Abstract:In order to further discuss the application effect of a new type of “up-drift-down” sounding data to data assimilation and numerical prediction, assimilation comparisons were carried out based on the WRF (Weather Research and Forecast) model and WRFDA (WRF data assimilation) system. In this paper, based on the quality evaluation and sparseness of the new radiosonde data, the descending data are combined and assimilated with the conventional observation data. The influence and causes of the simulated data on the rainstorm forecast quality in the middle and lower reaches of the Yangtze River are discussed. The main test results include that the accuracy of the latest test data is verified by cross-comparison of the test data with the FNL data and sounding data from the same station. The scheme of combining the characteristic layers with the specific layers can be used to sparse the ascending and descending segments of the new radiosonde, which can get better results. The rainstorm forecasting technique can be improved to some extent by assimilating the data in the descending section. The adjustment of wind field and humidity field is one of the important reasons for the improvement of rainstorm forecasting skills.