Marques et al. (2025) Integrating UAV Multi-Temporal Imagery and Machine Learning to Assess Biophysical Parameters of Douro Grapevines

⚠️ 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: Pedro Marques, Leilson Ferreira, Telmo Adão, Joaquim J. Sousa, Raul Morais, Emanuel Peres, Luís Pádua
• DOI: 10.3390/rs17233915

Research Groups

Not provided in the paper text.

Short Summary

This study investigates the use of UAV multispectral data and machine learning to estimate grapevine leaf area index, pruning wood biomass, and yield across mixed-variety vineyards, demonstrating that ML algorithms, especially with geometric features, provide accurate and scalable monitoring solutions.

Objective

• To evaluate the effectiveness of unmanned aerial vehicle (UAV) multispectral data combined with machine learning (ML) techniques for estimating grapevine leaf area index (LAI), pruning wood biomass, and yield in heterogeneous vineyards.

Study Configuration

• Spatial Scale: Mixed-variety vineyards in the Douro Region of Portugal.
• Temporal Scale: Three phenological stages, from veraison to maturation, within a growing season.

Methodology and Data

• Models used: Multiple Linear Regression (MLR) and five Machine Learning (ML) algorithms (including Random Forest). Feature selection (forward and backward) and logarithmic transformations were applied.
• Data sources: Unmanned Aerial Vehicle (UAV) multispectral data, photogrammetric elevation data (for geometric features).

Main Results

• Machine Learning algorithms generally provided better predictive performance than MLR, particularly when geometric features were included.
• At harvest-ready, Random Forest achieved the highest accuracy for LAI (R² = 0.83) and yield (R² = 0.75).
• MLR produced the most accurate estimates for pruning wood biomass (R² = 0.83).
• Canopy area was identified as the most informative geometric variable.
• The Modified Soil-Adjusted Vegetation Index (MSAVI) and the Soil-Adjusted Vegetation Index (SAVI) were the most relevant spectral features.
• The developed models performed well across different grapevine varieties.

Contributions

• Demonstrates a practical, non-invasive, and scalable approach for monitoring key grapevine biophysical parameters (LAI, pruning wood biomass, yield) in heterogeneous vineyards using UAV multispectral data and machine learning.
• Highlights the significant improvement in estimation accuracy when incorporating geometric features derived from photogrammetric data alongside spectral features.
• Identifies specific spectral indices (MSAVI, SAVI) and geometric variables (canopy area) as highly relevant for precision viticulture applications.

Funding

Not provided in the paper text.

Citation

@article{Marques2025Integrating,
  author = {Marques, Pedro and Ferreira, Leilson and Adão, Telmo and Sousa, Joaquim J. and Morais, Raul and Peres, Emanuel and Pádua, Luís},
  title = {Integrating UAV Multi-Temporal Imagery and Machine Learning to Assess Biophysical Parameters of Douro Grapevines},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17233915},
  url = {https://doi.org/10.3390/rs17233915}
}