Quintela et al. (2025) Refining European Crop Mapping Classification Through the Integration of Permanent Crops: A Case Study in Rapidly Transitioning Irrigated Landscapes Induced by Dam Construction

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

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-12-09
• Authors: Manuel Quintela, Manuel L. Campagnolo, Rui Figueira
• DOI: 10.3390/rs17243979

Research Groups

Not provided in the given text.

Short Summary

This study refines the EU Crop Map 2018 by developing an automated machine learning model integrating Sentinel-1 and Sentinel-2 imagery to distinguish permanent crop types in southern Portugal, achieving 91% overall accuracy and highlighting the critical need for balanced training data.

Objective

• To refine the EU Crop Map 2018 classification by distinguishing specific permanent crop types using an automated machine learning model integrating Sentinel-1 and Sentinel-2 imagery in a region undergoing rapid land use change.

Study Configuration

• Spatial Scale: Regional (Alqueva reservoir area, southern Portugal)
• Temporal Scale: Focus on land cover in 2018, using imagery from that period.

Methodology and Data

• Models used: Automated machine learning model.
• Data sources: Sentinel-1 (S1) imagery, Sentinel-2 (S2) imagery, large reference dataset, EU Crop Map 2018, Portuguese official land use product COS 2018.

Main Results

• The model achieved an overall accuracy of 91% in distinguishing permanent crops, forests, and other land occupations.
• It effectively identified almond groves (F1 score = 0.90) and distinguished three major olive grove cultivation systems (F1-score ≥ 0.78).
• Performance was lower for vineyards (F1 score = 0.71) and other permanent crops (F1 score = 0.48).
• Comparison with COS 2018 showed strong overall spatial alignment, despite inconsistencies, and a lower F1 score (0.60) in direct comparison.
• Overall accuracy remained above 83% even with only 5% of the training data, but underrepresented classes experienced significant performance degradation, emphasizing the need to address class imbalance.

Contributions

• Provides a refined classification of permanent crop types, improving upon the aggregated "shrublands and woodlands" class in the EU Crop Map 2018.
• Demonstrates the high potential of integrating Sentinel-1 and Sentinel-2 imagery with machine learning for detailed agricultural land cover mapping in dynamic regions.
• Quantifies the impact of training set size and class imbalance on classification accuracy, particularly for underrepresented classes, offering crucial insights for future mapping efforts.

Funding

Not provided in the given text.

Citation

@article{Quintela2025Refining,
  author = {Quintela, Manuel and Campagnolo, Manuel L. and Figueira, Rui},
  title = {Refining European Crop Mapping Classification Through the Integration of Permanent Crops: A Case Study in Rapidly Transitioning Irrigated Landscapes Induced by Dam Construction},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17243979},
  url = {https://doi.org/10.3390/rs17243979}
}