STATISTICAL DOWNSCALING MODELING FOR MONTHLY RAINFALL ESTIMATION USING GEOGRAPHICAL AND TEMPORAL WEIGHTED GAMMA REGRESSION

Statistical Downscaling modeling is a technique in climatology that uses statistical modeling to analyze the relationship between large-scale data (global) and small-scale data (local). General Circulation Model is a numerical model that produces many data from various climate parameters such as precipitation, temperature, and humidity for the need for climate forecasting. Statistical Downscaling modeling to estimate monthly rainfall in areas that have a monsoon rainfall pattern in Indonesia had been carried out using the L1/Lasso Regulation and Principal Component Analysis, Spatio Temporal Bayesian Regression, Spatio Temporal Generalized Linear Mixed Model, and Geographically and Temporally Weighted Regression. Monthly rainfall data are spatial and temporal heterogeneity and are not normally distributed because of non-negative values and skew to the right. One approach to analyze the data is using the Geographically and Temporally Weighted Gamma Regression method that was developed from Geographically and Temporally Weighted Regression using Gamma distribution and parameter estimation using the Maximum Likelihood Estimation method. This study will conduct this modelling using response variables of monthly rainfall data from 35 stations in West Java Province from January 2010 to December 2012, and predictor variables are monthly rainfall of the previous month, monthly precipitation from the General Circulation Model from the National Centers for Environmental Prediction in the form of a Climate Forecast System Reanalysis model. The study results show that Geographically and Temporally Weighted Gamma Regression modelling using the Gaussian kernel function and fixed bandwidth on Statistical Downscaling can predict monthly rainfall in a location in West Java Province for a certain period. Based on this model generated at a location, changes in a predictor variable can be seen in the value of monthly rainfall in a period. The combination of this model with the Kriging Spherical interpolation can estimate the monthly rainfall value in locations that are not observed.