Abstract: The multiple-scale time-frequency analysis based on wavelet decomposition is widely applied to modeling and forecasting of complicated time series, but a traditional time series model cannot effectively describe the structure of a series containing outliers. To solve this problem, we adopted robust regression estimation to improve the traditional wavelet model and developed a wavelet analysis-robust estimation hydrological time series model that can effectively eliminate the impacts of outliers in observed data by using a multiple-scale wavelet analysis theory and its efficient processing of non-stationary signals. This model was verified against the monthly runoff forecasts of the Auto-Regressive and Moving Average (ARMA) model and Back Propagation (BP) neural network model through comparative analysis. It shows that the model is not only more accurate but also has better reliability and stability and a wider application potential in hydrological forecasting.