Passive atmospheric wind profile retrieval via multi-region microwave radiometer network

Atmospheric Wind Profile (AWP) retrieval is essential for applications in meteorology, aerospace, and renewable energy. This study introduces MWR-WINet, a novel framework for passive AWP retrieval using ground-based Microwave Radiometer (MWR) data. To the best of our knowledge, this is the first study to apply passive MWR observations for wind profile estimation. The proposed approach incorporates a multi-district observation network to enhance spatial representation and introduces a composite loss function that combines Mean Squared Error (MSE) with Kullback–Leibler divergence (KL) to improve model performance. Applied to three districts in Xi'an—Weiyang, Chang'an, and Lintong—the model achieves retrieval errors of 1.63, 1.70, and 1.87 m/s, respectively. A joint three-district model further reduces the error to 1.27 m/s and enhances wind direction accuracy by 19.56 %, with a correlation gain of 0.08. These results demonstrate that a networked observational strategy significantly improves retrieval accuracy. This work overcomes the limitations of traditional observation methods and supports the broader application of MWR-based atmospheric profiling.