Autoregressive Model for Flood Forecasting

Abstract

The purpose of the present paper is to develop an autoregressive model for forecasting the frequency of occurrence of flood that India may countenance in a given year. The proposed model being autoregressive, prediction requires no other data than the temporal frequency of occurrence of flood. A comparative study is made among several orders of autoregressive models. The comparison is furnished on the basis of Akaike Information Criteria (AIC) and Bayesian Information Criteria (BIC) statistics for different orders of autoregressive equations. The result of the study reveals the suitability of third order autoregression as a predictive model. The result is qualitatively supported by learning the sample autocorrelations and autocorrelations theoretically implied by third order autoregressive model.