The runoff time series is characterized by complexity and uncertainty. R/S analysis and multifractal method were used to analyze the data collected from 1954 to 2006 at Shijiao hydrology station of the Beijiang River in this paper.Results show that the Hurst index of Shijiao station is largely higher than 0.5

indicating that the average annual runoff time series has obvious trend components

namely the average annual runoff of the Beijiang River shows obvious durability and longterm memory. According to the V statistical value

the average annual runoff persistence will disappear after 15-20 years. Using the multifractal method

the power spectrum and statistical moment function were analyzed. The average annual runoff of the Beijiang River is of multi-fractal features. These results are important for the runoff forecasting.