Classification: Documented

Hidden Amazonian geoglyphs: history beneath the canopy

Circles, squares, ditches and roads built by Indigenous peoples more than two thousand years ago have changed how western Amazonia is understood. LiDAR has revealed hundreds of previously undocumented earthworks; predictions of thousands more remain models, not confirmed finds. This dossier follows Acre’s sites, evidence of forest management and the questions still open.

Região de Rio Branco, Acre, Brasil · See stories from this place and nearby

The design a walker may not see

In eastern Acre, a person may cross uneven ground without realising it marks the edge of an ancient earthen square. From above the outline closes; elsewhere it forms circles, polygons or enclosures linked by roads. These are the so-called Amazonian geoglyphs: excavated ditches with the displaced soil piled into embankments. The term should not suggest a surface drawing like the Nasca lines. Here the figure is mostly constructed relief.

Brazil’s heritage institute, Iphan, places many recorded examples near the Acre and Iquiri rivers and recognises them as traces of long Indigenous occupation. This dossier’s map pin indicates only the Rio Branco region, not a particular archaeological site. The scale of the works is striking; harder to imagine is the repeated labour of opening, maintaining and linking these places in a landscape of forest, rivers and uplands across many generations.

Discovery did not begin with the laser

In the 1970s, archaeologist Ondemar Dias recorded structures in Acre through the Pronapaba programme, according to Iphan. Later research by Brazilian archaeologists and collaborators expanded the inventory. Aerial views and satellite imagery made outlines clearer than they are at ground level. Some of this growth in records had a troubling cause: modern deforestation removed vegetation and exposed ancient traces.

By 2007, Denise Schaan, Martti Pärssinen, Alceu Ranzi and Jacó Piccoli had published an analysis challenging the old image of Amazonian uplands as unable to sustain complex social organisation. Their paper did not uncover a ready-made “lost city” with a modern name. It brought together evidence of societies able to coordinate regional earthwork architecture, opening new questions about occupation, networks and collective labour.

What shapes and dates actually show

The enclosures combine ditches and banks in varied layouts; some join roads and others occur in clusters. A 2026 Nature study describes ditches reaching five metres deep in some cases and enclosures usually covering 0.35 to 14 hectares, although the largest known approaches 50 hectares. The variation matters: speaking of “the geoglyph” as a single plan erases centuries of changed technique and use.

The same research places construction of the geometric enclosures it calls Aquiry at roughly 600 BCE to 850 CE, on the basis of dates and modelling. Some structures in the region are later than 1000 CE, including mound villages; they must not be assigned the earlier period merely because they look similar from above. Radiocarbon dating, ceramics, stratigraphy and spatial comparison help untangle these histories. An aerial image reveals shape, not age or the builders’ identities by itself.

How LiDAR finds ground beneath leaves

LiDAR sends laser pulses from an aircraft and measures some of their returns. A fraction passes through gaps in vegetation to the ground; the data create a digital terrain model that visually reduces the tree cover. It can reveal relief previously hard to notice, but it is not a perfect X-ray. Dense canopy may leave too few returns from the soil. A form suggested by a model still needs archaeological assessment and, where possible, fieldwork.

On 2024 flights over parts of Acre and Amazonas, a team led by Pärssinen surveyed 4,497 km². In its 2026 paper it reported 432 earthworks in those data, 396 previously undocumented. Those are survey results under specified criteria; they do not mean all 396 sites have been excavated. The paper itself discusses limitations of ground visibility beneath the forest. Distinguishing detection, verification and dating is essential when reading “discovery” headlines.

From 432 signals to tens of thousands: the model’s leap

The 2026 team combined LiDAR samples, known records and satellite images to propose 24,000-30,000 earthworks within about 183,000 km² of southwestern Amazonia. That is an extrapolation, not an inventory of 30,000 visited sites. It depends on densities measured in flight strips, assumptions about their extension and what the canopy still conceals. The authors also estimated 1.25-3 million inhabitants at a regional peak in 100-300 CE. This rests on assumptions about how many centres operated simultaneously and how many people maintained them.

In 2023, a Science paper had predicted 10,272-23,648 large sites still undiscovered across the whole Amazon basin, using different sampling and modelling. The 2026 regional projection exceeds that older range. The numbers should not be added: they have different scopes, categories and premises, and the newer result calls for reassessing the older model. This tension is ongoing research, not proof of fraud or confirmation of every predicted site.

Gathering places, forts or homes?

What were the enclosures for? The leading interpretation in the 2026 paper is public, political and ceremonial gathering centres, connected by roads and used by dispersed communities without permanent residence within every figure. Size, labour investment, links and limited domestic evidence in some enclosures support the idea. They cannot reconstruct an exact ritual, a language or the governing authority of every community. A ditch alone does not prove a fortress.

The paper uses Aquiry for a broad multicultural archaeological tradition. It is a research category for organising evidence, not the attested name of a single empire or every people who lived there. Living Indigenous peoples are not relics of a vanished past: they participate in the landscape and its interpretation. The study mentions Apurinã collaborators when comparing ceremonial spaces and knowledge; ethnographic parallels can guide questions but do not automatically identify the builders of each enclosure.

The forest was managed, not simply removed

A 2017 PNAS study analysed sediments, charcoal and phytoliths (mineral microfossils of plants) near Acre geoglyphs. It found signs of long-term management, including increased palms, and local, short-lived clearings in the study area. It did not find extensive, persistent clearance comparable with modern deforestation. Monumental construction therefore does not require imagining a region entirely stripped of trees.

The 2026 study suggests repeated burning of bamboo-rich patches and crops such as maize and squash, alongside management of useful trees. The studies differ in scale, sites and sampling; extending one soil profile to all Amazonia would be rash. The documented point is richer than a choice between “untouched forest” and “destroyed forest”: people shaped parts of its vegetation while maintaining a mostly forested landscape.

Why an image cannot count a population

The 2026 authors report 162 radiocarbon dates from 32 sites and their 36 structures. This is a substantial dataset, but small next to the thousands of works their model predicts. The regional population was not counted house by house. It was inferred from required labour, probable overlap in centre use, comparisons with later settlements and spatial distribution. Change those assumptions and the result changes.

The study distinguishes the geoglyphs from later mound settlements. Combining every shape into one “city map” would turn different centuries into a simultaneous metropolis. The reach of forest urbanism and the degree of permanent occupation remain debated. The physical remains already demand that we abandon the idea that public architecture and linked social life were impossible away from floodplains.

Heritage facing roads and clearance

There is a bitter irony: land clearing that exposed many patterns can also destroy their embankments and erase relationships with older vegetation. Iphan reports that roads cross some sites and identifies their management as a challenge. The Jacó Sá site received formal protection in 2018. That protects a particular place; it does not mean every geoglyph has identical status or is secure.

Recording without conserving would lose future questions before they can be asked. There are reasons not to turn technical images into invitations for trespass or artifact collecting. Work with local communities, Indigenous peoples, archaeologists and heritage agencies can join research, education and protection. The forest still hides traces, but “hidden” does not make it empty land waiting for outside discoverers.

What remains beneath the canopy

Confirmed: Indigenous peoples built geometric earth enclosures in western Amazonia; records go back to the 1970s, sites have been studied in the field, and LiDAR has detected hundreds of additional structures. Strong but detailed interpretation remains open: gathering centres and a managed forest landscape. Models, not counts: tens of thousands of works and millions of inhabitants in the region defined by the 2026 authors. Open: each site’s precise use, the diversity of builders, local chronologies and the real number beneath the canopy.

Compare Iphan’s photograph with Nature’s terrain model, then read the 2017, 2023 and 2026 papers in order. Each method illuminates something different and leaves other questions dark. The deepest change is not finding one isolated “lost civilisation”: it is recognising complex Indigenous histories still present in the ground and forest even when modern interpretations failed to see them.

Open in Google Maps