Stochastic modelling for selecting an appropriate autoregressive model to predict average annual river flows into Shahid Veer Narayan Singh Dam

Abstract

This paper deals with the Stochastic modelling of annual River flows by using Autoregressive models. For present study 21 years data from 1994-2015 were used. On the basis of autocorrelation function (ACF) and Partial autocorrelation function (PACF) analysis of annual time series of River flows, AR model of different order 1, 2, 3 were tried and developed and their parameters were estimated. The adequacy test of selected model was carried out by Akaike Information Criteria, Portmantaeu lack of fit test. For calibration of all developed models about 70% of available data of annual River flows were used and the developed model were validated by prediction of rest of the data and compared with the observed average annual Rivers flow series. The forecast skills of different approaches were compared using a variety of measures of performance. Such as Root Mean Square Error (RMSE), Mean Relative Error (MRE), Absolute average error (AAE), Integral Square Error (ISE). Performance results indicated the high degree of model fitness which can represent the best possible forecast of annual River flows and also among the all developed models AR(1) i.e. Autoregressive model of order 1, can be used for best Prediction of average annual River flows of Kodar Nalla into Shahid Veer Narayan Singh Dam Situated in Mahasamund district of Chhattisgarh.