This study analyzes the climate simulation for the past 100 years over East Asia using the WRF-ARW (version 3.5.1) model, which is forced by the Twentieth Century Reanalysis (20CR-v2) data. The model continuously runs from 1900 to 2010 at 50 km horizontal resolution. Climatic means and seasonal cycles from the 20CR-v2 and WRF results are compared with observations in southern China for the last 50 years(1961-2010)with a focus on annual mean surface air temperature and precipitation and mean surface air temperature and precipitation in summer and winter. Results indicate that the dynamical downscaling method can reproduce the main spatial distributions and seasonal cycles of surface air temperature and precipitation. The WRF model outperforms the 20CR-v2 in regional details due to the high-resolution topography and land-use forcing. Compared with the 20CR-v2, biases of surface air temperature and rainfall are reduced in the dynamically downscaled WRF model experiment, especially in the annual mean and in summer. The pattern correlations between observations and WRF simulation for surface air temperature and precipitation are greater than 0.97 and 0.5, respectively. Furthermore, the simulated regionally averaged temperature and precipitation are close to observations in four periods (1914-1942, 1943-1971, 1972-2000, and 2001-2010) of the past 100 years over three sub-regions (South China, Central China, and Southwest China) of southern China. The WRF model has significantly improved the simulation in summer but temperature is underestimated in winter. Compared with the 20CR-v2, the WRF model has improved the simulation of temperature trend during the last 100 years, particularly that after the 1940s.