1.6.2.2 Biosphere: Amazon rainforest dieback

Amazon dieback as a tipping point is observed in some modelled climate change scenarios (1.3.2.1). One such study shows that temporal EWS, such as increases in AR(1) and variance, are not necessarily good indicators of Amazon dieback in a number of HadCM3 GCM runs (Boulton et al., 2013). This is most likely because the Amazon is forced too fast and non-linearly for these statistical measures to work. Because of this, a system-specific indicator was suggested, looking at the sensitivity of ecosystem productivity anomalies to temperature changes, and then as a real-world measurable signal, the sensitivity of atmospheric CO₂ anomalies to these temperature anomalies. Both of these indicators worked well across the ensemble of runs.

Further work observes an increase in drying in the Amazon region across the recent CMIP6 model suite (Ritchie et al., 2022). An increase in the sensitivity of the temperature seasonal cycle amplitude to global warming is observed to be more prominent in locations that subsequently experience abrupt dieback shifts. The increasing sensitivity of the temperature seasonal cycle amplitude to global warming, therefore, has the potential to be used as a system-specific EWS for future dieback in the Amazon rainforest (Parry et al., 2022).

Real-world observational data has shown different results regarding the generic indicators discussed in this chapter, particularly the use of vegetation optical depth (VOD), a remotely sensed product that is strongly correlated with the amount of water content in the trees. Using this, increases in AR(1) and variance particularly since the early 2000s have shown a loss of resilience in the Amazon rainforest (Boulton et al., 2022). Using this same dataset, while modelling the water recycling network across the region (1.3.2.1), a network approach shows similar losses of resilience (Blaschke et al., 2023).