Research Article
MFR-LSTM: A Robust Rainfall Estimation Model Based on Multi-feature Extraction Using Commercial Microwave Links
@INPROCEEDINGS{10.1007/978-3-032-33662-0_37, author={Zixuan Qin and Minghao Liang and Yong Liu and Biling Zhang and Caili Guo and Congzheng Han}, title={MFR-LSTM: A Robust Rainfall Estimation Model Based on Multi-feature Extraction Using Commercial Microwave Links}, proceedings={Communications and Networking. 20th EAI International Conference, ChinaCom 2025, Xi’an, China, November 8--9, 2025, Proceedings}, proceedings_a={CHINACOM}, year={2026}, month={10}, keywords={Rainfall Estimation Commercial Microwave Links Multi-Feature Extraction Long Short-Term Memory}, doi={10.1007/978-3-032-33662-0_37} }- Zixuan Qin
Minghao Liang
Yong Liu
Biling Zhang
Caili Guo
Congzheng Han
Year: 2026
MFR-LSTM: A Robust Rainfall Estimation Model Based on Multi-feature Extraction Using Commercial Microwave Links
CHINACOM
Springer
DOI: 10.1007/978-3-032-33662-0_37
Abstract
The microwaves signals are attenuated by the scattering and absorption of raindrops when they propagate through the rainfall, which makes commercial microwave links (CML) a unique opportunity for rainfall monitoring. However, traditional power law CML based methods for rainfall estimation encounter computational complexity and degraded accuracy under light rainfall conditions. To address these challenges, we propose Multi-Feature based and Robust LSTM (MFR-LSTM), a novel deep learning (DL) framework for rainfall estimation using CML. MFR-LSTM integrates an Inception module for spatial features extraction, a NoisyLinear layer for enhancing model robustness, and a LSTM module for capturing the temporal dependence of CML data. The experimental results show that compared to the baseline, MFR-LSTM are 8.25% higher in correlation coefficient and 1.82% higher in dry-wet classification accuracy. In addition, under the Gaussian noise, MFR-LSTM’s degradation in correlation coefficient is 6.2% less than the baseline, demonstrating its reliable potential in rainfall estimation and meteorological monitoring.