AN ON-ORBIT CALIBRATION METHOD FOR CORRECTION MICROWAVE RADIOMETER BASED ON PHYSICAL MODEL AND NEURAL NETWORKS

X. Zheng^1,2

¹National Space Science Center , ²Chinese Academy of Science

Correction microwave radiometer (CMR) plays a crucial role in providing wet tropospheric path delay for radar altimeters, a delay caused by atmospheric water vapor and cloud liquid water. The accuracy of geophysical retrievals from the CMR is fundamentally dependent on the quality of its calibration. This study proposes a novel method for selecting stable data to serve as the on-orbit calibration references. Subsequently, a novel parallel on-orbit calibration model that combines a physical model with a neural network (the P-N model) is introduced. In this model, the physical module calculates the antenna brightness temperature, while the neural network estimates the observed brightness temperature, the calibration parameters for the antenna brightness temperatures and the observed brightness temperatures are alternatively optimized. The proposed stable calibration reference and the P-N model are applied to the HY-2C CMR. The on-orbit calibration results are evaluated using the Observation Minus Background (O-B) methodology and through simultaneous nadir overpasses with the observed brightness temperatures from the Advanced Microwave Radiometer 2 (AMR-2) of Jason-3. The O-B residuals and the high-level consistency between the HY-2C CMR calibrated results and the Jason-3 AMR-2 measurements demonstrate the validity and accuracy of the stable calibration references and the proposed P-N model.