Bouaziz et al. (2026) Deep Learning for Spatio-Temporal Fusion in Land Surface Temperature Estimation: A Comprehensive Survey, Experimental Analysis, and Future Trends

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

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-01-15
• Authors: Sofiane Bouaziz, Adel Hafiane, Rachid Nedjaï
• DOI: 10.3390/rs18020289

Research Groups

Not explicitly mentioned in the provided text.

Short Summary

This study reviews deep learning-based spatio-temporal fusion (STF) methods for Land Surface Temperature (LST), addressing the limitations of existing surface reflectance-oriented techniques and introducing a new LST dataset for benchmarking.

Objective

• To provide a focused review of deep learning-based spatio-temporal fusion methods for Land Surface Temperature (LST), formally define the thermal fusion task, propose a refined taxonomy, and analyze necessary adaptations for existing models.

Study Configuration

• Spatial Scale: Satellite sensor resolutions, typically ranging from tens of metres (e.g., Landsat at 30 m) to kilometres (e.g., MODIS at 1 km).
• Temporal Scale: Satellite sensor revisit periods, typically ranging from daily (e.g., MODIS) to multi-day (e.g., Landsat at 16 days). The dataset spans 11 years (2013-2024).

Methodology and Data

• Models used: Deep learning-based spatio-temporal fusion (STF) methods (reviewed and evaluated).
• Data sources: 51 pairs of Land Surface Temperature (LST) data from Terra MODIS and Landsat satellites.

Main Results

• A formal mathematical definition of the thermal spatio-temporal fusion task.
• A refined taxonomy of deep learning methods relevant for LST spatio-temporal fusion.
• An analysis of modifications required to adapt surface reflectance-oriented models for LST.
• Introduction of a new dataset comprising 51 Terra MODIS-Landsat LST pairs covering 2013 to 2024.
• Evaluation of representative models to understand their performance on thermal data.

Contributions

• First focused review of deep learning-based spatio-temporal fusion methods specifically for Land Surface Temperature.
• Establishment of a formal mathematical definition for the LST thermal fusion task.
• Development of a refined taxonomy for deep learning methods in LST spatio-temporal fusion.
• Detailed analysis of necessary adaptations for existing surface reflectance-oriented models to LST.
• Creation and release of a new, publicly available dataset (51 Terra MODIS-Landsat LST pairs, 2013-2024) to facilitate reproducibility and benchmarking in LST spatio-temporal fusion research.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Bouaziz2026Deep,
  author = {Bouaziz, Sofiane and Hafiane, Adel and Canals, Raphaël and Nedjaï, Rachid},
  title = {Deep Learning for Spatio-Temporal Fusion in Land Surface Temperature Estimation: A Comprehensive Survey, Experimental Analysis, and Future Trends},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18020289},
  url = {https://doi.org/10.3390/rs18020289}
}