Development of rainfall forecasting model in Indonesia by using ASTAR, transfer function, and ARIMA methods

The aim of this research is to find the best method to most rainfall index data in Indonesia by comparing the forecast accuracy among ARIMA, ASTAR, Single-input Transfer Function, and Multi-input Transfer Function models. Three location of rainfall data at East Java are used as case study, i.e. Ngale, Karangjati, and Mantingan. In this research, Seasonal ARIMA, as the appropriate type for rainfall index data, is used. Three kinds of ASTAR models are used. Single-input Transfer Function model use Dipole Mode Index (DMI) and Sea Surface Temperature (SST) as the input one by one, and Multi-input Transfer Function model use these inputs simultaneously in the model. The results show that multi-input transfer function model yields better forecast at in-sample data in Ngale and Karangjati. The comparison of forecast accuracy at out-sample data show that singleinput transfer function model yields better forecast at these locations (Ngale and Karangjati). For rainfall data in Mantingan the best model is ASTAR model both in insample and out-sample data.