Based on the observation and a set of downscaling bias corrected model (NEX-GDDP-CMIP6) data,this study has investigated the model performance in simulating precipitation extremes over Southwest China using Taylor diagram and GEV fitting methods.Furthermore,the future changes of precipitation extremes that with different return intervals and their associated population exposure are also explored.Results show that NEX-GDDP-CMIP6 models and the multi-model ensemble (N-CMIP6-MME) can reasonably capture the spatial-temporal characteristics of changes in precipitation extremes over Southwest China,and N-CMIP6-MME out performs most of individual models.In the future,the precipitation extremes are expected to significant increases over most regions of Southwest China,including the extreme events that occurring once in 10 or 20 years,which will further increase the risk of population exposure to these extremes.Additionally,for the more extreme events (once in 20-year),the increasing magnitude of its occurring probability would be much greater and it would also exert a larger increase of its associated population exposure,when compared to the events of once in 10-year.Around the year of 2050,the occurring probabilities of the RX1day (RX5day) extremes that occurring once in 20-year are expected to increase by 175.2％(148.9％),216.0％(162.4％),210.9％(156.8％),and 274.3％(207.1％) under SSP1-2.6,SSP2-4.5,SSP3-7.0and SSP5-8.5 scenario,respectively.Correspondingly,thepopulation exposuresare projected to respective increase by 129.1％(118.8％),177.7％ (135.1％),182.4％(143.2％),and 237.5％(161.5％).Further analyses indicate that the increase of population exposure to the precipitation extreme over Southwest China is mainly due to the significant increase of precipitation extremes,and the changes in populations and their interaction generally present a negative contribution.