Optimization of Bias Correction Methods for RCM Precipitation Data and Their Effects on Extremes

Abstract

Studies on impact due to climate change on water resources are typically evaluated for regional scale and the evaluation is carried out at site-specific or local scale. General Circulation Models (GCMs) and Regional Climate Models (RCMs) are used to understand future climate changes. RCMs have higher resolution to understand the reliable estimation of local-scale climate variables. RCMs show critical biases in precipitation and therefore it is mandatory that bias correction is to be carried out so that they are usable for research. Six RCMs are considered and analysed in this study to reduce the errors in RCMs during the period from 1970 to 2005 over Amaravati region in Andhra Pradesh, India. Four statistical bias correction techniques, namely linear scaling, cumulative distributive transformation, quantile mapping using parametric transformation and quantile mapping using smooth spline methods are used. These bias-corrected datasets are compared with observed datasets using different relative errors, viz. standard error, mean absolute error, root mean square error and mean square error. Relative errors show the performance of simulated data with observed data. The results show that quantile mapping using parametric transformation technique gave optimum values for the results with minimum error compared to the other three methods. However, there is no generalized optimized technique available, at present, to reduce the bias in the datasets of the RCMs, and there is a need to reduce the errors for reducing the uncertainties in the climate impact studies either at the local or regional scale.