The Principal Component Linear Spline Quantile Regression Model in Statistical Downscaling for Rainfall Data

Abstract

Information regarding rainfall can be obtained from global data, namely the global climate model that can be accessed through the statistical downscaling approach. Linear spline quantile regression with principal component is a statistical method that can be employed in statistical downscaling to address multicollinearity and outliers in data by using nonparametric estimators. This method is applied to rainfall data in Pangkep Regency from January 2008 to December 2022 as the response variable and global climate model data as the predictor variable. The aim of this research is to obtain the best regression model used for predicting rainfall data. The results obtained indicate that statistical downscaling with two principal components at the 0.50 quantile with respective knot points of -10.20 and -0.30 is the best model with the lowest generalized cross-validation value. The forecasted rainfall data using this model shows a high level of accuracy with a correlation of 89%.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint