The central finding is stark: in the absence of human ignition, Amazonian forest fires rarely, if ever, start and spread on their own. Before people arrived in South America during the late Pleistocene, paleoecological records show little to no fire. Sediment cores from Serra Sul dos Carajás in drier eastern Amazonia lacked fire at some sites until the deglacial period and the onset of the Holocene, and where glacial-age fire did occur, charcoal abundances were far lower than in the last 10,000 years. Even fire-prone settings tell the same story. Lakes on the rainforest-savanna ecotone in Bolivia show very low and infrequent fire during the glacial period, high Andean lake sediments contain only minuscule pre-human charcoal, and a sediment core from 2,700 meters on the eastern Andean slopes of Ecuador recorded an absence of fire for 30,000 years until human activity began around 4,500 calibrated years before present.

To build the new reconstruction, the researchers compiled 1,361 accelerator mass spectrometry radiocarbon dates from 303 sites, discarding 32 modern post-1950 dates and analyzing the remaining 1,329 individually calibrated ages, which span the last 12,000 years. The dataset combined 815 dates from archaeological sites where the dated material was explicitly described as charcoal or charred, and 546 dates from charcoal fragments retrieved from soil cores, including 31 new fragments the team submitted from eight soil sites across three regions of northwestern and southwestern Amazonia. Each date was calibrated with the IntCal20 or SHCal20 curve, and the team used summed probability distributions alongside composite kernel density estimations, repeated 1,000 times on random draws of the data, to guard against overinterpreting noise and calibration-curve artifacts.

The results reveal a clear geographic trajectory for fire in the basin. The earliest fires cluster in peripheral areas and along the main Amazon River channel, mirroring the earliest archaeological sites in the region. Fire activity increased across most regions between roughly 6,000 and 4,000 years ago, a period that also saw a rise in dated occupation sites across South America and coincides with the earliest documented maize cultivation in northwestern and southwestern Amazonia. From there, human-ignited fire spread geographically, and by about 2,000 years ago it appears across all regions and deep into the interior of the forests, a timing consistent with many Amazonian societies transitioning to sedentary lifestyles reliant on cultivation and plant domestication.

Not every region followed the same script. Random mark permutation tests, which compared each region’s summed probability distribution against the basin-wide pattern 1,000 times, showed that central Amazonia lagged behind the rest of the basin, with significantly lower fire activity for most of the period from 7,000 to 2,000 years ago and only exceeding other regions in the last 1,500 years. Northwestern Amazonia showed higher activity than other regions around 4,000 to 2,000 years ago but lower activity during the last 800 years. The southwestern and central regions were even antiphased between about 1,500 and 500 years ago, with one rising as the other fell, underscoring that fire histories were not regionally synchronous.

The most striking pattern, however, emerges in the last millennium. Overall fire frequency peaked between 1,300 and 500 calibrated years before present, corresponding to roughly A.D. 700 to 1450, and then declined in a two-phase pattern. The first decline began around 700 to 500 years before present, several centuries before Europeans first traveled down the Amazon River in A.D. 1541, and appeared in every region of the basin, though not synchronously. A second wave of decline in the frequency and density of dated material followed after European arrival. The authors argue that these two waves support two distinct demographic hypotheses: an ‘early abandonment’ or major land-use change preceding contact, and the ‘Great Dying’ — the catastrophic collapse of Indigenous populations estimated at 90 to 95 percent through enslavement, disease, and warfare following colonization.

The pre-contact decline is particularly significant because it has been contested. An estimated 5 to 20 million Indigenous people lived in Amazonia before European arrival, and the Great Dying has been hypothesized to have driven large-scale site abandonment and even a measurable dip in atmospheric carbon dioxide as forests regrew and sequestered carbon. Yet a previous synthesis of lake sediment records found no signal of widespread abandonment and reforestation at the moment of European contact; instead, it documented larger changes in fire and vegetation several hundred years earlier. The new soil and archaeological dataset, with its much broader spatial coverage than lake records alone, independently confirms that the earlier signal is real, while the larger sample size now also reveals the subsequent post-contact decline. The reason for the initial pre-European abandonment or land-use change remains unknown.

Why does fire matter so much in this forest? Because Amazonian plants evolved for millions of years in landscapes where fire was essentially absent, most species lack any adaptation to survive burning. A single fire can trigger significant species turnover, and repeated burns can produce complete turnover, favoring the few lineages with some fire tolerance, such as certain palms. A recent synthesis of Holocene charcoal and pollen from lake sediments showed that vegetation responses to human activities were slow and gradual until the last 2,000 years, then accelerated sharply. The new radiocarbon synthesis captures that same acceleration, showing fire spreading across all regions during the final two millennia of the pre-Columbian era.

The scale of that transformation leads the authors to a provocative conclusion: the restructuring of Amazonian pyrogeography over the last 2,000 years has likely induced rates and magnitudes of vegetation change comparable to, or exceeding, major climatic events such as deglaciation or ecological upheavals like the extinction of Pleistocene megafauna. Because the pool of species able to return after fire remained biogeographically and evolutionarily Amazonian and essentially fire-sensitive, the overall impact on plant assemblages has been subtle — shifting the balance toward species that were already present but tolerated fire best — yet the fingerprint of millennia of human fire management is now understood to be embedded in the structure of the forests standing today.

The study also demonstrates the power of newly curated databases. The Neotoma Database houses paleoecological data from lake sediments, the Latin American Pollen Database catalogs published records, and the recently launched AmazonArch database documents more than 6,000 archaeological sites with locations and radiocarbon dates. By integrating these compilations with new soil-charcoal dates, the researchers show that the modern Amazon is not a pristine wilderness untouched by fire, but a landscape whose fire regime — and therefore whose forest composition — was shaped by Indigenous societies for millennia, then reshaped again by their demographic collapse. As modern fires increasingly penetrate standing rainforest, understanding this deep history of human-fire interaction becomes essential context for conserving what remains.

Subject of Research: Holocene fire history and human-driven pyrogeography of the Amazon basin reconstructed from radiocarbon-dated charcoal and archaeological material

Article Title: The pyrogeography of Amazonia: a Holocene perspective

Article References: McMichael, C. N. H., Heijink, B. M., Witteveen, N. H., Zwarts, A., & Bush, M. B. (2026). The pyrogeography of Amazonia: a Holocene perspective. Frontiers of Biogeography, 19, Article e169863. https://doi.org/10.21425/fob.19.169863

Image Credits: AI Generated

DOI: 10.21425/fob.19.169863

Keywords: Amazonia, pyrogeography, radiocarbon dating, charcoal, Holocene, Indigenous populations, Great Dying, early abandonment, paleoecology, fire history, archaeology, rainforest