‘Significant’ risk of Amazon forest dieback if global warming overshoots 1.5C

§ Warming pathways

As the planet warms, there is an increasing risk that parts of the Earth system will cross “tipping points” – critical thresholds that, if exceeded, could push a system into an entirely new state.

For example, a seminal 2022 study warned that five tipping elements – including the collapse of the West Antarctic ice sheet and abrupt permafrost thaw – are already within reach, while others are becoming increasingly more likely as temperatures rise.

One way to limit warming to 1.5C by the end of the century involves initially overshooting the threshold. However, research published last year warns that the longer the 1.5C threshold is breached – and the higher the peak temperature – the greater the risk of crossing tipping points.

The new study uses modelling to investigate the risks of overshoot for the Amazon and Siberian forests.

The paper considers three illustrative mitigation pathways taken from the Intergovernmental Panel on Climate Change’s (IPCC) mitigation report from its sixth assessment cycle, which was published in 2022.

Gregory Munday is an applied scientist at the UK Met Office Hadley Centre and lead author on the study. He tells Carbon brief that the authors selected “optimistic” pathways that “each have different relationships to the Paris Agreement goals”.

For each scenario, the authors assess a range of different climate sensitivities – a measure of the planet’s temperature response to a given increase in atmospheric CO2. The average outcome of each pathway is:

• The “renewables” scenario shows a future with reduced emissions and a heavy reliance on renewable energy, which keeps warming below 1.5C by 2100.
• The “negative emissions” pathway shows a world in which warming initially overshoots the 1.5C threshold, but extensive use of carbon removal sees warming drop back below 1.5C before 2100.
• The “gradual strengthening” pathway illustrates a strengthening of climate policies implemented in 2020, with rapid reductions mid-century and a reliance on net-negative emissions by the end of this century. This pathway sees global average temperatures reach 1.8C by 2100. 

The authors run the emissions pathways through a simple climate “emulator” model, which calculates the global temperatures associated with each emission pathway.

The charts below show cumulative CO2 emissions (left), atmospheric CO2 concentration (middle) and changes in global average surface temperature compared to the pre-industrial level (right), for the renewables (green), negative emissions (purple) and gradual strengthening (yellow) pathways until the year 2300.

The authors then use a different modelling framework to project the impacts of each emissions scenario. 

Study author Dr Chris Jones leads the UK Met Office Hadley Centre’s research into vegetation and carbon cycle modelling and their interactions with climate. He tells Carbon  Brief that the new study is the first application of this modelling framework, which he describes as a “rapid response tool”. 

He says the tool was developed to “rapidly look at a range of climate outcomes, both global and local, for new scenarios”, adding that it provides a “pretty good approximation” of what traditional global climate models would do.

Munday adds that the framework is able to produce results within days or weeks, rather than taking “months and months”.

Finally, the authors use land surface model JULES to assess forest health under the different scenarios. Overall, the authors produce 918 simulations each of Amazon and Siberian forest health.

§ Forest health

The authors assess forest health using two metrics. The first is the forest growth metric “net primary productivity”, a measure of the rate that energy is stored as biomass by plants, which can indicate forest productivity. The second metric, forest cover, is a way of measuring the forest’s long-term response.

The models show that rising CO2 levels causes net primary productivity to increase, due to the CO2 fertilisation effect, driving more rapid forest growth. Conversely, many of the impacts of climate change, such as increased heat and changes to rainfall patterns, can be detrimental to forests, damaging or killing trees.

To identify the impacts of overshooting 1.5C on the Amazon and Siberian forests, the authors compare the “renewables” and “negative emissions” pathways. Both of these scenarios reach a similar global average temperature by the year 2100, but the former does so without overshoot, while the latter overshoots 1.5C before temperatures come back down. 

The maps below show the difference in net primary productivity in the Amazon (left) and Siberian forests (right) between the two scenarios in the year 2100. Brown shading indicates that net primary productivity was higher in the non-overshoot scenario, while blue indicates that it was higher in the overshoot scenario. 

The maps show that “large areas of both Amazonian and Siberian forest show reduced net primary productivity” by 2100 due to overshoot, compared to a scenario with no overshoot, the paper says.

§ ‘High-risk zones’

From the three pathways, the authors generate 918 simulations of future climate and corresponding Amazon forest health. 

The authors use these results to identify which future temperature and rainfall conditions result in net forest “dieback”. This is when large numbers of trees die, shifting the rainforest into a dry savannah.

The plots below show which simulations result in Amazon dieback by the year 2100 (left) and 2300 (right), for different amounts of rainfall and temperature levels in the year 2100. Each graph is divided into four sections – hot and wet (top right), hot and dry (bottom right), cold and wet (top right) and cold and dry (bottom right). These sections are based on average regional temperature and rainfall in the year 2100.

Coloured dots indicate scenarios that see forest dieback. These are coloured by pathway, for renewables (green), negative emissions (purple) and gradual strengthening (yellow). Grey dots indicate scenarios without Amazon dieback. The red lines indicate “high-risk climatic zones”, above which there is “a significant risk of dieback”.

The study finds that most Amazon dieback scenarios happen in hot, dry conditions, the authors note.

Across all simulations where warming in 2100 is above 1.5C, 37% show “some amount of dieback” the study says. However, in these model runs, the risk increases further in the long term, the study notes, with “55% of simulations exhibiting dieback by 2300”.

Prof Nico Wunderling is a professor of computational Earth system science at the Potsdam Institute for Climate Impact Research and was not involved in the new research. He tells Carbon Brief it is significant that, according to this study, the Amazon will face impacts from climate change below the tipping point threshold of 2-6C, as assessed in the landmark 2022 tipping points paper.

The authors also carry out this analysis for Siberian forests. Instead of a drop in tree cover, they find a change in the composition of trees. Munday tells Carbon Brief that the vegetation shifts “from grassy surface types to lots more trees and shrubs” in a process called “woody encroachment”. 

Woody encroachment can have significant negative impacts on terrestrial carbon sequestration, the hydrological cycle and local biodiversity.

“The Siberian forest is probably committed to a long-term, and possibly substantial, expansion of tree cover,” the authors write.

§ High-risk scenarios

The greatest uncertainty in this study comes from the spread of climate sensitivities, Munday tells Carbon Brief.

He elaborates:

“This means that although we simulate the impacts from extremely optimistic mitigation scenarios, there is a chance that the Earth’s climate sensitivity is much higher than we expect, and so, small but significant risks of short- and long-term forest ecosystem impacts exist in spite of the choice of these strong-mitigation scenarios.”

In other words, if climate sensitivity is higher than expected, forests could face harmful impacts even under low emissions scenarios. 

Dr David McKay – a lecturer in geography, climate change and society at the University of Sussex – is the lead author of the 2022 study. He tells Carbon Brief that the new paper “shows the value in focusing not just on model averages, but also exploring a wide range of possible futures to capture potential ‘low probability, high impact’ outcomes”. He adds:

“[The study shows] how negative emissions to reduce warming might help restabilise these forests in future if we do overshoot 1.5C, but as such large-scale CO2 removal remains hypothetical, we shouldn’t assume we can rely on this in practice.”

However, McKay also notes some uncertainties in the models used. Mckay tells Carbon Brief that the vegetation model used in this study doesn’t include fire and “has some limitations around soil moisture stress and vegetation in the tundra”. These are “likely important for resolving potential tipping points in these biomes”.

Therefore, he adds, the study “doesn’t show how regional tipping points could potentially further amplify and lock-in these future forest shifts, even with negative emissions”. 

Dr David Lapola is researcher at the University of Campinas in Brazil and was not involved in the study. He also warns that vegetation models provide a “poor representation of how CO2 may affect these forests directly”. Lapola argues that scientists must “collect field data to make any new advancement with models”. 

Nevertheless, Lapola tells Carbon Brief that studies such as this will be “extremely useful” for the IPCC’s upcoming seventh assessment cycle, which will include a dedicated chapter on tipping points and other “low-likelihood high impact events” for the first time.

Study author Jones tells Carbon Brief that overshooting 1.5C leaves forest ecosystems “exposed to more risk than [they] need to be”. The findings show that “we can’t afford complacency”, he warns.