Sparse Pre-Columbian Human Habitation in Western Amazonia

See allHide authors and affiliations

Science  15 Jun 2012:
Vol. 336, Issue 6087, pp. 1429-1431
DOI: 10.1126/science.1219982


Locally extensive pre-Columbian human occupation and modification occurred in the forests of the central and eastern Amazon Basin, but whether comparable impacts extend westward and into the vast terra firme (interfluvial) zones, remains unclear. We analyzed soils from 55 sites across central and western Amazonia to assess the history of human occupation. Sparse occurrences of charcoal and the lack of phytoliths from agricultural and disturbance species in the soils during pre-Columbian times indicated that human impacts on interfluvial forests were small, infrequent, and highly localized. No human artifacts or modified soils were found at any site surveyed. Riverine bluff areas also appeared less heavily occupied and disturbed than similar settings elsewhere. Our data indicate that human impacts on Amazonian forests were heterogeneous across this vast landscape.

The Amazon Basin, an area approximately the size of the continental United States, is an important reservoir of biodiversity. A major recent question is the degree to which humans settled and modified Amazonian landscapes before European contact. It was initially thought that prehistoric Amazonia supported mainly small and highly mobile human populations, who exerted little impact on their environments (1, 2), but recent work has documented dense and complex human settlements in eastern Amazonia and on the river bluffs of the central Amazon. The evidence includes the presence of highly modified soils such as terra pretas (anthropogenic “black earth”) (3) and large-scale landscape alterations (Fig. 1) (4, 510). The evidence is impressive, but comes largely from riverine environments with abundant natural resources, especially river bluffs, or the driest parts of the eastern Amazon (Fig. 1).

Fig. 1

Sampled locations within western Amazonia (white squares) in relation to major pre-Columbian archaeological sites (1, Marajó Island; 2, Santarém; 3, Upper Xingu; 4, Central Amazon Project; 5, Bolivian Beni), known terra preta locations (brown circles) (3, 32, 33), and soil charcoal survey locations (black circles) (12, 22). Charcoal and phytolith data are presented from regions outlined in black (B, Barcelos; T, Tefe; PVM, Porto Velho-Manaus transect; I, Iquitos; Ac, Acre; LA, Los Amigos). The locations of Rio Madeira and associated terra pretas are shown. Here we define Amazonia as the region drained by the Amazon River and its tributaries.

The extent of this impact on terra firme settings has been uncertain. The terra firme forests of the interfluvial zone occupy 95% of Amazonia and have less-fertile soils and poorer-quality resources (11). Available data from several regions suggest that the prehistoric impacts on interfluvial landscapes were heterogeneous and highly localized (12, 13). Here we reconstruct histories of fire, vegetation, and soil modification from charcoal, phytolith, and geochemical data recovered from 247 soil cores collected from 55 locations, including sites with known impacts, across 3,000,000 km2 in western Amazonia (Fig. 1 and table S1) (14). We sampled soils from sites where the probability of past disturbances was high, such as river bluffs with known archaeological histories and nearby terra pretas, including Tefe, Barcelos, and Iquitos; from a previously unstudied river bluff at Los Amigos; and from terra firme sites, including Acre, Iquitos, Tefe, and a transect from Porto Velho to Manaus (PVM).

Natural fires in Amazonia are rare today (1517), but fire was a mainstay of prehistoric land use in the tropics (11, 18, 19). Consequently, charcoal recovered from soils can provide evidence of past human disturbances, and phytoliths, which document mature and disturbed vegetation, reflect the intensity of those occupations. In our samples, charcoal was most common in soils from riverine bluffs, especially in the central basin (Fig. 2, C to F). At Barcelos and Tefe, charcoal was present in many intervals in most cores, especially from 0 to 40 cm (Fig. 2, D and F). Charcoal dates ranged from ca. 500 to 2700 calendar years before the present (cal yr B.P.) at Tefe and from ca. 1200 to 1300 cal yr B.P. at Barcelos (table S2). The vegetation at Tefe appears to have been more heavily affected than that at Barcelos, which is in agreement with the longer span of documented occupation. In riverine settings, Tefe soil phytoliths contained elevated amounts of early successional herbaceous taxa (ESH, such as grasses, Heliconia, and sedges) and some grass phytoliths that were burned. These patterns probably reflect forest clearing and other human disturbances (see phytolith analyses in the supplementary materials and fig. S1). However, neither site yielded crop phytoliths. Arboreal-dominated phytolith assemblages and relatively sparse charcoal from riverine Iquitos sites indicate that the forest remained relatively undisturbed there, and nutrients and black carbon concentrations in soils from these sites were low. At Los Amigos, the charcoal dates ranged from 1000 to 4000 cal yr B.P. (table S2), but the soils were not enriched in nutrients and arboreal taxa dominated phytolith assemblages, which is consistent with a light and shifting human impact (table S4, Fig. 2E, and fig. S1).

Fig. 2

Regional maps, soil charcoal distributions, and phytolith percentages for soil cores from riverine (red squares and text) and interfluvial (black squares and text) sites in each region: Acre (A), PVM (B), Iquitos (C), Tefe (D), Los Amigos (E), and Barcelos (F). Areas of lower (darker) and higher (lighter) elevations illustrate drainage and rivers (from 90-m–resolution data from the Shuttle Radar Topography Mission) on each regional map. Colored boxes indicate charcoal results for each core within each site (see legend). Sites are listed in a north-to-south orientation. Soil cores with accompanying phytolith data are denoted with P. Phytolith percentages (column P) are listed to the right of the charcoal results. Geographic coordinates of all sites are provided in table S3.

We recovered little charcoal from soils at Acre or interfluvial Iquitos sites, indicating a lack of recurrent or extensive fires over the past several thousand years (Fig. 2, A and C, and table S2). Similar results were obtained from the phytolith records, which were dominated by forest taxa; ESH phytoliths were absent or rare (0 to 1%). No evidence for crops or burned phytoliths was found (fig. S1). Charcoal was more common in soils of the PVM transect than in the western interfluvial Iquitos or Acre sites (Fig. 2, A to C). However, phytolith records showed no signs of a significant human presence at most sites. ESH phytoliths were absent or scarce (0 to 6%), and burned tree phytoliths were nearly absent (Fig. 2B and fig. S1); forest taxa dominated in all samples. Site 121 contained evidence of maize cultivation and elevated frequencies of grass and Heliconia phyoliths, many of which were burned. No other crops, including squash (Cucurbita spp.), manioc (Manihot esculenta), arrowroot (Maranta arundinacea), and leren (Calathea allouia), were found. Because manioc produces fewer phytoliths than many other crops, we cannot state with the same confidence that it was not grown nearby.

We found no prehistoric ceramics, stone tools, or terra pretas in any of the 247 soil cores, and none of 184 samples analyzed for phytoliths contained evidence of intensive or persistent forest clearing. In many soil levels, no ESH phytoliths were observed in scans of >500 to 1000 additional phytoliths, underscoring the lack of disturbance that took place in these interfluvial forests. Together, the data suggest that human population densities in the sampled regions were low and highly localized, and were not consistent with major population centers with associated areas of widespread, extensive agriculture (20). Our data support the idea that humans had much less impact on interfluvial forests than on riverine environments (21) or in the drier eastern forests (22). However, even regions with known human sites and terra pretas (such as Barcelos and Tefe) were not subjected to continuous or large-scale forest clearing or intensive agriculture (Fig. 2), and show a lesser disturbance signature than found in modern slash-and-burn systems (see phytolith analyses in the supplementary materials). Forest clearings were probably small and short-lived, and the interior forests were apparently not permanently or intensively occupied by humans in prehistory. We found little indication that repeated fire, vegetational disturbance, and/or agriculture extended more than 5 km into the terra firme forests of the Tefe, PVM, Acre, and Iquitos regions (Fig. 2).

Our data imply that the disturbance signature was stronger in both riverine and interfluvial forests of the central basin than in the western basin (Fig. 2). Even in the PVM transect, however, evidence for disturbances was patchy and localized, despite being located 20 to 50 km from the Madeira River and within 100 to 200 km of dense concentrations of terra pretas (23) (Fig. 1). The frequency and distribution of terra pretas documented along the Madeira River (24) may have continued southward, parallel to our interfluvial transect. The resulting contrasting pattern of highly concentrated terra preta soils along the river, with localized and patchy disturbance 20 to 50 km into the uplands, illustrates how even in the central Amazon, intensive landscape modifications appear to be confined to near-riverine locations.

We interpret the charcoal presence along with low frequencies of burned tree phytoliths, and the dominance of forest over grass phytoliths, to mean that fires were mainly confined to the forest floor. The apparently infrequent and low-intensity fires do not appear to have penetrated canopies and altered forest structure substantially at most sites. Therefore, soil charcoal alone should not be taken to mean that fires were of sufficient intensity and duration to cause canopy disruption and major forest alteration [see also (12)].

It is likely that in some forests, edible or other useful fruit trees were planted or managed, resulting in an enrichment of those species (25). Palms such as peach palm (Bactris gasipaes) and Astrocaryum are economic mainstays in the Amazon and are prolific phytolith producers. We found no evidence for these species in most samples from every site studied (fig. S1 and palm distributions in the supplementary materials). There was no association between palm phytolith frequencies and other evidence of vegetation disturbance, and palm frequencies were never so high that they implied that a local grove was present. These data suggest that humans were not cultivating or selectively managing palms at most of our study sites. There was also no indication that many noneconomic species were selectively removed (26), because little change in forest composition was seen from the bottom to the top of the soil cores, including when early successional herbaceous taxa and/or charcoal were present.

Our data imply that the terra firme forests we studied in the western Amazon Basin were predominantly occupied by relatively small and shifting human populations during the pre-Columbian era. This has many implications for hypotheses about human effects on Amazonian forests. First, humans may have augmented the alpha-diversity of some Amazonian landscapes, but the hyperdiverse floras and faunas are more a product of long-term evolutionary and ecological processes (27) than anthropic landscape alteration (4, 26, 2830). Second, to the extent that prehistoric deforestation occurred, it was apparently primarily in the eastern Amazon, and this may have limited the proposed impact of post-Columbian population collapse and reforestation on atmospheric CO2 and CH4 levels (18, 31). Third, we cannot assume that Amazonian forests were resilient in the face of heavy pre-Columbian disturbance, because vast areas were probably never heavily disturbed. Prehistoric peoples settled most densely in habitats where resources were abundant and easily captured, fertile soils were available, and transportation routes were nearby, making ecological factors important in pre-Columbian settlement patterns.

Supplementary Materials

Materials and Methods

Supplementary Text

Fig. S1

Tables S1 to S4

References (3465)

References and Notes

  1. Site descriptions and materials and methods are available as supplementary materials on Science Online
  2. Acknowledgments: Field work and 14C dating of charcoal fragments were funded by the NSF Ecology Program (awards DEB 0742301 and DEB 0743666). Other funding was provided by the Florida Institute of Technology; the Smithsonian National Museum of Natural History, including a Restricted Endowment and Small Grant Award; and the Smithsonian Tropical Research Institute. All data will be deposited in the Neotoma Database ( We thank B. McMichael, A. Correa-Metrio, J. Hernandez, T. Harrison, and B. Rado for field assistance.
View Abstract

Navigate This Article