Comparison of Different Time Series Analysis Methods for Forecasting Monthly Discharge in Karkheh Watershed

Forecasting of hydrological variables are suitable tools for water resources management. On the other hand، using time series concepts in forecasting is very useful. Therefore in present research for this aim، measured monthly discharge in 21 hydrometric stations from 1984 1985 to 2004-2005 in Karkheh watershed used، Then different method in forecasting and modeling in time series analysis include Trend Analysis، Holt-Winters method and various Autoregressive Moving Average Models (ARMA) suggested by Box-Jenkins were fitted to data. Afterwards، condition and accuracy of fitted models in forecasting of future monthly discharges considered using Akaike Information Criterion (AIC) and Root Mean Squared Error (RMSE). The results showed that trend analysis have the best forecast and autoregressive moving average models with few differences are in next place. In addition، Akaike Information Criterion showed that ARMA models with higher orders have better accuracy in modeling، but in forecasting ARMA with lower orders have better efficiency. There are high error in Holt-Winters method and is not recommended in this area. Finally، forecasted values using expressed methods have an overestimation in comparison with observed data.