The world’s largest rainforest is losing its ability to bounce back from droughts and fires, pushing it farther toward a threshold where it could transform into arid savannah, releasing dangerous amounts of greenhouse gases in the process.

A study published Monday in the journal Nature Climate Change found that the Amazon has become less resilient as deforestation has continued and rising temperatures have worsened drought. The authors said the rainforest’s ability to recover from such events has diminished across three-quarters of its area in the last two decades, especially in parts that are closer to human activity, like urban areas and croplands.

“We’ve found a pronounced loss of Amazon rainforest resilience over the last 20 years,” said Chris Boulton, a climate scientist with the University of Exeter and a co-author of the report. “And when I say resilience, I mean the ability of the Amazon rainforest to restore itself back to a stable state.”

The study adds to a mounting pile of research projecting that the Amazon will reach a point when it rapidly converts into a different, drier ecosystem, although the timing of this “tipping point” remains uncertain and whether it will encompass the entire rainforest is also debated. The authors of the new study have said the change could come within decades. They were reluctant to be more specific, but said the tipping point could be sooner even than current models suggest.

“The core idea is, if the system is heading toward a tipping point, where by definition it’s getting less stable, this means that before that happens, it gets slower recovering from all these perturbations, like the drought events that are happening year to year,” said Tim Lenton, a professor of climate science at the University of Exeter and a co-author of the study.

The study report notes that there have been three “one-in-a-century” droughts in the region in recent years.

“If it gets to that tipping point and we commit to losing the Amazon rainforest, then we get significant feedback to global climate change,” Lenton said. “We’d lose about 90 billion tons of carbon dioxide, mosty from the trees and some from the soil, and that’s several years of emissions.”

That would, in turn, have “‘knock on’” effects elsewhere in the world,” Lenton said.

The authors of the study relied on satellite images taken over time to compare the forest before and after drought or fires. Using greenness as an indicator of health, they saw that the forest remained brown longer after these events, showing that it had not recovered.  They also measured naturally-occuring microwaves emitted by the plants to gauge changes in vegetation.

The loss of the forest’s ability to recover could mean that certain feedback loops could worsen. Previous research, the study noted,  has shown that fires worsen drought, which in turn makes fires more likely and could lead smaller fires to “tip” into “mega-fires.” 

Deforestation, meanwhile, leads to a loss of evapotranspiration, the process through which water moves from the earth’s surface to the atmosphere, which then leads to less rainfall. That, in turn, could have cascading effects across the Amazon basin.

“Once you tip part of the system the effects could spread to other parts of the forest,” Lenton said.  “If you start to shut down that rainfall cycle, you’re putting half of the forest at risk.”

A debate over an Amazonian tipping point stretches back two decades to the early 2000s, when researchers first said the rainforest could reach a threshold where it would convert to savannah. There is firm evidence now that this is already happening in parts of the forest, which have become net carbon emitters rather than carbon sinks.

Later research, also based on models, questioned the tipping point concept.

“There was a pendulum swing in the late 2000s that the models were just wrong, and the forest was more resilient than the modelers thought,” said Scott Denning, a climate scientist at Colorado State University.  “But now it’s swinging back into the scarier direction.”

Denning, who was not involved in the study, cautioned against the use of the term tipping point, saying it implied a wholesale, immediate loss of the rainforest.

“Our scientific debates don’t translate very well into pop culture memes like tipping points,” Denning said. “It’s sort of a mischaracterization that the scientists are certain that the Amazon is going to die.”

But Denning added that it’s very clear that the rainforest is in deep trouble, especially in the south and east, the so-called “Arc of Deforestation.”

“There’s an emerging set of different independent observations that does support that, especially along the southern and eastern margins, the forest is in decline,” Denning said. “It’s not bouncing back. It’s letting carbon out. It’s drying out. It’s dying back.”

The study’s authors emphasized that they were able to look at data over a longer period of time than was previously available, giving them a more complete picture that encompasses all the damage done to the forest in recent decades. 

“Humans have already deforested more than 20 percent of the pre-industrial rainforest,” said Niklas Boers, a co-author of the study and a researcher at the Potsdam Institute for Climate Impact Research. “The rainforest is millions of years old and, of course, rainfall regimes have changed over time.”

“But it has always survived,” Boers added. “The new component is the direct interventions that attack the ability of the rainforest to cope with the change. That’s the situation that’s unprecedented.”