Application of vector autoregressive models to estimate pan evaporation values at the Salt Lake Basin, Iran

Abstract

Abstract Thousands of billions of cubic meters of fresh water collected at great expense are evaporated annually from dams, and salts of evaporating water reduces water quality. In this study, the efficiency of the vector autoregressive model called VAR model has been examined on an annual scale using pan evaporation data in the salt lake basin, Iran, during the statistical period of 1996–2015. Since hydrologic modeling is concerned with the accuracy and efficiency of the model, therefore, we must try to evolve and improve the results of the models. In this study, VAR multivariable time series and nonlinear GARCH models have been used. The results of linear and nonlinear hybrid models in modeling the annual and monthly pan evaporation values of studied stations at the basin area of the salt lake indicated, that the pan evaporation values in the annual scale have the best fit with hybrid models. The results of the study of the accuracy of these models in modeling the pan evaporation values indicated, that the VAR-GARCH hybrid models have a high accuracy relative to the vector models and have been able to model the pan evaporation values with good accuracy and with the lowest error rate. Of the two models that have both annual nature (VAR and VAR-GARCH), the best model can be selected based on the estimation of the error values. In this study, we first examine the accuracy of the relatively new vector autoregressive model. The results of the estimation of error and efficiency of the model indicated the acceptable accuracy of this model in estimating the pan evaporation values in the annual scale. The 95% confidence interval confirmed the simulation results of the calibration step. Overall, the results showed that both VAR and VAR-GARCH models have high accuracy and correlation, and the model's performance criterion also confirms this. The percentage of improvement in the results from the model of the pan evaporation values in the annual scale using the VAR-GARCH model is about 4% relative to the VAR model. However, due to modeling the random section and reducing the uncertainty of the model, the results of modeling the pan evaporation values using the VAR-GARCH model are better than the VAR model. But due to the complexity of calculating the GARCH model, the VAR model can also be used.