Support Vector Regression Modelling for Rainfall Prediction in Dry Season Based on Southern Oscillation Index and NIN03.4

Abstract

Various c1imate disasters in lndonesia are mostly related to the El Nino Southern Oscillation (ENSO) phenomenon. The variability of etimate especially rainfall is strongly related to this phenomenon. Southern Oscillation Index (SOl) and sea surface temperature anomaly (SST A) at Nin03.4 region are two common indicators used to monitor phenomenon of El Nino and La Nina. Furthermore, SOl and NINO SSTA can be the indicator to find the rainfall probability in a particular season, related to the existing condition of c1imate irregularities. This research was conducted to estimate the rainfall during dry season at Indramayu district. The basic method used in this study was Support Vector Regression (SVR). Predictors used were SOl and NIN03A sea surface temperature (SST) data. The expcrtments wcre conducted by comparing the model performance and prediction rcsults, The training set was c1ustered in advance and then SVR model was generated using RBF kernel based on their c1ustering r esult. This research obtaincd an SVR model with correlation coefficient of 0.76 and NRMSE error value of 1.73.