Does CMIP6 inspire more confidence in projecting precipitation over China?

Abstract

Despite significant advancements in recent versions of general circulation models (GCMs), uncertainties persist in simulations of both historical and future precipitation. Notably, disparities exist between the simulated precipitation before 2015 and the projected precipitation in Coupled Model Intercomparison Project Phase 6 (CMIP6) models. In this study, the accuracy of projected precipitation in 12 CMIP6 models was evaluated before and after applying SD and EDCDFm methods, by comparing with observed precipitation data during 2015–2020. This study aims to address the gap in the comprehensive evaluation of projected precipitation against observational data. The results of the study showed that: (1) Downscaling and bias correction improved the skill of CMIP6 models in simulating spatial distribution of precipitation during 1961–2014, especially in east China. However, bias‐corrected projected precipitation during 2015–2020 exhibited a dry bias over most of China. (2) Bias correction improved the capability of CMIP6 models in simulating historical monthly variations in precipitation, but did not address the issue of projected precipitation following the simulated seasonal cycles. (3) The improvement of projected precipitation after bias correction was limited during 2015–2020, and for CanESM5, IPSL‐CM6A‐LR, and MIROC6 models, bias correction even exacerbated the bias of projected precipitation over the Song–Liao River, Yangtze River, and Southeast River basins. (4) Bias‐corrected CMIP6 models generally underestimated the increase in precipitation after 2015, but projected precipitation over China was still expected to increase during 2015–2099 after bias correction. These findings emphasize the need for more precise strategic suggestions to improve the projection of precipitation.