Comparison of performance of statistical models in forecasting monthly streamflow of Kizil River, China

This paper presented the application of autoregressive integrated moving average (ARIMA), seasonal ARIMA (SARIMA) and Jordan-Elman artificial neural networks (ANN) models in forecasting monthly streamflow in the Kizil River, Xinjiang, China. Two different types of monthly streamflow data (original and deseasonalized data) were used to develop time series and Jordan-Elman neural networks forecasting models using previous flow conditions as predictors. The one-month-ahead forecasting performance of all models for testing period (1998-2005) were compared using average monthly flow of Kalabeili Gaging Station on Kizil River, Xinjiang, China. The Jordan-Elman ANN models using previous flow conditions as inputs resulted no significant improvement in one-month-ahead forecasts over time series models. The results of this study suggested that simple time series models (ARIMA and SARIMA) models could be used in one-month-ahead streamflow forecasting at the study site with a simple, explicit model structure and similar model performance as the Jordan-Elman neural networks models.