Modelling on annual and monsoon rainfall of Cooch Behar district in West Bengal: An ARIMA approach

Abstract

The present study was conducted to identify the best fitted time series model to describe the annual and monsoon rainfall in Cooch Behar District of West Bengal. Rainfall data for 40 years, i.e. from 1978 to 2017 was collected from Indian Meteorological Department, Pune. The three stage Box Jenkins procedure for selection of Auto Regressive Integrated Moving Average model was performed. From the evaluation of model combinations, it was found that ARIMA (0,0,1) is the most appropriate model for fitting both monsoon as well as annual rainfall data. ARIMA (0,0,1) possessed lowest Bayesian Information Criteria and Root Mean Square Error. The Absolute Relative Percent Forecast Error was utilized for confirming the predictive power of the identified best fitted model. The same model was used for predicting the annual and monsoon rainfall for the next three years.