A non-local model output statistics approach for the downscaling of CMIP5 GCMs for the projection of rainfall in Peninsular Malaysia

Abstract

Abstract In this study, a non-local model output statistics (MOS) approach is proposed for the downscaling of daily rainfall of coupled model intercomparison project phase 5 (CMIP5) general circulation models (GCMs) for the projections of rainfall in Peninsular Malaysia for two representative concentration pathway (RCP) scenarios, RCP4.5 and RCP8.5. Projections of eight GCMs for both the mentioned RCPs were used for this purpose. The GCM simulations were downscaled at 19 observed stations distributed over Peninsular Malaysia. Random forest (RF) was used for the development of non-local regression-based MOS models. The results revealed a high accuracy of the models in downscaling rainfall at all the observed stations. The mean absolute error (MAE) of the models were found in the range of 0.8–0.39; normalized root mean square error (NRMSE) between 7.4 and 41.7, Percent Bias (PBIAS) between –0.3 and 10.1, Nash–Sutcliffe coefficient (NSE) between 0.81 and 0.99 and R2 between 0.89 and 0.99. The increase in annual rainfall was in the range of 7.3–29.5%. The increase was higher for RCP8.5 compared to RCP4.5. The maximum increase was observed in the northern part of Peninsular Malaysia in the range of 20.7–29.5%, while the minimum in the south-west region was in the range of 7.6–15.2%.