Stable isotopes show Homo sapiens dispersed into cold steppes ~45,000 years ago at Ilsenhöhle in Ranis, Germany

Abstract

The spread of Homo sapiens into new habitats across Eurasia ~45,000 years ago and the concurrent disappearance of Neanderthals represents a critical evolutionary turnover in our species' history. 'Transitional' technocomplexes, such as the Lincombian-Ranisian-Jerzmanowician (LRJ), characterize the European record during this period but their makers and evolutionary significance have long remained unclear. New evidence from Ilsenhöhle in Ranis, Germany, now provides a secure connection of the LRJ to H. sapiens remains dated to ~45,000 years ago, making it one of the earliest forays of our species to central Europe. Using many stable isotope records of climate produced from 16 serially sampled equid teeth spanning ~12,500 years of LRJ and Upper Palaeolithic human occupation at Ranis, we review the ability of early humans to adapt to different climate and habitat conditions. Results show that cold climates prevailed across LRJ occupations, with a temperature decrease culminating in a pronounced cold excursion at ~45,000-43,000 cal BP. Directly dated H. sapiens remains confirm that humans used the site even during this very cold phase. Together with recent evidence from the Initial Upper Palaeolithic, this demonstrates that humans operated in severe cold conditions during many distinct early dispersals into Europe and suggests pronounced adaptability.

Fig. 1. Oxygen, nitrogen, carbon and zinc stable isotope analyses of directly dated equid teeth show changes in climate and environment through the LRJ and Upper Palaeolithic sequence of Ranis.

Summer peak, mean annual and winter trough oxygen isotope values show low values throughout the sequence and a temperature decline from ~48 ka cal bp to a temperature minimum at ~45–43 ka cal bp. This oxygen isotope minimum coincides with high δ¹⁵N (dentine and mandible bone collagen) and δ⁶⁶Zn values, suggesting a hypergrazer niche of equids in open steppe environments or very dry soil conditions similarly indicative of an open environment. This is supported by high δ¹³C (dentine and mandible bone collagen) values consistent with a steppe or tundra biome. One individual has been marked with an asterisk as it has been excluded from climatic interpretations because ⁸⁷Sr/⁸⁶Sr and δ⁶⁶Zn seasonal amplitudes are high enough that a seasonal movement cannot be completely excluded. Oxygen isotope data points represent δ¹⁸O summer peak, winter trough and annual means of individual annual cycles represented in sinusoidal δ¹⁸O time series obtained from sequentially sampled tooth enamel (marked in Supplementary Fig. 2). Stable isotope data are presented as the mean ± measurement uncertainty based on sample replicates (1 s.d., n_replicates = 3 for δ¹⁸O and n_replicates = 2 for all other proxies where error bars are present; replicate measurements represent repeated isotopic measurements of aliquots of each single prepared sample). Measurement uncertainty for δ⁶⁶Zn is smaller than the symbol size. Horizontal error bars indicate the 95% calibrated age range of direct radiocarbon dates (n = 1 tooth sample for each data point). Symbol shapes indicate the excavation origin from either the Hülle (1932–1938, circles) or TLDA/MPI-EVA (2016–2022, triangles) campaigns. Collagen analysed for δ¹³C and δ¹⁵N was obtained from tooth dentine for all 1932–1938 samples and from adhering mandible bone for the two 2016–2022 samples marked by triangle shapes. Stable isotope delta values are reported in relation to the relevant scale-defining reference materials Vienna Mean Ocean Water (VSMOW), atmospheric N₂ (AIR), Vienna Pee Dee Belemnite (VPDB), and Johnson Mattey zinc metal (JMC Lyon).

Fig. 2. Air temperature estimates derived from δ¹⁸O measurements generally fall below modern-day conditions.

Lowest temperatures are observed in the ~45–43 ka cal bp interval, where they fall ~7–15 °C below modern day and mean annual temperatures below freezing. Oxygen isotope data from several individuals were grouped into time bins according to clusters of radiocarbon dates and δ¹⁸O measurements (Supplementary Table 6). Plotted points represent temperature estimates for each time bin. Error bars represent combined uncertainty for each temperature estimate, taking into account the uncertainty of each temperature calibration step (Supplementary Text 5). N_datapoints for each error bar varies by season and time bin and can be found in detail in Supplementary Table 6. In the time bins δ¹⁸O_dw estimates are based on a variable number of tooth specimens with n_{36–39 ka} = 7, n_{42–43 ka} = 3, n_{43–45 ka} = 2 and n_{45–48 ka} = 3. Summer and winter temperatures were estimated from inverse modelled δ¹⁸O time series, while annual means were derived directly from unmodelled δ¹⁸O (Supplementary Text 5). Lines and shaded ribbons of modern comparative data represent means and one standard deviation of modern climate observations (MAT, T_{coldest month}, T_{warmest month}) for 1961–2009 from the ClimateEU model).

Fig. 3. H. sapiens presence coincides with the coldest temperatures documented by equid δ¹⁸O data.

Comparison of equid δ¹⁸O data (top) with directly dated H. sapiens remains (bottom, turquoise symbols) demonstrates extensive overlap of H. sapiens presence with the coldest temperatures documented between ~45 and 43 ka cal bp (marked by blue shading). This coldest, low δ¹⁸O phase overlaps with the age ranges of both the LRJ layer 8 and the beginning of undiagnostic layer 7 (ref. ) (modelled 95% probability layer age ranges of the MPI-EVA/TLDA excavation in purple) but the direct dates of H. sapiens remain and faunal bone fragments with anthropogenic surface modifications clearly show that they endured the cold subarctic steppe conditions evidenced by the stable isotope data at this time. This also holds true independent of the calibration curve, as seen in the uncalibrated dates (Supplementary Fig. 13). Top panels show the relevant Greenland stadials (GS) and interstadials (GI) recorded in the NGRIP ice cores and the proportions of total arboreal pollen (dark brown) and Betula pollen (ochre) in the Füramoos pollen record from southwestern Germany^,. Data are presented as mean ± 95% calibrated age ranges (n = 1 bone or tooth enamel sample for each data point), while point shape indicates whether specimens were found in the Hülle (1932–1938, circles) or the TLDA/MPI-EVA (2016–2022, triangles) excavation collection. We argue that H. sapiens fragments from the Hülle collection (labelled IX–XI here) all originate from the LRJ deposits (layer X) and were sometimes assigned to a mixture of layer X and adjacent strata by the original excavators due to rough excavation methods (details in ref. ). We have pooled all these samples here to reflect this. Credits: equid silhouette by Mercedes Yrayzoz, vectorized by T. Michael Keesey (PhyloPic); human silhouette from NASA Pioneer plaque.

Extended Data Fig. 1. Location and stratigraphy of Ilsenhöhle in Ranis, Germany.

A - Location of Ilsenhöhle in Ranis, Thuringia, central Germany. B - Hydrotopographical setting of the area surrounding Ranis in the Orla valley. The Thuringian highlands can be seen to the south of the site. Other notable rivers include the Saale River passing south and west of Ranis. Elevation contour lines are spaced 200 m apart. Elevation data from the European Digital Elevation Model version 1.1. Waterways imported from OpenStreetMap. C - Schematic stratigraphy of the 1930s Hülle excavation and the 2016–2022 TLDA/MPI-EVA excavation with layer correlations (layer numbers in circles). Samples analysed here roughly cover the time period from the Lincombian–Ranisian-Jerzmanowician (LRJ) Layers 8 and 9 marked in red to the Upper Palaeolithic occupation of Layer 6. Rockfall events are marked as ‘R’.

Extended Data Fig. 2. Zinc stable isotope values across the Ranis food web.

Zinc stable isotope ratios of a range of taxa show typical trophic relationships with low δ⁶⁶Zn values for carnivores, high values for herbivores and intermediate omnivores and bone eating carnivores. Herbivores show a pattern with highest δ⁶⁶Zn values in equids and lower values in typical browser to mixed feeding cervid taxa such as Cervus elaphus. This is potentially consistent with grazer-browser δ⁶⁶Zn patterns observed in a European Pleistocene food web and a modern African food web and with limited studies of different plant parts. If confirmed, this suggests that higher δ⁶⁶Zn in some equids could be due to a hypergrazer feeding ecology (see Supplementary Text 4). Shapes indicate stages of tooth development, as teeth formed during nursing or in utero can exhibit higher δ⁶⁶Zn values (see Supplementary Text 4). It should be noted that for equids summer and winter δ⁶⁶Zn values are plotted (2 per tooth), while other taxa are represented by one measurement per specimen.

Extended Data Fig. 3. Summer, winter and mean annual estimates of environmental water oxygen isotope values compared to modern meteoric water sources.

Reconstructed oxygen isotope composition of drinking water (δ¹⁸O_dw) fall substantially below δ¹⁸O of modern-day water sources (precipitation - solid ribbons; rivers - hatched ribbons; springs - dotted ribbons), particularly for mean annual and winter values. This indicates that temperatures were substantially below modern-day conditions. Summer δ¹⁸O_dw reconstructions partially overlap with modern spring δ¹⁸O due to the pronounced seasonal buffering in groundwaters. Precipitation δ¹⁸O were obtained from estimates for the site location made using the OIPC). River δ¹⁸O data includes measurements from the Heiderbach, a small stream in the Rinne valley, the Bode, a Saale tributary in northern Thuringia and the Elbe close to the confluence of the Saale. Spring water δ¹⁸O data includes measurements from four small springs in the Rinne valley. Error bars represent the overall uncertainty introduced by the conversion to drinking water oxygen isotope values (see Supplementary Text 5). N_datapoints for each error bar varies by season and time bin and can be found in detail in Supplementary Table 6. In the time bins δ¹⁸O_dw estimates are based on a variable number of tooth specimens with n_{36-39 ka} = 7, n_{42-43 ka} = 3, n_{43-45 ka} = 2, n_{45-48 ka} = 3.