A stepwise cluster analysis (SCA) method is developed to downscale the daily mean surface temperature of 23 meteorological stations over the Beijing-Tianjin-Hebei (BTH) region in North China. Through comparisons of three evaluation indicators, including correction coefficients (CORRs), root mean square errors (RMSEs), and mean relative errors (MREs), between model simulations and observations, the SCA is proven to be capable of simulating the daily mean temperature in the baseline period (1961–2005). The projections of mean surface temperature under RCP4.5 and RCP8.5 scenarios are then developed using the SCA model. The results show that the mean surface temperature at most stations would keep increasing in future throughout the twenty-first century. The largest magnitude and rate of temperature increase are reported in the northwest mountains of the BTH region (especially obvious in winter), suggesting that the response to climate change in highland regions is more sensitive than that in low elevations. It is also reported that the projected changes in mean surface temperature in coastal regions are smaller than those in plain areas. This could be due to ocean’s moderating influence in the HadGEM2-ES. In terms of seasonal changes in surface temperature, the smallest magnitude of warming is expected in summer while the highest is likely to occur in winter, especially under RCP8.5 emission scenarios. Furthermore, the annual cycle analysis for temperature changes suggests that the highest amplitude of warming is expected in cold months (i.e., February or December) while the smallest is likely to occur in warm months (i.e., July or August).