Abstract: The wavelet transform has remarkable advantages in denoising of precipitation time-series data, and can effectively improve the accuracy of precipitation time-series prediction.In order to determine the selection of wavelet basis function, decomposition scale and threshold estimation method in the process of wavelet denoising of precipitation time-series and achieve optimal denoising effect, the daily precipitation time-series from 2008 to 2018 of the National Meteorological Science Data Center were used as the basis data, and the five provinces with different climate types in China were selected as the study areas.Based on the composite index T,the denoising effects of 57 kinds of wavelet basis functions were evaluated, and the possible decomposition scales and common threshold estimation methods in the denoising process were also evaluated.The results showed that the Daubechies of 7～10 order was the best wavelet basis function group, and the minimum T values ranged from 0.326 4 to 0.422 8.Wavelet functions from the Symlets wavelet family showed poor performance.Moreover, the optimal decomposition scale was 3-level, and the minimum T values were between 0.1844 and 0.2526.Heursure threshold and Stein unbiased risk estimation threshold had the best denoising effect, and the minimum T values were between 0.377 3 and 0.435 9.The research results can provide a reference for denoising methods for precipitation time-series in China and other hydrometeorological time series.