To solve many problems such as estimation of average monthly river inflow, it is necessary to consider a time-dependent phenomenon which is involved in several factors. So, a stochastic model has to be obtained to calculate the probability of a future amount of inflow. Studying of river behavior and the ability to forecast the future events is a prerequisite for the preparation of optimization models. In the present study, methods for creating, diagnosis and assessing the rate of compatibility with seasonal ARIMA time series models have been provided. It is also supposed that the seasonal classification based on their statistical parameter similarities, could lead to a better normalized series in comparison to the other transformation methods. The discussed methods are suitable for continuous systems. The ARIMA method was used to predict the future amounts of the inflow into the Karaj reservoir by using its previous and present values. For implementing models, first, it is necessary to normalize the observed data with a logical transformation method considering seasonalization. The results showed that the best fitted model is an annual series ARIMA (1, 0, 1) (2, 1, 1)12 with logarithmic transformation for forecasting models. It is also concluded that the 12 month forecast of inflow is better than 24 months in terms of the forecasted values.
