The Weather Research and Forecasting model (WRF) comprises a number of parameterization schemes, and therefore the different combined physical parameterization schemes should be discussed before determining optimal configurations. The regional climate of central Asia is simulated from 2003 to 2008 using the WRFv4.0.1 with nine physical combinations (microphysics [MP], cumulus [CU], planetary boundary layer [PBL]) on a 25‐km horizontal grid. The goal is to identify a reasonable configuration assemble of physical parameterization schemes for long‐term simulations through the sensitivity experiments of regional climate respond to microphysics, convection, and boundary layer processes. Compared to the observation data, all the groups mostly simulate the daily and seasonal variations of precipitation but with a shift from wet bias to dry bias based on the daily precipitation categories of the Tropical Rainfall Measuring Mission (TRMM). The large‐scale precipitation caused by the microphysics schemes is 3–4 times the amount of precipitation caused by the subscale precipitation generated by the cumulus schemes. The simulated difference of the vertical temperature shows results similar to those of the surface air temperature. The microphysics schemes play the dominant role among the three physical options, and the Thompson scheme addresses better precipitation and surface temperature simulations in contrast to the WDM6 scheme. By combining with the bias, correlation and root‐mean‐square error (RMSE) between the reproduced temperature and precipitation, it can be concluded that the best optimal physical schemes combination is the Thompson–Tiedtke–YSU, which can be used for long‐term simulations.