Rainfall forecasting by using autoregressive integrated moving average, single input and multi input transfer function

This study aims to compare performance of three methods for forecasting rainfall, i.e. Autoregressive Integrated Moving Average, Single Input and Multi Input Transfer Function. These methods are applied to meteorological data from Indonesian Agency for Meteorology Climatology and Geophysics, Juanda-Surabaya. For the first modeling, using single variable, i.e. rainfall time lags. While for single input and multi input transfer function, using seven local predictor variables, that are minimum, maximum, average temperature, humidity, solar radiation, wind speed and maximum wind speed. The five global predictor variables are added, i.e. Nino1.2, Nino3, Nino3.4, Nino4 and DMI. From the transfer function model, it will be known to any exogenous variables that significantly affect rainfall. Based on the model performance measurement, it is found that the best model is Multi Input Transfer Function with four exogenous predictors, followed by Single Input and the last is univariate Autoregressive Integrated Moving Average. In the time-series modeling, its accuracy will increase when it involves external predictor variables, including rainfall forecasting.