Mirtabatabaeipour et al. (2025) Distance Transform-Based Spatiotemporal Model for Approximating Missing NDVI from Satellite Data

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

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-10-10
• Authors: Amirhossein Mirtabatabaeipour, Lakin Wecker, Majid Amirfakhrian, Faramarz Samavati
• DOI: 10.3390/rs17203399

Research Groups

Not available from the provided text.

Short Summary

This paper proposes a novel spatiotemporal model to accurately approximate missing or contaminated Normalized Difference Vegetation Index (NDVI) data in satellite imagery due to clouds and shadows, demonstrating significant accuracy improvements over existing methods.

Objective

• To develop a robust model that integrates both temporal and spatial aspects of satellite data to accurately approximate missing or contaminated NDVI regions, thereby overcoming limitations posed by cloud and shadow contamination.

Study Configuration

• Spatial Scale: 16 farm fields in Western Canada.
• Temporal Scale: 6 years (2018 to 2023).

Methodology and Data

• Models used:
  • Proposed model: A spatiotemporal model combining spatial and temporal NDVI approximations using a distance transform and a novel decay function.
  • Comparison models: Spatial approximation, Temporal approximation, Simple combination, Spatiotemporal Kriging.
• Data sources:
  • High-resolution satellite images, specifically Normalized Difference Vegetation Index (NDVI) derived from publicly available sources like Sentinel.

Main Results

• The proposed spatiotemporal model significantly improves NDVI approximation accuracy.
• Achieved up to a 263% improvement in Root Mean Square Error (RMSE) compared to using only spatial or temporal approximations.
• Demonstrated up to a 51% improvement in RMSE compared to simple combination methods.
• Showed up to a 28% improvement in RMSE compared to Spatiotemporal Kriging.
• Empirically determined optimal parameters for the decay function and the distance-transform-based model.
• Successfully applied the model in a case study to improve the specification of the peak green day for numerous fields.

Contributions

• Introduction of a novel spatiotemporal model that effectively integrates spatial and temporal information for reconstructing missing satellite NDVI data.
• Development of a new decay function to control the transition between spatial and temporal approximations based on distance transform.
• Demonstrated superior performance and accuracy compared to individual spatial/temporal approximations, simple combination methods, and Spatiotemporal Kriging.
• Provided a practical application for improving agricultural monitoring, such as identifying peak green day.

Funding

Not available from the provided text.

Citation

@article{Mirtabatabaeipour2025Distance,
  author = {Mirtabatabaeipour, Amirhossein and Wecker, Lakin and Amirfakhrian, Majid and Samavati, Faramarz},
  title = {Distance Transform-Based Spatiotemporal Model for Approximating Missing NDVI from Satellite Data},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17203399},
  url = {https://doi.org/10.3390/rs17203399}
}