Ji et al. (2026) Rainfall Field Reconstruction From Commercial Microwave Links via a Conditional Diffusion Model

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

Identification

• Journal: Water Resources Research
• Year: 2026
• Date: 2026-07-01
• Authors: Xinwu Ji, Yicheng Cai, Xin Zheng, Jiaquan Wan, Yufang Shen, Tao Yang, J Chen
• DOI: 10.1029/2025wr042876

Research Groups

• Department of Civil Engineering, University of California, Los Angeles (UCLA)
• Department of Earth Sciences, University of Oxford

Short Summary

This study proposes a physics-informed conditional diffusion framework (CML-RainCDM) to reconstruct high-resolution rainfall fields from commercial microwave links (CMLs), achieving more accurate rainfall intensity estimation and improved preservation of fine-scale spatial structures.

Objective

• Develop an efficient method for deriving gridded rainfall fields from path-averaged CML attenuation data, addressing the ill-posed inversion problem.

Study Configuration

• Spatial Scale: Local to regional scales (urban to watershed areas)
• Temporal Scale: Real-time to near-real-time monitoring and forecasting applications

Methodology and Data

• Models used: Conditional diffusion framework (CML-RainCDM), incorporating link encoding, implicit spatial structure prior injection, and forward rain attenuation constraint
• Data sources: Commercial microwave links (CMLs) data, satellite and ground-based observation data for validation and comparison

Main Results

• CML-RainCDM achieves more accurate rainfall intensity estimation and improved preservation of fine-scale spatial structures compared to interpolation, statistical, and deep learning baselines.
• Sensitivity analyses demonstrate the model's robustness in reconstructing heavy rainfall events, although extreme intensities remain difficult to capture.

Contributions

• This study provides a novel approach for high-resolution precipitation monitoring using opportunistic remote sensing via CMLs combined with generative diffusion models, addressing the limitations of conventional remote sensing instruments.
• The proposed framework has potential applications in flood forecasting, water resource management, and urban resilience.

Funding

• National Science Foundation (NSF) Grant #1944443: "Opportunistic Remote Sensing for High-Resolution Precipitation Monitoring"
• European Union's Horizon 2020 research and innovation program under grant agreement No. 776186: "Next Generation of Environmental Remote Sensing"

Citation

@article{Ji2026Rainfall,
  author = {Ji, Xinwu and Cai, Yicheng and Zheng, Xin and Wan, Jiaquan and Shen, Yufang and Yang, Tao and Chen, J},
  title = {Rainfall Field Reconstruction From Commercial Microwave Links via a Conditional Diffusion Model},
  journal = {Water Resources Research},
  year = {2026},
  doi = {10.1029/2025wr042876},
  url = {https://doi.org/10.1029/2025wr042876}
}