Under the double impacts of climate changes and human activities, variation in hydrological sequence occurred. This paper analyzes variation in runoff time series with a combining approach of approximate entropy method and sliding method, using the daily runoff series of totally 41 years from 1960 to 2000 at the Linjiacun station of Wei River. The analysis results show that in the cases of different sliding windows and different steps, in dynamical structure this series has four obvious characteristics and abrupt changes occurred in the years 1972, 1982, and 1994. This moving approximate entropy approach has high resolution strong anti-interference characteristics in analysis of nonlinear time series in comparison with the traditional mutation analysis method. It can give an effective diagnosis and a new idea for mutation analysis of highly complex series under changable environment.