Based on runoff series in the past 50 years measured at Sanhuangmiao hydrologic station in the upstream of Tuojiang River, we adopted wavelet analysis to analyze the monthly runoff over years and the flood season flow in each year to find out the catastrophe points and period characteristics at multi-time scales. Wavelet variogram was used to identify dominant periods, and X-11 seasonal adjustment method was applied to extract the seasonal factors from monthly runoff series in order to obtain the trend and the seasonal impact. Results indicated that: the monthly runoff series exhibited three  dominant periods of 1-2 years, 11 years, and 21 years and two sub-dominant periods of 3 years and 5 years, while the flood season flow was featured by three obvious dominant periods of 3 years, 5 years, and 13 years. The impact of seasonal factors on monthly runoff in flood seasons were higher than that in other months. The seasonal factors in July increased steadily by years, and changed remarkably in August and September. Researches on runoff change would contribute to the exploitation of water resources and the optimal scheduling of water resources in canal irrigation areas.