Bias correction of CMIP6 GCMs for historical and future air temperatures across China

The CMIP6 global climate models (GCMs) yield a generally large bias in air temperature simulation over China, necessitating corrections before we can rely on their future projections. In this study, we performed air temperature corrections of 26 CMIP6 GCMs using two methods—bias correction method (BC-correction) and quantile mapping (QM-correction), with the CN05.1 observational data serving as the benchmark. We conducted a comparative assessment of the simulation performance of the CMIP6 GCMs and their multi-model ensemble means (MMEs) before and after corrections, considering both temporal and spatial scales across historical and future periods under multiple shared socioeconomic pathway (SSP) scenarios. The results show that the BC-correction method substantially rectifies the systematic underestimation of annual mean air temperature in CMIP6 GCMs across China, demonstrating superior performance compared to the QM approach, but with varying correction effects observed across different seasons. The MMEs show efficacy in capturing temporal-spatial variation patterns of air temperature recorded by the CN05.1 observation, however, certain discrepancies persist in specific trend magnitudes and locations. Moreover, compared to the original MME, the BC-corrected MME reveals projected enhanced future warming amplitudes across various timeframes and SSPs relative to the baseline period (1995–2014). This study emphasizes that uncorrected GCMs tend to underestimate projected climate warming across China.