TIME SERIES ANALYSIS, MODELING AND FORECASTING OF CLIMATE VARIABLE RAINFALL: A CASE STUDY OF RAJSHAHI DISTRICT IN BANGLADESH

Abstract

The purpose of the present study was to investigate the time series components, and to build an
appropriate model to forecasttherainfall of Rajshahi district in Bangladesh using the monthly
rainfall data over January, 1975 to June, 2012 collected from Bangladesh meteorological
department.The statistical software R with the packages ‘forecast’ and ‘zyp’ was used for
whole analysis. The descriptive statistics of rainfall showed high fluctuation from their mean,
positively skewed and platykurtic curve. The time series rainfall data was decomposedinto
stochastic trend, seasonal variations and random movements. The yearly rainfall data showed
decreasing trend by 11.25 mm/year. The moving average smoothing of monthly rainfall
intended to be decreased over May, 2007 to June, 2012. But, after first seasonal difference, the
rainfall data became stationary and that wasconformed using appropriate tests. The
SARIMA(0, 0, 0)(4, 1, 0)12 model was found as best model on the basis ofdiagnostic test,
stability and reliability. The forecasted values from July, 2012 to December, 2025 divulged
adecreasing pattern that may be a threat to the cultivators as well as to the nature as a whole.