Author(s): S. Suresh; M. V. Somasundaram; N. V. Pundarikanthan
Keywords: Time series; Forecasting; Stochastic; Autocovariance; Autoregressive; Trend and prediction
Abstract: Rainfall and flow prediction, adapting the Venkataraman single time series approach and Wiener multiple time series approach were conducted for Aralikottai tank system, and Kothamangalam tank system, Tamilnadu, India. The results indicated that the raw prediction of daily values is closer to actual values than trend identified predictions. The sister seasonal time series were more amenable for prediction than whole parent time series. Venkataraman single time approach was more suited for rainfall prediction. Wiener approach proved better for daily prediction of flow based on rainfall. The major conclusion is that the sister seasonal time series of rain and flow have identities of their own even though they form part of the whole parent time series. Further studies with other tropical small watersheds are necessary to establish this unique characteristic of independent but not exclusive behavior of seasonal stationary stochastic processes as compared to parent nonstationary stochastic processes.