Loss of Amazon Rainforest Resilience: Abstract and Introduction

Table of Links

• Abstract and Introduction
• Materials and Methods
• Data
• Results
• Conclusions, Open Research Section and Conflict of Interest/Competing Interests
• Author Contributions, Supplementary Information, Acknowledgments and References
• Additional Insight from the Conceptual Model
• AMSR2 until 2022 Excluded Due to Missing Human Land Use Data
• Investigation of Resilience Loss in AMSR2’s X-Band
• Results Proving Robustness
• Potential Driving Forces and Sources of Bias

Abstract

The Amazon rainforest (ARF) is threatened by deforestation and climate change, which could trigger a regime shift to a savanna-like state. Previous work suggesting declining resilience in recent decades was based only on local resilience indicators. Moreover, previous results are potentially biased by the employed multi-sensor and optical satellite data and undetected anthropogenic land-use change. Here, we show that the spatial correlation provides a more robust resilience indicator than local estimators and employ it to measure resilience changes in the ARF, based on single-sensor Vegetation Optical Depth data under conservative exclusion of human activity. Our results show an overall loss of resilience until around 2019, which is especially pronounced in the southwestern and northern Amazon for the time period from 2002 to 2011. The demonstrated reliability of spatial correlation in coupled systems suggests that in particular the southwest of the ARF has experienced pronounced resilience loss over the last two decades.

INTRODUCTION

The Amazon rainforest (ARF) is the most biodiverse region of our planet, and serves as a major carbon sink (Malhi et al., 2008; Gatti et al., 2021). Yet, the efficiency of its carbon uptake has been declining over the last decades (Malhi et al., 2008; Gatti et al., 2021; Brienen et al., 2015; Hubau et al., 2020), with the ARF becoming carbon neutral and even acting as a carbon source during the two one-in-a-century droughts in 2005 and 2010 (Phillips et al., 2009; Gatti et al., 2014; Feldpausch et al., 2016). The ARF’s important role in the global carbon cycle thus means that its existence and stability are crucial for climate change mitigation, especially as the planet continues to warm in the later parts of the century (IPCC, 2022).

Studies suggest that there is a critical mean annual precipitation (MAP) value at which parts of the forest might irreversibly transition into a savanna-like state (Hirota et al., 2011; Zemp et al., 2017a). In such a scenario, forest dieback would likely be self-amplifying, i.e. the non-linearity of such an abrupt transition would result from positive feedback mechanisms operating in the region. Besides fire (Brando et al., 2014), the main feedback mechanism that could amplify dieback in the ARF is related to moisture recycling (Staal et al., 2018; Salati et al., 1979). Moisture is transported at low atmospheric levels via the trade winds from the tropical Atlantic to the Amazon basin, where it precipitates. A substantial fraction is taken up by the vegetation and transpired back to the atmosphere, or evaporates from the complex surfaces of plants. This evapo-transpirated water is then transported further west and south over the Amazon and towards the Andes by low-level jets, sometimes called atmospheric rivers (Gimeno et al., 2016). The low-level circulation itself is amplified by condensational latent heating over the Amazon basin, strengthening the large-scale atmospheric heating gradient between ocean and land (Boers et al., 2017).

Two main mechanisms have been proposed that may activate positive feedback cycles and push the ARF towards a critical threshold (Lovejoy and Nobre, 2018, 2019). On the one hand, anthropogenic global warming will cause increased temperatures over the Amazon basin, which could lead to increased evapo-transpirative demand without a corresponding increase in water supply via precipitation, especially during a potentially intensifying and prolonging dry season (Malhi et al., 2009) and severe droughts (Vogel et al., 2020). This could additionally lead to decreased convection and a reduction of moisture inflow from the Atlantic (Pascale et al., 2019); moreover, models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) project an overall drying in tropical South America in response to increasing atmospheric greenhouse gas concentrations. Hence, global warming could drive the system towards destabilization (Zemp et al., 2017a). Furthermore, deforestation can lead to a critical decrease of evaporated moisture transported downstream, and to an additional reduction of moisture inflow due to a decreased heating gradient, further pushing the ARF toward a critical threshold (Boers et al., 2017; Zemp et al., 2017b). The decrease in precipitation that would occur beyond such a threshold would also cause degradation outside the ARF region.

Such vegetation carbon losses have been predicted by several CMIP6 models in parts of the Amazon basin, preceded by an increasing amplitude of seasonal temperatures (Parry et al., 2022). Furthermore, observations have shown that the Amazonian dry season is increasing in length (Marengo et al., 2017; Leite-Filho et al., 2019; Phillips et al., 2009), exacerbated by the three severe droughts that have occurred since 2005 (Feldpausch et al., 2016). In view of these projected and observed trends and the global relevance of the ARF, monitoring changes in its resilience is of great importance.

As the data-driven monitoring of resilience changes and the anticipation of critical transitions is important for many parts of our climate, including the ARF, a considerable amount of research has focused on developing and applying such methods. Existing methods focus on the detection of distinct signs of resilience loss, where resilience is defined as a system’s ability to recover from perturbations; the underlying mathematical concept is derived from dynamical system theory. Under the assumption that resilience loss can be dynamically represented by an approaching (codimension-1) bifurcation, the approach to the critical forcing value at which the bifurcation is accompanied by a weakening of the equilibrium restoring forces in the system, resulting in slower recovery from small perturbations. This is termed ‘critical slowing down’ (CSD). During CSD, the variance and lag-1 autocorrelation (AC1) of the system state increase, as they are directly linked to the recovery rate from perturbations (Scheffer et al., 2009; Dakos et al., 2008, 2012; Boers et al., 2022). Yet, in the case of spatio-temporal data, calculating the variance and AC1 of individual point locations does not exploit the information that is potentially encoded in the spatial dimensions, and in particular misses the interactions between different locations. In Dakos et al. (2010) the authors argue that when a system consists of several coupled units, a decrease in the units’ recovery rates causes increasing correlation between two coupled cells.

Variance and AC1 have recently been confirmed to quantify resilience empirically at global scale, by comparing theoretical estimates of the recovery rates based on classical CSD indicators to direct estimates of the recovery rate from VOD time series sections showing recovery from abrupt disturbances (Smith et al., 2022). Recent results in Boulton et al. (2022) have revealed that large parts of the ARF’s vegetation biomass show increasing AC1, implying a loss of resilience that has been especially pronounced since the early 2000s. The studies in Smith et al. (2022) as well as Boulton et al. (2022) used multi-satellite data to extend the period of time under study (Moesinger et al., 2020). Yet, Smith et al. (2023) subsequently showed that such merging of sensors can create spurious changes in higher-order statistics such as variance and AC1 when those are calculated from multi-instrument time series. The non-stationarity induced in the time series by the change of sensors with different orbits, intrinsic noise, and radiometric resolutions may result in statistical signals which can be misinterpreted as resilience changes. It is thus recommended to investigate the resilience changes in a system by using single-sensor instrument records when available. In this work we thus exclusively use singlesensor data: in particular, Amazon vegetation indices based on Vegetation Optical depth (VOD), see Fig. 1. While indicators of CSD could react differently to changes in the measurement process or to systemic changes that are not related to CSD, different indicators that are related to CSD via the recovery rate should behave consistently whenever resilience changes. We thus ensure the robustness of our results by comparing the calculated spatial correlation to corresponding estimates of the AC1 and variance. To further ensure robustness, we compare different data sources as recommended in Samanta et al. (2012).

It has been suggested that different areas of the ARF can be considered part of a coupled system connected by spatial interactions via evapo-transpiration, low-level winds, and moisture recycling (Staal et al., 2018). Since the trade winds transport moisture from east to west, we assume that coupling cells uni-directionally is a reasonable simplification of the plant-water moisture transport system in the Amazon (Boers et al., 2017). The almost laminar flows can be investigated as separate trajectories, thereby allowing a reduction to one average dimension. Based on this relationship, we set up a simple conceptual model with asymmetric interaction

(see Fig 2.) to validate variance, AC1, and spatial correlation as indicators of CSD and, hence, resilience loss in ARF vegetation. We then calculate all three resilience indicators for four singlesensor VOD satellite data sets, which quantify the ARF’s vegetation proxy of above-ground water content in biomass. Finally, based on these results we discuss changes in resilience of the ARF.