Abstract: Using the daily high/low observational temperature data of 740 stations in China from 1960 to 2005, the statistical characteristics of recordbreaking daily temperatures (RBDT) in the last 46 years are analyzed theoretically. Based on the Gaussian distribution of daily temperatures, and using the RBDT as initial conditions, the universal mean of the RBDT is obtained and compared with the result from the iterative Extreme Values Statistics (EVS), with the former being closer to the observational distribution. Then the future possible RBDT are predicted by the former method. In addition, the future distributions of theoretical maximum/minimum values of high/low RBDT are given as well. The results show that the amplitude of extreme temperatures variation change with geographic locations. The extreme high temperatures of Southwest China increase rapidly, and those of West China are lower than other regions of China. On the other hand, the extreme low temperatures decrease fast in the Northeast and Southeast China, while those of the Southwest China are relatively stable.