EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Quantifying biomass saturation in planted forests: integrating active and passive remote sensing with multi-dimensional structural features

Rula Sa, Wenyi Fan

Frontiers in Forests and Global Change · 02.07.2026 · 9

DOI: 10.3389/ffgc.2026.1873077 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Forests and Global Change, online erschienen am 02.07.2026. Verfasst von Rula Sa und 1 weiteren Autor:innen. Die Arbeit nennt 137 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Signal saturation in optical and synthetic aperture radar (SAR) data remains a critical challenge for accurate aboveground biomass (AGB) estimation in high-biomass forests. To address this, we developed a multi-source fusion framework integrating Landsat-8, Sentinel-2A, and ALOS-2 PALSAR data over the artificial coniferous forests of the Saihanba Forest in Hebei Province, China. The framework synergistically combines horizontal structure information from optical data with vertical structure information from SAR data, together with environmental factors. It systematically integrates three variable selection methods [Pearson correlation, Random Forest importance, and the least absolute shrinkage and selection operator (LASSO)] and three machine learning models (Random Forest, Support Vector Regression, Artificial Neural Network), evaluated under multiple cross-validation (CV) strategies (leave-one-out, repeated 10-fold, and repeated 5-fold). A novel method for quantifying the saturation point of combined variables is also introduced. Results demonstrate that horizontal structure explains AGB significantly better than vertical structure, and the proposed composite structural index simultaneously elevates the saturation point and improves estimation accuracy. The LASSO-ANN combination consistently achieves high performance across all three CV strategies (mean RMSE = 12.87 t/ha, MAE = 9.28 t/ha), attaining optimal accuracy under leave-one-out CV (RMSE = 5.14 t/ha, MAE = 2.44 t/ha) with a saturation point of 218.54 t/ha, effectively mitigating the saturation issue in medium-to-high-density coniferous forests. Coordinated optimization of feature selection, model architecture, validation strategy, and response variable is essential for improving AGB estimation accuracy and delaying saturation. The proposed framework provides a robust pathway for regional-scale forest biomass monitoring and carbon stock assessment, advancing the integration of multi-source remote sensing in carbon cycle research.

Zitieren

APA

Sa, R., & Fan, W. (2026). Quantifying biomass saturation in planted forests: integrating active and passive remote sensing with multi-dimensional structural features. Frontiers in Forests and Global Change, 9. https://doi.org/10.3389/ffgc.2026.1873077

Vancouver

Sa R, Fan W. Quantifying biomass saturation in planted forests: integrating active and passive remote sensing with multi-dimensional structural features. Frontiers in Forests and Global Change. 2026;9. doi:10.3389/ffgc.2026.1873077

BibTeX
@article{Sa2026,
  author = {Sa, Rula and Fan, Wenyi},
  title = {Quantifying biomass saturation in planted forests: integrating active and passive remote sensing with multi-dimensional structural features},
  journal = {Frontiers in Forests and Global Change},
  year = {2026},
  volume = {9},
  doi = {10.3389/ffgc.2026.1873077}
}
Kostenlos starten →