Retrieval of tropical cyclone temperature and humidity profiles from FY-3E MWTS and MWHS data using deep learning algorithm

Tropical cyclone (TC) disasters have a serious threat to social and economic development, so it is necessary to retrieve atmospheric temperature and humidity profiles in the typhoon region to provide timely and accurate information on the initial field of atmospheric humidity and temperature for numerical weather prediction, and to enhance the early warning capability of catastrophic typhoon weather. This paper addresses the retrieval of tropical cyclone temperature and humidity profiles over the western Pacific Ocean from the data acquired by the Microwave Temperature Sounder(MWTS) and the Microwave Humidity Sounder (MWHS) on Fengyun 3E (FY-3E) using the batch normalization and robust neural network (BRNN) algorithm. To improve retrieval accuracy, the FY-3E MWTS and MWHS observations in the TC region are classified according to different scattering conditions (clear, stratiform, convective), and the atmospheric temperature and humidity profiles in the TC region are retrieved by a deep learning method. The results show that, the root mean square errors (RMSEs) of the retrieved temperature profiles are less than 1.4 K, while the RMSEs of the derived humidity profile are less than 1.1 g/kg. In general, the retrieval algorithm can invert a reasonable TC thermal structure.