Intensifying Weathering and Land Use in Iron Age Central Africa

See allHide authors and affiliations

Science  09 Mar 2012:
Vol. 335, Issue 6073, pp. 1219-1222
DOI: 10.1126/science.1215400

A Price of Civilization

Large expanses of rainforests in parts of Central Africa were abruptly replaced by savannas around 3000 years ago, presumably because of climate change. However, that succession occurred at a time of expansion by Bantu tribes, from near the border of present-day Cameroon and Nigeria to the south and east, in a migration that brought with it agriculture and iron-smelting technologies. Bayon et al. (p. 1219, published online 9 February; see the Perspective by Dupont) analyzed the nearby marine sedimentary record and found that chemical weathering in Central Africa also increased markedly at this time. This increase in weathering could have been caused by forest clearing by the Bantu to create arable land and to fuel their smelters, rather than climate change alone.


About 3000 years ago, a major vegetation change occurred in Central Africa, when rainforest trees were abruptly replaced by savannas. Up to this point, the consensus of the scientific community has been that the forest disturbance was caused by climate change. We show here that chemical weathering in Central Africa, reconstructed from geochemical analyses of a marine sediment core, intensified abruptly at the same period, departing substantially from the long-term weathering fluctuations related to the Late Quaternary climate. Evidence that this weathering event was also contemporaneous with the migration of Bantu-speaking farmers across Central Africa suggests that human land-use intensification at that time had already made a major impact on the rainforest.

A major vegetation change occurred in Central Africa during the third millennium before the present (B.P.), when mature evergreen trees were abruptly replaced by savannas and secondary grasslands (14). The consensus is that the forest disturbance was caused by a regional climate change (14). However, this episode of forest clearance occurred contemporaneously with the migration of Bantu-speaking peoples from near the modern-day Nigeria-Cameroon border (59). The so-called Bantu expansion led to the diffusion of agriculture and iron-smelting technology across Central Africa, with potential impacts on the environment (10). Whether the Bantu farmers played an active role in the Central African deforestation event remains an open question.

To provide further constraints on this issue, we used a marine sediment record recovered off the mouth of the Congo River to reconstruct the late Quaternary history of chemical weathering in Central Africa (Fig. 1). This core (KZAI-01; 05°42′S, 11°14′E), collected at a water depth of 914 m, provides a continuous record of the Congo River sediment discharge for roughly the past 40,000 years [see supporting online material (SOM)]. Although changes in chemical weathering intensity on continents are driven primarily by natural factors—such as physical weathering rates, vegetation, rainfall, and temperature (11, 12)—intensive land use and accelerated soil denudation by increasing the surface area of minerals and rocks exposed to weathering can also dramatically lead to much higher rates of chemical alteration (13). The degree of chemical weathering of fine-grained sediments can be inferred from the ratio of aluminum to potassium (Al/K). Potassium is highly mobile during chemical weathering and typically depleted in soils, whereas aluminum is one of the most immobile elements, being incorporated into secondary clay minerals such as kaolinite (see SOM text). High Al/K ratios in Congo fan sediments are therefore considered to be indicative of periods of intense chemical weathering in the Congo basin (14). Because downcore variations of the bulk chemical composition can also reflect changes in the sediment source, we measured neodymium (Nd) and hafnium (Hf) isotopic ratios to discriminate between both weathering and provenance signals in our sediment record. The Nd isotopic signature of terrigenous sediments is retained during continental weathering and subsequent transport, thereby providing direct information on the geographical provenance of sediment (15). Hafnium isotopes exhibit globally similar behavior but are also prone to substantial fractionation during chemical weathering, because incongruent dissolution of silicate rocks leads to products of erosion having very distinctive but systematic Hf isotopic signatures (see SOM text) (16, 17).

Fig. 1

African satellite map showing the location of the studied core (KZAI-01). A major vegetation change was reported at a number of sites in Central Africa between ~3000 and 2000 years ago (yellow stars) (fig. S1), with evidence for a substantial loss of primary forest and expansion of savannas and other pioneer formations. This deforestation event was contemporaneous with the migration of Bantu-speaking agriculturalists originating from the Nigeria-Cameroon area. During the third millennium B.P., Bantu farmers spread both southward (across Atlantic equatorial Africa) and eastward (through the Congo watershed), reaching Angola and the Great Lakes region by ~2500 years B.P., respectively (thick orange arrows). Thin yellow arrows represent subsequent migration waves toward southern Africa. CAR, Central African Republic; DRC, Democratic Republic of the Congo.

In this study, we quantitatively determined the bulk sedimentary major element composition of KZAI-01 at a 5-cm sampling interval (Fig. 2), corresponding to a temporal resolution of ~100 to 400 years. The age model for KZAI-01 is based on accelerator mass spectrometry (AMS) radiocarbon measurements of mixed marine carbonate fractions and tuning to a well-dated nearby sediment record from the Gulf of Guinea (GeoB6518-1; see location in Fig. 1). We also obtained additional age constraints from 14C-AMS dating of bulk sediment organic carbon (Fig. 2). Figure 3A shows that the Nd isotopic composition of sediments deposited at site KZAI-01 (average εNd ~ –15.9 ± 0.6) is almost constant and very similar to that reported for present-day riverine particulates from the Congo basin (18). This indicates that the source of material delivered to the ocean by the Congo River has remained unchanged during the Late Quaternary. In contrast, Hf isotopes display significant downcore variations (from εHf ~ –6.8 to –13.9), which correlate well with the Al/K depth profile (Fig. 3, A and B). Because grain size is homogeneous in this core, with medians ranging from 4 to 6 μm (17), the large range of εHf values cannot be explained by changes in the relative proportions of mineral phases with distinct Hf isotope signatures. Therefore, these data show that downcore fluctuations of both εHf and Al/K ratios at site KZAI-01 reflect variations in chemical weathering intensity within the Congo River drainage basin, rather than changes in sediment provenance and/or grain size.

Fig. 2

CaO concentrations (weight percent) and Al/K ratios versus core depth (m) in core KZAI-01. The triangles indicate the position of AMS 14C dates for mixed marine carbonate fraction (black) and bulk organic matter (gray), as well as the age control points tuned to the well-dated nearby core GeoB6518-1 (white). The upper-left inset shows the depth versus calendar age plot for KZAI-01. LGM, Last Glacial Maximum; YD, Younger Dryas; kyr, thousand years.

Fig. 3

Proxy records for source provenance and chemical weathering intensity in core KZAI-01 and comparison with paleoclimatic and paleovegetation records. (A) Neodymium and hafnium isotopic composition in core KZAI-01 (expressed as εNd and εHf, respectively), as proxies for sediment provenance and chemical weathering intensity. (B) The Al/K record from core KZAI-01 indicating variations of chemical weathering intensity in Central Africa. (C) Plant-wax δD values (per mil) in core GeoB6518-1, as an index of precipitation changes in Central Africa (19). (D) The annual mean annual temperature (MAT) record of the Congo basin in core GeoB6518-1, based on biomarkers (21). (E) Abundance of pollen from herbaceous plants (mainly Gramineae) at the Lake Barombi Mbo (Cameroon), reflecting the relative presence of savannas versus forests in western equatorial Africa (3). ka, thousand years ago.

Comparison of our proxy data with organic geochemical and molecular records from core GeoB6518-1 suggests that much of the weathering signal at site KZAI-01 is driven by continental precipitation. From ~20,000 to 3500 years ago, our weathering record exhibits strong correlation with the precipitation signal from core GeoB6518-1, inferred from the deuterium composition of plant waxes (Fig. 3C) (19, 20). This observation suggests that chemical alteration in the Congo basin has responded quickly to regional climatic changes, at least for the time scales being considered here. The trends toward wetter conditions that are visible in the GeoB6518-1 deuterium record, between ~18 to 13 and ~12 to 9 ky B.P., coincide well with marked periods of intensifying chemical weathering. Similarly, the progressive onset of dryer conditions since ~6 ky B.P., which marks the end of the African Humid Period, is accompanied by lower weathering rates. Reduced weathering rates also occurred during the Younger Dryas, between ~12.8 and 11.5 ky B.P., a period characterized by lower precipitation levels in Central Africa (19). In comparison, the evolution of mean annual temperatures in Central Africa has been very gradual since the last deglaciation, rising smoothly from ~21° to 25°C (21). Most likely, this suggests that temperature only played a minor role in controlling past chemical weathering variations in the Congo basin during the Late Quaternary.

From ~3500 years B.P., an abrupt trend toward higher Al/K and Hf isotope values indicates rapidly intensifying chemical weathering. The weathering peak, centered at ~2500 years B.P., is characterized by the highest Al/K and εHf values measured throughout core KZAI-01, indicating that global weathering rates in the Congo basin during that period were higher than at any other time in the past 40,000 years. After ~2000 years B.P., chemical weathering intensity values decreased slightly but still remained at much higher levels than before 3000 years B.P. The weathering episode occurred contemporaneously with the major vegetation change in Central Africa during the third millennium B.P., illustrated in Fig. 3E by the sudden increase in the abundance of herbaceous pollen taxa (Gramineae) at Lake Barombi Mbo (see location in Fig. 1) (3). This event is well documented in numerous palynological and sedimentological records (14), from the equatorial Atlantic coastal area to the eastern border of the Congo basin, near the Great Lakes region of Africa (Fig. 1). At many sites, proxy records for past vegetation patterns indicate a major loss of primary forest between ~3000 and 2200 calendar years B.P. and its replacement by savannas and other pioneer formations. To some extent, this large-scale deforestation event shaped the African rain forest into its present-day vegetation patterns (1, 2). The cause usually invoked for the forest disturbance is a global shift toward seasonally dryer conditions in Central Africa (1, 4). This hypothesis is in agreement with the Late Holocene rainfall signals for tropical regions, which indicate reduced precipitation levels from ~4000 years B.P. (1, 19). At that time, one would expect the weathering signal at site KZAI-01 to follow the same way it evolved during the past 40 ky when continental climate became dryer; i.e., toward lower intensity levels (lower Al/K ratios). Instead, evidence that chemical alteration strongly intensified during the third millennium B.P., thus departing from the long-term weathering fluctuations related to the Late Quaternary climate, suggests that this weathering event was not triggered by natural climatic factors.

We are confident that this pulse of intense chemical weathering does not reflect reworking of sediments on the shelf, due to sea-level rise, for example, or denudation of strongly weathered ancient soils from the Congo basin. At the time of deposition, the global mean sea level in the oceans had already been close to modern values for several millennia at least (22). In addition, the calibrated ages for bulk organic matter samples in this part of core KZAI-01, which include a substantial continental organic fraction (23), agree well with those inferred from our age model (determined from radiocarbon dating of marine carbonate material) (see table S3). Importantly, this suggests that the suspended particles transported by the Congo River during that period were mainly derived from relatively young soils rather than from older tropical soils. Taken together, these observations clearly show that the anomalously high Al/K values in the upper part of core KZAI-01 correspond to a true contemporaneous signal of chemical weathering from the Congo basin.

In fact, recent archaeological surveys showed that the deforestation event in the third millennium B.P. coincided with the large-scale settlement of Bantu-speaking farmers in subequatorial Africa (4, 10, 24). The first Bantu speakers were cultivators in the eastern Nigeria and western Cameroon area, who began to spread eastward and southward ~4000 years ago (59). From the third millennium B.P., a major expansion wave was associated with the introduction of agriculture into the central African rainforest (5, 10, 24). This period coincides with marked increases in the abundance of oil palm pollen at numerous sites across West and Central Africa, interpreted as evidence for intensifying plant cultivation (25). In this region, numerous archeological sites containing ceramics, domesticated crop remains, oil palm nuts, and stone tools were dated between ~3000 and 2000 years ago (24, 2629). At that time, the cultivation of savanna crops, such as pearl millet and yams, was made possible by the onset of seasonality alternance between wet and dry seasons (4, 24). The discoveries of several iron-working furnaces and smelted iron artifacts in Cameroon, Gabon, the Central African Republic, and Congo, dating from the same period or even older, also indicated that the Bantu farmers were carrying the technology for iron metallurgy (2933).

One hypothesis to link the rainforest crisis to intensifying human activities in Central Africa during the third millennium B.P. has been to propose that the deforestation event created favorable conditions for the settlement of Bantu farmers across Central Africa, through opening of savanna corridors (4, 34). Alternatively, the introduction of novel agricultural practices and iron-smelting technology may also have led to intensive land clearance for shifting cultivation and charcoal production, thus being partly responsible for the major vegetation change ~2500 years ago (10). Because land use, anthropogenic deforestation, and agriculture would have increased rates of soil erosion and, as a consequence, chemical weathering, intensifying human activities in subequatorial Africa hence represent a plausible explanation for the weathering episode in the third millennium B.P. Based on our results, it is difficult to assess the degree to which human land use and/or climate change played a role during this Late Holocene deforestation event. However, evidence from our proxy record that chemical weathering rates at that time were unprecedented during the past 40,000 years clearly suggests that the environmental impact of the human population in the central African rainforest was already extensive ~2500 years ago, at least greater than that induced by the Late Quaternary climatic oscillations.

Supporting Online Material

Materials and Methods

SOM Text

Fig. S1

Tables S1 to S3

References (3567)

References and Notes

  1. Hydrogen isotope abundances (δD, in per mil relative to Vienna standard mean ocean water) in terrestrial plant waxes are controlled by the precipitation-evaporation balance and humidity levels in tropical areas, therefore reflecting predominantly continental rainfall fluctuations (19).
  2. See data compilation for rock samples at
Acknowledgments: This work was sponsored by the French National Research Agency (ANR), via the ECO-MIST project (2010 JCJC 609 01). Core KZAI-01 was collected during the Zaiango project funded by IFREMER and TOTAL (chief scientist: B. Savoye). We thank F. Jansen for providing five radiocarbon dates and giving access to the x-ray fluorescence core-scanner data set for core GeoB6518; J. Maley for stimulating discussion, sharing bibliographic resources, and providing Lake Barombi Mbo pollen data; Y. Germain for help in the laboratory; and W. F. Ruddiman, N. C. Chu, and five anonymous referees for their constructive comments on the manuscript. The data reported in this paper are listed in the SOM.

Stay Connected to Science

Navigate This Article