Abstract: This paper used the wavelet transform of the Mexican hat wavelet function to analyze the precipitation data for 45 years(1961-2005) in Xi'an,and calculated the variance yield of wavelet to study the time of precipitation variation and the characteristics of the periodic variation of precipitation in Xi'an.On the basis of the analysis of precipitation time series,the results showed that this paper revealed the different levels of precipitation in the structure and characteristics of abnormal variation and mutation and predicted the seasonal precipitation in Xi'an and the trend of the annual precipitation.Particularly,by analyzing the predicted results of wavelet,the large area of spring drought in 2009,2010 and water-logging from autumnal rains after 2005 in Guanzhong area were verified,and those results were of important reference value and significance to understand the precipitation characteristics of Xi'an,agricultural production in Xi'an and the flood prevention work of the Weihe river.In the past 45 years,the spring precipitation in Xi'an decreased significantly(P<0.05),however,the linear trend of other seasonal and annual precipitation was not significant.Nonetheless,seasonal and annual precipitation had the characteristics of multiple time scales,different time scales showed the different cyclic-alternating-patterns,the periodic variation on large scales was nesting the periodic variation on large-scales.Overall performance was from small-scale of severe shocks,no obvious regularity,to large-scale of obvious regularity and continue to enhance or first and then decrease.

Key words: Xi'an, Wavelet analysis, Precipitation