Retrieval of Global Total Precipitable Water over Sea Surfaces from MWHS-II/FY-3D Data Using the BP Neural Network

Total Precipitable water (TPW) is an important parameter to characterize the water vapor content of the atmosphere, and its distribution characteristics are of great significance to the understanding and prediction of global climate change. Satellite passive microwave remote sensing, with its all-weather and all-day observation capability and high global coverage, is currently the main means of atmospheric detection in meteorological research. The Advanced Microwave Humidity Sounder (MWHS-II) on the Chinese second-generation polarorbiting operational meteorological satellite Fengyun 3D (FY-3D) has an excellent capability of detecting the vertical structure of atmospheric humidity and precipitation parameters. In this work, a Back Propagation neural network (BPNN) algorithm is firstly developed to retrieve high-precision global TPW over sea surfaces from MWHS-II/FY-3D data. Because of high accuracy, the fifth-generation of European Centre for Medium-Range Weather Forecast atmospheric reanalysis (ERA5) data serve as ground truths, and the spatiotemporal matching points between the MWHS-II/FY-3D observations and ERA5 data are collected and used as samples to train the BPNN algorithm. The training results show that the root mean square error (RMSE) and correlation coefficient are 2.04mm and 0.99 for the test set, respectively. Then, global TPWs over sea surfaces are retrieved from the MWHS-II/FY-3D data in 2022. Finally, the retrieved TPWs are validated with the radiosonde data, and a RMSE of 3.46mm and a coefficient of 0.97 are obtained. The RMSE and coefficient of F18 SSMIS TPW product in 2022 against the radiosonde data are 4.24mm and 0.96, respectively. The validation results indicate that the TPWs retrieved in this work are more accurate than the F18 SSMIS TPW product.