Evaluation of Univariate, Multivariate and Combined Time Series Model to Prediction and Estimation the Mean Annual Sediment (Case Study: Sistan River)

Erosion, sediment transport and sediment estimate phenomenon with their damage in rivers is a one of the most importance point in river engineering. Correctly modeling and prediction of this parameter with involving the river flow discharge can be most useful in life of hydraulic structures and drainage networks. In fact, using the multivariate models and involving the effective other parameters such as flow discharge can be improved the modeling and prediction results. In this study using the common time series model (ARMA), multivariate model (CARMA) and combined models (ARMA-ARCH and CARMA-ARCH), mean annual sediment (ton.day) and mean annual flow discharge (m3/s) time series of Sistan River in period of 42 years (1970-2012) to estimating and prediction the mean annual sediment. By using the mentioned models, mean annual sediment in period of 1970-2012 was modeled. The results showed that with involving the mean annual flow discharge in multivariate model, the accuracy and model's error in validation phase compared the univariate models were improved almost of 8 and 50 percentages respectively. Also the results showed that among four mentioned models, the combined multivariate model have a lowest error. By using the multivariate model, the time series of mean annual sediment until the end of 2022 year were predicted. The prediction results showed that the mean annual sediment in prediction period was decreased compared the previous years.