bulk of the fauna. Their data, compiled in Table 1, show that herbivore remains represent 94.2% (1492/1578) of NISP and 78.8% (41/52) of MNI values for large mammals. These figures are close to those of VM3, 93.5% (6570/7027) and 84.4% (287/340), respectively (Table 1). A χ² test shows that these differences are not statistically significant (p > 0.3 in both cases). Among herbivores, Luzón et al.¹ indicate that E. altidens is the species most abundantly preserved, both in frequency of remains and number of individuals, followed by cervids, bison, caprines, and megaherbivores (i.e., elephant, rhino, and hippo). This is also the situation in VM3 according to data compiled in Table 1: for example, the NISP value of E. altidens represents 31.8% (124/390) of the remains of large mammal identified in VM4 and 49.6% (2937/5924) in VM3. Although this difference is statistically significant (χ² = 46.408, p < 0.0001), the frequencies of horse based on MNI estimates, 29.3% (12/41) in VM4 and 31.7% (91/287) in VM3, are similar (χ² = 0.096, p > 0.75). The difference based on NISP values seems high, but it falls within the range expected from variations in abundance data from different years for the ungulate prey more common in Serengeti, where the frequencies of Thomson’s gazelle, wildebeest, and zebra fluctuated in the late sixties between 18.9–56.3%, 21.3–42.8%, and 11.1–15.7%, respectively^15,16. Finally, P. brevirostris is the species most represented among carnivores in both assemblages according to NISP values (Table 1), 26.8% (15/56) in VM4 and 30.0% (122/407) in VM3 (χ² = 0.241, p > 0.6), followed by canids, ursids and felids.

The distribution of NISP and MNI values among taxa in VM4 and VM3 was further analysed using an approach based on contingency tables. The table for NISP values shows a significant χ² value (Table 2, left part). This results from some differences in taxa abundance between the assemblages compared, which are reflected in the adjusted residuals: remains of megaherbivores and carnivores (excluding hyaenas) are represented in VM4 by higher frequencies than those expected from a random, homogeneous distribution, while they are underrepresented in VM3. This applies to the estimates obtained for VM4 using the data of Luzón et al.¹ and our own data (Tables S1, S2). The NISP values estimated for P. brevirostris by Luzón et al.¹ suggest a higher frequency of this carnivore in VM4 than in VM3, as indicated by the adjusted residual. However, the abundance of hyaena remains in our dataset for VM4 does not depart significantly from the expectations, as happens in VM3. Given that the database of Luzón et al.¹ includes less than half of the remains of large mammals included in our database (Table 2), this suggests that the high frequency of P. brevirostris reported in VM4 results from poor sampling. The remains of other carnivores are more abundantly represented in VM4 than in VM3. However, it must be noted that a study of 24 dens of the three living hyaenas showed that the abundance of carnivore remains is highly variable, even among dens of the same species¹⁷. The distribution of MNI values among taxa in VM4 and VM3 (Table 2, right part) does not differ from the expectations of a random distribution according to the low χ² value of the contingency table. Only the adjusted residual for megaherbivores, which are slightly over-represented in VM4 according to the data of Luzón et al.¹, is statistically significant, while their abundance in VM3 is slightly lower than expected. Moreover, the probabilities of obtaining in the randomization tests the cumulative χ² values observed for the NISP and MNI values of each species (p < 0.001 and p > 0.97, respectively; Fig. S4) are equivalent to those obtained with their groupings in Table 2.

In summary, the comparison of the faunal assemblages from both excavation quarries (Tables 1, 2) only shows some minor differences in taxa abundance for horse, megaherbivores, and carnivores other than the hyaena, as well as the presence in VM3 of some remains of two small ungulates (a roe deer-sized cervid and a chamois-sized bovid) and two small carnivores (Table 1), which are not reported by Luzón et al.¹. Given their comparatively low number of specimens studied at VM4, it is reasonable to expect that the latter taxa, which are poorly represented in VM3, will also appear in VM4 during future excavations.

Age mortality profiles

Luzón et al.¹ indicate that two megaherbivores, elephant Mammuthus meridionalis and rhino Stephanorhinus aff. hundsheimensis, show frequencies of non-adults that are close to, or even higher than, those of adults, as happens in VM3 (Table 1). However, the low MNI counts for these species in VM4 do not allow to state this: for example, elephants are represented by a juvenile and an adult, which gives a frequency of 50% of non-adults; with a sample size of only two individuals, the 95% confidence interval calculated with a binomial approach for this percentage is 1.3–98.7% (Table 1). In S. hundsheimensis, the frequency of non-adults, 80% (4/5), has also a very wide confidence interval (28.4–99.5%). In three species of medium-to-large sized ungulates, E. altidens, the ancestor of water buffalo Hemibos aff. gracilis and P. verticornis, Luzón et al.¹ report similar frequencies of adults and non-adults, while they indicate that Bison sp. shows a lower frequency of juveniles (Table 1). This is true for horse and deer (58.3% and 42.9% of non-adults, respectively), but Hemibos is only recorded by one adult individual, which means that the percentage of non-adults for this species is not reliable. Luzón et al.¹ calculate the percentage of 33% non-adult bison over a sample of only three individuals, of which one is a juvenile: the confidence interval for non-adults (0.8–90.6%) comprises the frequencies for horse and megacerine deer (Table 1), which rules out their suggestion of a lower frequency of juveniles for this bovid. In contrast to VM4, the abundances of non-adult horse, bison and megacerine deer are similar in VM3 (Table 1), where they are represented by higher MNI counts (which makes their percentages reliable). A similar reasoning can be applied to the claim of Luzón et al.¹ that adults outnumber calves and juveniles among smaller herbivores such as the Ovibovini Soergelia minor, the Caprini Hemitraus albus and the cervid Metacervocerus rhenanus: in these species, MNI counts are very low to calculate reliably the percentage of juveniles (see their confidence intervals in Table 1). In fact, Luzón et al.¹ acknowledge this limitation when they write that “the total number of individuals in each species is too low to draw reliable conclusions on the resulting patterns” and “a prime-dominant, L- or U-shaped mortality profile cannot be clearly discerned”. The situation in VM3 is quite different (Table 1): MNI counts for the two ungulates better represented in the assemblage, E. altidens and P. verticornis, allowed to reconstruct U-shaped attritional mortality profiles (Fig. 2b), which evidenced that the hypercarnivores focused on young and old individuals in the case of large prey^6,7.

Patterns of skeletal abundance

The limitations and inaccuracies cited above result from the small sample analysed by Luzón et al.¹ in VM4 (1578 remains of large mammals of which only 420 could be determined taxonomically and anatomically, compared to 8150 and 6331 remains in VM3, respectively: Table 1). These limitations apply also to their inferences on the skeletal profiles of ungulates. For example, they indicate that species of herbivore size class 2 (50–125 kg: M. rhenanus, H. albus, and S. minor) show biased skeletal profiles, with a predominance of teeth and elements of the forelimb over those of the hindlimb. In VM3, these ungulates also show higher frequencies of teeth than of bones, which has been interpreted as evidence of the transport by P. brevirostris of small-to-medium sized ungulates as whole carcasses to their denning site, where the giant hyaenas fractured the bones for accessing their medullary cavities and this resulted in their underrepresentation compared to teeth^7,8,9,10. In the case of the major limb bones of these species in VM4, the elements of the forelimb (12.9%, 13 bones out of 101 determined remains) are twice as abundant as those of the hindlimb (6.9%, 7 bones), but these percentages do not differ statistically (χ² = 2.028, p = 0.1544), which indicates the effects of poor sampling. In the species of herbivore size class 3 (125–500 kg), Luzón et al.¹ indicate that they are well represented by all anatomical elements (e.g., craniodental elements account for ~ 30% of the remains, while both axial and appendicular elements show frequencies > 20%). This pattern is like the one reported in VM3 for medium-to-large sized ungulates^7,8,9,10. However, Luzón et al.¹ indicate a bias in the disproportionate amount of posterior limb remains compared to anterior limb specimens, which in their opinion contrasts with the more balanced representation of these elements observed in VM3. Specifically, the number of forelimb bones (13.8%, 54 out of 392 bones) is about half the abundance of hindlimb bones (25.3%, 99 bones). This difference is statistically significant (χ² = 16.460, p < 0.0001) because the sample size studied here is larger than in the former case. Moreover, Luzón et al.¹ do not report on the presence of astragali or calcanei, highly mineralised bones well represented in the faunal assemblage of VM3^5,7,9,10. Given that these elements are not absent from VM4 (our own dataset for VM4 includes nine calcanei and 20 astragali out of 78 autopodial bones: Table S1), their inclusion would further increase the frequency of hindlimb bones. On the other hand, they indicate the presence of 25 pelvises, which outnumber all major limb bones except the tibia (in our dataset, the number of pelvises is 16). Compared to the limb bones, the pelvis is a flat anatomical structure composed of three poorly mineralised bones, which uses to be fractured by the hyaenas for accessing its internal nutrients. As a result, it is usually recorded as fragments that preserve the acetabulum⁷. This suggests that the high number of pelvises reported by Luzón et al.¹ probably represent pelvis fragments, which explains in part the overabundance of hindlimb elements. In VM3, E. altidens and Bison sp. are better represented by limb bones than by cranial and axial elements⁸. The living hyaenas do not cooperate to transport large portions of a carcass, which limits the individuals in what they can move¹⁸. Given the size of the carcasses of the adult individuals of these species, which exceeds what even a hyaena as large as P. brevirostris could transport, the overabundance of limb elements of horse and bison indicates the dismemberment of the carcasses and the selective transport to the denning site of the limbs due to their high marrow yields^8,10. In the case of the horse, hindlimbs predominate over forelimbs in VM3, because the femur and tibia provide more marrow than the humerus and radius, respectively, and this resulted in the preferential transport of hindlimbs by the hyaenas to their den⁸. Therefore, the pattern of skeletal representation noted by Luzón et al.¹ in VM4 for the size category of the horse, the species better represented in the assemblage, agrees also with the expectations from VM3.

Figures 3 and 4 show examples of the patterns of preservation of cranial and postcranial remains of ungulates in VM4 and VM3, which are strikingly similar. The data compared above on species abundances and skeletal representation suggest that the small differences outlined by Luzón et al.¹ between both quarries result from random oscillations emerging from the small sample studied at VM4. Surprisingly, they only analyse taphonomically the fossils of large mammals of the years 2005 and 2019–2020 (1578 remains), but do not include those unearthed during the years 2013–2015 (3729 remains). Their inclusion would raise the number of specimens studied to 5338 specimens, a figure more in accordance with the information available for VM3 (the site they intend to compare their results with), which would make their analyses more robust. Moreover, it is difficult to understand why Luzón et al.¹ do include 4219 skeletal remains in the analysis of spatial patterns with random forest algorithms, which allowed them to assign confidently most of these fossils to any of the two bone accumulations identified at VM4, but do not use such data for analysing the patterns of species abundances and skeletal representation.

Patterns of bone weathering

Luzón et al.¹ indicate that 90.8% (1461/1609) of the specimens analysed of VM4 show weathering stage (ws) 0 (bones with no sign of cracking or flaking, preserving their cortical surface intact), which indicates less than 1 year of subaerial exposure before burial¹⁹, while the remaining 9.2% of elements show ws 1 (cracks poorly developed and longitudinally oriented in the long bones: 0–3 years) or ws 2 (outermost concentric layers of bone showing mosaic flaking and deeper split line cracks: 2–6 years)¹⁹. In the case of VM3, the estimates calculated over a substantially larger sample of skeletal remains (4921 bones, updated from Ref.⁹) are the following: 75.9% ws 0, 23.4% ws 1, 6.3% ws 2, and 0.04% ws 3 (bone surface with patches of rough, homogeneously weathered compact bone where all the external, concentrically layered bone has been exfoliated: 4–15 years of exposure)¹⁹. Half of the bones of VM3 classified within ws 1 show shallow split line cracks due to insolation in only one side or in a part or their outer surface, while the other shows ws 0 (Fig. 5d). For this reason, they could equally well have been classified within the latter category, which would raise the frequency of elements with ws 0 to ~ 88%, a figure more in accordance with the data of Luzón et al.¹. A similar proportion of the few remains classified in VM3 within ws 2 display ws 1 in a part or a side. If they are considered within the latter category, this would raise the percentage of bones with ws 1 to ~ 12%, a figure also closer to that of Luzón et al.¹. In summary, these estimates suggest that the skeletal remains of VM3 show a slightly more advanced degree of weathering than those of VM4, which tentatively indicates that part of the bones preserved at VM3 were exposed during a longer time before burial. However, most bones from both quarries were buried very shortly after the death of the animals, less than one year. Moreover, the medullary cavities of the bones fractured by the hyaenas in VM3 are infilled by mud flows, but those preserved complete do not show infillings, even in the areas close to large nutrient foramina. This indicates that they were buried with the greasy periosteum intact, which suggests a subaerial exposure of few months⁷.

Elements in anatomical connection

According to Luzón et al.¹, many bones of VM4 are anatomically connected (e.g., a humerus-radius and a femur-tibia, fibula, and talus of rhino S. hundsheimensis, both sets located some meters distant; a group of seven dorsal vertebrae of M. meridionalis; an almost complete forelimb and two complete hindlimbs of L. lycaonoides, also separated by a short distance; and two hemipelvises of E. altidens). During the field excavation seasons of 2013–2015, which they do not study, other elements of the same rhino skeleton (e.g., a humerus-radius and a femur-tibia, fibula, talus, and three metapodials as well as a cervical section including the axis and two cervical vertebrae; Fig. 6a–c) were found. In addition, some vertebrae of M. meridionalis (Fig. 6d), a complete forelimb of a small-sized felid (Fig. 6e), two complete hindlimbs of a large canid, probably L. lycaonoides (Fig. 6f), a skull of P. brevirostris, and two hemipelvises of E. altidens (Fig. 6g) were unearthed. Surprisingly, these fossils are not cited by Luzón et al.¹. Although there are also elements of VM3 in anatomical connection (e.g., several skulls showing the mandible articulated with the cranium, two hemipelvises of a juvenile horse, some groups of lumbar vertebrae and several groups of autopodial elements)^7,9,20, their frequency is lower than in VM4. However, many bones of VM3 are disarticulated but spatially associated to others with which they were anatomically connected in origin^7,9 (Fig. S3). Again, this difference suggests that the bones preserved at VM3 were exposed during a longer time before their burial, which resulted in more weathering and a higher degree of modification by the hyaenas. Most probably, the reason was that the breeding season in the denning site of P. brevirostris in VM3 prolonged a little longer than in VM4 before the rise of the water level of the lake covered the area and capped with limestone the bone assemblage.

Patterns of skeletal completeness

Luzón et al.¹ affirm that the skeletal remains of VM4 show a moderate degree of fragmentation. This claim is contradicted by data provided in their Table 2: ~ 36% of the remains measure < 3 cm and > 55% of limb bones show green fractures, which evidence their breakage in fresh state by the hyaenas. In VM3, a substantial proportion of the remains are also small shaft fragments and one third of them measure < 3 cm, like in VM4. Isolated epiphyses and diaphyseal bone shafts are very abundant and outnumber the bones preserved complete in VM3^9,10: 80.5% (963/1196) of the major limb bones of equids, bovids, and cervids are represented by isolated bone portions that show green fractures (mostly spiral or longitudinal fractures) and only 19.5% are preserved complete^7,9. These fractures were produced by the hyaenas during the biostratinomic stage, when the bones were still fresh and retained their marrow contents. Green fractures predominate over diagenetic fractures resulting from sediment compaction, which are found in 22.5% of the bones of VM3⁹. The latter tend to be orthogonally oriented to the major axis in the long bones and are delimited in all cases by both bone portions (i.e., there is no fracture defining the end of one specimen), which shows that the assemblage is not reworked⁷.

Bone transport

There is no evidence of transport by water currents in VM3, because the bones: (i) show a random pattern in their spatial orientation; (ii) have no rounded or polished edges; (iii) show no traces of abrasion from rolling or similar movements within a current; and (iv) detritics are nearly absent from the sediment, which composition is 90–98% pure micritic limestone precipitated in water ponds emplaced on a caliche palaeosoil of diagenetic origin^7,9. In the case of VM4, Luzón et al.¹ also indicate the absence of bones with rounded surfaces, which helps them to discard fluvial currents as responsible of the accumulation of skeletal remains. However, they affirm that ~ 40% of specimens show evidence of hydraulic abrasion. Given that it is difficult to explain the presence of bones that were abraded but not rounded, they suggest the following ad hoc hypothesis: while the currents were weak for displacing the remains, abrasion resulted from circulating waters that moved the sediment over the bones, which originated their abrasion¹. However, detritics are absent from the whole VM stratum⁷ and it is difficult to explain the finding at VM4 of microvertebrate remains, which should have been transported by water currents of low energy. Luzón et al.¹ indicate that the fossil remains of the two levels of accumulation described in VM4 show patterns of preferential orientations towards the NE. This is also difficult to explain, because they discard the role of hydraulic currents in the transport and deposition of bones. Concerning bone dip, only 4% of the remains show azimuth values over 45°, while 79% of bones are found relatively flat along the topography of the excavated surface¹. As a result, they suggest gravity as the likely cause for the observed patterns of bone inclination at VM4¹, a hypothesis previously suggested for VM3⁷. In the case of VM3, rose diagrams show no preferential patterns of alignments in the orientation of the skeletal remains, while dip angles greater than 30° are poorly represented and bones showing a vertical or subvertical inclination are very scarce⁹.

Carnivore tooth marks

Luzón et al.¹ indicate that carnivore alterations were only observed in 4.5% of those bones with a well-preserved cortical surface and that only three bones showed 3–6 tooth marks. The frequency of tooth-marked remains in the larger sample of VM4 analysed here is slightly higher, 5.5% (177/3227) (Table S3). The marks include scores, pits, notches, crenulated edges, and furrows made by the hyaenas (Table S4, Fig. 5). Of these marks, ~ 60% appear in limb bone shafts. In addition, one bone shows marks made by a porcupine (Fig. 5b). In the case of VM3, 29.4% (1555/5288) of the remains analysed show carnivore tooth marks⁹. These bones belong to all ungulate species identified in the site and there are also some tooth-marked bones of Pachycrocuta, both of adult and non-adult individuals. Many cranial fragments and most limb bones of VM3 show striations and biting marks, the preserved epiphyses have furrows and punctures, and the diaphyses, as well as the skull bones, show scoring and pitting. Pits, scores, and notches are the marks most frequently recorded, although crenulated edges and furrows are also abundant^7,9. The proportion of tooth-marked bones in VM3 was in all probability even higher, as many limb bones of the assemblage that do not preserve tooth marks show fracture patterns that evidence that they were broken in fresh state by the hyaenas⁹, and this also applies to VM4. A significant part of the tooth marks preserved in VM3 were probably produced by juvenile hyaenas, which deciduous teeth are more cutting than the permanent premolars of adults, which are progressively blunted by bone cracking, and this results in inconspicuous tooth marks⁷. However, a minor implication of some small or medium-sized carnivore like the wolf C. orcensis²¹ cannot be discarded. Among modern hyaena dens, the frequency of tooth marked bones is highly variable¹⁷: 29.0–53.5% in spotted hyaenas, 22.1–100% in brown hyaenas, and 6.0–56.2% in striped hyaenas (the lower limit for this species is contentious, because it corresponds to a den where the bones are highly weathered). These frequencies are higher than in VM3 and VM4, but many fossil bones of the site are pending of restoration and evidence of gnawing by juvenile hyenas is usually very subtle.

The higher proportion of tooth-marked bones in VM3 agrees with a longer exposure of the skeletal remains than in VM4, which would explain why they were exploited more thoroughly by the hyaenas. As discussed earlier, this is also suggested by the more advanced degree of bone weathering and by the lower frequency of articulated elements. According to Luzón et al.¹, bones with salivary and gastric alterations are absent in VM4, but they are recorded in VM3 at very low frequencies (0.34% and 0.15% of 5288 specimens analysed, respectively)⁹. This suggests that their absence from VM4 results from the small sample of remains studied. In fact, our analysis of the specimens from the excavation seasons not studied by Luzón et al.¹ showed the presence of two bones with evidence of digestion (0.06% of the specimens analysed; Fig. 5a) and two others with salivary alterations resulting from licking (Table S4). In the study of 24 hyaena dens cited above¹⁷, evidence of acid or gastric edging of bones was detected in only eleven skeletal remains (six from spotted hyaena, four from brown hyaena and one from striped hyaena assemblages). These low numbers are not surprising, because neither striped hyaenas nor brown hyaenas regurgitate bones.

Finally, Luzón et al.¹ indicate that hyaena coprolites are absent in VM4, but a small coprolite was unearthed during the excavation of 2014 (Fig. 5c). The hyaena den of VM3 also preserves some coprolites represented by isolated pellets with diameters of 3–6 cm⁷.

VM4 and VM3: coeval or successive bone accumulations?

According to Luzón et al.¹, VM4 is in the context of a series of short-time events (they identify two of them in VM4, while only one was recognized in VM3⁷) followed by rapid sedimentation, as indicated by the low degree of bone weathering, the low frequency of tooth-marked bones, and the presence of skeletal remains anatomically connected. In the case of VM3, the more advanced weathering, the higher frequency of bones with bite marks and the lower proportion of elements in anatomical connection suggest that the deposition of limestone that capped the bone assemblage after the rising of the water table of the Baza palaeolake was delayed compared to VM4.

The bone assemblage accumulated in the hyaena den of VM4 is positioned ~ 350 m distant from VM3 (Fig. 1f). This distance is very short for considering the possibility of two neighbouring hyaena clans: in Serengeti, where spotted hyaenas engage in prolonged clashes with neighbouring clans, the radius of the permanent territory defended by a clan around the communal den fluctuates between 2.6 and 5.7 km²². This suggests that VM4 and VM3 were not coeval but correspond to separate events of accumulation during different years. More specifically, a study of a spotted hyaena clan in Masai Mara showed that the hyaenas used 57 different sites for communal denning during a period of ten years, with an average distance between the dens used consecutively of 1.5 ± 0.1 km²³. The distance between VM4 and VM3 is five times shorter than the one expected when the adult spotted hyaenas move their cubs to a new denning site, which suggests that VM4 and VM3 do not represent a residential move of the same hyaena clan. Therefore, the most parsimonious interpretation is that the bone assemblages preserved at both denning sites were accumulated by the hyaenas during the dry seasons of different years in the emerged plain that surrounded the lake surface covered by permanent waters (Fig. 1f). During the interval of time in which the accumulation of the bone assemblage of VM3 took place, the rising of the water table in the rainy season probably occurred some months later than during the years that correspond to the two accumulations of bone remains detected in VM4¹. Following this interpretation, the remains accumulated at the denning site of VM4 were capped with limestone (which protected them from weathering) somewhat earlier than those of VM3, which explains the minor taphonomic differences between both sites cited above. This interpretation also helps to clarify why the whole VM stratum, which outcrops along ~ 2.5 km, seems to be littered with fossils of large mammals, as it is difficult to conceive a “megadenning site” of P. brevirostris that extended over several squared kilometres. In our model, the VM stratum would represent successive deposits of micritic limestones in the plain that surrounded the Baza palaeolake during several lowstand-highstand cycles, each corresponding to the dry and rainy seasons of one year (Fig. S2d). Each year, the hyena clans that inhabited the Baza Basin would randomly select their denning sites on this plain. After enough years, the surface seasonally submerged of the plain that surrounded the lake would be almost entirely covered by fossils of large mammals, which were preserved in the micritic limestones of the VM stratum.

Source: Ecology - nature.com