Penerapan Model ARIMA untuk Prediksi Anomali Curah Hujan Berdasarkan Indeks Monsun dan El Niño di Beberapa Kawasan di Indonesia.

Abstract

This study based on atmosphere-ocean interaction that is very important to be investigated because of its influences to the rainfall variability over Indonesia. This study aims to investigate Monsoon and Niño 3.4 interaction in influencing rainfall fluctuation over Indonesia. The study areas cover Lampung, Sumbawa Besar, Indramayu, Banjarbaru, and Pandeglang from Januari 1976 - Desember 2000. The results of Power Spectral Density (PSD) and Wavelet analysis on rainfall anomaly and monsoon index data show strong at 12 months. Dominant oscillation pattern of Niño 3.4 is about 60 months. The result of regression analyses shows more significant relation between rainfall anomaly and AUSMI (Australian Monsoon Index) and WNPMI (Western North Pacific Monsoon Index). Therefore, it is suggested that the Monsoon (AUSMI and WNPMI) and El-Niño(Niño 3.4) can be used for further analysis. This study was undertaken with assumption that Monsoon oscillation and El-Niño are interacted and the model is developed by using multivariate regression method: Y = a + bAUSMI + cWNPMI + dNiño3.4. By using Box-Jenkins method based on ARIMA (Autoregressive Integrated Moving Average), the prediction model that is close to observed time series data of rainfall anomaly is ARIMA (1,0,1)12 with model equation Zt = 0.9989Zt-12 - 0.9338at-12 + at (AUSMI), ARIMA (1,1,1)12 with model equation Zt = -0.0674Zt-12 + Zt-12 - Zt-24 - 0.9347at-12 + at,(WNPMI), dan ARIMA (2,0,2) with model equation Zt = 3.594Zt-1 -0.8362Zt-2 – 1.634at-1 - 0.1053at-2 + at (NINO3.4). Model equation shows that time series data forecast of rainfall anomaly depend on previous “t” months data and previous ”t” months error.