The comparison of bias correction methods in statistical downscaling of precipitation in Brunei Darussalam

Abstract

The rapid exponential increase of the level of greenhouse gases in the atmosphere has severely affected the climate globally, including Brunei Darussalam. It is thus vital to estimate the impacts of climate change and their future projections to minimize the effects not only for Brunei but worldwide. This study seeks to improve the statistical downscaling method of precipitation in Brunei Darussalam for efficient future projections by reducing bias correction errors. The study used the Canadian Earth System Model data under the Coupled Model Intercomparison Project Phase 5 (CMIP5) and found that six out of 26 predictors to be significant after using a Backward Stepwise Regression (BSR) to screen them. Five out of six predictors are from a box-grid location near the centre of the Borneo Vortex, which suggests that the vortex influences the formation of the precipitation in Brunei Darussalam. The regression equation from the BSR was used to downscale the precipitation for the calibration period from 2006 to 2013 and the validation period from 2014 to 2018. Bias correction methods using Quantile Mapping (QM) and Power Transformation (PT) were utilized and compared to evaluate the best estimator of the generated data. Based on the performance between the two bias corrections methods, QM performed better than PT in terms of matching with the characteristics of the observed historical precipitation which are the mean, standard deviation, number of rainy and non-rainy days, and maximum daily precipitation. However, PT performed better than QM for comparison of average monthly precipitations.