How well are the downscaled CMIP5 models able to reproduce the monsoon precipitation over seven homogeneous zones of India?

Abstract

Statistical downscaling through perfect prognosis (PP) method is widely utilized to bridge the gap between large‐scale global climate model (GCM) simulations and regional scale or local scale observed predictands. Present study has assessed the performances of PP‐based downscaled CMIP5 GCMs in simulating observed monsoon precipitation over seven homogeneous zones of India, namely, North Mountainous India (NMI), Northwest India (NWI), North Central India (NCI), Northeast India (NEI), West Peninsular India (WPI), East Peninsular India (EPI) and South Peninsular India (SPI). Firstly, PP models have been constructed through principal component regression (PCR) using large‐scale atmospheric predictors from National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis. Secondly, GCM predictors have been imposed on the PP models to downscale large scale GCM simulations at regional scale. Four performance metrics namely percent bias (PB), interquartile relative fractions (IRF), Perkins skill score (PS) and Kuiper metric (KM) have been considered to evaluate skills of downscaled GCMs in reproducing mean, variance, probability distribution function (PDF) and cumulative distribution functions (CDF) of observed precipitation, respectively. As per results of several metrics, PP models have performed relatively better over NCI and SPI zones. However, they have shown poor skills in reproducing the observed variance over all zones. Further to improve the performances of PP models, quantile mapping has been embedded to form hybrid (PPQM) models, which have shown superior skills over all the zones. In addition, PPQM models have also shown their applicability to provide more reliable added value information over sub‐regional scale compared to raw GCMs.