Development of ARIMA model for monthly rainfall forecasting over an Indian river basin

Abstract

The quantification of climate variability, especially in terms of rainfall patterns, is of vital concern all around the globe at present. Owing to the significant variation of rainfall distribution in both space and time, its precise estimation is crucial at river basin level. In this study, an autoregressive integrated moving average (ARIMA) model is developed for prediction of monthly rainfall over Betwa River Basin, India. The monthly precipitation data for 1960–2012 were obtained from the India Meteorological Department (IMD) Pune. The monthly average precipitation over the whole basin was estimated using Thiessen polygon method. The precipitation time series from 1960 to 2000 was utilized for training purposes in order to frame the most suitable model, and that of 2001 to 2012 was utilized for testing and validation purposes. The metrics viz., Akaike information criterion (AIC) and Bayesian information criterion (BIC) were used for selecting optimum values of the parameters; and the ARIMA (0, 0, 1) (1, 1, 2)12 was established as the parsimonious model. The Nash-Sutcliffe efficiency (NSE) and the coefficient of determination (R2) were used to evaluate the model efficiency. The model forecasts possessed an excellent match with the observed rainfall data (2001–2012). The outstanding accuracy of the model for predicting monthly rainfall for such a long duration of 23 years justifies its future application over the study region. This study will be helpful in developing a proper management framework for the core of Indian monsoon region. © ASCE.