Improvement of Seasonal ARIMA Model and Investigation of Model Performance for Monthly Precipitation Simulation

Precipitation is an important component of the hydrological cycle which links atmospheric and surficial process. Therefore, accurate modeling and estimation of parameter are needed for water resources management, irrigation scheduling, agricultural management and water allocation. SARIMA model is the most popular model for monthly precipitation simulation. The weakness of model is to ignore the inter-monthly variation within each year. Therefore, the aim of paper is the improvement of SARIMA model which takes into account the interannual, inter-monthly variation and comparison the performance of improved model with SARIMA model in Ardabil station. Ward's method for clustering of monthly precipitation time series and linear regression model for determination the relation between statistical characteristic of each cluster and monthly precipitation have been applied. 24.05%, 17.24% and 28.48% decreasing of RRMSE, RMSE and MAE from SARIMA to improved SARIMA model indicated the acceptable performance of improved model. The comparison of observed and simulated values showed the overestimation of improved model simulation. 51.16% RRMSE decreasing in seasonal precipitation simulation was expressive of accuracy increasing of improved model basis on the clustering analysis. The correlation coefficient between simulated of improved SARIMA model and observed precipitation data was increased and reached to the significant level. Therefore, the clustering analysis and improvement of SARIMA model, increased the accuracy of model.