Chen et al. (2026) Downscaling of satellite passive microwave brightness temperature through super-resolution reconstruction

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-04-03
• Authors: Jiaxin Chen, Ji Zhou, Tao Zhang, Shaojie Zhao, Ruyin Cao, Jin Ma, Wenbin Tang, Lin Feng
• DOI: 10.1016/j.rse.2026.115405

Research Groups

• School of Resources and Environment, University of Electronic Science and Technology of China, Chengdu, China
• State Key Laboratory of Earth Surface Processes and Hazards Risk Governance (ESPHR), Faculty of Geographical Science, Beijing Normal University, Beijing, China

Short Summary

This study develops an integrated land and atmospheric model approach (LEM-RTTOV) to simulate high-resolution passive microwave brightness temperature (PMW BT) and proposes a novel decoupled residual diffusion model (DRDiff) for super-resolution reconstruction to downscale satellite PMW BT, achieving accurate results for AMSR2 data.

Objective

• To address the limitations in PMW BT downscaling, specifically the lack of high-resolution BT references and the inadequacy of existing optical image super-resolution models for PMW BT spatial variations.
• To develop an approach by integrating land and atmospheric models to simulate high-resolution PMW BT.
• To propose a super-resolution reconstruction model (DRDiff) for PMW BT downscaling that effectively captures spatial variations.

Study Configuration

• Spatial Scale: Southwestern China and surrounding areas.
• Temporal Scale: Monthly (for AMSR2 BT downscaling).

Methodology and Data

• Models used: Land (LEM) model, atmospheric (RTTOV) model, LEM-RTTOV integrated approach, Decoupled Residual Diffusion Model (DRDiff), U-Net-like prediction network with large kernel attention.
• Data sources: Simulated test set, AMSR2 Brightness Temperature (BT) data.

Main Results

• An integrated land and atmospheric model approach (LEM-RTTOV) was successfully developed to simulate high-resolution PMW BT, providing a solid basis for BT downscaling.
• The proposed DRDiff model achieved a root mean square error (RMSE) of 0.90 K, a mean absolute error (MAE) of 0.68 K, and a mean coefficient of determination (R²) of 0.993 on the simulated test set.
• DRDiff effectively captures the spatial variations between high- and low-resolution BTs by numerically decoupling the residual and noise schedule.
• The model performed well on monthly AMSR2 BT downscaling, producing relatively accurate downscaled BTs.

Contributions

• Development of a novel simulated approach (LEM-RTTOV) integrating land and atmospheric models to generate high-resolution PMW BT references, addressing a critical data gap.
• Introduction of DRDiff, a new decoupled residual diffusion model specifically designed for PMW BT super-resolution reconstruction, which outperforms existing optical image SR models by effectively capturing PMW BT spatial patterns.
• Successful application and validation of DRDiff on real-world AMSR2 BT data, providing a robust tool for subsequent applied studies requiring high-resolution BTs, such as soil moisture and land surface temperature retrieval.

Funding

Not specified in the provided text.

Citation

@article{Chen2026Downscaling,
  author = {Chen, Jiaxin and Zhou, Ji and Zhang, Tao and Zhao, Shaojie and Cao, Ruyin and Ma, Jin and Tang, Wenbin and Feng, Lin},
  title = {Downscaling of satellite passive microwave brightness temperature through super-resolution reconstruction},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115405},
  url = {https://doi.org/10.1016/j.rse.2026.115405}
}