Runoff prediction using intelligent models

River runoff prediction because of its high importance in planning, reservoir operation and management of surface water has always attracted the attention of officials, planners and water engineers and water resources. On the other hand because of availab temporal and spatial changes, non-linear relationships and uncertainty, and many other factors to predict rainfall-runoff relationship is very difficult. But todays the use of intelligent systems can be useful for predicting such complex phenomena. In this study, using meteorological and hydrometric data for the period 1970-1971 to 2011-2012 to estimate runoff in the watershed Amameh using MLP, RBF, and ANFIS were used. The results showed that out of models ANFIS has the best function and can predict runoff very well. So that according errors, the structure model number 54 with eight inputs including rainfall and runoff to delay for two days and temperature, evaporation and relative humidity and cluster seperation and its errors was 0.001, 0.025 and 0.008 in training stage and 0.001, 0.026 and 0.008 in test stage was the best model in Amameh Watershed.