Environments and trypanosomiasis risks for early herders in the later Holocene of the Lake Victoria basin, Kenya

Significance

Herding was the earliest form of African food production and transformed local populations of people and animals. Herders migrated from eastern to southern Africa around 2,000 years ago, but only in small numbers. Zoonotic disease vectors, specifically the tsetse fly, which carries sleeping sickness, are thought to have impeded these movements. Archaeologists have argued that the presence of tsetse flies around Lake Victoria, Kenya, created a barrier that prevented migration and forced subsistence diversification. This study, using stable isotope analysis of animal teeth, reveals the existence of ancient grassy environments east of Lake Victoria, rather than tsetse-rich bushy environments. This overturns previous assumptions about environmental constraints on livestock management in a key area for southward movement of early herders.

Abstract

Specialized pastoralism developed ∼3 kya among Pastoral Neolithic Elmenteitan herders in eastern Africa. During this time, a mosaic of hunters and herders using diverse economic strategies flourished in southern Kenya. It has been argued that the risk for trypanosomiasis (sleeping sickness), carried by tsetse flies in bushy environments, had a significant influence on pastoral diversification and migration out of eastern Africa toward southern Africa ∼2 kya. Elmenteitan levels at Gogo Falls (ca. 1.9–1.6 kya) preserve a unique faunal record, including wild mammalian herbivores, domestic cattle and caprines, fish, and birds. It has been suggested that a bushy/woodland habitat that harbored tsetse fly constrained production of domestic herds and resulted in subsistence diversification. Stable isotope analysis of herbivore tooth enamel (n = 86) from this site reveals, instead, extensive C₄ grazing by both domesticates and the majority of wild herbivores. Integrated with other ecological proxies (pollen and leaf wax biomarkers), these data imply an abundance of C₄ grasses in the Lake Victoria basin at this time, and thus little risk for tsetse-related barriers to specialized pastoralism. These data provide empirical evidence for the existence of a grassy corridor through which small groups of herders could have passed to reach southern Africa.

Herding was the earliest form of food production in Africa, originating in the arid eastern Sahara ∼8 kya. Adoption of food production and increasing mobility in northeastern Africa allowed prehistoric people to manage resource availability amid a drastically changing climate (1) and transformed local populations of people and animals. Saharan herders and hunters spread southward to the Sahel, reaching eastern Africa by 4.5 kya, and eventually reaching southern Africa with sheep and cattle around 2 kya (1–8). The southern African data have been much debated, but recent genetic studies support at least limited movement of early herders from eastern to southern Africa (5, 7). Iron Age Bantu agriculturalist migrations and forager exchange processes also contributed to livestock spread. In contrast to Iron Age farmers, however, the sparse archaeological evidence suggests slow and small-scale southern pastoral migrations (9). Researchers have attributed the limited penetration of southern Africa by stone-using pastoralists to the prevalence of woodland habitats, the distribution of tsetse fly, and the influence of trypanosomiasis on livestock production from Lake Victoria and the Serengeti southward (10, 11). However, site-based paleoenvironmental data have proven difficult to obtain, and although often assumed, the proposition that ancient closed or bushy environments represented a barrier to herding has been seldom examined. Here we present stable isotope analysis of herbivore tooth enamel (n = 86) from Elmenteitan Pastoral Neolithic levels at Gogo Falls (ca. 1.9–1.6 kya) near Lake Victoria, one of the northernmost “woodland tsetse belts” modeled for Africa ca. 2,000 y ago (12, 13). Rather than providing evidence for the Lake Victoria basin as a wooded tsetse fly harboring habitat, these new fine-resolution paleoenvironmental data document extensive C₄ grazing by both domesticates and the majority of wild herbivores. A synthesis of lacustrine and terrestrial signals further supports the existence of grassy areas in the Lake Victoria basin at 2,000 B.P., indicating a change in ecology to bushy environments more recently than previously thought, and an ecosystem that would not have supported the tsetse fly around Lake Victoria.

Early herders in northern Kenya relied on sheep, goats, fish, and diverse wild vertebrates (1, 14). Research in western Kenya, and specifically the Lake Victoria basin (Fig. 1), has revealed southwestward movement of Elmenteitan herders into an area populated by complex ceramic-using, fishing hunter-gatherers and a more varied process of adoption of food production. Domestic sheep and goats appear in low numbers starting around 3.7 kya at Wadh Lang’o, and perhaps also at Gogo Falls and Usenge 3 (15), but not at other sites such as Siror (16), suggesting patchy adoption of herding. Specialized dependence on livestock in Africa is first documented in Elmenteitan sites, which date at the earliest to 3.1 kya at Njoro River Cave, east of Lake Victoria (17, 18). Elmenteitan sites occur to the north in Laikipia, on the Mau Escarpment, on the southern end of the Loita Hills and the northern Mara Plains, and at Gogo Falls (the westernmost extent) (18). Despite opportunities for hunting, fauna from Elmenteitan sites on the Mara plains are >90% domestic (19, 20). The exception to this is the site of Gogo Falls, which has a diverse and abundant faunal assemblage, including fish and wild herbivores (21). The unique faunal assemblages from Elmenteitan levels at Gogo Falls and from Prolonged Drift (22), which has a mixed Savanna Pastoral Neolithic stone tool assemblage, have sparked considerable debate over the significance of subsistence variability for understanding environmental risks and the dynamics of frontier pastoral-hunter-gather subsistence during the spread of food production. Foragers in the process of adopting herding, loss of stock by pastoralists, and environmental and disease constraints have all been considered possible factors leading to heavy reliance on wild resources (21–24). South of the Mara plains and Lake Victoria, ancient herders left fewer material traces, leading to arguments regarding the existence of a long-term pastoral-forager frontier, south of which herders did not thrive (25–27).

Fig. 1.

Location of Gogo Falls in relation to other Holocene archaeological sites (●) and towns (○) in the Lake Victoria basin.

The presence of tsetse flies in eastern Africa has been used as an explanation for observed patterns of subsistence diversity among early pastoralists in the Lake Victoria basin (21, 28) and for limited southward movement (15, 20, 29, 30). A range of zoonoses posed threats for early herders, but trypanosomiasis risks have an especially widespread effect on human communities, even today (10, 12, 13, 30). Mesic, bushy/wooded environments that may have harbored large numbers of tsetse fly (Glossina spp.) would have been poorly suited to extensive cattle stock-keeping. The tsetse fly carries trypanosome parasites that transfer to hosts during blood meals and eventually cause often-fatal sleeping sickness in ungulates (predominately domesticates) and humans. In modern and precolonial times, herders have controlled tsetse fly numbers by heavy grazing, burning and destruction of woody areas, and managed livestock to avoid areas with abundant tsetse (10, 31, 32). It is not unreasonable to assume, then, that bushy, tsetse-rich environments would have impeded heavy reliance on livestock by early herders and may have even created a boundary beyond which it was difficult for large numbers of herders to settle (20, 33). Despite an abundance of archaeological sites, however, terrestrial paleoecological data for the Holocene in Kenya are scarce and have been based primarily on lacustrine archives, rather than sites themselves. Pollen and leaf wax biomarkers from lacustrine records from Lake Victoria yield somewhat conflicting paleoenvironmental interpretations: pollen data indicate decreasing moist tree and shrub presence (indicated by Moraceae pollen) in the Lake Victoria basin until ∼2 kya, when grass pollen (Poaceae) increased (Fig. 2) (34), contrasting with stable carbon isotopes from lacustrine leaf wax biomarkers, which indicate persistence of C₄ grasses throughout the last 6 kya, with only a sharp decrease in C₄ around 3 kya (Fig. 2) (35). These contrasting results could be explained by differences in spatial integration, with lipids recording a more localized signal within the catchment and pollen recording a more regional vegetation signal (36), and likely also point to varied, fluctuating input from C₄ grasses during this period. Even so, pollen evidence suggests only minor (∼10%) moist forest in the region compared with ∼20–30% input from grasses. These records offer the important wider ecological context of regional (pollen) and basinal (leaf waxes) ecology, yet they are too coarse to understand specialized pastoralist expansions, and especially for local evidence (site-based reconstruction) of habitats harboring tsetse.

Fig. 2.

(A) Holocene pottery traditions present at archaeological sites in the Victoria basin. (B) Leaf wax δ¹³C and modeled %C₄ from lacustrine sedimentary cores in Lake Victoria (35). (C) Moraceae (tropical mesic trees and shrubs) pollen counts from lacustrine sedimentary cores in Lake Victoria (data from ref. 34). (D) Poaceae (grass) pollen counts from lacustrine sedimentary cores in Lake Victoria (data from ref. 34). B, C, and D modified with permission from Elsevier; www.sciencedirect.com/science/journal/02773791.

To determine whether or not shifting tsetse-rich environments were present at particular locales during specific periods, direct stable isotope-based paleoecological analysis at archaeological sites can provide a complement to faunal analysis (21, 37–39). Isotopic measurements of enamel carbon (δ¹³C) and oxygen (δ¹⁸O) can be used to understand diet, habitat, and climate (40, 41). Although bone and tooth collagen have been previously used to assess paleoenvironment at Holocene archaeological sites in Kenya (42), tooth enamel is preferred, as it resists digenetic alteration and can be easily compared with fauna from older periods of geologic time (43). In some instances, faunal tooth enamel isotopes may provide our best estimates of environmental change over time, because these signals are less temporally and spatially attenuated than other proxies, such as leaf wax biomarkers (44), which may be influenced by reservoir effects during catchment transport before deposition (45, 46). The environmental context of the Elmenteitan layers of Gogo Falls has been previously interpreted as a grassland/bushland mosaic, based on analysis of wild species diversity and the presence of modern browsing taxa such as roan or sable antelope (Tragelaphini) and bushpig (Potamochoerus), as well as grazers such as oribi (21). The relatively low proportions of domestic stock were interpreted as reflecting environmental constraints on pastoral productivity in the region. The presence of the tsetse fly in a presumed seasonally mesic, bushy/woody environment, it was argued, prevented herders from relying exclusively on domesticates, which resulted in seasonal fishing and hunting of large wild ungulates (21). These interpretations have not yet been ground-truthed with a local, terrestrial paleoecological indicator.

Results

Isotopic data are presented in in Fig. 3 and SI Appendix, Tables S1 and S3. For comparison, a compilation of modern comparable herbivore δ¹³C₁₇₅₀ values, primarily from southern Kenya and northern Tanzania, are presented in Fig. 4 (SI Appendix, Table S2) (47–53). The average δ¹³C values indicate a diet with >80% C₄ grass for 10 of 13 taxa from Gogo Falls. Potamochoerus sp. (δ¹³C = −3.86‰; n = 1), Sylvicapra sp. (δ¹³C = −8.81‰; n = 1), and Tragelaphini (δ¹³C = −6.6‰; n = 5) are the most ¹³C-depleted herbivores in the assemblage, but they are more ¹³C-enriched than their modern counterparts (Fig. 4; SI Appendix, Table S2). Modern hippos average −2.3 ± 1.7‰, whereas hippos at Gogo Falls average 0.6 ± 0.6‰. The presence of both large migratory and small, local bovids provides further evidence for both locally and basinally grassy environments (21).

Fig. 3.

Box plots of δ¹³C tooth enamel values from archaeological tooth enamel, Trench III, Gogo Falls.

Fig. 4.

Box plot of δ¹³C₁₇₅₀ tooth enamel values from modern comparative fauna from Kenya and Tanzania. *Neotragini and Sylvicapra sp. values were calculated to tooth enamel values from keratin, using ε_{keratin-enamel} values of 11.1‰ because of a lack of enamel values in the literature.

Domestic fauna are pure C₄ feeders at Gogo Falls. Bos taurus is within the range of pure C₄ diet at both Gogo Falls and in the modern environment. Tribe Caprini (Subfamily Caprinae), which includes both goats and sheep, is again overwhelmingly within the range of a pure C₄ diet (1.6 ± 1.4‰). Modern goat δ¹³C₁₇₅₀ values from locations across Kenya range from −2.6‰ to −12.3‰, whereas modern sheep range from −0.3‰ to −1.5‰ (SI Appendix, Fig. S1 and Table S4) because of the different dietary preferences of sheep and goats. Sheep are primarily grazers, whereas goats are mixed feeders that tend to browse, as their highly efficient digestive physiology confers an advantage for consuming low-quality woody forage (50, 54–57). Criteria for distinguishing domestic sheep and goat teeth have not been considered reliable in Africa (58), and Gogo Falls samples, at the time of identification, were simply identified as sheep or goat. Future studies may distinguish teeth and postcranial bones in African samples (59). If both sheep and goats were present at Gogo Falls, then these isotopic results indicate a much higher consumption of C₄ resources by modern goats than is typically observed (50, 56, 60). If the isotope data represent values exclusively from sheep, which is more likely, this level of C₄ intake is not unusual, yet the presence of sheep only at this site is unusual compared with current herding practices in East Africa.

Discussion

In comparison with the present day, isotopic analyses of herbivores indicate that during the Elmenteitan occupation at Gogo Falls, there were few C₃ browsers or mixed C₃–C₄ feeders, suggesting a landscape dominated by C₄ grasses. This interpretation is in general agreement with δ¹³C_wax values (Fig. 2) from lacustrine cores in Lake Victoria (which likely indicate vegetation locally, within the lake catchment), and to a lesser extent with pollen records, which indicate mixed environments regionally. These three lines of evidence, taken together, strongly imply an abundance of C₄ grasses in the basin during this period. Vegetation in and around the Winam Gulf today has been characterized as “evergreen and semi-evergreen bushland and thicket,” “edaphic wooded grasslands and grasslands on drainage impeded or seasonally flooded soils,” and “moist combretum wooded grassland” (which comprises the ecology area around the Kuja River, where the Gogo Falls is located) (61). Faunally based interpretations, primarily the presence of grazers and modern browsers such as Tragelaphini and Potamochoerus sp., of the ancient environment at Gogo Falls suggested a woodland/grassland/bushland mosaic that is climatologically and ecologically similar to today (21), which would have supported tsetse flies. This bushland interpretation is not supported by the isotopic data, which are dominated by C₄-grazing taxa with few C₃-browsing taxa, contradicting long-held assumptions of modern climate and ecology persisting over the last 3 millennia in most regions of eastern Africa (19, 21, 24).

Given the resolution of data available, it is not yet possible to quantify the relative proportions of grassy and woody vegetation; however, abundant C₄ grasses must have been present. Abundant grassy vegetation may have been the result of changes in rainfall seasonality in the later Holocene (62, 63), changes in total rainfall in the Lake Victoria basin, increased burning, or ecological factors, such as heavy grazing by domesticates or wild herbivores. Controlled grazing by cattle, sheep, and goats can stimulate new grass growth where grazing herbivores thrive (64, 65). Overgrazing, however, makes environments more bushy and reduces grass height (66). Burning promotes new grass and may have maintained grasslands if herders were regularly burning savanna for livestock, as is a common practice in modern eastern Africa (67–69). The presence of megaherbivores maintains grasslands as well, and when both burning and elephant or giraffe populations are high, grasslands expand (67, 68, 70). Given such complex environmental interactions in African grasslands today, the question arises whether ancient herders in the Lake Victoria basin helped create grasslands through grazing and fire, or whether they were created and maintained by other ecological and climatological factors (71, 72).

The isotopic data from Gogo Falls and leaf wax data from the Lake Victoria basin reveal an ideal environment for large grazing herds of domesticates in the immediate Lake Victoria basin catchment, one rich in C₄ grasses. Such grasslands were not an ideal environment for tsetse flies, which calls into question the maintenance of long-term areas of hunter-gatherer-pastoral interaction and the inhibition of southward movements of herders by stable and extensive bands of tsetse bush (2). This study fits with the findings of recent genetic research including Y-chromosome, lactase persistence, and livestock data that point to some migration of herders from eastern to southern Africa (5–7). Archaeologists arguing for successful early pastoral migrations from eastern to southern Africa have pointed to some similarities between Elmenteitan pottery and early herder ceramics in southern Africa (73). Our data imply that a tsetse-rich barrier preventing herders from moving through the Lake Victoria basin into northern Tanzania and other parts of sub-Saharan Africa (9, 24, 33, 74) was not widespread 2,000 y ago, bolstering the possibility of such connections.

On a local level, our data show that Elmenteitan herders at Gogo Falls were not forced into hunting and fishing because of ecological constraints on stock-keeping. The Elmenteitan mammalian fauna at Gogo Falls are made up of 53.5% domestic livestock comprising cattle and sheep (based on isotopes). Fishing was also a significant activity (21). Recent research in the Lake Victoria basin has shown that the nearby site of Wadh Lang’o shares similar dates of 1,950 ± 35 y B.P. for early Elmenteitan levels (27, 75) and a number of similarities with Gogo Falls: Wadh Lang’o is also a large open-air site with a long sequence containing Kansyore, Elmenteitan, and Iron Age horizons (27, 75). It is situated in a similar environmental setting, on the banks of a river flowing into Lake Victoria. There are as yet no direct paleoecological proxies from this site, but according to lacustrine paleoenvironmental archives (Fig. 2), it was also likely situated in a grassy environment. Fishing is a significant activity at both sites; however, the mammals in the Elmenteitan faunal horizons in Trench 1, Wadh Lang’o (below datum 140–190 cm), are predominantly (89–91%) domestic, suggesting limited hunting (75). Alcelaphines are represented, but the zebra and oribi present at Gogo Falls are absent at Wadh Lang’o. Herding strategies also differed between Wadh Lang’o and Gogo Falls, with cattle making up 1% or less of the domestic fauna at Wadh Lang’o (n = 88 below datum 140–190 cm; n = 156 below datum 140–160 cm; SI Appendix, Table S5) compared with 32% (n = 328) at Gogo Falls. In strong contrast with these sites in the Lake Victoria basin, there is no fishing and very little hunting in Elmenteitan sites on the plains of southern Kenya (19, 20). Cattle, sheep, and goats make up 97–100% of the fauna at the open air settlement sites in the Lemek-Mara, including Ngamuriak, Sugenya, and Oldorotua, as well as the Central Rift Valley rock shelter of Maasai Gorge (SI Appendix, Table S5). Reliance on cattle and some small stock, combined with fishing and significant hunting at Gogo Falls, is an unexpected contrast with nearby Elmenteitan sites and does not fit a typical Elmenteitan subsistence pattern. Nor do the fauna, material culture, and organization of this site suggest hunter-gatherers in the process of adopting herding. Gogo Falls appears to have been populated by a unique group of people, potentially a small group of herders, who took flexible approaches to subsistence opportunities and interactions with hunter-gatherer-fishers as they followed grassy corridors in eastern Africa.

Conclusions

Our results provide much-needed late Holocene paleoenvironmental data for the Lake Victoria basin and demonstrate that substantial ecological changes occurred over the last 2,000 y. Isotopic data reveal that the fauna were dominated by C₄ grazers with very few C₃ browsing taxa, suggesting a grassy paleoecological setting for the unique Elmenteitan site of Gogo Falls. This challenges interpretations based on macrofaunal remains and models of disease-depressed livestock production (21). The isotopic findings also draw attention to herder relations with hunter-gatherers, rather than ecology, in consumption of large wild ungulates in western Kenya. This suggests that social factors may have played a greater role than previously thought in subsistence diversity during the spread of pastoralism in Eastern Africa. High levels of hunting at Gogo Falls can no longer be attributed to closed or bushy habitats that inhibited livestock production and southward movement of herders (21) and contributed to long-term frontier interactions between pastoralists and hunter-gatherers (25, 27, 74). These data raise interesting questions about the interplay among herders, environmental management and adaptation, and ecological drivers for grassland maintenance. Finally, in light of the isotope data from Gogo Falls and lacustrine leaf wax data (35), the Lake Victoria basin may have been the setting for grassy corridors through which demic migration of pastoralists occurred out of eastern Africa and toward southern Africa (30).

Materials and Methods

Site Description.

The archaeological site of Gogo Falls is located on the banks of the Kuja River by a dam that bears the same name (34°21′E, 0°34′S; Fig. 1). The area receives ∼1,060 mm/y in rainfall, with bimodal rainfall seasonality, with rains occurring in March–May and October–November (28). At this time, the area surrounding the site has been ecologically altered by farming, but natural vegetation is characterized as a wooded grassland (61). Nearby local wild fauna include roan antelope, zebra, and giraffe, and historically, a diverse mammalian population including elephant, lion, cheetah, and rhinoceros (61). Browsers are rare, and thus woodland and thicket encroachment has been occurring in the area (76).

Gogo Falls was excavated by Robertshaw in 1980 and 1983 (28, 77), and teeth analyzed for this article were uncovered during the 1983 excavation. Five noncontinuous trenches were dug in the study area, each yielding abundant faunal remains, artifacts, and potsherds spanning three cultural traditions: Kansyore, Elmenteitan, and Urewe (28). The stratigraphic details of Trench III is presented by Robertshaw (28), but the highlight of this excavation is a 1.5-m ash and dung midden layer containing Elmenteitan pottery exclusively, which Robertshaw interpreted as a stock-keeping area. Radiocarbon dating of charcoal near the top of Trench III gave an uncalibrated age of 1,770 ± 80 y B.P., and another date from the bottom of the trench gave an age of 1,990 ± 80 y B.P. (28), with no evidence of stratigraphic mixing. These dates calibrate to 1,646 calibrated y B.P. (calBP) ± 96 and 1,900 calBP ± 100 at 95.4% confidence using ShCal13 (78).

Teeth were selected from faunal analysis by Marshall and Stewart (21). All the faunal material excavated from the 1.5-m ash midden is in excellent preservation. The fauna from this trench are diverse, containing fish, avian, and mammalian remains. Domestic caprines (goat and sheep) constitute the largest proportion of the assemblage (36.4%), followed by cattle (17.1%), topi/hartebeest (presented here at the tribal level of identification, “Alcelaphini”; 15.2%), oribi (presented as “Neotragini”; 11.8%), zebra (11.3%), reedbuck (“Reduncini”; 2.8%), warthog (1.8%), and eland and roan/sable (“T. oryx” and “Hippotragini”, respectively; 1.1%) (21). Stable isotope analysis was carried out on a subset of these teeth, including samples from rarer fauna as well (i.e., Potamochoerus sp. and Sylvicapra sp.).

Laboratory Analysis.

Following the sampling procedures outlined by the National Museums of Kenya, teeth were photographed and sampled along broken edges, using a Dremel tool and diamond drill bit. About 1 mg enamel powder was drilled from each sample. Enamel powder was prepared and analyzed at the University of Utah. Enamel powders were treated with 0.1 M buffered acetic acid (pH ∼5.3) for 30 min to remove labile carbonates (79, 80). After acid treatment, powders were rinsed three times with distilled water and dried at 60 °C overnight. Sample powders were weighed into silver capsules and dried under vacuum at 200 °C for 2 h before analysis. Samples were analyzed for δ¹³C and δ¹⁸O on a Finnigan MAT 252 coupled to a Carboflo dual-inlet carbonate device (common phosphoric acid bath at 90 °C for 15 min reaction time). Stable isotope ratios are reported as delta (δ) values relative to the international carbon isotope standard, Vienna Pee Dee Belemnite, using the standard permil (‰) notation, where δ¹³C = (R_sample/R_standard − 1) × 1,000 and R_sample and R_standard are the ¹³C/¹²C ratios of the sample and standard, respectively. SD of an internal carbonate standard (Carrara marble) was ±0.1‰.

Dietary designations for herbivores are based on estimated consumption of C₄ plants (tropical grasses) and C₃ plants (trees, shrubs, herbs), which is calculated using a hypothetical 100% C₄ diet and an isotope enrichment (ε*) between diet and herbivore tooth enamel of 14.1‰ (81). Estimated carbon isotope composition of tooth enamel for a 100% C₄ diet is based on the average isotopic value of modern C₄ plants in eastern and central Africa (−12.9‰; n = 764) (82). Changes in the isotopic composition of atmospheric CO₂ resulting from the combustion of fossil fuels between modern environments and preindustrial Holocene (presented here as δ ¹³C₁₇₅₀) were accounted for by adding 1.6‰ to the modern average δ¹³C of C₄ plants (81, 83, 84).