Monthly Rainfall Prediction Using Vector Autoregressive Models Based on ENSO and IOD Phenomena in Cilacap

A study of El Nino-Southern Oscillation (ENSO) and Indian Oscillation Dipole (IOD) as climate elements to predict extreme rainfall is critical to be carried out. This study aims to forecast monthly rainfall by using Vector Autoregressive (VAR) model in the Cilacap. Rainfall data were obtained by the Global Satellite Mapping of Precipitation (GSMaP) and ENSO and IOD by the National Oceanic and Atmospheric Administration (NOAA) from March 2000 to December 2018. The data was used to predict monthly rainfall for the next twelve months in 2019. The VAR model with a minimum lag of length 2 was considered to be the best lag and selected to model the rainfall in the study area. Response function reveals that changes in rainfall significantly affect changes in rainfall after some time lags. The results of the actual data accuracy test with predictive data from the Vector Autoregressive (VAR) model were conducted with the root mean square error (RMSE) and the mean absolute percentage error (MAPE). The RMSE value was 0.3291, 0.047, 0.061 and the MAPE value was 87.0671, 0.1845, 0.217. Rainfall in the Cilacap is predicted to be stable in the future. This condition is the same as the rainfall in the previous periods, which fluctuated at the same point.