FORECASTING OF STREAMFLOW USING TIME SERIES MODELLING

Abstract

For the proper management of any hydrological or water resources projects, the primary key is the early availability of the data associated with the project. One of the crucial tools regarding the same is an approach based on time series analysis. Time series analysis for the forecast of the monthly streamflow has vital importance in water resources engineering and act as a fundamental part in planning, designing and management of water resources systems. In this study, autoregressive integrated moving average (ARIMA) model has been used for forecasting the monthly discharge of the Sarda River at Banbassa, Uttarkhand, India. ARIMA model improves the performance of advance information for making planning and maintenance of the available water resources. The behaviour of the streamflow under different level of demand has been analyzed based on autoregressive integrated moving average (ARIMA) model, and it was found that the used model has great efficiency for the fitting and prediction. In order to implement the model application, a 32 years span of streamflow data from 1976 to 2007 has been used. The First 30 year’s data have been used for developing and trending a statistics related ARIMA model and the last two years streamflow data have been used for the validation of the generated model. The working procedure of the ARIMA model is based on the combine operation with various AR and MA orders. The developed model has been selected based on the tvalue and the residual of the autocorrelation function (ACF) and partial autocorrelation function (PACF). In this study, the statistical analysis for a developed model has been made with the help of IBM SPSS version 21. The prediction accuracy of various developed models has been examined by comparing their mean absolute percentage error (MAPE), and the coefficient of determination (R2 ) values and hence selected the best iterative model based on the above comparison. Furthermore, the selected iterative model has been used to forecast the stream flow up to 3 steps ahead in terms of MAPE. According to, above analysis, the generated model has been found the best solutions for the proper predictions and forecasting for the future usage of the streamflow resources and management.