TEKNIK STATISTICAL DOWNSCALING DENGAN REGRESI KOMPONEN UTAMA DAN REGRESI KUADRAT TERKECIL PARSIAL UNTUK PREDIKSI CURAH HUJAN PADA KONDISI EL NINO, LA NINA, DAN NORMAL

Abstract

This paper presents the results of validation of rainfall prediction models in Indramayu district using statistical approaches for downscaling, i.e. Principal Component Regression and Partial Least Square Regression, during El Nino, La Nina, and Normal conditions. Rainfall data from 6stations and the precipitation data from Global Circulation Model ECHAM3 are used in thisanalysis with the domain size 8x8 (1.4 ° S-18.1° S; 98.4 °-118.1° E), over the Indramayu region.Data are classified into each climatic anomaly condition based on the Oceanic Nino Index (ONI)which uses sea surface temperature data at the Nino 3.4 as the results of NOAA analysis. Theaverage value of RMSEP and correlation in El Nino conditions are 95.22 and 0.66 for PCR and102.52 and 0.62 for PLS,in La Nina conditions the values are 85.14 and 0.65 for the PCR, and 98.43and 0.69 for the PLS, and in normal conditions the values are 91.41 and 0.57 for PCR, and 85.37and 0.63 for PLS. In general PCR shows better performance than PLS in El Nino conditions, whilein La Nina and Normal conditions the PLS performance is better than PCR. The selection modeldepends on the coverage areas studied, whether representing the area around the rainfall station orrepresenting a district area.