Modeling of River Quality Parameters Using Hybrid Time Series Models (Case Study: Gedarchay River)

In this research, in order to study and evaluate the accuracy of single-variable time series models, multi-variables time series and hybrid models in modeling the river flow quality values in the Peygale hydrometric station located in the west of Lake Urmia from data The qualitative flow of the river and the flow river has been observed on an annual and monthly scale and during the statistical period of 1975-2016. The qualitative parameters studied in this study are EC, TDS and SAR values. In this study, EC and TDS values were modeled and studied. The ARMA model (1.0) was selected for EC and TDS values of the hydrometric station used as the superior model. The results showed that the accuracy of single-variable linear time series models in modeling the EC and TDS parameters of the studied station was not satisfactory, but acceptable. Similar results were obtained for monthly single-variable models. After studying the time series single-valued model, multi-variable time series models were investigated on the monthly and annual scale. In this model, the EC, TDS, SAR and flow data of the river were considered at the Peylgale hydrometric station as inputs of the model. The results of the error analysis of CARMA and MPAR models for EC values showed that, on average, the error values were 25 and 21 percent lower than the ARMA and PARMA models, which resulted in an improvement in the error values for the data TDS is 34% and 33% respectively. Overall, the results showed that in all stations studied, EC and TDS values are estimated to be better than single-variable models of ArmA family on annual and monthly basis using multiple time series models. Let's say Finally, applying the time series hybrid models, the error rate resulting from modeling of EC values in the monthly and annual scale was 46% and 10%, respectively, and for the TDS values it was 40% and 12%, respectively.