Rainfall time series forecasting using ARIMA model

Abstract

The Northeastern region of India, especially Assam and Meghalaya receive very heavy rainfall in the months of monsoon, which results in repeated loss of property, life and resources every year. This study aims to predict the rainfall distribution pattern over Assam and Meghalaya for the next five years. From the data available at Indian Meteorological Department’s website of the daily rainfall distribution pattern of the two regions, a time series has been created. This time series has been used to forecast the rainfall distribution pattern using an ‘Auto Regressive Integrated Moving Average’ (ARIMA) model. ARIMA (0,0,1)(2,1,2) was selected by comparing AICc values to forecast the data. This forecasting algorithm uses the past values of the series in predicting the future trend. The R programming language has been used for the entire study for precise statistical analysis. Thus, an accurate forecast of the rainfall pattern for the future 5 years can be extremely beneficial for the people of the region in planning their resources, crop patterns and managing disasters.