Improvement of temperature and humidity profile retrieval from brightness temperature data
摘要
To improve the retrieval accuracy of atmospheric temperature and humidity profile and promote the development of the domestic ground-based microwave radiometer (MWR), we developed an improved indirect sample retrieval algorithm. This study applied the improved algorithm to the HRA002 ground-based MWR produced by Wuhan Huameng Technology Co., Ltd and tested at Wuhan National Basic Meteorological Station. Compared with the conventional gradient descent method and the trainLM (Levenberg-Marquardt) algorithm, validation of the Bayesian regularization (BR) algorithm with the radiosonde at the same station showed a root-mean-square error (RMSE) decrease from 2.21 K, 2.59 K to 2.06 K for temperature, from 1.31 g/m3, 1.16 g/m3 to 1.04 g/m3 for water vapor density, and from 22.33%, 24.53–21.49% for relative humidity. On the basis of adopting the BR algorithm, compared with the ordinary network, the RMSE between the retrieved results of temperature, water vapor density and relative humidity of the improved network and the radiosonde data decreased from 2.06 K, 1.04 g/m3, and 21.49% to 1.89 K, 0.98 g/m3, and 19.41%, respectively. The retrieval accuracy was improved by the modifications made by this study to the algorithm. The research results in this paper show that the improved algorithm can further improve the retrieval accuracy on the basis of the existing indirect sample retrieval algorithm and can effectively promote the development of domestic ground-based MWR.