Main

Grassy glades—anthropogenic soil nutrient hotspots—on recent herder settlements increase biodiversity at a landscape scale and influence savannah ecosystem structure and function4,12,13,14,15. Although the processes creating these glades are well-understood7,8,9,12,13,14,15, the full time-depth of their creation and effects on African savannahs are as yet unexplored. To investigate the longevity of anthropogenic soil nutrient hotspots, we excavated three Pastoral Neolithic sites located west of the Rift Valley in Ntuka, Narok County, Kenya: Indapi Dapo (site code GvJh121), Oloika 1 (GvJh85), Oloika 2 (GvJh86) and sampled two (GvJm44 and GvJm48) at Lukenya Hill, located to the east of the Rift Valley (Fig. 1, Extended Data Fig. 1, Extended Data Table 1).The sites are located in the Loita–Mara–Serengeti ecosystem and Athi-Kapiti plains. Accelerated mass spectrometry radiocarbon dates for the Ntuka sites range from 2,450 to 2,000 cal. bp and the radiocarbon dates from the Lukenya sites range from 3,700 to 1,550 cal. bp, spanning the earliest to the latest phases of the Pastoral Neolithic20 (Extended Data Table 2). Lithic and ceramic technologies20 indicate that the Oloika sites are members of the Elmenteitan tradition of the Pastoral Neolithic; Indapi Dapo and Lukenya sites belong to the Savannah Pastoral Neolithic tradition (Supplementary Information). The archaeological sites are 60–140 m in diameter (Fig. 1f, g, Extended Data Table 1). All of the sites in the Ntuka study area are located in structurally open grassy patches within wooded savannah grassland. Glades, Pastoral Neolithic sites and abandoned modern settlements are visible as well-defined hectare-scale treeless grassy features on the ground and in satellite imagery (Fig. 1e–g, Extended Data Fig. 2).

Fig. 1: Study areas and sampled sites.
figure 1

a, Distribution of Pastoral Neolithic sites in southwestern Kenya. MASL, metres above sealevel. Digital elevation data from NASA SRTM. b–d, Dung deposits visible in profiles at 1,500–3,000-year-old herder settlements at Lukenya Hill (GvJm48, b), Oloika 2 (c) and Indapi Dapo (d). e, False-colour LANDSAT-7 image of Narok county Ntuka study area on 3 February 2003, Oloika 1, Oloika 2 and Indapi Dapo site glades are visible as white patches, and modern Maasai settlements as dark red patches. f, g, High nutrient legacies encourage open grassy plant succession relative to bushy off-site vegetation at Oloika 1 and Oloika 2 (f), and Indapi Dapo (g). Globe in a, CC BY 2.0 licence. b–d, Photographs by F.M. f, g, Imagery from Google Earth Pro, Digital Globe.

Full size image

A visually distinct grey fine-grained sediment layer, 15–30-cm thick (Fig. 1b–d, Extended Data Fig. 1), occurs in four sites (Oloika 1 and 2, Indapi Dapo and GvJm48) and is discontinuous at the oldest site (GvJm44) (Supplementary Note). Micromorphology shows this grey sediment originates from degraded dung (Extended Data Fig. 2). Colour, texture and structure differ between on-site and off-site sediments (Fig. 1b–d). Phytoliths and dung spherulites9 are present in Ntuka on-site sediments and absent in off-site sediments (Extended Data Fig. 3).

Particle size analysis demonstrates that on-site sediments are dominated by silt, relative to coarser sandy off-site samples, and that organic matter and carbonate percentages are higher on-site (Extended Data Fig. 4, Extended Data Table 1). Fourier transform infrared spectroscopy shows that opal and calcite are present in on-site samples at Ntuka sites. Opal originates from silt-sized grass phytoliths. Calcite appears in thin sections as dung spherulites or as microspar (Extended Data Fig. 3). Lukenya Hill sediments do not show substantial mineralogical differences from off-site samples. However, inductively coupled plasma mass spectrometry analyses demonstrate that phosphorous, nitrogen, magnesium and calcium are enriched by an order of magnitude in on- versus off-site samples (Fig. 2, Supplementary Table 1). In some cases, calcium concentrations were elevated by 200–1,000% in on-site sediments. These findings are consistent with the enrichment observed in contemporary pastoral settlements (Supplementary Table 2). Nitrogen and carbon isotope ratios are consistently higher in degraded dung deposits than in natural off-site soils, except at Lukenya site GvJm 48 (Fig. 3, Supplementary Table 1). Supplementary Table 2 summarizes metadata from Africa regarding nutrient elevation, and the distinctive vegetation and ecology of historical and Iron Age herder corrals.

Fig. 2: Elemental nitrogen and phosphorous concentrations in sediment from on- and off-site stratigraphic sections.
figure 2

Elemental nitrogen (top) and phosphorous (bottom) concentrations in sediments from on-site (black circles, n = 16 for N and n = 12 for P) and off-site (grey triangles, n = 20 for N and n = 12 for P) stratigraphic sections at the sampled archaeological sites. All samples are independent. Least-squares regressions plotted as dashed lines for on-site samples and solid lines for off-site samples.

Full size image
Fig. 3: δ13C and δ15N values measured in on-site and off-site sediment samples from five archaeological sites in East Africa.
figure 3

On-site samples, closed symbols (n = 16); off-site samples, open symbols (n = 20). All samples are independent.

Full size image

Our analyses of micromorphology, mineralogy, and chemical and isotopic composition reveal that elevated levels of nutrients persist for 3,000 years in decomposed dung at Neolithic herder sites in the grasslands of southern Kenya. Our interpretations of the archaeological data are based on ethno-archaeological and ecological studies of contemporary pastoral settlements that show enrichment in weight percentage (wt%) N and 15N of soil organic matter, grass phytoliths, dung spherulites and mineral nutrients (especially phosphorous and calcium), relative to off-site samples8,18,21,22. Nitrogen in cattle and sheep and goat dung is enriched in 15N because dung is composed of a mixture of both excreted undigested plant material and 15N-enriched proteinaceous material from the animals themselves18,22. Volatilization of 15N-depleted ammonia from dung and urine decomposition in semi-arid environments also increases soil δ15N21,23. Soil organic δ13C values on archaeological sites are significantly higher than off-site values, and the highest δ13C values are associated with the highest δ15N values. This pattern is consistent with soil organic carbon and nitrogen being derived from dung and urine excreted by herbivores that graze on C4 plants. Phosphorous originates from the organic component of dung24,25. Calcium carbonate is concentrated in decomposed dung in the form of dung spherulites, which elevate calcium levels. Bones are additional sources of phosphate, magnesium and calcium. The presence of microspar suggests that dung spherulites dissolved and calcite re-precipitated as microspar. This is supported by the presence of manganese-oxide florets in thin sections, suggesting occasional hydromorphic conditions at some sites. The precipitation of microbially mediated carbonates22 and translocation of carbonate-rich solutions down profiles9 resulted in the formation of basal calcitic crusts.

A previous study of a 40-year time series of abandoned Maasai pastoral settlements in Kenya demonstrated that organic matter content declines substantially after about 20–30 years9. Mineralogical cascades triggered by the products of organic matter (for example, formation of phosphate minerals) stabilize at this point in diagenesis, sites become more deeply buried and minerals and elements may persist for millennia9,25. Continued enrichment of pastoral corral sediments by wild ungulates and domestic herds attracted to palatable forage, and soil–plant–herbivore feedbacks may contribute to persistence of anthropogenic hotspots7,11,15. Our results, as well as geochemical analyses of the Pastoral Neolithic site of Sugenya18, reveal the persistence of nutrient-enriched dung-derived deposits over three millennia.

These findings reinforce the environmental importance of the fertile grassy patches created by the earliest southern Kenyan pastoralists. Widespread settlements generated nutrient-enriched, hectare-scale microhabitats. Neolithic, Iron Age and recent herder sites are visible in satellite images of the Ntuka area as glades (4,400–15,000 m2 in size) (Fig. 1e–g, Extended Data Fig. 2d). Research on the Laikipia Plateau of Central Kenya complements Pastoral Neolithic findings, extending our understanding of nutrient stabilization and glade landscapes into the Pastoral Iron Age. The seventeenth–eighteenth-century-ad settlement of Maili Sita preserves phytoliths, spherulites and elevated nutrients in dung deposits that support characteristic grass species19. The fifteenth-century-ad Maasai Plains and unexcavated sites on the Laikipia Plateau reveal a broad distribution of 15–45-ha pastoral glades26. Pastoral Neolithic and Iron Age sites in diverse Kenyan savannahs demonstrate the spatial influences of niche construction by pastoralists on soil nutrients and savannah heterogeneity, on timescales that range from five centuries to three millennia.

Influences of the settlements of ancient and recent mobile herders in eastern Africa create landscape palimpsests. Recent herders make similar choices about where to locate their settlements to ancient ones with regard to slope and distance from water27,28. As a result, contemporary Maasai herders settle on or near Pastoral Neolithic settlements in the Lemek Valley29 and Ntuka areas of southwest Kenya (Fig. 1g, Extended Data Fig. 5). Such settlement clusters are also attractive to modern herders because of the proximity of nutrient-rich grazing that supports growth and lactation (for example, for calves)3,7. Studies of Maasai settlements constructed over the last 60–100 years suggest that 1–20% of savannahs, and most land near water, has been settled over the past 100 years12,27.

East African findings draw attention to the temporal and spatial scale of pastoral legacies. Metadata on modern and ancient African pastoral settlements indicate that influences on nutrient enrichment and ecology are broadly dispersed (Supplementary Table 2). Ecological research on a South African Iron Age site in the Nylsvley, Nature Reserve Limpopo Province (about 700 years old), compares modern glade formation processes in South Africa, the Sahel and eastern Africa (marked on Fig. 1a) and documents increased biodiversity and forage quality on this ancient anthropogenic hotspot4. In the Limpopo Valley and eastern Botswana, corrals in Iron Age settlements (about 1,200–500 years old) (Supplementary Table 2) are characterized by grassy vegetation and eutrophic dung-derived soils16,17. Landscape-scale on- and off-site studies have the potential to resolve ancient patterns at finer scales. However, biogeochemical data on Neolithic and other East African sites, and ecological and nutrient data from widespread Iron Age and historic sites, indicate that herders have had a role in structuring and diversifying African savannah ecosystems for up to three millennia.

Outside Africa, reinforced nutrient enrichment related to Iron Age pastoral activity in arid environments has also been documented in the southern Levant30. Exploration of nutrient enrichment by ancient pastoral settlements in temperate and high-altitude grasslands in Central Asia31 and South America32 will yield insights into local and regional variability and the global importance of prehistoric herder influences on nutrient flows and grassland ecology.

Methods

No statistical methods were used to predetermine sample sizes.

Soil profiles and samples from excavations

Sites were identified in the Ntuka region during site surveys by S.H.A. before 2011. Surface exposures and animal burrows in Pastoral Neolithic sites were examined in 2011 for potential corral areas. Excavation proceeded by 5-cm spits or natural levels, where detected. All finds were screened through 5-mm mesh. Samples were taken within excavations for micromorphology, flotation and sediment analyses. Off-site samples were taken 30–40 m from the edges of archaeological sites, in bush-dominated microhabitats.

Sediment analyses

Elemental analysis of bulk sediment samples used an Agilent 7750 ICP-MS. Dried samples (~0.1 g) were digested in concentrated HNO3 in a Mars-6 microwave digester at 180 °C for one hour. After digestion, the supernatant was diluted to a 10× solution, filtered through a 0.22-μm filter and diluted again to make a final 100× solution. Internal reference standards and HNO3 blanks were used to calibrate the ICP-MS. Potential memory effects were monitored by running blanks after every 5 samples and drift monitored by running internal standards after every 15 samples. Particle size, loss on ignition, magnetic susceptibility and chemical composition analyses were conducted at the Geoarchaeology and Nano facilities at Washington University in St. Louis. Micromorphology and FTIR analyses were performed at the Laboratory for Sedimentary Archaeology, University of Haifa (see Supplementary Information).

Stable isotopes

Analyses on bulk sediment samples were carried out at Washington University in St Louis, using a Flash 2000 elemental analyser coupled to a Thermo Delta V Plus continuous-flow isotope ratio mass spectrometer. Samples were homogenized in an agate mortar, and pestle-and-sieved to a particle size of <250 μm. We aimed to analyse at least 20 μg of nitrogen, so 15–80 mg of sediment was weighed into 5 × 9-mm tin capsules. Samples for carbon isotope analysis were treated to remove carbonates with 2 M HCl until effervescence ceased (~24 h), rinsed 5 times with MQ water, dried in a 70 °C oven, and weighed into 5 × 9-mm tin capsules. Our results are expressed as δ15N and δ13C as parts per thousand (‰) relative to AIR and Vienna Pee Dee Belemnite standards, respectively. The average analytical precision for both C and N was <0.2 ‰ based on the standard deviation of 24 replicates of an in-house standard (Bob’s Red Mill millet flour) and 18 replicates of a second in-house standard (acetanilide). Weight percentage C and N are estimated based on standards of known elemental composition; wt% precision of these known compounds is better than 0.1%. We used the lme4 package in R to perform a linear mixed effects analysis of the relationship between sediment isotope values and the on-site presence of a dung profile (see Supplementary Information).

Radiocarbon dating

Faunal collagen and enamel apatite samples from Indapi Dapo and Oloika 2 (Extended Data Table 2) were prepared for radiocarbon dating at the Environmental Isotope Paleobiogeochemistry Laboratory, Department of Anthropology, University of Illinois, Urbana, and the Radiocarbon Laboratory of the Illinois State Geological Survey at the University of Illinois. To purify collagen, dentine was demineralized using 0.2 M HCl, rinsed 8× with dH2O, treated with 0.125 M NaOH to removed soil organic contaminants, rinsed 8× with dH2O and hydrolysed at 70 °C in 10–3 M HCl. Freeze-dried collagen was converted to CO2 using sealed-tube combustion, and cryogenically distilled for AMS dating at the UC Irvine Accelerator Mass Spectrometry radiocarbon laboratory. Enamel was separated from dentine and cementum and ground in an agate mortar. A ~400-mg sample was treated with 25 ml 2.63% NaOCl (sodium hypochlorite) to remove organic matter, rinsed 8× with dH2O, and reacted with 25 ml 0.1 M acetic acid under vacuum to remove adsorbed and diagenetic carbonate, alternating with return to atmospheric pressure with CO2-free N2. Cycling between vacuum and N2 continued at ~15–30-min intervals until the bubbling reaction ceased (~3–4 h). Samples were rinsed 5× in distilled water and freeze dried. Purified samples were reacted with 100% H3PO4 and CO2 from structural carbonate was purified by cryogenic distillation for AMS dating.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability

All data generated or analysed in this study are included in the paper and its Supplementary Information. Site descriptions are in Supplementary Information, radiocarbon dates in Extended Data Table 2 and particle size, ICPMS, FTIR, isotope and micromorphology data in Supplementary Table 1. Remaining soil samples are curated in the Archaeology Division of the National Museums of Kenya, Nairobi.