Assessing and reducing uncertainties in future mean and extreme precipitation projections over China

A large spread exists in future precipitation projections based on climate models, particularly at regional scales, and efforts are needed to reduce uncertainties for effective regional adaptation strategies. Using outputs of 27 CMIP6 models, this study investigates uncertainties in mean and extreme precipitation projections over China. It is shown that uncertainties in the fractional change of mean precipitation are proportional to the magnitude of projected changes, being higher in the northwest and lower in the southeast, while extreme precipitation exhibits a more uniform spatial distribution. An analysis of uncertainty decomposition reveals that internal variability dominates short-term projection uncertainty, particularly for extreme precipitation. By the end of the 21st century, model uncertainty emerges as the largest contributor to mean precipitation uncertainty and accounts for more than one-third of extreme precipitation uncertainty. Optimal model selection method can reduce model uncertainty up to about 40 %, and it is more effective for mean precipitation than for extreme precipitation. Besides, the effectiveness of the method varies largely with the evaluation metrics and regions considered, emphasizing the need for metric-specific and region-specific model evaluations. These findings highlight the importance of tailored strategies to improve the reliability of future regional precipitation projections.