Abstract:In this paper, we process and analyze, by wavelet decomposition, the time series of ground-based GNSS precipitable water vapor (PWV), atmospheric pressure and zenith tropospheric delay (ZTD). Based on the actual precipitation of rainstorms, the results show that the wavelet high-frequency decomposition coefficients of one-hour interval PWV and ZTD are close to each other, and the characteristic information of rainstorm prediction can be extracted from them. The high-frequency ZTD can be used to replace PWV for wavelet analysis. In ZTD with frequency between 30 minutes and 1 hour, the forecast time information should be searched at level 1~3, the frequency below 30 minutes should be searched at level 3~5, the forecast threshold of db4 wavelet decomposition PWV can be set to -1.2, the forecast threshold of db4 wavelet decomposition ZTD can be set to -0.007, and the forecast threshold of db2 wavelet decomposition ZTD can be set to -0.01.