Lutz et al. (2025) Estimating Plant Physiological Parameters for Vitis vinifera L. Using In Situ Hyperspectral Measurements and Ensemble Machine Learning

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

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-12-03
• Authors: Marco Lutz, Emilie Lüdicke, Daniel Heßdörfer, Tobias Ullmann, Melanie Brandmeier
• DOI: 10.3390/rs17233918

Research Groups

Experimental vineyard in Lower Franconia, Germany.

Short Summary

This study developed and evaluated an ensemble machine learning framework, integrating hyperspectral reflectance data with first derivative preprocessing, to accurately predict key photosynthetic parameters and water potential in grapevines, demonstrating its potential for precision viticulture.

Objective

• To develop and evaluate machine learning models, utilizing spectral reflectance data and preprocessing techniques, for the non-invasive prediction of photosynthetic parameters (assimilation rate, effective photosystem II quantum yield, electron transport rate) and water potential (stem and leaf) in Vitis vinifera.

Study Configuration

• Spatial Scale: Individual Vitis vinifera (cv. Müller-Thurgau) plants within an experimental vineyard.
• Temporal Scale: Three measurement days (25 July, 7 August, 12 August 2024) with diurnal observations.

Methodology and Data

• Models used: Support Vector Regression (SVR), Least Absolute Shrinkage and Selection Operator (Lasso), ElasticNet, Ridge Regression, Partial Least Squares Regression (PLSR), Artificial Neural Network (ANN), Random Forest. A stacking ensemble regressor with a Random Forest meta-learner was also employed. First derivative reflectance (FDR) preprocessing was applied.
• Data sources: Hyperspectral reflectance data (PSR+ hyperspectral spectroradiometer), photosynthetic parameters (LI-COR LI-6800 system), stem and leaf water potential.

Main Results

• First derivative reflectance (FDR) preprocessing significantly enhanced predictive performance, particularly for effective photosystem II quantum yield (ΦPSII) and electron transport rate (ETR).
• The ensemble machine learning approach achieved high predictive accuracy, with R² values up to 0.92 for ΦPSII and 0.85 for assimilation rate (A) at 1 nm spectral resolution.
• Predictive accuracy declined at coarser spectral resolutions, although FDR preprocessing provided some mitigation of this performance loss.
• Diurnal patterns indicated that morning to mid-morning measurements (between 9:00 and 11:00) were optimal for assessing vine vigor due to peak photosynthetic activity.
• Midday water potential declines were identified as favorable timing for irrigation scheduling.

Contributions

• Demonstrated the potential of integrating hyperspectral data with ensemble machine learning and FDR preprocessing for accurate, scalable, and high-throughput monitoring of grapevine physiology.
• Provided insights into optimal diurnal measurement timings for assessing vine vigor and informing irrigation scheduling.
• Supported the use of cost-effective sensors for real-time vineyard management under diverse environmental conditions.

Funding

Not specified in the provided text.

Citation

@article{Lutz2025Estimating,
  author = {Lutz, Marco and Lüdicke, Emilie and Heßdörfer, Daniel and Ullmann, Tobias and Brandmeier, Melanie},
  title = {Estimating Plant Physiological Parameters for Vitis vinifera L. Using In Situ Hyperspectral Measurements and Ensemble Machine Learning},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17233918},
  url = {https://doi.org/10.3390/rs17233918}
}