Fire may prevent future Amazon forest recovery after large-scale deforestation

The Amazon rainforest, encompassing roughly 40% of the world's tropical forests, plays a vital role in global climate stability and ecosystem services. Its vast carbon stores (approximately 10% of global forest carbon), substantial CO2 sequestration (around 5% of historical human emissions), and significant rainfall recycling (20–40%) make it a critical component of the Earth system. However, decades of human activity, including deforestation (reducing forest cover by about 20%), increasing droughts, temperature stress, logging, and slash-and-burn practices, have severely impacted the Amazon's functionality and resilience. The potential for large-scale Amazon dieback, a transition to savanna or treeless states, has been a subject of increasing concern. This transition could be accelerated by self-reinforcing feedback loops, particularly the coupling of climate change and intensified fire regimes, creating a "lock-in" effect preventing forest recovery. Similarly, the disruption of Amazonian moisture recycling through deforestation could reduce rainfall, further hindering tree regrowth. While CO2 fertilization could partially offset these negative impacts, its overall effect remains uncertain. Previous studies, often relying on remote sensing data, conceptual models, or simplified models, have addressed aspects of Amazon dieback and the role of fire, but comprehensive analyses integrating the complex interactions between fire, vegetation, and climate over multi-century timescales have been lacking. This study aims to fill this gap by using a sophisticated Earth system model to explore the long-term effects of fire on Amazon forest recovery after complete deforestation under varying atmospheric CO2 concentrations.

Existing research has extensively documented the threats to the Amazon rainforest, highlighting the roles of deforestation, climate change, and fire. Studies have shown a strong correlation between land use change and alterations in regional climate, including increased temperatures and reduced precipitation. The concept of a tipping point, where the Amazon rainforest could irreversibly transition to a savanna-like state, has been widely discussed. Various models and analyses have explored the potential for dieback, often emphasizing the self-reinforcing feedback mechanisms that could accelerate this process. The importance of fire in maintaining grassland ecosystems and its potential role in preventing forest recovery has been acknowledged, but the precise nature and extent of these interactions remain uncertain. The complexity of these interactions, including the long-term feedbacks between fire, vegetation dynamics, and climate, necessitate the use of sophisticated Earth system models that can simulate these processes over extended time horizons.

This research employed the fire-enabled Potsdam Earth Model (POEM), specifically the CM2Mc-LPJmL v1.0 configuration. This coupled Earth system model integrates a relatively coarse-resolution atmosphere and ocean model (CM2Mc) with a state-of-the-art dynamic global vegetation model (LPJmL), including the process-based fire model SPITFIRE. LPJmL simulates vegetation dynamics, including the establishment, growth, competition, and mortality of plant functional types (PFTs), considering bioclimatic limits and the impacts of productivity, heat, and fire. The model incorporates detailed biophysical couplings between the atmosphere and biosphere, exchanging variables such as canopy humidity, temperature, roughness length, and albedo. The study design involved three simulation phases: a spin-up phase to establish equilibrium, a grassland phase (250 years) simulating complete deforestation with only grass allowed to grow, and a recovery phase (250 years) allowing tree regrowth. These phases were conducted under four constant atmospheric CO2 concentrations (284, 450, 750, and 1200 ppm) with and without fire enabled. Control experiments, simulating natural vegetation without deforestation, were also performed under the same conditions. To simulate increased human ignitions expected after deforestation, the model used population density data from the Cerrado region, while lightning activity was reduced based on data from the Caatinga region. The model's spatial resolution (0.5° × 0.5°) offers a balance between detail and computational feasibility, allowing for the exploration of multi-century dynamics.

The simulations revealed a strong influence of fire on Amazon forest recovery. In experiments without fire, the forest largely recovered within 250 years, regardless of the CO2 level, except under the extreme 1200 ppm scenario. However, when fire was enabled, the results were drastically different. Forest recovery was significantly impeded, with only 40% biomass recovery at 284 ppm, 50% at 450 and 750 ppm, and 20% at 1200 ppm. Remarkably, 56–82% (353–515 million hectares) of the Amazon area remained locked in a grassland state at the end of the recovery phase, depending on the CO2 concentration. The warmer and drier conditions resulting from the grassland state increased fire activity, creating a positive feedback loop that stabilized the grassland state. The spatial pattern of forest recovery was heterogeneous, with faster regrowth in the western Amazon due to sustained moisture from the Andes. In the eastern Amazon, the re-establishment of moisture recycling was slower, contributing to a prolonged fire regime. While higher CO2 concentrations enhanced CO2 fertilization, the negative effects of increased heat stress ultimately limited forest recovery, especially at 1200 ppm. Comparing the recovery experiments with control experiments (no deforestation) demonstrated a clear bistability between forest and grassland states, highlighting the importance of initial conditions. The model showed that the same climate and fire forcing led to vastly different outcomes depending on whether the simulation started from an intact forest or a deforested grassland, emphasizing the irreversible nature of this transition under certain conditions.

The findings demonstrate the crucial role of fire in determining the fate of the Amazon rainforest after large-scale deforestation. The positive feedback loops between reduced evapotranspiration, decreased precipitation, increased temperatures, and intensified fire activity create a strong barrier to forest recovery. The bistability observed highlights the potential for irreversible transitions, where the system becomes locked into a stable grassland state, even if external conditions change. This underscores the urgent need for effective fire management strategies in conjunction with deforestation prevention to maintain the Amazon's resilience. The study’s findings provide compelling evidence for the significant contribution of fire to the potential for irreversible large-scale dieback, supporting previous concerns about the Amazon tipping point. The observed spatial heterogeneity in recovery suggests that regional differences in moisture dynamics and other factors must be considered in assessing future risks. While model limitations exist, the consistency of results across various CO2 levels strongly suggests the importance of fire dynamics in shaping the Amazon's future.

This study, using a comprehensive fire-enabled Earth system model, strongly indicates that fire is a critical factor in determining the Amazon rainforest's potential recovery after large-scale deforestation. The significant proportion of the Amazon that remains locked in a grassland state due to fire-vegetation-climate interactions highlights the irreversible nature of this transition under certain circumstances. This underscores the critical importance of preventing deforestation and mitigating climate change to safeguard the Amazon rainforest, emphasizing that future reforestation efforts alone may be insufficient without effective fire control. Future research could explore a wider range of deforestation scenarios and ignition patterns to further refine our understanding of this complex interplay and inform conservation strategies.

While this study utilizes a sophisticated coupled Earth system model, several limitations exist. The model's resolution might not fully capture the fine-scale heterogeneity of fire dynamics and vegetation patterns within grid cells. The representation of CO2 fertilization could be subject to uncertainties, and other factors not included in the model, such as nutrient availability or leaf cooling, could influence the results. The model's parameterizations and calibration to historical data introduce potential biases that could affect long-term projections. Additionally, the simplification of human ignition sources, while acknowledging model limitations, necessitates caution in directly applying the quantitative findings to specific real-world scenarios.