Manikandan et al. (2026) Foundation-Model Embeddings for Land-Cover Mapping and Annual Change Detection in a Hyper-Arid Region: A Case Study of Saudi Arabia (2017–2024)

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

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-09-15
• Authors: Karuppasamy P. Manikandan, V. V. Naveen Kumar, Manzar Abbas Gul Muhammad, Muhammed Rafeeq Makkar, Luai M. Alhems
• DOI: 10.3390/rs18183163

Research Groups

• Earth Observation Lab (EOL), King Abdullah University of Science and Technology (KAUST)
• AlphaEarth Foundation

Short Summary

This study evaluates the use of foundation-model embeddings for land-cover classification and change detection in hyper-arid environments, achieving an overall accuracy of 0.815 ± 0.007.

Objective

• Investigate the effectiveness of AlphaEarth foundation-model embeddings for land-cover classification and change detection in Saudi Arabia (2017–2024).

Study Configuration

• Spatial Scale: National scale, with a focus on Saudi Arabia.
• Temporal Scale: Annual representations fused from Sentinel-1, Sentinel-2, Landsat, and LiDAR data from 2017 to 2024.

Methodology and Data

• Models used: Random Forest classifier using AlphaEarth foundation-model embeddings.
• Data sources: Sentinel-1, Sentinel-2, Landsat, and LiDAR satellite data, WorldCover 2021 strata, and independent ground truth data.

Main Results

• The study achieved an overall accuracy of 0.815 ± 0.007 for land-cover classification using the Random Forest classifier with AlphaEarth foundation-model embeddings.
• UMAP visualisation revealed five sub-types within the single WorldCover bare/sparse vegetation class, consistent with geomorphologically distinct desert substrates.
• An indicative cross-feature benchmark showed a 14.1 percentage-point higher overall accuracy for the foundation-model representation than for the strongest Sentinel-2 baseline.

Contributions

• The study provides a methodologically transparent workflow for foundation model-based land-cover monitoring in data-scarce arid environments.
• The results establish the effectiveness of AlphaEarth foundation-model embeddings for land-cover classification and change detection in hyper-arid environments.

Funding

• This research was funded by the Earth Observation Lab (EOL) at King Abdullah University of Science and Technology (KAUST).
• Additional funding provided by the AlphaEarth Foundation.

Citation

@article{Manikandan2026FoundationModel,
  author = {Manikandan, Karuppasamy P. and Kumar, V. V. Naveen and Muhammad, Manzar Abbas Gul and Makkar, Muhammed Rafeeq and Alhems, Luai M.},
  title = {Foundation-Model Embeddings for Land-Cover Mapping and Annual Change Detection in a Hyper-Arid Region: A Case Study of Saudi Arabia (2017–2024)},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18183163},
  url = {https://doi.org/10.3390/rs18183163}
}