Abstract: The stochastic characteristics of meteorological factors play important roles in the dynamic variation and optimal regulation of soil moisture. Using 26 years' meteorological data of Xiaohe Irrigation Area, Shanxi Province, reference evapotranspiration per 10 days was calculated with Penman-Monteith equation suggested by FAO. The stochastic characteristics of reference evapotranspiration series and precipitation series were analyzed by time series analysis method, and each series was decomposed to trend component, periodic component and stationary stochastic component. The results suggest that the reference evapotranspiration has an increasing trend, while the precipitation series has a decreasing trend in the recent over 20 years. These two series are negatively correlated. Their mean values and standard deviation can be expressed with the second order Fourier series, and their stochastic component can be expressed with autoregression models. These models can be further used in the stochastic forecasting of soil moisture and stochastic optimization of irrigation scheduling.