24.-River flow forecasting using multivariate time series modelling

Abstract

Making efficient forecasts of floods is one of the major tasks in flood hydrology. In situations, where the major concern is to accurately predict the river flows or floods, a hydrologist may make use of time series modeling approach instead of developing a conceptual model for the basin. In this paper, river flow forecasting is approached assuming that daily flows follow an auto regressive moving average (ARMA) process. Based on auto. correlation and partial auto correlation functions, various ARMA models were considered and evaluated for their performance. The ARMA(3,1) model was found suitable, while others were discarded based on statistical analysis. The model has been used for forecasting daily flows using a one step ahead procedure. The forecasted series was analyzed for model performance and found satisfactory. The model was developed using the data of the Baitarani river basin, Orissa