The Amazon rainforest may be able to survive long-term drought caused by climate change, but adjusting to a drier, warmer world would exact a heavy toll, according to a new study.
The findings show adapting to the effects of climate change could see some parts of the Amazon rainforest lose many of its largest trees, releasing carbon stored in them to the air, and reducing the rainforest's carbon sink capacity.
Parts of the Amazon are expected to become drier and warmer as the climate changes, but long-term effects on the region's rainforests - which span more than 2 million square miles - are poorly understood.
Previous research has raised concerns that a combination of severe warming and drying, together with deforestation, could lead to lush rainforest degrading to a sparser forest or even savanna.
Now, findings from the world's longest-running drought study in tropical rainforest has revealed some of the profound changes the Amazon could undergo in a drier world.
Over a 22-year period, a one-hectare area of rainforest in north-eastern Amazonian Brazil - roughly the size of Trafalgar Square - has been subjected to long-term drought conditions.
The experiment began in 2002, with thousands of transparent panels installed above the ground to redirect roughly half of the rainfall to a system of gutters, taking it away from the trees.
Analysis by an international team of scientist shows most of the study area's largest trees died during the first 15 years of the experiment, after which the forest stabilised.
Their findings show seven years after the initial biomass losses the availability of water increased for the surviving trees. Tests on these remaining trees showed they were now no more drought-stressed than those in nearby rainforest not subjected to drought.
Dr Paulo Bittencourt, one of the study's co-authors from Cardiff University, said: "Researchers around the world, including us, were trying to understand if the Amazon Forest might become more resilient to future drought by the trees themselves acclimating to increased drought stress, which is not supported by the available data.
"In this study we found that after large tree mortality, the surviving trees had less competition for water."
While everyone was looking for Amazon forest resilience to arise from adjustments in tree physiology, our work shows Amazon Forest resilience is emerging from an ecosystem level feedback - less trees means more water is available for those remaining, leading the forest to not collapse but stabilise in a new state with less biomass.
Overall, the area lost more than one-third of its total biomass - the trunks, branches and stems and roots where carbon is stored in living vegetation.
Having lost carbon through excess tree deaths during the first 15 years of the study, surviving trees in the area are now making slight carbon gains, the team says.
While the study area has less woody biomass than normal rainforests in the Amazon, it still has more than many dry forests and savannas. This indicates the rainforest has some long-term resilience to the drier conditions it could experience due to climate change, but at a high cost.
Lead author Dr Pablo Sanchez Martinez of the University of Edinburgh's School of GeoSciences, added: "Our findings suggest that while some rainforests may be able to survive prolonged droughts brought on by climate change, their capacity to act as both a vital carbon store and carbon sink could be greatly diminished."
The amount of biomass the Amazon could lose, and the time required for it to stabilise, may be underestimated, as the study only assessed the effects of soil drought, the team warns.
Further research is needed to assess other likely impacts, such as changes to moisture in the air, temperature and the compounding effects of other climate-related factors such as storms or fires, they add.
Their study, published in the journal Nature Ecology and Evolution, was led by the University of Edinburgh, Universidade Federal do Pará and the Museu Paraense Emílio Goeldi, Brazil, and in collaboration with Cardiff University, the University of Exeter and CREAF in Spain.
The research was supported by the Natural Environment Research Council (NERC), the Royal Society and the UK Met Office Newton Fund.
