Historical Human Footprint on Modern Tree Species Composition in the Purus-Madeira Interfluve, Central Amazonia

Picture a map of the Amazon. The rivers are the highways, the settlements cluster along the banks, and the vast green interior between those rivers sits labeled, implicitly and explicitly, as untouched primary forest. That assumption has shaped conservation policy, land rights, and biodiversity science for decades. Then a field team went into that interior and found something in the soil that changes the map entirely. The prevailing view in Amazonian ecology held that pre-Columbian human impacts were concentrated on river bluffs and floodplains. The interfluvial zones — the upland land between river systems — were considered largely pristine wilderness. Part of what anchored this view was the spatial pattern of Amazonian Dark Earths, or Terra Preta de Índio: the ancient, charcoal-rich, artifact-laden soils that mark where Indigenous peoples lived and farmed. Those soils cluster along white-water river margins. If the soil signatures of past habitation hug the rivers, the thinking went, then the forest in between was probably left alone. But the picture was already fraying at the edges. Hundreds of geoglyphs — geometric earthworks — in the upper Purus-Madeira interfluve span an area roughly 250 kilometers north to south. Earthworks elsewhere in Amazonia turn up on interfluves, not riverbanks. Forests dominated by a handful of useful species, what researchers call oligarchic or anthropogenic forests, appear in unexpected places: castanhais, stands thick with Brazil nut trees, and caiauezais, groves of the palm Elaeis oleifera. Some of those Brazil nut trees exceed 220 centimeters in trunk diameter, making them probably older than the European colonization of Amazonia. Carolina Levis and colleagues went to the Purus-Madeira interfluve in Amazonas, Brazil, to ask a precise question: does the signal of past human forest management drop off sharply as you move away from rivers, or does it persist deep into what we call primary forest? To answer that, the team worked across six mature forest sites drawn from pre-existing monitoring installations. At each site, they walked into closed-canopy forest and measured every tree and palm with a trunk diameter of at least 10 centimeters in half-hectare plots. Larger trees — those above 30 centimeters in diameter — were sampled across full hectares. The 10-centimeter threshold was deliberate: stems that size may be descendants of pre-conquest planting and tending, recruits of trees that ancient peoples promoted and whose offspring still populate the understory. Botanical specimens were pre-identified in the field and confirmed against herbarium collections at Brazil's National Institute for Amazonian Research. The researchers compiled a list of useful species from the ethnobotanical literature: any species mentioned in at least two published studies, with documented food, medicinal, or material value, made the cut. The key metrics were the relative abundance of useful individuals, their relative basal area, and useful-species richness. To detect fire linked to past human activity, crews dug soil pits at 29 charcoal plots across the six sites, extracting undisturbed samples at 10-centimeter intervals down to 50 centimeters, drying them, and weighing the visible charcoal. Archaeological evidence — Dark Earths, mapped anthropogenic forests, artifacts — was located by observation and by asking local residents, then recorded with GPS. The principal independent variable was simple but powerful: the straight-line distance from each plot to the nearest major river. Now picture yourself walking away from one of those wide Amazonian waterways into the forest. Close to the water, the tree community still carries a strong human signature. Plots sitting 10 to 20 kilometers from a main river contained between 20 and 40 percent useful arboreal species and palms. Step farther out, between 20 and 40 kilometers, and that share falls to roughly 12 to 23 percent. Beyond 40 kilometers it slips below 15 percent. The relationship is a highly significant negative exponential curve: a steep drop within the first 20 kilometers, then a slower, shallower decline. Concrete sites anchor that gradient. Site M1, just 11 kilometers from a main river, recorded 195.5 useful tree and palm individuals per hectare. Site M2, at 36 kilometers, still held 131 useful individuals per hectare. These are not monopolies of a single species — they are suites of useful plants that vary by site, and the highest-concentration sites sit on paleo-várzeas, ancient floodplain terraces that predate the current hydrological regime. Simple regressions using distance to major and secondary rivers explained between 50 and 90 percent of the variation in useful-tree parameters. When the analysis zoomed in on secondary rivers alone, it still accounted for about half the variance. Palms deserve a separate sentence: they were the most abundant useful family, and when palms were removed from the analysis, the relationship for dicot trees fell to borderline significance. A multiple regression linking useful-palm abundance to distance from rivers and a hydrological gradient index produced an R-squared of 0.73, with distance to rivers the dominant term. The soil evidence tracks the same riverward pulse. Charcoal turned up in virtually every soil layer at every site, down to 50 centimeters. But ubiquity is not proof of farming. Levis and colleagues used three cross-checks to separate background fire from deliberate burning. First, depth: charcoal in the top 20 centimeters often originates from modern activity, while material deeper than 20 centimeters is probably pre-Columbian. Second, association: charcoal that co-occurs with phytoliths — microscopic silica bodies from plants — or with pollen of cultivated species is much stronger evidence of agricultural burning than charcoal alone. Only two of the 13 sites showed that pairing, and the team treats those as confirmation of past agriculture. Third, archaeological corroboration: plots with high charcoal masses that also sit near Dark Earths and dense useful-species populations make a converging case. The picture is site-specific. Site M2, 36 kilometers from the Solimões River and just 5 kilometers from the secondary Janauacá River, had unusually high charcoal across all soil layers, well above background values reported for upland forest soils without past human activity. Site M1's charcoal was concentrated in the top 20 centimeters and may partly reflect more recent burning by rubber tappers. Elsewhere, charcoal was common but not abundant, and crop phytoliths were scarce in the interfluve interior. The researchers are explicit about the limits: charcoal alone cannot testify to non-fire management — planting along trails, seed dispersal, selective weeding — and low-intensity fires without other indicators cannot be taken as evidence of extensive disturbance. Where multiple proxies align, they infer intensive past management; where proxies diverge, the conclusion is limited disturbance or unresolved origin, and the paper calls for further archaeological work to locate cultivated landscapes within the interfluve. What makes all of this matter beyond the forest ecology literature is the conservation implication. Forests that look primary are, in many places, legacies of long-term human management. The resource concentrations the team documented — enriched, oligarchic forests and Dark Earth soils — retain useful plants that Indigenous and local peoples depend on today for food and for commercially valuable non-timber forest products. The Brazilian National System of Conservation Units already recognizes people living within extractive-use forests. The argument Levis and colleagues make is that those human-enriched landscapes have a biodiversity role that standard ecological inventory, blind to history, will systematically underestimate. The authors are careful about scope. Six sites in one interfluve cannot speak for an entire basin. The Purus-Madeira region may not be representative of Amazonia broadly, and they call for wider sampling before anyone draws basin-scale conclusions. They also make a methodological argument: floristic inventories need to be combined with paleoecological and archaeological data, because charcoal, pollen, or phytolith analyses alone can miss the low-intensity, long-duration human influences that nonetheless shaped the forest. What should shift when you look at a map of the Amazon is something specific. See the rivers and secondary channels not merely as hydrographic features but as axes of past human shaping that extend tens of kilometers into upland forest. The gradient is real, it is measurable, and it is still written in the trees. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.