The Lost Cities of the Amazon: How epidemics swallowed a civilisation

The rainforest Europeans called empty was home to millions

For most of the twentieth-century, the dominant view among archaeologists and anthropologists was that the Amazon rainforest was too hostile an environment to support the kind of dense, complex societies that had arisen in the Andes or Mesoamerica.

The consensus held that the soils there were too thin and too easily exhausted by cultivation, its protein sources were too dispersed, and its ecology was too fragile to absorb the sustained pressure of large settled masses of people. Ergo, the theory went, the Amazon could only have been sparsely populated by various small mobile groups.

Of course, the Spanish conquistador Francisco de Orellana had sailed down the Amazon in 1542 and reported seeing cities stretching for miles along its banks – inhabited by millions of people and organised into complex chiefdoms with roads, plazas, and agricultural systems. But scholars dismissed his account as fantasy or self – the exaggeration of a man seeking to justify an expedition that had found no gold and needed another kind of glory. The jungle was pretty much empty. It had always been pretty much empty.

This consensus has collapsed. Over the past three decades, and with accelerating momentum in the past 15 years, a convergence of archaeological discoveries, remote sensing technology, soil science, and ancient DNA analysis has revealed that the Amazon basin was home before European contact to a population of extraordinary size and sophistication, organised into settlements of genuine urban scale, connected by road networks, and sustained by agricultural systems of remarkable ingenuity.

The forest Europeans found empty was empty for the sole reason that smallpox, measles, and influenza killed somewhere between 90 and 95 per cent of the indigenous population within a century of the conquistadors’ – reducing a civilisation of millions to scattered remnant communities before most European explorers had penetrated far enough inland to observe what they were destroying. Orellana told the truth, as one of the last outsiders to see the Amazon as it actually was, in the final years before catastrophe made its former reality invisible.

The soil was the first clue, and it remains the most compelling physical evidence for what the Amazon once contained. Scattered across millions of hectares of the basin, and concentrated particularly in the central Amazon around the confluence of the Rio Negro and the Amazon River itself, lie deposits of a remarkable material called terra preta – Portuguese for dark earth.

Whereas the natural soils of the Amazon are the colour of rust, thin, acidic, and largely infertile, terra preta is deep black, rich in charcoal, organic matter, and nutrients, and extraordinarily productive even after five centuries without human management. Nature does not produce it naturally. Human communities created this anthropogenic soil deliberately by mixing charcoal, bone, manure, and organic waste into the existing earth over generations of sustained habitation.

The charcoal giving terra preta its colour and fertility is the key to understanding how it was made and what it implies. Unlike the slash-and-burn agriculture of later Amazonian communities, which burns vegetation completely and deposits ash that washes away within a season or two, the communities that created terra preta used a process of low-temperature smouldering – essentially charring rather than burning – to produce stable carbon compounds improving the soil’s structure and nutrient retention. This intentional agricultural technology transformed the infertile soils of the Amazon basin into land capable of supporting permanent, dense settlement.

Brazilian archaeologist Anna Roosevelt first systematically investigated the settlements associated with terra preta deposits in the 1990s, when her excavations at Marajó Island at the Amazon’s mouth revealed evidence of a complex, stratified society of considerable antiquity, including the uncovering of ceramic traditions and monumental earthworks.

But the most dramatic early evidence for pre-Columbian Amazonian urbanism emerged in Brazil’s upper Xingu region. Working from clues in the accounts of a German ethnologist named Karl von den Steinen – who had visited the region in the 1880s and found indigenous people maintaining what they described as the remnants of a much larger ancestral civilisation – the American archaeologist Michael Heckenberger began systematic excavation in the Xingu in the late 1990s in collaboration with the Kuikuro people whose ancestors had built what he found.

Heckenberger and his colleagues documented a network of settlements connected by roads and organised around a consistent planning principle; he called them garden cities. These were intricately planned settlements, each arranged around a central plaza of consistent orientation, connected to satellite towns by roads up to 50 metres wide, and surrounded by palisades, moats, and agricultural zones indicating both defensive awareness and sustained landscape management across centuries. The builders engineered and maintained the roads between settlements as deliberate features, instead of paths worn by repeated use – implying a political authority capable of organising labour beyond the scale of individual communities.

Heckenberger identified at least 20 major settlements in the Xingu network, estimating a total population for the region of 50,000 people at minimum – while acknowledging that the true figure was probably considerably higher and that the surveyed area is only a fraction of the territory the culture occupied.

The implications of the Xingu discoveries were were dwarfed in scale by what remote sensing technology began to reveal across the wider Amazon basin from the early 2000s onward. Lidar – Light Detection and Ranging, a technology that fires laser pulses from aircraft and measures their return times to construct precise three-dimensional maps of the ground surface – transformed Amazonian archaeology in the same way it transformed the archaeology of Angkor in Cambodia and the Maya lowlands in Central America.

By penetrating the forest canopy and mapping the ground beneath it, lidar revealed earthworks, roads, and settlement patterns invisible to ground-based survey and undetectable from conventional aerial photography. The results, accumulating rapidly through the 2010s, have been extraordinary in their scale and in their implications for the understanding of pre-Columbian Amazonian civilisation.

In the Brazilian state of Acre next to the Bolivian border, lidar and ground survey have revealed over 450 geometric earthworks – circles, squares, and composite shapes defined by ditches up to eleven metres deep and banks up to four metres high – distributed across an area of 13,000 square kilometres. Indigenous groups constructed these structures, known as geoglyphs, between approximately 1000 BC and 1500 AD – and the effort was a sustained programme of landscape modification pointing to an advanced level of political organisation, shared cosmological principles, and a relationship with the land extending over two millennia of continuous cultural engagement.

In the Bolivian Amazon, the Llanos de Mojos – a seasonally flooded savannah region long assumed to have supported only sparse populations – has yielded lidar evidence for raised field agriculture covering millions of hectares, connected by a network of canals and causeways, associated with settlement mounds of considerable size.

The most dramatic single discovery of recent years came in January 2024, when researchers analysing lidar data from a region of the upper Amazon in Ecuador identified a network of settlements associated with the Upano culture, which flourished between roughly 500 BC and 600 AD. The Upano sites include a central urban area covering at least 300 hectares, connected by more than 2,500 roads and linear earthworks to a network of smaller communities across a surveyed area of roughly 300 square kilometres.

The central settlement contains evidence of monumental platform construction, public plazas, and the organised spatial planning characteristic of urban centres across the ancient world. The researchers estimated the total population of the Upano network at somewhere between 10,000 and 15,000 at minimum, with the caveat that the surveyed area is only a fraction of the region the culture occupied. This discovery pushed back the timeline for complex Amazonian urbanism by a millennium beyond previous estimates, confirming that the phenomenon extended into the deep roots of the basin’s history.

The Upano sites, particularly the Kunguints and Sangay complexes, exhibit a level of planning matching the great urban centres of the Old World. Builders constructed more than 6,000 earthen platforms, measuring up to 20 metres by 10 metres, arranged around rectangular plazas. These structures supported timber buildings and domestic hearths. A sophisticated drainage system of deep longitudinal ditches diverted the heavy Andean rainfall away from the living areas, while dug-out roads, some sinking several metres below the surrounding ground level, carved a permanent geometry into the clay.

In the Bolivian Llanos de Mojos, the Casarabe culture built massive water-control systems comprising hundreds of kilometres of canals and causeways. These features allowed for the intensive cultivation of maize, manioc, and squash in a landscape that cycled between extreme drought and total inundation. Again, the sheer volume of moved earth – millions of cubic metres – signifies a coordinated workforce directed by a central authority.

These Casarabe also managed vast “fish weirs”, zigzagging structures built across floodplains to trap migrating fish, providing a stable protein source that supported permanent, high-density populations throughout the entire calendar year.

The ecological implications of pre-Columbian Amazonian civilisation extend beyond archaeology into the contemporary understanding of the forest itself. The Amazon functions as a human artefact rather than a natural ecosystem that human beings inhabited without transforming. Studies of forest composition have found that tree species useful to humans – Brazil nuts, cacao, açaí palms, and dozens of others – occur at dramatically higher densities near archaeological sites and along ancient road corridors than in areas of the rainforest with no known occupation history. So the distribution of useful species in the Amazon reflects the planting decisions of people who lived there for thousands of years, managing the forest as an agricultural landscape instead of just foraging within it. The wild Amazon of European imagination was a garden whose gardeners had been killed before anyone thought to ask them how it worked.

The human cost of this civilisation’s destruction is difficult to hold in the mind. Recent scholarship estimates the population of the Amazon basin at the moment of European contact at between 8 and 50 million people – with most serious research converging on figures toward the upper end of this range.

Within a century of contact, disease had reduced this population by a proportion that has no parallel in recorded history outside of deliberate genocide. The speed and completeness of the demographic collapse meant that many European explorers who arrived in the generations after initial contact encountered a depopulated landscape and concluded that it had always been empty – that the forest was a wilderness that had never supported human complexity. They mistook the graveyard of vast masterworks of human endeavour for the natural state of the rainforest.