ARIMA method for development of a monthly rainfall forecasting model in Krishnanagar, West Bengal, India

Abstract:

Hydrological parameters are complicated phenomena, which require modern computer modelling and simulation techniques for accurate forecasting. Thorough understanding of the rainfall process is critical for the proper planning of integrated water resources management projects which relates agriculture, climate change, and natural hazards such as floods and droughts. In the present study, an auto-regressive integrated moving average (ARIMA) model was developed in MATLAB environment for simulating and forecasting the rainfall data for the study area using Box-Jenkins methodology. Krishnanagar (eastern India) region was selected as a study area. Rainfall data from 1971-2000 (30 years) were used for the development of the model while data from 2001 to 2010 was used to verify the developed model. Four basic steps i.e. model identification; estimation of model parameters; diagnostic checking for the identified model appropriateness for modeling and application of the model (i.e. forecasting) as suggested by Box and Jenkins were followed in the present study to develop a suitable rainfall prediction model for the study area. The ARIMA model (1,0,0)(1,1,1)12 was found to be the most suitable for forecasting monthly rainfall over the study area. The selected model can also be used for forecasting the monthly rainfall data for the up-coming years.

Keywords: Rainfall, time series, ARIMA, Uni-variate auto-regressive parameters