EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Evaluating Spectral Preprocessing and Machine Learning for Hyperspectral Estimation of Leaf Photosynthetic Capacity in Subtropical Forests

Miaomiao Wang, Shaoqiang Wang, Weikang Lin, Jinghua Chen, Hong Chen, Yong Liang, Jian Zhao

Forests · 10.11.2026 · 17 (11) · 1337

DOI: 10.3390/f17111337 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Forests, online erschienen am 10.11.2026. Verfasst von Miaomiao Wang und 6 weiteren Autor:innen. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Fujian Academy of Agricultural Sciences, Natural Science Foundation of Fujian Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate estimation of the maximum carboxylation rate (Vcmax) and maximum electron transport rate (Jmax) is important for improving the representation of photosynthetic capacity in terrestrial carbon-cycle models. Leaf hyperspectral spectroscopy provides a promising approach for estimating these physiological traits, but the effects of spectral preprocessing and modeling strategy on prediction performance remain insufficiently evaluated, particularly across different forest types. In this study, leaf hyperspectral reflectance (400–2400 nm) and gas-exchange measurements were collected from subtropical evergreen needleleaf forest (ENF) and evergreen broadleaf forest (EBF) samples to systematically evaluate the estimation of Vcmax and Jmax using six spectral preprocessing methods and three machine learning algorithms. ENF and EBF showed broadly similar spectral signatures and comparable distributions of Vcmax and Jmax. Across the tested model configurations, derivative-based preprocessing methods generally improved prediction performance, while random forest (RF) and extreme gradient boosting (XGBoost) generally outperformed partial least squares regression (PLSR). Using 99 leaf samples (36 ENF and 63 EBF samples) and internal leave-one-out cross-validation, the highest predictive performance was moderate, with R2 = 0.45 (RMSE = 9.89, MAE = 8.23) for Vcmax and R2 = 0.57 (RMSE = 15.50, MAE = 12.19) for Jmax. The models were evaluated using internal leave-one-out cross-validation (LOOCV). Although overall prediction performance was broadly comparable between ENF and EBF, spectral importance analyses revealed distinct forest-type-associated patterns in important wavelengths. Important wavelengths in ENF samples were concentrated mainly in the shortwave infrared (SWIR) region (1300–2400 nm), whereas EBFs showed greater importance of wavelengths in the near-infrared (NIR) region (750–1300 nm). These findings highlight the importance of spectral preprocessing when estimating leaf photosynthetic capacity from hyperspectral reflectance and indicate that forest types may exhibit different patterns of spectral importance. The identified patterns provide a basis for further investigation of hyperspectral approaches to photosynthetic trait estimation in subtropical forests.

Förderung

  • National Natural Science Foundation of China
  • Fujian Academy of Agricultural Sciences
  • Natural Science Foundation of Fujian Province

Zitieren

APA

Wang, M., Wang, S., Lin, W., Chen, J., Chen, H., Liang, Y., & Zhao, J. (2026). Evaluating Spectral Preprocessing and Machine Learning for Hyperspectral Estimation of Leaf Photosynthetic Capacity in Subtropical Forests. Forests, 17(11), 1337. https://doi.org/10.3390/f17111337

Vancouver

Wang M, Wang S, Lin W, Chen J, Chen H, Liang Y, et al. Evaluating Spectral Preprocessing and Machine Learning for Hyperspectral Estimation of Leaf Photosynthetic Capacity in Subtropical Forests. Forests. 2026;17(11):1337. doi:10.3390/f17111337

BibTeX
@article{Wang2026,
  author = {Wang, Miaomiao and Wang, Shaoqiang and Lin, Weikang and Chen, Jinghua and Chen, Hong and Liang, Yong and Zhao, Jian},
  title = {Evaluating Spectral Preprocessing and Machine Learning for Hyperspectral Estimation of Leaf Photosynthetic Capacity in Subtropical Forests},
  journal = {Forests},
  year = {2026},
  volume = {17},
  number = {11},
  pages = {1337},
  doi = {10.3390/f17111337}
}
Kostenlos starten →