Seasonal-to-interannual hydroclimatic variability in Afromontane environments during the late Early Pleistocene hominin occupations on the Southeastern Ethiopian Highlands

Abstract

Climate variability is a key factor influencing the distribution and adaptive response of mammalian species, including hominins. Here, we investigate seasonal and interannual hydroclimatic and environmental variability associated with the late Early Pleistocene hominin occupations on the Southeastern Ethiopian Highlands. We present stable carbon and oxygen isotope analyses of sequentially sampled tooth enamel from herbivores recovered from the archaeological contexts of Melka Wakena and Gadeb (2300–2400 m a.s.l.) and dated between 1.6 and ~ 1.0 Ma. Carbon isotope data indicate predominantly C₄ and mixed C₃–C₄ diets, consistent with mosaic vegetation as part of the Dry evergreen Afromontane Forest and grassland complex (DAF). Intra-tooth oxygen isotope profiles suggest seasonal variability and interannual fluctuations in water balance, reflecting changes in precipitation, evaporation, and local hydrology. Although the temporal resolution of individual tooth records precludes identification of specific climatic drivers, the isotopic data document recurring hydroclimatic variability within high-elevation environments characterized by broadly comparable ecological conditions. Herbivore diets show limited long-term change, suggesting sustained resource availability across seasons and over hundreds of thousands of years. These results indicate that the Ethiopian Highlands repeatedly supported mammalian communities and recurrent Acheulean hominin occupations during the late Early Pleistocene, providing sustained access to water and vegetation resources within an Afromontane environmental framework.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-60484-z.

Reconstructing past ecosystems and their connections to climate is essential for understanding the expansions and contractions of Plio-Pleistocene hominin populations within Africa and beyond. Paleoenvironmental data from northern¹, central^2,3, southern^4,5, and eastern Africa^6–16 emphasize the significance of heterogeneous habitats, microhabitats, and mosaic environments as key factors that shaped the geographic distribution of mammalian species^17–26.

On the Ethiopian highlands, at elevations of ~ 2000 m above sea level (m a.s.l.) and higher, palynological studies from Gadeb and Melka Kunture have documented Plio-Pleistocene environmental changes, revealing variable proportions of grasslands, wooded and forested vegetation within the Dry evergreen Afromontane Forest and grassland complex (DAF)^27,28. Temporal changes in pollen abundance are attributed to orbitally forced climatic events²⁸. Today, the DAF vegetation thrives primarily on the Ethiopian highlands at elevations above 1800 m a.s.l., such as in the Shashemene Forest (2010 m a.s.l.) and the Bale Mountains (4377 m a.s.l.)^27,28. In contrast, plain areas within this elevation range are largely modified by human activity, mainly for agriculture and grazing^27,29,30.

The modern climate of the Ethiopian highlands is shaped by the Indian Monsoon and the movement of air masses linked to the Intertropical Convergence Zone (ITCZ), producing a bimodal seasonal rainfall regime^31,32. During the dry season (Bega, from October to December/January), the region is influenced by a dry northeastern continental air mass. The ITCZ brings rainfall to the southern and southwestern areas between February/March and May (short rainy season, Belg). From June to September, the southwestern airstream extends over the Ethiopian highlands, generating the main rainy season (Kiremt)^32,33. The timing and movement of the ITCZ vary annually, contributing to pronounced interannual variability in rainfall across Ethiopia. These patterns are further modulated by the El Niño-Southern Oscillation (ENSO), with El Niño typically associated with reduced Kiremt rainfall during its warm phase, and La Niña linked to its cool phase and potentially enhanced rainfall^34–41.

Here, we employ stable carbon and oxygen isotope analysis of 312 sequential enamel samples from 14 teeth belonging to three mammalian families: Hippopotamidae, Bovidae, and Equidae (Table 1). The teeth were recovered from stratigraphically secure, dated sedimentary sequences at the Melka Wakena (MW) and Gadeb (GB) archaeological site-complex, situated in the Gadeb Plain (as defined in Hovers et al.⁴²) on the Southeastern Ethiopian Highlands between 2300 and 2400 m a.s.l. (Fig. 1) (see Supplementary Text A for dating, stratigraphy, and archaeological evidence)^42–47. The isotopic data enable reconstruction of hydroclimatic and local environmental conditions on a multi-seasonal scale between 1.6 and ~ 1.0 million years (Ma) ago. We compare our dataset with the published isotopic record from Melka Kunture (MK)^15,16, located on the western shoulder of the Main Ethiopian Rift at 2000–2200 m a.s.l (Fig. 1). By placing this information in a broader geographic framework, we aim to document the ecological dynamics that hominin groups associated with the Acheulean technocomplex (e.g., Homo erectus sensu lato), experienced during the late Early Pleistocene on the Ethiopian highlands.

Table 1.

Information on the faunal remains from Melka Wakena and Gadeb used in this study.

A	Fauna ID	Age (Ma)	Element	Family	Tribe	Genus	Species
	MW2-L3-B787	1.6	Left lower canine fragment	Hippopotamidae	Hippopotamini	Hippopotamus	Hippoptamus gorgops
B	MW n.d.	1.6	Canine	Hippopotamidae	Hippopotamini	Hippopotamus	Hippoptamus gorgops
C	MW2-L1&L2	1.45–1.34	Upper canine	Hippopotamidae	Hippopotamini	Hippopotamus	Hippoptamus gorgops
D	G8E-G64	~ 1.0	Lower left canine	Hippopotamidae	Hippopotamini	Hippopotamus	Hippoptamus cf. gorgops
E	G8D-G328-D1	~ 1.0	Lower right canine	Hippopotamidae	Hippopotamini	Hippopotamus	Hippoptamus cf. gorgops
F	G8D-G133-D2	~ 1.0	Lower left canine	Hippopotamidae	Hippopotamini	Hippopotamus	Hippoptamus cf. gorgops
G	MW2-L3-B695	1.6	Left lower m3	Bovidae	-	–	–
H	MW2-L3-B233	1.6	Right lower m2	Bovidae	-	–	–
I	MW2-L3-B579	1.6	Right upper M3	Bovidae	Tragelaphini	–	–
J	MW3	1.6	Right upper M3	Bovidae	Aepycerotini	Aepyceros	–
K	G2E-G507-K1	~ 1.45	Right upper M2	Bovidae	Alcelaphini	–	–
L	MW8	1.6	Left upper M3	Equidae	Equini	Equus	–
M	G2E-G496-L1	~ 1.45	Incisor	Equidae	Equini	Equus	–
N	G8-G49	~ 1.0	Lower left m2	Equidae	Equini	Equus	–

Fig. 1.

Archaeological and paleontological sites. Top panel: (A) Context of the study region in eastern Africa and location map of the Melka Wakena and Gadeb site-complexes on the Southeastern Ethiopian Highlands, including the location of Melka Kunture (MK) and Addis Ababa (Google Earth Pro 7.3.6); Bottom panel: (B) Stratigraphic section of Melka Wakena 2, showing the archaeological layers (red) and dated tephra deposits (dates rounded to the nearest decimal point, after Hovers et al.⁴²). Samples from localities MW3 and MW8 (Table S1) are stratigraphically coeval with MW2-L3; (C) Stratigraphy of Gadeb, including stratigraphic logs of localities 2 and 8 discussed in the text (modified after de la Torre⁴⁷. For additional details, readers are referred to Supplementary Text A.

Results

The δ¹³C and δ¹⁸O isotope measurements of enamel bioapatite are summarized in Table S1, where the data are grouped by faunal family and presented chronologically from the oldest to the youngest specimens, including the median for each isotopic profile. Figure 2 illustrates the overall isotopic ranges, whereas Fig. 3 displays the intra-tooth profiles of the analyzed taxa.

Fig. 2.

Solid box plots of δ¹³C (green) and δ¹⁸O (orange) intra-tooth values for (A–F) hippos, (G–K) bovids, and (L–N) equids from MW and GB. The isotopic data are grouped by faunal family, with each group arranged chronologically from the oldest (1.6 Ma) to the youngest (~ 1.0 Ma). Each box plot represents a single tooth. Specimen labels (A–N) correspond to the individual specimens listed in Table 1 and Table S1 (Excel file). The box represents the interquartile range, the vertical white line indicates the median, the whiskers extend to the most extreme non-outlier values, and the points correspond to outliers. Animal silhouettes are from https://www.phylopic.org.

Fig. 3.

Intra-tooth variation in stable carbon (δ¹³C, green circle) and oxygen (δ¹⁸O, orange circle) isotope ratios measured along the fossil teeth of hippos (A–F), bovids (G–K), and equids (L–N) from MW and GB. The isotopic data are grouped by faunal family, with each group arranged chronologically from the oldest (1.6 Ma) to the youngest (~ 1.0 Ma). Specimen labels (A–N) correspond to the individual specimens listed in Table 1 and Table S1 (Excel file). The time of tooth mineralization proceeds on the x-axis from left (older enamel–top of the tooth) to right (younger enamel – bottom of the tooth). Position on the x-axis refers to the location of each sample along the tooth growth axis (in millimeters). The samples collected from the teeth are scaled at 200 mm for hippos and 50 mm for bovids and equids. Animal silhouettes are from https://www.phylopic.org.

As shown in Fig. 2A–F, δ¹³C and δ¹⁸O values of six hippo isotopic profiles (n = 230) range from − 3.6‰ to + 2.09‰ (mean = − 0.08‰) and from − 7.9‰ to − 0.4‰ (mean = − 5.6‰), respectively. The δ¹³C values of five bovid teeth (n = 40) vary from − 0.4‰ to + 5.4‰ (mean = + 3‰), whereas δ¹⁸O values range from − 8.2‰ to − 0.5‰ (mean = − 4.4‰) (Fig. 2G–K). The three equid isotopic profiles (n = 42) show δ¹³C values ranging from − 2.9‰ to + 1.9‰ (mean = − 0.5‰) and δ¹⁸O values that range from − 5.9‰ to − 0.9‰ (mean = − 4.3‰) (Fig. 2L–N). Only three isotopic values (a single δ¹⁸O value in Fig. 2F; two δ¹³C values in Fig. 2G) deviate substantially from the overall data distribution and are therefore considered outliers.

Normal quantile plots and Shapiro-Wilk tests indicate non-normal distributions of δ¹³C and δ¹⁸O values for hippos (p = 0.0001 and p = 0.0038) and equids (p = 0.0069 and p = 0.0022) (Supplementary Fig. 2A and C). Bovids show non-normal δ¹³C values (p = 0.0262) but normally distributed δ¹⁸O values (p = 0.1878) (Supplementary Fig. 2B). Levene’s tests reveal unequal variances for both δ¹³C and δ¹⁸O values in hippos (p = 0.0001), and unequal δ¹³C but equal δ¹⁸O variances in bovids (p = 0.0033 and p = 0.0867) and equids (p = 0.0414 and p = 0.1033). Wilcox/Kruskal-Wallis tests confirm that all taxa differ significantly in both δ¹³C and δ¹⁸O rank (p = 0.0001).

Four of the six hippo specimens show significant cross-correlation between δ¹³C and δ¹⁸O profiles (R > 0.5) (Supplementary Fig. 3A, D, E and F), while two specimens exhibit no significant cross-correlation (R < 0.5) (Supplementary Fig. 3B and C). Among bovids, two specimens show significant cross-correlation (Supplementary Fig. 3G and H), whereas the remaining three do not (Supplementary Fig. 3I, J and K). For equids, two specimens also show significant cross-correlation (Supplementary Fig. 3L and M), and one shows no significant correlation (Supplementary Fig. 3N).

Discussion

Seasonal-to-interannual variability in δ¹³C and δ¹⁸O isotopic profiles

Stable carbon and oxygen isotope analyses of sequentially sampled herbivore tooth enamel document multi-seasonal variations in diet, drinking water, and local hydroclimatic and environmental conditions throughout the interval of enamel mineralization^48–55. In this context, seasonality refers to the cyclical fluctuations in rainfall throughout the year that influence ecosystem composition and resource availability.

Stable isotope analysis of hippos’ ever-growing canines and incisors (often referred to as tusks) provides high-resolution isotopic records across tooth enamel formation^56–59. Deciduous tusks erupt at birth and are shed at ~ 14 months, after which permanent open-rooted canines and incisors grow continuously throughout life⁶⁰. Interpreting temporal patterns along the tooth growth axis requires assumptions about enamel growth rates, which remain incompletely constrained and show substantial inter-individual variability. Passey et al.⁶¹ measured growth rates of 29 and 13 mm/year in the lower canines of two female common hippos of different ages, while Souron et al.⁵⁶ reported rates of 31–39 mm/year in modern females and noted less distinct isotopic cyclicity in fossil specimens, consistent with slower or more variable growth (20–30 mm/year)⁵⁶. Uno et al.⁶² documented additional variability related to sex and tooth position, with growth rates ranging from 19 to > 70 mm/year⁶². Because most published growth rates derive from modern, known-age individuals^56,61,62, their direct application to Pleistocene hippos assumes constancy in life history traits that may not hold under past environmental or selective pressures. While destructive approaches such as histological thin sections or die-saw dicing could potentially refine growth estimates^55,61–67, their application is limited in paleontological contexts.

In our dataset, two hippo canines from GB (Fig. 3D and E), for which a larger sample size is available (n = 61 and 51 samples, respectively), exhibit quasi-cyclical δ¹³C and δ¹⁸O variation patterns. Four of the six hippo intra-tooth profiles show δ¹³C variability within the C₄ range (Fig. 3A, B, D, and E), indicating consumption of grasses with differing C₄ photosynthetic pathways^68–71. By contrast, two specimens (Fig. 3C and F) display lower δ¹³C values consistent with mixed C₃–C₄ diets that incorporate C₃ vegetation such as cool-season grasses, browse, and aquatic plants^14–16,72.

As for the δ¹⁸O values, the specimen shown in Fig. 3E records a positive interannual shift of approximately 5 years, based on growth rates of ~ 30 mm/year in Pleistocene hippos⁵⁶. This trend is interpreted as reflecting increased evaporative water loss under relatively low rainfall, resulting in negative water balance and declining water levels during dry phases (see δ¹⁸O min-max values and seasonal cycles in Supplementary Fig. 4B). In contrast, negative interannual shifts in other isotopic profiles recorded in Fig. 3B and F suggest water bodies under positive water balance, where high rainfall and low evaporation cause rising water levels that occur during wet phases (Supplementary Fig. 4A and C).

Comparable patterns are observed in published modern and fossil hippo isotopic records⁵⁶ (Supplementary Fig. 5), supporting a hydrological interpretation for these variations. In particular, the modern hippo specimen is from the area of Sarh in southern Chad⁵⁶, where seasonal rainfall variability is influenced by ENSO phases^73,74. Based on this, we speculate that the interannual positive and negative trends recorded in our hippo isotopic profiles resemble the modern records of ENSO variability across the Horn of Africa^33–36, i.e., El Niño-like episodes corresponding to drier conditions, and La Niña-like episodes to wetter conditions. Further statistical and modeling analysis should be conducted to test this hypothesis, which we currently state with parsimony.

Cross-correlation analysis between intra-tooth δ¹³C and δ¹⁸O profiles clarifies the timing of ecological responses to hydroclimatic forcing. Some hippo individuals show strong positive correlations at zero lag (Supplementary Fig. 3E), indicating synchronous responses of water sources and vegetation to seasonal rainfall. In contrast, other individuals exhibit significant correlations at non-zero lags (Supplementary Fig. 3F), suggesting delayed dietary responses relative to hydrological inputs.

Among terrestrial herbivores, serially sampled bovid specimens (Fig. 3G–K) exhibit moderate intra-individual variability in both δ¹³C and δ¹⁸O values. Assuming published tooth growth rates of ~ 30–40 mm/year⁷⁵ for comparable taxa, one specimen (Fig. 3I) shows isotopic cyclicity that could be consistent with seasonal-scale variability, further supported by significant cross-correlation results.

Two equid specimens (Fig. 3L and N) record mixed C₃–C₄ diets with modest directional shifts in δ¹³C, whereas the equid profile in Fig. 3M indicates relatively stable C₄ grazing. In both bovids and equids, intra-tooth δ¹⁸O variability likely reflects a combination of seasonal climate signals, physiological differences, drinking behavior, and potential mobility rather than local climate alone ^{56,57,74–78}. These interpretations are necessarily constrained by several limitations of the dataset. First, specimen availability is restricted by the archaeological nature of the assemblage, which limits both the number of individuals and the taxonomic resolution of several specimens. In particular, two bovid teeth cannot be identified beyond the family level, preventing confident assignment to specific feeding guilds or drinking strategies. Second, the relatively small number of individuals analyzed restricts the possibility of evaluating population-level ecological patterns, which is a common issue in the literature^{1,8–10,56–58}. For these reasons, the interpretations presented here are framed conservatively and focus primarily on broad ecological trends rather than taxon-specific dietary or behavioral reconstructions.

Overall, the intra-tooth isotopic data from MW and GB indicate seasonal-to-interannual hydroclimatic variability within high-elevation Afromontane environments characterized by broadly similar ecological conditions across the sampled intervals. Frequent but variable coupling between δ¹³C and δ¹⁸O signals suggests that climatic fluctuations were propagated through local hydrological and vegetation systems, with no clear evidence for sustained changes in the analyzed herbivore diets within the limits of the available dataset.

Paleoecological inference

The dominance of C₄ isotopic signals in our dataset is consistent with expectations for elevations of MW and GB (2300–2400 m a.s.l.)²⁹. In modern eastern Africa, C₄ grasses are most abundant below 2000 m, while C₃ grasses dominate above 3000 m⁷⁹, although exceptions are documented⁸⁰. The mixed isotopic signals observed in our dataset, therefore, reflect the ecological heterogeneity expected at intermediate elevations rather than anomalous vegetation structure¹⁶.

Our results support the preliminary interpretation of the late Early Pleistocene faunal assemblages from MW and GB (Supplementary Text A). When Acheulean-making hominins occupied the Gadeb Plain, the local environment predominantly consisted of montane grasslands dominated by C₄ grasses, interspersed with C₃ vegetation, shrubs, and aquatic habitats^42,81. Studies of modern African ecosystems demonstrate that grazers commonly dominate herbivore assemblages even in landscapes with substantial woody cover^82,83. Accordingly, the abundance of grazers at MW and GB should not be interpreted as evidence for extensive open grasslands alone, and caution is warranted against simplistic vegetation reconstructions based solely on taxonomic composition^42,82–84.

Although no pollen or phytolith data are currently available from the archaeological horizons of MW and GB discussed here^27,42, the enamel carbon isotope data provide evidence of animal feeding preferences and associated habitats and, by extension, local environmental conditions during hominin occupations between 1.6 and ~ 1.0 Ma.

Because MW and GB share similar altitudinal and regional climatic settings with MK, we compared δ¹³C and δ¹⁸O values across these sites (Supplementary Text C; Supplementary Figs. 6–8; Supplementary Table 1). All three sites exhibit broadly overlapping isotopic ranges (Fig. 4A1–2), indicating comparable ecological conditions dominated by C₄ and mixed C₃–C₄ vegetation. This consistency suggests that broadly comparable Afromontane environmental conditions characterized the sampled high-elevation Ethiopian sites during the investigated intervals. We observe substantial overlap in δ¹⁸O values among the three sites, and a possible positive trend in minimum δ¹⁸O values between 1.6 and ~ 1.4 Ma (Fig. 5). This pattern is based on a limited number of measurements and is characterized by high variability and should be interpreted with caution. Rather than indicating a clear or systematic climatic shift, the data may reflect short-lived or local hydroclimatic fluctuations within the constraints of sparse sampling.

Fig. 4.

Scatter plots of δ¹³C and δ¹⁸O values (A) for (1) hippos, (2) bovids and (3) equids from MW, GB and MK; median δ¹³C and δ¹⁸O values (B) of the same taxa from high- (> 2000 m a.s.l.; MW, GB and MK) and low-elevation (≤ 1500 m a.s.l.; Oldupai Gorge, Lake Turkana Basin and Busidima Formation) eastern African sites. The isotopic data utilized for the graphs are provided in Tables S2 and S4. Animal silhouettes are from https://www.phylopic.org.

Fig. 5.

Scatter plots of δ¹⁸O min values of hippo intra-tooth profiles from MW-GB (pink circle) and MK (blue circle) plotted against chronology. Between 1.6 and ~ 1.4 Ma, positive stable oxygen isotopic values appear to indicate drier climatic conditions compared to those recorded at ~ 1.3 and 1.0 Ma. Median values are in yellow, with trends marked by a yellow dashed line.

Notably, niche modelling of the 1.6–1.4 Ma Ethiopian wolf (Canis simensis) from MW⁸⁵ generated Habitat Suitability Index (HSI) maps that are broadly consistent with the δ¹⁸O data. The model suggests higher HSI values at 1.6 Ma, associated with cooler conditions and a wider potential geographic range, and lower HSI values at 1.4 Ma, corresponding to warmer conditions and a reduced potential distribution⁸⁵. We emphasize that links between the isotopic signal and modeled habitat suitability would require additional, higher-resolution datasets for robust evaluation.

We compared the median δ¹³C and δ¹⁸O values (Table S4) from MW, GB, and MK (high-elevation sites > 2000 m a.s.l.) with those from equivalent taxa at lower-elevation sites (≤ 1500 m a.s.l.) in the East African Rift System, including Oldupai Gorge, Lake Turkana Basin, and Busidima Formation. As noted in Briatico et al.¹⁵, elevation does not appear to determine animal feeding behaviour in selecting vegetation, which remains dominated by C₄ resources across a wide altitudinal range. In contrast, high-elevation sites consistently yield lower δ¹⁸O values than low-elevation sites, reflecting the isotopic altitude effect^86–88 and predictably indicating cooler conditions in the Ethiopian highlands (Fig. 4B1–2)¹⁵. In particular, hippos depend on submersion in stable water bodies for thermoregulation, in the absence of sweat glands⁸⁹, and are not a seasonally migratory species. Thus, hippo isotopic signals primarily reflect local hydrology and water bodies rather than long-distance movement. For bovids and equids, however, seasonal altitudinal movements cannot be excluded based on carbon and oxygen isotope data alone.

The persistent, albeit intermittent, archaeological, paleontological, and anthropological evidence recorded in the Early Pleistocene multi-layered deposits of the Ethiopian highlands^{27–42,90−97} indicates that these environments repeatedly supported mammalian species, including Homo erectus sensu lato populations over extended periods. Rather than representing environmentally constrained settings, the Ethiopian highlands appear to have provided recurrent access to water and vegetation resources within the Afromontane environmental framework. Taken together, our results indicate that late Early Pleistocene high-elevation ecosystems were characterized by recurrent hydroclimatic variability within broadly comparable Afromontane environmental conditions, circumstances that may have promoted long-term hominin persistence. Distinct from the lowland Rift System, such environments may serve as informative analogues for the ecological settings faced by hominins during their dispersal out of Africa.

Methods

Stable isotope analysis

A total of 312 enamel samples were collected from 14 fossil teeth (Supplementary Fig. 1) excavated at various archaeological sites within the MW and GB areas (MW2, MW3, MW8; GB2, GB8), dated between 1.6 and ~ 1.0 Ma (Fig. 1). The fossil specimens include Hippopotamidae (n = 230 enamel samples), Bovidae (n = 40), and Equidae (n = 42) (Table S1). The specimens were purposely selected for intra-tooth sequential sampling in October/November 2023 at the Ethiopian Heritage Authority (EHA) facilities in Addis Ababa (Ethiopia), with the approval of EHA curators. Each sample lab code corresponds to the site abbreviation (MW = Melka Wakena; GB = Gadeb), followed by a progressive sample number (e.g., MW 1, MW 2; or GB 1, GB 2). First, the tooth surface was inspected to select the best sampling location. This area was systematically cleaned with a diamond drill bit (> 2.0 mm) to remove any extraneous material. Then, several enamel samples were taken with a diamond-tipped drill (< 2.0 mm) along the entire length of the tooth⁵⁶ (∼2–2.5 mm intervals) to obtain 12–15 mg of enamel powder per sample, allowing for potential material loss during subsequent laboratory procedures. After collection and export, pretreatment was conducted at the Stable Isotope Preparation Lab, Department of Anthropology of the University of Connecticut (UConn) (Storrs, Connecticut, U.S.A.), following the protocol by Balasse et al.⁹⁸. The powdered enamel samples were treated with 0.1 M acetic acid (CH₃COOH) for 4 h at room temperature to remove exogenous carbonate. The remaining samples were rinsed once or twice with MilliQ^® water (Millipore) until the solution reached neutrality; then, the reaction tubes were placed in a desiccating oven to dry overnight at 50 °C. Weighed samples (1 mg) were sent for carbon and oxygen isotope analysis at the Yale Analytical and Stable Isotope Center (YASIC) of Yale University (New Haven, Connecticut, U.S.A.), using an automated carbonate preparation device (Kiel-IV) coupled to an isotope ratio mass spectrometer (Thermo MAT 253). Isotope measurements were calibrated using repeated analysis of three in-house and two international standards (NBS-19 and NBS-18). The analytical precision is ± 0.1‰ for δ¹⁸O and ± 0.08‰ for δ¹³C (1 σ). Carbon and oxygen isotope compositions are reported in standard δ-notation relative to the Vienna Pee Dee Belemnite (VPDB) standard. Details regarding the isotopic signal interpretation are provided in Supplementary Text B.

Statistics and cross-correlation analysis

All statistical analyses were conducted using JMP Pro 18 (version 18.2.0; licensed by the Hebrew University of Jerusalem), with a significant threshold of p = 0.05. The δ¹³C and δ¹⁸O values were analyzed by faunal family to account for taxon-specific ecological variability. Normal quantile plots and the Shapiro-Wilk test were used to assess the assumption of normality. In this framework, the null hypothesis (H₀) states that the data are drawn from a normally distributed population, whereas the alternative hypothesis (H₁) indicates a non-normal distribution. Because group sizes are unequal, homogeneity of variance was evaluated using Levene’s test, which assesses whether the absolute deviation from group means differ among groups. Here, H₀ assumes equal variances across taxonomic groups, while H₁ indicates unequal variances.

When the assumptions of normality and/or homoscedasticity were violated, non-parametric tests were applied. Differences in δ¹³C and δ¹⁸O values among more than two independent taxonomic groups were evaluated using the Kruskal–Wallis test, which compares rank distributions across groups. When the Kruskal–Wallis test indicated a significant overall effect, pairwise post-hoc comparisons were conducted using Wilcoxon rank-sum tests. For comparisons involving two independent groups only, Wilcoxon rank-sum tests were used directly. The same analytical framework was applied when comparing δ¹³C and δ¹⁸O values across taxa and archaeological sites (MW, GB, and MK; Supplementary Text C). The degree of correlation between the δ¹³C and δ¹⁸O values within individual intra-tooth isotopic profiles (Supplementary Fig. 3) was assessed using cross-correlation analysis implemented with the “ccf” function in RStudio (version 2023.06.0 + 421). This analysis was used to evaluate whether changes in δ¹⁸O values could temporally precede or coincide with shifts in δ¹³C values. Cross-correlation coefficients (R) ≥ 0.5 were interpreted as strong correlations, whereas values < 0.5 were considered weak.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

G.B., G.H. and E.H. conceived and designed the research paper. The faunal fossil remains were identified by B.M.-N., N.K. and H.H. Sampling and sample processing for isotopic analysis were carried out by G.B. E.H. and T.G. are the principal investigators of the project and are responsible for the project research objectives, administration and funding acquisition. A.A., A.R., E.M.N., and P.R.R. provided contextual information. G.B. wrote the paper with inputs from G.H. and E.H. All the authors commented, edited and approved the final version of the submitted manuscript.

Funding

The research activities at MW, including stable isotope analysis, are funded by the Fritz Thyssen Foundation (grant number: #10.21.1.07AA), awarded to E.H. and T.G. B.M-N. is funded by María de Maeztu Grant 2024 (CEX2024-01485-M by MICIU/AEI/10.13039/501100011033) and Generalitat de Catalunya 2021SGR 01238 (AGAUR). This paper is part of the Lady Davis Postdoctoral Fellowship (2023–2025) awarded to G.B. at the Department of Prehistory, Institute of Archaeology of the Hebrew University of Jerusalem (Israel). G.B. was further funded by the INQUA (International Union for Quaternary Research) fellowship 2025. G.B. is currently supported by the project “POOL - Investigating cultural and biological scenarios of late Neanderthals and Homo sapiens in Poland” (FIS-2023-01196, CUP J53C24004280001), led by Prof. Andrea Picin and funded by the Italian Ministry of University and Research (D.D. MUR n. 23178 of 10/12/2024).

Data availability

The data generated and analyzed during the study are available in this published article and its Supplementary Information files. Original fossils are housed at the EHA facilities in Addis Ababa; the remaining enamel powder samples are stored at the Stable Isotope Preparation Lab (UConn, Department of Anthropology).

Code availability

No code has been created for this study.

Declarations

Competing interests

Prof. Asfawossen Asrat is an Editor of Scientific Reports but was not involved in the editorial handling or peer review of this manuscript. The remaining authors declare no competing interests.