Projection of extreme temperature events in megacity Beijing

This study leverages a comprehensive set of CMIP6 GCM outputs, downscaled via a convolutional neural network (CNN), to examine Beijing's temperature data spanning from 1961 to 2100. The ensemble of downscaled CMIP6 results indicates a consistent rise in Beijing's annual mean maximum (Tmax) and minimum (Tmin) temperatures from 1961 to 2100. Additionally, the potential impacts on local temperature-related extreme events are assessed using 8 indices of extreme climate events. The findings show that hot days (SU30) are projected to see a significant increase in both intensity (CSU30) and frequency (CSU30 frequency) over time and across scenarios, surpassing historical levels through a gradual step-up increase, a pattern similarly observed in warm nights (TR). The frequency and intensity of extreme temperatures associated with cold events (FD, ID) in Beijing are expected to decrease markedly, particularly in the central southern part of the study area. Regarding the temporal variation of urban-rural contrast, Tmin is on a downward trend under the SSP585 scenario, while that of Tmax is likely to remain stable under both the SSP245 and SSP585 scenarios in the future. The difference in the intensity of extreme weather events (SU30, CSU30, ID, FD, CFD, TR) between urban and suburban areas is growing, yet the frequency of extreme events (CSU30 frequency, CFD frequency) is on the decline. • Statistically downscaled CMIP6 GCMs data via CNN are generated to analyze Beijing's temperature trends from 1961 to 2100. • A significant increase is projected in hot days and warm nights with notable changes in both intensity and frequency. • Projected decline in urban-rural Tmin contrast under SSP585; stable Tmax contrast under both SSP245 and SSP585. • Urban-suburban contrast in weather extremes intensifying, whereas their frequency will show a slight decrease.