Dark soil is an effective greenhouse gas absorber

https://www.science.org/doi/10.1126/sciadv.adh8499 Intentional creation of carbon-rich dark earth soils in the Amazon The Terra Preta do Mangabal (TPM) archaeological site is located on a high forested bluff on the left bank of the Upper Tapajós River within the traditionally occupied territory of riverine (beiradeiro) communities (fig. S4). Much of the dark earth at the TPM site is concentrated on the edge of the bluff and in the most elevated area. The site was used as a homestead in the recent past and a small area on the edge of the bluff is currently being cultivated with bananas. It appears that cultivation covered a small part of the site along the bluff edge in the recent past. The northern half of the site consists of old-growth forest with canopy emergent trees, resulting in the high organic carbon levels near the surface in the middle of the soil transect due to the thick forest litter layer. We excavated 70 auger holes at distance intervals of 25 to 50 m to delimit the archaeological deposits at the site. Our excavations indicate a reduction of artifacts northward as the landscape transitions to grassland, indicating an estimated area of at least 20 ha. The TPM site contains a large quantity of ceramic and lithic remains, as well as wood charcoal, carbonized seeds, and faunal remains. The average depth of dark earth at the site is 50 cm, although areas of middens and mounded deposits contain deeper dark earth horizons (57, 58). Available radiocarbon ages range from 1260 to 940 cal BP (fig. S28 and table S7) (57, 58), while available OSL dates range between 1572 ± 188 before present (BP) and 1135 ± 81 BP (table S8). These dates are interpreted to be from a single, continuous occupation that has been related to Tupian speakers ancestral to the Munduruku people (59). The Mangangá archaeological site is a forested site in a valley in the Carajás Mountains located along the Sossego River (a mountain stream with headwaters on the nearby plateau) near the confluence of a small tributary (fig. S5) (60). The riverbank is a few meters high with a narrow floodplain, 20 to 30 m wide, on the southeast and south side of the site and an upper terrace where most of the archaeological deposits were found. The transect presented here is 100 m long with sampled profiles every 10 m. It begins at the river’s edge, crosses the narrow floodplain (20 m), and traverses the slope and upper terrace through archaeological deposits with dark earth (61, 62). Radiocarbon dates on and near the transect range from 3700 to 500 BP, but the lower levels of excavations in other areas of the site were dated to as early as 11,800 BP, including early Holocene soil enrichment (fig. S29 and table S7) (60). Mapping Mapping of archaeological features and excavations in the Upper Xingu was carried out with a Trimble XRS Global Positioning System (GPS) receiver with real-time correction (41, 42). Mapped features include ditches, plazas, roads, and water access locations. Plazas and roads are bordered by linear mounds up to 1 m high. These features were mapped by collecting points at intervals of several meters in the approximate center of the mound or ditch. Additional sample locations were recorded with a Garmin hand-held GPS. At the TPM site, sample locations were mapped with a total station and georeferenced with a Garmin hand-held GPS. Contours were derived from the Multi-Error-Removed Improved-Terrain (MERIT) digital elevation model (63). At Mangangá, the topography and sample locations were mapped with a total station and georeferenced with a Garmin hand-held GPS. Soil sample collection Soil samples were collected during archaeological excavations or in transects using a bucket auger. Excavations included 1-m-wide trenches that bisect archaeological features, 1-m2 excavation units (including block excavations), or 50 × 50 cm test pits. Samples were collected from excavation walls with a trowel in a vertical column in 5- or 10-cm increments. Additional samples were collected at 1-m intervals in transects within or outside excavations using an 8-cm bucket auger to extract a core in 5- or 10-cm depth intervals up to 2-m deep. At Kuikuro II, samples were collected from four test pits along a 60-m transect beginning in a backyard refuse disposal area and ending in a manioc field outside of the village (Fig. 1B). Additional samples were collected at 1-m intervals on transects within village zones (plaza, house, backyard, and refuse middens) and activity areas (hearths and manioc processing) (27). Samples were collected from a transect in the center of and parallel to an old midden that was formerly on the edge of a backyard at the Ipatse village site (occupied ca. 1920–1940). At the historic village site Kuikuro I (occupied ca. 1973–1983), samples were collected in the former plaza, domestic areas, middens, and trails (27). One 52-m transect, with samples at 1-m intervals to a depth of 30 cm, began in the plaza, passed through a former house and backyard, and lastly over a mounded midden (fig. S3). At Seku, a transect with seven test pits begins in the mound surrounding the plaza and extends for 970 m between two major roads (Fig. 1C). At Akagahütü, we sampled a transect traversing the site from the edge of the floodplain, adjacent to a probable excavated pond, to the peripheral earthwork (ditch), and four additional test pits were excavated beginning on the outside of the ditch and leading away from the site between two major roads (fig. S1A). At Ngokugu, a 100-m transect, with cores every 5 m, begins on the outer edge of the circular central plaza, traverses the plaza mound, and crosses a residential area before terminating near the inner ditch (fig. S1B). Additional test pit transects traverse residential areas within the inner ditch and between the inner and outer ditches (27). Test pit transects at Heulugihütü pass through residential areas outside of the central plaza (fig. S1C) (27). At TPM, the 400-m transect begins on the upper slope of a steep bluff overlooking the Tapajós River, heads inland (north-northwest) crossing the relatively flat central area of the site until it leaves the forest and enters an adjacent savannah (fig. S4). The 100-m transect at Mangangá begins at the river’s edge, crosses a narrow floodplain, goes up a low slope with deposited dark earth, crosses a flat area devoid of dark earth, and then passes through a second deposit of dark earth before lastly entering an area of decreased enrichment beyond the second deposit (fig. S5) (61, 62). Soil laboratory analysis We analyzed 3532 soil samples from 1176 individual locations (dataset S1). Each sample corresponds to a discrete depth range (e.g., 10 to 20 cm) from the excavation, test pit, or auger core. Laboratory analyses of soil samples were carried out at EMBRAPA Soils in Rio de Janeiro, EMBRAPA Amazonia Oriental in Belém, the Luiz de Queiroz College of Agriculture (ESALQ)/University of São Paulo in Piracicaba, the Department of Ecology at the Emilio Goeldi Museum (MPEG) in Belém, and the environmental laboratory of Eletronorte in Belém. Samples were air-dried and screened through 2-mm mesh in preparation for chemical and physical analyses. For selected samples, particle-size analysis was performed on the <2 mm fraction. The sand fraction was measured by wet sieving, and the pipette method was used with 20 g of soil in 100 ml of distilled water plus 10 ml of 1 M sodium hydroxide (NaOH) for measuring clay and silt fractions. Physical analyses included measurement of magnetic susceptibility (MS) and apparent electrical conductivity (ECa) using a Terraplus (Canada) model KT10 SC instrument. To standardize the samples for analysis of MS and ECa, samples were placed in petri dishes 9 cm in diameter and 1.7 cm deep, holding approximately 150 g of soil. All samples were analyzed for SOC using the modified Walkley-Black method, and soil pH was determined in distilled water (1:2.5 soil:solution) (64). A total of 193 samples were analyzed for fertility including measurements of pH in potassium chloride (KCl), exchangeable Al, Ca, and Mg by 1 M KCl extraction, and available P, K, Na, Cu, Fe, Mn, and Zn extracted with the Mehlich-1 solution [0.05 M hydrochloric acid (HCl) and 0.0125 M sulfuric acid, (H2SO4)] method (64, 65). For 3339 samples, a standard hydrofluoric acid (HF) digestion was used in a closed-vessel microwave system to extract total elements from 0.1 g of sample (27, 66). The mass concentration Cm of Al, Ba, Ca, Co, Cr, Cu, Fe, K, Mg, Mn, Na, Ni, P, Pb, Sr, Ti, V, and Zn was measured by inductively coupled plasma atomic emission spectroscopy (Varian Vista Pro simultaneous) with axial viewing, a radio frequency of 40 MHz, and charge-coupled device detection. Soil data analysis Fence diagrams of SOC, pH, and elemental mass concentration (Fig. 2 and figs. S2, S3, and S6 to S15) were generated by linearly interpolating along transects between sampled test pits. We assumed that each sample is representative of its associated depth range and that quantities are uniform across that range. For one missing sample (Seku, 970 m along transect, 30- to 40-cm depth), we estimated values by averaging the samples immediately above and below in the same test pit. PCA was performed separately for each site using soil data normalized to a common mean and variance. PCA of the Kuikuro II, Seku, Mangangá, and TPM transects included pH, SOC, and available or extractable elemental concentrations. PCA of Akagahütü, Ngokugu, and Kuikuro I transects included pH, SOC, and total element concentrations. We plot fence diagrams of the first principal component as described above and give the weighting coefficients in tables S1 and S2. To estimate SOC and phosphorous inventories, we used the average concentration in the upper 1 m of soil at each site. To compute this average, we used a depth-integrated approach. At each depth horizon between 0 and 1 m, we computed the average of all samples whose depth range includes this horizon. We combined these averages to estimate an average depth-concentration curve for each site (figs. S16 to S26), which we integrated to a depth of 1 m to compute the average concentration. Because of the nature of the sampling, this approach typically results in fewer samples representing deeper levels than shallower levels. We performed a similar calculation for samples collected outside dark earth sites to compute background soil properties. This calculation yields the average mass concentration Cm (M/M, dimensionless) in the upper 1 m of the soil. We report these values in tables S3 and S4. We convert the mass fraction Cm to a volumetric concentration Cv (M/L3) by multiplying by the bulk density rb (M/L3), Cv = rbCm, assuming a soil bulk density of 1100 kg/m3 (67). This expression gives the average mass per unit volume of a soil quantity (e.g., SOC or P). By multiplying by a depth of 1 m, we calculate the areal density (M/L2); this is the average mass per unit area contained within the upper 1 m. To estimate the total mass (M) contained within an archaeological site, we then multiplied this average by the area of the site (L2), which we estimated using a combination of field mapping, test pits, earthworks, and vegetation patterns in satellite imagery (tables S3 and S4). For the modern Kuikuro II village, we calculated carbon and phosphorus inventories from measured concentrations and mapped areas of middens in 2002 (27). In the historic Kuikuro I village, we used measured concentrations and mapped areas of middens in 1993 (41). We report the areal densities, the mapped areas, and the total inventories in tables S3 and S4. Upper Xingu sites differ in forest cover and recent land use history, as many of the ancient dark earth sites have been used for cultivating crops within living memory. Each site was designated as forested or deforested; in this case, only Seku was designated as a forested site. To account for the lower naturally occurring SOC and nutrient concentrations in deforested settings, we computed background concentrations separately for forested and deforested samples away from archaeological sites (9.2 g/kg SOC and 856 mg/kg total P in forested areas; 6.8 g/kg SOC and 277 mg/kg total P in deforested areas). We subtracted the appropriate value from each dark earth sample to estimate the anthropogenic contribution (tables S3 and S4). Mass concentration data in the supplementary table were standardized to mg kg−1. Results that were reported in cmolc or mmolc were converted to mg by multiplying mmolc by the atomic weight of the appropriate element. For results reported in volumetric units (dm3) (Mangangá samples), a pedofunction was used that estimates the fine earth density (<2-mm grain size) based on the quantity of organic carbon (68). Geochronological analysis We collected samples for radiocarbon dating from charcoal in situ in archaeological test pits and excavations at Akagahütü, Seku, and areas between archaeological sites (table S7). Samples were measured by accelerator mass spectrometer at Beta Analytic in Miami, Florida. We converted radiocarbon dates to calibrated ages with the SHCal20 calibration curve (69) using OxCal 4.4 (70). We also compiled previously published radiocarbon dates from Ngokugu, Heulugihütü, Kuhikugu (9), Mangabal (57), and Mangangá (60) and recalibrated these dates with the updated calibration curve. We report all radiocarbon dates and calibrated ages in figures S27 and S29 and table S7. Optically stimulated luminescence (OSL) dating (table S8) was performed at the Laboratory of Gamma Spectrometry and Luminescence at the Institute of Geosciences, University of São Paulo. The dose rate was estimated by gamma spectrometry with a high-purity germanium detector using ultralow background shielding. The dose equivalent was determined by single-aliquot regenerative-dose protocols with multigrain aliquots of quartz. The OSL measurements were carried out with a Lexsyg Smart detector equipped with a beta radiation source (Sr/Y) with a dose rate of 0.116 Gy/s. The preparation of quartz aliquots included the following steps: First, detrital grains in the size range of 180 to 250 μm and 125 to 250 μm (sample 5522) fractions were recovered by wet sieving; second, the target fraction was treated with hydrogen peroxide (H2O2, 27%) to eliminate organic matter and hydrochloric acid (HCl, 10%) to remove carbonate minerals; third, a heavy liquid separation with lithium metatungstate (LMT) was used to separate heavy and light minerals (LMT = 2.75 g/cm3) and quartz (LMT = 2.62 g/cm3); fourth, the samples were etched in HF (30%) for 40 min to eliminate the external layer of quartz grains and feldspar remnants. Equivalent doses of samples were calculated using the Central Age Model, Minimum Age Model (overdispersion > 30%), and simple mean average (aliquots with dose saturation) (sample 5522). Only aliquots with a recycling ratio between 0.9 and 1.1, a recuperation <5%, and no contamination of feldspar (IR signal) were considered for the calculation of equivalent dose. A dose recovery test was made on sample 5024 (preheating to 220°C, administering doses of 2.5, 5, and 10 Gy). Ethnographic research Ethnographic and ethnoarchaeological research consisted of observations, mapping, sampling, and recording interviews carried out over 12 months of fieldwork between 2002 and 2019 in collaboration with the Kuikuro community. Informed consent was obtained from all study participants. Observations were used to determine the spatial distribution of activities in the village, which were then mapped using GPS. Soil cores were collected and analyzed in the different activity areas. Interviews were carried out with elder agricultural specialists in the community in the native Kuikuro language. Video recordings of nine interviews were translated into Portuguese by experienced Kuikuro translators and then translated to English (text S2). Portuguese text received minor edits to improve readability but was otherwise left in the translator’s words. We analyzed the interview texts by excerpting and tabulating interviewee responses related to two topics: soil management (table S5) and dark earth fertility and cultivation (table S6). We scored each response related to soil management according to whether it supports the hypothesis of intentional dark earth creation, contradicts intentionality, or neither supports nor contradicts intentionality (table S5). Text S2 provides additional information on the interviews, a glossary of key terms in the Kuikuro language, and the complete translations of the interviews in Portuguese and English. Acknowledgments W