Huidobro et al. (2026) Re‐Evaluating Inherited SIF Anomaly Detection Methods for the 2024 Amazon Drought: A 3D Convolutional Autoencoder Approach

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Geophysical Research Letters
• Year: 2026
• Date: 2026-08-28
• Authors: Carmen Oliver Huidobro, Gerbrand Koren
• DOI: 10.1029/2026gl122340

Research Groups

• Amazon Forest Observatory (AMFO)
• Remote Sensing Laboratory, University of São Paulo
• Department of Earth Sciences, University of Brasília

Short Summary

This study compares two anomaly detection methods for Solar‐Induced Chlorophyll Fluorescence (SIF) data to identify vegetation stress in the Amazon during severe droughts. The 3D Convolutional Autoencoder (CAE) method outperforms a classical z-score approach and aligns with drought patterns detected by SPEI and MSB.

Objective

• Investigate the effectiveness of anomaly detection methods for identifying vegetation stress using SIF data

Study Configuration

• Spatial Scale: Regional scale, focusing on the Amazon rainforest
• Temporal Scale: Long-term (2010-2024), with a focus on severe drought events

Methodology and Data

• Models used: TROPOSIF data, 3D Convolutional Autoencoder (CAE)
• Data sources: Satellite-based SIF data, Standardised Precipitation‐Evapotranspiration Index (SPEI) metric, Brazil's Monitor de Secas (MSB)

Main Results

• The CAE method identifies broader and more persistent anomalies than the z-score approach
• The CAE method aligns with drought patterns detected by SPEI and MSB, particularly during peak dry periods

Contributions

• This study highlights the importance of selecting appropriate anomaly detection methodologies for identifying vegetation stress
• The results demonstrate the potential of the 3D Convolutional Autoencoder (CAE) method for detecting drought patterns in SIF data

Funding

• This research was funded by the Amazon Forest Observatory (AMFO), grant number FAPESP 2019/05024-8
• Additional support provided by the National Council for Scientific and Technological Development (CNPq), grant number 305.456/2020-2

Citation

@article{Huidobro2026ReEvaluating,
  author = {Huidobro, Carmen Oliver and Koren, Gerbrand},
  title = {Re‐Evaluating Inherited SIF Anomaly Detection Methods for the 2024 Amazon Drought: A 3D Convolutional Autoencoder Approach},
  journal = {Geophysical Research Letters},
  year = {2026},
  doi = {10.1029/2026gl122340},
  url = {https://doi.org/10.1029/2026gl122340}
}