Abstract: Under the combined influence of global climate change and high intensity human activities, the natural water cycle in many river basins has been disrupted. The runoff series show obvious non-stationary characteristics, which brings certain challenges to water resource planning, management, forecast and regulation. Revealing the non-stationary characteristics of runoff series can effectively deal with complex water problems under global climate change, and is important for reducing the difficulty of hydrological analysis and improving the accuracy of runoff prediction. This study takes Lancun Station in the upper reaches of Fenhe River as the research object to analyze whether the annual and monthly runoff series of the station from 1958 to 2016 are stationary. Firstly, from the perspective of stochastic hydrology, MannKendall test method and wavelet analysis method are used to identify the trend, abrupt change and periodic characteristics of the runoff series. On this basis, the Ng-Perron unit root test is introduced from the perspective of statistical hydrology. A suitable test equation is selected by Mann-Kendall trend test and scatterplot method, and Generalized Least Squares（GLS） regression potential is applied to the runoff series. The Modified information criterion（MIC） is used to calculate the optimal time lag order and determine whether the runoff series is nonstationary. The results show that there are trends, abrupt changes and periodic components in the runoff series, which are non-stationary runoff series. At the same time, the Ng-Perron unit root test shows that the annual and monthly runoff series of the station has non-stationary characteristics at 1% significance level. Compared with the traditional unit root test method, Ng-Perron unit root test adopts more robust modified test statistics to significantly adjust the horizontal distortion in the case of small samples, which has better test level and efficacy, and can obtain more reasonable test results. The research results provide reference for the further improvement of the non-stationarity test theory of runoff series and the development and application of runoff prediction model.