A 20-year experiment in southeastern Amazonia finds that even heavily degraded, grass-invaded forests retain the capacity to bounce back from fires, but not without costs to edges, drought resilience, and species left behind.
In 2004, an international team of researchers began setting fire to three 50-hectare plots of Amazon forest in Mato Grosso, Brazil. They were asking a question that has only become more urgent in the two decades since: What happens to a tropical forest when it burns, again and again, during the worst droughts in living memory? Yale School of the Environment Associate Professor Paulo Brando has led a team of scientists to address this question.
The answer, recently published in a study by the Proceedings of the National Academy of Sciences, is complicated.
The forests did not collapse into savanna. Even after repeated burning, severe droughts, and destructive windstorms, they retained what the researchers call a “fundamental capacity to remain forests.” Once the experimental fires stopped in 2010, recovery began. Within roughly a decade, interior forests had regained much of their structural complexity and species richness.
“Our study shows that even under enormous pressure, our experimental forests have not lost their fundamental capacity to remain forests,” Brando said. “Despite dramatic collapses in biodiversity and structural complexity, they have retained the ability to recover some of their fundamental characteristics within a decade or so. That is a meaningful finding, and cautiously, yes — it is a reason for optimism.”
That optimism, however, comes with significant caveats.
The forests that grew back after the fires stopped were not the same forests that had stood before. The trees that returned fastest were generalists, species with broad geographic ranges that can survive in both forest and savanna environments, rather than the specialist species characteristic of the humid Amazonian forest. Long-lived, slow-growing species that store carbon for centuries were largely absent from the recovering plots; instead, fast-growing but short-lived species filled the canopy.
Leandro Maracahipes, a research scientist at YSE and the study’s lead author, described this process as ecological homogenization, and warned that a forest that appears recovered can mask losses beneath the canopy.
“Even when this highly degraded forest appears to have recovered, this homogenization driven by disturbance tends to result in the loss of several key species and ecosystem services,” Maracahipes said. “Some of the tree species lost through homogenization play an essential role in maintaining local wildlife, and their exclusion by disturbance also has direct impacts on the long-term persistence of fauna that depend on them. The forest is recovering, but key ecosystem services have been lost and are still being lost.”
The shift was most dramatic at forest edges — the first 200 meters bordering agricultural fields. While in interior forests recovery was relatively swift, species richness in the most heavily burned edge forest plots was still more than 50% below pre-fire levels as of 2024, 14 years after the last experimental burn.
The contrast between forest edges and interiors has implications for how scientists and policymakers think about deforestation.
“Deforestation doesn't only eliminate the forest that is removed; it also exposes the remaining forest to a fundamentally different set of conditions that make it far more vulnerable to fire,” Brando said. “The fact that edge recovery lags so far behind interior recovery is a reminder that the damage from deforestation extends well beyond the cleared area itself, with consequences that persist for many decades.”
Even forests under enormous pressure, losing key characteristics, may not be destined to collapse into something unrecognizable.”
The experimental fires explained much of what the researchers observed — but not everything. Even in the unburned control plot, never subjected to experimental burning, the researchers found compositional shifts over 20 years as new species arrived following natural disturbances.”
This finding could be a signal of broader regional pressure that fire alone cannot explain, Maracahipes noted.
“Even the control plots, which were not directly affected by fire, are changing over time with the arrival of new species following natural disturbances such as strong winds and severe droughts,” he said. “The remaining forest areas are becoming hotter and drier, with the presence of exotic grasses. As a result, forests that are naturally not very flammable can become flammable ecosystems that are highly vulnerable to future natural or human-driven disturbances.” One of the most closely watched questions in Amazon science is whether fire-degraded forests can tip permanently into savanna, an alternative stable state that, once established, could prove very difficult to reverse. The study, though, found little evidence that this occurred, at least not yet.
Biweekly, we highlight three news and research stories about the work we’re doing at Yale School of the Environment.
The study's 20-year timeframe captured something that shorter studies could not: the slow, nonlinear arc of forest recovery, but the researchers are careful not to overstate what recovery means.
“This highlights the importance of preserving and conserving intact Amazon forests,” Maracahipes said. “The forest is recovering, but it will likely require much longer than 14 years to recover all forest properties.”
The researchers emphasized that had burning continued or had fire ignition sources remained present, exotic grasses likely would have expanded across the experimental area, and consequently a human-derived savanna state could have taken hold.
“Broadly, even forests under enormous pressure, losing key characteristics, may not be destined to collapse into something unrecognizable,” Brando said. “That is the important message that pushes back against the notion of an inescapable tipping point for the Amazon. But only if we prevent the fires from recurring.”
Associate Director of Strategic Communications and Engagement
