Need to criteria for the evolution of performance and categorize models in hydrology

A lot of models have been developed from the past until now to analysis rainfall-runoff process and the complexity of the models are increasing. Artificial Intelligent technique is one of the most developed methods as a branch of computer science. In these methods, ANN is widely used among those which are employed alone or in combination with others. In this study, daily runoff time series of 60 stations from different parts of Iran are simulated using both auto regressive moving-average (ARMA) as a simple and classic model, and ANN as an advanced model. The results indicate that in most of the time series, there is no significant difference between the results of two models. In the absence of clear criteria for evaluating the model performance to prediction, only complexities of the models were added daily. There has been no significant change in the results of the models and both of the models encountered problems with the prediction of some datasets.