Abstract: The study approach of runoff change complexity based on the de-noising of wavelet transform from symbolic dynamics viewpoint is present in the paper.First, the runoff sequences are dis-noised by the wavelet transform and the noise is filtered.Then the complexity of runoff change is calculated with the Lempel-Ziv algorithm.Finally some cases studies of the Yangtze River and the Yellow River are investigated.The results are as following: (1) Runoff change complexity in the Yellow River is higher than that of the Yangtze River.(2) The mean of runoff change complexity in the Yellow River basin is 8.3, and the trend of complexity from upstream to downstream has no evidence change.(3) Runoff change complexity in the Yellow River basin has no evidence change from time axial direction.