Chen et al. (2026) Geoinformation-explicit retrieval of canopy live and surface dead fuel moisture from Sentinel-1 SAR data via a physics-guided machine learning framework

Identification

• Journal: Remote Sensing Applications Society and Environment
• Year: 2026
• Date: 2026-09-22
• Authors: Zhenyu Chen, Cui Zhou, Zhiwei Liu, Junxiang Zhang, Jianjun Zhu, Zengyuan Li
• DOI: 10.1016/j.rsase.2026.102282

Research Groups

• School of Advanced Interdisciplinary Studies, Central South University of Forestry and Technology (China)
• School of Geosciences and Info-Physics, Central South University (China)
• Research Institute of Forest Resource Information Techniques, Chinese Academy of Forestry (China)

Short Summary

This paper proposes a physics-guided machine learning framework for retrieving canopy live and surface dead fuel moisture from Sentinel-1 SAR data. The framework significantly outperforms existing methods in accuracy.

Objective

• Investigate the feasibility of using Sentinel-1 SAR data to retrieve forest fuel moisture content (FFMC) across vertical strata, including canopy live fuel moisture content (LFMC) and surface dead fuel moisture content (DFMC).

Study Configuration

• Spatial Scale: Regional-scale forest fuel moisture monitoring in China.
• Temporal Scale: Monthly fuel moisture maps for 2024.

Methodology and Data

• Models used:
  • Polarimetric Decomposition
  • Semi-empirical Water Cloud Model (WCM)
  • Machine learning techniques:
    • Random Forest for LFMC
    • Multilayer Perceptron for DFMC
• Data sources: Sentinel-1 dual-polarization SAR data, field measurements from 33 sites in California, and auxiliary geoinformation such as leaf area index, slope, and aspect.

Main Results

• The proposed framework significantly outperforms the WCM alone in accuracy.
• Normalized root-mean-square error (NRMSE) for LFMC decreases from 17.84% to 14.62% (R2 = 0.75).
• NRMSE for DFMC decreases from 31.17% to 23.83% (R2 = 0.72).

Contributions

• The proposed framework offers a practical, data-driven solution for regional-scale forest fuel moisture monitoring.
• It supports adaptive forest management and wildfire early warning under climate change.

Funding

• This research was funded by the National Natural Science Foundation of China (Grant No. 52177021).
• The authors acknowledge the support from the Chinese Academy of Forestry.

Citation

@article{Chen2026Geoinformationexplicit,
  author = {Chen, Zhenyu and Zhou, Cui and Liu, Zhiwei and Zhang, Junxiang and Zhu, Jianjun and Li, Zengyuan},
  title = {Geoinformation-explicit retrieval of canopy live and surface dead fuel moisture from Sentinel-1 SAR data via a physics-guided machine learning framework},
  journal = {Remote Sensing Applications Society and Environment},
  year = {2026},
  doi = {10.1016/j.rsase.2026.102282},
  url = {https://doi.org/10.1016/j.rsase.2026.102282}
}