Dietary versatility of Early Pleistocene hominins.

New geochemical data from the Malawi Rift (Chiwondo Beds, Karonga Basin) fill a major spatial gap in our knowledge of hominin adaptations on a continental scale. Oxygen (δ18O), carbon (δ13C), and clumped (Δ47) isotope data on paleosols, hominins, and selected fauna elucidate an unexpected diversity in the Pleistocene hominin diet in the various habitats of the East African Rift System (EARS). Food sources of early Homo and Paranthropus thriving in relatively cool and wet wooded savanna ecosystems along the western shore of paleolake Malawi contained a large fraction of C3 plant material. Complementary water consumption reconstructions suggest that ca. 2.4 Ma, early Homo (Homo rudolfensis) and Paranthropus (Paranthropus boisei) remained rather stationary near freshwater sources along the lake margins. Time-equivalent Paranthropus aethiopicus from the Eastern Rift further north in the EARS consumed a higher fraction of C4 resources, an adaptation that grew more pronounced with increasing openness of the savanna setting after 2 Ma, while Homo maintained a high versatility. However, southern African Paranthropus robustus had, similar to the Malawi Rift individuals, C3-dominated feeding strategies throughout the Early Pleistocene. Collectively, the stable isotope and faunal data presented here document that early Homo and Paranthropus were dietary opportunists and able to cope with a wide range of paleohabitats, which clearly demonstrates their high behavioral flexibility in the African Early Pleistocene.

Dietary adaptations are responsible for significant behavioral and ecological differences among humans and other primates (1, 2). Early hominin diets in Africa before 4 Ma were dominated by C3 resources and diversified over time (3). Increasing contributions of C4 plants were triggered by biomes gradually shifting to more open C4 grasslands since the Late Miocene (4–8). In eastern Africa, dietary versatility was very likely an integral part of early hominin adaptations to open landscapes (7, 9–12), yet it is unclear whether this assumption holds for African Early Pleistocene hominin evolution in general.

Here, we present multiproxy carbon (δ13C), oxygen (δ18O), and clumped (Δ47) isotope data focusing on paleoenvironmental patterns and diets of Paranthropus boisei and Homo rudolfensis. Both hominin taxa coexisted around 2.4 Ma in the Western (Malawi) Branch, the southern part of the East African Rift System (EARS), a region representing a large geographical gap in our knowledge of Early Pleistocene hominin adaptations on a continental scale (13–16). Paleosol development characterizes the lacustrine and deltaic deposits of the Chiwondo Beds (Karonga Basin, northern Malawi) (16). These deposits yielded remains of H. rudolfensis at Uraha (17) and Mwenirondo (15), as well as Paranthropus remains at Malema, which were assigned to P. boisei due to, for example, tooth morphology and palatal height (18). Collectively, these hominin localities of the Chiwondo Beds are less than 50 km apart and are situated today in the Zambesi Ecozone of the African Savannas just south of the boundary to the Somali-Masai Ecozone, and hence outlining the southernmost extent of the Intertropical Convergence Zone. Given this unique location, the Chiwondo Beds data are crucial for understanding hominin–environment interactions and are broadening our view on habitat flexibility and dietary adaptation in early hominins.

The Karonga Basin lies in the southeastern African hominin corridor region, connecting eastern and southern African endemic faunal zones (19). Hominins thriving in these faunal zones indicate different behaviors: Eastern African Paranthropus robustus had a mixed to C4-dominated diet around 2.4 Ma, while P. boisei, younger than 2 Ma, provides robust evidence for C4-dominated consumption in this region. This clearly distinguishes them from age-equivalent southern African P. robustus, which shows mixed or C3-dominated diets (20–23). Early Homo had a highly variable diet, including C3 and C4 resources (3, 6).

In contrast to the mostly open C4 grasslands of the northern EARS, the Malawi Rift was dominantly covered by persistent wooded savannas throughout the Plio-Pleistocene (24). Despite exhibiting a woody cover exceeding 60% and only regional patches of open C4 grasslands (24, 25), this part of the EARS was populated by early P. boisei and H. rudolfensis, pointing to a much broader dietary versatility of early East African P. boisei than previously assumed.

We utilize a multiproxy approach for reconstructing habitats and diets during the earliest phases of coexisting Homo and Paranthropus in the Malawi Rift. Our study of δ13C and δ18O values in tooth enamel of H. rudolfensis and P. boisei provides insight into dietary preferences, flexibility, and adaptation in variably open and closed savanna environments. In contrast, intratooth δ13C and δ18O time series from contemporaneous equids (Eurygnathohippus sp.) and bovids [Alcelaphinae (Alc.) Megalotragus sp.], in addition to δ13C and δ18O data from pedogenic carbonate of the hominin fossil sites, allow the reconstruction of vegetation patterns (25, 26). The Δ47 data from paleosols of the hominin fossil sites are from the southern part of the EARS and provide insight into regional temperature differences during the time of early hominin evolution. Our results show that early hominins must have been able to adapt to different environmental settings and diets as early as 2.4 Ma, indicating high behavioral flexibility already in the early stages of hominin evolution, as first indicated by, for example, Turkana Basin Kenyanthropus platyops individuals at 3.4–3.0 Ma (3).

Results

We report δ13C and δ18O data from tooth enamel of all three Chiwondo Bed hominin individuals, as well as age-equivalent bovid (n = 2) and equid (n = 3) specimens [sample IDs stated according to the Hominin Corridor Research Project (HCRP), with location ID followed by the individual sample number]. We complement these by δ13C and δ18O (n = 199 samples) as well as Δ47 thermometry (n = 12 samples) data of pedogenic carbonate. All samples originate from unit 3-A2 (ca. 2.5–2.3 Ma) of the Chiwondo Beds (27) in the Karonga Basin at the northwest margin of Lake Malawi. In addition, we present modern soil temperatures that were monitored at three locations (full shade, partial shade, and full sun) throughout a complete year (n = 26,280 measurements) in the depth of typical pedogenic carbonate precipitation within the region (∼40 cm below the surface). Results are shown in Figs. 1–3 and listed in .

Fig. 1.

Intratooth δ13C and δ18O variations in equid and bovid teeth vs. distance from the crown; hence, time of tooth growth. The hominin tooth enamel sample quantity was too small to determine intratooth patterns; each sample averages almost the entire interval of enamel development. Bovids and equids were coexistent with the early hominins, and fossils were collected at the same localities. Different patterns indicate different migratory behavior of the individuals and/or seasonal changes. Eur., Eurygnathohippus; gen. indet., genus indeterminate.

Fig. 3.

Mean monthly soil temperatures (circles) measured for this study (2016/2017) with day and night air temperatures (diamonds), mean precipitation (blue), and average sunshine hours (orange) (32). Soil temperatures (columns on the left represent the complete temperature span of the full year in the respective color) generally follow air temperature patterns and correlate with sunshine hours, especially in soil directly exposed to the sun; here, the soil is usually warmer than the air due to the effect of direct solar radiation. Bars at the left of the figure show the maximum range of measured soil temperatures at the different locations. Temp., temperature.

Fig. 4.

Comparison of environmental proxies between the Malawi Rift and Eastern Rift. Biome cover, relative proportions of biomes based on paleosol δ13C values (5), and boundaries between biomes are described in . The Malawi Rift was dominated by a more wooded environment (in a generally moister and cooler climate; Fig. 5) during the time of early hominin evolution compared with the Eastern Rift.

Hominin and Herbivore Tooth Enamel δ13C and δ18O Data.

H. rudolfensis.

Each tooth shows only a small range in δ13C and δ18O values, but the two individuals show a clear geochemical distinction in their tooth enamel, with a mean difference of 3.8‰ in δ13C and 2.5‰ in δ18O (Fig. 1).

The δ13C and δ18O values from the Uraha individual HCRP-U18-501 (n = 3) are −6.3‰, −6.0‰, and −5.3‰ (mean = −5.9‰, σ = 0.5‰, C3/C4 ratio = ∼65:35) and 25.9‰, 26.5‰, and 26.9‰ (mean = 26.4‰, σ = 0.5‰), respectively.

The δ13C and δ18O values from the Mwenirondo specimen HCRP-MR10-1106 (n = 3) are more negative, with δ13C values of −9.9‰, −9.8‰, and −9.5‰ (mean = −9.7‰, σ = 0.2‰, C3/C4 ratio = ∼95:5) and δ18O values of 23.6‰, 23.9‰, and 24.1‰ (mean = 23.9‰, σ = 0.2‰).

P. boisei.

The molar from the Malema individual HCRP-RC11-911 (n = 3) shows δ13C values of −7.0‰, −6.9‰, and −6.8‰ (mean = −6.9‰, σ = 0.1‰, C3/C4 ratio = ∼70:30). The δ18O values are 26.3‰, 26.8‰, and 27.2‰ (mean = 26.8‰, σ = 0.5‰).

Bovids (Alcelaphinae).

The Alcelaphini Megalotragus sp. molar HCRP-U18-401 (n = 9) from the H. rudolfensis fossil site has much more positive δ13C values compared with the hominins, with a range between −2.6‰ and 0.8‰ (mean = −1.7‰, σ = 1.3‰, C3/C4 ratio = ∼30:70). The δ18O values range between 26.3‰ and 28.4‰ (mean = 27.3‰, σ = 0.7‰).

The δ13C and δ18O values of Alc. genus and sp. indeterminate HCRP-RC11-595 (n = 19) from the P. boisei locality at Malema are generally even more positive than the Megalotragus sp. data; δ13C values range from 0.7 to 2.0‰ (mean = 1.5‰, σ = 0.4‰, C3/C4 ratio = ∼2:98). Only a slight positive shift of 1.3‰ is present in this very narrow range. The δ18O values show a curved pattern between 27.5‰ and 29.5‰ (mean = 28.5‰, σ = 0.5‰).

Equids (Eurygnathohippus sp.).

Similar to Alcelaphinae, a very narrow range in δ13C values and a larger variability in δ18O values characterize the stable isotope data of the three sampled Eurygnathohippus sp. teeth from different individuals, which show a clear geochemical distinction, especially in δ13C (Fig. 1).

The premolar of HCRP-U17-390 (n = 11) (location U17 is the same age and spatially correlated with H. rudolfensis from U18) has δ13C values between −1.2‰ and 0.2‰ (mean = −0.2‰, σ = 0.5‰, C3/C4 ratio = ∼15:85). The δ18O values are between 27.5‰ and 28.7‰ (mean = 28.1‰, σ = 0.3). HCRP-RC11-528 (n = 12) from the P. boisei site has δ13C values between −2.4‰ and −1.3‰ (mean = −1.9‰, σ = 0.4‰, C3/C4 ratio = ∼30:70) and δ18O values from 27.2 to 28.6‰ (mean = 28.1‰, σ = 0.5‰). HCRP-RC11-545 (n = 14) also shows more negative δ13C values than the other Eurygnathohippus sp., with a very narrow range from −5.7 to −5.0‰ (mean = −5.3‰, σ = 0.2‰, C3/C4 ratio = ∼50:50), while the δ18O values are generally more positive, between 29.0‰ and 30.3‰ (mean = 29.8‰, σ = 0.4‰).

Pedogenic Carbonate Δ47 Temperatures.

Pedogenic carbonate Δ47 temperatures from the H. rudolfensis site U18 (Uraha) range from 19 to 38 °C (mean = 28 °C, σ = 8.3 °C; n = 6). The dataset seems to be bimodal, with values grouping at relatively high temperatures (mean = 35 °C, σ = 2.2 °C; n = 3) and another cluster of low temperatures (mean = 21 °C, σ = 1.8 °C; n = 3). A trend over time (i.e., across the sampled fossil horizon) is not present.

Paleosol temperatures from the P. boisei site RC11 (Malema) show a smaller variation, with values between 22 °C and 29 °C (mean = 26 °C, σ = 2.9 °C; n = 6).

Karonga Basin soil water δ18O values (δ18OSW) were calculated using the δ18O values of the carbonate and their Δ47 temperatures according to the methods of Kim et al. (28) and Kim and O’Neil (29), resulting in δ18OSW values between −5.4‰ and −1.5‰ (mean = 3.6‰, σ = 1.2‰ °C; n = 12).

Pedogenic Carbonate δ13C and δ18O Values.

Uraha.

The δ13C values of pedogenic carbonate from a soil profile at the H. rudolfensis site U18 (n = 52) range between −11.0‰ and −7.0‰ (mean = −9.1‰, σ = 0.8‰). The estimated fraction of woody cover (fwc) (5) ranges between 0.5 and 0.8. The corresponding δ18O values lie between 23.6‰ and 24.9‰ (mean = 24.3‰, σ = 0.3‰).

Malema.

The δ13C values of pedogenic carbonate from the P. boisei site RC11 (n = 147) cover a similar range as at the Uraha site, with δ13C values between −10.7‰ and −6.3‰ (mean = 8.5‰, σ = 1.1‰). The estimated fwc values range from 0.4 to 0.8. The δ18O values show a slightly larger variation than at the Uraha site, with values between 23.3‰ and 24.7‰, excluding one outlier with a value of 25.2‰ (mean = 24.1‰, σ = 0.4‰).

Correlation Between Stable and Clumped Isotope Data.

(Negative) correlation between datasets was determined by using the Pearson correlation coefficient. Statistically (negative) correlation is significant at the 95% level.

We observe a statistically relevant positive correlation between δ18O and δ13C values in the Uraha (P = 0.5) and Malema (P = 0.8) datasets, suggesting a higher fraction of woody cover during periods of wetter climates.

Reconstructed Δ47 soil temperatures and δ18O and δ13C values show a strong negative correlation at the Malema site RC11 (P = −0.9 for δ18O and P = −0.8 for δ13C); hence, high (soil) temperatures coincide with a moister climate, which produces an even denser canopy cover serving as a buffer and resulting in less thermal loss at night. This is complemented by the small range in soil temperatures at this site (Fig. 5).

Fig. 5.

Clumped isotope paleosol temperatures (○, with black lines representing the mean) from paleosol carbonates of the Karonga Basin (n = 12) vs. the Turkana Basin (n = 8) (30) between 2.8 Ma and 1.8 Ma. For direct comparison, the Turkana Basin ∆47 data were recalculated to the absolute reference frame using the technique of Uno et al. (49) and the calibration of Cerling et al. (50) (). Gray bars represent today’s minimum to maximum soil temperatures in 40 cm (Karonga Basin) or 50 cm (Turkana Basin), with the thick line as the overall mean. Turkana Basin soil temperatures are measured under sparse deciduous bush vegetation, while the thin lines in the Karonga Basin data represent mean temperatures of soils exposed to full sun (top), partial shade (middle), and full shade (bottom). Black stars show MAATs. Similar to modern data, reconstructed soil temperatures are generally lower in the Karonga Basin, which created a landscape with different (food) resources than in the hotter Turkana Basin. Temp., temperature.

Modern Soil Temperatures.

Soil temperatures were measured hourly 40 cm below the surface under conditions of full shade, partial shade, and full sun (Fig. 3).

Full shade.

Over the course of 1 y (August 1, 2016 to July 31, 2017), soil temperatures in sandy soil under dense tree cover range from ca. 23 to 27 °C (mean = 26 °C, σ = 0.9 °C; n = 8,760). Temperature differences between day and night, as well as between dry and wet seasons, are relatively small (generally ca. 0.5 °C and 2 °C, respectively).

Partial shade.

Recorded soil temperatures (August 1, 2016 to July 31, 2017) show a range of 21 to 32 °C (mean = 27 °C, σ = 2.1 °C; n = 8,760) ∼50 m away from the full-shade location. Here, temperatures below 24 °C are limited to a few outliers (n = 98; i.e., 1%) during the wet season, probably due to extreme rainfall and resulting soil water saturation. Day/night differences at around 0.5 °C are comparable to the full-shade site, but the difference between dry and wet seasons is more pronounced, with almost 6 °C warmer soil temperatures during the end of the dry season.

Full sun.

Soil temperature loggers were put directly in the Chiwondo Bed sandy deltaic deposits at the open habitat of the Malema P. boisei site (RC11). The recorded temperatures (August 15, 2016 to August 14, 2017) are higher than at the locations of full shade and partial shade and range from 25 to 37 °C (mean = 31 °C, σ = 3.2 C; n = 8,760). Day/night temperature differences are >2 °C and increase during the rainy season, with maximum day temperatures of generally 36 °C contrasting with night temperatures, which regularly drop below 32 °C.

Discussion

Paleoenvironmental Reconstruction of H. rudolfensis and P. boisei Sites in the Karonga Basin.

The results of paleosol stable isotope geochemistry from the Karonga Basin hominin sites indicate woodland, bushland, and scrubland to grassy woodland environments characterized by 40–80% woody cover (Figs. 3 and 4 and ) during the time of early hominin occupation at ca. 2.5–2.3 Ma. Reconstructed mean paleosol Δ47 temperatures attain 26 °C for the Malema P. boisei (Malema) site and 28 °C at the Uraha H. rudolfensis (Uraha) site, which are similar to today’s partial shade soil temperatures in the region (Figs. 3 and 5). Especially the Malema Δ47 temperatures (usually reflecting soil temperatures below 30 cm), which are considered to have a strong seasonal bias toward dry and warm periods when carbonate precipitation predominantly occurs, are relatively cool compared with the Eastern Rift paleosol temperature data of 2.8–1.8 Ma, which are typically in excess of 35 °C (30) (e.g., Turkana; Fig. 5), indicating that hominins thrived under different temperature conditions in different parts of the EARS during the Early Pleistocene.

The present-day Karonga Basin is characterized by a tropical savanna climate (31) with mean annual air temperatures (MAATs) of ∼25 °C and mean annual precipitation (MAP) of ∼1,170 mm·y−1 (32), which results in the wooded savanna ecosystem of the Zambezian woodland savanna. The Turkana Basin, in contrast, is characterized by a hot desert/semiarid climate (31), with MAATs exceeding 29 °C and MAP < 200 mm (32), creating open savanna settings with dominantly C4 grasses, similar to reconstructed landscapes in the Early Pleistocene (Fig. 4). Soils exposed to direct sunlight at both localities exhibit soil temperatures (>35 °C) that are typically warmer than measured air temperatures (5, 30) (Figs. 3 and 5), reflecting the effect of radiative heating under sparse canopy cover. This indicates that the generally lower Karonga soil temperatures during the Early Pleistocene were triggered, among other factors, by a higher fraction of woody cover in the southern part of the EARS. Complementary calculated Karonga Basin δ18Osoilwater data (depending on δ18O values of the carbonate and their Δ47 temperature) show ca. 5‰ more negative average values than the Turkana data (30), pointing to a smaller effect of evaporation of the soil water in the southern part of the EARS. The δ18O values from Karonga Basin pedogenic carbonates are generally lower than data observed at most other African hominin sites (ref. 33 and references therein), which suggests more humid conditions and less influence of evaporation in the cooler and more wooded habitats of the Malawi Rift in contrast to those of eastern Africa.

Consequently, larger fractions of woody cover imply more shade, shelter, and dietary resources for early hominins in the Zambesi Ecozone in the Malawi Rift, accompanied by lower air temperatures during the dry season. Ecosystem diversity of the Pleistocene Karonga Basin is reflected by the high (<30 °C) reconstructed soil temperatures of individual pedogenic carbonates from the Uraha site, pointing to temporary direct sun exposure of the soil in a possibly patchy landscape (also ref. 25).

Intratooth δ13C variability (−5.7 to 2.0‰) of herbivorous mammals coexisting with H. rudolfensis and P. boisei indicate that both C3 and C4 vegetation was available for consumption, which permitted highly variable foraging strategies for migrating mammals (25). Alcelaphini and equids were able to either feed almost selectively on C4 resources or exhibit mixed grazing and browsing foraging strategies (25) (Fig. 1). These migrating and often specialized feeders, however, do not necessarily reflect the dominant paleoenvironment, but typically select their dietary niche. On the contrary, (Pleistocene) suids are considered generalists, and suid enamel δ13C values hence average local resource δ13C variability. The δ13C tooth enamel data for Karonga Basin Plio-Pleistocene suids indicate a predominantly C3 browsing diet during the time of early Homo sp. evolution (24) (Fig. 6). This suggests that early Karonga Basin hominins had access to a wide range of vegetation types, such as gallery forest near the freshwater sources and open C4 grasslands in the more elevated or arid regions at the rift shoulders, providing a diverse spectrum in food supply and shelter (cf. refs. 24, 25).

Fig. 6.

δ13C values of hominin tooth enamel with respect to the age of the fossils. Karonga Basin Malawi H. rudolfensis and P. boisei of this study are compared with Homo sp., P. boisei, and P. aethiopicus of the Turkana Basin (3) and P. robustus from South Africa (20–23). The percentage of C4 consumption is calculated using the method of Cerling et al. (5) and is explained in . Especially P. boisei of the Karonga Basin therefore had a diet much more influenced by C3 intake than their younger Eastern Rift relatives, indicating an adaptation to more open environments in the Eastern Rift. Gray areas indicate δ13C values of coexiting suids for the Karonga Basin (16) (Left) and the Eastern Rift (49–56) (Right), reflecting a dominantly C3 diet of Karonga Basin omnivores and a highly C4-influenced diet of Eastern Rift ones, which, in turn, reflects the main vegetation type of the hominin sites.

Diets of Early H. rudolfensis and P. boisei in the Malawi Rift.

The δ13C values of the Karonga Basin hominins at ca. 2.4 Ma suggest variable, mixed diets with only small portions of C4 intake (<40%; Fig. 6). Stable isotope geochemical data suggest that the two H. rudolfensis individuals showed different foraging strategies. HCRP-U18-501 was a mixed feeder with 60–70% C3 resources, while HCRP-MR10-1106 fed almost exclusively on C3 resources and had less than 6% C4 intake (Fig. 6). This suggests an adaptation to versatile ecosystem parameters with varying types of food supply and/or selective foraging strategies. The coexisting P. boisei individual reflects a similar diet, where C3 resources amounted to ca. 70%. Early Paranthropus in the Karonga Basin therefore relied heavily on the presence of woodland habitats.

The two analyzed hominin taxa are geochemically indistinguishable due to the relatively large spread in δ13C and δ18O values measured for the two individuals of H. rudolfensis and the fact that P. boisei isotope data lie within the same range (Figs. 1 and 6).

The δ13C values of Homo and Paranthropus point to food intake that consisted dominantly of C3 resources, such as forest foods, while xeric plant foods [e.g., C4 underground storage organs (USOs), such as tubers, corms, roots, and bulbs, but also sedges, termites (20, 34, 35), or C4 grass leaves (36)] were not dominant. However, δ13C data alone may not decipher C3 and C4 resources of dietary patterns because plant-based, meat-based, and omnivore diets cannot be distinguished with this method (e.g., ref. 37).

The narrow range of soil water and paleosol carbonate δ18O values from the Karonga Basin hominin sites (Fig. 2) suggests relatively constant climatic conditions in the vicinity of paleolake Malawi. H. rudolfensis individual HCRP-MR-1106 shows the lowest δ18O values, while the δ18O values of H. rudolfensis HCRP-UR-501 and P. boisei HCRP-RC-911 overlap. The relatively low δ18O values in concert with the δ13C data render it likely that both hominin taxa had access to drinking water only little affected by evaporation (and associated 18O-enrichment). The δ18O values of the hominin teeth are generally lower than the δ18O values from coexisting bovids and equids (Fig. 1). The spread in δ18O values of tooth enamel of both Megalotragus sp. and Eurygnathohippus sp. is larger than the scatter measured for the hominin individuals. Migrating herbivores had access to different drinking water sources of distinct oxygen isotope compositions, whereas hominins probably remained rather stationary near freshwater sources, such as tributaries of paleolake Malawi.

Fig. 2.

δ13C, δ18O, and ∆47 values for Karonga Basin paleosol carbonate from the 2.5- to 2.3-Ma-old hominin fossil sites of Malema (A) and Uraha (B). Definitions of biome abbreviations are provided in Fig. 4. The black line shows modern MAAT.

Comparison with Other African Hominins.

Stable carbon isotope analysis of African hominins suggests an increasing intake of C4 biomass between ca. 4.1 Ma and 1.4 Ma throughout eastern, southern, and central Africa (Fig. 6). While hominins prior to 4 Ma (e.g., Ardipithecus ramidus and Australopithecus anamensis) depended almost fully on C3 resources, an expansion toward mixed diets that included C4 resources is observed starting at 3.75 Ma [e.g., Australopithecus bahrelghazali, Australopithecus afarensis, K. platyops, Hominini indet (3, 4, 7, 20, 38–42)]. By ca. 2 Ma, the genus Homo had a highly variable mixed C3/C4 diet in the Turkana Basin, while P. boisei was characterized by a strongly C4-dominated diet with only <25% C3 intake (Fig. 6), pointing to a diet characterized by USOs and other C4 resources. The evolution of (open) wooded C4 grassland savannas in the Eastern Rift presents a very different ecological setting from the persistent wooded savanna environments of the Karonga Basin (24, 25) (Fig. 4). This indicates different feeding strategies of Homo sp. and Paranthropus sp. in the increasingly open grassland savannas of the Eastern Rift (3, 6), and is a potential key for understanding the dietary evolution of early hominins. H. rudolfensis HCRP-U18-501 had a similar diet as younger (ca. 2.3–1.4 Ma) Homo species from the Turkana Basin (3). The almost exclusive C3 diet of individual HCRP-MR10-1106, however, is rarely observed in Pleistocene Turkana hominins; the widely variable diet of eastern African Homo sp. rarely includes such selected feeders (3) (Fig. 6), indicating an adjustment to the denser wooded landscape of the Malawi Rift and/or to the more open savanna landscapes in the Eastern Rift.

Based on the existing spatial and temporal coverage of P. boisei remains, P. boisei adopted a significantly different diet in the diverging biomes of the Eastern Rift and Malawi Rift, supporting an unexpected versatile adaptive behavior. This predicts significantly different feeding strategies of P. boisei in the Karonga Basin at ca. 2.4 Ma compared with Turkana Basin individuals dated <2 Ma, which consumed predominantly C4 biomass (>40%) (6). The latter most likely relied heavily on the intake of sedges (37, 43). In contrast, the Karonga Basin P. boisei foraging strategies even involved a higher fraction of C3 food resources than coexisting Turkana Basin Paranthropus aethiopicus, which, in turn, consumed less C4 plants than the younger P. boisei in the same region (3, 6). South African P. robustus (Sterkfontein Valley and Gauteng; ca. 1.8–1.0 Ma) probably also lived in a woodland environment and had a similarly mixed diet like P. boisei in the Karonga Basin (3, 20–23).

Collectively, this points to an adaptive behavior toward increasing C4 biomass consumption [e.g., sedges, termites (20, 37) if these were easily accessible] but an enduring C3-dominated diet in wooded settings throughout the time of early hominin evolution. However, due to the lack of Karonga Basin hominins younger than 2 Ma and/or sufficient data of Eastern Rift hominins older than 2 Ma that thrived in a woodland-dominated ecosystem, it is difficult to determine if the adaptation could be a supraregional effect or the result of local ecosystem pressure.

Conclusions

Stable carbon and oxygen isotope data of hominin fossil tooth enamel from the Malawi Rift show that early (ca. 2.4 Ma) H. rudolfensis and P. boisei included a large fraction of C3 food resources in their diets. Abundant C3 resources were provided by relatively cool and well-watered wooded savanna ecosystems in the vicinity of paleolake Malawi in the southern part of the EARS. Younger (<2 Ma) Paranthropus individuals from the Eastern Rift incorporated an increasing amount of C4 resources in hotter, more arid, and open savanna settings, while P. robustus maintained a C3-dominated diet in the wetter and more mesic environments of southern Africa. Throughout the Early Pleistocene, Homo shows a high versatility in the diverse habitats of the EARS. H. rudolfensis and P. boisei were therefore dietary generalists, able to adapt to different paleohabitats, successfully utilizing a broad range of ecosystems, including freshwater environments near the tributaries to paleolake Malawi.

Methods

Teeth of three hominins, three equids, and two bovids temporarily housed in the Senckenberg Research Institute and from the Cultural and Museum Centre Karonga were sampled using a high-speed, rotary, diamond-tip drill to obtain 2.5–5 mg of enamel powder from each of the up to 19 samples per tooth. To remove organic and potential diagenetic carbonate, enamel was pretreated with 2% sodium hypochlorite solution for 24 h, followed by treatment with 1 M Ca-acetate acetic buffer solution for another 24 h (44). Additionally, 199 pedogenic carbonate nodules were analyzed. Then, 600–1,200 μg of pretreated enamel and 100–160 μg of untreated pedogenic carbonate material were reacted with 99% H3PO4 for 90 min at 70 °C in continuous flow mode using a Thermo Finnigan 253 mass spectrometer interfaced to a Thermo GasBench II. All analyses were performed at the Goethe University Frankfurt–Senckenberg Biodiversity and Climate Research Centre Stable Isotope Facility. Analytical procedures followed the technique of Spötl and Vennemann (44). Final isotopic ratios are reported versus VPDB (Vienna Pee Dee Belemnite; δ13C) and VSMOW (Vienna Standard Mean Ocean Water; δ18O); overall analytical uncertainties are better than 0.3‰.

Homogenized powder of 12 selected soil nodules measured for stable isotopes was additionally used for clumped isotope analyses, which were performed in the same laboratory. Untreated carbonate powder (5–15 mg) was digested in ≥106% H3PO4 at 90 °C for 15–30 min, using a semiautomated acid bath (45, 46). The produced CO2 was cleaned by flowing through cryogenic traps at −80 °C before and after passage through a Porapak Q-packed gas chromatography column to remove traces of water and hydrocarbons (cf. refs. 45, 46). The purified CO2 was analyzed using a Thermo Scientific MAT 253 gas source isotope ratio mass spectrometer dedicated to the determination of masses 44–49 in 10 acquisitions consisting of 10 cycles each, with an ion integration time of 20 s per cycle. Five to six replicates were run per carbonate sample. The best precision that can be achieved under these conditions is represented by the shot noise limit, which is 0.004‰ for n = 5 and 0.003‰ for n = 6. The Δ47 values are reported in the absolute reference frame (47) and were processed according to the protocol of Fiebig et al. (48). Apparent carbonate crystallization temperatures were computed from measured Δ47 values using the empirical calibration technique of Wacker et al. (46).

Soil temperatures (quoted accuracy of ±0.1 °C) were measured hourly using Voltcraft DL-101T USB-Temperature loggers, which were buried within polyethylene-plastic bottles ca. 40 cm below the surface in narrow trenches (∼15 cm wide), which were subsequently backfilled with the soil removed during excavation.

includes additional information on analytical procedures, data processing, dating, time interval recorded in enamel and pedogenic carbonate, isotope enrichment factors in mammals, stratigraphic information, and biome classifications used in this work.