Global forests are expected to contribute significantly to climate change mitigation efforts outlined in the 2015 Paris Agreement, primarily through limiting deforestation and promoting forest regrowth. The Brazilian Amazon, the world's largest continuous tropical forest, plays a crucial role in this global effort. It stores a substantial amount of global forest carbon and historically acted as a significant carbon sink, offsetting deforestation emissions. However, deforestation rates have fluctuated, posing a risk to this critical carbon sink. Brazil's Nationally Determined Contributions (NDCs) under the Paris Agreement commit to restoring and reforesting millions of hectares by 2030, a goal partially achievable through the natural regeneration of secondary forests. Secondary forests, regrowing on abandoned deforested land, have shown impressive carbon sequestration rates, potentially exceeding those of old-growth forests. Previous studies have provided estimates of this carbon uptake, but these have limitations in capturing the spatial variations in regrowth rates due to varying environmental and anthropogenic factors. These drivers include factors like shortwave radiation, precipitation, soil fertility, fire, and repeated deforestation cycles, all influencing the amount and location of carbon sequestration. These are not static; they affect both the rate and magnitude of carbon sequestration, therefore impacting the overall contribution of secondary forests to climate change mitigation. The research employs remote sensing data to analyze and quantify these spatial variations and to assess the future carbon sequestration potential of secondary forests in the Brazilian Amazon.

Existing literature highlights the significant carbon sequestration potential of secondary forests in the tropics. Studies show that these forests can accumulate carbon at rates much higher than old-growth forests, potentially up to 20 times faster. This rapid carbon uptake has made them a focus of research and conservation efforts. However, these initial estimates have generally been based on limited field data and haven't fully captured the regional variations and the influence of various environmental and human-induced disturbances. Previous works have touched upon some of these drivers, such as water deficit and soil fertility, demonstrating that regional differences matter. The influence of disturbances like fire and repeated deforestation cycles on carbon sequestration in secondary forests needs more systematic investigation. Existing studies provide a foundation but lack the detail to fully understand and accurately predict the carbon sequestration rates. Several studies have analyzed the carbon stocks and dynamics in Amazonian forests, including the influence of climate variability and deforestation. However, there is a need for spatially explicit analysis that integrates environmental drivers and disturbance factors to improve estimates of secondary forest regrowth and carbon sequestration.

This study uses a spatially explicit approach, leveraging the power of remote sensing data to analyze the carbon sequestration potential of secondary forests in the Brazilian Amazon. It integrates multiple datasets with high spatial resolution to achieve this goal. Firstly, it identifies secondary forests and their ages using the MapBiomas Collection 3.1 land-use and land-cover product. This product provides annual land cover maps at 30m resolution spanning 1985-2017, allowing tracking of deforestation and regrowth events. A methodology similar to past studies was implemented to define and classify secondary forests based on transitions between land uses. To address misclassifications in MapBiomas, the study incorporates data from the TerraClass-2014 dataset, improving the accuracy of secondary forest identification. Aboveground carbon (AGC) was estimated for 2017 using the European Space Agency Climate Change Initiative (ESA-CCI) Aboveground Biomass product at 100m resolution. The AGC was derived by converting biomass to carbon using a 2:1 ratio. The study analyzed the influence of six key drivers on AGC accumulation, including both environmental and anthropogenic factors. Environmental drivers encompass shortwave radiation, precipitation, and forest water deficit and soil fertility. Anthropogenic disturbances are represented by fire occurrences and the frequency of past deforestation events. All these drivers have varying spatial resolutions and were resampled to the 30m resolution of the secondary forest map. The study uses the Chapman-Richard model to represent AGC accumulation as a function of forest age and the different drivers. This model uses AGC data from old-growth forests, extracted from ESA-CCI, as asymptotes, which is a value of the AGC that is being reached. A bias correction was done on median AGC values to adjust for the variations in the values by subtracting the smallest median value from all values. To determine the relative importance of each driver, a conditional random forest model was implemented, and then the importance ranking was used for each of the driver to better account for correlations between the variables. The study further classifies Amazonia into four regions based on climatic conditions and disturbance patterns, creating region-specific regrowth models to capture spatial heterogeneity. These models are validated with field AGC data from 284 samples across 33 locations. Finally, the study estimates the 2017 carbon stock and projects future carbon sinks under different preservation scenarios to assess the contribution of secondary forests to Brazil's NDC goals.

The study reveals significant spatial variation in secondary forest regrowth rates and carbon sequestration across the Brazilian Amazon. Western regions exhibit substantially higher regrowth rates (3.0 ± 1.0 Mg C ha⁻¹ yr⁻¹) compared to eastern regions (1.3 ± 0.3 Mg C ha⁻¹ yr⁻¹). Shortwave radiation emerges as the most critical environmental driver, with lower radiation areas exhibiting faster regrowth. Maximum Cumulative Water Deficit (MCWD) is another significant factor, showing that areas with less water deficit have better carbon accumulation. Anthropogenic disturbances, particularly fire, substantially reduce regrowth rates. Fire alone can lead to declines of up to 75% in carbon accumulation. Repeated deforestation shows a less significant impact overall but still leads to a decrease in regrowth rates. The study identifies four distinct climate regions in the Amazon, each showing unique regrowth patterns and importance rankings of drivers. The North-West region, with high precipitation, low radiation, and minimal water deficit, presents the highest regrowth rates, less affected by disturbances. The North-East and South-East regions have lower overall regrowth rates, more significantly affected by fires and deforestation. The analysis finds the total carbon stored in Amazonian secondary forests in 2017 to be 293.7 Tg C. The potential stock, had no disturbances occurred, would have been 319.7 Tg C, indicating an 8% reduction due to disturbances. Modeling future carbon sequestration under various preservation scenarios reveals a substantial potential contribution to Brazil's NDC. Maintaining the 2017 secondary forest area could accumulate 19.0 Tg C yr⁻¹ by 2030, representing a significant reduction in net emissions. However, if only forests older than 20 years are preserved, the potential is drastically reduced. Comparison with previous models reveals that the approach used in this study is more accurate in assessing secondary forest regrowth, potentially correcting for past overestimation.

The findings address the research question by providing a spatially explicit analysis of secondary forest regrowth, demonstrating the significance of both environmental and anthropogenic factors. The substantial difference in regrowth rates between western and eastern regions emphasizes the importance of regional context. The dominant role of shortwave radiation highlights the importance of understanding microclimatic variations and their influence on carbon sequestration. The significant negative impact of anthropogenic disturbances, particularly fire, underscores the need for effective fire prevention and management strategies. These results are highly significant for the field, as they refine estimates of the carbon sink potential of secondary forests and offer a new framework for spatial prioritization of forest restoration and conservation efforts. The high-resolution spatial analysis distinguishes the study from prior research and clarifies the complex interaction of drivers influencing regrowth. The model validation and comparison with existing estimations strengthen the findings and improve understanding of secondary forest carbon dynamics. The quantitative projection of future carbon accumulation provides valuable input for policymakers in Brazil and informs discussions around climate change mitigation and land-use management.

This research provides a novel spatially-explicit assessment of the carbon sequestration potential of secondary forests in the Brazilian Amazon, demonstrating substantial regional variation influenced by both environmental and anthropogenic factors. The findings highlight the critical need for targeted conservation and restoration strategies that account for regional differences and minimize human-induced disturbances. The results contribute significantly to carbon modelling and policy by providing high-resolution maps and models of regrowth rates that can inform future land-use planning and emission reduction targets. Future research should focus on incorporating additional factors, such as proximity to other forests and the specific land-use history before abandonment. This study provides a valuable methodological framework for analyzing secondary forest dynamics that could be easily adapted and applied in other regions globally.

The study's limitations primarily stem from the inherent uncertainties and limitations of the remote sensing data utilized. While efforts were made to mitigate issues such as misclassifications and spatial resolution differences, some inaccuracies might still exist. The space-for-time substitution methodology, relying on the age distribution of standing secondary forests in 2017, assumes consistent growth conditions over time. This could be challenged by unforeseen disturbances and climate change. The temporal scope of the study, covering 32 years of secondary forest data and one year of AGC data, also limits the long-term projection of the models. The model's accuracy in projecting far beyond the available data might be reduced. Additional variables such as species composition and proximity to other forests, while relevant, were not explicitly included in this study due to data availability. Despite these limitations, this study significantly advances our understanding of secondary forest dynamics.