Modeling the River Flow Discharge by Using the Combined Multivariate Time Series Models

For over three decades, hydrologists were recommended multivariate models to describe and modeling the complex hydrology data. While recently the multivariate models in hydrology is discussed. In multivariate models, the modeling and predicting various parameters can improve by involving other factors. Also Since nonlinear models with conditional variance, the remaining portion of the linear models to adequately model, we expect that the combination of linear and nonlinear models, partly to increase the accuracy of modeling and predictions. In this study, two multivariate periodic ARMA and combined multivariate with conditional variance models were investigated to modeling monthly discharge of Nazloochai, Babolrood and Hamoon Rivers that located in West Azerbaijan, Mazanderan and Sistan-Balochestan Provinces respectively during the period of 1962-2011 (50 years) under effective the precipitation and temperature of mentioned basin synoptic stations. The results of evaluation and verification models (Root mean square error) showed that booth models have a more accuracy to modeling the river flow rate. Also the results showed that the combined multivariate with conditional variance model has the more accurately than multivariate periodic ARMA model. Also the results indicated that with combined two mentioned models, the model’s error in modeling the Nazloochai, Babolrood and Hamoon rivers flow discharge will be better amount 30, 17 and 1 percentage respectively. Finally the results indicated that the combined model has a more accuracy in the moderate zones of Iran.