RAINFALL BIAS CORRECTION FROM REGIONAL CLIMATE MODEL CCAM NN JAVA ISLAND

• Main Author: Kartika, Fenny
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/70808
• Institution: Institut Teknologi Bandung

Currently the climate has undergone many changes in tropical regions such as Java Island. One of the climate change adaptation efforts is with climate modeling studies. BRIN has carrie out dynamical downscaled the CCAM model into regional climate models, but CCAM's current regional climate external model data still having an uncorrected bias in order to be applied to climate change analysis and hydrology required reliable quality data. Therefore, the correction of rainfall bias of CCAM regional climate models using three bias correction methods needs to be carried out to reduce bias so that the data is more accurate and determine the performance of the best correction method for the Java Island region. The study used SA-OBS observation rainfall data: A Daily Gridded from 1991 to 2005, as well as CCAM regional climate model rainfall data from 1991 to 2005. The data is then processed by being corrected using three methods (Quantile Mapping, Linear Scaling and Local Intensity Scaling) and then evaluated using general statistics, namely average, correlation and RMSE as well as verification of probability of detection (POD) and False Alarm Ratio (FAR) dichotomous. The results showed that the rainfall of the CCAM model was very biased towards observation data. Rainfall correction can increase the CCAM model close to the observed value evidenced by an increase in mean and correlation, and a decrease in RMSE value for all three methods. Rainfall after correction is able to more accurately detect rainfall with a threshold of 0.5 mm and extreme rainfall of the percentile threshold of 90 with a POD value close to 1. The Quantile mapping method has the best performance for all general statistical evaluations and dichotomous verification, while the Local Intensity Scaling method shows poor results in almost all statistical evaluations.