Rainfall-runoff modeling is important in flood forecasting systems. Although, a wide variety rainfall-runoff models has been developed and applied, but it cannot be claimed that there is one model which can perform satisfactorily at all times or under all conditions. Instead of switching models from one to another, this study proposes combining the simple linear rainfall-runoff model results. This study presents the development of five combination methods, simple average method (SAM), the weighted average method (WAM), the fuzzy system method (FSM), the neural network method (NNM) and the adaptive neuro-fuzzy inference system method (ANFISM) to combine the simulated results of three different rainfall-runoff models called single linear model (SLM), Linear Perturbation Model (LPM) and Linearly Varying Gain Factor Model (LVGFM) on four catchments. Comparison of the estimated runoff results reveals that the ANFIS combination method performs better than the other combination methods and is the best individual rainfall-runoff model. Furthermore, the ANFIS combination method providesimproved flood estimates and is recommended for use as the combination system for flood forecastingthat can also be used by engineers and hydrologists.