A Principle Component Analysis Cascade with Multivariate Regression for Statistical Downscaling Technique :A Case Study in Indramayu District

Abstract

This research is focused on the development of statistical downscaling model using principle component analysis (PCA) as feature extraction cascade with multivariate regression as model prediction. Sea Surface Temperature (SST) and the General Circular Model (GCM) derived from 18 ensemble members of ECHAM5 model under AID scenario are used as predictors, and the September, October, November, and december (SOND) rainfall in each station in Indramayu as the response variables. SST domains were reduced using the PCA that explain 80%, 85%, and 95% of the data variability, which were divided into three sets, namely SST lag 1 (August), lag 2 (July), and lag 3 (June). GCM domain-sized 5x5 with 2.8x2.8 km resolution is reduced by PCA. The number of components that are taken are such that the variables must still explain 90% of original data variability. The new variables yielded by peA are then take as the input of the multivariate regression with 13 observation data (from 13 station) as the output. The regression parameters are estimated by using the least sum of square error criteria. By assigning 0.4 as the correlation boundary to select the domain SST, the SST lag 3 (June) yields the best regression model with a correlation of about 0.7. But if the limit is increased to 0.55, only the SST lag I (August) that meet, and produce a model with a correlation above 0.7.