Reliable river flow forecasting is a key element in achieving sustainable water resources and environmental management. Accurate short term and long term river flow forecasts are particularly essential for the design of hydraulic structures, flood and drought analysis, irrigation scheduling, reservoir operation and environmental planning. Due to stochastic characteristics of hydrological events, forecasting the future condition of surface water is always associated with uncertainty. In this work an attempt is made to develop highly accurate river flow forecasting models. Wavelet multi-resolution analysis is applied in conjunction with various types and structure of computational intelligence models. Research outcomes indicate that forecasting reliability is significantly improved by applying proposed hybrid models, especially for longer lead time and peak values.