Projections of Autumn Precipitation Extremes in Southwest China Based on a Statistically Downscaled CMIP6 Dataset

This study evaluates the performance of 15 CMIP6 models (before and after downscaling) in simulating autumn extreme precipitation in Southwest China based on a high-resolution, statistically downscaled CMIP6 dataset, using the CN05.1 dataset as a reference. The BCCAQ (Bias-Corrected Constructed Analogues with Quantile Mapping Reordering) method used in deriving the downscaled CMIP6 dataset significantly enhanced the models’ ability to reproduce the spatial patterns of the extreme precipitation indices, particularly for total precipitation, R10 (number of moderate rain days), and R20 (heavy rain days). Notable improvements were also observed for RX1 (maximum 1-day precipitation), RX5 (maximum 5-day precipitation), and SDII (simple daily intensity), alongside reduced inter-model spread and systematic biases. Bias correction also improved the simulation of interannual variability, substantially reducing RMSE for total precipitation, R10, and R20. Increased interannual variability in the future is expected for certain indices, spatially concentrated for RX1 and RX5 in the south and R20 in the east. Projections using bias-corrected multi-model ensemble under SSP2-4.5 and SSP5-8.5 scenarios indicate a significant intensification of autumn extreme precipitation (intensity and frequency indices), especially in southern Southwest China by the 2080s, with precipitation becoming more concentrated in heavier events. While changes in consecutive dry days (CDD) are spatially variable and increase in the southeast, consecutive wet days (CWD) show no significant change. These findings highlight an increased risk of intensified autumn rainfall and altered precipitation patterns in the region under future climate change.