Year-Long Monthly Rainfall Forecasting for a Coastal Environment of Bangladesh

Abstract

Forecasting rainfall plays an important role to develop, planning and management a sustainable water resource system. In this study stochastic Seasonal Auto Regressive Integrated Moving Average (SARIMA) were used to forecast monthly rainfall of Teknaf for 12 month lead time. The best SARIMA (0, 0, 0) (1, 1, 1) model was selected based on Normalized BIC (Bayesian Information Criteria) and R-squared. Diagnostic check was then conducted for the best fitted model to check if the residuals are white noise. The predicted rainfall amount from the best fitted model was compared with the observed data. The predicted values shows reasonably good result. Thus the model can be used for future rainfall prediction.

Keywords: Bangladesh, Teknaf, Rainfall, ARIMA, Forecast