Wu et al. (2026) DLMSR-Transformer-MoE: A novel method for synchronous retrieval of land surface temperature and emissivity

Identification

• Journal: Mendeley Data
• Year: 2026
• Date: 2026-06-11
• Authors: Chenhao Wu, Kebiao Mao, Jiancheng Shi, Sayed M. Bateni
• DOI: 10.17632/f9sf94wvd4.1

Research Groups

• Chinese Academy of Agricultural Sciences, Beijing

Short Summary

The study introduces DLMSR-Transformer-MoE, a deep learning method designed for the synchronous retrieval of land surface temperature (LST) and land surface emissivity (LSE) using multi-channel thermal infrared brightness temperatures.

Objective

• To develop a novel retrieval method that learns the nonlinear relationship between LST and LSE by incorporating physical solvability constraints, product-level refinement, and surface-adaptive modeling.

Study Configuration

• Spatial Scale: Global
• Temporal Scale: Multi-temporal

Methodology and Data

• Models used: DLMSR-Transformer-MoE (a deep learning architecture combining Transformers and Mixture of Experts).
• Data sources: NASA Earthdata (Aqua/MODIS thermal infrared brightness temperatures, geolocation information, cloud masks, and original LST and LSE products).

Main Results

• Established a standardized preprocessing workflow including quality control, cloud screening, geometric correction, spatial resampling, and global-scale mosaicking.
• Developed a retrieval workflow capable of synchronous LST and LSE estimation, validated through global multi-temporal cross-validation, in situ LST measurements, and classification-based LSE references.

Contributions

• Proposes a novel deep learning framework (DLMSR-Transformer-MoE) that overcomes traditional retrieval limitations by integrating physical constraints and surface-adaptive modeling for the simultaneous estimation of LST and LSE.

Funding

• Science and Technology Department of Ningxia Yinchuan (Grant ID: No. 2024AC02032)

Citation

@article{Wu2026DLMSRTransformerMoE,
  author = {Wu, Chenhao and Mao, Kebiao and Shi, Jiancheng and Bateni, Sayed M.},
  title = {DLMSR-Transformer-MoE: A novel method for synchronous retrieval of land surface temperature and emissivity},
  journal = {Mendeley Data},
  year = {2026},
  doi = {10.17632/f9sf94wvd4.1},
  url = {https://doi.org/10.17632/f9sf94wvd4.1}
}