A refined chronology of the Naumann’s elephant (Palaeoloxodon naumanni) provides a new insight on factors of their extinction

Many large mammal species on each continent became extinct during the Late Pleistocene¹. Because the Late Pleistocene is also the time that Homo sapiens dispersed from Africa, previous studies from diverse environmental contexts have focused on the impacts of human activity and climate change as the principal drivers of their extinction^{2,3,4,5,6,7,8,9}. While it has been argued that the arrival of humans and the extinction of large mammals occurred almost simultaneously in Australia and North and South America, the timing also coincided with climatic shifts of the Terminal Pleistocene^10,11. Furthermore, several other factors have been proposed for the extinction including disease^12,13 and a decrease in genetic diversity^14,15. Although population dynamics might be reconstructed from genetic analyses¹⁶, the impacts of climatic shifts, human hunting pressure, and their combined effects remain important factors to be tested when investigating the extinction of large mammals in the Late Pleistocene¹⁷.

The key issue in discussing the impact of human and climate change on the extinction of large mammals concerns accurate estimates of the chronology of their extinction and human arrival. The timing of the extinction of large mammals has been estimated based on radiocarbon dating of fossil bone collagen. To help avoid the bias associated with incomplete extinction records (the Signor–Lipps effect¹⁸) and to estimate the exact time period of the extinction, the Gaussian-Resampled and Calibration-Resampled Inverse-Weighted McInerny Methods (GRIWM and CRIWM, respectively) have been proposed^19,20. Using the GRIWM method, it was estimated that the extinction of large mammals occurred during the rapid warming periods of the Dansgaard–Oeschger interstadial cycles⁴.

Naumann’s elephant (Palaeoloxodon naumanni) dispersed from Africa about one million years ago, arrived in the Japanese Archipelago 330 k years ago, and survived until the end of the Late Pleistocene^21,22,23,24. Naumann’s elephant preferred to live in mixed deciduous and conifer forests under a temperate climate²⁴, which helped enlarge their habitat. Although the drivers of their extinction are thought to be climatic shifts^5,25,26 or human impact²², most researchers consider both factors and advocate the need for further research^21,24,27. Radiocarbon dating of fossil bone collagen from Naumann’s elephant from two major Japanese islands (Honshu and Shikoku islands) has been attempted to estimate the timing of their extinction^{28,29,30,31,32}. Previously, it was thought that their extinction occurred around 20 k years ago during the Last Glacial Maximum (LGM). However, studies based on compiled reported dates have estimated the timing back to ca. 30 k cal BP^25,33. Recently, based on the compiled dates of Iwase et al.²⁵, the statistical estimates of their extinction using the GRIWM and CRIWM methods proposed that their extinction occurred around 24.6–23.5 k cal BP²⁰.

However, previous radiocarbon dating in Japan has mostly adopted the gelatinization method on bone collagen. A small amount of contaminated young carbon in collagen can have a large effect on the radiocarbon dates of samples aged around 50–30 k cal BP^34,35, and this can critically affect debates regarding megafaunal extinction. The ultrafiltration method, which separates collagen with larger peptides (> 30 kDa) from that with shorter peptides, has been proposed to achieve more accurate measurements of old samples³⁶. This method removes low-molecular-weight contaminants from final products and enables reports to include older dates than previous studies on fossils from the Late Pleistocene^{37,38,39,40,41}. However, ultrafiltration does not necessarily give accurate dates when involving contamination from the ultrafilter or unseparated contamination^42,43. Glycerol on the ultrafilter can be removed with appropriate washing and is typically contaminated in the low-molecular-weight fraction^44,45. Humic substances from sediments are a possible source of the contamination^36,45. Collagen-humic cross-link complexes might be contaminated in the high-molecular-weight fraction^43,46,47. Another possible source in the case of salvaged fossils is dissolved organic carbon originating from seawater. Most dissolved organic carbon shows a low molecular weight⁴⁸, which can be separated using an ultrafilter. The ultrafiltration method might work well for bone samples with poorly preserved collagen that are contaminated with other sources⁴⁹. Radiocarbon dating on previously reported samples and cross-dating have provided updated ages⁵⁰. Further, compound-specific dating, which can fully remove humic substances, is applied to fossils^46,51,52,53. Therefore, a proper collagen extraction method should be applied to improve the accuracy of fossil age from the Late Pleistocene.

To evaluate the human impact on the extinction of large mammals persuasively, the arrival timing and intensity of human activity should also be properly estimated. Human arrival on the Japanese Archipelago is considered to have occurred at around 38 k cal BP using the IntCal13 calibration curve^54,55,56. This is based on the radiocarbon ages of charcoals carefully sampled from layers bearing the oldest lithic tools. An increasing number of sites until 30 k cal BP has been proposed to have caused hunting pressure on large mammals²². Given sufficient radiocarbon dating, the summed probability distributions (SPDs) of radiocarbon dates from archaeological sites can be considered as the proxy of population densities over time^57,58,59,60. The rise and fall of the SPDs related to human population change should therefore indicate the intensity of human predatory behaviour on large mammals.

Given this background, the present study aimed to determine the accurate radiocarbon ages of Naumann’s elephant fossils from Honshu and Shikoku Islands in Japan. Previously published materials younger than 38 k cal BP were re-measured using updated methods. Additionally, a representative Pleistocene mammalian fossil collection in Japan that was recovered from the seabed off Hakata Island in the Seto Inland Sea in Imabari City, Ehime Prefecture, was newly analysed. The SPDs of radiocarbon dates were plotted, and then the extinction times were estimated using the CRIWM method. In addition, radiocarbon dates of archaeological sites from the Upper Palaeolithic were compiled to illustrate the diachronic change of population densities. The proxy of climatic data and the SPDs of the radiocarbon dates of Naumann’s elephant and archaeological sites were then plotted over time to investigate the intensity of climatic shifts as a driver of extinction.

Radiocarbon dates of Palaeoloxodon naumanni

A total of six P. naumanni fossil samples from Honshu and Shikoku were re-dated (Fig. 1: LN-1–6). Five samples that were younger than 38 k cal BP were selected from the 57 compiled radiocarbon dates by Iwase et al.²⁵. Another sample is from a report by Takahashi³¹. Use of the ultrafiltration collagen extraction method on these six samples yielded well-preserved collagen (GL fraction > 30 kDa). The radiocarbon dating of six fossils from Honshu and Shikoku gave 42.7–31.7 k BP. The results of the ultrafiltration method were older than the reported gelatinization method by 8.3 ± 4.2 k years on average (Fig. 1).

Furthermore, a total of 21 P. naumanni fossils from Imabari City in Ehime Prefecture were used for collagen extraction by ultrafiltration and gelatinization methods, among which, six samples exhibited well-preserved collagen with high collagen yield (> 1%). Collagen from the ultrafiltration method showed 53.5–38.2 k BP and was older than the dates of the gelatinization method by 5.9 ± 2.5 k years (Fig. 1).

In sum, a total of 12 P. naumanni samples presented calibrated radiocarbon dates of 55.6–35.7 k cal BP. Two samples were above the range of the calibration curve. The six reported radiocarbon dates before 38 k cal BP from the compiled list were also included to calculate the SPDs of 16 Naumann’s elephant samples (Fig. 2; see Supplementary Tables S1 and S2 online). The estimate of the time of extinction on 10 samples from earlier than 50 k cal BP using the CRIWM method was 34–32.3 k (median 33.2 k) cal BP. When the additional six reported radiocarbon dates older than 38 k cal BP from the compiled list were included, the estimate of the time of extinction was 34.1–32.7 k (median 33.5 k) cal BP.

Radiocarbon dates of archaeological sites

The SPDs of radiocarbon dates from the archaeological sites gave the range of 39–23.5 k cal BP (Figs. 2 and 3). The intensity of the SPDs gradually increased from 39 to 35 k cal BP. In addition, the intensity of the SPDs decreased at around 31 k cal BP, and then increased until 29 k cal BP. The estimate of the arrival time of humans using the CRIWM method was 39.1–38.1 k (median 38.7 k) cal BP. Most of the dates during 39 k to 35 k cal BP consisted of the southern Chubu and Kyushu regions, and subsequently, the dates increased in regions west of the Kanto region. Only a small number of dates was found from the Tohoku and Chugoku regions during this period.

The findings of the present study revealed that the radiocarbon age of Naumann’s elephant was older than 35.7 k cal BP. A comparison of radiocarbon dates obtained using the ultrafiltration and gelatinization methods yielded distinctive results: the dates obtained from the ultrafiltration method were older. This suggests that the ultrafilter separated low-molecular-weight contaminants of earlier carbon. The contaminants might include humic substances from sediments and dissolved organic carbon from seawater in the salvaged samples. This is in accordance with the proposition that a small amount of contamination can affect radiocarbon dates of particularly old materials of 50–30 k cal BP^34,35,39,40. The youngest date (last appearance datum [LAD]) was 36.4–35.7 k cal BP, but the fossil material would not be truly the last surviving individual. To solve the problem of the Signor–Lipps effect, a statistical estimate of the extinction time with the CRIWM method was applied. The statistical estimate of the time of the extinction was 34.1–32.7 k cal BP. Because the time span of the LAD and the one sample before LAD were long, the dates of the LAD and the extinction differ by 1.6 k years. The CRIWM estimate based on the re-dating of Naumann’s elephant fossils was about 9 k years older than the previous estimate of 24.6–23.5 k cal BP^4,20.

The SPDs of the radiocarbon dates of Naumann’s elephant showed peaks at about 45 and 42 k cal BP, and most samples indicated dates before 42 k cal BP. The four re-dated samples from Shimane and Aomori Prefectures (Chugoku and Tohoku regions, respectively) exhibited 40–35 k cal BP, which partially overlapped with the SPDs of the archaeological sites. These were the only four well-preserved samples among 57 specimens of the compiled list. Collagen was not preserved in many fossil materials, with only six well-preserved samples among 21 newly collected specimens salvaged from the Seto Inland Sea, which showed dates before 42 k cal BP. Only a few fossils of Naumann’s elephant corresponded to the time after the arrival of humans.

The SPDs of the archaeological sites showed a distribution starting from about 38 k cal BP (Fig. 2). The estimate of the arrival time using the CRIWM method indicated 39.1–38.1 k cal BP. The intensity of the SPDs increased gradually between 39 and 35 k cal BP, and remained steady until 32 k cal BP. When compared with the SPDs of Naumann’s elephant, the maximum coexistence time of Naumann’s elephant and humans was 39.1–32.7 k cal BP. When the date of the LAD was used, the time became shorter between 39.1 and 35.7 k cal BP. These results indicate that Naumann’s elephant and humans coexisted in the Japanese Archipelago for 6–4 k years, which is shorter than previous estimates. In the cases of North and South America, however, an even shorter period of 1.5 k years of coexistence is emphasized as evidence supporting the human-induced extinction of large mammals^11,61, classically known as ‘the Pleistocene overkill’ model⁶². The coexistence interval shown in our results is longer than that of the overkill model and therefore may not support a similar model in Japan.

Rather, our data further weaken the impact of hunting pressure by Upper Palaeolithic foragers. It should be noted that the archaeological SPDs around 35 k cal BP mainly consists of radiocarbon dates from the Chubu, Kanto, and Kyushu regions (Fig. 2). Sites from the Chugoku and Tohoku regions of this period are remarkably scarce. On the contrary, younger fossil samples of Naumann’s elephant dated to 40–35 k cal BP have been found outside the areas already inhabited by humans at that time. Currently observed phenomena would indicate that the sites of the Upper Palaeolithic do not spatiotemporally correlate with the habitats of the Naumann’s elephant, likely implying that human hunting had only a small impact on megafaunal extinction.

Furthermore, the period of 39–35 k cal BP is the time of the early Upper Palaeolithic when tiny trapezoids and edge-ground axes were mostly used⁵⁵. The use of such unformal small trapezoids cannot be persuasively said to have enabled humans to hunt large mammals. Unlike in North America, where Paleoindian groups equipped with large, thin, and symmetrical bifacial points would have hunted megafauna, including mastodon and bison⁶³, trapezoids from the early Upper Palaeolithic in Japan were supposedly used as arrowheads⁶⁴. Extensive ethnographic research on elephant hunting worldwide has found that spear hunting and trapping predominate, with only a few instances of bow hunting that typically involved the use of poisoned arrows⁶⁵. Blade points as large as the bifacial projectile points in North America became widespread as hunting weapons only around 33–32 k cal BP⁵⁵, after the extinction of Naumann’s elephant. Though the past hunting technology should be further investigated, we did not consider the lithic tool kit used until 35 k cal BP to support the dissemination of elephant hunting.

The results of this study suggest that the extinction of Naumann’s elephant occurred during Greenland Stadial (GS)-8 to GS-6 in Marine Isotope Stage 3 (MIS3) (Fig. 3). These results do not support the idea that the effect of catastrophic environmental deterioration caused by the Aira-Tn (AT) tephra or cold climate during the LGM was a driver of extinction. The stadials and interstadials in the Dansgaard–Oeschger cycles found in the Greenland ice core were also found in the oxygen isotope ratios of stalagmites that reflect the intensities of the East Asian summer monsoon^66,67 and in the results of pollen analysis of the sediment core of Lake Nojiri, Nagano Prefecture, Central Japan⁶⁸. Naumann’s elephant would have persisted through climatic shifts until GI-9 and the long cold GS-9 stadial (Heinrich event 4), which corresponds to the arrival of humans. Then, they became extinct at 35–33 k cal BP, when relatively warm interstadials and cold stadials occurred repeatedly in MIS3⁶⁹. Naumann’s elephant survived during the cold GS-8–6 and the warm GI-8–6, whereas their population and habitats would have shrunk in repeated cold stadials^25,27,33. Given the current evidence, it might be reasonable to conclude that this species, adapted to temperate zones, gradually declined in population during GS-8–6 in MIS3. Such population dynamics might be reconstructed in future research using ancient DNA analysis.

Radiocarbon dating of reported and new fossils yielded a short duration of coexistence time of Naumann’s elephant and Upper Palaeolithic humans. The spatiotemporally unrelated tendency between them along with the nature of hunting weaponry of humans may not support the large impact of human hunting on their extinction. The ultrafiltration method cannot exclude high-molecular-weight contaminants, and compound-specific radiocarbon dating (individual amino acids)^46,51,52,53 on the LAD sample would be a good practice in future research. Additional dating to increase the number of samples is still needed to improve the accuracy of the extinction date. Although Naumann’s elephant is the most prominent fossil in the Late Pleistocene fauna, the dates of other species, including Yabe’s giant deer (Sinomegaceros yabei), Bison sp., and woolly mammoth (Mammuthus primigenius), should be tested with re-dating. The approach of compiling the dates and re-dating the fossils may prompt reconsideration of the extinction of the large mammals even at the continental level. The findings of this study highlight the use of accurate radiocarbon dates of fossils and archaeological sites to clarify the time of extinction of large mammals of the Late Pleistocene.

Materials

Among the compiled radiocarbon dates of Naumann’s elephant fossils²⁵, 12 samples had well-preserved collagen based on the following criteria: collagen yield > 1%⁷⁰ and C/N ratio within 2.9–3.6⁷¹. Five samples that showed dates earlier than 38 k cal BP were resampled for radiocarbon dating using the ultrafiltration method. Another sample was also resampled³¹. The fossil LN-1 is a molar stored in Yawatahama City, Ehime Prefecture, that was salvaged from the western Seto Inland Sea. Radiocarbon dates of 24.9 k BP have been reported by use of the gelatinization method^30,72. The sampling of LN-1 was conducted under the authorization of the Educational Board of Yawatahama City. LN-2 (Kyoto No.2) is an incisor stored in the Department of Geology and Mineralogy, Graduate School of Science, Kyoto University, that was reported as 29 k BP without alkali wash and gelatinization²⁸. The sampling of LN-2 was conducted under the authorization of the Department of Geology and Mineralogy, Graduate School of Science, Kyoto University. LN-3 is a molar salvaged from the sea off the Saigo Port of Shimane Prefecture in the Japan Sea and stored in the Oki Islands Geopark Museum. A radiocarbon date of 28.3 k BP was reported for LN-3³¹. The sampling of LN-3 was conducted under the authorization of the Oki Islands Geopark Museum. LN-4–6 are molars originally found in the limestone mine at Shikkari, Aomori Prefecture, and stored in the Aomori Prefectural Museum⁷³, with reported radiocarbon dates of 31.3–23.6 k BP³². The sampling of LN-4–6 was conducted under the authorization of the Aomori Prefectural Museum.

The new fossil collection, salvaged from the Seto Inland Sea (east of Hakata Island) and stored in the Murakami Kaizoku Museum of Imabari City, Ehime Prefecture, was researched. The results of species identification gave 64 fossils of Naumann’s elephant and six deer antlers (Cervidae). Nine dentine molar samples and 12 bone samples for collagen extraction were selected. About 1-g samples were cut with a diamond cutting disc and hand drill. The curation and sampling were conducted with the permission of the Murakami Kaizoku Museum of Imabari City.

Collagen extraction

Collagen extraction from the fossil samples was performed using the gelatinization and ultrafiltration methods at the laboratory in Tokai University. Gelatinization followed the method of Longin⁷⁴. The samples were ultrasonically cleaned and soaked in NaOH (0.2 mol/L) for about 20 h. Lyophilized samples were crushed with a hammer and mill. Then, powdered samples were closed in a cellulose tube (> 14 kDa) and demineralized in HCl (0.5 mol/L) for about 20 h. The samples were then centrifuged in glass tubes. Precipitation fraction with ultrapure water was heated with a block bath at 90 °C for 20 h. The supernatant fraction from the centrifuge was filtered, and lyophilization yielded gelatinized collagen.

The ultrafiltration method followed the method of Brown et al.³⁶. Fossil samples were ultrasonically cleaned and lyophilized. Samples were crushed with a hammer and mill. The samples in the cellulose tube were demineralized with HCl (0.5 mol/L) for about 20 h. The samples were centrifuged, and the supernatant was separated as the soluble collagen fraction (S fraction). Precipitation fractions were washed with NaOH (0.1 mol/L) for 30 min and washed twice with ultrapure water. Then, samples were washed with HCl (0.5 mol/L) for 30 min and then twice with ultrapure water. Next, the samples were heated at 75 °C for 20 h in HCl (pH 3). The gelatinized collagen was filtered into a washed ultrafiltration tube (Vivaspin 30 kDa; Sartorius AG, Germany). Centrifugation of the ultrafiltration tube separated the GL fraction (> 30 kDa) from the GS fraction (< 30 kDa). Lyophilization yielded the final products of collagen GL and GS. There is a problem of contamination from the ultrafilter with small samples in the GL fraction; thus, the sample weight was adjusted to give sufficient weight in collagen GL. The ultrafiltration tubes were washed before centrifuging to remove glycerol on the ultrafilter (ca. 30 ka^44,45). Ultrapure water was centrifuged twice in ultrafiltration tubes before use. The ultrafiltration tubes were sonicated in ultrapure water for 1 h. Then, ultrapure water was centrifuged twice again.

The carbon and nitrogen contents of the collagen were measured using an elemental analyser (EA IsoLink: Thermo Fisher Scientific Inc., USA; Tokai University, Japan). Samples with amino acid standards were measured in duplicate, and averages were reported.

The graphitization of collagen by a standard protocol and measurements of radiocarbon dates were performed at Paleo Lab⁷⁵. Radiocarbon dates were measured with compact accelerator mass spectrometer (AMS) (1.5 SDH; National Electrostatics Corp., USA). The measurements were standardized with oxalic acid (NIST-9582) and checked by IAEA-C7. The averages of the background samples gave 49.4 ± 2.7 k BP (N = 7).

Calibration of the radiocarbon dates was performed using the OxCal4.4 programme⁷⁶ and the IntCal20 terrestrial calibration curve⁷⁷.

The plot of SPDs and estimates of the extinction and arrival times were calculated using R (R, ver. 4.5.0) and the R packages ‘rcarbon’ and ‘Rextinct’⁵⁸.

SPDs of radiocarbon dates from archaeological sites

To test the coexistence time of Naumann’s elephant and humans, radiocarbon dates of charcoal from Upper Palaeolithic sites were compiled from the radiocarbon database of National Museum of Japanese History^78,79, throughout Paleo-Honshu Island (Honshu, Shikoku, Kyushu islands), and AMS dates older than 20 k BP were selected. Because bones are rarely found in this area because of the acidic soil resulting from volcanic activity, all data were based on AMS dates measured from charcoal and all radiocarbon dates were measured from charcoal closely associated with lithic concentrations; samples lacking this condition were rejected.

The compiled list provided radiocarbon dates of 494 samples from 100 archaeological sites. The list was analysed using ‘rcarbon’⁵⁸. The calibration was done by IntCal20⁷⁷ and SPDs were plotted. ‘Rextinct’ was also used to estimate the arrival time of humans.