One of the most important and fundamental issues that Iran has faced in recent years is water crisis. This problem mostly hits the country in years with drought. The minimum water flow in the river in fact is the low flow. Environmentally, the shrinkage of low flow to its minimum value. Increases the related contamination and lowers the dissolved oxygen in water which causes the death of aquatics in the rivers. As directorial view, this minimum of low flow has special importance for study the fields of urban, industrial and agricultural water supply. Therefore prediction of low flow in the water field and environmental issues must be heeded. In this study, prediction of 7-day and 30-day low flow has been considered which data of daily flows came from staions -Ghale Shahrokh and Eskandari in Isfahan province and Chamriz and Chenarsookhte in Fars province- was used. For this purpose, two methods of usual time series modeling and modeling of time series using wavelet (wavelet-series) were used. The approached, presented by Box and Jenkins was considered for the modeling of time series which consists of three steps: namely model, parameter estimation and goodness of fit test or time independency. For the prediction of low flow, modeling of time series was done for main, logarithmic and seasonal series. The results indicated the advantage of logarithmic series in all of the stations. One of the procedures which recently has been considered in hydrologic field is utilizing wavelet as an innovative and effective method in time series analysis. In the wavelet-time series modeling with Haar wavelet theory the considered series was decomposed. According to the considered series, decomposition was performance at levels 5 and 6 in mentioned stations. In data, there are results of analyzing wavelet including ‘A’ approximation which has main essence of data and details –containing white noise. Then the time series modeling steps were performed for the approximate. Finally, the series analysis has done simply with use of by obtaining correlation coefficient, mean rooted squared error and mean absolute deviation in modeling and predicting for two methods and also by considering that wavelet decomposition simplifies the series which simplifies the analysis of the series, the wavelet-time series method was proved to be the most appropriate method to predict the amount of low flow. Keywords: Box and Jenkins, forecast, Haar, low flow, time series, Wavelet