Investigating the existence of a trend and reaching a static time series is one of the essential issues in modeling the hydrological time series, because the trend reduces the time series forecasting model's accuracy. The purpose of the present study was to compare the different methods for detrending the river discharge time series to increase the accuracy of forecasting models. The monthly trend in discharge at 15 stations in the Urmia Lake basin was investigated using the Mann–Kendall test. The test results showed a significant decrease in trend with 99% reliability in the monthly discharge of most stations. After detecting the trend, time series detrending was performed using methods based on line slope, Cindex, cumulative, standardization, and differencing. In addition, for methods of line slope and Cindex, the time series trend was removed from the change point. Then, the Autoregressive Integrated Moving Average models were fitted to each time series. The best model was selected based on the Akaike Information Criterion, Root Mean Square Error, and Standard Error criteria. As a result, the detrending operation increases the accuracy of the forecasting. The results showed that the method of line slope detrending had less error and more correlation than the other methods for discharge forecasting. Moreover, results for the cumulative method are worst for forecasting. Cindex and Difference methods provide very slightly improved results than the case without detrending.
