Human influence on Amazon’s aboveground carbon dynamics intensified over the last decade

Arthur Nicolaus Fendrich, Yu Feng, Jean‐Pierre Wigneron, Jérôme Chave, Arnan Araza, Zheyuan Li, Martin Herold, Jean Pierre Ometto, Luiz E. O. C. Aragão, Isabel Martínez, Lei Zhu, Yidi Xu, Philippe CiaisNature Communications 2025 · DOI

TL;DR

This study tackles the uncertainty in Amazonian aboveground carbon dynamics by developing a novel method to disaggregate coarse-resolution satellite data, providing high-resolution annual maps of carbon stock changes from 2010-2020. The findings reveal that human activities, particularly deforestation and degradation, have intensified over the decade, leading to a net carbon loss for the Amazon, underscoring the urgent need for conservation policies.

What Problem Does This Paper Address?

The Amazon rainforest is a critical component of the global carbon cycle, storing a vast amount of carbon in its aboveground biomass. However, accurately quantifying annual changes in this carbon stock has been challenging due to the complex interplay of natural processes (like climate variability and natural disturbances) and human-induced activities (deforestation, degradation, and regrowth). Existing satellite-based methods, while valuable, often suffer from coarse spatial resolution, limiting our understanding of fine-scale dynamics and the attribution of carbon changes to specific drivers or land use types. This lack of detailed, high-resolution information hinders effective policy implementation for forest conservation and accurate reporting of greenhouse gas emissions.

Methods

The core innovation of this paper lies in its approach to disaggregate low-frequency passive L-band microwave data (specifically, Vegetation Optical Depth or L-VOD) from its native coarse resolution (~25 km) to a high resolution (~100 m). This was achieved by developing a statistical disaggregation model that leverages relationships between L-VOD and benchmark aboveground biomass carbon (AGC) datasets. The model incorporates dynamic factors like climate variables (temperature, water deficit) and land cover change data (deforestation, degradation, regrowth) to reconstruct annual AGC change maps. This approach allows for a more detailed spatial and temporal analysis of carbon dynamics, enabling the attribution of changes to human-influenced versus natural lands.

Datasets

  • L-VOD (Vegetation Optical Depth): Passive L-band microwave data from the SMOS satellite, used as the primary input for disaggregation.
  • Benchmark AGC Datasets: Multiple existing global AGC maps (e.g., Avitabile, Baccini, CCI Biomass, JPL, Zarin) were used to calibrate the L-VOD to AGC relationship.
  • Land Cover Change Data: European Commission's Tropical Moist Forests (TMF) dataset, providing high-resolution information on deforestation, degradation, and regrowth.
  • Climate Data: ERA5-Land (temperature) and CHIRPS (water deficit) datasets.
  • Validation Data:
    • Forest plot data aggregated to 0.1° spatial resolution.
  • High-resolution LiDAR data from the Brazilian Amazon.
  • Other global AGC products (JPL, CCI, WRI) for uncertainty analysis.

The paper does not introduce a new benchmark dataset but rather focuses on improving the resolution and attribution of existing data.

Key Insights and Ideas

  • Human influence is intensifying: The study clearly demonstrates that human activities, particularly deforestation and degradation, have significantly intensified over the last decade, outweighing the carbon gains from forest regrowth.
  • High-resolution mapping is crucial: Disaggregating coarse satellite data to a fine scale is essential for understanding the nuanced dynamics of carbon loss and gain, and for attributing these changes to specific drivers.
  • El Niño years exacerbate losses: Human-influenced land use practices lead to a disproportionately larger share of gross carbon losses during El Niño years, highlighting the vulnerability of these areas to climate extremes.
  • Degradation is a major contributor: Forest degradation, often overlooked, is shown to be a substantial source of carbon loss, contributing significantly to the overall negative carbon balance of the Amazon.
  • Protected areas are not immune: The analysis reveals an intensification of AGC losses within Brazilian protected areas and indigenous lands, raising concerns about the effectiveness of current conservation measures.

Key Results

  • The Amazon rainforest experienced a net loss of -0.37 ± 0.17 PgC in aboveground biomass carbon (AGC) between 2010 and 2020.
  • This net loss was driven by significant losses from deforestation (-0.55 ± 0.04 PgC) and degradation (-0.42 ± 0.08 PgC), which were only partially offset by gains from undisturbed forests (0.33 ± 0.13 PgC) and secondary forest growth (0.33 ± 0.05 PgC).
  • Losses in human-influenced land accounted for 60% of all gross losses during El Niño years.
  • Human-influenced lands were responsible for an annual gross loss of -242 ± 73 TgC year⁻¹ and a gain of 170 ± 65 TgC year⁻¹, representing 43.0 ± 11.1% and 33.5 ± 10.5% of the annual gross losses and gains, respectively.
  • The study's disaggregation method reduced uncertainty in AGC change estimates across 81.5% of the study domain.

Limitations and Future Work

  • Attribution challenges: The disaggregation model cannot definitively distinguish between direct human activity and drought-triggered tree mortality events, making it difficult to precisely attribute all AGC changes.
  • Data limitations: The accuracy of the disaggregation is dependent on the quality and resolution of the input land cover change datasets.
  • Extrapolation caution: The authors recommend using extrapolated predictions with extra care, as they are derived from areas not covered by the calibration procedure.

Future work could focus on improving the attribution of AGC changes by integrating more detailed data on specific human activities and natural disturbances. Further refinement of the disaggregation model and validation with more extensive ground-truth data would also be beneficial. The authors emphasize the need for stronger implementation of policies to control forest degradation and deforestation, supported by near real-time and trusted mapping products.

Summary updated

Abstract

The Amazon rainforest is crucial for the global carbon cycle, yet annual changes in its aboveground biomass carbon (AGC) stock remain highly uncertain. Natural and local anthropogenic drivers such as deforestation, forest degradation, and regrowth following deforestation interact with large-scale climate variability to determine AGC dynamics. Here, we propose an approach to disaggregate low-frequency passive L-band microwave data over 2010-2020 and reconstruct maps of annual change. We show that the Amazon lost -0.37 ± 0.17 PgC, with gains by undisturbed (0.33 ± 0.13 PgC) and secondary forest growth (0.33 ± 0.05 PgC) outweighed by losses by deforestation (-0.55 ± 0.04 PgC), degradation (-0.42 ± 0.08 PgC), and agricultural areas (-0.06 ± 0.03 PgC). Losses in human-influenced land intensified over time and amounted to 60% of all gross losses in El Niño years. Our study reinforces the need for stronger implementation of policies and effective actions to control forest degradation.
BibTeX
@inproceedings{fendrich2025human,
  author = {Arthur Nicolaus Fendrich and Yu Feng and Jean‐Pierre Wigneron and Jérôme Chave and Arnan Araza and Zheyuan Li and Martin Herold and Jean Pierre Ometto and Luiz E. O. C. Aragão and Isabel Martínez and Lei Zhu and Yidi Xu and Philippe Ciais},
  title = {{Human influence on Amazon’s aboveground carbon dynamics intensified over the last decade}},
  booktitle = {Nature Communications},
  year = {2025},
  doi = {10.1038/s41467-025-61856-1},
}
Figure