An Interpolation Iterative Inversion Method for Sparse Background Field of Microwave Radiometer Based on BP Neural Network

As an all-weather passive remote sensing instrument, microwave radiometer plays an important role in meteorological observation. The radiometer uses the inversion of water vapor and water vapor radiance temperature to measure the atmospheric profile. However, the radiometer inversion accuracy is highly dependent on historical data, which can cause large errors in observation points lacking historical data. To solve this problem, we propose a composite method that uses the NCEP data to form a high-resolution background field and improves the accuracy based on physical iterative inversion. To verify the effectiveness of this method, we carried out comparative experiments on high-resolution background field and physical iterative inversion using the historical meteorological data for Taizhou, China. The results show that, when the polar air mass impacts the seasonal climate, the atmospheric profile oscillates violently. Compared with the previous method, the average error for the whole year was reduced by 13%. However, for winter, when the strongest profile oscillation is present, our method exhibited the best accuracy, with error reduction by 30%. The method proposed in this paper facilitates improved tracking of the atmospheric profile, as well as improved data accuracy and remote sensing accuracy of microwave radiometer data.