Projection of future climate change scenarios provides the scientific basis for addressing climate change and for proposing strategies of adapting climate change. This study used the simulation data of 30 climate models, which were evolved in the Coupled Model Intercomparison Project Phase 5 (CMIP5). Through evaluating the performance of each model on simulating the historical climate change, the preferred climate models were selected. Then, using the outputs of preferred climate models as independent variables and using ground measurements as dependent variables, the partial least squares regression (PLS) models were constructed for temperature and precipitation, respectively, of each region of China. By analyzing the ensemble predictions of regional temperature and precipitation changes, we found that the PLS ensemble mean of preferred climate models is closer to the ground measurements than the PLS ensemble mean of all of the climate models and the traditionally arithmetic average-based ensemble mean. The PLS ensemble projections of preferred climate model showed that climate warming would generally continue during the 21st century, which would be stronger in the cold half-year and in the northern regions than that in the warm half-year and in the southern regions. Under the scenario of RCP 4.5,the climate warming would be stronger in the first half of the 21st century and weaker in the second half. Under the scenario of RCP 8.5, the climate warming would keep nearly constant rate and, by the end of 21st century, the temperature would rise by two folds of that under the scenario of RCP 4.5. The increasing trend of precipitation would be stronger under the scenario of RCP 8.5 than that under the scenario of RCP 4.5 and would be stronger in the dry regions than that in the rainy regions with decadal oscillations. Finally, the equal weighting ensemble projections of all of the climate models exhibit that climate warming would be stronger in summer than in winter and that precipitation would increase linearly without decadal oscillations. These findings are opposite to the primary characters of climate changes that climate warming is stronger in winter than in summer and precipitation has strong inter-decadal variability. Thereby, the PSL-based ensemble mean of preferred climate model may provide reasonable projections of future temperature and precipitation changes.