Multi-Band Weather Radar Polarization Information Conversion and Data Consistency Verification Based on Neural Network

Polarimetric weather radar measurements vary nonlinearly with changes in radar frequency and scanning elevation. In comparative observation experiments between Ku-band weather radar and CINRAD/SA radar, there were systematic errors in the polarimetric data of the two radars, the reason was the difference of frequency and elevation angles of them. Because the observation data of the ground-based Ku-band weather radar is small, it cannot cover most of the precipitation. So the T-matrix method was used to simulate the differential reflectivity factors of the Ku-band and S-band radars at different elevation angles, and the BP neural network was trained based on the simulation data to realize the conversion of the differential reflectivity factor from S-band to Ku-band to correct the systematic errors. Using the measured data, it was verified that the BP neural network can correct the systematic errors between the differential reflectivity factors of the two radars, improve the data consistency of them, and provide possibility for data fusion of S-band and Ku-band weather radars.